CONFERENCE

PROCEEDINGS
Table of Contents
Papers by session

Monday, October 2, 2023 Wednesday, October 4, 2023

10:30 am - 12:00 pm 8:30 am - 10:00 am
Foundations I Case Studies I
Geohazards I Foundations III
Stabilisation and Reinforcementent Numerical Models II
Tailings and Liquefactions Tailings and Rock
Unsaturated Soils
3:00 pm - 4:30 pm
Cover Systems 10:30 am - 12:00 pm
Geoenvironmental II Advanced Testing II
Geohazards II Cold Regions
Hydrogeology II Dams and Embankments II
REDI (Special Session) Geoenvironmental III
Ground Improvement

Tuesday, October 3, 2023 1:30 pm - 3:00 pm

10:30 am - 12:00 pm Advanced Testing III
Geohazards III Case Studies II
Insitu Investigations Geohazards IV
Pipeline Geotechnics Innovative Geotechnical
Rock Mechanics II
1:30 pm - 3:00 pm
Numerical Models I
Rock Mechanics I

3:30 pm - 5:00 pm
Advanced Testing I
Contaminant Transport
Dams and Embankments I
Foundations II
Monday, October 2, 2023

FOUNDATIONS I
Preliminary results from field monitoring of
frost heave effects on steel H-piles
Joash Bryan Q. Adajar & Marolo C. Alfaro
University of Manitoba, Winnipeg, MB, Canada
Lukas U. Arenson
BGC Engineering Ltd., Vancouver, BC, Canada

ABSTRACT
Frost heave in soils during winter can significantly affect pile foundations. Piles are prone to experience significant upward
displacement due to frost heave forces. A part of these frost heave forces, acting on pile foundations in response to the
freezing of the surrounding soil, are contributed by adfreeze stresses along the pile shaft. Research is ongoing to improve
the understanding of adfreeze stresses on piles. This paper discusses a unique field instrumentation program designed to
quantify adfreeze stresses and to better understand the mechanisms involved with their development on piles due to frost
heave. Strain gauges were installed on the pile shaft to measure adfreeze stresses. Piezometers, thermistor strings,
magnetic sensors, Shape Accel Arrays, and heaving pins/plates were installed in the vicinity of the pile to monitor
parameters that can be beneficial in characterizing the mobilization of adfreeze stresses on piles. Preliminary data collected
from the first monitored winter period are also presented in this paper.

RÉSUMÉ
Le soulèvement par le gel des sols pendant l'hiver peut affecter considérablement les fondations sur pieux. Les pieux sont
susceptibles de subir un déplacement vers le haut important en raison des forces de soulèvement dû au gel. Une partie
de ces forces de soulèvement par le gel, agissant sur les fondations des pieux en réponse au gel du sol environnant, est
due aux contraintes de gel le long du fût du pieu. Actuellement, il existe une lacune dans la recherche dans la
compréhension des contraintes de gel sur les pieux. Cet article présente un programme unique d'instrumentation de terrain
conçu pour quantifier les contraintes de gel et pour mieux comprendre les mécanismes impliqués dans leur développement
sur les pieux. Des jauges de contrainte ont été installées sur le fût du pieu pour mesurer les contraintes de gel. Des
piézomètres, des chaînes de thermistances, des capteurs magnétiques, des réseaux Shape Accel et des broches/plaques
d'attrape ont été installés à proximité du pieu pour surveiller les paramètres qui peuvent être bénéfiques pour caractériser
le développement des contraintes de gel sur les pieux. Les données préliminaires recueillies à partir de la 1ère période
hivernale surveillée sont également présentées dans cet article.

1 INTRODUCTION as it still adopts the method from the Canadian Foundation

Engineering Manual, where values were derived from a
Piles are among the most common type of foundations study 30 years ago.
used for different structures in the province of Manitoba This paper presents the field instrumentation and
due to soft clay and silt deposits in the area. When soils monitoring program undertaken on a research test pile to
within the active layer freeze in winter, adfreezing occurs quantify adfreeze stresses in the field and to better
at the pile-soil interface. As frost heave develops, pile understand the mechanisms involved with the
foundations are prone to experience frost heave uplift development of adfreeze stresses on pile foundations in
forces due to adfreeze stresses along the pile shaft that cold regions. Among the methods available to measure
can be enough to induce significant upward movements, adfreeze stresses in the field, the strain gauge method
known as frost jacking. This uplift movement of the piles (SGM) was chosen as it presents the advantage of
decreases the overall stability of the structure and measuring adfreeze stresses at different sections of the
compromises their capacity. pile shaft (Johnson and Buska, 1988). The SGM also
Pile foundations in cold regions are designed and avoids underestimating measured adfreeze stresses as
built to withstand such adfreeze stresses. If these forces the monitored strains are isolated from the shaft resistance
are not properly accounted for during pile design, the below the active layer. The instrumentation and monitoring
structural integrity of the structure will be at risk, and failure program are discussed in the succeeding sections, and the
may occur once significant upward movement develops. preliminary data collected from monitoring during the first
There are different approaches presented in literature to winter period are presented.
estimate adfreeze in piles. Current available methods only
capture a small portion of the entire process in properly
measuring these stresses, and they show significant 2 SITE AND SUBSURFACE DESCRIPTION
variability. This variability requires a better understanding
of the mobilization of adfreeze stresses along the pile shaft. The project site is a 3x12 m research area in Brandon,
Additionally, the current approach in considering adfreeze Manitoba (49°51'N, 99°58'W). It is situated on the west side
stresses for pile design here in Manitoba seems outdated of Daly Overpass along the Provincial Trunk Highway
 Strain Gauge (SG) Piezometer (PZ) Temperature Sensor (TS)
Heaving Pin/Plate (HP) Snow Rod (SR) Magnetic Sensor (MHS)
Shape Accel Array (SAA) Camera

TOP VIEW

at 0.5 m sensor
interval

CROSS-SECTIONAL VIEW

Figure 1. Instrumentation list and plan (top and cross-sectional view).

(PTH) 10. Based on the geotechnical report of the area, 3 INSTRUMENTATION

which involves borehole logs, cone penetration tests, and
pressuremeter tests, the stratigraphy generally consists of The research test pile is a steel H-pile (HP310x110) with a
clay (till) soil. length of 17.5 m embedded into the ground. The instrument
Index tests have revealed that the moisture list and layout are shown in Figure 1. The purpose of each
content varies from 15% to 25%, with an average specific instrument is discussed in the succeeding sections. The
gravity of 2.69. Atterberg limits tests showed liquid limit instrumentation setup was developed to monitor the
values ranging from 32% to 45%, with plasticity indexes following parameters: i) adfreeze stress on piles; ii) depth
from 16 to 26. The soil can be classified as Lean Clay (CL) of groundwater table; iii) ground temperatures; iv) frost
with a frost susceptibility classification of F3 (medium to heave at the surface and at different depths; and v) depth
high) based on the U.S. Corps of Engineers Frost Design of snow cover.
Soil Classification.
3.1 Strain Gauges

Strain gauges were attached on the pile shaft to monitor

the changes in axial strains on the pile as the soil heaves.
Axial strains are converted to axial forces, which are then
used to calculate the equivalent adfreeze or tangential
heave stresses. A total of 10 strain gauges were installed
on the pile shaft. Two strain gauges were installed at each
depth for redundancy. Vibrating wire (VW) strain gauges
were used as they are significantly more robust compared
to other gauges. VW strain gauges have been used in
multiple studies to monitor stresses in piles induced by the
surrounding soil (Yang et al., 2006; Zhang and Wang,
2007; Budge and Dasenbrock, 2010, 2011; Drbe et al.,
2017; Bartz and Blatz, 2022). These strain gauges have
been recommended due to their longevity and high
survivability rate during pile driving, allowing longer
monitoring periods. Figure 2a shows how the VW strain
gauge was welded on the pile shaft. A welding blanket was
placed over the strain gauge before welding a protective
steel angle (3’’x3’’x1/8’’) (Figures 2b and 2c). The welding Figure 2. Installation of (a) strain gauge and placement of
blanket reduces the risk of burning or melting the strain (b) welding blanket and (c) steel angle as a protective
measure.
gauge and cables while welding the protective steel angle.
Installing a steel angle over the strain gauge is a protective
measure to reduce the potential for damage to the strain
gauge during pile driving (Drbe et al., 2017; Bartz and 3’’ x 3’’ x 3/8’’
Blatz, 2022). It was proven effective as all strain gauges Steel angle Cross-sectional
still function well after driving the pile. A schematic of the area: 0.0164 m2
cross-section of the pile after the installation of strain
gauges and steel angles is shown in Figure 3. The VW
Surface area per
strain gauge has a working range of 4000 microstrain (με) meter: 1.92 m2/m
and was initially set to experience a slight compression to
give it enough space to expand due to tensile uplift forces Strain Gauge
from adfreeze stresses. Steel HP 310x110

3.2 Piezometers Figure 3. Schematic diagram of the cross-section of the h-

pile after strain gauge and steel angle installation.
Vibrating wire piezometers were installed near the pile
(Figure 4). The plan was to install two piezometers in a
single borehole at different depths. The two piezometers
were supposed to monitor the upward flow of groundwater
at the frozen fringe. However, only one survived after
grouting the borehole. So far, we were able to track the
changes in the depth below ground surface of the
groundwater table.

3.3 Thermistor Strings

Thermistor strings were installed vertically, with the Figure 4. Vibrating Wire Piezometer.
temperature sensors positioned at different depths (Figure
5). A total of two thermistor strings were installed; one is
positioned closer to the pile, while the other is farther. Two
strings were installed to compare the difference in
temperature between the section close and far from the
pile. Temperature sensors provided information on
changes in ground temperatures with depth. This
information was used to determine the frost penetration
depth during the winter.

Figure 5. Thermistor Strings

3.4 Shape Accel Array (SAA) be analyzed from the photos to measure snow cover depth
at the time of the photo. Survey points were also attached
A four-meter Shape Accel Array (SAA) is horizontally to the snow rods. This is a redundancy of the SAA for
installed near the ground surface (0.15 m depth) and measuring surface heave close to the pile.
oriented radially outward from the pile. The interval length
per sensor in the SAA is 0.5 m. The purpose of the SAA is
to measure the deformation (heave and settlement) of the
soil surface during winter. It is installed radially to monitor
the soil deformation with distance from the pile. SAAs have
been used to measure ground movements with great
success in other research projects (Kurz et al., 2016; De
Guzman et al., 2021). Budge and Dasenbrock (2010), for
example, specifically used SAAs to measure deformation
around driven piles. A survey marker was placed on one Figure 7. Magnetic Sensors
end of the SAA. This marker was regularly surveyed to
track its vertical displacement and adjust the readings for
the other sensors along the SAA accordingly (Figure 6). As
a protective measure, the SAA was encased in a 1’’ PVC
pipe before burying. Each SAA sensor has a built-in
thermistor that provided information on the temperature
near the ground surface.

Figure 8. (a) Snow rod and (b) trail camera.

3.7 Frost Heave Plates and Pins

Complementary survey pins and plates were installed in

the surrounding area (Figures 9a and 9b). They were
installed at different depths (0.5, 1.0, 1.5, 2.0 m), where
their tip serves as a survey point. They serve as a
redundancy for the magnetic sensors as they were
surveyed regularly to monitor frost heave at different
depths.
Figure 6. Shape Accel Array (SAA) enclosed in a PVC pipe
with a survey marker.

3.5 Magnetic Sensors

A magnetic heave system was installed near the pile.

Magnetic sensors were attached around an inclinometer
casing at different depths (Figure 7). The position of the
magnetic sensors was tracked during winter to monitor
frost heave at different depths within the active layer.

3.6 Snow Rods and Trail Camera

Snow rods are simple steel rods with yellow and red
reflective stickers attached (Figure 8a). These snow rods
were utilized to measure and monitor the depth of snow
accumulation/cover on the research site. A trail camera
was installed in addition to the snow rods (Figure 8b). The
trail camera regularly took photos of the site to monitor
changes in snow cover. Reflective sticker marks can easily Figure 9. Frost Heave (a) plates and (b) pins.
3.8 Data Acquisition System 0
0.2

Depth below ground

Strain gauges, piezometers, temperature sensors, and 0.4
Shape Accel Array (SAA) were all connected to a data 0.6

surface (m)
acquisition system to collect readings at a specific interval. 0.8
All cables connected from the mentioned instruments were 1
run through a PVC pipe to serve as an additional protective
1.2
measure before burying. Magnetic sensor reading and
1.4
general surveying activities were performed at least once
a month during winter. 1.6
1.8
3.9 Research Site 2

A photo of the research site after all the instruments have

been installed is shown in Figure 10. The research site was Figure 11. Changes in depth below ground surface of
cleared of snow every site visit to maximize the heat loss groundwater table over time.
of the ground and promote maximum frost penetration.
Temperature ( C)
-10 -5 0 5 10 15 20 25
0

4
Max frost depth
around 1.5 m
6
Depth (m) 8

10

12

14
7/30/2022 0:00 8/30/2022 0:00 9/30/2022 0:00
16 10/30/2022 0:00 11/29/2022 9:00 12/18/2022 0:00
Figure 10. Research site during winter. 1/21/2023 0:00 2/17/2023 0:00 3/31/2023 9:00
18
Figure 12. Ground temperatures with depth.
4 PRELIMINARY FIELD MONITORING RESULTS
Figure 13 shows the air temperature recorded from
Presented below are some of the results from the first the weather station near the site, compared with the
winter monitoring period (2022 – 2023). Most recent measured ground temperature closest to the surface,
reading included was March 31, 2023. which is 0.15 meters below ground surface (mbgs). The
apparent delay of heat loss/extraction or change in
4.1 Groundwater Depth temperature on the ground compared to air temperature
can be observed as the first negative mean air temperature
The change in the depth below ground surface of the was recorded on November 2, while the first negative
groundwater table over time is plotted in Figure 11. The ground temperature was recorded on November 16. This
decrease in depth during seasons before winter matches time lag can be amplified if thicker snow cover develops on
the instances of significant rainfall, confirming that the the site.
piezometer is functioning well. The groundwater table was
40
observed to gradually become deeper during winter. Air Temperature
30
Ground Temp (0.15 mbgs)
4.2 Temperature Readings
Temperature ( C)

20
10
Figure 12 shows the graph of temperatures at different
depths. They are plotted at different times/dates. The 0
trumpet-shaped curves represent the maximum variation -10
or the range of subsurface temperatures observed during
-20
the monitoring period. As we get closer to the height of
winter, plots start to lean more to the left or toward negative -30
temperatures. From the graph, the maximum frost -40
penetration depth was identified to be around 1.5 m.

Figure 13. Comparison of air and ground temperatures.

4.3 Frost Heave

Surface heave measured at different distances from the

pile is shown in Figure 14. Based on the most recent
reading, the maximum heave ranges from 1.6 to 2.2 cm.
The general trend is that surface heave is higher at
locations farther from the pile. The existence of the pile and
the development of adfreeze bond and stresses at the
frozen soil-pile interface provides resistance against the
upward movement of the soil layer close to the pile.

3
29-Nov-22 18-Dec-22 4-Jan-23
21-Jan-23 28-Feb-23 31-Mar-23
2.4
Surface Heave (cm)

1.8

1.2

0.6

0
away from pile
-0.6
0.00 1.00 2.00 3.00 4.00 5.00
Distance from the Pile (m)
Figure 14. Surface heave at different distances from the
pile. Figure 16. Trail camera photos showing (a) before and (b)
after a significant snowfall event.
Figure 15 shows the measured surface heave in
comparison to the heave measured at a depth of 0.5 m. No 4.5 Adfreeze Forces and Stresses
heaving was measured at deeper locations despite the
frost depth penetrating as deep as 1.5 m. Survey data and Adfreeze stresses were determined using the internal axial
magnetic sensor readings are still being analyzed to verify strains measured from the strain gauges installed on the
and confirm these initial frost heave measurements. pile shaft (Johnson and Buska, 1988). First, the internal
axial stresses were determined using the axial strains. The
2.5 axial forces inducing those axial stresses were then
Surface 0.5 mbgs calculated. Lastly, the axial force values were utilized to
2 calculate the equivalent adfreeze stresses along specific
Frost Heave (cm)

portions of the pile shaft. It is important to note that the

cross-sectional area and surface area utilized in the
1.5
calculations included the attached steel angle on the H-pile
to protect the strain gauges (Figure 3).
1
160
SG01 (0.5 m)
0.5 140
SG02 (1 m)
Axial Forces (kN)

120 SG03 (1.5 m)

0 100 SG04 (2.25 m)
SG05 (3 m)
80
60
Figure 15. Frost heave at different depths.
40
20
4.4 Snow Accumulation 0
-20
Figure 16 shows the difference between the trail camera -40
photos taken when there is no snow accumulation and after
a significant snowfall. Using the snow rods, the maximum
snow cover depth was identified as 20 cm. Snow cover did Figure 17. Axial forces calculated from recorded strains.
not accumulate higher than this, as the site was cleared of
snow during visits to maximize frost penetration depth in Figure 17 shows the calculated axial forces based
the area. on the strain gauge readings. Temperature corrections
were applied to account for the thermal
expansion/compression of steel. Positive and negative
force values represent tensile and compressive forces, The temperature, axial force, and tangential
respectively. The increase in axial forces is apparent during stress distribution on the pile with depth at specific times
winter. This indicates that the pile is experiencing uplift are all plotted in Figure 18. Figures 18a to 18c present
adfreeze forces as the soil heaves. The first noticeable information from December 26, 2022. The temperature
increase in axial force occurred on November 20, 2022, a graph (Figure 18a) shows that frost penetration depth is 1
few days after the first negative ground temperature was mbgs. The blue shade in the graphs represents the depth
recorded. Peak axial force was measured on February 4, of frost penetration. The total axial force increased within
2023, which coincides with the instance when the coldest the frost depth and eventually decreased in the unfrozen
ground temperature was recorded. Axial forces start to zone below (Figure 18b). This increase in axial force was
decrease as the end of March 2023 approaches. caused by the development of adfreeze stresses at the
frozen soil-pile interface and soil frost heave within the frost

Dec. 26, 2022

Temperature ( C) Total Axial Force (kN) Tangential Stress (kPa)
-15 -10 -5 0 5 10 -100 -50 0 50 100 150 200 -100 -50 0 50 100 150
0 0 0.00 0
69.14
0.5 0.5 66.38 0.5
22.99
1 1 88.45 1
-12.29
Depth (m)

Depth (m)

Depth (m)
1.5 1.5 76.65 1.5

-9.75
2 2 2
62.61
2.5 2.5 2.5
-22.08

3 3 30.80 3 Uplift Stress

(a) Temperature (b) Total Force (c) Resisting Stress
3.5 3.5 3.5

Feb. 4, 2023
Temperature ( C) Total Axial Force (kN) Tangential Stress (kPa)
-15 -10 -5 0 5 10 -100 -50 0 50 100 150 200 -100 -50 0 50 100 150
0 0 0.00 0
103.70

0.5 0.5 99.56 0.5

51.73
1 1 149.23 1
-14.17
Depth (m)

Depth (m)
Depth (m)

1.5 1.5 135.62 1.5

-13.24
2 2 2
116.56
2.5 2.5 2.5
-30.86

3 3 72.12 3 Uplift Stress

(d) Temperature (e) Total Force (f) Resisting Stress
3.5 3.5 3.5

Mar. 15, 2023

Temperature ( C) Total Axial Force (kN) Tangential Stress (kPa)
-15 -10 -5 0 5 10 -100 -50 0 50 100 150 200 -100 -50 0 50 100 150
0 0 0.00 0
67.63
0.5 0.5 64.93 0.5
23.54
1 1 87.54 1
-10.86
Depth (m)

Depth (m)
Depth (m)

1.5 1.5 77.11 1.5

-10.34
2 2 2
62.22
2.5 2.5 2.5
-31.13

3 3 17.39 3 Uplift Stress

(g) Temperature (h) Total Force (i) Resisting Stress
3.5 3.5 3.5

Figure 18. Temperature, axial force, and tangential stress distribution along the pile shaft with depth at different times:
(a-c) Dec. 26, 2022, (d-f) Feb. 4, 2023, and (g-i) Mar. 15, 2023.
depth. The calculated tangential stresses on the pile shaft 6 ACKNOWLEDGMENT
indicate that adfreeze stresses (uplift) exist within the frost
depth, and resisting stresses below (Figure 18c). This trend This research work was funded by Research Manitoba.
of axial force and tangential stress in relation to the frost The authors would like to acknowledge the personnel,
depth was observed most times during winter. technicians, and engineers from Manitoba Transportation
A peak adfreeze stress of 103.7 kPa was and Infrastructure (MTI), TREK Geotechnical Inc., Green
recorded on February 4, 2023 (Figure 18f). This was Infrastructure Partners Inc. (GIPI), and RST Instruments
calculated within the soil layer closest to the ground surface Ltd. for their assistance and expertise in the field
(0 – 0.5 mbgs). The average adfreeze stress when instrumentation activities. The services of Kerry Lynch,
considering the entire frost depth in the calculation is 77.8 Liam Soufi, Sohit Garg, and Marvin Malonzo on all the
kPa. It is worth noting that during this time, despite the fieldwork done are also highly appreciated.
monitored frost depth being at 1.25 mbgs, no increase in
axial force (no uplift adfreeze stress) was observed in the 7 REFERENCES
soil layer between 1.0 to 1.5 mbgs. A possible reason is
that despite the development of adfreeze bonds in the Aldaeef, A. A. and Rayhani, M. T. 2019. Interface shear
area, the soil layer might not have heaved yet. Additionally, strength characteristics of steel piles in frozen clay
we are evaluating the possibility that this may be related to under varying exposure temperature, Soils and
the calibration and temperature correction factors used for Foundations. Japanese Geotechnical Society, 59(6):
the force and stress calculations. 2110–2124.
On March 15, 2023, as ground temperature Bartz, J. R. and Blatz, J. A. 2022. Considerations for
increase, axial forces and adfreeze stresses eventually measuring residual stresses in driven piles with
decrease (Figure 18g to 18i). There was a decrease in axial vibrating wire strain gauges, Canadian Geotechnical
force and adfreeze stress despite some frost heave still Journal, 59(3): 441–446.
being recorded nearby. This can imply that the soil freezing Budge, A. S. and Dasenbrock, D. D. 2011. Performance
temperature significantly impacts the magnitude of Data Collected from Instrumentation on a Mn/DOT
adfreeze stresses at the frozen soil-pile interface. Soil Bridge Abutment Foundation Subject to Downdrag,
freezing temperatures may have affected the lateral earth Geo-Frontiers 2011, ASCE, Reston, VA, USA, 273–
pressure applied on the pile shaft by the frozen soil; which, 282.
in turn, affected the magnitude of adfreeze stresses as they Budge, A. S. and Dasenbrock, D. E. 2010. Installation of
are sensitive to the normal pressure applied at the interface Downdrag Instrumentation on a Bridge Abutment
(Ladanyi and Theriault, 1990; Aldaeef and Rayhani, 2019). Foundation: Lessons Learned, GeoFlorida 2010,
The shear rate and displacement may also play a role in ASCE, Reston, VA, USA, 1237–1245.
this observed force and stress decrease. Drbe, O. et al. 2017. Instrumentation Monitoring Program
to Measure the Magnitude, Distribution and Time
Dependency of Drag Load on Abutment Piles: A Case
5 SUMMARY AND CONCLUSIONS Study, GeoOttawa2017, CGS, Ottawa, ON, Canada.
De Guzman, E. M. B. et al. 2021. Performance of highway
An instrumentation program was developed to quantify embankments in the Arctic constructed under winter
adfreeze stresses on pile foundations in the field and conditions, Canadian Geotechnical Journal, 58(5):
identify parameters that may characterize their mobilization 722–736.
on piles. After one winter, the installed instruments have Johnson, J. B. and Buska, J. S. 1988. Measurement of frost
reasonably worked well in determining and monitoring heave forces on H-piles and pipe piles, CRREL-88-21,
adfreeze stress, depth of groundwater table, ground U.S. Army Cold Regions Research and Engineering
temperatures, frost heave at the surface and at different Laboratory.
depths, and snow cover. For the first monitored winter Kurz, D. et al. 2016. Seasonal deformations under a road
period, a maximum frost penetration depth of 1.5 mbgs was embankment on degrading permafrost in Northern
recorded based on the temperature readings. A surface Canada, Environmental Geotechnics, 7(3): 163–174.
heave of 1.6 to 2.2 cm was measured at different radial Ladanyi, B. and Theriault, A. 1990. A Study of Some
distances from the pile. Smaller surface heave was Factors Affecting the Adfreeze Bond of Piles in
measured closer to the pile as the pile resisted upward Permafrost, Geotechnical Engineering Congress GSP,
displacement. Peak and average adfreeze stress values of 213–224.
103.7 kPa and 77.8 kPa were calculated, respectively. The Yang, J. et al. 2006. Observed Performance of Long Steel
peak value is close to the recommended values of the H-Piles Jacked into Sandy Soils, Journal of
Canadian Foundation Engineering Manual (CFEM) for Geotechnical and Geoenvironmental Engineering,
steel piles, which is 100 kPa. Further analysis is currently 132(1): 24–35.
being done on the collected field data. Instrumentation Zhang, L. M. and Wang, H. 2007. Development of Residual
monitoring will continue in the succeeding years to get Forces in Long Driven Piles in Weathered Soils, Journal
more data and better understand the mobilization of of Geotechnical and Geoenvironmental Engineering,
adfreeze stresses on pile foundations. 133(10): 1216–1228.
Use of CFA piles in Southern Ontario –
review and case studies
Laifa Cao
WSP, Toronto, Ontario, Canada

ABSTRACT
High-capacity continuous flight auger (CFA) piles have been successfully applied to support building foundations in
Southern Ontario. The paper reviews the design methods for CFA piles and summarizes results of pile load tests published
in the literature. Two compressive load tests conducted in Markham are presented. The ultimate resistance of CFA piles
estimated from semi-empirical analysis are compared with that measured from the pile load tests. Methods used to
estimate the factored resistances of CFA piles at the ultimate limit states (ULS) and serviceability limit states (SLS) are
recommended.

RÉSUMÉ
Des pieux à tarière à vol continu (CFA) de grande capacité ont été appliqués avec succès pour soutenir les fondations de
bâtiments dans le sud de l'Ontario. L'article passe en revue les méthodes de conception des pieux CFA et résume les
résultats des essais de charge de pieux publiés dans la littérature. Deux essais de charge en compression effectués à
Markham sont présentés. La résistance ultime des pieux CFA estimée à partir d'une analyse semi-empirique est comparée
à celle mesurée à partir des essais de charge de pieux. Les méthodes utilisées pour estimer les résistances pondérées
des pieux CFA aux états limites ultimes (ELU) et aux états limites de service (ELS) sont recommandées.

1 INTRODUCTION 2 REVIEW OF CFA PILES IN SOUTHERN ONTARIO

The need for pile foundations results from soft or loose soil CFA piles are cast-in-situ concrete piles. During
conditions at shallow depth that would result in installation, a hollow-stem continuous flight auger is rotated
unacceptable settlement or inadequate bearing resistance and pushed into the ground at controlled speeds to the
from shallow foundations. The pile size and length are target depth and then concrete/grout is continuously
designed based on the ultimate limit states (ULS, which pumped under a positive pressure through the hollow stem
considers the load resistance) and serviceability limit states of the auger to fill the borehole from auger toe during auger
(SLS, which considers the deformations or settlements), retrieval. A reinforcing cage is lowered by its self-weight
and the selection of the pile type is usually based on into the fresh concrete/grout to the designed depth once
subsurface conditions as well as local experience and auger retrieval.
practice. When the site is accessible, removal and disposal With availability of self-compacting concrete and high
of spoils material generated from pile installation is strength grout, development of powerful CFA drill rigs,
allowable, and strict quality control and supervision are application of data acquisition system, 500 to 610 mm
available, continuous flight auger (CFA) piles can be diameter CFA piles installed 9 to 29 m below pile cap level
considered due to several advantages such as the have been successfully used as deep foundations to
relatively low cost, faster installation, higher load capacity, support high-rise buildings in Southern Ontario.
and lower levels of vibration and noise.
High-capacity CFA piles, which have been successfully 2.1 CFA Pile Resistance
installed in Southern Ontario, could not be designed using
conventional methods. The paper reviews the design For the design of CFA piles, the geotechnical pile ultimate
methods for CFA piles and summarizes results of pile load resistance or capacity (Ru) is usually estimated by
tests published in the literature. Two compressive load summing the shaft friction and end bearing capacity as
tests conducted in Markham are presented. The ultimate follows:
resistances of CFA piles estimated from semi-empirical
analysis are compared with that measured from the pile Ru = qsAs + qtAt - Wp [1]
load tests. Methods used to estimate the factored
resistances of CFA piles at the ultimate limit states (ULS) where qs is unit shaft friction, As is pile embedded shaft
and serviceability limit states (SLS) are recommended. area, qt is end bearing capacity, At is pile toe area, and Wp
is pile weight.
The qs value may be estimated from effective in-situ
vertical stress (σ’v) times shaft resistance factor (β). The β
value is related to the soil type and strength, ranging from
0.25 to 1.2 (FHWA, 2007). For cohesive soil, qs may be
estimated from the undrained shear strength (su) times sand deposits. The groundwater table was at
adhesion coefficient (α). The α value is 0.55su for su ≤ 150 approximately 3 m below the existing grade. A 510 mm
kPa and varies linearly from 0.55su to 0.45su for su values diameter CPA pile with 35 MPa grout (1 part of Portland
varying from 150 kPa to 250 kPa (FHWA, 2007). cement, 2 parts of sand, and 2 parts of water) was installed
The qs value may be equal to 57.3N60 for N60 ≤ 75 and 12.7 m below the basement level including 6.7 m into the
4300 kPa for N60 > 75 for cohesionless soils, where N60 is dense to very dense sandy silty to silty sand deposits and
is the standard penetration test (SPT) N-value at 60% tested to a maximum load of 3800 kN with a settlement at
hammer efficiency. For cohesive soil, the qb value might be pile top of 11.6 mm. The creep settlement of the pile under
estimated from su times the bearing capacity factor (Nc) the maximum test load was 3.2 mm over a period of 18
values. The Nc ranges from 6.5 to 9, increasing with su hours. The settlement at the maximum test load is less than
values varying from 25 kPa to 200 kPa (FHWA, 2007). 26 mm, i.e. 5% of the pile diameter (FHWA, 2007) and thus
The qs and qb values might also be estimated from SPT test pile was not considered failure under the maximum
N vales using Decourt’s method (1995) recommended in load.
CGS (2006) as follows: Almeida, et al. (2021) presented seven load test results
on sever CFA piles at four sites as follows:
qs = α(2.8N60 +10 ) (kPa) [2] (1) At a 11-sotorey building site in Markham, the
subsurface consists of 3.6 to 6.6 m thick fill and
qb = KbNb [3] very soft to stiff silty clay till over very loose to very
dense silty sand till. 500 mm diameter CFA piles
where α is equal to 1 for clay and 0.5 to 0.6 for granular were installed 18 to 20 m below grade to support
soil, N60 is the average SPT N-value at 60% hammer the building. Two static load tests were conducted
efficiency along pile shaft, Nb is the average SPT-N value on 17 m long CFA piles to maximum test loads of
at 60% hammer efficiency in the vicinity of the pile toe, and 6000 and 7250 kN, respectively with settlement at
Kb is 165 for sand, 115 for sandy silt, 100 for clayey silt and the pile top of 36 and 20 mm. The test pile with a
80 for clay. settlement of 36 mm was considered as failure.
The CFA pile resistance should be confirmed by static This pile settled 25 mm (5% of the diameter) at the
load tests due to uncertainty in the soil parameters used to test load of 5000 kN which was considered as the
estimate the pile resistance. The static load test procedure ultimate resistance.
and set-up for CFA pile generally follow the test methods (2) At a 29-storey building site in Toronto, 500 mm
recommended by ASTM (2020). The acceptance criterion diameter CFA piles were installed to shale bedrock
for CFA pile static load test is that the geotechnical ultimate to support the podium foundations. The subface
limit state is defined as the load resistance at a consists of 0.7 to 6.8 m thick very soft to firm/loose
displacement equal to 5% of the pile diameter (FHWA, fill over stiff to hard silty clay underlain by shale
2007), considering that the skin friction capacity is fully bedrock. Static load test on a 15.2 m long CFA pile
mobilized with relatively small pile settlement, typically 5 to was tested to a maximum load of 3000 kN, at which
10 mm and the settlement required to fully mobilize toe the settlement at pile top was 9.2 mm.
capacity can be assumed to be approximately 5% of the (3) At a 21-storey building site in Toronto, a firm to
pile diameter (Reese and O’Neil, 1988; AASHTO, 2006). hard (SPT N-value of 5 to 36) clayey silt extended
Davisson criterion (or offset limit load) commonly used for to a depth of 11 m below the ground surface and
driven pile, which is sometimes underestimate the ultimate underlain by a dense to vey dense (SPT N-value
resistance, is not appropriate for CFA piles. It should be of 33 to greater than 50) sand till with occasionally
noted that the pile resistance at the geotechnical ultimate interbeds of loose to compact (SPT N-value of 6 to
limit state in FHWA (2007) is not considered as the factored 19) sand. The groundwater level was near the
pile resistance at the ULS. A resistance factor of 0.6 should ground surface. Static load test was conducted on
be used for a factored ULS resistance if a pile load test is 600 mm diameter CPA piles installed 21 to 29 m
conducted as recommended in CGS (2006). below ground surface, respectively. A 6 m de-
The creep criterion recommended by ASTM (2020) is bonded casing was installed in the upper section
generally applied for CFA piles, which requires maintaining for the 21 m long to account for the effective pile
the maximum test load for 12 hours until the axial length of 15m. The maximum test load was 6000
movement measured over a period of 1 hour does not kN for the two load tests and the settlement at the
exceed 0.25 mm. pile top was 28 and 21 mm, respectively for the 15
and 21 m long CFA piles. The settlement at the
2.2 Published Static Pile Load Tests maximum test load included 4.5 mm and 2.5 mm
creep movement from 1 to 60-minute maintenance
Zhu and Peaker (2018) presented a case study using CFA for the 21 and 29 m long CFA piles, respectively.
piles to support a multi-storey residential building with one (4) At Barrier site, the subsurface soils consist of 8 m
level of basement approximately 3 m below ground surface thick loose to compact sand over dense to vey
in Markham. The soils below the basement level consisted dense sand. 500 mm diameter CFA piles were
of a 6 m thick layer of firm (standard penetration test, SPT installed into the dese to very dense sand to a 10-
N-value ranging from 3 to 6 and undrained shear strength sorey building. Two static load tests were
ranging from 24 to 52 kPa) silty clay deposits overlying conducted on 9 and 15 m long CFA piles to a
dense to very dense (SPT N-value > 30) sandy silty to silty maximum test load of 4500 kN, at which the
 settlement at the pile top was 37 and 20 mm, (SPT N values of 1 to 20 and field vane su of 45 to 60 kPa)
respectively for the 9 and 15 m long CFA piles. The silty clay to clayey silt. The silty clay to clayey silt extends
9 m long pile settled 25 mm (5% of the diameter) to 6.1 to 11.6 m below the existing ground surface, which
at the load of 3900 kN which was considered as are underlain by very loose to very dense (SPT N values of
the ultimate resistance. 0 to greater 50) sandy till and/or very soft to hard (SPT
Table 1 summarizes the CFA pile resistance based on values of 0 to greater than 50) clayey till. The till deposits
published results of static load tests on CFA piles (Zhu and are interbedded with and underlain by compact to very
Peaker, 2018; Almeida, et al. 2021) as described dense silty sand to sand and gravel. The soil deposits at a
previously and the cases described in this paper. The depth of 13.5 m below the existing ground surface are
factored resistances at the ultimate limit state (ULS) using generally in a dese to very dens compactness condition or
a resistance factor of 0.6 as recommended in CGS (2006) a hard consistency. The groundwater table lies at 4.1 to 5.4
are also shown in Table 1. m below the existing ground surface.
Since the shallow soil deposits could not support the
building foundations, deep foundations were commended.
Table 1. CFA Pile Resistances in South Ontario Driving H-piles were not recommended due to the concern
of noise/vibration and ground settlement caused by pile
Location L (m) D (m) TL (KN) STL R (kN) SR driving. Drilled caissons were also not recommended due
(mm) (mm) to less control of end bearing capacity under high
Markham 12.71 510 3800 14.8 >2280 <8 groundwater table during caisson installation.
17.02 500 6000 36.0 3000 10 Two 610 mm diameter CFA piles was installed for static
17.02 500 7250 22.0 >4350 <12 load tests. Test pile No. 1 installed 27.5 m below grade was
27.53 610 7200 12.4 >4320 <8
tested to a maximum load of 7200 kN and test pile No. 2
22.03 610 5000 38.8 3000 8
Toronto2 15.22 500 3000 9.2 >1800 <4 installed 22 m below grade was tested to a maximum load
15.02 600 6000 28.0 3600 8 of 5000 kN. The maximum test load is 2 times the design
29.02 600 6000 21.0 >3600 <6 load that was estimated based theoretical analysis. The
Barrie 92 500 4500 37.0 2340 14 static load testing was performed generally in accordance
152 500 4500 20.0 >2700 <8 with ASTM (2020) using maintained test procedure. Load
increments/decrements were 900 kN and 625 kN (i.e. 25%
L: Pile length; D: Pile diameter, TL: maximum test load; STL: Pile of design load) for test pile No. 1 and test pile No. 2,
settlement on TL; R: Factored resistance at ULS using a respectively. Each load was maintained ranging from 20 to
resistance factor of 0.6; SR: Pile settlement on R 80 minutes, generally in accordance with the maintained
1
Zhu and Peaker (2018)
2
Almeida et al. (2021)
test method of ASTM (2020). For test pile No. 1, the test
3
Cases described in this paper load of 6300 kN (175% of design load) was maintained for
12 hours due to a concern of structural stability of steel test
beam and then tested to the maximum load of 7200 kN
The ultimate resistances obtained from these pile load maintained for 12 hours. For test pile no. 2, the maximum
tests summarized in Table 1 are significantly higher than test load of 5000 kN was maintained for a period of 20
the theoretical estimations using the method hours.
recommended in FHWA (2007). This could be due to Figure 1 shows the results of compressive load test on
following two reasons: test pile No. 1. During loading, the measured total
(1) The volume of grout is usually greater than that of settlement on pile top under a test load of 6300 KN was
the augered hole, leading the effective pile 11.6 mm including a creep movement of 2.3 mm over a
diameter and toe area greater than the designed
values; and
(2) Low-pressure injection is usually used for CFA pile
installed, increase the soil strength along pile shaft
as well as below pile toe in sandy soils.
It is also noted that the pile settlement at the factored
ULS pile resistance less than 15 mm. This settlement is
less than the allowable settlement of 25 mm at the
serviceability limit states (SLS). Thus, the ULS resistance
controls the design of CFA piles.

3 CASE STUDY

This site was redeveloped with mixed-use twenty-four to

twenty-eight story residential buildings with one
underground level for parking and amenities in Markham,
Ontario. Subsurface investigation using drilled boreholes
indicates that the site stratigraphy is made up of 1.4 to 1.8 Figure 1. Results of static compressive load test of test pile
m thick recent deposits of fill overlying very soft to very stiff No. 1 in Markham
period of about 16 hours. Under the maximum load of 7200 theoretical ultimate resistances of a single CFA pile are
kN, the settlement was 12.3 mm. The maximum load was compared with the ultimate resistances obtained from pile
maintained for 12 hours. During first hour, the creep load test and summarized in Table 2.
movement was 1 mm. No creep movement was measured
during the following 2 hours and then the total settlement For the 12.7 m long CFA presented by Zhu and Peaker
decreased from 12.4 mm to 11.3 mm) during the remained (2018), the ultimate resistance (RuFHWA) estimated using
9 hours. The total settlement under the maximum test load soil parameters recommended in FHWA is significantly
is less than 5% of pile diameter, indicating the pile load test lower than the ultimate resistance (Rum) that measured in
did not reach the geotechnical ultimate resistance of the the pile load test; and the ultimate resistance (RuCGS) using
tested CFA pile. The movement of pile head was less than the correlation of pile resistance with SPT N values for
the pile elastic compression under the load test load, displacement pile developed by Decourt (1995) as
indicating the effective pile diameter could be greater than recommended in CGS (2006) is slightly lower than Rum.
the design value. Zhu and Scott (2018) also noted that the estimated from
Figure 2 shows the results of compressive load test on the method recommended in FHWA (2007)
test pile No. 2. During loading, the measured total underestimated the ultimate resistance.
settlement on pile top under the maximum test load of 5000 For the 17 m long CFA pile presented by Almeida, et al.
KN was 38.8 mm including a creep movement of 14.6 mm (2021), the value of RuFHWA is lower than the Rum that
over a period of 21 hours. The total settlement under the measured in the pile load test at pile settlement of 5% of
maximum test load is greater that 5% of pile diameter of the pile diameter; and the value of RuCGS is comparable with
30.5 mm, indicating the pile load test reach the the value of Rum.
geotechnical ultimate resistance of the tested CFA pile. For the 22 m long CFA pile presented in this paper, the
During the 12 hours while maintaining the maximum test value of RuFHWA is comparable with the value of Rum,
load, the rate of settlement of 0.7 mm/hour is greater than whereas the value of RuCGS is sightly higher than the value
0.25 mm/hour, indicating the pile was failure during the of Rum.
creep test. The movement of pile head was less than the Note that a resistance factor of 0.4 is used for the Ru
pile elastic compression under the load test load, indicating estimated from theoretical analysis and 0.6 for the Ru
the effective pile diameter could be greater than the design obtained from the load the test at the ULS. The factored
value. The settlement of the pile head was less than the resistance at the ULS estimated from the SPT N value
theoretical elastic compression of the pile when the test using Decourt’s correlation (1995) can be used for the
load was less than 3125 kN, indicating the effective pile preliminary design, whereas the factored resistance at the
diameter might not be greater than the design value for test ULS estimated from the method recommended in FHWA
pile no. 2. (2007) is generally underestimated the factored resistance
at the ULS.

Table 2. CFA Pile Resistances in South Ontario

Location L (m) D (m) Rum RuFHWA RuCGS

(KN) (kN) (kN)
Markham 12.71 510 3800 2100 3520
17.02 500 5000 3105 5020
22.03 610 5000 5250 6250
L: Pile length; D: Pile diameter, Rum: ultimate resistance measured
from pile load test; RuFHWA: ultimate resistance estimated using the
method recommended in FHWA (2007); RuFHWA: ultimate
resistance estimated using the method recommended in CGS
(2006) for displacement piles.
1
Zhu and Peaker (2018)
2
Almeida et al. (2021)
3
Case described in this paper

Figure 2. Results of static compressive load test of test pile

No. 2 in Markham 5. CONCLUSIONS

This paper reviews the published results of static load tests

Comparing the results of test pile Nos. 1 and 2, it is on CFA pile at varies sites and presents the results of two
found that the CFA pile capacity is dependent on the static load tests at one site in Southern Ontario. The
installation quality. following conclusions and recommendations can be made
for CFA pile design:
4 DISCUSSIONS (1) A factored resistance of 2000 kN in ULS can be
considered for a 500 mm dia. CFA pile installed 5
Based on available borehole information and according to m into dense to very dense soil.
the methods presented in Section 2.1, the estimated
 (2) A factored resistance of 3000 kN in ULS can be
considered for a 600 mm dia. CFA pile installed 5
m into dense to very dense soil.
(3) The settlement under the load of factored
resistance at the ULS for 500 to 610 mm diameter
CFA piles is generally less than 15 mm. The
resistance at the SLS is not control the design of
CFA piles. The resistance at the SLS can be
taken as the same as the factored at the factored
resistance at the ULS.
(4) The factored ultimate resistance of the CFA at the
ULS estimated from the method recommended in
FHWA (2007) is generally underestimated.
(5) The factored resistance at the ULS estimated
from the SPT N value using Decourt’s correlation
(1995) for displacement pile can be used for the
preliminary design,

REFERENCES

AASHTO, 2006. LRFD Bridge Design Specifications, 4th

Edition, American Association of State Highway and
Transportation Officials, Washington, D.C.
Almeida, A., Gurpersaud, N., and Lee, H. 2021.
Application of High-capacity Continuous Flight
Augercase (CFA) Pies in the Greater Toronto Area,
GeoNiagara 2021, Niagara, Canada.
ASTM, 2020. Standard Test Methods for Deep
Foundations under Static Axial Compressive Load,
D1143/D1143M–20, American Society of Testing and
Materials, Philadelphia, PA.
Canadian Geotechnical Society (CGS), 2006. Canadian
Foundation Engineering Manual. 4th Edition, BiTech
Publishers Ltd., Vancouver, 488 p.
Decourt, L. (1995). Prediction of load-settlement
relationships for foundations on the basis of the
SPT, Cieco de Conferencias Internationale,
Leonardo Zeevaert, UNAM, Mexico, pp. 85-104.
FHWA, 2007. Design and Construction of Continuous
Flight Auger (CFA) Piles, Geotechnical
Engineering Circular No. 8, Report No. FHWA-HIF-
07-03, Federal Highway Administration (FHWA),
U.S. Department of Transportation.
Reese, L. C., and O’Neill M. W. 1988. Drilled Shaft:
Construction Procedures and Design Methods,
FHWAHI-88-042, Federal Highway Administration,
Washington, D.C.
Zhu, F. and Peaker, S. 2018. Field Load Testing and
Integrity Tests of CFA Piles in Sandy Deposits,
GeoEdmonton 2018, Edmonton, Canada.
Instrumentation and Associated Monitoring
Program to Measure Drag Force on Driven
H-Piles Subjected to Negative Skin Friction
Sepehr Chalajour, James A. Blatz
University of Manitoba, Winnipeg, Manitoba, Canada

ABSTRACT

The need for infrastructure has increased due to population growth to facilitate economic activities. Piles transfer
infrastructure loads to competent ground layers, especially where the surface soil lacks adequate bearing capacity. One
of the challenges that may arise in cohesive soils is negative skin friction (NSF), occurring due to the settlement of the
surrounding soil relative to the piles. NSF results in an additional axial force on the pile, referred to as drag force which
has caused several failure cases in pile-supported structures. This study discusses an instrumentation program, data
monitoring and analysis on a 36-meter driven H-Pile located at the edge of the bridge abutment’s foundation on an active
project to arrange a measurement approach to better understand the load transfer mechanism along the pile. Associated
interpretation, data corrections, and lessons learned are discussed to provide an approach to calculate the axial force
distribution.

RÉSUMÉ

Le besoin d'infrastructures a augmenté en raison de la croissance démographique afin de faciliter les activités
économiques. Les pieux transfèrent les charges d'infrastructure aux couches de sol compétentes, en particulier là où le
sol de surface manque de capacité portante adéquate. L'un des défis qui peuvent survenir dans les sols cohésifs est la
friction négative de la peau (FNP), qui se produit en raison du tassement du sol environnant par rapport aux pieux. La FNP
entraîne une force axiale supplémentaire sur le pieu, appelée force de traînée, qui a causé plusieurs cas d'échec dans les
structures supportées par des pieux. Cette étude porte sur un programme d'instrumentation, de surveillance des données
et d'analyse sur un pieu H enfoncé de 36 mètres situé sur le bord de la fondation de l'about de pont sur un projet en cours
pour mettre en place une approche de mesure afin de mieux comprendre le mécanisme de transfert de charge le long du
pieu. L'interprétation associée, les corrections de données et les leçons apprises sont discutées pour fournir une approche
de calcul de la distribution des forces axiales.

1 INTRODUCTION MnDOT (2021) classifies induced drag force into two

loading conditions: unfavorable and favorable. Unfavorable
In Manitoba, there are many locations where bridges are drag force conditions occur when construction excavation
built on sites with compressible native soft clay soil and backfill areas for pile caps, side-hill or embankment fills
deposits. Deep foundation units, such as driven piles or (less than 1.2m), and bridge pier construction are small,
drilled shafts, are typically used to support these bridges. resulting in a low magnitude of drag force. On the other
These units rely on axial resistance from shaft and/or tip. hand, favorable conditions arise from loading due to
Constructing on compressible soft soils poses challenges embankment fills or side hill fills placed over loose or
for geotechnical engineers. Developing negative skin compressible soils, new structures with shallow
friction (NSF) is a significant factor in evaluating the foundations located near existing or new deep foundations,
structural and geotechnical performance of piles in soft sites with significant dewatering, and sites susceptible to
soils. NSF refers to the downward shearing force exerted consolidation or compaction due to vibration or seismic
on the pile when the surrounding soil undergoes vertical effects. Therefore, the drag force can be developed along
deformation relative to the pile's shaft over time and depth. all piles as a result of the relative movement between the
The integration of NSF along the shaft is known as drag soil and pile caused by construction activities.
force, which is an additional axial force applied to the pile. Many case studies have documented failures caused by
In compressible soils, immediate and long-term the drag force, including Chellis (1951), Garlanger (1974),
settlements can occur due to embankment loading from the and Davisson (1993). Therefore, it is essential to take into
approach fill. This settlement can lead to the relative account the effects of drag force on pile foundations and
movement between the soil surrounding the deep evaluate this phenomenon, especially in bridge
foundation units and the piles themselves. The distribution construction sites.
of drag force with depth also depends on the relative This research includes developed instrumentation,
stiffness of the soil and pile (Alawneh et al. 2001). dynamic testing and a monitoring program on a driven steel
Even minor grading changes around piles, such as work H-Pile driven to a 36-meter depth located at the edge of the
platforms, temporary excavations, and backfilling around bridge abutment’s foundation system on an active bridge
abutments and pile caps, as well as final build-out and construction site, named Daly Overpass, in Brandon,
grading near foundations, can induce this force (MnDOT Manitoba, Canada, to evaluate the developed axial force
Geotechnical Manual 2021). along the pile’s shaft under service loads. Associated
interpretation and data corrections, in addition to lessons less than 60mm in diameter, is encountered. This is
learned from the instrumentation program, are discussed followed by layers of stiff lean clay, and lean clay till that
to provide an approach to calculate axial force distribution. has been divided into five layers, labeled till 5 to 1, based
Recommendations from this study can be utilized by on the soil resistivity test, extending from the ground
engineers and practitioners to negotiate the challenges surface to a depth of 41 meters. The properties of the soil
concerning NSF in any bridge construction site. layers were classified through a comprehensive range of
in-situ and laboratory tests, such as Standard Penetration
2 PROJECT SITE DESCRIPTION Test (SPT), Cone Penetration Test (CPT), Pressuremeter
Test (PMT), Geophysics, and Sensitivity tests. In addition,
The research site is situated at the bridge construction site sieve analysis, hydrometer, Index tests, and oedometer
on PTH 10 (18th Street), which serves as the primary tests were performed to supplement the evaluation of the
North-South corridor in Western Manitoba, Canada and a soil properties.
crucial transportation route through the City of Brandon. Based on the test results, the moisture content of the soil
The Daly Overpass, a new four-lane bridge, will replace the layers ranged from 12 to 22%. The liquid limit and plastic
50-year-old bridge that currently stands. The old three-lane limit values were found to be within the range of 27 to 42%
overpass has been subject to traffic congestion and and 14 to 17%, respectively. The unit weight of the clay till
bottlenecking, making it imperative to replace it to was determined to be approximately 21.4 kN/m3, while for
accommodate the region's growth. the clayey sand layer, the unit weight was found to be 20
The production pile chosen for instrumentation and kN/m3. Furthermore, the shear strength parameters of the
monitoring is positioned on the southern edge of the new soil layers were evaluated, with the friction angle ranging
bridge abutment foundation adjacent to the mechanically from 30 to 35 degrees and the cohesion value ranging from
stabilized earth (MSE) wall. Figure 1 illustrates the location 5 to 10 kPa.
of the research site and production instrumented pile. The elastic modulus of the clay till was determined using a
3% strain from PMT results. The modulus values ranged
from approximately 150 MPa for the uppermost till layer to
70 MPa for the lowermost till layer. The compression and
recompression indexes of the soil layers were also
calculated and found to be approximately 0.11 and 0.011,
respectively from the oedometer tests.

3 INSTRUMENTATION PROGRAM

The instrumentation plan for this study consists of installing

vibrating wire strain gauges as well as a strain transducer
and accelerator for conducting dynamic testing with a Pile
Driving Analyzer (PDA). Vibrating wire piezometers were
also installed to measure the pore water pressure by the
project consultant, TREK Geotechnical Inc., and the data
was provided.
The upper half portion of a steel H-pile 310 x 110 pile was
instrumented with arc-weldable vibrating wire strain
gauges (RST Instruments model VWSG-A). The gauges
were equipped with thermistors that allowed for the
measurement of temperature variations. Various models of
Figure 1. Research site and instrumented pile location
strain gauges have been utilized to measure the developed
strains, including semiconductor, electrical resistance, foil
2.1 Site stratigraphy
strain gauges and vibrating wire types.
Semiconductor strain gauges have several limitations that
The geological composition of the site is characterized by
should be taken into consideration. Firstly, they are not
a complex fill layer that overlays thick deposits of native
suitable for measuring large strains or deformations as they
clay till containing intercalated sand and gravel lenses,
have a limited measurement range. Additionally, they can be
followed by a shale layer. The native clay is medium
fragile and easily damaged, especially during installation and
plastic, moist, and is composed of silty, sandy, and pebbly
use. These factors should be weighed against the benefits
materials. At the surface, the clay exhibits a hard to very
of using semiconductor strain gauges in specific applications
stiff consistency, but it gradually softens with increasing
including the long-term measurement of strain along the
depth. The sand and gravel lenses are poorly graded and
piles.
are fully saturated.
Electrical resistance strain gauges were deemed unsuitable
During on-site investigations, the groundwater level was
for long-term use due to their high sensitivity to humidity and
determined to be approximately 5 meters below the ground
other weather-related effects. Also, high sensitivity to
surface. The soil stratigraphy commences with a 1.5-meter
environmental factors, particularly temperature fluctuations,
layer of clayey sand, classified using the unified soil
which can cause changes in the electrical resistance and
classification system (USCS). Subsequently, a layer of
affect the accuracy of the measurements (Bica et al. 2014).
hard to very stiff lean clay, containing mostly gravel traces
Vibrating wire strain gauges are the preferred sensors for
the long-term measurement of strain in driven piles, as they
operate on the vibrating wire principle, offering survivability
and long-term stability (Bartz and Blatz 2021). Compared
to other types of strain gauges, they are more robust and
have fewer technical and practical issues. Additionally,
they can accommodate long cable lengths without affecting
measurements and perform well in excessively moist
conditions (Suits et al. 2004).
Vibrating wire strain gauges consist of a high-tensile steel
wire anchored and tensioned between two mounts within
the instrument's body. The wire is protected from the
external environment by an encapsulating, sealed
stainless-steel tube. The strain measurement concept
involves tensioning a steel wire between two mounting
blocks that are welded to the surface of the pile. As the wire
Figure 3. Pre-tensioning range definition with slope
expands or contracts due to changes in the pile load, the
indicator vibrating wire data recorder for strain gauges
natural frequency changes. This frequency change
corresponds to a strain on the pile, which can be utilized to
A total of 20 vibrating wire strain gauges were installed over
calculate the change in the applied load on the pile (Budge
the first 18 meters of the pile. Figure 4 provides an overview
and Dasenbrock 2010). Figure 2 shows the strain gauge
of the soil profile and the arrangement of the strain gauges.
and mounting blocks attached on the pile’s flange as well
In order to minimize the effect of potential bending-related
as layout.
strains, two strain gauges were placed on opposite sides
of the pile flanges, below the top level, at each elevation.
Moreover, four additional strain gauges were installed at
the first 3 meters from the pile head to better measure the
pile head deflection and applied load. The remaining strain
gauges were placed at 1.8-meter intervals along the pile,
up to the halfway point.

Figure 2. Layout of strain gauge and mounting block

Sensors were also set up within the manufacturer's

recommended pre-tensioning range. After securely
fastening the strain gauges to the blocks, the
recommended pre-tensioning values were set between
3000 to 3500 micro-strains for monitoring expected
compressive strains and 1500 to 2000 micro-strains for
tensile strains, respectively (RST Instruments 2020).
The slope indicator vibrating wire data recorder device was
used to record the pre tensioning readings. Figure 3
displays the definition of the pre-tensioning range for
installed strain gauges with slope indicator vibrating wire
data recorder.
 during welding, and expansive foam was injected into the
flanges to fill the voids and firmly secure the cables after
welding. After the pile driving was completed, the wires
were enclosed in PVC pipes and buried underground to
protect them from the subsequent construction activities.
The protection measures are depicted in Figure 6.

Figure 4. Arrangement of the strain gauges layout along a) b)

the pile shaft

3.1 Instrument Protection Plan

To safeguard the sensors and wires during pile driving, two

steel angle bars (76.2 mm x 76.2 mm x 9.5mm thick) were
welded along the flanges of the pile and over the strain
gauges. The protective covering had a cross-sectional area
of approximately 26 mm2.
Sections of the attached protective covering's window were
cut along the protective angle bars to install the strain
gauges and nuts. The flange within the window was then
brushed with a grinder to create a clean surface. The cut
portion of the angle was preserved to be welded back in
place at a later time. Figure 5 illustrates the layout of the
cut portion dimensions.
c) d)

Figure 6. Protection plan measures a) Zip tied and fished

cables, b) Welding blanket over the strain gauges, c)
Expansive foam, d) Fished wires through PVC pipes and
buried under the ground

3.2 Pile Dynamic Testing

During the pile installation process, dynamic testing with a

PDA was conducted to estimate the pile capacities and
resistance distribution along the shaft. Prior to the pile
driving, PDA gauges were installed 1m from the pile top on
the web, as illustrated in Fig. 7.
Dynamic testing was carried out for each hammer blow
Figure 5. Cut portion dimensions
until the end of initial driving, while additional restrike
dynamic tests were performed to observe the potential
Following the installation of the sensors, the cables were
increase in shaft resistance of the pile.
secured with zip ties and fished through the protective
angle irons. To prevent loosening from the impact of
hammer pounding, the nut's screws were glued with
threadlocker to keep them firmly in place and safe. Then,
welding blankets were placed to shield the instruments
 5.1 Data interpretation

To calculate the strain changes caused by construction

activities, a baseline reading is required. The baseline
reading can be selected at any time to observe the impact
of the desired activity on the pile. In this study, to calculate
the axial force applied on the pile while considering the
effect of the pile installation, the baseline reading was
selected one day prior to the pile installation when there
was no construction activity in the vicinity of the pile, and
the strain distribution was relatively stable.
The adopted strain gauges and thermistors were labelled
as P10 to P1 from top to bottom. In this research, positive
strains indicate compression or shortening, while negative
strains indicate tensile strains or elongation. In addition, the
average of the values recorded by each pair of strain
Figure 7. Installed PDA sensors on the pile’s web gauges is taken into consideration for the calculations. The
following Eq. 1 to Eq. 6 can be used to convert the
4 SURVIVEABILITY measured strains to stress, axial and shear force as well as
the bending moment at each depth along the pile.
An important aspect to consider when selecting It should be noted that the adopted relationships for
instruments is their ability to survive the installation process calculating stresses/forces based on monitored values of
and their suitability for long-term monitoring. After strains, follow the principles of elasticity. Therefore,
conducting the instrumentation program, it was observed measures must be taken into account before interpreting
that only two out of 20 strain gauges ceased to function the data. Prior to the installation of strain gauges, a
after around 85 minutes of driving the pile with 6 a ton calibration process is performed to ensure the accuracy
hammer for installation, resulting in a survival rate of 90%, and reliability of the gauges. The calibration factor is
which is the desired survival rate. These 18 strain gauges typically provided by the sensor’s manufacturer, or it can
are continuing to function properly. be calculated by subjecting the gauges to known strains
and verifying their responses. Moreover, it is
5 DATA MONITORING AND ANALYSIS recommended to test the strain gauges to ensure they
return to zero strain when the applied strain is removed.
The data logger utilized in this research study was the Additionally, during the installation of the strain gauges,
RST-DT2040, as shown in Figure 7. This data logger has monitoring is carried out to ensure that the installation
the ability to monitor up to 40 sensors, which can be a process itself does not induce any plastic deformation. This
combination of vibrating wire sensors and thermistors. The approach prevents local plastic strains from being included
data logger was initially set to record data at 40-second in the recorded values, ensuring the accuracy of the
intervals for the first three days of the experiment, and then measurements. To account for thermal strains, thermistors
switched to 15-minute intervals for the subsequent 20 may be installed alongside or incorporated with the strain
days. For the long-term measurement, it was programmed gauges. By monitoring temperature variations and applying
to record data at 1-hour intervals. appropriate compensation, separating the effects of
temperature-induced strains from the elastic strains
becomes possible.

𝜀 = (𝜀1 − 𝜀2 ) [1]
𝜎 = 𝐸 .𝜀 [2]
𝐹 = 𝜎 .𝐴 [3]
(𝜀 − 𝜀2 )⁄ [4]
𝑘 = 1 𝑑
𝑀 = 𝐸 .𝐼 .𝑘 [5]
𝑑
𝑄 = .𝑀 [6]
𝑑𝑧
Where 𝜀1 and 𝜀2 are the strains for each side, 𝜎 is the axial
stress, 𝐸 is the pile elastic modulus, 𝜀 is the measured
average strain after pile installation, 𝐴 and 𝐼 are the total
cross-sectional area and moment of inertia of the pile and
protective covering, 𝑘 is the curvature, 𝑑 is the distance
between two opposite strain gauges at each level, 𝑀 is the
Figure 7. Data logger utilized to record the strains bending moment and 𝑄 is the shear force.
The elastic deformation from strain gauges is also can be
150 25
estimated by integrating the strains along the length of the
piles according to Eq. 7. 145
20

Temperature ( Celicius)
𝑛−1 140

Microstrain (με)
𝜀𝑖 + 𝜀𝑖+1
∆ = ∑( ) (𝑙𝑖+1 − 𝑙𝑖 ) [7] 135 15
2
𝑖=1 130
Where n is number of levels along the pile, 𝜀𝑖 and 𝜀𝑖+1 are 125 10
the strain measurement at level (i) and (i+1), and (𝑙𝑖+1 − 𝑙𝑖 ) 120
is the length of the pile between the two levels. 5
115
5.2 Data correction 110 0
0 2 4 6 8 10 12 14 16 18
To obtain accurate results, two factors must be considered Time (Hour)
when correcting the strain distribution along the pile, taking SG 2 Thr 2
into account the effect of locked-in manufacturing strains in
the steel pile and temperature variations after pile
installation. Figure 8. Strain and temperature variations for strain gauge
Bartz and Blatz (2021) conducted a study on a technique number 2 located at 16m from the ground level
proposed by Randolph et al. (1994), which involved striking
the pile to a steel plate to eliminate the influence of locked- 6 DISCUSSION
in manufacturing strains. Their findings revealed a
difference of approximately 5 microstrains before and after To accurately measure the axial force distribution along the
the experiment. Furthermore, Bartz and Blatz (2021) noted shaft, it is necessary to take into account the stress exerted
that the prior handling of the pile before installation may on the pile during installation (Bartz and Blatz 2021).
also remove the locked-in manufacturing strains in the Unfortunately, the strain gauge recordings are affected by
steel pile. The present study assessed the preinstallation the hammer's rebound, which has made it difficult to
process and determined that the recommended range for measure this phenomenon despite several attempts in the
considering the locked-in strains was around 6.5 to 7 literature. Rieke and Crowser (1987) have noted that
microstrains. Consequently, the recorded data was during each hammer blow, the soil surrounding the shaft
adjusted by adding 7 microstrains. and underneath the pile tip undergoes plastic deformation,
Following the installation of the pile in the cooler ground, a leading to the development of stresses that hinder further
decrease in strain was observed where there was no pile penetration. These stresses, also known as residual
external load on the pile, indicating the development of soil stresses, persist even after the end of driving, resulting
tensile forces. This phenomenon can be reversed when the in the pile experiencing residual forces. The term "residual"
pile is installed in the ground with a higher temperature is commonly used to describe the axial force that develops
than the surrounding atmosphere. Hence, it is necessary prior to the static load test (Fellenius 2002). Hence, for this
to correct the measured strains by accounting for the study, the axial force generated during pile driving is
variations caused by a temperature change until the referred to as drag force.
temperature stabilizes. To measure this force, it is necessary to take into account
In this study, After the installation of the pile in April 2022, the baseline reading before pile installation. Figure 9
into the cooler ground, a notable decrease in strain was illustrates the axial force distribution along the pile while
observed for the strain gauges located deeper down the taking into account the drag force resulting from pile
pile from the ground surface over the 17.5 hour after driving installation for two weeks after driving. The force
where there was no external load applied on the pile. The distribution has also been corrected according to the
decreasing trend observed was almost linear, with a suggested corrections. The maximum value of the drag
negligible strain change observed in the top sensors and force, which occurred at 14m from the pile head, was
over 75 microstrains observed in the strain gauges located measured to be 2075 kN after two weeks of pile installation.
at a depth of 16 meters from the ground surface. It is However, if the baseline reading is selected after pile
believed that this decrease in strain was solely due to driving when the driving-induced strains have stabilized as
temperature decrease, which reached over 11 degrees suggested by Drbe et al. (2017), the maximum drag force
Celsius during the 17.5-hour measurement period for the will occur similarly at 14m, but its magnitude will be
deepest sensors. Figure 8 indicates the records for the approximately 420 kN. In this scenario, the maximum drag
representative strain gauges and incorporated thermistor force is about 20% of the actual magnitude. Figure 10
(Number 2 from P1) located at the deep down of the pile. shows the comparison of these two scenarios.
The jumps in the figures are related to the time of the The selection of the baseline reading after pile installation
restrike tests. can also be influenced by the soil type. In cohesive soils,
the process of selecting the baseline reading may take
longer due to the dissipation of pore water pressure and
consolidation processes, which affect the required time for
the piles to settle and stabilize. Strain gauges in such soil
types may initially record higher values because of the
soil's adhesion and cohesive nature. During pile
installation, the soil can adhere to the pile surface, creating Axial Force (kN)
additional frictional resistance. However, as the pile settles -1000 -500 0 500 1000 1500 2000 2500
and excess pore pressure dissipates, the strain gauge 0
readings may gradually decrease and stabilize over time.
Furthermore, piles driven into cohesive soils susceptible to 2
creep can experience stress relaxation, resulting in
diminishing residual forces over time (Poulos 1987).
4
In cohesionless soils, strain gauges attached to piles may
show relatively lower initial readings compared to cohesive
soils. This is because cohesionless soils allow for easier 6
penetration and reduced friction during pile installation. The
initial readings may stabilize quickly as the excess pore

Depth (m)
8
pressure dissipates rapidly in these soils.
10

Axial Force (kN) 12

-1000 -500 0 500 1000 1500 2000 2500
0 14

2 16

4 18

6 2 Weeks after pile driving, Corrected

2 Weeks after driving, Baseline after installation
Depth (m)

10 Figure 10. Comparison of axial force along the pile

12
Additionally, to determine the pile shaft and tip bearing
capacity, signal matching with CAPWAP was performed
using data obtained from the end of the initial driving
14
hammer blow. The total capacity of the pile was calculated
to be approximately 3000 kN, with the shaft and tip
16 contributing 2850 kN and 150 kN, respectively.

18 7 CONCLUSION AND LESSONS LEARNED

2 Weeks after pile driving, Corrected This paper describes the instrumentation program with
2 Weeks after pile driving, Apparent vibrating wire strain gauges installed along a pile to
measure the long-term axial force distribution. Details of
the instrumentation in addition to the analysis and
Figure 9. Corrected and apparent axial force distribution correction of the collected data were discussed as well.
along the pile for two weeks after installation Furthermore, pile dynamic analyses were conducted using
the PDA and CAPWAP signal matching techniques to
accurately calculate the shaft and tip resistance of the pile.
Due to the heavy construction that has been performed
around the production instrumented pile, it was not
possible to install other types of instruments, such as
displacement transducers, extensometers, pressure cells,
etc.
Data analysis revealed that in order to accurately measure
the drag force resulting from pile installation, strain gauges
should be recorded and attached to the data logger before
pile driving. In this scenario, the maximum drag force is
observed to be more than 80% greater than when the
baseline reading is selected after pile driving and the
induced driving strains have stabilized. Also, to obtain
accurate results, it is also necessary to apply temperature
correction and take into account the locked-in
manufacturer strains.
8 ACKNOWLEDGMENT Randolph, M.F., Dolwin, R., and Beck, R. 1994. Design of
driven piles in sand. Géotechnique, 44(3): 427–448.
The authors would like to thank the following parties for Thomas Telford Ltd.
their contributions to the study: Manitoba Transportation doi:10.1680/geot.1994.44.3.427.
and Infrastructure and Mitacs for providing financial Rieke, R.D., and Crowser, J.C. 1987. Interpretation of pile
support, Dart welding services to install instrumentation load test considering residual stresses. Journal of
and pile installation, and TREK Geotechnical Inc. for Geotechnical Engineering, 113(4): 320–334.
providing dynamic testing services and site investigation American Society of Civil Engineers.
data.

9 REFERENCE

Alawneh, A.S., Nusier, O., Husein Malkawi, A.I., and Al-

Kateeb, M. 2001. Axial compressive capacity of
driven piles in sand: a method including post-driving
residual stresses. Canadian Geotechnical Journal,
38(2): 364–377. NRC Research Press Ottawa,
Canada. doi:10.1139/t00-104.
Bartz, J.R., and Blatz, J.A. 2021. Considerations for
Measuring Residual Stresses in Driven Piles with
Vibrating Wire Strain Gauges. Canadian
Geotechnical Journal, (ja). doi:10.1139/cgj-2020-
0614.
Bica, A.V.D., Prezzi, M., Seo, H., Salgado, R., and Kim, D.
2014. Instrumentation and axial load testing of
displacement piles. Proceedings of the Institution of
Civil Engineers-Geotechnical Engineering, 167(3):
238–252. Thomas Telford Ltd.
Budge, A.S., and Dasenbrock, D.E. 2010. Installation of
Downdrag Instrumentation on a Bridge Abutment
Foundation: Lessons Learned. In GeoFlorida 2010.
American Society of Civil Engineers, Reston, VA. pp.
1237–1245.
Chellis, R.D. 1951. Pile foundations: theory, design,
practice, New York: McGraw Hill.
David Suits, L., Sheahan, T., Paikowsky, S., and Hajduk,
E. 2004. Design and Construction of Three
Instrumented Test Piles to Examine Time Dependent
Pile Capacity Gain. Geotechnical Testing Journal,
27(6): 12191. ASTM International.
doi:10.1520/GTJ12191.
Davisson, M.T. 1993. Negative skin friction in piles and
design decisions. In Proc., 3rd Int. Conf. Case
Histories Geotechnical Engineering. pp. 1793–1801.
Drbe, O., Sadrekarimi, A., Naggar, H.E.L., Sangiuliano, T.,
Moen, R., and Ng, J. 2017. Instrumentation
Monitoring Program to Measure the Magnitude,
Distribution and Time Dependency of Drag Load on
Abutment Piles: A Case Study. GeoOttawa.
Fellenius, B.H. 2002. Determining the True Distributions of
Load in Instrumented Piles. In Deep Foundations
2002. American Society of Civil Engineers, Reston,
VA. pp. 1455–1470.
Garlanger, J.E. 1974. Measurement of pile downdrag
beneath a bridge abutment. Transportation
Research Record, (517).
MnDOT Geotechnical Manual. 2021. MnDOT
Geotechnical Manual, APPENDIX F: Downdrag
Load (Dragload, Drag Force) and Downdrag.
Poulos, H.G., 1987. Analysis of residual stress effects in
piles. Journal of Geotechnical Engineering, 113(3),
pp.216-229.
Centrifuge modeling tests of seismic
behaviour of helical piles in dry sand
Chunhui Liu & Lijun Deng
Department of Civil and Environment Engineering – University of
Alberta, Edmonton, Alberta, Canada
Cheng Wang
School of Civil Engineering – Yantai University, Yantai, Shandong,
China
Yongzhi Wang
Institute of Engineering Mechanics – China Earthquake Administration, Hebei, China

ABSTRACT
Helical piles are often used in earthquake-prone zones, but the seismic performance of helical piles remains relatively
understudied. This research presents a centrifuge shaking table test program for both helical piles and smooth piles
embedded in layered dry sand. The specific objectives of this study encompass investigating the axial load redistribution
caused by seismic shaking and examining the lateral response behavior of helical piles. Four single-helix piles were
installed, where the helical plates were embedded into a dense sand layer to mimic the practical working condition of
helical piles. Accelerometers, displacement transducers, and strain gauges were utilized to measure the dynamic response
of both helical piles and the surrounding soil during earthquakes. Based on the experimental results, the predominant
periods of soil and helical piles were evaluated, and the response of both soil and piles was interpreted. Soil and pile
settlements during earthquakes were observed, along with variations in axial loads during the shaking sequence.
Additionally, the distribution of bending moments was presented to illuminate the distinctive behavior of helical piles
compared to conventional smooth piles.
RÉSUMÉ
Les pieux hélicoïdaux sont souvent utilisés dans les zones sujettes aux tremblements de terre, mais les performances
sismiques des pieux hélicoïdaux restent relativement peu étudiées. Cette recherche présente un programme d'essai de
table vibrante centrifugeuse pour les pieux hélicoïdaux et les pieux lisses intégrés dans du sable sec en couches. Les
objectifs spécifiques de cette étude englobent l'étude de la redistribution de la charge axiale causée par les secousses
sismiques et l'examen du comportement de réponse latérale des pieux hélicoïdaux. Quatre pieux à simple hélice ont été
installés, où les plaques hélicoïdales ont été encastrées dans une couche de sable dense pour imiter les conditions de
travail pratiques des pieux hélicoïdaux. Des accéléromètres, des transducteurs de déplacement et des jauges de
contrainte ont été utilisés pour mesurer la réponse dynamique des pieux hélicoïdaux et du sol environnant pendant les
tremblements de terre. Sur la base des résultats expérimentaux, les périodes prédominantes du sol et des pieux
hélicoïdaux ont été évaluées, et la réponse du sol et des pieux a été interprétée. Des tassements de sol et de pieux
pendant les tremblements de terre ont été observés, ainsi que des variations des charges axiales pendant la séquence de
secousses. De plus, la distribution des moments de flexion a été présentée pour éclairer le comportement distinctif des
pieux hélicoïdaux par rapport aux pieux lisses conventionnels.

1 INTRODUCTION investigate the seismic behavior of helical pile during

earthquakes.
Pile foundations, having advantages of high resistance and Helical piles are comprised of a steel pipe shaft and one or
seismic performance, have gained extensive use in critical more helices welded onto the pile shaft. Their axial
infrastructure within seismically active regions (Liam Finn performance under static loading has been extensively
et al., 2005; Liu et al., 2022). Driven steel and precast investigated (e.g. Li and Deng, 2019; Li et al., 2018;
concrete piles, as well as drilled cast-in-place concrete Harnish and El Naggar, 2017). The conclusions drawn from
piles, are popular choices for such projects. However, model tests, field tests, and numerical modeling
some of these pile options did not perform well and demonstrate that the compressive and tensile bearing
suffered damage during recent earthquakes (Ishihara et al. capacities of helical piles are significantly improved when
1990, Liu et al. 2017). One the other hand, the case studies compared to traditional cast-in-place piles and
from recent earthquakes in Christchurch, New Zealand prefabricated piles of the same diameter. This
(Cerato et al. 2017) demonstrated that helical piles and improvement is attributed to the helices that increase the
superstructures supported on helical piles withstood contact area between the pile and the soil. Based on these
earthquakes with limited or no structural damage. This studies, design procedures have been summarized in
suggests that helical pile foundations could be a viable current design guides, enabling helical piles to be designed
choice in seismically active regions. However, the seismic for various structures (Ridgley, 2015). Nonetheless, these
performance of helical piles remains relatively guides do not provide specific details regarding the
understudied. As a result, there is a motivation to application of helical piles in high seismic areas.
Helical piles have been utilized as the foundation for The soil layers were constructed using dry sand, with a
significant infrastructure projects. The performance of relative density of about 50%, overlying dry sand with a
helical piles under seismic loading has garnered attention relative density of about 80%. Settlements in both the soil
from engineers, especially due to their widespread use in and piles during earthquakes were observed and utilized to
British Columbia, Quebec, and Ontario. The seismic interpret the changes in axial loads after the earthquake.
behavior of helical piles was investigated by subjecting the Lateral displacement and distribution of bending moment
pile head to cyclic and dynamic loads, which primarily were presented to highlight the superior performance of
simulated the inertial load from the superstructure. helical piles in comparison to conventional smooth piles.
However, this approach neglected the kinematic
interaction between the soil and helical pile (El Sharnouby 2 DESCRIPTION OF THE CENTRIFUGE MODEL
and El Naggar, 2018; Elkasabgy and El Naggar, 2018).
Nevertheless, this method was still considered a 2.1 Centrifuge and soil container
reasonable simulation for studying the behavior of piles
under dynamic loading. Based on these studies, it has A centrifuge shaking table tests was conducted on the
been concluded that the dynamic behavior of helical piles large centrifuge at the Institute of Engineering Mechanics,
is essentially similar to that of driven piles, with the helices China Earthquake Administration. The centrifuge model
having an insignificant effect. contained 4 single piles with various helices and masses of
Large-scale shaking table tests can effectively account for superstructure (Figure 1). The tests were performed at a
the kinematic soil-pile interaction induced by soil response. centrifugal acceleration of 50 g. All the quantities used in
Elsawy et al. (2019) conducted a large-scale shaking table this analysis have been converted to prototype according
test on single helical piles in dry sand, where ten steel piles to the scaling laws described by Garnier et al. (2007)
with different configurations and pile head masses were unless otherwise noted.
installed. The results demonstrated that the response of
helical piles closely resembled that of driven piles,
highlighting the helical piles' capability to perform on par
with conventional driven piles under seismic loading.
Shahbazi et al. (2020) conducted a large-scale shaking
table test on a group of helical piles in dense sand. This
study analyzed the effects of strain magnitude, pile-
structure connection, pile slenderness ratio, and soil depth
on damping characteristics. Based on observations from
the aforementioned single and grouped helical pile large-
scale table tests, Fayez et al. (2022) assessed the Soil-
Structure Interaction (SSI) effects on the stiffness and
natural frequencies of both single and grouped helical
piles. They proposed an analytical approach to simulate
the dynamic characteristics of helical piles. Fayez et al.
further evaluated the effects of earthquake characteristics
and pile head connection on the seismic performance of
both single and grouped helical piles.Orang et al. (2021)
evaluated the application of helical piles in reducing
liquefaction-induced foundation settlement. They
investigated the seismic performance of helical piles in
liquefiable grounds through two large-scale shake table
test series. The results indicated that foundations
supported on helical piles experienced minimal differential
settlement and tilt. Additionally, the post-shaking
Figure 1. Photograph of the centrifuge shaking table model
liquefaction-induced settlement mechanisms did not
significantly affect the helical pile foundation settlement.
Dynamic centrifuge model tests offer the advantage of
The centrifuge model was constructed in a laminar shear
accurately simulating in-situ confining stress and soil
container, which was composed of independent layered
dynamic response, in comparison to 1-g shaking table
rectangular frames. The rectangular frames were
tests. In this study, a centrifuge test program was
sandwiched using rubber interfaces providing lateral
conducted on helical piles in dry sand. The objectives were
flexibility and simulating a periodic boundary condition. The
to investigate the axial and lateral dynamic response of
internal dimension of the laminar shear container at the
soil-pile systems, compare the performance of helical piles
model scale was 1.2 × 0.5 × 0.6 m in length width, and
and smooth piles, and lay the groundwork for subsequent
depth respectively (Figure 1).
research on piles installed in liquefiable soils. The dynamic
centrifuge model tests conducted in this study can be
2.2 Soil and pile properties
considered among the first of their kind.
The centrifuge model consists of four individual piles,
The model was built as follows. First, the piles were fixed
comprising three helical piles and one smooth pipe pile.
vertically on beams of laminar shear container, then the
sand layers were constructed using dry pluviation method The centrifuge model consisted four single piles, i.e., two
to achieve the target relative densities. After the double helical piles, one single helical pile, and one smooth
construction of soil layers, the fixity was removed. The soil pipe pile. The outer diameters of pile shafts are all 80 cm,
profile of the centrifuge model contained two soil lays both and the diameters of the helices (D) were 200 cm (2.5 times
sieved from the Fujian sand. The relative density (Dr) of the the diameter of pile shaft). For double helical piles (HP1
upper soil layer is about 50% overlaying the dense sand and HP2), the interhelix spacings are both equal to 2 D
layer with a relative density about 80%. The physical (400 cm). The wall thickness of the pile shaft was 10 cm,
properties of the sands of the two layers are summarized and the length of the piles were all 27.5 m. The physical
in Table 1. The thicknesses of the two soil layers are 10 m properties of the piles are listed in Table 2. The thickness
for upper loose sand layer and 17.5 m for the bottom dense of the soil below the pile tip is 6 m (7.5 times the outer
sand layer at the prototype scale (Figure 2). The diameters of the pile shaft), which is large enough to avoid
relationship between drop distance and Dr was verified the boundary effect. The piles of the centrifuge model were
before the construction of soil layers. manufactured using 6061 aluminum alloy hollow pipes. For
the helical piles, the helices were welded on the pile shafts.
To avoid the pile collapse during the spun-up from 0 to 50
g centrifuge acceleration, the pile axial capacity was
considered. For the SP4 (smooth pipe pile), the axial
capacity was calculated based on the recommendation of
API. For the helical piles (HP1, HP2, and HP3), individual
bearing and cylindrical shear failure mechanisms can be
used the pile ultimate capacity. In this study, the method
reported by Elsherbiny and Naggar (2013) was used to
calculated the pile ultimate capacity. The added masses at
pile heads were obtained from the ultimate capacities by
considering various factor of safety and summarized in
Table 2. It should be noted, the helical pile was usually
installed using the method of applying a torque on pile head
in engineering practice, however, this installation method
was not adopted in the centrifuge model to facilitate the
(a) model construction.

Table 2. Physical properties of piles at the prototype scale

Pile Pile type Helix Mass at Axial Outer

ID embed pile load diameter,
ment head ratio1 Wall
(m) (Mg) thickness,
Helix
diameter
(cm)
HP1 Double 15.5, 500 0.4 80,10,200
19.5
HP2 Double 15.5, 1000 0.8 80,10,200
9.5
(b) HP3 Single 19.5 500 0.4 80,10,200
Figure 2. Typical centrifuge test layout with the locations of SP4 Smooth n.a. 100 0.08 80,10/-
the transducers at the prototype scale: (a) front view, (b) Note: 1 Axial load ratio equal to weight of structure over axial
capacity of pile. The axial capacity of the pile can be calculated by
vertical view yield strength times cross-sectional area

2.3 Model Instrumentation

Table 1. Properties of Fujian standard sand
The strain gauges were placed at key locations of piles,
Parameters Dense sand Loose sand as shown in Figure 2 to record the axial and lateral
Specific gravity, Gs 2.645 2.645 response of piles during the excitations. Each pile had eight
or seven strain gauge elevations, with two opposing strain
Relative density, Dr 80% 50%
gauges at each elevation. The strain gauges were
Coefficient of curvature, Cc 0.926 0.987 connected as half-bridges to reduce the effect of
Coefficient of uniformity, Cu 1.748 1.508 temperature-induced measurement errors. The
accelerometers and linear variable displacement
D60 (mm) 0.526 0.199
transducers (LVDT) were placed on pile heads to record
D30 (mm) 0.383 0.161 the accelerations of piles and displacements in lateral and
D10 (mm) 0.301 0.132 vertical directions of the piles. The accelerometers and
LVDT were instrumented in soil profiles to measure the
accelerations of soil and settlement of ground surface. In
addition, an accelerometer was mounted at the base of the
laminar container to measure the applied excitations.

2.4 Input motions

Two different shaking schemes were applied to the

centrifuge model, i.e., white noise and earthquake motions.
The white noise was a random signal with a constant
intensity for a range of frequencies. The white noise used
in the tests had a peak acceleration of 0.01 g and a
bandwidth of 0-10 Hz. The intensity of this white noise is
enough to produce the soil and pile deformation to be
recorded by various sensors without resulting in non-linear (b)
behavior in the soil and piles. The earthquake records from Figure 3. Input motions: (a) Time history of El Centro
Imperial Valley earthquake (El Centro motion) and Kobe motion; (b) Time history of Kobe motion
earthquake (Kobe motion) were input to the centrifuge
model to analysis the dynamic response of pile and soil. 3 EXPERIMENTAL RESULTS
The normalized time histories of the motions are shown in
Figure 3. The Fourier spectral analysis was conducted on 3.1 Site response
the motions, and it is found that the most energy was
concentrated in 1-3 Hz. These two earthquake motions The distance between the sensors in the middle of the
were scale to various amplitudes, as shown in Table 3 to model and surrounding piles is 10 m (12.5 times pile shaft
investigate the nonlinear dynamic response of soil and pile. diameter, and far larger than 6 times pile shaft diameter),
Before and after each earthquake motion, one white noise and the effect of pile on soil can be ignored. To facilitate to
was input from the container base to evaluate the natural illustrate the dynamic results of soil, this part of soil was
period of soil and piles. The HP2 was damaged during 0.1 termed as free field. For the soil between HP1 and HP2
g El Centro motion, thus, the test was paused and the HP2 was termed as near field (Figure 2). Figure 4(a) is 5%-
was removed from the centrifuge model to ensure the damped acceleration response spectra (ARS) of El Centro
safety. After that, the following motions in Table 3 were and Kobe earthquakes at container base, free field surface
applied. Noted that, the earthquake motions shown in and near field surface. It is seen that Kobe earthquake at
Figure 3 were measured and normalized from the the base has a peak at the 0.35 s, while the Kobe
accelerometer installed at the base of the laminar earthquake at the ground surface of free field and near field
container. both have a peak at 0.69 s. The frequency of earthquake
motions was changed, which shows significant site effect.
The ARS amplitude of near field is larger than that of near
Table 3. Input earthquake motions field caused by the reinforcement of pile. For the El Centro
earthquake, it is seen that a peak was found at 0.42 s for
Serial the ARS at the container base, while the peaks were found
Dynamic loadings Peak amplitude
number at 0.52 s for the ARS at the ground surface of free field and
1 0.06 g near field.
2 El Centro ground 0.1 g
3 motion 0.19 g
4 0.31 g
5 0.07 g
6 0.12 g
Kobe ground motion
7 0.23 g
8 0.32 g

(a)

(a)
 input from about 0.06 g increasing to 0.3 g. The loose soil
tended to behave contraction.

3.2 Soil and pile settlement

The soil and pile settlements after earthquakes were

plotted in Figure 6 to analyze the axial behavior of piles
during earthquakes. The x axis corresponding to the input
earthquake motion serial number listed in Table 3. It is
seen that pile settlements increased as the earthquake
sequences and then tended to be constant during Kobe
earthquake for HP1, HP3 and SP4. HP2 collapsed during
0.1 g El Centro. The soil settlement increased as the
increase of input acceleration. It shows that the soil
(b) settlement is larger than piles except HP2, which shows
Figure 4. (a) 5%-damped acceleration response spectra negative frictional force along the pile shafts was
(ARS) of El Centro and Kobe waves at container base, free developed during the earthquakes. It is quite interesting
field surface and near field surface; (b) Spectral that the settlement of smooth pile (HP4) is small than that
acceleration ration of EI Centro and Kobe at free field of helical piles. It might be caused that the mass at pile
surface and near field surface head of HP4 is far less than that of helical pile.

To understand the site effect on the El Centro and Kobe

earthquake motions. The spectral acceleration ratio (α)
was analyzed and shown in Figure 4(b). The α was defined
as the ratio of the ARS at the ground surface of near field
and free field to the ARS at the container base for the given
period. It is seen that the El Centro and Kobe motions were
strongly influenced by the local site condition. At 0.58-s
period, the α of Kobe motion reached 4.4 and 5 for the free
field and near field, which show a significant amplification.
At 0.69 s-period, the α of El Centro motion reached 3.65
and 4 for the free field and near field, which also showed a
significant amplification, although the amplification effect is
less than that of Kobe motion. In addition, comparing to the Figure 6. Soil and pile settlements after earthquakes
free field, the amplification effect was more significant at
the near field for both El Centro and Kobe motions. The pile masses on HP1 and HP3 are the same, however,
the settlement of HP1 is larger than that of HP3. It shows
that the settlement of helical pile with double helices is
larger than that of helical pile with single helix. It is caused
that the soil settlement is larger than pile settlement (except
HP2), thus the more downward drag force was developed
on double helical piles comparing to single helical pile. The
settlement of HP2 is larger than the settlements of soil and
HP1, HP3 and SP4, which might be caused by the heavy
mass of superstructure.

3.3 Natural frequency of pile

The natural frequency of pile was evaluated by the pile

head acceleration response under white noise. The pile’s
Figure 5. Relationship of spectral acceleration ratio versus natural frequency was obtained by performing a fast
input acceleration Fourier transform of the measured response time history.
These results assist in further understanding the piles’
To investigate effect of the site effect on input response to earthquake motions with different frequency
accelerations, the relationship of spectral acceleration ratio contents. Figure 7 shows the Fourier spectra using
versus input peak acceleration was shown in Figure 5. It response from accelerometer readings of HP1, HP3 and
shows that the α increased then decreased as the increase SP4. For the Fourier spectrums of HP1 and HP3, several
of input peak acceleration, and reached a peak at 0.19 g peaks were observed, while for the Fourier spectrums of
for El Centro motion. While for the Kobe motion, the α tends SP4, the peak was narrowed in 1-2 Hz. The first peak
to decrease as the increase of input peak acceleration. It corresponds to the natural frequency of the horizontal
should be noted, as shown in Table 3, the motions were degree of freedom (first mode) of the soil–pile system,
while the second peak corresponds to the natural layer. Thus, negative shaft fraction was developed in loose
frequency of the rotational degree of freedom (second sand layer, and positive shaft fraction was developed in
mode). It is found that the first peaks of HP1 and HP3 are dense sand layer. From Figure 8 (b), it is also observed
closed, which are 0.32 and 0.34 Hz. that the negative shaft fraction was developed in upper
The natural frequency of piles can also be estimated region of SP4, and positive shaft fraction was developed in
approximately using the following equations: the lower region. The critical depth is 5 m which is
𝑘 shallower than that of HP1. Another interesting observation
𝑓=√ [1] from Figure 8 is the initial load near the pile tip of HP1 is
𝑚
where m is the total mass of superstructure mounted at pile slightly larger than 0 kN, but the inertial load near the pile
head. k is the stiffness of the soil–pile system, and can be tip of SP4 is about 910 kN, which means large tip
3𝐸𝐼 resistance was developed for SP4.
obtained by 𝑘 = √ , in which EI is the bending stiffness By comparing the axial load along the pile after the
𝑙3
of pile, and l is considered simply as the length of pile earthquakes and initial load, the mechanism of
above ground surface in this study. Based on equation [1], redistribution of axial load can be obtained. It is seen from
the natural frequency of HP1 and HP3 are both 0.833 Hz, Figure 8 (a) that the axial load along the pile above the
which is larger than the values obtained by acceleration bottom helix increased after the earthquakes, but
readings at pile head. The natural frequency of SP4 is 1.86 decreased below the bottom helix and reduced to around
Hz, which is slightly larger than that from the measured 0 kN. This is caused by that the settlement of soil is larger
result (about 1.72 Hz as shown in Figure 7.). As the actual than that of pile in the upper region, thus downward drag
l is larger than the length of pile above ground surface, force was developed on upper helix and pile shaft portion
which resulted in the calculated values larger than that from above the bottom helix during earthquake. The settlement
the measured accelerations. of sand is smaller than that of pile for the portion below the
bottom helix, thus, upward force was develop on bottom
helix and pile shaft below the bottom helix during
earthquakes. It is also seen that the negative shaft fraction
on pile was quite small for the shallow portion, and most
part of negative shaft fraction was developed on pile from
10 m to 19 m during earthquakes.

Figure 7. Fourier spectra using response from

accelerometer readings of HP1, HP3and SP4

3.4 The axial load redistribution after earthquakes

During the spin-up of the centrifuge and application of

earthquakes, the axial loads were observed to be changed,
and Figure 8 shows the axial load along the pile length (a)
before and after earthquakes. The axial loads along pile
after spin-up to 50 g but before applying earthquake
motions were termed as the initial load, which are relative
to the that immediately prior to swing-up of the centrifuge.
Although there might be some axial load on pile, owing to
the low confining pressure in the centrifuge a 1 g, this axial
load was small, and the most changes of axial load were
caused by the spin-up of the centrifuge.
From Figure 8 (a), it is observed that the initial load
increased as the pile depth and reached a peak then
decreased, and the critical depth (the depth of the natural
plane) is 10 m. It is shows that negative shaft friction in the
upper regions of the HP1, whereas towards the bottom of
this pile positive shaft fraction is recorded. When the
centrifuge was spined up, the soil and pile settlement
(b)
occurred. In the loose sand layer (upper 10 m), the
Figure 8. Axial loads of piles after earthquakes: (a) HP1;
settlement of soil is larger than that of pile, while the
(b) SP4
settlement of soil is smaller than that of pile in dense sand
The axial load redistribution after earthquakes of SP4 is rotational degree of freedom of HP1 and HP3 are both
different with HP1. It is seen form Figure 8 (b) that the axial about 1 Hz which are closer to the predominant frequency
load along the pile decreased after the earthquakes, which of Kobe motion comparing to the horizontal degree of
shows the pile settlement of smooth pile was larger than freedom. By comparing the bending moments at 5 meters
the surrounding soil and positive fraction was developed depth, it is seen that the bending response of HP1 is larger
along the pile shaft. Besides that, axial load close to pile tip than HP3, which shows that more soil was mobilized for
is about 510 kN (far smaller than initial load) after 0.31 g El HP1 during earthquakes to constrain the pile.
Centro motion which show large tip resistance was
developed.

3.5 Lateral response of piles

The accelerometer was mounted at the pile head to record

the acceleration response of HP1, HP3 and SP4 under
earthquakes. The acceleration time histories at pile heads
under 0.06g El Centro motion was shown in Figure 9. It
shows that the frequency content of acceleration of all piles
are similar with that of base. The peak acceleration of HP1
and HP3 are similar with that of base, while the peak
acceleration of SP4 is 0.1 g, which is twice of base peak
acceleration. It is known that the seismic response of a pile
to a certain earthquake motion can be evaluated in terms
of frequencies, i.e., comparing the earthquake’s
predominant frequency and natural frequency of pile. The (a)
closer the two frequencies are, the higher is the resonance
effect and, consequently the pile response. As shown in
Figure 3 (b), the most energy of El Centro motion is
distributed in 1-3 Hz, and as addressed above, the natural
frequency of HP1 and HP3 are both around 0.3 Hz, while
the natural frequency is 1.72. Thus, higher peak
acceleration was observed from SP4.

(b)

Figure 9. Acceleration time histories at pile heads under

0.06g El Centro motion

The lateral response of helical pile and smooth pile can be

investigated using the bending moment response. The
bending moment profiles of HP1, HP3 and SP4 under
earthquakes were shown in Figure 10. It is shown that the Figure 10. Bending moment profiles under 0.07 g Kobe
bending moment response of single helical pile and double motion: (a) HP1; (b) HP3; (c) SP4
helical pile were similar. The maximum bending moments
of HP1 and HP3 were both slightly larger than 400 kN-m,
and the maximum bending moments were both observed Figure 10 shows that the bending moment response of
at 1 m below the pile head mass rather than that at ground smooth pile is quite different with helical pile. The maximum
surface level. It is caused that the natural frequency of the bending moment of smooth pile is larger than that of helical
pile. Besides that, the maximum bending moment of SP4 Fayez, A.F., El Naggar, M.H., Cerato, A.B. and Elgamal,
under 0.07 g Kobe motion was observed at ground surface A., 2022. Seismic response of helical pile groups from
rather than pile head, which can be explained that the shake table experiments. Soil Dynamics and
natural frequency of the horizontal degree of freedom of Earthquake Engineering, 152: 107008.
SP4 is close to the predominant frequency of Kobe motion. Garnier, J., Gaudin, C., Springman, S.M., Culligan, P.J.,
The bending moment below 9 m depth of SP4 is negligible Goodings, D., Konig, D., Kutter, B., Phillips, R.,
It should be noted that although the bending moment Randolph, M.F. and Thorel, L., 2007. Catalogue of
response of helical pile and smooth pile is different, it might scaling laws and similitude questions in geotechnical
be caused by the natural frequency of pile rather than centrifuge modelling. International Journal of Physical
helices. Modelling in Geotechnics, 7(3): 01-23.
Harnish, J. and El Naggar, M.H., 2017. Large-diameter
4 CONCLUSION helical pile capacity–torque correlations. Canadian
Geotechnical Journal, 54(7): 968-986.
A centrifuge test on helix piles in dry sand was performed. Ishihara, K., Okusa, S., Oyagi, N. and Ischuk, A., 1990.
The site response was analyzed based on the acceleration Liquefaction-induced flow slide in the collapsible loess
response of soil, And the axial and lateral dynamic deposit in Soviet Tajik. Soils and foundations, 30(4): 73-
response of piles was investigated. The following 89.
conclusions may be drawn: Li, W. and Deng, L., 2019. Axial load tests and numerical
1. The ARS amplitude of near field is larger than that modeling of single-helix piles in cohesive and
of near field caused by the reinforcement of pile, and a cohesionless soils. Acta Geotechnica, 14(2): 461-475.
significant amplification was observed in both near field Li, W., Zhang, D.J.Y., Sego, D.C. and Deng, L., 2018. Field
and free field. testing of axial performance of large-diameter helical
2. The settlement of ground surface is larger than piles at two soil sites. Journal of geotechnical and
that of piles after earthquakes, and the settlement of helical geoenvironmental engineering, 144(3): 06017021.
pile with double helices is larger than that of helical pile with Liam Finn, W.D., 2005. A study of piles during
single helix, which might be caused by large down drag earthquakes: issues of design and analysis. Bulletin of
force on the upper helix of double helices pile. Earthquake Engineering, 3: 141-234.
3. The axial load along the pile shaft above the Liu, C., Tang, L., Ling, X., Deng, L., Su, L. and Zhang, X.,
bottom helix increased for double helical pile after the 2017. Investigation of liquefaction-induced lateral load
earthquakes, but decreased below the bottom helix and on pile group behind quay wall. Soil Dynamics and
reduced to around 0 kN; the pile settlement of smooth pile Earthquake Engineering, 102: 56-64.
is larger than the surrounding soil, and the axial load along Liu, C., Wang, C., Fang, Q. and Ling, X., 2022. Soil-pile-
pile shaft of smooth pile decreases after earthquakes. quay wall interaction in liquefaction-induced lateral
4. The maximum bending moment for helical piles spreading ground. Ocean Engineering, 264: 112592.
occurs at the pile head, whereas for smooth piles, it occurs Orang, M.J., Boushehri, R., Motamed, R., Prabhakaran, A.
at the ground surface. and Elgamal, A., 2021. An experimental evaluation of
helical piles as a liquefaction-induced building
References settlement mitigation measure. Soil Dynamics and
Earthquake Engineering, 151: 106994.
Cerato, A.B., Vargas, T.M. and Allred, S.M., 2017. A critical Ridgley, N., 2015. Practice note 28: Screw piles:
review: State of knowledge in seismic behaviour of Guidelines for design, construction & installation. The
helical piles. DFI Journal-The Journal of the Deep Institution of Professional Engineers New Zealand Inc.
Foundations Institute, 11(1): 39-87. Shahbazi, M., Cerato, A.B., El Naggar, M.H. and Elgamal,
El Sharnouby, M.M. and El Naggar, M.H., 2018. Field A., 2020. Evaluation of seismic soil–structure
investigation of lateral monotonic and cyclic interaction of full-scale grouped helical piles in dense
performance of reinforced helical pulldown micropiles. sand. International Journal of Geomechanics, 20(12):
Canadian Geotechnical Journal, 55(10): 1405-1420. 04020228.
Elkasabgy, M. and El Naggar, M.H., 2018. Lateral vibration Ingold, T.S. and Miller, K.S. 1983. Drained Axisymmetric
of helical and driven steel piles installed in clayey soil. Loading of Reinforced Clay, Journal of Geotechnical
Journal of geotechnical and geoenvironmental Engineering, ASCE, 109: 883-898.
engineering, 144(9): 06018009.
Elsawy, M.K., El Naggar, M.H., Cerato, A. and Elgamal, A.,
2019. Seismic performance of helical piles in dry sand
from large-scale shaking table tests. Géotechnique,
69(12): 1071-1085.
Elsherbiny, Z.H. and El Naggar, M.H., 2013. Axial
compressive capacity of helical piles from field tests
and numerical study. Canadian Geotechnical Journal,
50(12): 1191-1203.
Strength Gain of Driven Steel Piles in
Variable Ground Conditions in Calgary,
Alberta
Imran Shah, Anwar Majid and Faris Alobaidy
AECOM Canada Ltd., Calgary, Alberta, Canada

ABSTRACT
A large number of driven steel H piles were installed in cohesive soils (clay and clay till) with layers of sand and silt to
support bridges for a major transportation project in Calgary, Alberta. The total number of driven steel piles installed were
871. Due to significant variability in subsurface conditions across the site and at each bridge site, the risk for increased
piling quantities and splicing existed which would have impacted the project schedule and cost. To manage this risk, it was
decided at the onset of the project to rely on the time dependent strength gain of soils (set up). This set up is attributed to
increase in effective stresses resulting from dissipation of porewater pressure in cohesive soils after pile driving. For this
purpose, Pile Dynamic Analyzer (PDA) testing was performed at End of Initial Drive (EOID) and at restrike followed by
Case Pile Wave Analysis Program (CAPWAP) analyses and calibration of analysis using Wave Equation Analysis Program
(WEAP). The EOID PDA testing was performed on at least 15 % of the total piles at each foundation element and restrike
testing was performed on at least one third of the PDA tested piles at EOID. The set up was found to be variable consistent
with variable subsurface conditions. The risk with respect to increased piling quantities and splices was also managed
using the set up.

This paper presents the strength gain of soils measured between the EOID and restrike PDA testing on large number of
piles installed in variable ground conditions.

RÉSUMÉ
Un grand nombre de pieux en H en acier battus ont été installés dans des sols cohésifs (argile et till argileux) avec des
couches de sable et de limon pour soutenir des ponts pour un important projet de transport à Calgary, en Alberta. Le
nombre total de pieux en acier enfoncés installés était de 871. En raison de la variabilité importante des conditions du
sous-sol sur le site et sur chaque site de pont, il existait un risque d'augmentation des quantités de pieux et d'épissage, ce
qui aurait eu un impact sur le calendrier et le coût du projet. Pour gérer ce risque, il a été décidé en début de projet de
s'appuyer sur le gain de résistance des sols en fonction du temps (mise en place). Cette configuration est attribuée à
l'augmentation des contraintes effectives résultant de la dissipation de la pression interstitielle dans les sols cohérents
après le battage des pieux. À cette fin, des tests de l'analyseur dynamique de pile (PDA) ont été effectués à la fin de
l'entraînement initial (EOID) et au réamorçage, suivis des analyses du programme d'analyse des ondes de pile de cas
(CAPWAP) et de l'étalonnage de l'analyse à l'aide du programme d'analyse des équations d'onde (WEAP). Les tests PDA
EOID ont été effectués sur au moins 15 % du nombre total de pieux à chaque élément de fondation et les tests de
réamorçage ont été effectués sur au moins un tiers des pieux testés PDA à EOID. La configuration s'est avérée variable
en fonction des conditions variables du sous-sol. Le risque lié à l'augmentation des quantités de pieux et des épissures a
également été géré à l'aide de la configuration.

Cet article présente le gain de résistance des sols mesuré entre l'EOID et le test PDA de réamorçage sur un grand nombre
de pieux installés dans des conditions de sol variables.

1 INTRODUCTION During preliminary and detailed design stages, the

ultimate axial capacities of the piles were estimated
A major transportation project in Calgary included considering skin friction and end bearing following
installation of driven steel piles to support twenty (20) Canadian Foundation Engineering Manual (CFEM, 2006)
bridge structures. The bridge locations are shown on and Robert (1997). A total of 871 piles were installed in
Figure 1. Five bridges (STR1.1 to STR1.5) were variable ground conditions. The pile design lengths varied
constructed as part of Valley Ridge Interchange and fifteen from 27 m to 51 m. The total design pile lengths were
bridges (STR2.1 to STR3.5) were constructed as part of 31,702 m. The subsurface conditions generally consisted
Systems Interchange (Intersection of Stoney Trail and 16th of clay till with layers of silt, sand and clay at variable
Avenue NW). Driven steel piles HP360×132 and depths and of variable thicknesses. During installation of
HP360×152 were used as foundations for the bridges. The piles for STR1.4 (the bridge to be constructed), the
design unfactored geotechnical resistances of the piles installed pile lengths extended below the design pile tip
varied from 2200 kN to 3600 kN. elevations to achieve the required ultimate capacity. This
was concerning because increased pile lengths would
require more splices as well impacting the project cost and element and restrike testing on at least one third of the PDA
schedule significantly. The project specifications allowed tested piles at EOID if the piles did not achieve the required
terminating piles at shorter lengths by relying on time- capacity.
dependent strength gain of piles (i.e., set up) provided it Sometimes, the project schedule did not allow for EOID
can be confirmed that the piles can achieve the required testing, therefore PDA testing was performed one or two
capacity based on set up. Therefore, it was decided to days after the EOID PDA testing. The PDA testing one or
manage this risk by relying on the set up of piles after two days after the EOID testing was referred to as End of
driving. Pile Dynamic Analyzer (PDA) testing was Redrive (EOR) testing. A total of 216 piles were PDA
performed on the selected piles at the End of Initial Drive tested, including 77 piles at EOID, 139 piles at EOR, and
(EOID) and restrike testing. The restrike testing was 103 piles at restrike, exceeding the project requirements.
required to be performed a minimum of seven days after This paper discusses the piles set up estimated from
EOID PDA testing. The PDA testing was followed by Case PDA testing results and application of the estimated set up
Pile Wave Analysis Program (CAPWAP) analyses and to the EOID capacity to manage the risk with respect to
calibration of initially performed analyses using Wave increased piling quantities and splices.
Equation Analysis Program (WEAP).
The project specifications required performing EOID
testing on at least 15 % of the total piles at each foundation

Figure 1. Bridge Locations and Surficial Geology

2 REGIONAL GEOLOGY glacial sediments overlying fairly thick pre-glacial sands

and gravels, underlain by bedrock.
The project is within the northern extent of the Porcupine On the escarpment slope itself, thin glacial deposits
Hills physiographic region, located just east of the Rocky (predominantly clay till) are underlain by bedrock.
Mountain Foothills. The project site is bound by an At the base of the Paskapoo escarpment where most
escarpment to the south and Bow River to the north; of the bridges were constructed, alluvial-fan sediment of
therefore, the site surficial geology is variable. bedded gravelly loam overly glacial pebble-loam (till). In
The project site extends from the southern side of the addition, potentially water bearing fan sediments exist at
pre-glacial Bow River Valley, up the Paskapoo the base of drainage channels running down the
escarpment, to the relatively level Springbank upland. The escarpment slope. These fan sediments include
upland geology consists of relatively thin, predominantly interbedded fine to coarse grained materials.
 The northern portion of the project is located within the 𝛼 = 1 (Su ≤ 32 kPa) Equation 2
Bow River Valley which is comprised of fluvial channel
sediment (gravel with minor sand) overlying non-marine 𝛼 = 0.35 + 170𝑆𝑢−1.6 (Su > 32 kPa) Equation 3
bedrock.
Bedrock underlying the project area consists The toe resistance of the piles in cohesive soils were
predominantly of mudstone (claystone) interbedded with determined using Equation 4 (CFEM, 2006).
cross-stratified siltstone and sandstone belonging to the
non-marine Porcupine Hills Formation (Prior et al. 2013). 𝑞𝑡 = 𝑁𝑡 𝑆𝑢 Equation 4
The surficial geology of project site at the bridge where;
locations is shown on Figure 1. qt = unit end bearing resistance;
Nt = bearing capacity factor
3 SUBSURFACE CONDITIONS = 9 (for pile diameter less than 0.5 m)

A geotechnical investigation program consisting of drilling For cohesionless soils (sand and gravel), Equations 5
testholes at the bridge abutment and pier locations was and 6 (Robert, 1997) were used to estimate the skin friction
carried out for all the bridges. The depth of testholes at the and end bearing.
bridge locations varied from approximately 35.5 meters
below ground surface (mBGS) to 43.1 mBGS. Standard 𝑓𝑠 = 1.9𝑁 Equation 5
Penetration Tests (SPTs) were conducted at 1.5 m
intervals to provide an indication of in-situ soil consistency and
and relative density. The SPT blow counts for 300 mm
penetration of the split spoon (SPT N-values) were also 𝑞𝑡 = 190𝑁 Equation 6
recorded. Samples were collected at regular intervals and
where stratigraphy changed. Disturbed samples were where;
collected from the augers and split spoon sampler for N = SPT N-value
laboratory testing and examination. Relatively undisturbed
Shelby tube samples were also collected from testholes for 5 PILE INSTALLATION AND MONITORING
strength testing.
The subsurface conditions at the Valley Ridge Pileco D30-32, Pileco D36-32, Pileco D62-22 and APE
Interchange consisted of alluvial fan sediment (silt and D46-42 diesel hammers were used to drive the H-piles at
sand) overlying glacial sediment (clay till and silt till) or the bridge foundation locations. Hammer driving energy
sediment deposited at and beyond the toe of a slope by and pile driving resistance were recorded in terms of blows
ephemeral streams. The glacial till (silt till and clay till) were per minute (BPM) and blows per 250 mm of pile
underlain by claystone and sandstone. The clay till penetration, respectively during pile monitoring. The pile
contained variable gravel, sand and silt contents. driving records also included start and finish dates and
The subsurface conditions at the Systems Interchange times for pile driving, pile plumbness (verticality), and pile
consisted of fluvial channel sediment (sand) and alluvial splice locations for the piles. Full-time construction
fan sediment (silt and sand) overlying glacial sediment monitoring was performed during pile driving.
(clay till and silt till). Variable energy is transferred to the piles during pile
driving when using the diesel hammers, depending on the
4 PILE DESIGN METHOD soil resistance. This variability in hammer energy was
measured in terms of BPMs which in turn provides the
The geotechnical resistance of the driven steel piles is hammer stroke lengths. Higher the BPM, lower the
derived from the skin friction and end bearing. The skin transferred energy, and hence smaller the stroke length.
friction and end bearing are contributed from the pile shaft At the beginning of the project, to satisfy the pile
and toe resistances, respectively. Two common methods termination criteria which was estimated using WEAP, the
used for estimating skin friction and end bearing are total installed pile lengths exceeded the design lengths.
stress and effective stress methods. For this project, the Therefore, it was decided to rely on PDA testing to verify
total stress method was used for cohesive soils (clay till, silt the pile capacities instead of termination criteria obtained
till and clay). The skin friction was calculated using from WEAP analyses. The project specifications allowed to
Equation 1 (CFEM, 2006). terminate piles above the design pile tip elevations, and
rely on set up, provided the set up is confirmed by PDA
𝑓𝑠 = 𝛼𝑆𝑢 Equation 1 testing, i.e., perform EOID PDA testing on 15 % of the piles
and perform restrike PDA testing on at least one third of
where; EOID tested piles to estimate set up.
fs = unit skin friction PDA testing was performed on selected piles
α = adhesion coefficient considering the BPMs and penetration resistances at each
Su = undrained shear strength of soil bridge foundation element. The piles with lowest resistance
(i.e., lowest blow counts for 250 mm for final 1000 mm
The adhesion coefficient (α) was determined using penetration) were selected for a range of BPMs observed
Equations 2 and 3 (Coduto, 1994, Robert, 1997). at each foundation element.
 The PDA testing included applying the driving hammer
stroke and measuring the subsequent force and velocity
response of the pile using two pairs of transducers and
accelerometers installed on to the pile. The PDA testing
resulted in evaluation of the driving hammer performance;
and provided estimated maximum tensile and compressive
stresses during pile driving; evaluated the integrity of the
piles; and estimated the piles static bearing capacities at
the time of testing.
As discussed earlier, EOR testing was performed one
to two days after the pile driving when the construction
schedule did not allow for EOID testing. Majority of the piles
achieved the required geotechnical resistances at EOID or
EOR. However, the piles that could not achieve the
required geotechnical resistances were allowed a resting
period of at least 7 days before performing the restrike Figure 2. Soil Setup Phases (after Komurka et al., 2003)
testing.
The CAPWAP analyses were performed on the PDA At the beginning of the project, the pile termination
tested piles that combined the measured force and velocity criteria was developed based on the WEAP analyses only.
data with wave equation analysis to estimate the soil Since the results obtained from WEAP analyses were
resistances (skin friction and end bearing) and based only on the static resistance and did not account for
compressive stresses of the piles. The CAPWAP results the piles set up, the actual pile lengths exceeded the
were incorporated to calibrate the WEAP analyses, design pile lengths.
calculate the soil set up and estimate the geotechnical
resistances of the remaining piles that were not PDA tested 7 DISCUSSION
at each foundation element.
The subsurface conditions at the site predominantly
6 TIME-DEPENDENT STRENGTH GAIN (SET UP) consisted of clay till with variable proportions of gravel,
sand and silt. In clay till and silt till, the excess pore water
The time-dependent strength gain (set up) of driven steel pressure due to pile driving dissipates at a slower rate and
piles is the ratio of the resistance at any time after pile the set up predominately corresponds to both phases 1 and
driving to the resistance at EOID (Stuart and Sandford, 2 and may also correspond to phase 3 (Figure 2). The
2015) and is predominately related to an increase in soil effect of pile type on set up is associated to pile/soil
resistance on a pile shaft (Komurka et al., 2003). Excess interface coefficient of friction (Preim et al., 1989). The set
pore water pressures are generated during pile driving up rate is also reported to decrease with increasing pile
resulting in reduction of effective stresses. Soil set up takes sizes (Camp and Parmar, 1999).
place as a result of excess pore water pressure dissipation Several empirical relationships have been used to
and corresponding increase in shear strength of soil. estimate the set up including Skov and Denver (1988) that
Based on Komurka et al. (2003), the soil set up is reported presents the set up linearly in relation to the log of time.
to have three phases as depicted in Figure 2.
First phase includes logarithmically nonlinear rate of 𝑄𝑡
excess pore water pressure dissipation due to soil ⁄𝑄 = 1 + 𝐴 [𝑙𝑜𝑔10 (𝑡⁄𝑡 )] Equation 7
0 0
disturbance. The dissipation of excess pore pressure is not
linear in this phase for some time after driving. The duration where;
of phase 1 depends on the type of soil and pile properties. Qt = pile capacity at time t after pile driving;
In fine grained (cohesive) soils, the dissipation of excess Q0 = pile capacity at time t0;
pore water pressure takes much longer compared to A = pile set up factor
granular (cohesionless) soils. t0 = time corresponding to start of phase 2
Second phase constitutes a logarithmically linear rate = (an empirical initial time value).
of excess pore water pressure dissipation and
corresponding set up. In this phase, the rate of pore water Skove and Denver (1988) recommended the pile set up
dissipation becomes constant resulting in gaining shear factor (A) of 0.6 for clays and 0.2 for sands. The t0 values
strength. The start of this phase is referred to initial time, of 1 day and 0.5 day were also suggested by Skove and
to . Denver (1988). The soil set up obtained from Equation 7
Aging is the third phase of set up where the rate of set does not differentiate between skin friction and toe
up reduces substantially and continues even after excess resistance contributions, and instead uses combined
pore water pressure dissipation and is understood to be resistances.
independent of effective stress. Based on Svinkin and Skov (2000), the set up can be
estimated from Equation 8.

𝑄𝑡
⁄𝑄 = 1 + 𝐵[1 + 𝑙𝑜𝑔10 (𝑡)] Equation 8
𝑜
 Where B is set up factor assumed to be equivalent to A It was found that start time of logarithmically linear set up
in Equation 7. was highly variable for set ups calculated from both skin
As mentioned earlier, a total of 77 piles were tested at frictions and total resistances.
EOID, 139 piles at EOR, and 103 piles at restrike. Majority The t0 varied from 0.1 days to 29.3 days for soil set up
of PDA tested piles achieved the required capacities at contributed by skin friction, while t0 corresponding to the
EOID or EOR. However; in instances where required total soil set up varied from 0.2 days to 22.4 days. This
capacities were not achieved during EOID or EOR, restrike represents the subsurface variability and corresponding
testing was performed. The soil set ups were calculated for non-linear porewater pressure dissipation after the pile
the piles by comparing the measured pile capacities at driving. Figure 3 presents the variation of t0 based on the
EOID and EOR and restrike stages. The minimum measured soil set ups.
calculated set ups were used for the remaining piles at
each foundation element to validate the achievement of
required capacities. This validation process included
calibration of the WEAP analyses for a range of BPMs and
stroke lengths since diesel hammers do not operate at
fixed stroke lengths. The BPMs and stroke lengths of diesel
hammers fluctuated based on the soil resistance.
Table 1 presents the soil set ups calculated for the piles
based on EOID and restrike tests only. Both the skin friction
and toe resistance were measured for each tested pile that
indicated the relative contribution of these resistances to
the total resistance. For majority of piles, skin friction was
the predominant contributor to the total piles resistances.
The average skin friction and toe resistance contributions
to the total resistance were found to be approximately 68
% and 32 %, respectively. The soil set ups were calculated
by comparing total resistances as well as the resistances Figure 3. Time Related to Start of Soil Set Up Phase 2
contributed by skin friction during the PDA testing.
The soil set ups from skin friction contributions varied 8 CONCLUSIONS
from 1.0 to 1.97, while the total resistance soil set ups
ranged from 1.05 to 1.76, with an average of 1.30. • The estimated set ups were variable consistent with
The measured soil set ups at each foundation element variability in subsurface conditions.
were used to calibrate the WEAP analyses. A range of • The risk with respect to increase in piling quantities
curves corresponding to various BPMs were developed to and number of splices was managed by terminating
verify that remaining piles have achieved the required pile tips at elevations higher than the design pile tip
capacities. elevations.
A comparison of the measured soil set ups and the set • The use of pile set ups resulted in substantial savings
ups computed from the empirical equations provided by in piling quantities and splices. It was estimated that
Skove and Denver (1988) and Svin and Skov (2000) was pile quantities were reduced by approximately 10 % .
also performed. The results indicate that both empirical • The empirical equations developed by Skove &
relationships over predicted the soil set ups for the piles at Denver (1988) and Svinkin & Skov (2000) were found
this project. The Skov and Denver (1988) overpredicted the to over predict the soil set ups at this site.
soil set ups between 2 % and 69 % with an average of 28 • The times corresponding to start of phase 2 of soil set
%. Similarly, the relationship proposed by Svinkin and Skov up (t0) were also found to be variable consistent with
(2000) overpredicted the soil set up in the range of 24 % to the variability in subsurface conditions.
119 % with an average over prediction of 71 %.
In addition, an effort was made to predict the time The soil set ups depend on site specific subsurface
corresponding to the start of phase 2 (i.e., t 0) for the soils conditions; therefore, PDA testing should be performed to
encountered at site. Pile set up factor (A) was normalized determine site specific set ups.
to account for soil proportions. The t0 was estimated based
on the normalized pile set up factor and measured set up.
Table 1. Measured Piles Set Ups between EOID and Restrike PDA Testing

Computed Resistance (kN)

Test Embedment
Structure Pile No. Set Up - Shaft Set Up - Total
(EOID/R) (m) Total Shaft Toe

EOID 32.00 1614 944 670

A1-2 1.37 1.20
R 32.40 1933 1290 642
Bridge 1.1
EOID 36.25 2595 1505 1090
A1-7 1.60 1.22
R 36.30 3175 2411 764
EOID 30.50 2700 2100 600
Bridge 1.2 A1-2 1.00 1.11
R 30.50 3000 2100 900
EOID 26.85 2449 1677 772
A1-2 1.01 1.05
R 26.90 2564 1698 865
EOID 26.60 2817 2001 815
Bridge 1.3 A1-4 1.08 1.40
R 27.25 3950 2159 1791
EOID 27.10 2248 1506 743
A2-2 1.15 1.13
R 27.20 2541 1729 812
EOID 30.50 2900 1700 1200
A1-14 1.26 1.29
R 30.50 3750 2150 1600
EOID 31.00 2950 1400 1550
A1-26 1.61 1.34
R 31.00 3950 2250 1700
EOID 37.10 2050 1350 700
A2-01 1.78 1.76
R 37.10 3600 2400 1200
Bridge 1.4
EOID 37.65 2390 1480 910
A2-05 1.86 1.55
R 37.75 3700 2750 950
EOID 33.90 2500 1800 700
A2-09 1.47 1.38
R 33.90 3450 2650 800
EOID 37.10 2200 1500 700
A2-13 1.57 1.57
R 37.10 3450 2350 1100
EOID 34.75 2504 1678 826
A1-8 1.29 1.25
R 34.90 3122 2172 950
EOID 35.00 1748 1010 738
Bridge 2.1 A1-16 1.21 1.15
R 35.00 2008 1221 787
EOID 34.75 1638 823 814
A1-17 1.97 1.44
R 35.00 2354 1625 728
EOID 25.75 3213 2557 656
Bridge 2.8 P-10 1.10 1.06
R 25.90 3417 2819 598
EOID 29.75 2465 2299 167
A2-24 1.15 1.41
R 29.75 3484 2638 847
Bridge 2.9
EOID 29.75 2503 2274 229
A2-25 1.21 1.37
R 29.75 3418 2745 673
EOID 29.00 2700 2050 650
Bridge 2.10 P-16 1.42 1.34
R 29.00 3620 2920 700
EOID 33.80 3300 2600 700
A1-3 1.06 1.06
R 34.00 3500 2750 750
EOID 34.00 3200 2500 700
A1-5 1.20 1.19
R 34.00 3800 3000 800
EOID 34.00 3000 2200 800
Bridge 2.12 A1-8 1.31 1.23
R 34.00 3700 2875 825
EOID 33.75 3300 2550 750
A2-5 1.31 1.24
R 34.00 4100 3350 750
EOID 34.75 3200 2350 850
A2-8 1.34 1.25
R 34.75 4000 3150 850
EOID 34.75 2580 1522 1058
A2-8B 1.49 1.38
R 35.10 3559 2274 1284
Bridge 3.3
EOID 34.80 2871 1795 1054
A2-9B 1.40 1.28
R 34.85 3650 2507 1143
8.1 References

Prior, G.J., et.al. 2013. Bedrock geology of Alberta; Alberta

Energy Regulator, AER/AGS (Map 600).
Canadian Geotechnical Society 2006. Canadian
Foundation Engineering Manual, 4th Edition, BiTech
Publishers Ltd., Vancouver, BC.
Camp, W.M., and Parmar, H.S. 1999. Characterization of
pile capacity with time in the Cooper Marl: Study of
applicability of a past approach to predict long-term pile
capacity, Transportation Research Record, 1663(1):
16-24.
Coduto, D.P. 1994. Foundation design, principles and
practices. Prentice-Hall Inc., Englewood Clifs, N.J.
Komurka, E. V., Wagner, B. A., and Edil, B. T. 2003, A
review of pile set-up, Proceedings, 51st Annual
Geotechnical Engineering Conference, University of
Minnesota, St. Paul, Minnesota, February 21, 2003:
105-130.
Komurka, E. V., Wagner, B. A., and Edil, B. T. 2003.
Estimating soil/pile set-up. Wisconsin Highway
Research Program # 0092-00-14.
Robert, Y. 1997. A few comments on pile design. Canadian
Geotechnical Journal (34): 560-567.
Preim, M.J., March, R., and Hussein, M. 1989. Bearing
capacity of piles in soils with time dependent
characteristics, Piling and Deep Foundations, Volume
1: 363-370.
Skov, R. and Denver, H. 1988. Time-dependence of
bearing capacity of piles. Proceedings, 3rd
International Conference on application of stress-wave
theory to piles, Ottawa, Ontario, Canada, May 25-27,
1988: 879-888.
Stuart, C. J. and Sandford, T. C. 2015. Set-up factors for
driven piles through the Presumpscot Formation. 2nd
Symposium on the Presumpscot Formation, Portland,
Maine, USA, October 28, 2015.
Svinkin, M. R. and Skov, R. 2000. Set-up effect of cohesive
soils in pile capacity, Proceeding, 6th International
Conference on application of stress-wave theory to
piles, Sao Paulo, Brazil, September 11-13, 2000: 107-
111.
Performance of composite caisson-pile
foundation (CCPF) under axial and
monotonic lateral loading
Rajan KC
Department of Civil Engineering, Pulchowk Campus, IoE, TU,
Nepal
Keshab Sharma
BGC Engineering Inc., Fredericton, NB, Canada
Indra Prasad Acharya
Department of Civil Engineering, Pulchowk Campus, IoE, TU, Nepal
Prishati Raychowdhury
Indian Institute of Technology (IIT), Kanpur, India

ABSTRACT
Composite Caisson-Pile Foundation (CCPF) is a novel type of hybrid foundation developed from the combination of the
caisson and piles foundation to take advantage of both foundation types. CCPF minimizes construction time and cost and
offers innovative solutions to challenging site circumstances in deep water. The CCPF may decrease the length and
embedment depth of the caisson and pile foundation compared to traditional foundations. However, there are still a few
crucial hurdles to fully accepting CCPF foundations as primary foundations in deep water with complex geology.
Comprehensive research is required to evaluate the performance of the CCPF and to increase the confidence level in the
use of CCPF, despite the lack of research in the direct performance analysis of the CCPF under various loading and soil
conditions. A series of reduced-scale tests were performed on CCPF. The load settlement, load improvement ratio (LIR),
settlement reduction ratio (SRR), and load sharing characteristics of the CCPF were investigated at varying relative sand
densities and pile configurations to comprehend the effect of various influencing factors on the behavior of CCPF. The test
results indicate that adding piles under the caisson can effectively increase load capacity and decrease the settlement of
the foundation system. A three-dimensional finite element model of CCPF was prepared in PLAXIS 3D and calibrated
using the reduced-scale laboratory test results.

RÉSUMÉ
Composite Caisson-Pile Foundation (CCPF) est un nouveau type de fondation hybride développé à partir de la
combinaison de la fondation caisson et pieux pour tirer parti des deux types de fondation. Le CCPF minimise le temps et
les coûts de construction et propose des solutions innovantes aux conditions de chantier difficiles en eau profonde. Le
CCPF peut réduire la longueur et la profondeur d'encastrement de la fondation en caisson et sur pieux par rapport aux
fondations traditionnelles. Cependant, il reste encore quelques obstacles cruciaux à l'acceptation complète des fondations
CCPF en tant que fondations primaires en eau profonde avec une géologie complexe. Des recherches approfondies sont
nécessaires pour évaluer les performances du CCPF et pour augmenter le niveau de confiance dans l'utilisation du CCPF,
malgré le manque de recherche dans l'analyse directe des performances du CCPF dans diverses conditions de charge et
de sol. Une série d'essais à échelle réduite a été effectuée sur le CCPF. Le tassement de charge, le rapport d'amélioration
de charge (LIR), le rapport de réduction de tassement (SRR) et les caractéristiques de partage de charge du CCPF ont
été étudiés à différentes densités relatives de sable et configuration de pieu pour comprendre l'effet de divers facteurs
d'influence sur le comportement du CCPF. Les résultats des tests indiquent que l'ajout de pieux sous le caisson peut
effectivement augmenter la capacité de charge et diminuer le tassement du système de fondation. Un modèle d'éléments
finis en trois dimensions de CCPF a été préparé dans PLAXIS 3D et calibré à l'aide des résultats des tests de laboratoire
à échelle réduite.

1 INTRODUCTION Moreover, caissons are preferable for major bridges

built on liquefiable soils. However, challenges arise during
The caisson foundation, also known as a well foundation, construction due to difficulties in sinking the caissons to the
is an ideal choice for ensuring stability and functionality in desired depth, leading to frequent obstacles (Abdrabbo
the case of long-span bridges (Chowdhury et al. 2017, Bag and Gaaver 2012). For instance, a 35-meter-long bridge
et al. 2021). It is particularly suitable for rivers where the with five spans in Lumbini, Nepal, faced a significant delay
depth of alluvial deposits is significant (Mukherjee et al. of over ten years in completing its caisson foundations due
2015) and when there is a risk of deep scour around the to unexpected hard strata encountered within the design
pier foundations (Kothyari 2007). With its high section depth (MoPIT 2020). Similar issues have been observed in
modulus, the caisson foundation can withstand substantial various bridge projects, such as the Sunkoshi Bridge
horizontal and vertical stresses, even when the (Khurkot), Kaligandaki Bridge (Ridi), and Sunkoshi Bridge
unsupported length is extensive in scouring riverbeds. (Ghurmi) in Nepal (Misra et al. 2022, KC et al. 2023). India
also experiences similar challenges with bridges like the Despite the need for more research in directly analyzing
Brahmaputra Bridge (Tezpur), Ganga Bridge (Bhagalpur), the performance of CCPF under different loading and soil
Pasighat Bridge (Arunachal), and Tapi Bridge conditions, there are still significant challenges to
(Maharashtra) (Bag et al. 2021). To address these overcome before fully adopting CCPF foundations as the
problems, the concept of a caisson-pile composite primary choice in deep water with complex geology. It is
foundation (CCPF) has been introduced. Recently, Nepal necessary to conduct extensive research to assess the
constructed a highway bridge using CCPF (Figure 1), performance of CCPF and enhance the level of confidence
drawing on the bridge construction experience of the in its utilization.
Qiongzhou Straits Bridge in China (KC et al. 2022b) and
the Longna Highway Yangtze River Bridge (Gong et al. In this study, reduced-scale tests were conducted on
2020). Cross-sea bridges like the Benicia Martinez Bridge the CCPF to examine its load settlement, load
in the United States and the Hitsuishi-Jima Bridge in Japan improvement ratio (LIR), settlement reduction ratio (SRR),
have also utilized CCPF (Tu et al. 2021). and load sharing characteristics. The test outcomes
demonstrate that incorporating piles beneath the caisson
effectively enhances load capacity and reduces foundation
settlement. Additionally, a three-dimensional finite element
model in PLAXIS 3D was prepared to compare with the
results of the reduced-scale laboratory tests.

2 METHODOLOGY

The experimental model tests conducted at the laboratory

setup of the Indian Institute of Technology Kanpur (IITK),
Kanpur, India, utilized uniformly graded Ganga sand
sourced from the Ganga River as the soil bed. The physical
and mechanical properties of the Ganga sand were
Figure 1. Tinau River Bridge (Nepal) is suffering from determined. The soil was classified as SP (poorly graded
sinking issues in caisson foundations and the introduction sand) according to the USCS classification, with a sand
of CCPF as solution (KC et al. 2022a) fraction of 98.19%. The specific gravity and unit weight of
the sand was measured as 2.67 and 14.07 kN/m3,
Furthermore, it is crucial to consider the issues and respectively. The angle of internal friction (ɸ) was
challenges associated with pile foundations, along with determined to be 32.5°, while the maximum and minimum
some relevant case examples. The CCPF is a recently void ratios were obtained as 0.99 and 0.70, respectively.
developed innovative concept in bridge foundation
engineering (Figure 2) that combines the benefits of two
common deep foundation types: pile groups and caisson Table 1. Characteristics of the caisson and pile prototype
foundations. It was initially proposed for a highway crossing and model properties.
the Qiongzhou Straits in the South China Sea, aiming to
provide effective solutions for complex site conditions in
deep-water areas and reduce construction time and costs Properties Prototype Model
(Zhong and Huang 2014, Huang et al. 2018). Compared to
conventional foundations, the CCPF allows for shorter and Modulus of elasticity,
27,386 2,10,000
less deeply embedded caissons and piles, thereby E (N/mm2)
simplifying construction and minimizing risks.
Poisson’s ratio, υ 0.2 0.3

External diameter of
7000 90
caisson (mm)

Thickness of caisson (mm) 900 18

Caisson depth (mm) 9000 180

Pile diameter (mm) 800 10

Pile depth (mm) 23000 460

Figure 2. A schematic diagram illustrating the sectional

view of the CCPF system.
 For the experimental loading tests, the design of sand in a defined volume to achieve the desired density
dimensions of the Tinau River Bridge in Lumbini Province, through compaction. A total of 12 sensors were utilized to
Nepal, as designed by the Department of Roads, Nepal, measure various responses of the foundation and soil. All
were used as a prototype. A similitude philosophy based the sensors were connected to a data acquisition (DAQ)
on 1-g load tests, as described by Iai (1989), was applied. system. LabView 2017 software was employed to program
A scaling factor of 50 was determined for the prototype the system and automate the data collection process.
M30 grade RCC (reinforced cement concrete) caisson and
piles. Mild steel was used to construct the models, with a 3 RESULTS AND DISCUSSIONS
modulus of elasticity of 210 GPa and a Poisson's ratio of
0.3. The properties of both the prototype and model Figure 4 illustrates the load versus settlement plots for the
caisson and piles are presented in Table 1. caisson only, CCPFtop, and CCPFbottom under two relative
density conditions: loose and dense sand. At a settlement
of 10 mm (equivalent to 10% of the caisson diameter), the
ultimate load-carrying capacities for loose sand are
determined as 188 N for the caisson only, 267 N for
CCPFbottom, and 334 N for CCPFtop. CCPFtop exhibits the
highest load-carrying capacity (334 N), while the caisson
alone demonstrates the lowest (188 N). In the case of
dense sand, the ultimate load-carrying capacities at 10 mm
settlement are measured as 904 N for the caisson only,
1186 N for CCPFbottom, and 1213 N for CCPFtop. These
results highlight that the load-carrying capacity of CCPF
increases with an increase in the density of the soil bed for
all examined scenarios. Additionally, it is observed that the
vertical load-carrying capacity of CCPF surpasses that of
the caisson alone, emphasizing the beneficial effect of
incorporating piles to enhance the load capacity of the
caisson.

Figure 3. Experimental loading setup for CCPF

A cubical tank with sides measuring 0.6 m was

employed for the model experiments. To minimize the
influence of boundaries, the side boundaries of the tank
were maintained at a distance greater than 2.5 times the
caisson diameter, while the bottom boundary was
maintained at a distance greater than ten times the pile
diameter from the top of the pile.

The loading tests on the CCPF were conducted at the

structural laboratory of the Indian Institute of Technology
(IIT), Kanpur, India (Figure 3). The experimental setup
consisted of the CCPF with piles connecting at the base of
caisson (CCPFbottom) and the CCPF with piles extending up
to the top of the caisson (CCPFtop). In both cases, a
hydraulic jack with a capacity of 10 kN was used to apply Figure 4. Load-displacement curves illustrate the behavior
vertical load on the model foundation. The loading frame of CCPF in loose (40%) and dense (70%) sand.
consisted of four vertical columns measuring 1.5 m in
height, with two columns on each side and two horizontal The load-improvement ratio (LIR) is a non-
beams. A calibrated load cell with a capacity of 500 kN was dimensional parameter that quantifies the enhancement of
attached to the jack to measure the load. the external load-carrying capacity of the caisson when
compared to the caisson alone at a specific settlement. It
In the model setup, the sand bed was prepared to is expressed as a fraction of the load supported by the
achieve relative densities of 40% (loose) and 70% (dense). CCPF relative to the load supported by the caisson alone.
Calibration tests determined that a fall height of 15 cm was Figure 5a presents the load improvement observed in
optimal for attaining the specified relative density. A relative different CCPF models under loose and dense sand
density of 70% was also achieved using the hammer blow conditions. It is evident that, for both sand densities, CCPF
method. The total height of the tank (60 cm) was divided with piles extending to the top of the caisson (CCPFtop)
into equal intervals marked by signs every 10 cm along the exhibits higher LIR values compared to CCPFbottom.
inner side, allowing for the placement of a specific weight
 The reciprocal of the Load Improvement Ratio (LIR)
can be interpreted as the percentage of the total applied
load that is borne by the caisson alone. This parameter can
be helpful in preliminary assessments for understanding
the load-settlement behavior of the CCPF system. The
distribution of load sharing between the caisson and piles
is a critical design consideration for the CCPF system.

Figure 6 illustrates the proportion of the load carried

by the caisson as the vertical load increases. It is observed
that, initially, the load-bearing capacity of the caisson
increases with the applied load. However, beyond a certain
point, when the load exceeds a certain threshold, the load-
bearing ratio of the caisson starts to decrease.

This information provides valuable insights into the

load sharing characteristics of the CCPF system, aiding in
the understanding of its performance under varying load
levels.

Figure 5. (a) Load improvement ratio and (b) Settlement

reduction ratio of CCPF for loose (40 %) and dense (70%) Figure 6. Load sharing analysis of caisson for vertical
sand in vertical loading loading in loose (40%) and dense (70%) sand.

The settlement reduction ratio (SRR), also known as In order to investigate the response of CCPF under
the settlement efficiency ratio, is a non-dimensional combined vertical and lateral loading conditions, we
parameter that represents the ratio of settlement between conducted a series of tests using a combination of these
the CCPF and the caisson alone at a specific load. It is two loads on the CCPF model. Figure 7 illustrates the
calculated as the settlement of the caisson alone (δ c) vertical load versus displacement graph for four different
divided by the settlement of the CCPF (δccpf) at that vertical loading scenarios (No load, 300 N, 600 N, and 900
particular load. N vertical loads), specifically in dense sand conditions. The
CCPF configuration with piles extending to the bottom of
Figure 5b illustrates the settlement performance of the caisson (CCPFbottom) was employed for these tests.
the caisson alone, CCPFtop, and CCPFbottom in two relative
densities, presented as the SRR. As the settlement of the From the graph, it can be observed that as the vertical
caisson alone (δc) is always greater than the settlement of overburden load increases, the lateral displacement is
the CCPF (δccpf), the SRR values range between 0 and 1. significantly reduced when subjected to monotonic lateral
A higher SRR value indicates that the CCPF is more load. This indicates that the application of increased
efficient in reducing settlement at a given load level. It was vertical load results in a notable decrease in lateral
observed that the SRR varies in an almost nonlinear displacement of the CCPF.
manner with the applied load. Moreover, the CCPFs in
loose sand demonstrate higher settlement reduction
compared to the CCPFs in dense sand across the entire
range of considered load levels.
 The experimental tests conducted on CCPF with
sand in this study were utilized to compare with the 3D
numerical model (Figure 8a) prepared in PLAXIS 3D. The
caisson and piles were represented in the model using
plate and embedded beam elements, respectively. To
assess the accuracy of the numerical model, the vertical
load displacements of the CCPF with piles extending to the
bottom were obtained through numerical simulations for
dense sand conditions. These results were then compared
to the experimental measurements reported in this study,
as shown in Figure 8b. The comparison reveals that the
experimental and numerical outcomes are within an
acceptable range of variation, indicating a reasonable level
of agreement between the two.

Figure 7. Variation of lateral load with lateral displacement 4 CONCLUSIONS

of CCPF in dense sand (70%) under different vertical load
conditions.
The performance of composite caisson-pile foundation
(CCPF) under axial and monotonic lateral loading was
evaluated using reduced-scale laboratory tests and
numerical modeling. Conclusions of this research are as
follows.

The CCPF effectively maximizes the vertical load carrying

capacity of the soil while controlling settlement. In loose
sand, the ultimate vertical load-carrying capacity increased
by 42% and 77% for CCPFbottom and CCPFtop, respectively.
In dense sand, the increment percentages were 31% and
34%.

The inclusion of piles beneath the caisson foundation

enhances the lateral load bearing capacity. In loose sand,
the ultimate lateral load-carrying capacity increased by
141% and 190% for CCPFbottom and CCPFtop, respectively.
In dense sand, the increment percentages were 97% and
197%. Both the vertical and lateral load-carrying capacity
of CCPF increased approximately 5 and 2 times,
respectively, as the density of the soil bed increased from
40% to 70%.

With an increase in vertical overburden load, the lateral

displacement significantly decreased when subjected to
monotonic lateral load. Increasing the vertically applied
load from 300 N to 900 N increased the ultimate capacity
by 1.4 times. The load improvement ratio (LIR) exhibits a
nonlinear variation with settlement in both vertical and
lateral loading conditions. CCPFtop demonstrates higher
LIR values than CCPFbottom.

The strength reduction ratio (SRR) values for CCPFbottom

are higher than those of CCPFtop across all load levels. The
load bearing ratios of the caisson initially increase with an
increasing load but decrease when the load exceeds the
highest point. The load sharing curve exhibits nonlinear
behavior.
Figure 8. Figure showing (a) 3D numerical model of CCPF
in PLAXIS 3D and (b) comparison 3D model with the
vertical test results.
 Hydrology: Regional Studies, 4: 1–14. Elsevier.
REFERENCES doi:10.1016/J.EJRH.2015.03.005.
Tu, W., Gu, X., Chen, H.P., Fang, T., and Geng, D. 2021.
Abdrabbo, F., and Gaaver, K. 2012. Challenges and Time domain nonlinear kinematic seismic response
Uncertainties Relating to Open Caissons. DFI of composite caisson-piles foundation for bridge in
Journal - The Journal of the Deep Foundations deep water. Ocean Engineering, 235: 109398.
Institute, 6(1): 21–32. doi:10.1179/dfi.2012.002. Pergamon.
Bag, R., Mandal, M., and Singhal, V. 2021. Rectification of doi:10.1016/J.OCEANENG.2021.109398.
the Tilt and Shift of Well Foundation: A Numerical Zhong, R., and Huang, M. 2014. Winkler model for dynamic
and Analytical Solution. Indian Geotechnical Journal, response of composite caisson-piles foundations:
51(5): 1011–1025. Springer India. Seismic response. Soil Dynamics and Earthquake
doi:10.1007/s40098-020-00477-5. Engineering, 66: 241–251. Elsevier Ltd.
Chowdhury, I., Tarafdar, R., Ghosh, A., and Dasgupta, S.P. doi:10.1016/j.soildyn.2014.07.005.
2017. Dynamic soil structure interaction of bridge
piers supported on well foundation. Soil Dynamics
and Earthquake Engineering, 97: 251–265. Elsevier.
doi:10.1016/J.SOILDYN.2017.03.005.
Gong, W.M., Wang, Z.Z., Dai, G.L., Liu, X.G., Zhu, M.X.,
Zhao, X.X., and Guo, H. 2020. Foundations of
Yangtze River mainstream bridges in China.
Proceedings of the Institution of Civil Engineers:
Forensic Engineering, 173(1): 13–24.
doi:10.1680/jfoen.19.00010.
Huang, M., Tu, W., and Gu, X. 2018. Time domain
nonlinear lateral response of dynamically loaded
composite caisson-piles foundations in layered
cohesive soils. Soil Dynamics and Earthquake
Engineering, 106(December 2016): 113–130.
Elsevier Ltd. doi:10.1016/j.soildyn.2017.12.015.
Iai, S. 1989. Similitude for Shaking Table Tests on Soil-
Structure-Fluid Model in 1g Gravitational Field. Soils
and Foundations, 29(1): 105–118. Elsevier.
doi:10.3208/SANDF1972.29.105.
KC, R., Acharya, I.P., and Sharma, K. 2022a. A Review on
the Behavior of Composite Well-Pile Foundation
(CWPF) under Static and Dynamic Loading. In
Proceedings of 11th IOE Graduate Conference.
Lalitpur, Nepal.
KC, R., Sharma, K., Burnwal, M.L., Raychowdhury, P.,
Acharya, I.P., and Misra, J. 2022b. Load-settlement
behaviour of composite caisson-pile foundation
(CCPF) in sand. In IGC 2022.
KC, R., Sharma, K., Raychowdhury, P., Acharya, I.P.,
Burnwal, M.L., and Misra, J. 2023. State-of-the-art
review of composite caisson-pile foundation (CCPF).
Applied Ocean Research, 136: 103571.
doi:https://doi.org/10.1016/j.apor.2023.103571.
Kothyari, U.C. 2007. Indian practice on estimation of scour
around bridge piers - A comment. Sadhana -
Academy Proceedings in Engineering Sciences,
32(3): 187–197. doi:10.1007/s12046-007-0017-7.
Ministry of Physical Infrastructure and Transport, G. of N.
2020. Yearly Progress Report F.Y. 077/078.
Misra, J., Subedi, M., and Sharma, K. 2022. Design of
composite well-pile foundation (CWPF): A case
study of bridge foundation in Nepal. In Proceedings
of the 20th International Conference on Soil
Mechanics and Geotechnical Engineering, Sydney
2021.
Mukherjee, A., Saha, D., Harvey, C.F., Taylor, R.G.,
Ahmed, K.M., and Bhanja, S.N. 2015. Groundwater
systems of the Indian Sub-Continent. Journal of
Monday, October 2, 2023

GEOHAZARDS I
A Case Study: Post construction
performance of rockfill columns installed in
clay shale – Theodore Dam Right Abutment

Cody Anderson, David Elwood

University of Saskatchewan, Saskatoon, Saskatchewan, Canada
Keisha Hertz
Water Security Agency, Regina, Saskatchewan, Canada
Imteaz Bhuiyan
City of Regina, Regina, Saskatchewan, Canada

ABSTRACT
Rockfill columns replace weaker native soils with a granular material of high frictional strength to develop a new shear
strength comprised of both the native soil and the replacement material. This new strength from the combined materials is
typically analyzed as an equivalent shear key. Application of rockfill columns for slope stabilization provides the ability to
extend through zones of weak soil with minimal disturbance to the surrounding ground, resulting in a general improvement
in factor of safety throughout construction. However, before this strength is realized, deformations are required to mobilize
the friction of the rockfill columns with deformations typically ceasing within two years. In November 2020, 72 rockfill
columns were installed into the natural valley wall in the right abutment to Theodore dam. The site consists of a clay till
underlain by a pre-sheared, weathered clay shale and then by intact clay shale. Displacements have been shown to take
place along an existing shear zone within the disturbed clay shale. Perched groundwater discharges into the clay till and
disturbed clay shale from upslope. This perched groundwater is heavily influenced by precipitation. To provide a full
understanding of movements within this slope, four shape accelerometer arrays (SAA) were installed before and after the
construction in order to observe deformations. Two SAAs were installed three months before construction located above
and below the rockfill column placement to observe overall site deformations. After construction, two additional SAAs were
installed, each in a separate rockfill column to observe deformations within the rock column shear key. The deformation
monitoring post-construction indicates continued deformations are driven by seasonal fluctuations. This paper provides a
summary of the observations made from instrumentation and provides comments on the performance of the remediation
method with correlation to laboratory data conducted on select clay shale specimens from the site.

RÉSUMÉ
Les colonnes d'enrochement remplacent les sols natifs plus faibles par un matériau granulaire à haute résistance au
frottement pour développer une nouvelle résistance au cisaillement composée à la fois du sol natif et du matériau de
remplacement. Cette nouvelle résistance des matériaux combinés est généralement analysée comme une clé de
cisaillement équivalente. L'application de colonnes d'enrochement pour la stabilisation des pentes offre la possibilité de
s'étendre à travers des zones de sol faible avec une perturbation minimale du sol environnant, ce qui entraîne une
amélioration générale du facteur de sécurité tout au long de la construction. Cependant, avant que cette résistance ne
soit réalisée, des déformations sont nécessaires pour mobiliser le frottement des colonnes d'enrochement, les
déformations cessant généralement en deux ans. En novembre 2020, 72 colonnes d'enrochement ont été installées
dans le mur naturel de la vallée dans la culée droite du barrage Théodore. Le site consiste en un till argileux reposant
sur un schiste argileux altéré et pré-cisaillé, puis sur un schiste argileux intact. Il a été démontré que des déplacements
ont lieu le long d'une zone de cisaillement existante dans le schiste argileux perturbé. Les eaux souterraines perchées
se déversent dans le till argileux et le schiste argileux perturbé depuis le haut de la pente. Cette nappe phréatique
perchée est fortement influencée par les précipitations. Pour fournir une compréhension complète des mouvements au
sein de cette pente, quatre réseaux d'accéléromètres de forme (SAA) ont été installés avant et après la construction
pour observer les déformations. Deux SAA ont été installés trois mois avant la construction situés au-dessus et au-
dessous de la mise en place de la colonne d'enrochement pour observer les déformations globales du site. Après la
construction, deux SAA supplémentaires ont été installées, chacune dans une colonne d'enrochement distincte pour
observer les déformations dans la clé de cisaillement de la colonne rocheuse. La surveillance de la déformation après la
construction indique que les déformations continues sont entraînées par les fluctuations saisonnières. Cet article fournit
un résumé des observations faites à partir de l'instrumentation et fournit des commentaires sur les performances de la
méthode d'assainissement avec une corrélation avec les données de laboratoire menées sur des spécimens de schiste
argileux sélectionnés.
1.0 INTRODUCTION concentration within the water bath throughout the testing
process.
Rockfill columns as a slope stabilization method intercept
weaker soil zones by removing weak soil and replacing it Laboratory tests conducted on rockfill aggregate indicate
with a high frictional strength material the same way as a the importance of densification to achieve high peak shear
shear key does. The major difference is that not all native strength. Tests indicate higher densification effort allowed
soil will be replaced. The resulting resistance is created by for higher peak shear strengths to be realized with less
both the aggregate referred to as rockfill and the native soil, strain, ultimately allowing for less deformations before
with strength properties determined with the area resistance of rockfill was engaged.
replacement ratio. This ratio provides an equivalent
strength of the native soil and aggregate (Thiessen 2011). 1.1 Background

Instrumentation prior to abutment remediation reported Theodore Dam impounds the Whitesand River
that deformations were occurring along a slide plane approximately 40 km northeast of Yorkton, Saskatchewan.
located at approximately 490 meters above sea level The dam’s construction was completed 1964 for the
(masl). The resulting movements were consistent with a purpose of flood control and has since been utilized for
translational failure mode on a weak, residual plane within irrigation and recreation. Theodore Dam has a riparian
the native Cretaceous clay shale. The primary concerns of outlet located in the middle of the embankment, and a
the Saskatchewan Water Security Agency (WSA) were to concrete spillway constructed in the right abutment as seen
cease (or reduce the movement significantly in an in Figure 1. Instabilities within the right abutment were first
acceptable level of risks) displacements along the existing documented in 1997. Site observations revealed the
slip surface and stabilization was required to address presence of localized circular slump blocks that emerged
potential concerns of impacts to the spillway structure. from the middle of the slope on the natural valley wall.
Concerns driving the slip surface below the designed shear
key prompted a design selecting the rockfill column Figure 1: Right Abutment of Theodore Dam, Dec 2020
termination depth where the factor of safety increase from
the shear zone would be the same if a slip surface Cross Section D
developed below the columns.

Another major concern was the potential for longer-term

strength reduction of clay shale e.g., softening of the clay
shale at depth due to meteoric water infiltration and
swelling. The stone column solution would change the
site’s hydrogeologic flow by introducing relatively high Efforts over the next decade to stabilize the right abutment
permeable materials that penetrated an aquifer perched on included installing French drains and regrading with a mid
the shale surface, and directly introduced fresh water to slope bench. These improvements were unsuccessful in
lower horizons. Because the Cretaceous shale in improving slope stability and prompted installation of a
Saskatchewan is composed of minerals and pore fluids mechanically stabilized earth wall (MSE) of free draining
prone to ionic exchange in the presence of fresh water, material at the toe of the slope near to the existing spillway.
there was the potential for the stone columns to indicate After the construction of the MSE wall, the slope was
short-term stabilization, followed by longer-term instrumented with inclinometers and piezometers in the
movements following seasonal impact to the shale right abutment to monitor performance of the remediation.
geochemistry. This paper discusses a series of laboratory Movements associated with the unloading of the toe for the
tests to determine the swell potential of the clay shale when construction of the MSE wall were not monitored.
subjected to deionized water, measurement of the residual Additional progressive failure within the clay shale may
shear strength in a series of ring shear tests; direct shear have occurred at the time of construction, which in turn
testing of the aggregate used for construction of the rockfill would have potentially rendered the implemented remedial
columns. method ineffective.

The results of laboratory testing indicated clay shale The post-MSE wall instrumentation indicated continued
reconstituted with deionized water had the potential to deformations within the right abutment, with movement
dramatically impact the measured residual strength values taking place at approximately at 490 masl or 9 m below the
of the clay shale. Once this residual shear value was previous ground surface. This movement imposed a risk to
obtained on specimens reconstituted with deionized water the spillway and outlet channel and prompted the most
were reintroduced to a synthetic pore fluid representative recent stabilization effort by rockfill columns in 2020.
of in situ pore fluid an improvement in the residual shear
strength was observed. The improvement in residual shear 1.2 Site Geology
strength indicates an ongoing ion exchange between the
water bath and the test specimen, highlighting the The right abutment consists of a natural valley wall with
importance of monitoring the water bath chemistry and three strata: intact Cretaceous clay shale (Pierre Shale),
making regular adjustments to maintain the desired salt disturbed clay, and an overlying silty clay till. A perched
water table resides within the silty clay till. The disturbed
clay appears brecciated and weathered and is vibro-lance was lowered tip first into each rockfill column to
approximately 3 m in thickness and is encountered at the termination depth. The aggregate was densified in 1-
approximately 488 to 492.5 masl. The weathered shale lies meter lifts for 1 minute intervals or until the frequency
on top of the intact clay shale. Analysis of the dropped to the point of tip stall. At this point, the lance was
displacements throughout the site indicates a general raised by 1 m until the entire column was densified.
movement direction in alignment with cross section D as Densification of the rockfill was specified at 15% of initial
shown on Figure 1. density. The improvement was assessed by measuring the
drop in rockfill before and after densification . The resulting
Soil classification further supports a disturbed layer as the change in height from densification was conical with most
material appears 40% mottled distinct blocky clay chunks settlement occurring at the centre of the column.
by colours of light grey, dark grey and brown within the Measurements to assess percent increase in density were
disturbed clay shale. The intact clay shale is consistent with required around the perimeter and centre of a given column
Pierre shale, based on mapped geology of the area. to obtain an average settlement of the placed rockfill.
Repeated glaciation has eroded the valley down to the
marine clay, creating the disturbed clay layer above the 1.3.1.1 Variation in column densification
intact clay. Based on the observed contact elevations of the
weathered/disturbed clay shale, the disturbed layer is Six columns were densified without water and limited the
consistent with where deformations within the slope are densification efficiency to a maximum of 8-10%. Because
concentrated. the target densification was not being consistently
The groundwater level has been measured in a number of achieved, water was introduced as to facilitate ground
open standpipes and vibrating wire piezometers located improvement. In cases where densification specifications
throughout the slope. The groundwater levels were found were not achieved, the column aggregate would be
to range from EL 491.2 to 500.7 masl. The higher excavated halfway, and densification was completed in two
groundwater levels correspond with piezometers installed lifts; densification in two lifts allowed the column to meet
near to the crest of the abutment, while the lower values specifications but introduced a risk of a poor interfacing
correspond with piezometers installed near the toe of the between the lifts. To reduce this risk of a weak interface
abutment. Considering the ground surface elevation of the between the successive lifts, the vibro-lance was inserted
various piezometers, the depth to the groundwater level 1 to 2 m into the previous lift while maintaining vibration.
below the ground surface ranged from 1.9 to 7.4 m with an Generally, densification in 1 lift with the addition of water
average depth of 4.5 m. met and often exceeded the design specifications.

There appeared to be slightly elevated pore-water Rockfill used in this site consisted of crushed field rock from
pressures within the clay shale when considering nearby farmland rockpiles. The specifications required 50
hydrostatic pressures based on those measured in the mm diameter minus aggregate with 25% to 60% of the
overlying till. When the hydrostatic projection from the aggregate within 5 mm size and 2% to 8% of the aggregate
overburden to the deeper piezometers in the shale is within 0.08 mm. The resulting granular material would be
considered, there would appear to be only an average classified as a silty gravel with sand.
difference of 1.3 m of head between the two formations,
despite the tip elevations typically being greater than 5 m Upon completion of installation, the rockfill columns were
apart. This suggests that the pore-water pressures in the capped with a clay berm to achieve a slope grading of 4:1
shale are greater than hydrostatic assuming the phreatic and to limit surface water infiltration. The cap consisted of
surface exists in the overlying till. native silty clay till that was excavated during the rockfill
column construction. The berm was placed in 25 mm thick
1.3 Rockfill Column Construction lifts compaction with 8-10 passes with a sheepsfoot roller.
The capping berm also acts to increase the confining
Rockfill column construction in 2020 was completed in pressures on the rockfill columns, which in turn also
three stages, first the upper slope was offloaded by increases the normal stress to increase shear resistance
excavating ~10,000 m3 of overburden. Following unloading within the slope. Aggregate samples were collected using
of the abutment crest, a work platform was constructed at a sampling pad following ASTM standards (Committee
the mid slope for the installation of 72 - 2.1-meter diameter D04.30 2019), and periodically during backfilling of
rockfill columns. The rockfill columns were designed with a columns.
termination depth of 480 masl or 15-meter-long columns.
These columns were excavated sequentially using a bored 1.4 Instrumentation
pile rig to the target depth. The columns were then
backfilled with rockfill aggregate by loader and densified by Three shape accelerometer arrays (SAAs) were installed
vibro-lance. within the critical cross section. The critical section was
determined during the shear key design at Section D as
The vibro-lance used consisted of a hollow steel cylinder shown in Figure 2. SAA5910 was placed in the mid-slope
that applied a high frequency vibration by means of a and spans from 482 to 503 masl. It was installed
pneumatic hammer located at the tip of the lance. The approximately 31 meters upslope of rockfill column nine.
vibro-lance had a penetration length of 18 meters and was SAA5520 was positioned downslope of the rockfill
maneuvered by crane. Vibration was activated and the columns, and immediately upslope of the spillway stilling
basin. The SAA had an instrumentation length of 14.5 representing the vertical effective stress at the termination
meters and was installed through elevations 490 to 475 depth of the columns. Due to the compressive loading the
masl. It was located 28 meters downslope of rockfill column normal load is applied by flexible bladder with air and
nine. Both SAA5910 and SAA5520 were installed six monitored during shearing with pressure transducer. A
months before construction commenced with sensor plastic plate was used between the bladder and the
intervals every 0.5 meters. aggregate to prevent bladder damage, limiting the ability to
monitor vertical displacements during consolidation and
SAARC09, the SAA for rockfill column nine, was installed shearing.
immediately upslope of the column shaft for construction
convenience. It was oriented in the anticipated direction of 2.2 Clay Shale Testing
movement; towards SAA5520. The casing for SAARC09
was installed during column construction but remained Calculation of the effect of the shear key on stabilization of
uninstrumented until the final stage, which involved the the right abutment required additional understanding of the
construction of the 3 m thick capping berm. To ensure a clay shale parameters. The residual (large strain) shear
secure anchor for the instrument, this column was strength of the clay shale is relevant to this study given the
extended 2 meters below the termination depth of the large displacements previously realized along the slide
rockfill column. plane. To determine the large strain characteristics of the
clay shale, ring shear testing was carried out on
Figure 2: Cross Section reconstituted samples of the Pierre Shale. Disturbed
samples of the Pierre Shale were recovered from auger
cuttings generated as part of the rockfill column
construction. Two test specimens were reconstituted one
at 25% with groundwater sampled in September 2022 and
the other at 21% with deionized water. The reconstituted
soils were then placed within the ring shear device for
consolidation and testing.

The Bromhead Ring Shear provides the ability to determine

residual shear strength for materials by continuous one
directional shearing of soil specimens. This testing
provides peak and residual drained shear strengths and
corresponding friction angle of the clay. Bishop et al. (1971)
2.0 LABORATORY TESTING discusses the use of the ring shear and its ability to
2.1.1 Rockfill Testing in Large Direct Shear accurately measure the residual shear strength (despite its
limitations of variable stress throughout the sample)
The objective of the large direct shear apparatus was to relative to back and forth shearing of a conventional direct
determine peak and residual shear strengths, as well as shear test.
the stress strain behaviour of the rockfill aggregate. The
large direct shear apparatus selected for material testing is 2.2.1 Bromhead Ring Shear Apparatus
manufactured by Durham Geo Slope Indicator (DGSI); has The soil specimen container has a circular ring shape with
a cross-sectional shear face of 305 mm by 305 mm and a inner and outer diameters of 70 mm and 100 mm
depth of 102 mm. Typically, the equipment is used to respectfully. The maximum specimen thickness permitted
measure interface shear strength between a soil and is 5 mm, resulting in a shear area of 40.05 cm 2. The
geosynthetic. The dimensions of the equipment are dimensions were designed to minimize errors related to
conducive for larger diameter aggregate shear testing that non-uniform shearing and confining stresses within the
would otherwise not be possible in conventional direct sample. The vertical normal stress is applied using a 10:1
shear equipment. The apparatus has a fixed box on the top mechanical advantage with a maximum operating stress of
and a movable base to apply the shear stress. The 1000 kPa and a maximum applied shear stress of 500 kPa.
segments are separated by an adjustable gap which is To maintain saturation, throughout the consolidation and
controlled by the grain size of the test material. shearing phases, the test specimen is submerged in a
water bath. Because the clay shale has a relatively high
The apparatus applies a horizontal (shear) load at a salinity, the water bath was regularly circulated during
constant rate of 0.003 mm/min. The shear rate was consolidation and shearing to monitor the electrical
selected based on grain size to allow for any porewater conductivity and temperature within the water bath. It is
pressures to dissipate. A constant shear area of 0.093 m2 assumed that any changes to the total dissolved solids
is maintained throughout testing. Horizontal displacements (TDS) and electrical conductivity (EC) of the water bath are
are monitored with a Linear Voltage Displacement solely a result of salts leaching (or being adsorbed) by the
Transducer (LVDT) while the shear force is measured by clay shale specimen.
S-type load cell. A logging interval of 5 seconds was
selected for both the LVDT and load cell. The confining The top and bottom of each specimen was confined by
pressure is provided by a pre-determined vertical load porous stones with shear grooves to minimize the risk of
acting on the top plate. The vertical loads were selected as slippage occurring at the soil-porous stone interface. The
50, 100 and 400 kPa, with the 400 kPa load approximately
shearing rate could be selected as either mm/min or Mitchell and Soga (2005) determined three mechanisms of
deg/min. The shear forces were applied by variable speed cation exchange occurs in clay soils. The first observed
motor (Controls Group 2023), with a selected shear rate of higher valance cations replacing lower valence cations.
0.003 mm/min. The shear strength throughout a given test The second mechanism related the size of the dehydrated
was measured indirectly by two load cells measuring the radii of the cations, with larger radii replacing smaller radii.
applied torque of the loading ring. The vertical normal The third mechanism is dependent on the concentrations
stress was applied by weights applied to the loading arm of cations in solution, as higher concentrations will replace
while the vertical displacement was measured by high lower concentration cations.
precision LVDTs. The corresponding shear rate was
verified by periodic manual measurements of the rotated In an attempt to address the issues posed by using
angle on the apparatus frame. Figure 3 shows a profile deionized water in the ring shear water bath, a synthetic
view of the apparatus with load cells on the left and right pore fluid was produced. The geochemistry of the Pierre
and vertical load applied by the top onto the load platen Shale pore fluid was measured as part of this research. A
and lines circulating the water bath. clay shale sample from 481 masl of rockfill column 09 was
compressed using the same procedure as Hendry, et. al.
Figure 3: Modified Bromhead Ring Shear Side View (2013) and the expelled pore fluid was analyzed for salt
and metals concentrations.

The synthetic pore fluid used in the subsequent

experiments were based on the geochemical test results of
the squeezed soil found in Table 1.

Table 1: Squeezed Soil Composition

Element in Squeezed
mg/L Soil

Chloride 342
Sulfate 6830
Nitrate 14.9
Calcium 493
Current ASTM standards do not account for variation in the
pore fluid composition or consider changes to the pore fluid Magnesium 758
(or water bath) chemistry with time. Previous work by Maio Potassium 34.2
(1995); Maio et al. (2014) and Leik (2020), observed an Sodium 1490
increase in the residual/critical state shear strength of a
Sulfur 3300
clay when the concentration of salts within the pore fluid or
water bath changes with time. This implies that strict
control over the water bath concentration is critical to define Deviations from the naturally occurring (in-situ) chemistry
the actual residual shear strength of the Prairie clay shales. occur in the synthetic fluid. The differences occur due to
Moreover, preliminary testing in this study indicated that the limitation of adding sulfur to the solution. Sulfur is
the temperature of the water bath also played an important volatile in solution and its predisposition to create sulfate
role in the overall performance of the ring shear testing. As when introduced to oxygen or sulfide gas when introduced
a result, the water bath was circulated throughout a given to water was not considered safe and therefore was not
test (and each phase of the test) to monitor conductivity included in the synthetic pore fluid. It is not expected the
and temperature during testing. sulfur charge(s) to impact the clay particles. The
2.2.1.1 Synthetic Pore Fluid Composition composition and corresponding concentrations of the
synthetic fluid is provided in Table 2 below.
The pore water chemistry will directly affect a clay’s soil
fabric by altering the mineralogical structure. An increase Table 2: Synthetic Fluid Composition
of residual shear strength was observed by Leik (2020) Molar
when a Regina clay (high montmorillonite content) sample Compound Name
Mass
Mass in Solution (g/L) Composition
was subjected to a water bath containing guanidinium salts
(GuCl). Leik (2020) observed an increase in the residual Sodium Sulfate 144.666 4.43432 43%
shear strength to values similar to the initial peak strength Magnesium Sulfate 136.134 2.57472 45%
value. X-ray diffraction (XRD) and scanning electron Potassium Sulfate 160.186 0.06618 0%
microscopy (SEM) of the soil after treatment with GuCl Calcium Chloride 136.28 0.65737 7%
found that the clay particles appeared to increase in size Potassium Nitrate 94.069 0.02261 0%
and were more rotund looking based on SEM imagery of Calcium Sulfate 161.436 0.43860 4%
the treated soils. In addition to the synthetic pore fluid, sampling of
groundwater within monitoring wells installed during the
rockfill column construction was also carried out. The
purpose of the groundwater collection was two-fold; first, Figures 6 and 7 display the shear stress and shear strain
the chemistry of the groundwater within the rockfill columns curves obtained from tests conducted on rockfill aggregate
could be assessed after two years of service; and, excess at densifications of 15% and 10% respectively. Under a
samples of groundwater could be used as a more confining pressure of 400 kPa, the material densified at
representative water bath for subsequent laboratory 15% exhibited a peak strength 1.8 times higher than its
testing. At this time, the groundwater has not been post-peak strength, while the 10% densified material
analysed other than for TDS and EC. Additional testing is showed a 1.3 times greater strength. It should be noted that
ongoing. strengths near strains of 20% were influenced by edge
2.2.2 Determination of clay strength effects from the apparatus, and therefore, these increases
cannot be considered representative of the critical state
Drained internal friction angles have been determined condition of the rockfill aggregate. Due to limitations in
assuming a Mohr-Coulomb failure criterion and a zero deformation measurements, it was not possible to
cohesion. Drained conditions apply based on the determine conclusively if the aggregate had achieved
assumption that the rate is sufficiently slow to permit critical state strength by the end of the testing.
dissipation of excess porewater pressures during shearing.
Figure 6: Stress Strain Profile of Rockfill aggregate
3.0 LABORATORY TEST RESULTS densified to approx. 15%
3.1.1 Rockfill Shear Strength
Large sample direct shear testing of the rockfill column
aggregate indicate a trend of increasing friction angle with
lower confining stress as seen with Figure 4 and 5 with
supporting data in table 3 summarizing the results of six
direct shear tests under different confining stresses and
percent increase in density. The precent increase in
density was determined by measuring the settlement of the
rockfill during compaction within the shear box.

Figure 4: Rockfill Peak Friction Angle

Figure 7: Stress Strain of rockfill densified to 10%

Figure 5: Rockfill Post-Peak Friction Angle

Table 3: Rockfill Test Peak and Post Peak Strength

Normal Stress Shear Strength (kPa) Friction Angle (degrees)
Densification (%)
(kPa) Peak Post Peak Peak Post Peak Comparison of peak strength to the two different
14.5% 400 524.8 279 52.7 34.9 densifications of 10% and 15% observe a 1.5 time
14.6% 100 251.3 124.4 68.3 51.2 improvement in peak shear strength and further supports
15.2% 50 266.9 196.4 79.4 75.7 finding of (Alfaro et al., 2009) where higher relative
10.7% 400 337 257 40.1 32.7 densities lead to a greater shearing resistance.
9.4% 100 119.1 102.7 50.0 45.8
9.8% 50 69.4 60.4 54.2 50.4 Rockfill tests conducted at lower normal stresses was
found to have an increased friction angle. Vertical
displacements during shearing were not monitored due to The results obtained from ring shear testing on
the presence of the internal air bladder and the separating reconstituted Pierre Shale further support the findings of
plastic sheet providing confining pressure. Chattapoadhyay (1972), Moore (1991) and Leik (2020). As
is shown in Figure 10 below, an increase in salt
3.1.2 Ring Shear Testing of Clay Shale concentration, specifically cation concentrations, increases
3.1.2.1 Consolidation the shear strength of the soil relative to tests carried out
with deionized water. Comparative tests were conducted
A high susceptibility to swelling was observed during using deionized water on reconstituted clays and using
consolidation of the clay shale when reconstituted with groundwater sampled in September 2022 (RS Sept 2022).
deionized water (RS DI). The swelling of a representative The peak shear strength of RS Sept. 2022 was almost
sample is shown in Figure 8. double of RS DI with the only differences being the
concentration of salts present in the water and the final
During consolidation, temperature fluctuations were also water content of each sample. A final water content of 22%
observed to affect swelling of the samples, however, the for RS Sept 2022 and 29% for RS DI were observed by
swelling is related to expansion of the sample and appear drying each specimen after shearing was ceased. A higher
to be recoverable. In a test reconstituted with sampled water content in the deionized test is likely attributed to the
groundwater, the swell associated with ion exchange was swell of the sample indicating an affinity to water. This
found to be negligible. However, minor changes to the affinity to water results softening, attributing to the lower
laboratory temperature resulted in fluctuations to the shear strength.
sample and appear as the saw-toothed noise observed in
Figure 9. Upon reaching the residual state during the test using
sampled groundwater RS Sept 2022, the water bath was
flushed with deionized water to simulate a loss of salt
concentration within the groundwater (spring snow melt).
Once the shear strength had stabilized, a synthetic pore
fluid was introduced to the water bath. The introduction of
the synthetic pore fluid resulted in a 60% improvement in
the measured residual shear strength. The stages of water
bath flushing are shown in Figure 10 and are only for the
test reconstituted using sampled groundwater. A summary
of the test results is provided in Table 3 below.

Figure 10: Ring Shear tests on Reconstituted clay with

different pore fluid

Figure 8: Consolidation stage of reconstituted specimen

with deionized water at 50 kPa

Table 3: Clay Shale Ring Shear Test Results

Peak Peak Residual Residual
Avg. Salt
Reconstituted Pore Shear Friction Shear Friction
Content
Fluid Strength Angle Strength Angle
(g/L)
(kPa) (degrees) (kPa) (degrees)
Deionized 0 59.1 11 37.2 7
Column 09

1.08 119.5 21.7 61.8 11.6

Sept 2022

DI Flush 0 N/A N/A 56.4 10.6

Synthetic
8.19 N/A N/A 66.3 12.4
Flush

4.0 INSTRUMENTATION DATA

Figure 9: Consolidation stage of reconstituted specimen
with sampled groundwater at 40 kPa Instrumentation data from the SAAs installed through the
shear zone between EL 486 to 490 masl are shown in
3.1.2.2 Clay Shearing Figure 12. These plots indicate that movements appear to
be seasonally driven, with an increase in shear rate
occurring over the winter months and decreasing in the the SAA indicates a clear temperature gradient with depth.
summer months. Evaluation of the upslope groundwater As expected, the greatest fluctuation in temperatures occur
table monitored by vibrating wire piezometer referred to near the ground surface 489.6 masl and reduce with depth.
P6013 indicates that there is an increase in loading from The ring shear testing demonstrated the impact of
freeze/thaw cycles and increased groundwater during temperature changes on the consolidation of the
spring and summer seasons. The fluctuations are reconstituted clay shale, and it is possible that these
consistent with seasonal variations in the area, which are temperature affects are also playing a role in the measured
typically characterised by cold winters with accumulation of response of the SAA. Though the temperature fluctuations
snow up to 2 m in depth, followed by spring snow melt and to the shear zone elevation are relatively minor (~4 to 7 C),
moderately dry summers. it is possible that this may be sufficient to result in volume
change of the clay shale and the measured cyclic
At this time, it is not clear whether the movements displacements.
observed in the SAAs upslope and downslope of Rockfill
Column 9 are influenced by the changes in groundwater Figure 12 shows that the displacement in SA5520
level, groundwater chemistry or temperature fluctuations. increases roughly around the maximum measured
Based on the data presented in Figure 11, the SAA temperature and then decreases thereafter. Further
installed downslope of the Rockfill Columns (SAA5520) is evidence that the measured displacements are related to
moving prior to any increase in groundwater levels. It is temperature is given by the recovery of the measured
possible that the location of the SAA is more permeable movements. If the displacements were a result of actual
and is responding to changes in groundwater sooner than ground deformation related to elevated pore-water
the piezometer is. It is also possible, that this SAA may be pressures, the displacement would be cumulative and not
influenced by the level of the reservoir. It is not clear at this recovered. The indication of recovered displacements is
time what is driving the cyclic changes in the data. like the consolidation data.

The SAA installed upslope (SAA5910) has not shown any

appreciable displacement outside of one event between
May and July of 2022. It is important to note, that the period
of spring and early summer of 2022 was extremely wet with
very high rates of precipitation occurring throughout the
season. Most of the precipitation events however were of
relatively short duration and of high intensity. It is also
interesting to note that there were no obvious changes in
displacement measured in the downslope SAA at this time.

When the SAA installed into Column 9 (SARC09) is

examined, there was considerable increase in the
Figure 12: SA5520 Shear Zone 489.6 - 486.6 masl
displacement during the spring of 2021 and a muted
displacement in the spring of 2022. It is common for rockfill Temperature Influenced Deformations
columns to undergo ongoing movements for a period of 2-
5.0 CONCLUSION
3 years post installation while the shear strength of the
granular material is being mobilized. Based on the data
The paper presents experimental findings demonstrating
shown below, it appears that the rate of displacement
the relationship between densification percentages and
within the columns is still decreasing with time.
stress-strain behaviour of tested rockfill. The results show
that higher densification leads to stress-strain profiles
resembling that of an overconsolidated soil strength profile,
while lower densification results in resembling a normally
consolidated profile. These observations indicate the direct
advantages of higher densification in limiting strains and
achieving greater frictional resistance. The results of rockfill
testing align with findings from Razaq (2007), indicating the
potential for elevated frictional resistance under reduced
confining stresses.

The results of ring shear testing presented supports the

Figure 11: SAA Displacements in Shear Zone 490 -486.5 influence of pore water chemistry on the strength
masl properties of clay. Additionally, the paper discusses the
notable effect of temperature variation on this clay’s
It should be noted in SA5520 appears to be influenced by swelling potential. Specifically, temperature changes
seasonal effects, as is shown in Figure 11. It is clear the exceeding 4 degrees Celsius were found to cause
ground begins to displace when ground temperatures drop significant volume changes.
and are mostly recovered when temperature increase.
Preliminary analysis of the temperature data measured in
Furthermore, the paper discusses the use of Maio, C Di, G Scaringi, and R. Vassallo. 2014. "Residual
instrumentation for monitoring deformations of a rockfill strength and creep behaviour on the slop surface
column shear key remediation. The findings establish a of specimens of a landslide in marine orgin clay
strong correlation between deformations and seasonal shales: influence fof pore fluid composition."
driven temperature changes at this site. This correlation Landslides 657-667.
further supports the influence of temperature-driven heave.
In conclusion, the findings presented in this paper provide Maio, C. DI. 1995. "Exposure of bentonite to salt solution:
valuable technical insights into the design of rockfill osmotic and mechanical effects." Geotechnique
columns considering seasonal fluctuations. 695-707.
5.1 Acknowledgements Mitchell, J. K., and K Soga. 2005. Fundamentals of Soil
Behavior. Hobolen, N.J: John Wiley & Sons. X.
The authors are grateful for the funding from
Saskatchewan Water Security Agency for this research. Moore, R. 1991. "The chemical and mineralogical controls
This paper would not be possible without the sampled upon the residual strength of pure and natural
aggregate and clay shale during construction, and access clays." Geotechnique (G) 35-47.
to the site and instrumentation data post construction.
Prairie Farm Rehabilitation Adminstration . 1967. Report
on construction of Theodore dam project
Whitesand river. Document of Record, Regina:
6.0 REFERENCES Saskatchewan Regional Division.
Bishop, A. W., Green, G. E., Garga, V. K., Andresen, A., & Razaq, Wisam F. Abdul. 2007. Evaluation of riverbank
Brown, J. D. (1971). A new ring shear apparatus stabilization using rockfill and soil-cement
and its application to the measurement of residual columns. PhD, Winnipeg: University of Manitoba.
strength. Geotechnique, 21(4), 273-328.
Tang, Yuk Lun Yolanda. 2020. Experimental
Alfaro, M. C., Blatz, J. A., Abdulrazaq, W. F., & Chang- characterization of the smipermeable membrane
Seok, K. (2009). Evaluating shear mobilization in behaviour of Pierre Shale at Southern
rockfill columns used for riverbank stabilization.
Saskatchewan. MSc Thesis, Saskatoon:
Canadian Geotechnical Journal, 976-986.
University of Saskatchewan.
Chattopadhyay, Pradyot Kumar. 1972. Residual Shear Thiessen, Kendall. 2011. Stabilization of Natural Clay
Strength of some pure clay minerals. PhD Riverbanks with Rockfill Columns: A full-Scale
Thesis, Edmonton: University of Alberta. Field Test and Numerical Vertification. Thesis,
Committee D04.30. 2019. Practice for Sampling Winnipeg: University of Manitoba.
Aggregates. Book of Standards Volume 04.03, Trek Geotechnical. 2019. "Stone Column Slope
Pennsylvania: ASTM International. Remediation of the Right Abutment at Theodore
Controls Group. 2023. "Ring Shear Testing Machine - Dam." Regina.
TORSHEAR EmS." Controls. Accessed 06 01,
2023. https://www.controls-group.com/usa/direct-
residual-shear/ring-shear-machine-torshear-
ems.php.

D18 Committee. 2021. ASTM D 5321 - Standard Test

Method for Determining the Coefficient of Soil
and Geosynthetic or Geosynthetic and
Geosynthetic Friction by the Direct Shear
Method. ASTM International.

Hendry, Jim, S L Barbour, K Novakowski, and L I

Wassenaar. 2013. "Paleohydrogeology of the
sediments of the Williston Basin using stable
isotopes of water." Water Resources Research
4580-4592.

Leik, Adam. 2020. Laboratory scale evaluation of the

feasibility of using guanidinium salts as a means
of insitu soil strengthening for weak clays. M.Sc
Thesis, Saskatoon: University of Saskatchewan.
CASE STUDY OF AN OIL AND GAS DEVELOPMENT IN
LANDSLIDE TERRAIN
Catlan J Dallaire Jasyn Henry, & Allen Kelly
Clifton Engineering Group Inc., Regina, SK, Canada
Md Abu Zahid, Kayne Ulmer, & Geoff Witwicki
Thyssen Mining Construction of Canada Ltd., RM of Sherwood, SK, Canada

ABSTRACT
Terrain analysis is critical in the development and protection of oil and gas assets. Regional geohazards, such as
landslides, are identified, risk assessments completed, and planning revisions made prior to construction. This case study
is located along a major river valley in southern Alberta, among ancient landslides with rates of movement as low as
mm/year to catastrophically high. Analysis of terrain showed the uplands area was a dead ice moraine. Nearing the valley,
glacial drift was eroded by glacial meltwater. The river valley terrain was complex and had many historical retrogressive
landslides. A large headscarp was observed east of the site where the valley wall met the surrounding uplands. Graben
features trapped runoff causing ponding, locally increasing groundwater levels and accelerating landslide movement.
Geological maps indicated the ground surface was near the boundary of the Paskapoo Formation, the Scollard Formation
and the Edmonton Group, which included the Battle Creek Formation and the Whitemud Formation. Stratigraphy of the
valley walls in this area consisted of landslide debris, with an interbedded and eroded mixture of overburden soil and
bedrock. In 2014, an existing well began to show irregularities in operation. Downhole measurements indicated the well
casing had deformed approximately 35 m below ground surface. The site was partially decommissioned in early 2015 by
removing surface and shallow subsurface structures. A geotechnical investigation was completed to assist in definition of
the subsurface materials and for design parameters to be used in a freeze shaft mining technique for non-routine
abandonment.

Ce site d'étude de cas est situé le long d'une grande vallée de la rivière dans le sud de l'Alberta, parmi les anciens
glissements de terrain avec des vitesses de déplacement aussi bas que mm / an à un niveau catastrophique. L'analyse
du terrain a montré que la superficie des hautes terres près du site du puits était une moraine de glace morts. Près de la
vallée, la dérive glaciaire a été érodée par l'eau de fonte glaciaire. La vallée de la rivière a connu de nombreux glissements
de terrain historiques rétrogradés. Un grand voile a été observé à l'est du site où la paroi de la vallée a rencontré les hautes
terres environnantes. Graben comporte un ruissellement piégé entraînant des étés, des niveaux d'eau souterraine
localisés et des mouvements de glissements de terrain accélérés. En 2014, le puits a commencé à montrer des
irrégularités en exploitation. Les mesures du fond du trou indiquent que le boîtier du puits s'est déformé à environ 35 m
sous la surface du sol. Le site a été partiellement déclassé au début de 2015 en supprimant les structures superficielles
superficielles et peu profondes. Une enquête géotechnique a été complétée pour aider à la définition des déformations de
la sous-surface, des matériaux et des paramètres de conception à utiliser dans une technique d'exploitation de l'arbre à
congélation pour l'abandon non routinier. Cet article décrira la caractérisation du site, l'étude géotechnique et l'abandon
de bien-être unique qui a été utilisé sur ce site.[CD1]

1 BACKGROUND / INTRODUCTION partially decommissioned in early February 2015 by

Wells like the one described above are not uncommon in removing surface structures, some subsurface structures,
the prairies. River valleys carved into the underlying clay and the gas collection line.
rich bedrock present challenges for pipelines, wells, and
other facilities built on their walls. The damage caused by A geotechnical investigation was completed to assist in
landslides can be catastrophic to the infrastructure and the definition of the subsurface deformations, materials, and
environment. This paper describes the case study of one for design parameters to be used in a freeze shaft mining
such well and the unique methods employed to regain technique for non-routine abandonment. Prior to mobilizing
control of it for conventional decommissioning. It also aims to site, terrain analysis was undertaken to characterize the
to point out how preliminary terrain analysis for siting and local landforms and stratigraphy.
pre-design could earn substantial savings in the long term.
TERRAIN ANALYSIS
The well in this case study was a natural gas well situated It is a common practice in geotechnical engineering to
on a large landslide. In late 2014, the well began to show identify key surficial landforms with the use of air photo
irregularities in operation. Downhole measurements interpretations (API) based on aerial stereo photography
indicated the well casing had deformed at a depth of (Sauer, 1980). Now, more advanced technologies
approximately 35 m below ground surface. On that basis, including satellite images, radar images, LiDAR, etc. are
the well casing was confirmed to be sheared and the loss being incorporated. Terrain Analysis deepens the site
of well integrity was reported to the regulators. The site was understanding by pairing API with other mapping sources
 Site Features along the Red Deer River

to provide a more in-depth evaluation of the physical Air Photo Interpretation (API)
geography and geology of a region (Schroder, 2014). Analysis of the terrain near the well site using API showed
Additional information often includes details on vegetation, that the uplands area near the well site was a dead ice
soil, land use and/or resources, hydrology, geology and moraine. Nearing the river valley, the drift and till became
landforms. A thorough analysis can identify key thin as it was eroded by glacial meltwater flows. It was
geotechnical risks and greatly reduce potential long term evident that the river valley had many historical landslides
issues with respect to drainage, slope movement, (see figure above). A large headscarp was observed east
subsidence, soft soils, and any other geotechnical “red- of the site where the valley wall rises to meet the
flags.” surrounding uplands. Multiple retrogressive landslides
were identified in the area. Graben features trapped local
4 PHYSICAL ENVIRONMENT runoff causing ponding and localized wetlands that could
The site was located in a major river valley in southern cause local increase in groundwater levels and accelerate
Alberta which is approximately 1.6 km wide with a south landslide movement.
flowing river at that location. The elevation of the river was
approximately 740 m asl, the top of valley elevation was Stratigraphy
approximately 850 m asl to 860 m asl, and the well was The site was located within a massive retrogressive
approximately 783 m asl. landslide with signs of historical movement. Stratigraphy at
the site varied significantly but generally consisted of
Geological maps of the site indicated that the ground landslide debris, with an interbedded and eroded mixture
surface was near the boundary of the Paskapoo Formation, of overburden soil and bedrock. The geotechnical
the Scollard Formation and the underlying Edmonton investigation determined the site stratigraphy was
Group, which included the Battle Creek Formation and the complex, but generally consisted of fill, and non-lithified
Whitemud Formation (Dawson, et al., 1994). Generally, bedrock consisting of clay shale, overlying interbedded
these units dipped west. Stratigraphy of the valley walls in mudstones and sandstones to the depth of exploration.
this area
Bedrock was defined as all sediments that were pre-glacial. Sandstone was fine grained and interbedded to laminated
Clay shale is a non-lithified, cretaceous, marine clay with clay and silt. Generally, it was homogeneous with coal
deposit that is often associated with slope instability in inclusions, but some zones exhibited bedding and
Western Canada river valleys. While the surficial materials laminations, including cross-bedding, flaser laminations
were highly variable, they were generally interpreted as and well-cemented zones.
eroded and intermixed remnants of the Paskapoo
Formation and upper Edmonton Group formations. 5 MONITORING
Stratigraphy became consistent below 36.6 m to the depth Piezometers
of exploration. Vibrating wire piezometers were installed at depths of
16.8 m bgs, 27.1 m bgs, and 48.9 m bgs. The piezometers
The borehole logs from this were grouted in place and measurements were collected
site identified zones of every 2 hours. Standpipe piezometers were installed to
slickensides and monitor groundwater levels and to allow determination of
disturbance to a depth of hydraulic conductivity in what was assumed to be the most
about 36 m. It is important to permeable strata of the overburden and bedrock. Overall,
note that these features are the data showed there was a net downward gradient.
not always associated with
landslide movement. 5.1.1 Slug Test
Slickensides can be found in A slug test was performed on the well installed in BH103
jointed strata, caused by determine in situ hydraulic conductivity. The slug test took
mechanical disturbance (drill approximately 6 hours and was terminated when 80%
spinning), and also by recovery was achieved. The slug test was analyzed with
natural disturbance. At sites Aquifer Test software using the Hvorslev analysis method.
as heavily disturbed as this, The hydraulic conductivity of the sandstone formation,
it is not uncommon to see approximately 44 m bgs (739 m asl), was calculated to be
slickensides. 2.2x10-8m/s. The purpose of this test was to test the
hydraulic conductivity of the most permeable member of
The strata between 32 m the lithified bedrock for freeze shaft design.
and 36 m bgs was
considered to be the most Slope Inclinometers
disturbed based on the core Slope inclinometer (SI) casing was installed to the depth of
samples and in correlation investigation (66.7 m bgs). The data showed a slip surface
with the known gas well at 32.2 m bgs (751.5 m asl). During the 60 day monitoring
deflection point. The core period in early spring 2015, the landslide progressed
recovered from this zone approximately 12 mm with an increased rate of movement
was highly disturbed upon from approximately 0.1 mm/day to 0.25 mm/day between
removal from the core barrel the March and April monitoring events. It is typical to see
due to the swelling nature of increased movement in the spring season due to increased
marine clays upon their infiltration from snow melt and increased groundwater
exposure to fresh water levels.
used in drilling. Shale below
a depth of 36 m transitioned 6 PROPOSED CONSTRUCTION SOLUTION
into lithified bedrock. In our The geotechnical data presented above was collected with
experience, it is typical to the purpose of facilitating a non-conventional well
observe landslide movement abandonment technique; excavation of a frozen shaft to the
at or near the well failure point. The following sections will discuss the
oxidized/unoxidized shale construction activities associated with the freeze shaft
contact(s) and in this case installation and well decommissioning.
was shown to be true.
Construction Methodology
The condition of bedrock As a result of successful implementation of freeze shaft
was fairly consistent below a technique at Penn West Petroleum’s Swan Hills Oil Well
depth of approximately 36 Casing Repair Project in 2012, a similar method was
Generalized site m bgs. The bedrock employed here. The contract was awarded in early 2015
stratigraphy for project area consisted of interbedded and began with engineering efforts related to the setup and
(from Figure 24.3 by mudstone and sandstone, sinking of a small diameter, steel lined shaft through frozen
Dawson et al. (1994). fining-upwards in a overburden down to the projected shear zone.
lithological sequence.
Generally, the mudstone had varying silt, clay and very In light of the nature of the water levels observed in the
fine-grained sand contents. It was unoxidized, light gray to overburden, a small freeze project was the best suited
olive gray, laminated to bedded, and had varying amounts application. The overall plan consisted of ground freezing
of organic inclusions. It was generally competent. to just below the anticipated shear zone depth, followed by
sequential excavation and steel lining of the frozen was to provide a means of mucking the shaft during the
overburden. Construction was planned to utilize a pre-sink. The pre-sink depth was designed to ensure the
combination of a shaft mucker and mucker tower for the Galloway work stage could be safely installed and remain
pre-sink, and a shaft mucker, hoisting plant and headframe in the shaft. During the pre-sink, the vertical shaft mucking
arrangement for the main sink. The frozen ground around unit was used to load a 2.0 m³ sinking bucket. Once loaded,
the well head was to be excavated and steel lined to a a mobile crane was used to hoist the bucket to surface to
depth of approximately 50 m. Upon successful excavation dump the muck in the designated waste storage area.
to depth, the sheared well casing would be replaced and
advanced back to surface elevation for traditional Shaft Excavation – Pre-Sink
decommissioning. As the shaft was excavated, corrugated steel liner plate
and ring beams were installed to in order to provide ground
Freeze Hole Drilling support. As an additional means of ground support, the
The design consisted of freeze holes to be drilled in a annulus between the excavation and the corrugated steel
circular manner centered around the well casing. Additional liner plate was backfilled with a lean concrete mix.
holes were to be located slightly outside the freeze hole
perimeter and used as temperature monitoring holes.

Each freeze hole was drilled to a final approximate depth

of 60 m and was cased with standard J55 steel casing.
Vertical drilling accuracy was very important, as even a
slight deviation over great lengths, can cause the freeze
holes to drift into the excavated shaft area. Upon
completion of the cased freeze holes, a wire line survey
was carried out to ensure their accuracy. The steel casings
were then capped and pressure tested to 1.0 MPa, ready
to be connected to the freeze headers, brine distribution
system and mobile freeze plant.

Shaft during Pre-Sink

Pre-sink operations continued to a depth of approximately

20 m. As the pre-sink advanced, the gas well casing was
cut and removed in 2.0 m sections.

Clam Tower Structure used during Pre-Sink Concurrent with the pre-sink, a Galloway work stage was
assembled on surface. Once the pre-sink was completed,
Collar Construction the Clam Tower structure was removed and the Galloway
Before ground freezing could begin, the shaft collar area work stage was placed on shaft bottom. A Portable
was excavated to a depth of 2.5 m, formed, and poured Headframe structure was erected over the shaft and a hoist
with 30MPa concrete to establish both a freeze cellar and and two (2) stage winches were installed and
a foundation for a Clam Tower and Portable Headframe commissioned accordingly. Guide ropes from the two (2)
structure. The freeze pipes were then outfitted with stage winches were used to suspend the Galloway in the
headers and connected to a brine distribution system. A shaft and doubled down as a means of conveyance
refrigeration sub-contractor was utilized to assist with the guidance. A sinking bucket was connected to the main
installation, commissioning, maintenance, and continual hoist as the primary means of mucking and egress from the
monitoring of the freeze system during ground freezing and shaft. The Galloway work stage provided a second means
shaft sinking operations. of egress and a small emergency hoist was installed as a
tertiary means of egress at the client’s request. A winch
Once the freeze system was turned on, a Clam Tower was installed on the collar steel and used to suspend the
structure was installed on its foundation near the edge of Cryderman mucker in the shaft during the main sink
the shaft. The purpose of the Clam Tower, in conjunction mucking operations.
with a Cryderman Shaft Mucker, crane and sinking bucket,
 Shaft Excavation – Main Sink 8 REFERENCES
The main sink continued to the location of the well casing Dawson, F.M., Evans, C.G., Marsh, & R., Richardson. R.
shear zone. The shear zone was reached at a depth of 1994. Uppermost Cretaceous and Tertiary Strata of the
approximately 35 m. After client inspection of the well Western Canada Sedimentary Basin, in Geological
casing at the shear zone a new well casing was installed Atlas of the Western Canada Sedimentary Basin, G.D.
and backfilled, utilizing excavated material, to surface. Mossop and I. Shetsen (comp.), URL
https://ags.aer.ca/atlas-the-western-canada-
Post Construction sedimentary-basin/chapter-24-uppermost-cretaceous-
At the conclusion of the well recovery project, the Alberta and-tertiary-strata (last accessed 15 May 2023).
Energy Regulators were presented with a report outlining Sauer, E.K. 1980. Air-Photo Interpretation of Glacial
the project from the design and planning phases to the Stratigraphy and the Engineering Assessment of
onsite execution and abandonment phases. The intent of Terrain in Southern Saskatchewan. Rock and
this report was to help establish a precedent in Alberta, Landform Jigsaw Puzzles, Photogrammetria, 35:207-
amongst all oil and gas companies, for both the recovery 232.
and non-routine abandonment of damaged or deformed Shroder, J.F. 2014. Rock and Landform Jigsaw Puzzles,
well casings utilizing the Thyssen Mining ground freezing Natural Resources in Afghanistan, Elsevier, 2014:16-
and shaft excavation methodology. 52.

7 DISCUSSION / CONCLUSIONS
The terrain analysis conducted as part of this study clearly
demonstrated visual evidence of historic slope instabilities
(headscarps, eroded stratigraphy) and the increased risk
of on-going movement due to inhibited ground and surface
water movement (graben features, wetlands). Using API,
these features were identified and applying the principles
of terrain analysis augmented to increase the
understanding of the site geohazards and potential
instability mechanisms.

The subsequent geotechnical drilling program was

targeted to confirm the terrain analysis assumptions and
obtain the necessary soil properties for designing a
solution. Ultimately, it was the geotechnical investigation
that correlated the infrastructure failure to the physical
environment and provided Thyssen Mining with the key
properties required to design an innovative
decommissioning method.

These geohazards at the site and in the adjacent river

valley should have been identified early in the planning
process. This would have allowed for a thorough risk
assessment of the construction and operation hazards
which may have avoided the need for untimely well
decommissioning. Terrain analysis would have provided a
cost-effective early warning of these geotechnical “red
flags”.

While the chosen construction solution proved effective in

decommissioning the damaged gas well, the engineering
intensive and costly solution may have been avoided if
terrain analysis has been incorporated in the original well
siting and pipeline routing. Several other wells continue to
operate in the Red Deer River Valley and without terrain
analysis to guide future infrastructure planning, ongoing
landslide activity may implore this innovative solution
again.
Saskatoon River Valley Geohazards – An
Update and Modelling

Vanessa Heilman1 & David Elwood2

1Geotechnical Engineering Specialist, City of Saskatoon, Canada
2Associate Professor, University of Saskatchewan, Canada

ABSTRACT
The City of Saskatoon has had a long, but relatively unknown history of geohazards throughout the city limits owing to
highly variable surficial geology and a moderately incised river valley. Landsliding activity ranges from large, retrogressive
slides, translational to conventional circular slides as well as instabilities related to internal erosion. The complex glacial
history of the soils in Saskatoon plays a key role in the type and nature of the instabilities. The presence of elevated
groundwater levels within the uppermost surficial stratified drift and the relative location of the till layer and the river level
are typically controlling factors. Other factors like possible artesian pressures within the lowermost Empress Sand and
river levels that are reflective of the flow through the Gardiner Dam also play a lesser role.
This paper will provide a brief synopsis of the surficial geology of the City of Saskatoon as well as a history of landslide
activity on the east bank of the South Saskatchewan River. The landslide summary is an update to a previous publication
that documented and described previous slide activity within the Saskatoon city limits. Lastly, a brief discussion on the
current methods of slide remediation is discussed.
RÉSUMÉ
La ville de Saskatoon a une longue histoire, mais relativement inconnue, de risques géologiques dans toutes les limites
de la ville en raison de la géologie superficielle très variable et d'une vallée fluviale modérément incisée. L'activité des
glissements de terrain va des grands glissements régressifs, des glissements translationnels aux glissements circulaires
conventionnels ainsi que des instabilités liées à l'érosion interne. L'histoire glaciaire complexe des sols de Saskatoon
joue un rôle clé dans le type et la nature des instabilités. La présence de niveaux élevés d'eau souterraine dans la dérive
stratifiée superficielle la plus élevée et l'emplacement relatif de la couche de till et du niveau de la rivière sont
généralement des facteurs déterminants. L'érosion progressive de la rive est par la rivière Saskatchewan Nord entraîne
une pente continue des parois de la vallée, ce qui entraîne une instabilité. D'autres facteurs tels que les pressions
artésiennes possibles dans le sable Empress le plus bas et les niveaux de la rivière qui reflètent le débit à travers le
barrage Gardiner jouent également un rôle moindre.
Cet article fournira un bref résumé de la géologie superficielle de la ville de Saskatoon ainsi qu'un historique de l'activité
des glissements de terrain sur la rive est de la rivière Saskatchewan Sud. Le résumé des glissements de terrain est une
mise à jour d'une publication précédente qui documentait et décrivait les activités de glissement antérieures dans les
limites de la ville de Saskatoon. Enfin, une brève discussion sur les méthodes actuelles de remédiation des glissements
est abordée.
Abramson et al. (2002) indicated that there are many
factors that affect selection of the appropriate method for
1 INTRODUCTION slope stabilization. These factors include the risk
The purpose of this paper is to provide a historical review associated with the selected construction methods; cost
of landslide hazards and slope failures within the City of and time required for installation. The most effective
Saskatoon, SK. The evaluation considers the current method of stabilization is one that provides improvement
topography of the site; the in-situ geologic conditions; within the shortest time frame and at the least cost.
previous stabilization methods employed; location of any Other factors that can affect selection of the stabilization
failure planes that may form the current failure surface; method are:
groundwater conditions; surface hydrology; as well as any
other external factors that might contribute to slope 1. The slope type (whether it is natural or man-
movements and the works necessary for stabilization and made) and associated risk of failure;
remediation. 2. Soil type; (cohesive versus cohesionless soils);
3. Topographic relief;
General recommendations regarding construction 4. Groundwater level;
methodology and best practices will be discussed to 5. The shear strength of the ground in situ;
provide designers information when considering 6. Allowable movements of the slope during and/or
implementation of different stabilization methods for after construction;
landslide and top of bank displacements.
 7. Movements required to mobilize the strength of index is rather low due to the sand content and is typically
the reinforcement; around 6 to 7% with the natural water content and liquid
8. Resilience of the stabilization relative to the limit being 10 to 12%; and 19% respectively. This range of
project lifetime; water content indicates that the natural water content is at
9. Availability of the materials or equipment. or slightly below the plastic limit implying a slightly
overconsolidated state and that it is nearly saturated.
In addition to the technical constraints, site access, Shear strength testing of the Battleford till suggest an
environmental requirements, aesthetics, scheduling, and
internal angle of friction of 30 to 32 and an effective
other factors such as politics and local tradition should be
cohesion of 20 to 30 kPa is representative. MacDonald and
considered.
Sauer (1970) illustrate one triaxial compression test (CIUC)
2 SURFICIAL GEOLOGY AND GEOTECHNICAL on the Battleford till. The test data shows that the Battleford
CONSIDERATIONS OF SASKATOON AND AREA till is typically strain hardening with contractive volume
change as indicated by positive pore-water pressures. This
The surficial geology of the Saskatoon Area is remarkably suggests that the material is relatively close to normally
complex owing to at least four previous glaciations consolidated or lightly overconsolidated despite being a
(Christiansen, 1970). The drift in the area is variable in basal till. Christiansen and Sauer (1998) report that the
thickness and consists of (from bottom up) Empress Group transition from contractile behaviour to dilative occurs at
sands and gravels; the Sutherland Group; and the around an overconsolidation ratio (OCR) of 4. Residual
Saskatoon Group. Within the Saskatoon Group, the group (large strain) shear strengths are not reported and
is divided into the Floral Formation and Battleford therefore are not discussed herein.
Formation tills and the Surficial Stratified Drift (SSD).
During the retreat of the Wisconsin Glacier approximately The SSD is generally present in the CoS above the
12,000 years ago, Lake Saskatoon formed in the Battleford till south of the 25th Street bridge. Moving
depressed area where Saskatoon now exists. The glacial upstream of the South Saskatchewan River, the thickness
lake was bounded in the east by Strawberry Hills, the north of the SSD increases as the contact with the underlying till
by the retreating glacier, and west by the gradual increase approaches the river level near the south end of the city
in topography. Glacial Lake Saskatoon was thought to Haug et al. (1977), Chursinoff (1980) and Yoshida and
have drained to the northeast through the Watrous Krahn (1984). The physical characteristics of the SSD is
Spillway, which forms the present-day South best described by Christiansen and Sauer (1998). They
Saskatchewan River valley. It is not known whether water report that the peak shear strength of the SSD clays are
from Lake Saskatchewan also flowed south towards the typically around 26 to 30 and an effective cohesion of
Qu’Appelle Spillway. It was at the height of glacial Lake approximately 5 kPa. From this perspective, there is little
Saskatoon that the basal silts and clays that comprise the difference between the peak effective stress parameters of
lower reaches of the SSD were deposited. the SSD and the underlying Battleford Till. One of the key
differences is the effect of the SSD during fluctuations in
With respect to landslide hazards in the City of Saskatoon pore-water pressure. Increased pore-water pressure on
(CoS), only the Battleford Till and the SSD are of concern the SSD results in extreme reductions in effective stress
and therefore only these two formations will be discussed and therefore shear resistance. Yoshida and Krahn (1984)
in detail. If additional information is required for the other report that the residual friction angle from laboratory and
formations, please refer to Christiansen (1970); Whitaker back analysis of the Beaver Creek landslide is
and Christiansen (1972); Christiansen and Sauer (1984,
approximately 6 to 8. Christiansen and Sauer (1998)
1998) and Eckel et al. (2002). Christiansen (1970);
recommend that a design value for slope stability should
MacDonald and Sauer (1970a and b); Haug et al. (1977)
be taken as 7. Hamilton and Tao (1977) report that the
and Yoshida and Krahn (1984) which provide an excellent
grain size of the SSD consists of fine sands and silts with
outline of the geology and engineering characteristics of
roughly 30 to 45% sand; 28 to 40% silt and 25 to 30% clay
the various formations throughout the city and surrounding
percentages in the upper strata. Below the “cohesionless”
area. Fredlund (1970) statistically summarizes the soil
formation, the clayey SSD typically is characterised by low
characteristics for a wide range of studies carried out
sand contents (0 to 12%); 23 to 45% silt and 45 to 78% clay
through the CoS.
content. The liquid limits and plasticity indices range from
The Battleford till is typically between 0 to 45 m in thickness approximately 44 to 84% (typically greater than 60%) and
but is typically only around 3 m thick in the downtown 45% respectively. The natural water contents are generally
Saskatoon area. It is generally soft, friable and massive less than 20% in the upper sand and between 35 and 40%
(unoxidized) with frequent boulders. Typically, the boulders in the clayey SSD. The Atterberg limits and natural water
are present as lag near the contact with the underlying contents suggest that the SSD is generally normally to
Floral Formation. The boulders are reportedly soled and lightly overconsolidated and saturated. The presence of the
striated from glacial action. The Battleford till is generally sand layer near the surface is conducive to infiltration and
coarse grained (relative to the underlying tills) and is deeper percolation to the surface of the clay SSD. This
typically comprised of 50% sand and 50% fines. phenomenon was well reported by Fredlund (1970) and
Christiansen (1970) and MacDonald and Sauer (1970b) Hamilton and Tao (1977) who illustrated that the
summarize the gradations from several sites as being groundwater level in the SSD is strongly dictated by the
approximately 50% sand, 30% silt and 20% clay contents. degree of precipitation (or irrigation) at the ground surface.
Despite having a relatively high clay content, the plasticity Hamilton and Tao (1977) report that the annual
groundwater fluctuations in the SSD can vary from 1.5 to 2 slope failures. The presence of increased groundwater
m depending on the availability of water for infiltration. within the SSD results in reduced effective stress within the
SSD and the starting of slope movements. Once these
3 LANDSLIDE HISTORY IN THE CITY OF movements start, there is a progressive softening of the
SASKATOON SSD soils at the till interface, resulting in a thin zone of
This section provides a brief history on the landslide activity residual soils.
in the CoS and the surrounding area. Details of historical
landslides in the CoS have been well documented by
Clifton et al. (1981). In this paper, an inventory of the nature
of the landslides and some of the basic remedial measures
taken are presented.

Photo: Slope failure on Saskatchewan Crescent, close to

intersection with 16th Street, Saskatoon, 2016
Historical failures of this nature consist of the Cosmopolitan
Park; 17th and 18th Street slides; McRaney Slide; 15th
Street Slide; the Broadway Bridge slide; Eastlake
Crescent; Cherry Lane; Long Hill; Nutana Collegiate;
Queen’s House of Retreat; Diefenbaker Park and Beaver
Photo: Extract from Saskatoon StarPhoenix, May 3, 2018 Creek.
Additional information on landslides in the CoS and the
surrounding area are provided by Christiansen and
Mendeley (1970); Haug et al. (1977); Yoshida and Krahn
(1984); Christiansen and Sauer (1984, 1998) and Eckel et
al. (2002). These reports detail two primary modes of
failure with the predominant mode being translational,
retrogressive slides between SSD and the underlying
Battleford Till. These types of failures are common along
the eastern bank of the South Saskatchewan River south
of the 25th Street bridge. Along this portion of the riverbank,
the SSD is present and generally thickens to the south.
Where the SSD is thicker, the formation (Grasswoods
Formation) tends to consist of silty sand overlying a high
plasticity silty clay. The uppermost sand and silt layer acts
as an unconfined aquifer and tend to possess perched
groundwater. Hamilton and Tao (1977) indicate the key
factor associated with these types of landslips stems from
Photo: Tension cracks along Saskatchewan Crescent at
excessive groundwater within that perched aquifer.
intersection with 17th Street, 2012
Hamilton and Tao observed that regular watering of lawns
in Grosvenor Park and the University Campus could result Many of these failures are solely a result of increase pore-
in considerable increases in the groundwater table with water pressures within the slope as illustrated by the
relatively minor watering. Their study found that by adding failures occurring well above the river level. Exceptions to
approximately 25 mm of water weekly, that the this mode of failure are the Beaver Creek slide and the 15 th
groundwater level in the SSD could rise approximately 150 Street Slide. These two slides consisted of the whole slope
to 300 mm. Prior to the President’s House landslide, lawn and at least in the case of the Beaver Creek slide, was
and garden watering was adding approximately 63.5 mm retrogressive in nature. Retrogressive slides are different
of water a week. Cessation of irrigation resulted in a from the typical surficial slope failure common to the CoS
decrease in groundwater level of approximately 1.8 m in that they typically involve failure of the slope further down
locally. Hamilton and Tao (1977); Clifton et al. (1981); slope (near the river) and move upslope as the previous
Christiansen and Sauer (1998); and Eckel et al. (2002) toe fails.
clearly illustrate the effect of groundwater on this nature of
 Photo: Petursson’s Ravine, 2022
Toe erosion and progressive failures related to seasonal
Photo: Beaver Creek landslide system, 2016 high river levels were of considerable concern for years
Attempts have been made to locally dewater the SSD by with respect to landslide hazard within the city. However,
placement of lateral drains as well as drainage galleries. since the opening of Gardiner Dam and a weir near the
Eckel et al. 2002 discuss the historical construction of a north end of the downtown area, seasonal fluctuations in
series of lateral drains installed at the Diefenbaker Park river level and the rate of erosion along the outside bends
Landslide, Rotary Park and hand mined tunnels of the river have reduced. In historical areas of concern or
constructed below Saskatchewan Crescent with lateral areas where structures impact the natural flow of the river,
drains discharging into the South Saskatchewan. The hand heavy rock revetments have been constructed and have
mined tunnels consisted of a series of drilled shafts further minimized the ongoing erosion of slope.
extending into the till (depths of around 15 m bgs) and the 4 MODELLING OF LANDSLIDES IN THE COS
installation of perforated pipes between the shafts. A
schematic from Eckel et al. 2002 is shown below. Modelling of landslide activity in the CoS is rather complex
in of itself in that there are several factors that need to be
considered. Because the predominant failure mechanism
DRILLED is the SSD sliding on the contact with the underlying
SHAFTS Battleford till, this results in a translational slide. This in of
itself is not difficult to reproduce, what is difficult is using
the measured strength values to achieve a suitable back-
analysis of a given failure. Typically, the Battleford till is
assigned an infinite strength and a thin veneer of residual
strength SSD is assigned at the interface between the two
soils. This assumption is an oversimplification that
DRAINAGE
achieves the desired results to reproduce measured failure
GALLERY planes. This is an oversimplification due to the very nature
of a Limit Equilibrium model (LEM) is the factor of safety
Saskatchewan Crescent subdrain system (adapted from reported is an average factor of safety for each slice
Eckel et al. 2002 through the failure plane. This does not mean that all slices
in the slide mass have that factor of safety, rather that the
The second mode of failure occurs north of the 25 th Street majority have that value. In addition, the LEM is not
Bridge and is consistent with piping failures. In this area, capable of truly representing the failure mechanisms that
the SSD is not generally present and the till and associated occur within a given slope. With more complex modelling
intra-till sands are present near the ground surface. It is the techniques, critical structures are moving away from LEM
presence of these sand layers that tend to result in analysis and towards a more strain-based model that
elevated pore-water pressures and piping related failures. allows for progressive failure to take place. This type of
Examples of these failures are found at the Ski Jump and modelling is still in its infancy and has a long way to go, but
Petursson’s Ravine. it is now being required for more complex projects. One of
the major drawbacks of the strain-based models, is the
need for an accurate stiffness input for the peak and
residual soils. Advances in field and laboratory testing has
made the determination of a non-linear stiffness model
easier to obtain and therefore implement numerically.
Further discussion of the numerical modelling strengths
and weaknesses is provided below.
Modelling of the slope failures in the CoS is further
complicated by the occasional presence of retrogressive
slides in the city. Beaver Creek (Haug et al., 1977 and
Yoshida and Krahn, 1984) and the 15th Street slide (Clifton versus the resisting forces. As a result, the strains needed
et al., 1981) encompassed the entire slope to the river. It is to mobilize friction cannot be considered.
not clear if 15th Street slide was retrogressive, but Beaver
Creek certainly is. These failures typically occur when the In addition, the LEM makes another critical error in its
entire slope is relatively steep, and the contact of the SSD calculations. LEM considers an average factor of safety
is below or near to the river elevation. In these instances, along a given slip surface. This means that the model
river erosion can remove the toe support, which triggers a cannot consider either softened conditions at the toe due
landslide near the riverbank, which then triggers another to progressive failure or the strength of the remedial
landslide above it and so on until the slope angle is method separately from each other. Finally, 2D LEM
sufficiently flat to come to a metastable state. Haug et al. cannot indicate the extent laterally that a toe berm or shear
(1977) indicate that these slope failures should be key must be extended to stabilize the whole slope.
modelled as a series of blocks to successfully back analyse Because the slope as treated as a 2D plane strain problem,
a given failed slope. Yoshida and Krahn (1984) however it assumes that the whole slope is equal to the selected
demonstrated that by modelling the entire slide mass using cross-section. Three-dimensional edge effects are not
LEM, then a reasonable solution may be obtained. considered. As a result, the area of the stabilization design
Considering the findings of Yoshida and Krahn, wherever can sometimes fall short or extend far beyond what is
the slope is relatively steep from the riverbank to the slope needed.
crest, it should be standard practice to consider the failure To deal with this, it is suggested that the actual factor of
of the entire slope considering retrogressive conditions. safety is checked using a 3D LEM model to determine the
The practice of modelling the entire slope should also be effective extents of the area to be stabilized.
considered in cases where there may be internal drainage
to the slope; sewer outfalls; or on the outside of bends in 4.2 Finite Element Method (FEM)
the South Saskatchewan River where the SSD contact with Hassiotis et al. (1997) demonstrated that the FEM can be
the Battleford till is near to the river level. used in the analysis and design of stabilized slopes. The
Because the slope failures within the SSD are typically methods used are limited as with the LEM to 2D plane
initiated by changes in the groundwater level, an analysis strain conditions, and therefore some assumptions need to
that assesses the typical seasonal groundwater be made. Thiessen (2011) demonstrates that the results
fluctuations within the SSD should be conducted. Typical are generally in agreement with measured field results.
fluctuations under high infiltration conditions are reported FEM analysis can be used to analyze the stability of
by Hamilton and Tao (1977); Christiansen (1970) and unreinforced and reinforced slopes to achieve an optimal
Fredlund (1970). Transient analyses that reproduce the design. Abdul Razaq (2007); Abdul Razaq et al. (2010);
typical saw-toothed groundwater fluctuations are easily and Thiessen (2011) also demonstrate effective FEM
produced and can help understand seasonal design of slope stabilization.
displacements with time of a given slope and remedial Dawson et al. (1999) and Hazzard et al. (2010)
solution. demonstrated that the factor of safety of a slope can be
4.1 Limit Equilibrium Method (LEM) computed accurately in an FEM environment using the
shear strength reduction (SSR) method. Duncan et al.
The use of LEM is convenient mainly due to the ease of (2014) stated that the factor of safety for slopes can be
use and the clear indication of a factor of safety prior to and calculated iteratively by dividing the shear strength of soil
following implementation of the stabilization method. by a factor to bring the slope to the verge of failure. This
These factors of safety give a scalable improvement to the scenario is defined as the shear strength reduction method.
slope based on the implementation of a given stabilization The factor used for the iteration is called the Shear
method. Reduction Factor (SRF). Dawson et al. (1999) pointed out
Because 2D LEM applications are based on two- that the factor of safety for slopes computed with the finite
dimensional plane strain analysis, any non-plane strain element method using an SRF was comparable to factor of
remedial methods cannot be explicitly modelled and safety obtained from the traditional LEM analysis. As an
therefore any solutions other than toe berms or shear keys example, if a back analysis is carried out for a failed slope
are transformed into an equivalent shear key of appropriate using the LEM, then the parameters that are obtained from
width. The shear key is then wished into place in the the analysis result in a factor of safety of 1. The strength
stability model with assigned strength properties based on values obtained are typically either a softened, post-peak
the expected behaviour of the selected remedial method. or residual strength value. An SSR works by inputting the
The influence of the stabilization method on the factor of most probable peak strength values and reducing those
safety is calculated by determining the factor of safety of peak strengths, via the SRF until the model can no longer
the improved slope when compared to that of the original converge on a unique solution. In theory, if the values
slope. As stated above, this gives a very clear value that obtained from the LEM back analysis are used as the input
can be assigned to the degree of improvement of the slope, parameters for the SSR, then the SRF will be equal to 1.
i.e., a 30 or 50% improvement is clear and easily Alternatively, if the peak values are still valid and the slope
visualized. The strength of the LEM, (its simplicity) is has not failed, the SRF will coincide with the stable LEM
however also its weakness. Because of the nature of LEM analysis factor of safety. SSR defines failure of the slope
analysis, the models cannot consider the displacements of as an inability of the model to converge on a solution. An
the slope and only consider the statics of the driving forces inability for a model to converge is usually as a result of
excessive yielding and large strains occurring within the geohazards in Saskatchewan and the Saskatoon area
domain. Non-convergence does not necessarily mean that predominately driven by the complexity of the surficial
the model is incorrect, rather just that the tolerance of the geology. Several glaciations have basally deposited thick
calculations is not within the user defined acceptable limits. layers of till throughout the region, which have been
In most cases, the stresses within the slope will likely be overlain by relatively thin layers of pro-glacial sediments.
correct, however it is the displacements that should be This clear unconformity between the underlying, relatively
called into question whenever a model fails to converge on strong glacial sediments and the upper SSD formations
a solution. The inability to rely upon the displacement presents a preferential slide plane along much of the east
calculations, is a considerable drawback as use of an FEM riverbank of Saskatoon. Furthermore, the varying energy
is primarily to determine the strains necessary for environment for which most of the SSD was deposited has
mobilization of shear strength as well as the long-term resulted in a fining downwards structure, where a layer soft
displacements. Swan and Seo (1999) also demonstrated clay is overlain by a relatively permeable silt and sand
that the SSR appears well suited for analyzing the stability layer.
of existing slopes in which unconfined seepage is present.
Despite the benefits of using an FEM for seepage, stress Studies have demonstrated that even minor addition of
and strain analysis, Krahn (2006) recommended that the water to the upper strata results in large increases in the
use of SSR needed careful attention because of its groundwater level above the till unit. This suggests that
inherent limitations. The second limitation as Krahn (2006) watering of gardens and lawns should be kept to a
pointed out is that the strength of soil is reduced equally minimum if in relatively close proximity to the east bank of
throughout the slope. This means that the local safety the South Saskatchewan River.
factor is taken to be constant along the entire slip surface Numerical modelling (either LEM or FEM) are particularly
similar to the LEM. In reality, the local factor of safety varies difficult to conduct in the city given the limitations and
considerably along a given slip surface. These reasons are assumptions of both models. Though LEM is regularly
why there is currently no direct implementation in Slope/W carried out, it is not capable of calculating progressive
of the SSR method. Krahn (2006) is concerned about the failure of the slope, nor is it very efficient at determining the
use of SSR if a designer is unaware of the limitations and efficacy of a given remedial measure. By not being able to
decisions may be made based on an inherently flawed determine acceptable displacements post construction,
model. Griffiths and Fenton (2004); Hazzard et al. (2010); ongoing movements may imply to the general public that a
indicated that these limitations can be minimized by taking given method is not working as intended. Alternatively, FE
a stochastic approach to the FEM analysis. Usage of a models can accurately represent progressive failure, but
probabilistic evaluation of the factor of safety allows for an the accuracy of the inputs and test data typically hinder its
assessment of the range of strains and factors of safety use in most practical applications. Additionally, models that
that can be expected for a given slope using SSR. As a assess the onset of failure in the FEM may be misleading
result, a stochastic approach can provide a more useful in that they often present problems with convergence near
probability distribution of both the likelihood of failure as to the limiting state. Unlike the LEM, instability may appear
well as the associated strains following implementation of to give results that are more exaggerated than are likely the
the selected stabilization method. case. These problems point to the need for advanced site
Displacements are notoriously difficult to model in lateral characterisation as well as ongoing monitoring of the
scenarios using 2D FE models. To more accurately model existing slopes and known hot spots to ascertain suitable
lateral displacements, more involved constitutive models models for any remedial works going forward.
are required. Typically, non-linear elasto-plastic models 6. REFERENCES
like the Hardening Soil (HS) or Hardening Soil with
softening are required to accurately represent the lateral Abdul Razaq, W. F. (2007). Evaluation of riverbank
movements. These more complex constitutive models stabilization using rockfill and soil-cement columns.
require additional information in terms of the loss of Unpublished PhD Dissertation, University of Manitoba
stiffness with strain. It is well understood that as a soil mass (Canada).
displaces, the stiffness (modulus) decreases with strain. Abdul Razaq, W. F., Alfaro, M. C., & Blatz, J. A. (2010).
This in turn leads to additional displacement, which further Evaluating the stability of riverbanks reinforced with rockfill
reduces the stiffness of the soil mass. The loss in stiffness columns. Lowland technology international: the official
continues until a state of constant volume (effective stress journal of the International Association of Lowland
analysis) or pure plasticity (total stress analysis) is Technology (IALT)/Institute of Lowland Technology, Saga
achieved. The HS models predict lateral displacements far University, 12(2), 1-8.
more precisely than the simpler models that are commonly
used in practice (Burland, 1989; Simpson, 1992; Benz et Abramson, L. W., Lee, T. S., Sharma, S., & Boyce, G. M.
al., 2009; Kelln et al., 2009). The HS model requires (2002). Slope stability and stabilization methods. John
detailed stiffness degradation data typically obtained from Wiley & Sons.
triaxial and seismic field testing.
Benz, T., Vermeer, P. A., & Schwab, R. (2009). A small‐
5. CONCLUSIONS strain overlay model. International journal for numerical
and analytical methods in geomechanics, 33(1), 25-44.
Though commonly thought as being a relatively flat area
with little relief, there are a surprising number of landslides Burland, J. B. (1989). Ninth Laurits Bjerrum Memorial
throughout the Prairies. The key factors of most Lecture:" Small is beautiful"—the stiffness of soils at small
strains. Canadian geotechnical journal, 26(4), 499- Hazzard, J., Hammah, R., Curran, J. (2010) Advances in
516.Christiansen, E.A. (1970). Geology. in Physical probabilistic slope stability analysis using the finite element
Environment of Saskatoon, Canada, EA Christiansen method. Proc. of the 63rd Canadian Geotechnical
(Ed.), Saskatchewan Research Council in co-operation Conference & 6th Canadian Permafrost Conference,
with the National Research Council of Canada, NRC Calgary AB.
Publication 11378. 3-17.
Kelln, C., Sharma, J., Hughes, D., & Graham, J. (2009).
Christiansen, E.A., & Meneley, W.A. (1970). Slope Finite element analysis of an embankment on a soft
Stability. in Physical Environment of Saskatoon, Canada, estuarine deposit using an elastic–viscoplastic soil model.
EA Christiansen (Ed.), Saskatchewan Research Council in Canadian Geotechnical Journal, 46(3), 357-368.
co-operation with the National Research Council of
Canada, NRC Publication 11378. 51-52. Krahn, J. (2006). The limitations of the strength reduction
approach. GEO-SLOPE Direct Contact Newsletter.
Christiansen, E. A., & Sauer, E. K. (1984, September).
Landslide styles in the Saskatchewan River plain: a MacDonald, A. B., & Sauer, E. K. (1970a). The engineering
geological appraisal. In Proceedings of the 37th Canadian significance of Pleistocene stratigraphy in the Saskatoon
Geotechnical Conference. Toronto, Ont. Canadian area, Saskatchewan, Canada. Canadian Geotechnical
Geotechnical Society, Alliston, Ont (pp. 35-48). Journal, 7(2), 116-126.

Christiansen, E. A., Sauer, E. K., Karrow, P. F., & White, MacDonald, A. B., & Sauer, E. K. (1970b). Engineering
O. L. (1998). Geotechnique of Saskatoon and surrounding Properties of Tills. in Physical Environment of Saskatoon,
area, Saskatchewan, Canada. Geol. Assoc. Can. Spec. Canada, EA Christiansen (Ed.), Saskatchewan Research
Pap, 42, 117-145. Council in co-operation with the National Research Council
of Canada, NRC Publication 11378. 53-54.
Chursinoff, R. W. (1980). Measuring Ground Movement by
Precise Survey (MSc dissertation, University of Simpson, B. (1992). Thirty-second Rankine Lecture:
Saskatchewan). Retaining structures: displacement and design.
Geotechnique, 42(4), 541-576.
Clifton, A. W., Krahn, J., & Fredlund, D. G. (1981).
Riverbank instability and development control in Swan, C. C., & Seo, Y. K. (1999). Limit state analysis of
Saskatoon. Canadian Geotechnical Journal, 18(1), 95-105. earthen slopes using dual continuum/FEM approaches.
International Journal for Numerical and Analytical Methods
Dawson, E. M., Roth, W. H., & Drescher, A. (1999). Slope in Geomechanics, 23(12), 1359-1371.
stability analysis by strength reduction. Geotechnique,
49(6), 835-840. Thiessen, K. (2011). Stabilization of Natural Clay
Riverbanks with Rockfill Columns: A Full Scale Field Test
Duncan, J. M., Wright, S. G., & Brandon, T. L. (2014). Soil and Numerical Verification. Unpublished PhD Dissertation,
strength and slope stability. John Wiley & Sons. University of Manitoba (Canada).
Eckel, B., Christiansen, E.A., Richardson, N., & Schreiner, Whitaker, S. H., & Christiansen, E. A. (1972). The Empress
B. (2002) Trip B7- Riverbank instability in the City of group in southern Saskatchewan. Canadian Journal of
Saskatoon. Saskatoon 2002 Field Trip Guidebook. Earth Sciences, 9(4), 353-360.
Geological Association of Canada Joint Annual Meeting.
Yoshida, R. T., & Krahn, J. (1985). Movement and stability
Fredlund, D.G. (1970) Guide to Foundation Design. Report analysis of the Beaver Creek landslide, Saskatchewan,
to the University of Saskatchewan Buildings and Grounds Canada. Canadian Geotechnical Journal, 22(3), 277-285.
Department, Saskatoon, Canada.
Griffiths, D. V., & Fenton, G. A. (2004). Probabilistic slope
stability analysis by finite elements. Journal of geotechnical
and geoenvironmental engineering, 130(5), 507-518.
Hamilton, J. J., & Tao, S. S. (1977). Impact of urban
development on groundwater in glacial deposits. In
Proceedings of the 30th Canadian Geotechnical
Conference, Saskatoon (Vol. 2, pp. 1-35).
Hassiotis, S., Chameau, J. L., & Gunaratne, M. (1997).
Design method for stabilization of slopes with piles. Journal
of Geotechnical and Geoenvironmental Engineering,
123(4), 314-323.
Haug, M. D., Sauer, E. K., & Fredlund, D. G. (1977).
Retrogressive slope failures at Beaver Creek, south of
Saskatoon, Saskatchewan, Canada. Canadian
Geotechnical Journal, 14(3), 288-301.
Design and Construction of Concrete Pile
Retaining Walls for Slope Stabilization along
a River Valley Trail in Edmonton, Alberta
Graeme Law, Robin Tweedie, & Tamer Elshimi
Thurber Engineering Ltd., Edmonton, Alberta, Canada
Gordon Pauls & Paul R. Lach
City of Edmonton, Edmonton, Alberta, Canada

ABSTRACT
A landslide occurred along the North Saskatchewan River’s south valley slopes during the summer of 2020, causing the
collapse of three timber retaining walls and damaging an important shared use trail southwest of the Capilano Bridge in
Edmonton, Alberta. The failed slopes required stabilization involving the construction of two parallel cast-in-place concrete
pile retaining walls, and reconstruction of the above-ground retaining walls and trail, in order to reopen the shared use trail
for cyclists and pedestrians. A geotechnical investigation was carried out to evaluate the overall slope failure and provide
adequate information for the trail rehabilitation. This involved slope stability analysis and finite element analysis using
PLAXIS 2D. The results of these analyses were then used in structural analyses to design the twin buried concrete pile
retaining walls. Slope movements before and during construction were monitored using slope inclinometers. The
management of construction risk was also supplemented by survey monitoring of slope stakes used as surface
monuments.

RÉSUMÉ
Un glissement de terrain s'est produit le long des pentes sud de la vallée de la rivière Saskatchewan Nord au cours de
l'été 2020, provoquant l'effondrement de trois murs de soutènement en bois et endommageant un important chemin à
usage partagé au sud-ouest du pont Capilano à Edmonton, en Alberta. Les pentes défaillantes ont nécessité une
stabilisation impliquant la construction de deux murs de soutènement parallèles sur pieux en béton coulé sur place et la
reconstruction des murs de soutènement hors sol et du sentier, afin de rouvrir le sentier à usage partagé pour les cyclistes
et les piétons. Une enquête géotechnique a été effectuée pour évaluer la rupture globale de la pente et fournir des
informations adéquates pour la réhabilitation du sentier. Cela impliquait une analyse de la stabilité des pentes et une
analyse par éléments finis à l'aide de PLAXIS 2D. Les résultats de ces analyses ont ensuite été utilisés dans des analyses
structurelles pour concevoir les murs de soutènement à pieux en béton enterrés jumelés. Les mouvements de pente avant
et pendant la construction ont été surveillés à l'aide d'inclinomètres de pente. La gestion du risque construction a également
été complétée par un suivi topographique des piquets de talus utilisés comme bornes de surface.

1 INTRODUCTION

A landslide occurred in the lower portion of the North

Saskatchewan River south valley slope during the summer
of 2020, causing the collapse of three timber retaining walls
and damaging an important shared use trail southwest of
the Capilano Bridge in Edmonton, Alberta (Figure 1). The
landslide impacted a 100 m long section of the Riverside
Trail, impeding public mobility and restricting the
connectivity of the trail system. The sliding surface is within
the uppermost portion of the weathered bedrock. The
landslide is mostly composed of colluvium, a mixture of
surficial deposits and the uppermost weathered bedrock.
In 2021, a geotechnical investigation was conducted to
characterize the soil and bedrock conditions, determine the
mechanism of the landslide, and establish an appropriate
remediation strategy to stabilize the slope and allow for the
re-opening of the trail system. Results of the geotechnical
investigation indicated that the preferred stabilization
option consisted of the installation of twin buried retaining
walls to reinforce the slope and retain the upslope soils. Figure 1. Key Plan Showing Project Location

This paper provides an overview of the geotechnical

investigation, engineering design, construction
remediation, and slope monitoring before and during slide
repair.

2 PROJECT SETTING

2.1 Site Conditions

The site is located about 250 m southwest of the Capilano

Bridge which carries Wayne Gretzky Drive NW over the
North Saskatchewan River in Edmonton, Alberta
(Figure 1). At this location, the river valley is approximately
50 m deep with an average slope of about 20 degrees.
A shared use trail was constructed in the lower part of
the slope in the 1970’s, providing connectivity through the
North Saskatchewan River valley. A series of low timber
retaining walls were constructed on the upslope side of the
trail to retain slope cuts up to about 3 m high.
A landslide occurred at this site in 2020, affecting about
a 100 m long section of the Riverside Trail starting about
200 m west of the Capilano Bridge. The landslide extended
above and below the asphalt trail and caused the collapse
of three of the existing timber retaining walls on the upslope
(south) side of the trail (Figures 2 and 4). Several tension
cracks with opening widths of up to 150 mm and soil
subsidence/heave of up to 500 mm were observed along
the trail alignment within the landslide area.
Head scarps up to 4 m in height were noted at different
elevations above and below the trail level along Cross-
section A-A’ (Figure 2, 4 and 5). A gully that extends from
the top of the slope down to the trail elevation was also
evident, which resulted in surface water collecting and
draining towards the failed retaining wall area. The plan
extents of the observed landslide location and features are
shown in Figure 2 and Figure 4 below.
Figure 2. Surficial Geology Map
2.2 Geology
Two colluvial units can be distinguished in the study
The North Saskatchewan River valley downcut rapidly area: an upper unit comprised of thin colluvium covering
during the last 10,000 years and is incised into bedrock of the major scarps and composed of weathered
the Upper Cretaceous Horseshoe Canyon Formation, glaciolacustrine deposits; the lower unit is colluvium
comprised of deltaic and fluvial deposits of interbedded composed of moved bedrock and surficial deposits. The
fresh and brackish water sandstone, siltstone, and shale. lower unit derives from large and deep-seated landslide
Typical sediments consist of grey, greenish, and white movements, which now appear to be less active or very
weathered non-cemented bentonitic feldspathic slow-moving. The major discontinuity or lower sliding
sandstone, brown bentonitic clay shales, coal seams, and surface is along the relatively flat-lying bentonite layer at
bentonite layers. Less common are concretions and beds approximately 606 to 607 m at or near the riverbed level.
of hard calcareous sandstone, and thin nodular beds of This bentonite layer is located above the main seam of a
red-brown ironstone. historical coal mine.
The bedrock is overlain by surficial deposits composed
of Tertiary and late Quaternary Period deposits, shown in 2.3 Coal Mining
Figure 2. Tertiary deposits in the Edmonton area consist of
preglacial sand and gravel of the Empress Formation, Coal mining was prevalent throughout the North
which was deposited in the buried valley whose thalweg is Saskatchewan River valley in the last century, and access
located east of the study area. The buried valley is a to the coal seams was frequently by inclined adits within
tributary to the Beverly Valley and trends in a south-north the valley slopes. A coal seam located approximately 55 m
direction. The thickness of the Empress sand formation below the uplands elevation (approximate coal seam
decreases towards the landslide location. elevation of 605 m, Cross-section A-A’ in Figure 5) was
Deep seated landslides are common throughout the mined from 1931 to 1951 in the Red Hot Mine (Alberta
river valley and have resulted in the valley slopes being Energy Regulator, Taylor 1971) in the area of the landslide
mantled with colluvial deposits, consisting of weathered site.
surficial or bedrock deposits. Evidence of the former coal mining activities is visible
along the North Saskatchewan River valley slope below the
active slide area. There are visible cuts into the slope along
the riverbank, which were most likely used for coal loading
onto a barge, and indications of earthwork at the crest of
the slope to facilitate access to the coal mine (traces shown
in Figure 3).
Evidence of the impacts of the former coal mining has
also been found in several test holes completed by the City
and others. This includes distorted bedrock, non-horizontal
bedding planes, voids, and evidence of mine workings. In
addition, there are reports of major subsidence within and
above the workings in 1942. The coal mining has resulted
in weakening of the overall valley slopes and contributed to
deep-seated landslides elsewhere in the area, including
the Grierson Hill Slide (Wang, Tweedie, and Lach 2018).

Figure 3. 3D view of the site (2019 LiDAR imagery)

3 SITE INVESTIGATION

A geotechnical investigation was carried out in 2021

which included drilling three test holes along the existing
trail (Thurber Engineering Ltd., 2021). Three slope
inclinometer casings, SI20-1, SI20-2, and SI20-3, were
installed in the completed bore holes and extended to an
elevation of about 603 m, terminating below a coal layer, to
allow for future monitoring of potential slope movements at
the test hole locations and to identify the depth of the slip
surface(s). Two nested vibrating wire piezometers were
also installed in each test hole to permit future monitoring
of the groundwater levels in the overburden and bedrock at
this site. A standpipe piezometer was also installed in a
shallow test hole drilled about 1 m east of TH20-2 to a Figure 4. Site plan showing the slide scarps prior to
depth of about 5 m below the ground surface. Locations of construction, cross-section location, instrumentation
the instrumentation are shown in Figure 4. One direct shear location, and proposed pile wall alignment
test was performed on a clay shale sample obtained from
about 15 m below grade. 4 STRATIGRAPHY
Evidence of the former coal mining activities, including
the presence of wood (former mine supports) and voids The general stratigraphy below the trail consisted of thin fill
(evidence of coal extraction), were encountered during the layers, overlying colluvium (thin layers of clay till and sand),
drilling of Test Hole TH20-1 at about 17 m below grade, followed by clay shale and sandstone bedrock which was
corresponding to an approximate elevation of 606.5 m. encountered at approximate elevations of 619 to 620 m at
the test hole locations. Based on the 2019 LiDAR
(Figure 3), 2021 geotechnical investigation, and known
slope morphology, a representative cross-section of the
soil stratigraphy expected at the site location was
developed and illustrated in Figure 5.
Figure 5. Generalized cross-section A-A’ showing the location of estimated failure planes A (deep) and B (shallow)

Colluvial materials were encountered at the test hole Table 1: Piezometric Readings
locations below the fill and were comprised of thin layers of Tip Max. June
Material at
glacial clay till and sands. Bedrock was encountered in the Piezometer Depth
Tip
Level 2021
three test holes at depths ranging from 2.5 to 4.3 m below (mBGS) (mBGS) (mBGS)
the asphalt trail surface. The bedrock consisted primarily of VW20-1A 4.5 Clay Till 1.4 1.4
bentonitic and carbonaceous clay shale with siltstone and VW20-1B 13.5 Clay Shale Damaged
VW20-2A 10.0 Sandstone 6.3 6.4
sandstone layers. The upper 1 to 2 m of the bedrock was
VW20-2B 18.5 Sandstone 10.4 10.4
highly disturbed, with Standard Penetration Test (SPT) ‘N’ SP20-2 5.0 Sand/Bedrock 2.8 2.8
values ranging from six to seven blows per 300 mm, VW20-3A 8.0 Sandstone 4.8 7.1
indicating a firm consistency. Bentonite and coal layers VW10-3B 16.0 Clay Shale 9.1 9.1
were encountered in all three test holes at elevations of 606 Note: mBGS is metres below the ground surface.
to 607 m. Wood pieces (remnants of the Red Hot Mine
workings) were also encountered above the coal layers in 6 SLOPE STABILITY ASSESSMENT
TH20-1 and TH20-2. A possible void, indicative of former
coal extraction, was noted in test hole TH20-1 at an 6.1 MECHANISM OF FAILURE
elevation of 606 m. The water contents of the bedrock
samples varied widely from 8 to 55 percent. Atterberg Results of the geotechnical investigation, supplemented by
Limits tests performed on selected clay shale samples geological interpretation, indicate two separate slope
yielded liquid limits ranging from 70 to 427 percent and failure mechanisms, as follows:
plastic limits ranging from 27 to 38 percent, indicating high
plasticity. ● A deep-seated landslide extending from the crest of
One direct shear test conducted on a selected clay the valley slope to the toe of the slope, with the lower
shale sample indicated peak and residual friction angles of translational slide surface through coal and bentonite
26° and 11°, respectively. The peak and residual cohesion layers at elevations of approximately 606 to 607 m.
were 13 kPa and 0 kPa, respectively. The deep-seated slide appears to be inactive (SI20-1
and SI20-3) or slow-moving (SI20-2) at present.
5 INSTRUMENTATION READINGS ● A local (shallow) landslide through the uppermost
weathered bedrock at elevations of approximately 618
The porewater pressures measured in the vibrating wire to 619 m. This landslide is currently active and is
piezometers (VZ) and standpipe are summarized in causing the noted distress to the existing trail and
Table 1 below. Slope inclinometer readings taken before, timber retaining walls within the study area.
during, and after construction over a period from October
2020 to April 2023, and are presented in Table 2.
 Table 2: Slope Inclinometer Readings Before and During Construction
During Construction to
Pre-Construction During Construction
SI Movement Current Date
Pile No. (Oct 2020 to Jun 2021) (Sep to Dec 2022)
(Tip Zone (Sep 2022 to Apr 2023)
(Location )
1
Depth) (mBGS) Inc. Rate Cumul. Inc. Rate Cumul. Inc. Rate
Cumul. (mm)
(mm/yr) (mm) (mm/yr) (mm) (mm/yr)
SI20-1 N/A (32m) 4 – 6.4 8.5 12.2 - - - -
SI20-2 N/A (55m) 4 – 6.4 7.0 10.0 - - - -
SI20-3 N/A (96m) 2.5 - 4 8.4 12.1 - - - -
SI22-4 TP232 (102m) 1 – 15 - - 9.6 18.5 6.5 15.2
SI22-5 TP102 (45m) 1 – 15 - - 14.4 19.3 7.5 17.5
SI22-6 S65 (94m) N/A3 - - - - <1 <1
SI22-7 N48 (70m) N/A3 - - - - <1 <1
SI22-8 N32 (47m) N/A 3
- - - - <1 <1
SI22-9 N16 (22m) N/A3 - - - - <1 <1
mBGS is metres below ground surface; “Inc. Rate” refers to the incremental movement rate determined since the previous reading.
1
Approximate distance from westernmost piles, Pile N1 and Pile S2.
2
Temporary shoring pile. 3
No discernable movement at depth, less than 1mm.
It is suspected that the recent shallower landslide surface location. The soil parameters determined by the
affecting the trail is a result of gradual weakening of the back-analysis are summarized in Table 3.
weathered bedrock due to ongoing slow creep movements, Two potential slip surfaces were identified as
and possibly exacerbated by rising groundwater levels representative and used as the basis for evaluating the
during high precipitation events noticed over the few years remedial options. The first surface (denoted ‘A’ in Figure 5)
preceding the sliding. Based on the available rainfall assumed a deeper failure plane (through the softened clay
records for the City of Edmonton, the total precipitation in shale) which encompassed the shallow movement and
2019 and 2020 was about 22 and 32 percent, respectively, resulted in a back-calculated factor of safety (FS) ranging
higher than the average precipitation in the last 20 years. from 1.05 to 1.13. The second surface (denoted ‘B’ in
As noted, historical coal mining activity has also generated Figure 5) had a shorter overall failure plane which resulted
weakened zones in the upper bedrock, contributing to past in a back-calculated factor of safety (FS) ranging from 0.90
slope instability at this location. to 0.97. These analyses were expected to represent the
range of likely failure surfaces operative at this location.
6.2 REPAIR OPTIONS The selected remedial option consisted of the installation
of two rows of buried tangent concrete pile walls (one row
Several options for rehabilitating the trail that were of piles on each side of the trail) extending into the
originally evaluated included the following: undisturbed bedrock to achieve a target factor of safety of
1.3 at the trail location, equivalent to a 30 percent increase
● Locally reinforcing the slope above and below the trail in the slope stability factor of safety. Since slope
with two buried concrete pile retaining walls to improve movements were observed upslope and downslope of the
the stability of the slope along the existing trail trail, two walls were required to provide the required
alignment within the study area. stabilization of the trail and to avoid significant earthwork in
● Realignment of the trail to a lower, relatively more the landslide area.
stable bench closer to the river. Based on the results of the preliminary stability
analyses, the piles were required to extend into the
After reviewing these options, the retaining walls undisturbed bedrock to an elevation of 611 m (pile depth of
stabilization option was selected in order to maintain the approximately 12 m). The length of the pile walls installed
trail on its existing alignment and to better manage the were about 120 m (a total length of 240 m for the two walls).
longer-term hazard risk. No significant earthwork nor The minimum required shear resistance of the pile walls
environmental disturbance would be required with this ranged between 120 kN/m and 280 kN/m for different
option, and the piles were to be designed assuming that locations along the trail. Piles on the east side of the
the terrain downslope of the trail will continue to experience landslide required larger shear resistance to provide the
shallow slope movement. required slope reinforcement.
Drainage measures were also required upslope and
6.3 Stability Analysis and Preliminary Design downslope of the retaining walls to control the groundwater
pressures behind the walls. These measures included
A back-analysis of the existing conditions was initially subdrains behind the upper wall and drainage channels to
undertaken to determine appropriate geotechnical drain the collected water towards the river downslope of the
parameters for evaluating the slope stabilization options. trail. The channels were lined with non-woven geotextile
Initial soil and groundwater parameters were selected and riprap for erosion protection.
based on the available subsurface information, It should be noted that this remedial option was
groundwater readings, and previous experience on similar designed to stabilize the local shallow slide directly
sites, and the back-analyses were used to refine the affecting the trail and retaining walls, and not to stabilize
parameters to obtain a factor of safety near unity. The the overall valley slope.
cracks observed on the trail were used to control the slip
Table 3: Summary of Soil Properties used in the Stability and Deformation Analyses

γ c’ φ’ E ν
Soil Type
(kN/m3) (kPa) (°) (MPa) (-)

Clay 19 1 23 15 0.35
Clay Till 20 5 28 25 0.30
Colluvium 19 3 18 15 0.30
Sand 19 0 30 30 0.30
Disturbed Coal / Bentonite 18 0 11 20 0.30
Softened Clay Shale 20 1 16 20 0.30
Clay Shale / Sandstone (Upper) 21 25 25 100 0.35
Clay Shale / Sandstone (Lower) 21 50 25 100 0.30
Note: γ – Unit Weight; c’ – Effective Cohesion; φ’ – Effective Friction Angle; E – Elastic Modulus, ν – Poisson’s ratio

6.4 Deformation Analysis and Detailed Design ● A temporary retaining wall was installed, where
necessary.
Deformation analyses were carried out for the proposed ● Simulation of the short-term performance of the
pile walls using the finite element software PLAXIS 2D. temporary retaining wall was estimated by
This software was developed specifically for the analysis of removing the material from the downslope side of
two-dimensional geomechanics and soil-structure the wall to create a working area for the
interaction problems using the finite element method. installation of the cast-in-place concrete piles.
● The cast-in-place concrete pile walls were then
6.4.1 Methodology and Assumptions added to the model.
● Backfilling the area between the temporary
The geometry of each cross-section was developed using retaining wall and the south concrete pile wall was
survey data from 2021. A combination of field observations then modeled.
and the 2019 LiDAR data were used to estimate the ● Simulation of the long-term performance of the
geometry of the slope outside the limits of the surveyed pile walls was estimated assuming that the
area. landslide downslope of the north pile wall will
The analyses were carried out assuming two rows of retrogress to the north pile wall location.
cast-in-place concrete pile walls, one on each side of the To minimize the effects of the model boundary
trail, with a distance of about 4.5 m (centre to centre) conditions, a two-dimensional geometric model 35 m high
between them to allow the reconstruction of the 3 m wide and 115 m wide was adopted in the analyses. The finite
asphalt trail. element mesh of the computational domain is shown in
A temporary pile wall, consisting of H-piles and timber Figure 6.
lagging, was also considered in the analysis to provide The response of different soils was simulated using a
short-term support to the slope while removing the existing linear elastic, perfectly plastic Mohr-Coulomb model (MC
failed timber retaining walls. The temporary pile wall was model). Soil parameters used to define the MC model were
assumed to be placed at about 2 m upslope of the south estimated based on the results of the geotechnical
cast-in-place concrete pile wall. investigation and are summarized in Table 2. The
The anticipated sequence of construction and proposed retaining walls were simulated using structural
operation which was simulated is as follows: plate elements, with a pile material modulus of elasticity of
25 GPa and 200 GPa for the concrete and steel elements,
● The in-situ stress field of the slope was first respectively.
established in the initial computation phase.

Figure 6. Generalized cross-section showing the finite element mesh used in the analyses
 (a) (b)
Figure 7a. Contours of long-term horizontal deformations with a downslope landslide retrogression
Figure 7b. Estimated maximum shear forces and bending moments in the north pile wall used for design

6.5 Analysis Results and was carried out in the following sequence, largely from
east to west:
It was determined that two pile walls were required to
provide long-term local slope stability at the trail location. ● Baseline and ongoing survey of the slide extents.
The pile walls consisted of cast-in-place concrete tangent ● Removal of the existing trail.
piles with a minimum diameter of 0.9 m and a minimum ● Removal of the existing timber retaining structure.
embedment of 12 m. The concrete piles were installed at a ● To mitigate the risk of further slope
spacing of less than two times the pile diameter (1.5 m destabilization, the permanent north pile wall was
centre to centre) as shown in Figure 8. installed first to allow for construction to proceed
The estimated maximum lateral deformation of the in a stable working area.
proposed 0.9 m diameter cast-in-place concrete piles ● Construction of a temporary retaining structure
ranged from 10 to 47 mm. The estimated maximum shear with two temporary slope inclinometers.
forces and bending moments for the concrete piles range ● Construction of the south permanent retaining
from 66 to 201 kN/m and 159 to 568 kN-m/m, respectively. structure.
To maintain the recommended factor of safety of at least ● Trail paving and landscaping restoration to final
1.3 from a stability perspective, the cast-in-place concrete grade and design.
piles were designed to provide a minimum shear resistance
of 280 kN/m. The contours of estimated horizontal soil The demolition of the original timber retaining wall and
deformation are shown in Figure 7a, and the maximum the installation of the temporary retaining wall was
estimated shear forces and bending moments are shown completed segmentally where only a small section of the
in Figure 7b. upslope material was left unsupported at one time. As part
Construction of a temporary retaining system was of the temporary shoring design, an observational
required to retain the failed soils upslope of the trail at the approach was taken using the daily slope survey and
locations of the failed timber retaining walls. The temporary temporary slope inclinometer results. Groundwater was
retaining system also allowed for the construction of a controlled with a subdrain system surrounded with gravel,
working platform to install the cast-in-place pile walls. connected to culverts discharging water downslope, away
The estimated deflection, shear forces, and bending from the permanent wall.
moments of both the temporary and permanent walls were In total, 158 permanent concrete cast-in-place piles
then used in the structural analyses to design the two pile were installed in 2022. Four slope inclinometers (SI22-6,
walls with soldier pile stickup sections to retain upslope SI22-7, SI22-8, and SI22-9) were also installed in the
soils. The permanent pile walls consisted of 0.9 m diameter permanent pile walls during construction. The backfill
reinforced concrete piles, spaced 1.5 m apart centre-to- behind the wall and final grading, paving, and landscaping
centre, and embedded about 12 m below trail level was completed in the summer of 2023. The approximate
(Figure 8). construction cost of this slope stabilization was about
$4,500,000.
7 CONSTRUCTION

7.1 Summary

The slide remediation construction work was undertaken in

2022 and included the installation of additional
instrumentation at the site. The construction generally
adhered to the design specifications as shown in Figure 8
 the installation of temporary shoring in July 2022. The
control point was destroyed on August 8, 2022.

Figure 9. Slope survey results during construction.

8 ACKNOWLEDGEMENTS
Figure 8: Tangent Pile Wall Design Section
The authors thank and acknowledge the City of Edmonton
7.2 Monitoring during Construction for providing third-party information utilized in preparing
this paper and for their permission to publish. We
Baseline readings were measured from the slope specifically acknowledge Mr. Gordon Pauls and Dr. Paul
inclinometers installed in 2020 prior to construction. Lach, for their review of this paper. We would also like to
Routine slope inclinometer readings were taken during thank Dr. Zichao Wu for his assistance with the structural
construction when accessible and are summarized in design of the pile wall section.
Table 2. It was a positive indication that the rates of
movement measured in the slope inclinometers installed in 9 REFERENCES
the permanent pile wall were lower than the movements
estimated from the slope inclinometers installed in 2020 Alberta Energy Regulator / Alberta Geological Survey.
(SI20-1, SI20-2, and SI20-3). Total deflections measured 2013. Bedrock Geology of Alberta. Map 600.
in the permanent pile walls up until April 2023 were less Alberta Government. Alberta Energy Regulator. Coal Mine
than 5 m, far less than the design horizontal movement of Viewer
about 50 mm. Further readings of the slope inclinometers https://extmapviewer.aer.ca/AERCoalMine/Index.html.
will be monitored on an ongoing basis to assess future Andriashek, L.D. 1987. Bedrock Topography and Valley
movement. Thalwegs of the Edmonton Map Area. NTS 83H.
Twenty-eight (28) surficial slope monitoring points were Alberta Research Council.
installed throughout the landslide area prior to construction Bayrock, L.A. 1972. Surficial Geology Edmonton. NTS
and measured daily (Figure 9). An observational approach 83H. Alberta Research Council.
was employed by using the daily surficial survey data as City of Edmonton. 1987. Proposed Sewer Shaft 75 Street
construction progressed to assess the effect that the and Rowland Road. Materials Engineering Section –
construction activities had on the stability of the slope and Engineering Division – Transportation.
to balance the risk with the project coordination and Kathol, C.P. and McPherson, R.A. 1975. Urban Geology of
schedule of construction on an unstable slope. Prior to Edmonton. Bulletin 32. Alberta Research Council.
construction, stop-work threshold values were developed, Taylor, Richard Spence. 1971. Atlas: Coal-mine Workings
and mitigation measures were established in the event that of the Edmonton Area.
movements exceeding tolerable limits were observed. Thurber Engineering Ltd. 2021. Riverside Trail Instability
The surficial points generally moved less than 50 mm West of Capilano Bridge in Edmonton, Alberta –
downslope throughout construction. Notable slope Geotechnical Investigation and Preliminary
movements were measured downslope of Pile N65 (black Assessment. Report submitted to the City of Edmonton,
marker in Figure 9), which corresponded with the piling Thurber Engineering Ltd.
drilling rig installing the permanent piles in July where up to Wang, X. Tweedie, R.W., and Lach. P.R. 2018. Grierson
200 mm of movement was measured. After the permanent Hill Slide and Stabilization Strategies for Developments
piles were installed at this section, the movement stabilized in the Area. GeoEdmonton Conference Proceedings.
over the duration of construction as shown. Another
instance of notable slope movement was measured
upslope of Pile S15 (red marker in Figure 9), which
corresponded to the excavation at that location required for
The Borden Bridges – Remedial Design
for a Large Scale Deep Seated Landslide
Jon Osback, Allen Kelly, Wayne Clifton
Clifton Engineering Group, Saskatoon, SK, Canada

ABSTRACT
The east abutments of the Borden bridges, located on Saskatchewan
Highway No. 16 across the North Saskatchewan River, were
constructed atop ancient landslide debris in 1937, 1985 and 1997. The
slip surface is located within bedrock clay shale at a depth between 38 m and 44 m and has been sheared to residual
strengths. Post construction displacements were minimal, until 2010 – 2012, when higher than average precipitation led
to increased porewater pressures and subsequent displacement in the range of 30 mm/yr, resulting in the shearing off of
some slope inclinometers. The risk of future damage to the bridge was a significant concern. This case study paper
summarizes the 2D and 3D stability analysis undertaken to assess remedial options and produce a remedial design, along
with the key impacts to the factor of safety at this site.

RÉSUMÉ
Les culées est des ponts Borden, situées sur la route no 16 de la Saskatchewan à travers la rivière Saskatchewan Nord,
ont été construites sur d'anciens débris de glissement de terrain en 1937, 1985 et 1997. La surface de glissement est
située dans le schiste argileux du substratum rocheux à une profondeur comprise entre 38 m et 44 m et a été cisaillé aux
résistances résiduelles. Les déplacements après la construction ont été minimes, jusqu'en 2010-2012, lorsque des
précipitations supérieures à la moyenne ont entraîné une augmentation des pressions interstitielles et un déplacement
subséquent de l'ordre de 30 mm / an, entraînant le cisaillement de certains inclinomètres de pente. Le risque de dommages
futurs au pont était une préoccupation importante. Ce document d'étude de cas résume l'analyse de stabilité 2D et 3D
entreprise pour évaluer les options d'assainissement et produire une conception d'assainissement, ainsi que les principaux
impacts sur le facteur de sécurité sur ce site.

1. INTRODUCTION remedial design that could be immediately implemented

should accelerated movements occur again.
Background The focus of this paper is to discuss the stability model
development, the assessment of remedial options and the
The Borden Bridges, owned and operated by the resulting remedial design. Key impacts to the calculated
Saskatchewan Ministry of Highways (MoH), are located factor of safety (FoS) at this site are also discussed.
across the North Saskatchewan River approximately 7 km
south east of Borden, SK on Highway No. 16 Control
Section 16-23 as shown in Figure 1. The first southmost
bridge was constructed in 1937, and decommissioned and
replaced in 1985. In 1997, the newest (middle) bridge was
built to accommodate two lanes of traffic in both the
eastbound and westbound directions.
Constructed on a deep-seated historical landslide, the
east abutment has been instrumented with piezometers
and slope inclinometers (SIs) and monitored since
construction of the second bridge, with post construction
slope movement being observed in the range of 0.8 mm to
2 mm per year. In 2010 to 2012, higher than average
precipitation (up to 640 mm/yr compared to about 350
mm/yr on average) was realized, and porewater pressure
head at the shear zone increased by up to 4 m, resulting in
accelerated landslide movement of up to 30 mm/year and
shearing off of some SIs. MoH was concerned about the
integrity of this major asset and initiated this project.
Panesar et al. (2014) provided recommendations for Figure 1 - Project Location
further study at this site; those which were included in this
work were analysis of river hydraulics, additional site
investigations and instrumentation installations and 3D
stability analysis. Other project components included
historical review, updated 2D stability analysis and detailed
remedial design. The intent of the project was to develop a
 Geologic Setting 44 m, corresponding to elevations between 412 m and 418
masl.
The geology of the Borden Bridges area is described in
Christiansen (1968) and Christiansen (1970) - in ascending
order, the local stratigraphy is comprised of bedrock clay
shale of the Lea Park and Judith River formations; glacial
till of the Saskatoon Group; and, stratified drift. Bedrock
clay shale was deposited as Cretaceous inland sea
sediments. The Saskatoon Group tills were deposited
throughout a series of glaciations, with the glacial activity
resulting in significant shearing, disturbance and glacial
thrusting of the upper shale. Upper stratified drift is present
from alluvial deposition throughout the lifetime of the river.
Figure 2 outlines site stratigraphy and piezometric
conditions.

Figure 3 - Disturbed shale sampled from 38.5 m to 39.5 m

Figure 2 - Site stratigraphy

The North Saskatchewan River valley was formed via

erosion by post glacial meltwater, which eroded through
the till into the shale, destabilizing the valley wall and
resulting in large-scale, deep-seated retrogressive
landslides that are common throughout the Saskatchewan
River valleys that are eroded into clay shale. Typical
characteristics of these landslides include relatively flat
valley wall slopes which correspond closely with the
residual friction angle of regional bedrock clay shales of
5.6o to 8.7o (Eckel et al. 1986; Sauer 1982); composite
failure surfaces resulting in differential movement at
surface and between landslide blocks, with a significant Figure 4 - Slickensided shale at 38.4 m
length of the shear zone being close to horizontal (Kelly et
al. 1995). Horst and graben topography is typical and
discussed frequently in the literature; this feature is not 2 INSTRUMENTATION DATA
obvious at Borden in most aerial imagery but the complex,
retrogressive composition of the landslide is apparent in Many site investigation and instrumentation programs have
geological cross sections as shown in Figure 2. As such, been conducted at the site and have consisted primarily of
determination of the landslide extents has primarily relied collection of shale core and installation of SIs and vibrating
on 3D model predictions. wire piezometers (VWs) - Figure 5 shows the borehole
The shear zone(s) of these landslides are typically locations drilled in the vicinity of the east abutment. The
located within relatively thin layers of bentonitic clay within instruments used for this study included SIs installed in
the shale, or along the till/shale interface, where the upper 1995 during investigations for the 1997 bridge construction,
shale has been sheared and softened by glacio-tectonic and VWs installed in 2014. Locations of these instruments
movement. Figure 3 and Figure 4 show shale core samples are shown on the stability figures in later sections.
collected from the shear zone during the 2014 drilling
program showing the highly disturbed condition of the
upper shale and slickensided shear surfaces from ongoing
landslide movement.
Primary movement at the Borden Bridge east
abutment has been measured between depths of 38 m and
 Figure 5 - Borehole and Instrument Locations

Piezometers and Groundwater Regime

The North Saskatchewan River valley acts as a regional groundwater drain; piezometric levels follow a characteristic
drawdown from the uplands towards the river. Nests of three piezometers were installed at each of the 500-series SIs in
2014: at the base of the borehole, within intact shale; within the shear zone, or at the shale/till contact; within sand layers
in the till, or within the upper drift. As shown in Figure 2, two distinct piezometric surfaces were identifiable: one acting on
the shear zone, and one within the upper drift. In 1995, the upper piezometric surface was observed at a depth of about
10 m; however, more recent measurements from the 2014 VWs indicated the piezometric surface had risen to a depth of
about 6 m. Potentiometric contours showing the heads acting on the shear zone are shown in Figure 6.
 Figure 7 - Movement direction and magnitude from SI's

Bathymetry

In 2014, a new bathymetric survey was conducted and

revealed an area beneath the bridges near the landslide
toe where 2 m of scour had occurred (Figure 8). The scour
Figure 6 - Piezometric surface at the shear zone likely formed during the high river flows in 2010 – 2012
which resulted from extreme precipitation events.
Identification of the scour and assessment of backfilling
Slope Inclinometers options during remedial design was an integral part of the
2019 analysis.
SIs were typically installed at least 3 m into intact shale and
grouted in place, with a maximum depth in the range of
60 m. Since construction of the third bridge in 1997,
landslide displacements were relatively minor: 0.8 mm to
2 mm per year. From 2010 to 2012, movements in the
range of 30 mm were measured. Since 2014, the landslide
movement has been negligible; recent SI data shows minor
drift with no discernable movement trend.
Movement was measured in both the A and B
directions, with the average resultant movement oriented
about 15o clockwise from the bridge centreline. Some SIs
were sheared off in the period of accelerated movement
between 2010 and 2012, indicating lateral displacement of
50 mm or more, which would correspond with shear strains
in the range of 10%, indicating that at least a portion of the
shear zone had been sheared to its residual state. Figure
7 shows movement directions in the SIs, with the length of
arrows corresponding to magnitude of measured
displacement.
Figure 8 - Scour location

3 EVOLUTION OF STABILITY MODELS: 1995 TO

2019

Slope stability analysis for this site has gone through

various iterations since the 1985 construction; the most
recent are discussed herein, with a summary of back
calculated strength parameters for shale summarized in
Table 1. Slight variations in the strength parameters in till
and drift did not result in significant changes to the FoS
compared to variation in shale strength, so are therefore
not discussed in this paper.
 unknown lower global FoS. By designing to a minimum
Table 1. Summary of Shale Strength Parameters from Back Analysis FoS of 1.30 for local and global stability, this risk was
Date Shear Zone Disturbed Shale Method reasonably mitigated. This approach has been supported
’ c’ ’ c’ by regional practice by the authors where acceptable long-
term performance of remediated slopes on river valley
1995 6.5 2 - - 2D LEM
landslides founded in shale has been achieved at a target
2013 5.4 0 8.7 4 Optimized 2D LEM, FoS of 1.30.
refined direction
2019 6.0 0 8.7 4 2D LEM with
updated bathymetry 2D Back Analysis
5.4 0 6.5 0 3D LEM
The 2019 2D back analysis was completed using Slope/W
software and the Morgenstern-Price method with a half-
A 2D Limit Equilibrium Method (LEM) stability analysis
sine interslice force function. Initial strength parameters
was completed in 1995 for the new bridge construction and
were adopted from the 2013 analysis and back calculated
was based on a topographic survey of the above-river
to achieve a FoS of 1.00 while using the elevated
valley wall, and data from the “SI” series instruments. The
porewater pressure data from 2014. Figure 9 below shows
critical section was located through the bridge abutments,
the 2D back analysis that corresponds with Section A-A’ in
parallel to the bridges centerline.
Figure 2.
The 1995 analysis was revisited in 2013, with two key
differences: introduction of the resultant landslide
orientation of 15o measured clockwise from the bridge
centerline, and slip surface optimization. The 2013 analysis
resulted in refined strength parameters and showed that
the FoS varied significantly even for closely spaced parallel
cross sections in the direction of measured movement,
which prompted further site investigation and
instrumentation in the summer of 2014. The intent of the
field program was to replace damaged instruments and
provide more complete stratigraphy, piezometric and
movement data. The 2D and 3D stability analysis
conducted in 2019 used all available data from 1995
onward, which included new bathymetric survey data
collected in the summer of 2014. Figure 9 - 2D Back analysis
Direct shear testing of cores recovered during the 2014
drilling resulted in anomalously high strength values; this 3D Back Analysis
data was not used. Similar to the 2D back analysis, initial parameters for the
As can be seen in Figure 12 through Figure 14, the 3D model were adopted from the 2013 work, with the
global (riverbank) and local (bridge abutment) FoS varied model being calibrated to a FoS of 1.0, using the
significantly. Landslide displacements were greatest near movement direction calculated from the SIs and
the centroid, with the magnitude decreasing towards the piezometric conditions interpreted from the 2014 VWs. The
edges of the landslide mass where edge effects are upper and lower piezometric surfaces were related to the
greater, Figure 7. This effect is also indicated in Figure 11 upper, disturbed shale and drift (alluvial deposits and till),
through Figure 14 with the local FoS typically being higher and the shear zone, respectively. SVSlope 3D Professional
than the global FoS. 2009 software was used with the Morgenstern Price
The risk of overestimating the FoS based on a local method and half-sine intercolumn force function. The
assessment of the bridge abutment area using 2D methods model domain was separated into 2500 columns (50 by 50)
was identified in 2013. Large river valley landslides seated to achieve a reasonable balance between result accuracy
in shale commonly exhibit creep-style behavior, with and calculation time. Figure 10 below shows the 3D back
ongoing displacements occurring despite an apparent FoS analysis. Figure 11 shows the landslide extents estimated
greater than unity, sometimes up to a FoS as high as 1.2 by the 3D back analysis with the critical section A-A’, along
(Henry, 2014). A 3D LEM analysis provides a more realistic with an additional local section that was considered for the
FoS. A thorough assessment of these landslides is difficult bridge abutments. In subsequent figures, the landslide
as the progressive nature of these failures mean the strains footprint predicted from section A-A’ was kept constant to
along the shear plane are variable, and the residual state compare the effectiveness of each remedial option shown
exists only along a portion of the shear plane (Jaspar and in red as Riverbank landslide extents.
Peters, 1979). A more thorough assessment may have
been possible using finite element methods; however, the
lack of accurate up-to-date SI data precluded this as a
viable option.
The ability to consider local and global FoS in 3D
addressed the risk of potentially assigning a site FoS based
on a 2D assessment of the abutment area with a potentially
 40 m. To increase the local stability (bridge abutment
section) FoS to 1.25, over 3,000 – 800 mm diameter stone
columns at a 1 m spacing would be required, while the FoS
for the larger riverbank landslide would be 1.12. The stone
column option was discounted early in the initial analysis
due to the expected difficult construction conditions and
high cost. Figure 12 shows the stability analysis results for
stone columns.

Figure 10 - 3D Back analysis

Figure 12 - Stone columns option

Structural Reinforcement
Structural reinforcement options including tiebacks, driven
H-piles with and without tiebacks, and a retaining wall with
deadman anchors were considered. It was not possible to
achieve a FoS greater than 1.11 using structural options
Figure 11 - Landslide extents from 3D back analysis combined with commonly available materials. The
requirement for detailed analysis of deadman and tieback
The landslide was estimated to be approximately 210 m in performance in shale, monitoring and maintenance for the
length from headscarp to toe, and about 400 m wide. The deadman and tiebacks, the required 40 m pile length and
critical Section A-A’ was located about 45 m north of the the low FoS precluded the use of structural reinforcement
westbound bridge, skewed at an angle of 15° clockwise options at this site.
from the bridge centreline. The predicted toe of the slide
mass extended about 100 m into the river. The elevation Earthworks
difference from the riverbed to headscarp was about 23 m. Earthworks based remedial options that were considered
included berming, a subcut along the riverbank crest,
backfilling of the riverbed scour, and lightweight fill at the
4 ASSESSMENT OF REMEDIAL OPTIONS abutments.
Lightweight fill replacement of the highway
The target FoS for remediation was 1.30, which was embankment approach was deemed a potential alternative
consistent with the MoH standard at the time of project using 2D analysis resulting in an estimated FoS of 1.28.
award in 2013. Constructability, traffic impact, However, given the relatively large extents of the landslide
environmental impact and construction zone safety were footprint predicted in the 3D model, compared to the
key considerations in the final evaluation of feasible relatively small area of the highway embankment, the
remedial options. lightweight fill option resulted in a negligible increase to the
global (riverbank) FoS (FoS = 1.02) in the 3D analysis.
Stone Columns Significant impact to traffic operations would also be
required to construct this option; therefore, lightweight fill
Stone columns have successfully been used for river valley was not considered a feasible alternative.
landslide remediation by MoH so were considered for As mentioned previously and shown in Figure 8, the
Borden. Stone columns were modeled as an equivalent 2014 bathymetric survey identified previously undetected
shear key extending 5 m below the shear zone, which scour in the region of the most easterly in-river piers at both
would have required penetration depths in the range of bridges. Combining scour backfill, a toe berm, and a subcut
along the riverbank crest resulted in an acceptable global refined to achieve a local and global FoS of at least 1.30
and local FoS. Analysis showed that scour backfill would while avoiding impact to highway operations, utilities
produce a positive impact to the FoS, particularly for the running along the south edge of the subcut and rare plants
local stability of the bridge abutments, where the FoS was near the north end of the subcut. The remedial design is
increased by about 20 percent. Dimensions of the berm shown in Figure 15. The potential effect to river hydraulics
and subcut were adjusted to obtain a reasonable footprint was considered during remedial design development. 2D
and height while maintaining local stability of those HEC-RAS models were used to assess the impact to river
components while scour backfill dimensions were adjusted flows as a result of permanent berm construction as well as
to minimize footprint and provide scour protection around short – term cofferdam construction for placement of the
the piers while maintaining an acceptable FoS. Figure 13 scour backfill. No significant changes to river flows, water
and Figure 14 show the initial remedial concept and draft elevation or scour conditions were predicted.
remedial design, respectively, and highlights the
optimization of the scour backfill.

Figure 15 - Remedial design drawing

Figure 13 - Initial scour backfill assessment

A sensitivity analysis was completed for optimization of the
berm height. As shown in Figure 16, the increase to FoS
with respect to berm height was most significant up to a
height of 5 m. Stability checks for local FoS of the berm
face showed decreased stability beyond a 5 m height,
hence a 5 m berm height was selected.

Figure 14 – Partially refined remedial design

5 RECOMMENDED REMEDIAL DESIGN Figure 16 - Berm height sensitivity analysis

The remedial design consisted of: scour backfill 2 m deep;
a toe berm 5 m high; and a riverbank crest subcut 6 m to
7 m deep. The geometry of the remedial components was
6 CONSTRUCTION CONSIDERATIONS Jaspar, J. L., & Peters, N. (1979). Foundation performance
of Gardiner dam. Canadian Geotechnical
Construction was initially scheduled for the winter of Journal, 16(4), 758-788.
2020/2021; however, the Department of Fisheries and
Oceans gave notice in early 2021 that full project Kelly, A. J., Sauer, E. K., Barbour, S. L., Christiansen, E.
authorization would be necessary, requiring re-scheduling A., & Widger, R. A. (1995). Deformation of the Deer
to allow for preparation of offsetting plans and revisions to Creek bridge by an active landslide in clay
drawings and the construction tender package. Remedial shale. Canadian Geotechnical Journal, 32(4), 701-724.
construction is tentatively scheduled for the fall/winter of
2021. Panesar, H., Kelly, A., Clifton, W., & Bushman, R. (2014).
Slopes in the Highly Plastic Clay Shales of the
Canadian Prairies – A Case History of the Three
7 CONCLUSION Borden Bridges. CGS Annual Conference GeoRegina
2014.
Key impacts affecting the stability of the east abutments of
the Borden Bridges were the presence of a scour hole in Sauer, E. K. (1983). The Denholm landslide,
the riverbed identified by 2014 bathymetric surveys and the Saskatchewan. Part II: analysis. Canadian
rise in porewater pressures on the deep-seated shear Geotechnical Journal, 20(2), 208-220.
plane on the ancient landslide. Special considerations in
the stability analyses were impacts on both the local
stability of the abutments and the global stability of the
valley wall; the selection of critical section and its
orientation; and, variation of porewater pressures at the
shear zone. Variations in FoS during 2D LEM modeling for
relatively small adjustments in section location and
orientation indicated that 3D modeling was required to
provide an understanding of the landslide behaviour.
Development of an appropriate remedial design for this
complex large-scale landslide was completed using 3D
LEM analysis to simultaneously assess local and global
stability. Consideration of a multitude of remedial options
led to the recommended composite remedial design
including backfilling of the scour hole, a toe berm, and
unloading of the riverbank crest.
The geometry of the remedial design was refined to
minimize construction impact while achieving a minimum
global and local FoS of 1.30. It was possible to limit the
remedial footprint to beyond the highway right-of-way,
thereby avoiding disruption to traffic and impact to public
safety.

8 REFERENCES

Christiansen, E. A. (1968). Pleistocene stratigraphy of the

Saskatoon area, Saskatchewan, Canada. Canadian
Journal of Earth Sciences, 5(5), 1167-1173.

Christiansen, E. A. (1970). Physical Environment of

Saskatoon, Canada.

Eckel, B. F., Sauer, E. K., & Christiansen, E. A. (1987). The

Petrofka landslide, Saskatchewan. Canadian
Geotechnical Journal, 24(1), 81-99.

Henry, J. (2014). A Novel Framework for the Analysis of

Low Factor of Safety Slopes in the Highly Plastic Clays
of the Canadian Prairies (Doctoral dissertation,
University of Saskatchewan).
 Mitigation Response on Hwy 570 Landslide
along the Red Deer River Valley in Alberta.
Rodriguez, Jorge; Grapel, Chris, Duxbury, Jason
Klohn Crippen Berger, Edmonton, Alberta, Canada
Macciotta, Renato
Department of Civil and Environmental Engineering – University of
Alberta, Edmonton, Alberta, Canada
Penney, Tony; Skirrow, Roger; Tappenden, Kristen
Ministry of Transportation and Economic Corridors, Edmonton, Alberta, Canada

ABSTRACT
Assessing the potential impact of a landslide to a transportation corridor requires an understanding of the kinematics and
triggers of movement. Klohn Crippen Berger Ltd. (KCB) is tasked with providing geohazard assessment services as part
of the Geohazard Risk Management Program (GRMP) for Alberta Transportation and Economic Corridors (TEC). On May
31, 2022, KCB and TEC conducted an inspection of a deep-seated landslide on the north side of Hwy 570:01 in central
Alberta. The observed landslide features included a series of backscarps at mid-slope and open ground cracks in the upper
section of the slope, and highly weathered and fractured rock blocks. This paper details the characteristics and
mechanisms of the landslide, as well as the initial assessment methodology, recommendations and mitigation strategies
adopted.

RÉSUMÉ
L'évaluation du risque de glissement de terrain à côté des corridors de transport nécessite de développer une
compréhension de la cinématique et des déclencheurs de mouvement. Klohn Crippen Berger Ltd. (KCB) a été chargé de
fournir des services d'évaluation des Géorisques dans le cadre du Geohazard Risk Management Program (GRMP) pour
la région centrale par le Minister of Transportation and Economic Corridors (TEC) de l'Alberta. Le 31 mai 2022, KCB et
TEC ont effectué une inspection où un glissement de terrain profond a été récemment identifié du côté nord de Hwy 570:01.
Les caractéristiques observées dans le glissement de terrain comprenaient une série de escarpe arrières à mi-pente et
des fissures de sol ouvertes dans la partie supérieure de la pente, et des blocs de roche fortement altérés et fracturés. Cet
article détaille les caractéristiques et les mécanismes de cette nouvelle étude de cas de glissement de terrain, ainsi que
les méthodes pour fournir une évaluation initiale et une recommandation des stratégies d'atténuation adoptées.

1 INTRODUCTION resulted in periodic traffic disruptions due to the deposition

of rockfalls, debris flows, and frozen blocks of soil in the
Transportation corridors in Alberta traverse regions with winter (Rodriguez et al. 2020). New research lead by the
varying geological characteristics. In the East Central University of Alberta, KCB, and TEC is starting to provide
Plains of Alberta, the provincial highway network crosses quantified evidence of the relationship between slope
the badlands regions. The badlands slopes along the Red failures along the RDR valley and climatic factors and
Deer River (RDR) valley are characterized by having little weather events (Mirhadi et al. 2021).
to no vegetation cover and are highly susceptible to erosion On May 31, 2022, KCB and TEC conducted an
and weathering (Roustaei et al. 2020). This is an arid to inspection where a deep-seated landslide had been
semi-arid region with temperatures ranging from -18.6 °C recently identified on the north side of Highway 570:01
in the winter and 26.7 °C in the summer, and precipitation (Figure 1). The features observed in the landslide included
averaging 370 mm of combined snow and rainfall per year a series of backscarps at mid-slope and open ground
(Government of Canada 2023). A provincial-scale cracks in the upper section of the slope. Importantly, highly
landslide susceptibility model by Pawley et al. (2016) weathered and fractured rock blocks were observed. The
shows that the slopes along the RDR valley are precarious geometry of the rock blocks required an
categorized as highly susceptible to landslides. A immediate response to avoid the release of large rockfalls
landslides inventory review from TEC shows 13 active sites onto the highway.
within the RDR valley that have historically impacted the Due to the unknown history of the site and the
highway infrastructure due to ongoing weathering precarious conditions of the rock blocks, KCB conducted a
processes, landslide activity, or a combination of both geotechnical investigation that involved collecting historical
(Government of Alberta 2023). air photos and past LiDAR data, and conducting two
Active landslides in the area are subject to annual unmanned aerial vehicle (UAV) surveys on June 02 and
inspections, and on occasion, multiple interventions June 10, 2022, to assess the historical and current activity
throughout the year to keep the highways open. An of the landslide. The UAV surveys were used to complete
example case study previously documented and published photogrammetry reconstruction of the ground surface and
is the C018 site, in which multiple failure mechanisms have perform a change detection analysis with the new and

Classification: Protected A
 a) b)

Bird Nest

85 m

Figure 1 Panoramic view of the Hwy 570 landslide and the site location (a), photo taken on June 2, 2022. Location of the
landslide (b).

historical point cloud data. The analysis provides a first higher amounts of traffic can occur during the summer
estimate of the level of activity of the landslide and rock tourist season.
blocks. The slope height is 110 m vertical from the highway to
After the initial assessment and within a short time span the crest, and the active zone is 85 m vertical from the
(29 days after the geohazard was initially observed), the highway to the upper scarp. The width of the active zone is
scaling and disposal of three rock blocks and one rock slab approximately 250 m at the highway. The slope angle at
that posed a potential rockfall hazard to the section of the the active zone can be discretized in three sections: the
highway traversing the base of the slope was completed. highway back slope is 29° from the toe of the slope to 35 m
As part of the execution plan an environmental overview above the highway; at the middle section of the slope, the
assessment and site surveys were undertaken to limit angle reduces to 20° from 35 m to 67 m above the
liability from potential spread of regulated and invasive highway; and from 67 m to 85 m above the highway, the
weeds and the destruction of protected plants or wildlife angle is 9°. Also, in the middle section there is a 31 m
while removing the rockfall hazard. After the completion of vertical rock scarp with a 52° angle on the west side of the
the scaling activities UAV surveys were completed on landslide. The slope is bounded by two gullies that are
September 15, and October 28, 2022. 20 m apart at the top and 264 m apart at the toe, with a
This paper presents the characteristics of the Hwy 570 slope angle of approximately 30° (see Figure 1).
landslide, the methods used for the initial assessment, and
the mitigation strategies completed by TEC to reduce the 2.1 Geologic Context and Weathering Susceptibility
risks to the traveling public.
The surficial geology and bedrock stratigraphy have been
identified using Alberta Geological Survey geological
2 HIGHWAY 570 LANDSLIDE maps. The bedrock units at the site are part of the
Horseshoe Canyon formation and Bearpaw formation
Highway 570 is a paved, two-lane highway orientated in the (Prior et al., 2013). These formations are generally
northwest-southeast direction near the hamlet of Dorothy, composed of horizontally bedded layers of sandstone,
in south-central Alberta. Most of the slope and the upland siltstone, bentonite, mudstone, and coal seams. The
area of the site is on private land, located in Special Area surficial geology at the site consists of colluvial deposits
No 2 (Figure 1). Part of the land where the landslide is that may contain bedrock, till, glaciolacustrine, glaciofluvial,
located is managed by a ranch and the Government of and eolian sediments that are generally poorly sorted
Alberta under the Special Areas Act. The highway was (Fenton et al., 2013).
constructed over a former railway line that was part of the The weathering processes associated with the geology
Drumheller Railway, which was built in 1911 and at this valley and the presence of dispersive soils produce
abandoned in 1983 (FJHS, 2022). the development of rill erosion, deep erosion gullies,
The 2021 Average Annual Daily Traffic along Hwy 570 sinkholes, and weak runoff marks with pipe or tunnel
below the landslide site was 410 and 510 vehicles per day erosion features (Roustaei et al. 2020). These weathering
going east and west, respectively (AT 2021). However, processes are intensified primarily by precipitation events,

Classification: Protected A
 water seepage, ice accumulation and snow melt that can KCB’s project team deployed UAV photogrammetry
result in differential disintegration of the rock mass. surveys on June 02, June 10, 2022, to expediently assess
The failed material accumulated at the toe of the slope the level of activity of the site, which also helped all parties
includes fine-grained soils in two talus fans, a portion of the involved to visually assess the deformations and critical
slope that appeared to have bulged out towards the areas on the generally unvegetated slope (Table 1).
highway in a manner consistent with deep-seated sliding, Following the rock scaling activities, follow-up UAV surveys
and an accumulation of cobble- to boulder-sized angular were completed on September 15, and October 28, 2022,
particles at various locations across the base of the slope. by KCB.
The length of the zone of accumulation of failed materials The four UAV surveys at the Hwy 570 landslide were
was estimated to be approximately 120 m along the completed using a DJI Mavic 2 Pro with a Hasselblad L1D-
highway. 20c camera. Each UAV survey at the site required taking a
series of oblique photographs of the ground surface. In
2.2 Historical Air Photos addition, ground control points (GCP) and scale references
were placed on the ground to refine the photogrammetry
A review of historical air photos obtained from Alberta processing. The photos from the 2022 UAV surveys were
Environment and Parks between 1950 (Figure 2a) and taken between 50 m and 110 m above the crest of the
2005 (Figure 2g) indicates that a railway was in place along landslide. The take-off point for the UAV in the June 02
the toe of the slope in 1950 (Figure 2a). Also, the photos survey was from the highway as no access permit had
indicate that the highway was constructed sometime been granted to the crest of the slope at the time by the
between 1981 (Figure 2d) and 1986 (Figure 2e). The scale private landowner. The take-off point for June 10,
of the air photos does not permit a determination of September 15, and October 28, was from the upper plain
whether highway construction involved excavating the toe above the crest of the slope to cover the active zone, the
of the natural valley slope at the site; however, the slope side slopes and gullies, and the upland area. The general
deformation preceded the highway construction. The characteristics of each 2022 UAV survey are shown in
earliest evidence of cracking and slope deformation in the Table 1.
upper part of the slope is from 1963 (Figure 2b). Evidence
of multiple tension cracks on the slope surface is observed Table 1 Characteristics of photos in the UAV surveys
in the air photo from 1975 (Figure 2c). 2022.
A discussion with one of TEC’s retired Maintenance
Contract Inspectors (MCI), indicated that the "pad" of fill Quantity Photo
UAV Photo GSD1
across the highway was a former ridge that was used as a Survey
of
Angle2
Overla
(cm/pixel)
borrow area for the railway embankment and construction Photos p
of the highway. The railway initially constructed a double Jun. 02 1284 80° 80% 2.91
cut through the ridge, and gradually fill borrowing reduced Jun. 10 1129 80° 80% 2.41
Sept. 15 1285 80° 80% 3.10
the ridge to its current level on the south side of the
Oct. 28 1003 80° 80% 2.40
highway. The air photos from 1963 (Figure 2b) and 1975
(Figure 2c) shows signs of landslide displacement from the
1
Ground Sampling Distance
2
Angle from horizontal.
development of tension cracks on the back slope, and the
air photos appear to show that failed material on the back
slope was removed and placed at the "pad". The surveys conducted on June 10, September 15, and
October 28 were completed using GCPs with eight to ten
targets located near the highway and near the crest of the
3 GROUND SURFACE INFORMATION slope. After the first two UAV surveys, the GPCs from
September 15 and October 28 were surveyed with
3.1 Historical Aerial LiDAR Sparkfun RTK GPS with a local level of precision of 0.02 m
to improve the precision of the change detection analysis.
Historical aerial LiDAR of the Hwy 570 landslide dated
October 28, 2007, was obtained to assess ground surface 3.2.1 Photogrammetry
information. The LiDAR data was collected by North West
Geomatics own by Hexagon AB. The data was collected The digital photogrammetric reconstruction was completed
using a Leica LiDAR sensor ALS50 with a scan rate of 30 for each survey using PIX4D Mapper Desktop (Pix4D,
hz. at approximately 1820 m above ground surface. The 2022). The reconstruction provided an XYZ point cloud
data included point spacing every 2.0 m with a horizontal data set for each survey (see Figure 3 and Figure 4) that
accuracy of ± 50 cm and vertical accuracy of ± 30 cm. was used for the change detection analysis. The
reconstruction for the September 15 and October 28
3.2 UAV survey surveys was corrected with the surveyed GCPs. The
average root mean square error (RMSE) of the GPC for
High-quality LiDAR survey data can be considered more northing, easting and elevation coordinates is shown in
accurate, especially for vegetated areas, than UAV survey Table 2.
data. However, UAV surveys using photogrammetry are The geolocation and scale of the June surveys were
currently less expensive and faster to deploy, even in areas corrected using the Sept. 15 survey with GCPs and known
with difficult access, or where vegetation is negligible. coordinates to improve the precision and accuracy of the

Classification: Protected A
 survey. The corrections were applied by matching multiple point cloud from the September 15 survey. The
points of known stable zones in the point cloud from replacement section was made to complete the gaps in the
June 02 and 10 with respect to the September 15 survey point cloud after reprocessing to improve the accuracy in
using CloudCompare (LGPL, 2022). Also, a section of the the active zone.
June 2 point cloud in the upper scarp was replaced with the

Classification: Protected A
 Table 2 Average Root Mean Square Error of the GPC for The level of detection was assessed from a cloud-to-
Northing, Easting and Elevation Coordinates cloud comparison, using CloudCompare, of the areas
outside the landslide area. The lower level of detection was
Easting Northing Elevation ±0.4 m for the June surveys due to lack of referenced
UAV Survey
RMSE1 (m) RMSE1 (m) RME1 (m) GCPs, as shown in Table 3.
Jun. 02 - - -
Jun. 10 - - - Table 3 Level of Detection based on cloud to cloud
Sept. 15 0.008 0.004 0.02 comparison
Oct. 28 0.02 0.02 0.08
1 The RMSE of the GCPs is a local indicator of the difference Analysis Mean Standard Deviation
between the initial and the computed position of the GCPs after A 0.13 0.12
reconstruction and indicates the precision of the point cloud fitted B 0.39 0.27
to the coordinates of the measured GCPs (Pix4D, 2022). C 0.1 0.1

4 LANDSLIDE AND ROCKFALL HAZARD Analysis “A” provided a quantitative estimation of the
ASSESSMENT displacement at the Hwy 570 landslide in the last 15 years
(see Figure 3). The result indicates that the extent of the
A combination of site observation, historical air photos landslide has not change significantly in the last 15 years;
review and change detection analysis were used to assess however, mid-slope the results provide signs of a typical
the level of activity of the landslide and the rock blocks. slide failure, where the upper portion has a downslope
The historic tension cracks well above the brow of the displacement of approximately 3.0 m (i.e., material loss),
slope appear to indicate that the slope is failing due to and the toe of the slope has a bulged displacement or
deep-seated sliding, most likely on a weak plane in the accumulation of failed material of up to 4.0 m.
bedrock. The depth of the failure plane is unknown but Analysis “B” provided a quantitative measure of the
appears to be horizontal based on the cracking and changes after the rock scaling activity (see Figure 4a), as
displacement pattern observed near the crest of the slope, described in Section 5. In the areas where no rock scaling
and the toe of the slope. Also, no signs of distress were activity occurred there were no significant displacements
observed on the pavement, which could indicate that the within a detectable change of ± 0.4 m.
toe of the failure plane could be just below the highway Analysis “C” provided a quantitative measure of the
level or just above the highway and obscured by talus. The short-term (i.e., 43 days) erosion and slide process (see
weathering susceptibility of the material slope has left Figure 4b). The analysis after the rock scaling activity
some blocks of bedrock mid slope that presented a rockfall shows that no significant displacements were occurring in
hazard to the public travelling on Hwy 570. Disaggregated most areas of the Hwy 570 landslide. Nevertheless, there
material and rock blocks that fell previously were observed are small areas (less than 5 m2) where rock scaling
in the ditch and across the highway during the inspection. activities occurred that continue to erode into the back
Although the ditch has intercepted most rockfall, the slope of Highway 570.
accumulated material had filled up most of the catchment The results from the three change detection analyses
ditch at the toe of the slope, impeding proper drainage and provide an indication of the level of activity of the
ponding water near the toe of the slope. Highway 570 landslide. However, the results did show
significant noise at the edge of the active zone, such as
4.1 Change Detection Analysis areas with displacement (e.g., material loss) which are
expected to be stable, as shown in Figure 4a. Two factors
As part of the preliminary assessment of the Hwy 570 contributed to the noise in the results from the UAV surveys
landslide, change detection analyses were completed in June 2022: the absence of surveyed GCPs and the
using the available point cloud data from the surveys. The difficulty of filtering out the dense vegetation in the gullies.
change detection was calculated using the M3C2 method
(Lague, 2013) found on CloudCompare (LGPL, 2022) to
assess the activity level of the landslide. The M3C2 method 5 RISK MITIGATION MEASURES
estimates the distance between the two-point clouds based
on the orientation of normal vectors and a discretization of The site mitigation measures included short- and long-term
the point cloud. The methodology used for the comparison actions to protect motorists and other users on Hwy 570.
follows a similar method implemented at the C018 site on These measures were based on the site inspection and
Highway 837 (Rodriguez et al. 2020). The three change initial landslide assessment. As part of short-term
detection analyses presented in this paper are: mitigation measures, TEC immediately installed a
temporary protection barrier, between the highway
• Analysis A: LiDAR point cloud from October 28, 2007, shoulder and the backslope, and placed traffic warning
with respect to the October 28, 2022, UAV survey. signs at each end of the Highway 570 landslide. The barrier
• Analysis B: UAV survey point cloud from June 10, 2022, was meant to reduce the potential for displaced material to
with respect to the UAV survey from October 28, 2022.
• Analysis C: UAV survey point cloud from
September 15, 2022, with respect to the UAV survey
from October 28, 2022.

Classification: Protected A
 a) Oblique View b) Plan View

Hwy 570

Material Loss Material Gain

-5.0 -4.0 -3.0 0 3.0 4.0 5.0 100 m
-2.0 -1.0 1.0 2.0
Landslide area
Figure 3 Ground surface changes after 15 years, change detection results from analysis “A”, a) results shown in oblique
view, b) results shown in plan view.

a) June 2022 Vs. October 2022 b) September 2022 Vs. October 2022

Material Loss Material Gain

-1.2 -0.8 -0.6 -0.4 -0.2 0.2 0.4 0.6 100 m
-1.0 0 0.8 1.0 1.2
Landslide area
Figure 4 Ground surface changes in 2022. a) results from before rock scaling activities, analysis “B”; and b) results after
rock scaling activities, analysis “C”.

reach the highway. Shortly after, on June 28 and June 29, 5.1 Rock Scaling
2022, TEC completed the removal of the identified unstable
rock blocks and vertical rock slab (Figure 1). In terms of The project execution rock scaling work at the
long-term mitigation measures, due to the limited Highway 570 landslide included the innovative use of a
information of the landslide activity the site has been “spider” excavator. This unique equipment had minimal
included in the TEC GRMP as a new monitoring site in the disturbance to the site, when compared to alternative
Central Region, and will be inspected on an annual basis methods.
or after a significant precipitation event. The spider excavator used in the rock scaling activities
was an adaptable and efficient piece of equipment that
reached the rock blocks by “walking down” from the top of

Classification: Protected A
 the slope using four extension legs. The access path at the removal work. As an example of the well-orchestrated and
site had areas with grades of approximately 40% to 100%, expedited regulatory process, the AEPA authorization was
which were too steep for traditional excavation equipment digitally provided.
to traverse without causing significant disturbance to the The expedited authorization of the rock scaling also
site and risking operator safety (see Figure 5a). The spider exposed a gap in the MBCA regulatory permitting. ECCC
equipment included two all-terrain paws at the front of the has a permit that can be used in a situation where migratory
excavator that accommodated the roughness of the terrain birds and or nests have created a hazard to human health.
and provided stable footing while accessing the site and However, as was discovered during the authorization of the
removing the rock blocks. Additionally, a roped-access Hwy 570 landslide, a permit did not exist for the
crew accustomed to working on steep slopes disturbance of a nest where the risk to human health is not
supplemented the project team and provided inspection the birds or nests themselves, but the substrate on which
services. Due to the size of the site, communication the birds are nesting. In the case of the Highway 570
between the excavator, scaling team and base was landslide, the substrate included the rock blocks and bluffs
accomplished with radios. The implementation of a spider that were at risk of falling onto a highway. A risk to human
excavator at the Highway 570 landslide demonstrated the health was imminent and the rock blocks needed to be
benefits of this versatile equipment for working on sites with removed. Since the birds or the nests were not the primary
limited and/or difficult access (see Figure 5b). Future risk to human health, the MBCA permit did not apply and
implementation could help TEC with mitigating other ECCC could not offer any other permit to fit the situation.
challenging rockfall hazards that could threaten highway ECCC noted the gap in permitting and worked
safety and efficiency, such as areas in the Rocky Mountain collaboratively with KCB to quickly develop a stop-gap
and Rocky Mountain Foothills Regions of Alberta. method for the project team to demonstrate due diligence.
A week after the successful permitting of the
5.2 Regulatory requirements Highway 570 landslide, a meeting between ECCC and the
Mining Association of Canada (MAC) was held to discuss
During environmental overview assessment and site the changes related to the Modernization of the MBCA
surveys to assess wildlife and vegetation at the Regulations. At the meeting, KCB noted the absence of
Highway 570 landslide, two Violet-green Swallow nests such a permit. It was noted by ECCC and recognized as a
were identified on the back slope (see Figure 1). Although gap in the federal permitting. Since rock faces adjacent to
the rock scaling work was initially planned for areas away roadways are commonplace in mining operations, MAC
from the location of two nests, a significant tension crack requested that a permit applicable to the situation faced by
behind a rock slab located next to one of the bird nests on
the west side of the landslide was identified on June 28, a)
2022. KCB engaged Environment Canada and Climate
Change (ECCC), and Alberta Environment to maintain
transparency and to seek guidance on how to best proceed
with the construction work while remaining in compliance
with legislation. On June 29, 2022, a discussion between
KCB and ECCC led to the conclusion that ECCC did not
have a federal Migratory Birds Convention Act (MBCA)
nest disturbance permit for a situation where human health
was at risk due to non-wildlife concerns (i.e., rockfall
hazards). No permit was available to be issued so KCB
immediately worked with ECCC to develop an innovative
due diligence plan for potential loss of active nests and b)
approval to proceed with the work was granted. An
authorization to remove the nests was also granted by
Alberta Environment and Protected Areas (AEPA) under
Section 36 (2) subsection (c) of the Wildlife Act.
The extremely short timeframe required to obtain
regulatory authorizations for the rock scaling activities as
well as complete a design, procure a contractor, and
complete the work was extraordinary. Normal permitting
processes would have required waiting months until
nesting season was over. Normal design and contractor
procurement timelines would also have been measured in
months. A key to expediting the regulatory process was the
consultant’s ability to maintain open communication and
properly inform regulators of the risk to human safety. KCB
Figure 5 Rock scaling work with spider excavator at the
was able to provide an evaluation of those risks against the
potential environmental effects of the proposed mitigations. Hwy 570 landslide on June 28, 2022. a) photos of the
Fully informed, the federal and provincial regulators were access route through the east side gully used by the
able to make timely decisions and authorize the rock spider excavator; b) photo of the start of scaling of the
rock blocks located on the east side.

Classification: Protected A
 the Highway 570 landslide be added to the permitting Also, the author acknowledges the contribution by:
options. ECCC is currently reviewing the recommendation. Bison Historical Services Ltd. a KCB sub-consultant
regarding regulatory requirements applications; Emcon
Services, the Prime Contractor; BAT construction Ltd. and
6 CONCLUSIONS Spidex All Terrain Excavating Inc the sub-contractors for
the rock scaling work.; and sub-contractor ATS Traffic that
The use of UAVs continues to increase in engineering provided temporary traffic accommodation during the rock
practice to assess geohazards quickly and easily where the scaling activities.
risks are unquantified. The initial identification of the
Highway 570 landslide was made from ground 8 REFERENCE
observations where a wire fence along the ditch of the
highway was covered by displaced material from the slope. Alberta Transportation (AT). 2020. Traffic Counts
However, photos and videos from an immediate UAV flight Reference No. 116220 and No. 997198. Retrieved July
revealed the presence of highly weathered and fractured 05, 2022, from:
rock blocks at the crest of the back slope which posed a http://www.transportation.alberta.ca/mapping/ .
threat to highway safety. Also, it allowed for the first-time Fenton, M.M., Waters, E.J., Pawley, S.M., Atkinson, N.,
observation of the extent of the landslide area and the Utting, D.J. and Mckay, K. 2013. Surficial geology of
upper scarp. Alberta; Alberta Energy Regulator, AER/AGS Map 601,
Determining the current state of activity of the landslide scale 1:1 000 000.
required a review of historical information that helped Forth Junction Heritage Society (FJHS). 2022. "Canadian
assess the time and locations of ground deformation at the Northern Railway in Central Alberta". Access July 14,
site. The implementation of UAVs to survey the site using 2022. https://forthjunction.ca/canadian-northern-
photogrammetric methods provided a quick, simple, and railway.htm
relatively low-cost method to acquire change-detection Government of Canada. 2023. Historical Weather Data
information. The relatively high accuracy and resolution of from Drumheller. http://climate.weather.gc.ca/.
the data from the UAV surveys helped to quantify and Accessed April 15, 2023.
locate the ground deformations in a short period of time. Government of Alberta. 2023. Geotechnical and Erosion
The UAV change detection surveys allowed for an Control, Landslide Assessment, Reg-Central Region.
assessment of the long-term ground deformation from the Ahttp://www.transportation.alberta.ca/PlanningTools/G
previous 15 years when compared to the 2007 LiDAR data. MS/Annual%20Landslides%20Assessments/.
UAV surveys were also used to assess the changes from Accessed April 15, 2023.
the rock scaling work when comparing a survey before and Lague, D., Brodu, N., Leroux, J. 2013. Accurate 3D
after the rock blocks were removed. Future UAV surveys comparison of complex topography with terrestrial laser
to quantify the magnitude of displacements and the scanner: application to the Rangitikei canyon (N-Z).
location of critical zones at the Highway 570 landslide ISPRS Journal of Photogrammetry and Remote
would help to assess level of activity and enhance long- Sensing, Volume 82, 2013, Pages 10-26,ISSN 0924-
term mitigation actions that reduce the risk to highway 2716, https://doi.org/10.1016/j.isprsjprs.2013.04.009
safety and efficiency. LGPL. 2022. CloudCompare V2 [GNU GPL software]
The use of a spider excavator allowed the work to avoid (2.12).
much of the breeding bird habitat at the site with its ability Mirhadi, N., Macciotta, R., Grapel, C., Skirrow, R.,
to move stepwise into and across the landslide area. While Tappenden, K. 2022. “Antecedent weather signatures
the ECCC and Wildlife Act authorizations to remove the for various landslide failure modes at a 60-m-high rock
two bird nests were granted, the methods used in the slope near Drumheller, AB. Conference of
rockfall mitigation resulted in no harm to the second nest. Geohazards 8.
The geohazard was successfully mitigated in 29 days from Pix4D S.A. 2022. Pix4Dmapper Desktop (4.8.0).
the time that the threat to public was identified. This https://www.pix4d.com/
timeframe includes initial response and set up of hazard Prior, G.J., Hathway, B., Glombick, P.M., Pană, D.I.,
mitigations, engineering assessments, environmental Banks, C.J., Hay, D.C., Schneider, C.L., Grobe, M.,
studies and permitting, contractor procurement, and Elgr, R. and Weiss, J.A. 2013: Bedrock geology of
construction work, including extra, “just-in-time permitting” Alberta; Alberta Energy Regulator, AER/AGS Map 600,
associated with nest removal during construction. scale 1:1 000 000.
Rodriguez, J., Macciotta, R., Hendry, M. T., Roustaei, M.,
Gräpel, C., & Skirrow, R. 2020. UAVs for monitoring,
7 ACKNOWLEDMENTS investigation, and mitigation design of a rock slope with
multiple failure mechanisms—a case study. Landslides
This work was made possible by Alberta Transportation Journal. https://doi.org/10.1007/s10346-020-01416-4
and Economic Corridors. The authors would like to Roustaei, M., Macciotta, R., Hendry, M. T., Rodriguez, J.,
acknowledge the Highway Geohazard Research Program Grapel, C., & Skirrow, R. 2020. Characterisation of a
between the University of Alberta, Klohn Crippen Berger rock slope showing three weather-dominated failure
Ltd., and Alberta Transportation and Economic Corridors modes. 427–438.
for proving additional resources and review of the work https://doi.org/10.36487/acg_repo/2025_24
presented in this paper.

Classification: Protected A
 Monday, October 2, 2023

STABILIZATION AND
REINFORCMENT
SOLUTIONS FOR THE REINFORCED FILL
OVER A VOID PROBLEM CONSIDERING A
NEW GEOSYNTHETIC STIFFNESS LIMIT
STATE
F.M. Naftchali & R.J. Bathurst
GeoEngineering Centre at Queen’s-RMC, Kingston, Canada

ABSTRACT
Several closed-form analytical methods for the design of a geosynthetic-reinforced fill spanning a void appear in national
standards and codes in the UK, France, Germany and in the Eurocode. In this paper, the reinforced void problem is
examined using the UK BS 8006-1 (2010) method. A simple hyperbolic stiffness model for reinforcement geosynthetics
that is applicable to low load levels, is used to model the constant-load isochronous creep-strain behaviour of these
materials for fills under operational (serviceability) conditions. Vertical deformation, reinforcement strain, reinforcement
strength, and reinforcement stiffness limit states are identified in the paper with special focus on the stiffness limit state
which has not been identified in design codes as a separate limit state. This limit state can control design outcomes as
demonstrated in the paper.

RÉSUMÉ
Plusieurs méthodes analytiques de forme fermée pour la conception d’un remplissage renforcé de géosynthétiques
couvrant un vide apparaissent dans les normes et codes nationaux au Royaume-Uni, en France, en Allemagne et dans
l’Eurocode. Dans cet article, le problème du vide renforcé est examiné à l’aide de la méthode UK BS 8006-1 (2010). Un
modèle simple de rigidité hyperbolique pour les géosynthétiques de renforcement applicable aux faibles niveaux de charge
est utilisé pour modéliser le comportement de fluage-déformation isochrone à charge constante de ces matériaux pour les
remplissages dans des conditions opérationnelles (de service). Les états limites de déformation verticale, de déformation
d’armature, de résistance d’armature et de rigidité d’armature sont identifiés dans le document en mettant l’accent sur
l’état limite de rigidité qui n’a pas été identifié dans les codes de conception comme un état limite distinct. Cet état limite
peut contrôler les résultats de conception, comme démontré dans l’article.

1 INTRODUCTION noted earlier and is amenable to implementation in

spreadsheets.
Voids may form under thin fills and embankments due to
karstic terrain, mining operations, permafrost thawing, pipe 1.1 General problem
failure or local subsidence at the surface of a landfill
closure. A strategy that can be used in anticipation of void Figure 1 introduces the general problem. A reinforcement
formation is to introduce a geosynthetic reinforcement layer geosynthetic (typically a geogrid) is placed at or close to
at the base of the fill at the time of construction that can the base of a fill of height H. The fill is assumed to be a
support the fill over the void. There are currently four frictional soil with unit weight  and friction angle . The
analytical design methods for the problem of a void forming geosynthetic layer serves to prevent the fill from collapsing
below a geosynthetic reinforced fill. A notable shortcoming
of each method is the omission of a geosynthetic stiffness
limit state to ensure that the reinforcement has adequate
stiffness to carry the maximum tensile load in the
reinforcement at the maximum predicted strain. The design
of these systems is further complicated by the load-strain-
time (creep) properties of candidate geosynthetic sheet
reinforcement materials. The paper introduces a two-
component hyperbolic isochronous secant stiffness model
that has proven useful to capture the rate-dependent
behaviour of these materials under low operational strains.
The basic features of the stiffness model are explained in
the paper. A generic design flow chart is presented that
includes all four design limit states for the void problem
(vertical deformation of the reinforcement above the centre
of the void, reinforcement strain, tensile strength and the
new reinforcement stiffness limit state). The design flow
chart can be used with any of the four design methods Figure 1. Reinforced thin fill over a circular or long-shape
void (diameter D or width b, respectively).
into a void which develops after construction of the fill. The Assuming that d = ds×Ds/b and the geometry in Figure 1,
void may have circular or long–void shape with diameter or leads to:
width Ds at the surface. The diameter of the void or width
of a long-void at the elevation of the geosynthetic is D and
2
b, respectively. In the BS8006-1 (2010) method used in the 8d 
example to follow, the void slope angle d = .  max =   [3]
3 b 
1.2 Limit states

The limit states for design of a reinforced fill over a void are: In order to focus on the mechanics of the general problem,
partial factors that appear in the BS 8006-1 (2010) method
1. Deformation serviceability limit state for the maximum are taken as 1 in this study corresponding to operational
allowable vertical deformation at the surface of the soil (serviceability) limit states. The mobilised tensile stiffness
layer above the void (i.e., ds < prescribed tolerable is computed as:
limit).
2. Reinforcement strain serviceability limit state to keep Jmob = Tmax/max [4]
the maximum strain (max) in the reinforcement within a
prescribed tolerable limit (allow) (i.e., max < allow) or, The focus of this paper is on this parameter which must not
equivalently, the value of d/b (or d/D) within a exceed the available stiffness using a candidate
prescribed limit. reinforcement material. This limit state has not been
3. Tensile strength limit state to ensure that the maximum considered in the prior related work with the exception of
computed tensile load (Tmax) in the reinforcement does recent work by writers in the journal literature (i.e., Naftchali
not exceed the allowable tensile capacity of the and Bathurst 2023; Bathurst and Naftchali 2023).
material.
4. Stiffness serviceability limit state to ensure that the
reinforcement has sufficient tensile stiffness to carry the 3 REINFORCEMENT STIFFNESS
maximum load in the reinforcement at the maximum
tensile strain. In order to examine the stiffness limit state it is necessary
to quantify geosynthetic stiffness in a consistent and
rigorous manner. This is a challenge because polymeric
2 ANALYTICAL MODEL (geosynthetic) reinforcement materials are rate-dependent
materials meaning that their tensile properties are load,
The methods of Giroud et al. (1990), BS 8006-1 (2010) and strain and time-dependent. To quantify this behaviour
EBGEO (2011) are the three most common analytical under monotonic increasing tensile loading the writers
methods to compute the maximum strain (max) and developed an isochronous two-component hyperbolic
maximum load Tmax in the reinforcement. The three model for the secant stiffness of the reinforcement J(, t)
methods vary in details but all are based on an assumed that can be expressed as:
geometry of the deflected reinforcement layer over the void
and simple force equilibrium. Details of the methods can be
1
found in the paper by Naftchali and Bathurst (2023). In this J(, t) =
paper, we use the BS 8006-1 (2010) method. 1 1 [5]
+ 
J o (t) (t)
2.1 Load, strain and stiffness limit states

The maximum tensile load in the reinforcement layer using Parameter, Jo (t) is the initial tangent stiffness of the
the method of BS 8006-1 (2010) for a long void is computed isochronous curve for time t, and (t) is a parameter that
assuming that the catenary shape of the deflected captures the curvature of the isochronous load-strain and
geosynthetic profile is usefully approximated by a parabola stiffness curves with strain. Figure 2 shows examples of
leading to: stiffness versus strain for three different products at three
different isochronous times. These plots show that the
1 stiffness of the reinforcement decreases with time and
Tmax = 0.5( H + q)b 1 + [1]
6 max strain as may be expected. As strain level increases, the
isochronous stiffness curves tend to converge.
For a given strain and isochronous time, these curves
The maximum strain for a long void is calculated using the can be used to compute the tensile load in the
following approximation: reinforcement as:
2 4
d   2H  T(,t) = J(,t)× [6]
8  s   b + 
Ds   tan d  [2]
 max =  The isochronous curves may be used to estimate the
3b 4 tensile load in the reinforcement shortly after void formation
(say t = 10 and 100 h) and at 1000 h after the reinforcement and are thus non-conservative (unsafe) for stiffness limit
and fill have come to equilibrium over the void. state design and should not be used.
Constant load creep tests (ASTM D5262-97) or Creep test data may not always be available. Bathurst
stepped-isothermal creep tests (ASTM D6992-03) are and Naftchali (2021) collected data from many tests and
required to compute the stiffness values seen in these plots developed useful linear relationships between isochronous
and to fit the model curves. Also shown in these figures are
the stiffness values inferred from in-air constant rate-of-
strain tests carried out at 10% stain/min in accordance with
ASTM D4595-17 or ASTM D6637/D6637M-15 methods.
These curves fall above the creep-adjusted stiffness values

a) t = 10 h

a) Uniaxial high density polyethylene (HDPE) geogrid

b) t = 100 h

b) Biaxial polypropylene (PP) geogrid

c) t = 1000 h

Figure 3. Isochronous secant stiffness values versus

c) Woven polyester (PET) geogrid ultimate tensile strength of reinforcement based on in-
air 10% strain/min tests (i.e., ASTM D4595-17; ASTM
Figure 2. Isochronous secant stiffness curves. D6637/D6637M-15).
 normalised deflection (d/b) at the reinforcement elevation.
The maximum strain is then computed from Equation 2 or
3. The computed strain (max) must not exceed a prescribed
(allowable) strain for the reinforcement (strain limit state)
so that the primary creep stage of the material is not
exceeded. A typical value is allow = 5%, but for PP geogrids
can be much larger.
Next, the maximum load in the reinforcement (Tmax) is
computed using Equation 1. For the same max value, Tmax
varies with the ratio of fill height to void width (H/b). The
maximum load in the reinforcement must not exceed the
maximum allowable strength of the reinforcement (T al)
where:

Tal = Tult/RF [7]

Parameter Tult is the ultimate tensile strength of the

reinforcement from constant rate-of-strain (CRS) tests
(e.g., ASTM D4595-17; ASTM D6637/D6637M-15) and RF
is a dimensionless factor that accounts for loss of strength
due to installation damage, creep and chemical durability
mechanisms. A total strength reduction factor of RF = 3 has
been used for polyester (PET) geogrids. However, this
value can vary in practice. For example, a value of RF = 5
may be appropriate for polyolefin geogrid materials that
tend to exhibit greater creep than PET geogrids of similar
strength and when all other conditions are the same. Tult
may also be divided by an additional overall (global) factor
of safety (Fs) at the discretion of the designer (Giroud et al.
1990).
The final limit state to be examined is the stiffness limit
state where:

Jmob ≤ Javail [8]

Figure 4. Design flow chart illustrating calculation steps
for a reinforced fill layer over a long void using BS 8006- Here Javail is the available isochronous secant stiffness for
1 (2010) method: a) deformation and reinforcement a candidate reinforcement product from plots of the type
strain limit states, b) reinforcement tensile strength limit shown in Figure 2, or using a linear approximation with Tult
state, and c) reinforcement stiffness limit state. Problem of the type shown in the plots of Figure 3.
parameters: b = 1 m, H = 0.5, 1, 1.5 and 3 m, q = 0,  =
18 kN/m3 and  = 35°.
5 EXAMPLE DESIGNS
stiffness values at different isochronous times and strain,
and ultimate tensile strength for different categories Figure 5 repeats the calculations for the previous example
(groups) of geosynthetic sheet reinforcement products. but extends the outcomes to a void width of b = 2 m).
Examples are shown in Figure 3. There are detectable Figure 5a demonstrates the strain limit state. The
differences in the slopes of these curves between products maximum strain must not exceed the reinforcement
and between isochronous times. Bathurst and Naftchali serviceability limit state strain. In this example we assume
(2021) showed that based on all available (pooled) data for that max = 5% controls; this strain corresponds to a
sheet reinforcement types, R = J/Tult = 4.5. For practical deflection ratio at the elevation of the geosynthetic of d/b =
preliminary design purposes a value of R = 5 is judged to 0.137 which is assumed to fall within project tolerances.
be convenient. Figure 5b shows computed maximum reinforcement
The influence of reinforcement stiffness on closed-form load (Tmax) versus strain () curves for the problem
solutions for MSE wall applications (i.e., reinforcement parameters shown in the figure caption. The right vertical
loads and wall displacements) can be found in the paper axis shows the minimum required ultimate tensile strength
by Bathurst and Naftchali (2023). (Tult) assuming that Tmax = Tal = Tult / RF and RF = 3.
Increasing the width of the void and the height of the
embankment results in an increase in the required Tult. For
4 DESIGN FLOW CHART example, the minimum required tensile strength for b = 2 m
and H/b = 3 is Tult = 674 kN/m, and Tult = 169 kN/m for b =
Figure 4 shows a design flow chart for the BS 8006-1 1 m and H/b = 3.
method. The design process starts with an assumed Figure 5c addresses the tensile stiffness limit state. The
 limit state. These values are greater than those required to
satisfy the stiffness limit state and thus the tensile strength
limit state would control design for this case.
Figure 6 summarises the minimum required ultimate
tensile strength to satisfy the stiffness limit state for
stiffness at 1000 h and 5% strain for group-specific
reinforcement types and using all pooled data, void widths,
b = 1 m and b = 2 m, and different H/b ratios. R values
correspond to different reinforcement types, three of which
have been shown in Figure 3. All values can be found in
the paper by Bathurst and Naftchali (2021). These values
are useful at the time of design when a project-specific
sheet reinforcement geosynthetic has not been decided but
the type of reinforcement has been specified. The pooled
data value R = 4.5 can be used when the type of sheet
reinforcement is not known at design time and for
preliminary design. The minimum Tult is computed for each
reinforcement type based on R values. The bar chart
shows that PP woven geotextiles (R = 2.4) require the
maximum Tult, whereas PET woven geotextiles (R = 5.0)
required the minimum Tult for all H/b cases. With the
exception of PP woven geotextile and PET woven and
knitted geogrid, the use of pooled data is judged to be
conservative (safe) for design for practical purposes.
Finally, if a specific product is proposed for a project
and creep-stiffness data are available, then a product-
specific design flow chart can be constructed (e.g., Figure
3) and the tensile strength and stiffness limit states
checked directly for that product.

6 CONCLUSIONS

This paper examines vertical deformation of the

Figure 5. Design flow chart for problem examples using
reinforcement, reinforcement strain, tensile strength, and
BS 8006-1 (2010) method: a) deformation and
stiffness limit states for the analysis and design of a
reinforcement strain limit states, b) reinforcement tensile
geosynthetic reinforced fill over a void under operational
strength limit state, and c) reinforcement stiffness limit
conditions. The general approach uses the BS 8006-1
state with Tult = J/4.5. Problem parameters: b = 1 and 2
(2010) method. A simple two-parameter hyperbolic
m, H = 0.5, 1, 1.5 and 3 m, q = 0,  = 18 kN/m3 and  =
isochronous load-strain model described by Bathurst and
35°.
Naftchali (2021) is used to characterise the stiffness of the
reinforcement as a function of elapsed time and strain.
available (isochronous) stiffness is equated to the ultimate
strength of the reinforcement and is computed as J = R ×
Tult where R = 4.5 is an empirical coefficient described Table 1. Ultimate tensile strength to satisfy tensile
earlier for pooled data for isochronous stiffness at 1000 strength and stiffness limit states for long-shape void
hours and 5% strain. This allows Tult to be conveniently with b = 1 and 2 m using BS 8006-1 (2010) method.
used to check that the tensile stiffness limit state is
satisfied. Other values of R corresponding to group-specific Width of void, b (m)
data and different isochronous times shown in Figure 3,
could also be used. Figure 5c shows that as the width of 1 2
the void and the height of the embankment increase, both
the mobilised stiffness and the minimum required tensile
strength to meet the stiffness limit state also increase. H/b Strength Stiffness Strength Stiffness
Table 1 summarises the results of analysis for both limit state limit state limit state limit state
strength and stiffness limit states. This table and Figures
5b and 5c show that for all fill heights, the stiffness limit Tult
state controls (i.e., required Tult for the strength limit state Tult (kN/m) Tult (kN/m) Tult (kN/m)
(kN/m)
is greater than the required Tult for the stiffness limit state).
However, this is not always the case. For example, 0.5 28 42 112 166
selecting RF = 5 would lead to different designs to satisfy 1 56 83 225 333
the tensile strength limit state for all H/b ratios; specifically, 1.5 84 125 337 499
larger Tult values would be required for the tensile strength 3 169 250 674 998
 4. The paper demonstrates conditions when the stiffness
of the reinforcement will control design and when the
strength of the reinforcement will control.

ACKNOWLEDGMENT

Financial support for this study was provided by the Natural

Sciences and Engineering Research Council (NSERC) of
Canada through Grant Number RGPIN-2018-04076 held
by the second author.

REFERENCES

ASTM D4595-17. Standard Test Method for Tensile

a) void width b = 1 m Properties of Geotextiles by the Wide-Width Strip
Method. ASTM International, West Conshohocken, PA,
USA.
ASTM D5262-97. Standard Test Method for Evaluating the
Unconfined Tension Creep Behavior of Geosynthetics.
ASTM International, West Conshohocken, PA, USA.
ASTM D6637/D6637M-15. Standard Test Method for
Determining Tensile Properties of Geogrids by the
Single or Multi-Rib Tensile Method. ASTM
International, West Conshohocken, PA, USA.
ASTM D6992-03. Standard Test Method for Accelerated
Tensile Creep and Creep-Rupture of Geosynthetic
Materials Based on Time-Temperature Superposition
Using the Stepped Isothermal Method. ASTM
International, West Conshohocken, PA, USA.
Bathurst, R.J. and Naftchali, F.M. 2021. Geosynthetic
reinforcement stiffness for analytical and numerical
modelling of reinforced soil structures. Geotextiles and
b) void width b = 2 m Geomembranes, 49(4): 921-940.
Bathurst, R.J. and Naftchali, F.M. 2023. Influence of
Figure 6. Ultimate tensile strength to satisfy stiffness uncertainty in geosynthetic stiffness on deterministic
limit state for long-shape void with a) b = 1, and b) b = 2 and probabilistic analyses using analytical solutions for
m using BS 8006-1 (2010) method and different group- three reinforced soil problems. Geotextiles and
specific reinforcement types and pooled data from all Geomembranes, 51(1): 117-130.
reinforcement types corresponding to stiffness at 1000 BS 8006-1. 2010. Code of practice for
h and 5% strain. strengthened/reinforced soils and other fills
(+A1:2016). British Standards Institution (BSI). UK
The important points from this study are: EBGEO. 2011. Recommendations for design and analysis
of earth structures using geosynthetic reinforcements.
1. Convenient design flow charts can be constructed for ISBN 978-3-433-60093-1.
all four limit states including the stiffness limit state. Giroud, J.P., Bonaparte, R., Beech, J.F. and Gross, B.A.
1990. Design of soil layer-geosynthetic systems
2. The stiffness limit state can be conveniently linked to overlying voids. Geotextiles and Geomembranes, 9(1):
the ultimate strength of the reinforcement using 11-50.
product-specific test data (if available) or default values Naftchali, F.M. and Bathurst, R.J. 2023. Influence of
taken from a large database compiled by the authors geosynthetic stiffness on analytical solutions for
(Bathurst and Naftchali 2021). reinforced fill over void. Geosynthetics International,
30(1): 95-107.
3. The choice of coefficient R that equates reinforcement
stiffness to the ultimate strength of the reinforcement,
and the reduction factor RF applied to the
reinforcement strength will decide which limit state
controls design.
Incremental permanent deformation from
repeated heavy vehicle loads on
embankment dams: Field measurements
Maxime Blanchette, J.P. Bilodeau & Erdrick Perez-Gonzalez,
Department of Civil and Water Engineering, Université Laval,
Quebec, QC, Canada

Valérie Fréchette & Steven Doré-Richard

Hydro-Québec, Montréal, QC, Canada

ABSTRACT
Embankment dams are not typically designed for heavy vehicle traffic on their crests. Nevertheless, the demand for such
use is growing, particularly from electricity producers, mining, and forestry industries. Establishing traffic lane on the dam
crests would ensure accessibility to the structures and surrounding territories. Progressive damage associated with vehicle
traffic is likely to affect dyke and dam performance and therefore raises certain concerns among dam owners and
operators. They could benefit from tools and experimental data to improve understanding and characterize the dam's
mechanical response when subjected to passing of heavy vehicles. This tool aims to offer guidance in making informed
decisions regarding heavy vehicle traffic on a dam’s crest, given uncertainties associated with these unconsidered
solicitations. A dyke located in eastern Quebec has been instrumented to register stress states and deformations caused
by a heavy vehicle simulator. This paper presents the evolution of permanent deformation and stress states of an
embankment dyke core subjected to a high number of heavy load cycles. Field measurements allow quantifying the
magnitude of damage and impact of load repetitions on the core of dams.

RÉSUMÉ
Les barrages en remblai ne sont généralement pas conçus pour la circulation de véhicules lourds sur leurs crêtes.
Néanmoins, la demande pour une telle utilisation est croissante, en particulier de la part des producteurs d'électricité, ainsi
que des industries minières et forestières. L'établissement d'une voie de circulation sur la crête des barrages permettrait
d'assurer l'accessibilité aux structures et aux territoires environnants. Les dommages progressifs liés à la circulation des
véhicules sont également susceptibles d'affecter la performance des digues et des barrages et suscitent donc certaines
inquiétudes chez les gestionnaires et les opérateurs. Ils pourraient bénéficier d'outils et de données expérimentales pour
mieux comprendre et caractériser la réponse mécanique du barrage lorsqu'il est soumis au passage de véhicules lourds.
Cela pourrait aider à prendre des décisions éclairées sur l'autorisation ou la restriction de la circulation des véhicules
lourds sur la crête du barrage, compte tenu des incertitudes associées à ces sollicitations inconsidérées. Une digue
localisée dans l’est du Québec a été instrumentée pour enregistrer les états de contrainte et les déformations provoquées
par un simulateur de véhicule lourd. Cet article présente l'évolution de la déformation permanente et des états de
contraintes du noyau d'une digue en remblai soumise à un grand nombre de cycles de charges lourdes. Les mesures sur
le terrain permettent de quantifier l'ampleur des dommages et l'impact des répétitions de charges sur le noyau des
barrages.

1 INTRODUCTION dams since they were not designed specifically for

repeated heavy-load traffic. Dam owners face technical
Hydroelectric developments and exploitation of mineral challenges to provide authorizations and permits for the
and forestry resources require adequate and efficient passage of heavy vehicles on the dam’s crest given
access to surrounding territories. As a result, heavy and uncertainties related to the effect of the applied stresses.
non-standard vehicle traffic is increasing in the areas Therefore, new tools are needed to facilitate decision-
where dams and dykes are located. Using dam crests as a making based on the impact of such traffic on dam crests.
means of transportation infrastructure for heavy vehicles This paper aims to quantify the magnitude of
can result in significant economic and environmental permanent damage caused by increased heavy traffic
benefits, including reduced travel times and costs, reduced loads on embankment dams based on experimental field
fuel consumption, and avoided bridge construction. In data. A dyke located in eastern Québec has been
addition, Hydro-Québec is well engaged in a rehabilitation instrumented to register stress states and deformations
phase of some of its power generation facilities, which will caused by a heavy vehicle simulator.
require the use of non-standard vehicles on access roads This article will begin with preliminary concepts that will
to the dam crest. provide context on the current state of the art. The
However, there is concern about the impact of heavy methodology section will outline parameters of the field
vehicles on the safety and effectiveness of embankment
measurement experiment. Results of the experiment will 𝑝𝑐
is strain rate differentiation in a post-compaction state (𝜀̇𝑝 )
then be presented and discussed. and in a shakedown state (𝜀̇𝑝𝑠𝑠 ). The differentiation of the
former allows describing an initial phase with a high
2 PRELIMINARY CONCEPTS deformation development during the first repeated load
applications, and the latter allows describing the
The behavior of unbound granular materials and soils used permanent deformation behavior after a high number of
in dam construction is reasonably well-understood from a load applications.
soil mechanics perspective for static loading conditions, Despite the above, there are still gaps in the
even in northern climates subject to freeze-thaw cycles and understanding of how embankment dam cores respond to
frost heave. Deformations resulting from static loading are transient stresses caused by heavy and non-standard
caused by volumetric changes, lateral spreading, and vehicles. In addition, new damage laws may be needed to
shear displacements and can lead to some cracking in the assess how the resulting loads and deformation may affect
dam's core. Cracks in the core material can lead to the the mechanical and hydraulic properties of the dam, and
development of internal erosion, by the formation of empty possibly make it more susceptible to other events (i.e.
spaces, which can create preferential pathways for water seismic events, internal erosion, etc.). In this context, to
to flow through the dam or its foundation (Talebi et al., improve knowledge and develop management tools, this
2013). Therefore, it is important to monitor the stress and study presents the results of research on the mechanical
strain states in an embankment dam’s core to ensure its response and behavior of an experimental dyke subjected
proper performance, and consequently the dam’s safety. to accelerated traffic tests.
The core of an embankment dam plays a crucial role in
water retention and the associated reservoir due to its 3 METHODOLOGY
ability to resist seepage (Fell et al., 2005). Its primary
function is to ensure the dam’s imperviousness and control This section presents the structure of the dyke used for the
seepage through the dam, thereby ensuring the stability of experimental field measurements, the instrumentation
the downstream slope and preventing internal erosion. In utilized to record the mechanical response, as well as the
the province of Quebec, the materials typically used to experimental test protocol that was followed.
construct the core are clayey soils or glacial till.
Material properties such as gradation, mineralogy, 3.1 Structure of the dyke
particle shape and arrangement, moisture content, density,
and stress distribution within the embankment and Figure 1 shows the structure of the dyke used to conduct
foundation play an important role in controlling the experimental field measurements. To select the section
deformations. Furthermore, stress transfers (arching) from where the test would be conducted, criteria were
soft to hard zones throughout the structure influence the established due to the variable composition/geometry of
vertical settlement. The occurrence of excessive the structure along its length. One criterion was to choose
differential settlement can result in internal deformation of a linear area with enough width to accommodate the
the impervious zone and lead crack formation in the core. simulator and allow for maneuvering space on the sides.
Conversely, uniform settlement can decrease freeboard, The area should also provide enough space for
which is the distance between the water level and the top acceleration and braking of other vehicles (tests with real
of the dam (Chugh, 1990). Although horizontal deformation loaded heavy trucks). Additionally, the location should be
is typically small compared to vertical settlement, its where the dyke is almost at full thickness/height, as there
proportion depends on the geometry, embankment zones, are areas where the dyke gradually decreases in thickness.
and material properties. If the vertical settlement is within The structure of this dyke can be divided into several zones
an acceptable range, it is generally assumed that the with different materials having various properties and
horizontal displacement will also be acceptable. functions (Fell et al., 2005). Zone 1 of an embankment dam
However, when analyzing behavior of unbound is the core. Its main purpose is to provide impermeability
granular materials and soils used in dam construction and stability to the structure. In the dyke evaluated for this
under dynamic loads, such as those transmitted by heavy study, the core is a compacted mixture of clay and sand.
vehicle traffic, there are variations and unique conditions to The core is the main area of concern for studying the
consider. Recent research have used a mechanistic- potential damage caused by heavy vehicle traffic on the
empirical approach to predict deformation caused by dyke crest, as discussed by (Pérez-González et al., 2021).
vehicles passing on embankment dams (Pérez-González Zone 2 is composed of sand and gravel for shoulder and
et al., 2021, 2022). In this, stress states calculated from a transitions. Zone 2 functions as a protective filter for Zone
linear-elastic model are used as inputs to an empirical 1. In addition to providing filtration, zone 2 also contributes
model calibrated in the laboratory, in order to predict to the overall stability of the dyke and serves as a transition
deformation rates associated with vehicle magnitude and element between Zone 1 and Zone 3. Zone 4 is a riprap
velocity. and functions as a protective layer against wave and ice
Pérez-González et al., (2021) used the principles erosion on the upper upstream face of the dyke.
described by the shakedown theory for the deformation Zone 3 is divided into three subzones of different
analysis, and concepts such as the depth of influence (𝐷 ̃ ),
materials: Zone 3A, which corresponds to the riprap
defined as the depth measured from the top of the core cushion whose maximum particle size is 225 mm and is
where the material will be susceptible to develop plastic used for transitions; Zone 3B used as upstream revetment
deformation, are introduced. Another highlighted concept of the dam and comprises a particle material with a
maximum size of 760 mm; Zone 3C used as downstream their range are up to 100 and 50 kPa, respectively. In zone
revetment, comprising particles with a maximum size of 3, one vertical 200 kPa capacity SPT is positioned. Figure
380 mm. In the remainder of this article, when referring to 2a and 2b show the position of the strain gauges and the
Zone 3, reference is made to Zone 3A. SPTs in the dyke. Table 1 and 2 presents the position of
In addition to the areas described above, a thin layer of the strain gauges and the SPTs in zone 1 and zone 3
MG-20 (Bureau de normalisation du Québec, 2002), a respectively.
material used for pavement unbound granular base,
recovers the top of the crest to provide a smoother surface
and reduce the potential for damage to the dyke structure
caused by vehicle traffic.
Figure 1 shows the geometry of the dyke evaluated
during this study. The red quadrilateral represents the
instrumentation island that was installed to monitor and
measure damage magnitude caused by heavy vehicle
traffic on the dyke’s crest.

Figure 2a. Cross-section Figure 2b. Top view of the dyke

of the dyke with the with the instrumentation
instrumentation.

Table 1. Position of the SPTs and strain gauges in zone

1 of the dyke
Figure 1. Structure of the dyke. Adapted from (Groupe
ABBDL, 1979)
Zone 1 Depth Position relative to the MDD
(core) Z (mm) X (mm) Y (mm)
3.2 Instrumentation of the dyke
SPT vertical -2000 450 -450
Instrumentation was positioned at the top of the dam’s core SPT horizontal -1900 480 370
to assess vertical and horizontal stresses and vertical Strain gauge -1900 -270 270
displacements along a referential axis at different depths in
zone 3A and zone 1 of the dyke. The multi-depth
deflectometer (MDD) is one of the devices used for Table 2. Position of the SPTs and strain gauges in zone
measuring the deformation of soil under load. It consists of 3 of the dyke
four circular tubes with circular plates at the end. The tubes
have different lengths and diameters, with the longest tube Zone 3 Depth Position relative to the MDD
having the smallest diameter and the other tubes having (surface)
Z (mm) X (mm) Y (mm)
increasing diameters so that each tube can fit inside the
SPT vertical -705 450 -450
other. The plates are located at different depths and the
ends of the rods are at the same elevation, allowing their Strain gauge -700 -270 270
relative displacements to be measured using displacement
sensors (LVDTs). The MDD can measure both elastic and 3.3 Experimental test protocol
plastic deformation of the soil between each pair of plates
and the distance between a pair of plates is 300 mm Figure 3 presents the heavy vehicle simulator that was
(Bordeleau et al., 2021). Figure 2a shows the position in used to conduct the experimental test on the dyke. This
the dyke of the four circular plates which allow to measure system loads the crest with a half-axle equipped with
the deformation in zone 1 and in zone 3. MDD 2 and MDD standard dual tires moving at 9 km/h over a test track of 6
1 are positioned at the top of the core of the embankment m long.
dyke (zone 1), at a depth from the surface of 1950 mm and
2250 mm respectively. MDD 4 and MDD 3 are positioned
in the middle of zone 3 at a depth of 575 mm and 875 mm
respectively.
Strain gauges were also used to measure the
deformations during the loading tests. These consist of a
high-precision sensors installed in a 150 mm long sealed
tube attached to two circular stainless-steel pads, one at
each end, used to record the elastic and plastic
deformations in the dam body. Soil pressure transducer
(SPT) are used to measure the variations of stresses
during the passage of heavy load. In zone 1, two SPTs are Figure 3. Heavy vehicle simulator on the crest of the
positioned to measure vertical and horizontal stresses, embankment dyke
 Table 3 presents the general protocol followed during
the full-scale testing using the heavy vehicle simulator.
Measurements of the mechanical response of the dyke
(stresses and strains) are taken periodically, with a higher
frequency of measurements at the beginning of the cycles
and a progressively lower frequency over time at each load
stage. The experimental test covers a range of loads up to Figure 5. Signal from the vertical SPT in zone 3 to measure
1.5 times the maximum legal load for a half single axle in the stress induced by the heavy vehicle simulator (5000 kg,
Quebec during the normal period, which is 5000 kg 2000 cycles)
(Ministère des Transports du Québec, 2013). A higher
number of load repetitions was selected for the load of Figure 6 shows the average vertical and horizontal
5000 kg, as it is considered the reference load. The test stresses measured in zone 3 (surface) and zone 1 (core)
covers a wide range of axle loads, from moderate to very as a function of the load applied.
heavy, and provides valuable data for dam owners to
understand the mechanical response and performance of
the dam associated with the passage of heavy vehicles.

Table 3. General loading sequence and protocol during full-

scale testing using a heavy vehicle simulator.

Condition Load applied (kg) 1 Number of load cycle

A 3500 5000
B 5000 10 000
C 6500 5000
D 7500 2500
1
half single axle load

4 RESULTS

This section presents the results that were obtained Figure 6. Average stresses measured as a function of the
following the field measurements. Stress and strain signals load applied for zones 3 and 1
from the sensors were recorded periodically at specific
number of cycles during the accelerated traffic tests. An Figure 7 shows the average elastic strain measured in
example of strain signals recorded is shown in figure 4. For zones 3 zone 1 as a function of the load applied.
the MDDs, the strain magnitude was calculated from the
relative displacement and the distance between the plates.
Plates 2 and 1 were used to record displacements in the
core, and plates 3 and 4, displacements in zone 3.

Figure 7. Average elastic strain measured as a function of

the load applied for zone 3 and zone 1
Figure 4. Signals from MDD 3 and the strain gauge in zone
3 to measure the deformation (5000 kg, 2000 cycles) Figures 8 and 9 show the cumulative permanent strain
as a function of the number of cycles and the load applied
An example of the stress signals recorded is shown in for zones 1 and 3 respectively.
figure 5.
 Table 4. Initial geostatic stress measured for zone 3 and 1

Zones Initial geostatic stress (kPa)

Vertical (zone 3) 19
Vertical (zone1) 45
Horizontal (zone 1) 23

Table 5 and 6 present the comparison between the

geostatic stress of the embankment dyke structure and the
average dynamic transient stress induced by each stage
load. The "% of increase" refers to the percentage of
additional stress added by the heavy vehicle simulator in
relation to the initial geostatic stress.

Table 5. Comparison between the geostatic stress and the

average vertical and horizontal stresses induced by the
Figure 8. Cumulative permanent strain in zone 1 as a heavy vehicle simulator for zone 1
function of the number of cycles and the half-axle load
applied
Cond. Load Vertical Horizontal
applied (Zone 1) (Zone 1)
(kg)
Induced % of Induced % of
stress increase stress increase
(kPa) (kPa)
A 3500 4 9 0.9 4
B 5000 7 16 1.3 6
C 6500 9 20 1.8 8
D 7500 10 22 2.1 9

Table 6. Comparison between the geostatic stress and the

average vertical stresses induced by the heavy vehicle
simulator for zone 3

Cond. Load Vertical

applied (Zone 3)
(kg)
Figure 9. Cumulative permanent strain in zone 3 as a Induced % of
stress increase
function of the number of cycles and the half-axle load (kPa)
applied
A 3500 9 47
5 DISCUSSION B 5000 12 63
C 6500 15 79
Conducting the field tests and collecting the data on the D 7500 18 95
mechanical response of the dyke subjected to heavy
vehicle traffic is necessary to quantify the magnitude of In zone 3, where the stress recorded are higher, the
damage and the impact of load repetitions on the core of average vertical stress measured for the higher load stage
dams. The data collected can be analyzed to assess the (7500 kg) is 18 kPa corresponding to 95% of the initial
level of damage caused by the load repetitions. geostatic pressure. In contrast, the core exhibits lower
stress levels during the higher load stage, with
5.1 Stress analysis measurements of 10 kPa (22% of the initial geostatic
pressure) for vertical stress and 2.1 kPa (9%) for horizontal
To quantify the magnitude of damage and impact of load stress.
repetitions on dams, it is possible to compare the stress
induced by the heavy load simulator with the initial 5.2 Strain analysis
geostatic stress of the embankment dyke. The initial
geostatic stress refers to the stress caused by the weight To quantify the impact of each individual load stage on the
of the overlying soil before any external loads are applied. accumulation of permanent strain, the time hardening
Table 4 presents the vertical geostatic stress measured for concept is among the most efficient method (Rahman &
zone 3 and the vertical and horizontal geostatic stress Erlingsson, 2014). A model considering the time hardening
measured for zone 1. concept has the following form:
 ip = a(Ni0+Nieq)b [1] Table 7. Parameters for the permanent deformation model
in zone 1 of the dyke
Where, ip (microstrains) is the permanent strain in the
stress path i, Ni0 is the number of cycles in the stress path Cond. Load applied (kg) a b Neq R2adj
i starting from zero, Nieq is an adjustment factor for a A 3500 1.385 0.367 0 0.72
number of cycles and a and b are regression parameters
B 5000 15.523 0.367 0 0.90
(Pérez-González et al., 2021).
Figures 10 and 11 present the application of the time C 6500 15.523 0.388 5003 0.93
hardening model to the cumulative permanent strain D 7500 15.523 0.410 7098 0.96
measured on the field.
Table 8. Parameters for the permanent deformation model
in zone 3 of the dyke

Cond. Load applied (kg) a b Neq R2adj

A 3500 0.131 0.690 0 0.95
B 5000 0.491 0.690 0 0.91
C 6500 0.491 0.763 4864 0.98
D 7500 0.491 0.787 7057 0.89

The modeling of the cumulative permanent strain

measured on the field allows quantifying the magnitude of
damage of each of the four load stages, by analyzing them
with the Shakedown theory. The Shakedown theory
provides a framework to describe the permanent
deformation behavior of a material under repeated loading.
According to this theory, the material response can be
Figure 10. Cumulative permanent strain in zone 1 as a
classified into three stages: plastic shakedown, plastic
function of the number of cycles and the load applied
creep, and incremental collapse. Plastic shakedown is
characterized by a plastic response for a finite number of
load repetitions, followed by complete resilience. Plastic
creep refers to a stable long-term response for permanent
strain rate and accumulation. Incremental collapse is
characterized by a continuous increase in permanent
strains with each load cycle and eventual failure
(Werkmeister et al., 2005).
To define the boundaries of these stages, triaxial
repetitive loading tests are typically used. The European
specifications use criteria based on the difference between
the strains at 5000 and 3000 loading cycles (𝜀p5000 – 𝜀p3000).
If the calculated difference is less than 45 microstrains, the
material typically behaves in the Plastic shakedown range
(range A). If the difference exceeds 400 microstrains, the
material typically behaves as Incremental collapse range
(range C). Plastic creep is between these two values
(range B) (CEN-European Committee for Standardization,
Figure 11. Cumulative permanent strain in zone 3 as a 2004). Tables 9 and 10 present the analysis of the
function of the number of cycles and the load applied cumulative permanent strain according to Shakedown
theory in zones 1 and 3 respectively. For each load stage,
Tables 7 and 8 present the results of the time hardening the soil material stays in a stable mechanical response for
model calibration. The parameters of the time-hardening permanent strain rate and accumulation.
model (Equation 1) were determined by minimizing
quadratic errors. To maintain consistency in the behavior
described by the time-hardening model, constraints were
imposed to ensure that the optimization process converges
to a point closer to a global minimum than to a local
minimum. For this, the values of a, b and Neq must be equal
to or greater than those defined by a previous stress level.
Table 9. Analysis of cumulative permanent strain in zone 1 Table 11. Regression parameters for the models of zones
of the dyke with Shakedown theory 1 and 3

Load Induced p5000-p3000 Zones Model equations R2adj

Cond applied stress (kPa) () Range Zone 1 y = 0.0001x +0.057x+8.35
2
0.98
(kg)
Zone 3 y = 0.0002x2+0.049x+3.7 0.99
A 3500 4 5 A
B 5000 7 60 B
Table 12 presents the critical induced stress for the soil
C 6500 9 76 B to stay in range A for both zones 1 and 3.
D 7500 10 96 B
Table 12. Critical induced stress by the heavy vehicle to
Table 10. Analysis of the cumulative permanent strain in stay in Plastic shakedown state (range A).
zone 3 of the dyke with Shakedown theory
Zones Critical induced stress (kPa)
Cond Load Induced p5000-p3000 Range Zone 1 6.3
applied stress (kPa) () Zone 3 11.1
(kg)
A 3500 9 14 A
The determination of critical induced stress is an
B 5000 12 52 B important parameter that can offer decision-making tools to
C 6500 15 105 B dam owners. With the data collected on this embankment
D 7500 18 132 B dyke and also others, numerical modeling will be used to
obtain the induced stress for different combinations of
It is also possible to model the transient stress induced embankment dam geometry and heavy truck load. These
by the heavy vehicle simulator as a function of the induced stress values will then be compared to the
permanent strain accumulated between cycle 5000 and corresponding critical induced stress, providing a useful
3000. This model can be useful in determining the critical tool for dam owners.
induced stress that causes the soil to transition from range
A to B. Figure 12 present the model developed for zones 1 5.3 Limitations of the study
and 3 as well as the transition between range A and B. Both
models were not extended to determine the critical induced For the field measurements and the strain analysis, no
stress before the soil transitioning from range B to C wheel wandering was used during the testing for safety
because the field measurements obtained were not close reasons. The use of wander systems would have ensured
enough to range C for the predictions to be reliable. that the wheel does not always go by the same path during
the load cycles, representing more precisely the normal
distribution of the paths taken by the heavy vehicle’s
wheels while circulating on the dam’s crest. The non-
utilization of wander can result in the soil gaining stiffness,
but also represents the worst-case loading scenario.
The Shakedown theory was originally developed for
laboratory experiments using triaxial loading conditions.
which may not necessarily represent the real-world
Plastic
Creep
conditions of a dam under repeated loading. However, the
work of Pérez-González et al., (2021) demonstrated the
Plastic
Shakedown
efficiency of this analysis tool and its applicability to
embankment dams engineering.

6 CONCLUSION

The objective of this paper was to determine the extent of

impact resulting from repeated heavy loads on
Figure 12. Model predicting the critical induced stress that embankment dams by presenting field-experiment data.
causes the soil to transition from range A to B Data have been collected to analyze the mechanical
response of two zones; the core of a dyke (zone 1) which
Table 11 presents the equation for the modeling of the is the main zone of interest because of its crucial functions
transient stress induced by the heavy vehicle simulator as and zone 3 which is closer to the surface. Stress and strain
a function of the permanent strain accumulated between have been quantified to evaluate the magnitude of damage
cycle 5000 and 3000. associated with heavy vehicles passing on the crest of
dams.
In zone 3, stress levels are higher with an average
vertical stress of 18 kPa measured for the highest load
stage (7500 kg), corresponding to 95% of the initial
geostatic pressure. In comparison, the core (zone 1) has Pérez-González, E., Bilodeau, J.-P., Doré, G., & Doré-
lower stress levels during the highest load stage, with Richard, S. (2021). Assessment of the permanent
measurements of 10 kPa (22% of the initial geostatic deformation at the earth core of a rockfill dam under
pressure) for vertical stress and 2.1 kPa (9%) for horizontal heavy vehicle loading. Canadian Geotechnical Journal,
stress. 58(2), 165–175. https://doi.org/10.1139/cgj-2019-0449
As for the strain, the material appeared to exhibit a Pérez-González, E., Bilodeau, J.-P., Doré-Richard, S., &
stable mechanical response for permanent strain rate and Fréchette, V. (2022). Embankment dams under
accumulation at each load stage and in both zones, in dynamic heavy traffic loading: Challenges and
accordance with the Shakedown theory. It has been perspective.
suggested that the critical induced stress for the material to Rahman, M. S., & Erlingsson, S. (2014). Permanent
stay in range A of the Shakedown theory is 6.3 kPa and deformation models of unbound granular materials: A
11.1 kPa for zones 1 and 3 respectively. comparative study based on multi-stage repeated-load
The field measurements and strain analysis conducted triaxial tests.
during the study had certain limitations. Firstly, no wander Talebi, M., Vahedifard, F., & Meehan, C. L. (2013). Effect
was used during testing for safety reasons, which could of geomechanical and geometrical factors on soil
have resulted in soil gaining stiffness and affecting the arching in zoned embankment dams. 1056–1065.
accuracy of the results obtained from strain analysis. Werkmeister, S., Dawson, A. R., & Wellner, F. (2005).
Secondly, the Shakedown theory was developed for Permanent deformation behaviour of granular
laboratory experiments using triaxial loading conditions. materials. Road Materials and Pavement Design, 6(1),
The use of this theory for real-world conditions of a dam 31–51.
under repeated loading is relatively new but has been
successfully applied in the past (Pérez-González et al.
2021).
Other analyses are underway. In addition to the heavy
vehicle simulator, field tests were also conducted with
loaded heavy trucks and a super heavy crane on various
embankment dykes. These tests will allow to quantify the
damage caused by other parameters such as the speed of
the vehicles and the effects of the climate and materials, in
addition to providing more data to quantify the damage and
magnitude caused by heavy vehicle traffic at the dam crest.

7 ACKNOWLEDGMENT

The authors would like to acknowledge Hydro-Quebec and

NSERC for their interest and their support in this research.

8 REFERENCES

Bordeleau, C.-A., Doré, G. 1956-, Côté, J. 1971-, &

Université Laval. Faculté des sciences et de génie.
(2021). Effet des charges lourdes sur les barrages en
remblai: Déformation permanente d’un noyau d’argile
[Université Laval]. WorldCat.org.
http://hdl.handle.net/20.500.11794/69366
Bureau de normalisation du Québec. (2002). NQ 2560‐
114/2002. Travaux de Génie Civil–Granulats.
CEN-European Committee for Standardization. (2004).
Cyclic load triaxial test for unbound mixtures. European
Standard.
Chugh, A. K. (1990). Evaluation Of Embankment Dam
Stability and Deformation.
Fell, R., MacGregor, P., Stapledon, D., Bell, G., & Foster,
M. (2005). Geotechnical Engineering of Dams. Taylor
& Francis. https://doi.org/10.1201/NOE0415364409
Groupe ABBDL. (1979). Rapport de conception digues et
barrage principal.
Ministère des Transports du Québec. (2013). Guide des
normes de charges et dimensions des véhicules
routiers.
Field experimental investigation of
geosynthetics-stabilized bases in Sturgeon
County, Alberta
Mian Huang , Jiming Liu & Cheng Lin
Department of Civil Engineering – University of Victoria, Victoria,
British Columbia, Canada
Sanat Pokharel, Taylor Dagenais & Arghya Chatterjee
Stratum Logics Inc., Acheson, Alberta, Canada

ABSTRACT
Geosynthetics have been widely used to enhance the performance of roadways in seasonal frost regions, but their
effectiveness against freeze-thaw induced deterioration in pavements is not well appreciated. This study introduces full-
scale experiments on the freeze-thaw (F-T) performance of geosynthetics-stabilized bases in a flexible pavement located
in Sturgeon County, Alberta. The test was composed of three sections with the base courses stabilized by geocomposites
and two types of novel polymeric alloy (NPA) geocells (100-mm Type-D and 150-mm Type-C), respectively. The pavement
sections were fully instrumented with earth pressure cells, moisture sensors, and thermal sensors. Plate loading tests and
trafficking tests were performed on each section before the ground was frozen in November 2022, and will be conducted
after each seasonal F-T cycle. This paper presents the first stage of the experimental study, focusing on the instrumentation
in the road sections and plate loading and trafficking tests. Some preliminary observations on the performance of the three
test sections are discussed.

RÉSUMÉ
Les géosynthétiques ont été largement utilisés pour améliorer les performances des routes dans les régions sujettes au
gel saisonnier, mais leur efficacité contre la détérioration due au gel-dégel dans les revêtements routiers n'est pas bien
comprise. Cette étude présente des expériences sur le terrain portant sur les performances de géosynthétiques renforçant
les bases dans un revêtement souple situé dans le comté de Sturgeon, en Alberta. Le test est composé de trois sections
dont les couches de base sont renforcées par des géocomposites et deux types de géocellules en alliage polymère
novateur (NPA) (Type-D de 100-mm et Type-C de 150-mm), respectivement. Les sections de chaussée ont été
entièrement instrumentées avec des cellules de pression des sols, des capteurs d'humidité et des capteurs thermiques
pour surveiller leurs réponses mécaniques et environnementales. Des essais de chargement sur plaque et des essais de
circulation ont été réalisés sur chaque section avant que le sol ne soit gelé en novembre 2022, et seront effectués après
chaque cycle saisonnier de gel-dégel. Cet article présente la première étape de l'étude expérimentale, en mettant l'accent
sur l'instrumentation dans les sections routières, ainsi que sur les essais de chargement sur plaque et de circulation.
Certaines leçons tirées de l'instrumentation et des observations préliminaires sur les performances des trois sections de
test sont discutées.

1 INTRODUCTION Geocells, a type of three-dimensional geosynthetics with

honeycomb-like interconnected cells, have been
In cold regions, freeze-thaw (F-T) cycles are the common demonstrated to be an effective solution to improving F-T
cause of roadway distress induced by the seasonal actions performance of roads at limited field trials (Edil et al., 2002;
of frost heave and thaw weakening. In the process of Henry et al., 2005; Pokharel et al., 2017). However, such a
freezing, the combined effects of the migration of water and benefit of geocell stabilization has not been quantified on
the formation of ice lenses in soils alter the physical full scale road sections. For the high stiffness/strength
properties and structure of soils-i.e., void ratio, particle Novel Polymetric Alloy (NPA) geocells, no measurable
bonding, and soil fabric (Qi et al., 2006; Viklander, 1998). benefits have been reported yet. This study aims to fill
The reversed thawing process further modifies soil these gaps by introducing a long-term full-scale
structure, resulting in the mechanical degradation of bases experimental program for the highly instrumented road
and subgrades, which is responsible for roadway sections in Alberta. This paper will share some preliminary
destabilization (Kawabata et al., 2016; Simonsen and results from this program.
Isacsson, 1999).
Various practical countermeasures have been applied 1.1 Background
to mitigate F-T induced damage and improve the durability
of roadways. These include the replacement of frost- The test road sections were part of reconstructed asphalt-
susceptible soils, installation of a drainage system, paved highway, located at the intersection of Township Rd
chemical stabilization of road bases with cement or lime, 544 and Range Rd 260 in Sturgeon County, Alberta, as
and physical stabilization (e.g., with geosynthetics). shown in Figure 1. The test sections included three
sections stabilized with three types of geosynthetics: Type- 2). Perform short-term tests, i.e., plate loading tests and
D NPA geocells, Type-C NPA geocells, and Combigrid® trafficking tests, before and after seasonal F-T cycles.
geocomposite. The geocomposite consists of geotextile 3). Monitor the long-term performance of the well-
welded within geogrid structure. Table 1 and 2 summarize instrumented pavement sections during seasonal freeze
the basic properties of geocells and geocomposites, and thaw cycles.
respectively. This paper presents the first stage of the full-scale
experimental investigation, introducing the instrumentation
in the pavement sections, and presenting the preliminary
results of plate loading tests and trafficking tests before the
first F-T season.

2 CONSTRUCTION AND INSTRUMENTATION

The construction started in June 2022 and ended in

November 2022. The instrumentation, field testing and
monitoring were implemented by the team of the University
of Victoria (UVic) and Stratum Logics Inc. Note that the
design of the test sections was undertaken by a third party,
without the involvement by UVic and Stratum Logics.

2.1 Pavement construction

Figure 1. The satellite map of the test sections A stretch of the old roadway was fully removed up to the
depth below the native subgrade, and then new soil
materials were filled back and compacted, followed by the
Table 1. Properties of NPA geocells reconstruction of a two-lane flexible pavement. The top-
down structure of the rehabilitated pavement consisted of:
Values (A) 90-mm hot mix asphalt (HMA) layer, (B) 300-mm or
Properties
Type-C Type-D 350-mm base courses, (C) 300-mm subbases, (D) 1000-
Tensile strength, kN/m 19 22 mm pit run backfill underlain by (E) native clayey subgrade.
Dynamic modulus, MPa Note the original design by a third party referred Layer (C)
+30°C 775 800 as subgrade preparation; this paper renames it as
+45°C 675 700 subbases as the material used was a gravel and clay
+60°C 525 650
Distance between seams, mm 330 330
mixed, which after compaction could achieve comparable
Open cell dimensions, mm × mm 245 × 210 245 × 210 stiffness as the base courses.
Figure 2 illustrates the process of pavement
construction in three sections. In the geocell-stabilized
Table 2. Properties of Combigrid® geocomposite sections, a layer of non-woven geotextile was placed on
the surface of the subbases, and both 100-mm Type-D and
150-mm Type-C geocell were added on the geotextiles,
Values
Properties then backfilled with 300 mm thick base courses. In the
Geogrid Geotextile Composite
geocomposite-stabilized section, the geocomposite layer
Tensile strength
30/30 7.5/11.0 - was directly placed at the interface of the subbases and
(md/cmd)1, kN/m
Grab tensile base courses, and the base courses in this section was 50
- - 0.55 mm thicker than the bases in geocell-stabilized sections.
strength (md), kN
Trapezoidal tear After the compaction of base courses, hot mixed asphalt
0.26
strength, kN (HMA) layer was placed.
Aperture size
32 × 32 - -
(md/cmd), mm × mm
Characteristic
- 90 -
opening size, µm
1
md is machine direction; cmd is cross machine direction.

1.2 Objectives

The objective was to evaluate and quantify the efficacy of

geocell reinforcement in improving the F-T performance of
base courses in a full-scale test. To achieve the objective,
the following tasks would be carried out.
1). Instrument various types of sensors in three (a)
sections of the flexible pavement.
 (b)

(a)

(c)
Figure 2. The pavement construction: (a) geocell-stabilized
base courses, (b) geocomposite-stabilized base courses,
and (c) the finished asphalt layer

2.2 Instrumentation

LPTPC12-S earth pressure cells (RST Instruments®),

Type-T thermocouples [Omega® (blue PVC insulation
peel) and Thermo-Kinetics® (brown Teflon insulation
(b)
peel)], and EC-5 moisture sensors (Meter Group®), as
Figure 3. Photographs of: (a) sensors and (b) weather
shown in Figure 3(a) were installed. A weather station
(made by Campbell Scientific®) was installed at the site, station and data acquisition system
which included a TE525M tipping rain gauge and a Type
109 thermistor to monitor the precipitation and air
All sensors were embedded in the south (or eastbound)
temperatures on site, respectively. A remote data
acquisition system (also manufactured by Campbell lane in each section [e.g., Figure 4(a)]. Horizontally, earth
Scientific®) was set up to view or download collected data pressure cells, thermocouples, and moisture sensors were
spaced at distances of 1200 mm and 500 mm to avoid
from the sensors and weather station in a computer. As
interference between each other while utilizing different
shown in Figure 3(b), the system consisted of a datalogger
(i.e., CR1000X with the extension module AM16/32B measurements at the same location. Figure 4 presents the
multiplexer), a remote communication system (i.e., a instrumentation plan for the Type-D geocell section, where
two earth pressure cells with different measurement
C2446 antenna and a CELL205 cellular module), and a
capacities were placed at the interface of subbases and
power supply unit (comprising a BP26 12-V rechargeable
battery and a SP50-L 50-W solar panel). The datalogger, base courses [Figure 4(a)] with the larger measurement
cellular module, and rechargeable battery were housed in capacity cell (2 MPa) located below the inner wheel path
and the smaller capacity one (i.e., 500 kPa) below the outer
a fiberglass case, which was enclosed in a wooden
wheel path. The two cells were spaced at a distance of
container.
2.22 m, which was equivalent to the common axle width of
a trailer in Alberta. Figure 4(c) and 4(d) respectively
presents the layout of six Type-T thermocouples and four
EC-5 moisture sensors vertically embedded in base
courses and subbases. Two thermocouples were placed
in the base courses and the other four in the subbases.
One moisture sensor was placed in the base courses
below the centerline of the south lane, while the remaining
three moisture sensors were placed in the subbases. It is
worth pointing out that the temperature and moisture
sensors in the subbases were installed by drilling a hole in
the subbases while those in the base courses were placed
during the construction of the base courses. The vertical
positions of the sensors were surveyed in reference to the
surface of the subbases so their depths with respect to the
asphalt surface could be determined. The sensor layout in
the Type-C geocell section and the geocomposite section
was almost identical. Each of them had one more moisture
sensor installed in the base course at the pavement edge
different than that in the Type-D geocell section.

North a b c

South Traffic direction Pavement edge

Earth pressure cell Center line
2 MPa 1200 mm 500 mm
Thermocouple
Inner wheel
Wheel line
Outer wheel
Moisture sensor
Earth pressure cell a' b' c'
500 kPa
(a)

(d)
Figure 4. Sketch of sensor location in Type-D geocell
section of: (a) top view, (b) earth pressure cells, (c)
thermocouples and (d) moisture sensors

3 FIELD TESTS

Plate loading tests and trafficking tests were conducted on

November 3 and 4, 2022 when the ambient temperatures
were 4℃ to -9℃. Soil temperature in 150-mm Type-C
geocell and geocomposite sections was measured before
sunrise on November 3, 2022. The results are presented
in Figure 5. It was noted that the thickness of base courses
(b)
at the 150-mm Type-C geocell and geocomposite sections
was approximately 0.3 and 0.35 m, respectively. The
temperature of the soil appeared to increase with the
increase in depth. They were greater than 2°C, which
indicated soil layers in ground were still unfrozen when the
field tests were performed, although the above-grade
temperatures were below zero degree Celsius.

Temperature (°C)
0 2 4 6 8 10 12 14
0.0
Geocomposite
Type-C geocell
0.2
Depth below base coursesurface (m)

0.4

0.6

0.8

(c) 1.0

1.2

1.4
Figure 5. Temperature profiles in 150-mm Type-C geocell
and geocomposite sections
3.1 Testing methodology

Plate loading tests (PLTs) were conducted in the three

sections to investigate the mechanical behavior. Three
types of PLTs were implemented in the field. Here, one of
the PLTs is described and results presented. This type of
PLT imposed one cycle of incremental loading and
unloading following the procedure outlined by Huang et al.
(2023) and Pokharel et al. (2010). The loading proceeded
by applying the incremental plate pressure of 200 kPa, 400
kPa, 800 kPa and 900 kPa at the loading stage, followed
by unloading with the decremental pressure of 450 kPa, (b)
225 kPa and 0 kPa. Figure 6(a) illustrates the setup of a Figure 6. Photographs of: (a) plate loading tests and (b)
PLT. The PLTs were carried out on the asphalt surface with trafficking tests
the testing spot located exactly over the 2 MPa earth
pressure cell. The distributed stress over the subbases
could be measured at each loading stage of the plate After the completion of the plate loading tests at each
loading tests. A thin leveling layer of sand was spread on section, trafficking tests were conducted to investigate the
the asphalt surface at the loading spot before the 300 mm dynamic responses of geosynthetic-stabilized pavement
diameter loading plate of was placed over it. The loads under traffic loading (Timm and Priest, 2004; White et al.,
were generated by a hydraulic jack with one end sitting on 2010). A 40-ton trailer with eight axles was run for two
the center of the loading plate and the other end acting passes in each section in the forward-backward direction,
against the bottom frame of the truck that served as the as illustrated in Figure 6(b). The trailer was guided to
reaction beam. The jack was connected to a hydraulic ensure its wheels passed over the locations of two earth
pump with a high-pressure hose. A Dwyer® DPG-111 pressure cells. The axle width was the same as the
digital pressure gauge was mounted on the pump to distance of two earth pressure cells, so the pressure
measure the applied pressure, and three mechanical dial induced by two wheels of an axle could be measured and
gauges were placed on the loading plate at 120º to each recorded by earth pressure cells. The forward and
other to measure the displacement. Prior to the PLTs, the backward speeds were controlled at 5 km/h and 2 to 3
sand pad was preloaded with 10 kPa for 30 seconds for the km/h, respectively. During the tests, the datalogger
initial adjustment. During this process, any displacement CR1000X continuously took the readings of earth pressure
induced was recorded. The PLTs were then commenced cells at a sampling rate of 100 Hz.
by applying the predetermined incremental plate loads. At
each increment of loads, the load was maintained until the
displacement rate was less than 0.03 mm/min for three 1200
Applied, Type-D geocell
minutes before the next loading step was applied. At each Applied, Type-C geocell
incremental loading stage, the applied pressure and 1000 Applied, geocomposite
displacement were recorded, and the readings of the earth Measured, Type-D geocell
pressure cell were also measured and taken by the 800
Measured, Type-C geocell
datalogger CR1000X with the sampling rate of 100 Hz at Measured, geocomposite
Pressure (kPa)

the same time. The capacity of digital pressure gauge and

600
mechanical dial gauge was 34.5 MPa (5000 psi) and 25
mm, respectively. The corresponding accuracy was 0.25%
of the full-scale value and 0.01 mm. 400

200

0
1 2 3 4
Number of load increment
(a)

(a)
 60 about 4° larger distribution angle than the geocomposite
Type-D geocell
Type-C geocell section. This indicates that the geocells provide better
55 stabilization than the planar reinforcements (geocomposite
Geocomposite
in this case) for the road sections with the comparable
50
stiffness.
Distribution angle (°)

45
3.2.2 Trafficking test results
40
Figure 8(a) shows the typical time-history responses of the
35 trafficking tests for both the forward and backward passes
in the 150-mm Type-C geocell-stabilized road section.
30 When the wheels were passing over the earth pressure
cells, the recorded pressure behaved with a loading wave
25 where the peak value represented the distributed stress
when the cells were just underneath the tire contact areas.
20 The pavement performance was determined by the
0 200 400 600 800 1000
effectiveness in distributing the vehicle-induced stress over
Applied pressure (kPa) the subbase-base interface. The measured peak pressure
(b) was the major index for the evaluation. Therefore, peak
Figure 7. Plate loading test results of: (a) applied and pressure recorded by two earth pressure cells were chosen
measured pressure and (b) distribution angles in the three at each pass. The average peak values could be calculated
sections to evaluate the benefits of geosynthetics under traffic
loading. The calculated average peak pressure at each
pass is shown in Figure 8(b). Like the PLT result in Figure
To normalize the effect of base course thickness and to 7(b), the peak pressure transferred to the subbases was
quantify and estimate the effectiveness of geosynthetics- similar between the 150-mm Type-C geocell and the
stabilized base courses in distributing applied pressure geocomposite sections-e.g., 50.7 kPa and 50.4 kPa,
over subbases, the stress distribution angle was defined. respectively even though the geocell sections had 50 mm
The distribution angle, θ, was determined using Equation thinner base courses than the geocomposite section.
1. Moreover, the 100-mm Type-D geocell section effectively
𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 =
𝑏𝑏 𝑝𝑝
�� 0 − 1� [1] reduced distributed traffic loads over subbases with the
2ℎ 𝑝𝑝
1 peak pressure approximately 88% as compared to the
where, other two sections. However, the tire pressure applied to
𝜃𝜃 is the distribution angle, º. the pavement was not measured separately, so the load
𝑏𝑏 is the diameter of the loading plate, mm. transfer ratio and distribution angles could not be
ℎ is the depth between the pavement surface and the calculated in this case. This important parameter will be
subbase surface, mm. obtained in the next stage of field tests after the first
𝑝𝑝0 is the applied pressure by the loading plate, kPa. seasonal F-T cycle.
𝑝𝑝1 is the measured stress by the earth pressure cell,
kPa.
It should be noted that the effective geocell-reinforced 80
150-mm Type-C geocell, Inner wheel
zone is the geocell reinforced thickness plus 20 mm above
70 7th wheel
the geocell top (Pokharel et al., 2010), so the actual
distribution angle within effective geocell-reinforced section 3rd wheel 4th to 6th wheel 8th wheel
60 2nd wheel
is larger than the equivalent distribution angle. Figure 7(b)
Measured pressure (kPa)

presents the calculated equivalent distribution angle in the 50

three sections. The 100-mm Type-D geocell section (that Front
wheel
has 125 mm of effective geocell reinforced thickness) 40
showed the largest distribution angle of about 46° among
the three test sections. In contrast, the corresponding 30
angle was 38° and 34° in the 150-mm Type-C geocell and
20
geocomposite sections, respectively. The larger modulus
of the base courses in the 100-mm Type-D geocell section 10
as compared to the other sections contributed to the wider
stress distribution. Another reason might be the weaker 0
subbases overlain by the strong base courses in the 100-
mm Type-D geocell section, effectively mobilizing the 0 5 10 15 20 25 30 35 40 45 50 55 60
beam effect of geocell reinforcement (Dash et al., 2004; Time (Second)
Pokharel et al., 2011, White et al., 2010), which contrasted (a)
with stronger subbases than base courses in the other two
sections. It was interesting that the 150-mm Type-C geocell
section, having a 50 mm thinner base course layer, had
 90 Edil, T.B., Benson, C.H., Bin-Shafique, M., Tanyu, B.F.,
1st forward
1st backward Kim, W.H. and Senol, A., 2002. Field evaluation of
80
2nd forward construction alternatives for roadways over soft
70 2nd backward subgrade. Transportation Research Record, 1786(1),
Average peak pressure (kPa)

Averaged pressure pp.36-48.

60 Henry, K.S., Olson, J.P., Farrington, S.P., Lens, J. (2005).
Improved performance of unpaved roads during spring
50
thaw. Research Report, ERDC/CRREL TR-05-1,
40 Engineer Research and Development Center,
Hanover, New Hampshire, USA.
30 Huang, M., Lin, C. and Pokharel, S.K., 2023. Freeze-thaw
behavior of geocell-stabilized bases considering
20
different fines contents. Geosynthetics International,
10 pp.1-18.
Kawabata, S., Ishikawa, T. and Kameyama, S., 2016.
0 Effects of freeze-thaw history on bearing capacity of
100-mm Type-D 150-mm Type-C Geocomposite
geocell geocell
granular base course materials. Procedia engineering,
Pavement section 143, pp.828-835.
(b) Pokharel, S., Norouzi, M., Eng, P., Breault, M. (2017). New
Figure 8. Trafficking test results: (a) typical time-history advances in novel polymeric alloy geocell-
responses for the forward pass and (b) averaged peak stabilizedbase course for paved roads. Proceedings of
pressure recorded by earth pressure cells. 2017 Conference of the Transportation Association of
Canada, St. John’s, Newfoundland, Canada,
4 CONCLUSIONS September 2017, pp.1-16.
Pokharel, S.K., Han, J., Leshchinsky, D., Parsons, R.L. and
This paper presents the instrumentation and preliminary Halahmi, I., 2010. Investigation of factors influencing
results of a full-scale experiment to investigate the F-T behavior of single geocell-stabilized bases under static
performance of geosynthetics-stabilized flexible pavement loading. Geotextiles and Geomembranes, 28(6),
in Sturgeon County, Alberta. The sensor instrumentation pp.570-578.
proved to be effective as there were no damages to any Pokharel, S.K., Han, J., Manandhar, C., Yang, X.,
sensors. The preliminary results of field tests show the Leshchinsky, D., Halahmi, I. and Parsons, R.L., 2011.
effectiveness of geocell reinforcement in improving the Accelerated pavement testing of geocell-reinforced
performance of the flexible pavement. Even though the unpaved roads over weak subgrade. Transportation
150-mm Type-C geocell section had a lower degree of Research Record, 2204(1), pp.67-75.
compaction achieved and had 50 mm thinner base Qi, J., Vermeer, P.A. and Cheng, G., 2006. A review of the
courses, its distribution angle was 2° to 4° greater than the influence of freeze‐thaw cycles on soil geotechnical
geocomposite section. The effectiveness of the properties. Permafrost and periglacial processes,
geosynthetic reinforcement could have been more evident 17(3), pp.245-252.
had the native clayey subgrade layer was not replaced Simonsen, E. and Isacsson, U., 1999. Thaw weakening of
with gravelly soils. It is anticipated the long-term pavement structures in cold regions. Cold regions
performance monitoring and testing after several seasonal science and technology, 29(2), pp.135-151.
F-T cycles would help manifest the effectiveness of the Timm, D.H. and Priest, A.L., 2004. Dynamic pavement
geosynthetics. response data collection and processing at the NCAT
test track. NCAT Report, 4(03).
5. ACKNOWLEDGEMENT Viklander, P., 1998. Permeability and volume changes in
till due to cyclic freeze/thaw. Canadian Geotechnical
This research is financially supported by Natural Sciences Journal, 35(3), pp.471-477.
and Engineering Research Council (NSERC), MITACS White, D.J., Gieselman, H.H., Douglas, C., Zhang, J. and
and the Geosynthetic Institute (GSI). The authors are Vennapusa, P., 2010. In-Situ Compaction
grateful for their support. Mr. Daniel Breault, Mr. Arnold Measurements for Geosynthetic Stabilized Subbase:
Gonzales and Mr. Murtaza Ali from Stratum Logics Weirton, West Virginia. EERC Publication ER10-05.
provided assistance during the project. Their assistance in
this study is greatly appreciated.

6. REFERENCES

Dash, S.K., Rajagopal, K. and Krishnaswamy, N.R., 2004.

Performance of different geosynthetic reinforcement
materials in sand foundations. Geosynthetics
International, 11(1), pp.35-42.
Evaluating the Performance of Geogrid-
Stabilized Aggregate Layers for Low-
Maintenance Roadways through
Accelerated Traffic Loading
Ethan Landry, Chelsey Yesnik, Adam Hammerlindl, Ian Fleming &
Haithem Soliman
Department of Civil, Geological, and Environmental Engineering
University of Saskatchewan, Saskatoon, SK, Canada

ABSTRACT
The depletion of high-quality aggregates and soft subgrade soils are challenges for transportation agencies when building
low-maintenance roads. Unsurfaced roads rely on stiff aggregate layers over soft subgrade to resist a bearing capacity
failure from excessive traffic loads. However, the stiffness of the aggregate layers can be extended through mechanical
stabilisation with geosynthetics. A full-scale study, using a newly designed apparatus, has been conducted to quantify the
performance of geogrid-stabilised aggregate over a soft subgrade under accelerated traffic loading. Field studies can be
replaced by this apparatus, and it provides a controlled laboratory setting. The apparatus also has a keen focus on the
characterization of the subgrade material over the course of testing. Monitoring and controlling suction throughout the
subgrade effectively provide consistent conditions in each test section. Comparison testing focusing on the change in
material stiffness and deformations provide insight into the benefits of geosynthetic stabilisation. Agencies often seek to
minimize risk by increasing the lifespan of their construction, and geosynthetics can provide an effective method of
prolonging the stiffness of the aggregate layer and reducing the degree of rutting over repeated loading.

RÉSUMÉ
L'épuisement des granulats de haute qualité et des sols de fondation mous sont des défis pour les agences de transport
lors de la construction de routes à faible entretien. Les routes non revêtues reposent sur des couches de granulats rigides
sur un sous-sol souple pour résister à une défaillance de la capacité portante due à des charges de trafic excessives.
Cependant, la rigidité des couches de granulats peut être prolongée par une stabilisation mécanique avec des
géosynthétiques. Une étude à grande échelle, à l'aide d'un appareil nouvellement conçu, a été menée pour quantifier la
performance d'un agrégat stabilisé par une géogrille sur une plate-forme souple sous une charge de trafic accélérée. Les
études sur le terrain peuvent être remplacées par cet appareil, et il fournit un cadre plus contrôlé en laboratoire. L'appareil
se concentre également sur la caractérisation du matériau de fondation au cours des tests. La surveillance et le contrôle
de l'aspiration tout au long du sous-sol fournissent efficacement des conditions cohérentes dans chaque section d'essai.
Des tests de comparaison axés sur le changement de rigidité et les déformations des matériaux donnent un aperçu des
avantages de la stabilisation géosynthétique. Les agences cherchent souvent à minimiser les risques en augmentant la
durée de vie de leur construction, et les géosynthétiques peuvent fournir une méthode efficace pour prolonger la rigidité
de la couche d'agrégats et réduire le degré d'orniérage lors de chargements répétés.

1 INTRODUCTION excessive deformations have occurred, which are

indicative of structural failure. Base stabilisation relies on
1.1 Background particle interlock in the geogrid apertures. The interlocking
mechanism increases shear strength within the base
As the availability of aggregate becomes increasingly material, which laterally distributes the applied load.
scarce, road agencies must seek out alternative methods Wayne and others completed a study to test the
of prolonging the lifespan of roadways and providing load distribution capabilities of geogrids in road structures
adequate load support over the native subgrade soil. The (2013). Traffic loading was simulated using service
Canadian Prairies present unique challenges to road equipment over a mechanically stabilized layer (MSL)
construction, such as harsh weather conditions and poor using a triangular aperture shaped geogrid. This study
foundation materials. Due to these factors, it is imperative showed that stresses in the subgrade decreased by over
to find a solution to the long-term sustainability of 50% in the geogrid-stabilised section, as compared to the
roadways. unstabilised section.
Geosynthetics can be incorporated into road To further evaluate the performance of geogrids in
structures to provide increased load support and service roadway applications, an acclerated traffic loading device
life. More specifically, geogrids can improve soil conditions was constructed. An overview of this apparatus is
through stabilization and reinforcement. Stabilisation is discussed in this paper. For more details on the apparatus
mobilized at lower strains (i.e.,<75mm of rutting), and design and construction, refer to the paper by Landry et al.
reinforcement is mobilized at higher strains (i.e., >100mm (2022).
of rutting) (Giroud and Han 2016). Stabilization is the
primary function of a geogrid included in roadway
structures. The reinforcement function is not triggered until
1.2 Experimental Program includes 300 mm layer of aggregate and approximately 600
mm of subgrade. In consideration of bias between test
The USASK Wheel Loader Apparatus was constructed in sections, the location of mechanically stabilised sections is
the Multi-purpOse Slope Testing (MOST) Facility within the randomized for each trial. A schematic of the test setup is
University of Saskatchewan. The apparatus can be seen shown below in Figure 1. The properties of the materials
below in Figure 1. This apparatus accommodates four test are discussed in subsequent sections.
sections, each 1.5m in width, with a trafficking distance of
2.8 m. A one-half equivalent single axle load (ESAL) is
applied to the dual tire set using two 8” pneumatic
cylinders. A programmable logic controller (PLC) operates
the 7.5 kW servo motor which powers the apparatus. A
pressure transducer is installed in-line with the air feed to
monitor the pneumatic load supply.

Figure 2. USASK Wheel Loader Schematic.

Figure 1. USASK Wheel Loader Apparatus.

Data collection for this research consists of both

1.3.1 Aggregate
intermediate and continuous measurements of elastic and
plastic strains. Intermediate measurements include
To date, five trials have been conducted utilizing two
changes in dynamic deformation modulus (i.e., stiffness)
different types of base coarse aggregate. Initially, a local
and rutting of the aggregate surface. Changes in stiffness
aggregate was tested as to represent a typical base used
are measured using a light weight deflectometer (LWD).
for road construction in Saskatchewan. To represent an
LWD data has been collected using a 10 kg and 15 kg
aggregate more widely used worldwide in roadway
hammer. Deflections are recorded by two linear variable
construction, a crushed rock sourced from a pit near Flin
differential transducers (LVDTs) installed on both sides of
Flon, MB. In comparison to the Saskatchewan aggregate
the axle. Rutting is recorded manually using a rut slide. In
material, the crushed rock is coarser and more angular with
another companion paper for this research, multi-channel
100% fracture. The grain-size distribution of each
analysis of surface waves (MASW) has been completed to
aggregate type is shown below in Figure 3. Aggregate is
evaluate the changes in shear wave velocity and small-
placed in three 4” lifts to comprise a 12” (300 mm) layer of
strain shear modulus with increasing number of load cycles
base. Lifts are compacted using a standard plate tamper
(Yesnik et al. 2023). The current test plan includes
and checked for adequate compaction by nuclear
intermediate measurements at 150, 450, 1350, 2700,
densometer testing. Proctor tests were conducted
4000, and 7500 load cycles. A plastic deformation limit of
following ASTM D698 (2021) and used a rock correction as
75 mm has been chosen to limit the duration of testing on
prescribed by Houston and Walsh (1993) for proctors on
excessively deformed test sections. Continuous monitoring
materials with a large gravel content.
observes changes in subgrade suction and vertical stress.
Subgrade suction is monitored using 18 Teros 31
tensiometers inserted throughout the sidewalls of the
apparatus. Four vibrating wire total stress cells are used to
monitor changes in vertical stress. These total stress cells
are positioned under the center of two lanes and at the
interface of adjacent lanes. This was done to observe the
reduction in vertical stresses both directly beneath the
applied load and at a 0.75 m offset from the load.
Additionally, the California Bearing Ratio (CBR) of the
subgrade material was collected using a MEXE cone
penetrometer before and after testing.

1.3 Material Description

Each test setup is constructed to represent a typical

unsurfaced road for the Canadian Prairies. Each test setup Figure 3. Aggregate Grain Size Distribution.
 2.40 Tempe Cell WP4C Van G Fit
30.0%
2.35
Dry Density (T/m3)

2.30
8.00%, 2.320 25.0%

Grav. MC
2.25

2.20
20.0%

2.15
2.00% 4.00% 6.00% 8.00% 10.00% 12.00%
Water Content (%)
15.0%
1 10 100 1000
Suction (kPa)
Figure 4. Proctor for Flin Flon, MB Aggregate.
Figure 6. Pottery Clay SWCC.
1.3.2 Subgrade
Using a lab vane apparatus (ASTM 2019) a
Pottery clay, sourced from Medicine Hat, AB, was used as relationship between unsaturated undrained shear
artificial subgrade material. This material is a low plasticity strength and gravimetric water content was evaluated.
clay which is vacuum pugged in 10 kg blocks. These blocks These results are shown in Figure 7. By logging the change
are hand placed tightly together and are packed using hand in suction in the subgrade and using the relationships
tools. To achieve a 24” layer of subgrade, nearly 2000 shown in Figure 6 and Figure 7, the undrained shear
blocks were used. After each test setup, approximately half strength can be determined for any measured water
the material was excavated and replaced for the next test content. These calculated values of undrained shear
setup. An image of the installed subgrade can be seen strength are compared with the results from MEXE cone
below in Figure 5. After placing the subgrade layer, the clay testing. As CBR closely relates to the unsaturated
surface was smoothed by hand and levelled using a undrained shear strength, results from calculated values
construction level. and measured should be similar.
Subgrade characterization was completed to
60
determine the Atterberg limits, soil-water characteristic
curve (SWCC), and the relationship between unsaturated 50
Undrained Shear

undrained shear strength and water content. Following

Strength (su)

40
ASTM D4318 (2017), the plastic and liquid limit of the 30
subgrade was determined as 17.3% and 34.4%,
20
respectively.
10
0
18.0% 20.5% 23.0% 25.5% 28.0%
Water Content (%)
Figure 7. Relationship of Lab Vane Unsaturated Undrained
Shear Strength and Water Content.

1.3.3 Drainage Layer

Figure 5. Installed subgrade prior to smoothing. Fine sand was acquired by sieving from a bulk sourced
material. Approximately 1 m3 of sand passing the no. 100
sieve was used for this apparatus. This fine sand was
Using a Tempe cell at low pressures and a WP4C chosen to aid in the control of head throughout the
soil-water potential meter at higher pressures, the SWCC subgrade layer. By adjusting the level of the reservoir,
plot was generated. The results of these laboratory testing water pressure can be applied evenly across the bottom of
are shown below in Figure 6. the subgrade to maintain saturation of the overlying
subgrade.
 Top BC - Port 1 Middle BC - Port 2
Bottom BC - Port 3 Top CD - Port 4
Middle CD - Port 5 Bottom CD - Port 6
West Top AB - Port 1 West Mid AB - Port 2
West Bottom AB - Port 3

-25.00

-20.00

-15.00

Figure 8. Sand layer in place. -10.00

1.3.4 Geosynthetics -5.00

To construct this soil geosynthetic composite (SGC), three 0.00

different geogrids are used. These geogrids are a punched 2023-03-30 2023-04-03 2023-04-07 2023-04-11
and drawn polypropylene geogrid. These geogrids vary in Figure 10. West Wall Suction Data.
rib thickness. For the purposes of this paper, geogrid 1,
geogrid 2, and geogrid 3 indicate the geogrids in increasing
thickness respectively. These geogrids are all variable Top BC - Port 1 Middle BC - Port 2
aperture shaped geogrids (VASG), an example can be Bottom BC - Port 3
Middle CD - Port 5
Top CD - Port 4
Bottom CD - Port 6
seen below in Figure 9. In addition, a separator layer of East Top AB - Port 4 East Mid AB - Port 5
East Bottom AB - Port 6
three 8 oz. non-woven geotextiles are installed below the
-25.00
sand layer. This was done solely to ensure that clogging is
prevented by migration of fines from the drainage layer into -20.00
the reservoir system.
-15.00

-10.00

-5.00

0.00
2023-03-30 2023-04-03 2023-04-07 2023-04-11
Figure 11. East Wall Suction Data.

MEXE cone penetrometer tests downwards from

the surface of the subgrade was taken prior to aggregate
placement and following the exhumation aggregate. CBR
Figure 9. Sample of Variable Aperture Shaped Geogrid measurements were completed in a 3 by 5 grid in each test
(VASG). section to a depth of approximately 24”. The results of this
testing can be seen below in Figure 12. Pre-and post-
testing show a gradual increase in subgrade CBR after
2 RESULTS loading. This response is expected, as CBR will increase
as the subgrade is compacted under traffic load.
2.1 Subgrade Monitoring

The data presented in this paper contains results from the

fourth trial of this research, other trials will be presented in
future publications. Continuous monitoring of suction
throughout the duration of testing was completed. This data
was logged using METER ZL6 data acquisition systems on
a 1-hour measurement interval. An indicator maximum
suction of 25 kPa was monitored for the course of testing.
This threshold ensured no desiccation occurred in the
subgrade by desaturation. Rapid increases in suction
would indicate desaturation of the underlying subgrade. Figure 12. Top Down MEXE Cone Testing.
The goal was to have suction at each lane location stay
mostly constant. Figure 10 and Figure 11 show the suction Due to the duration of testing and test section
data measured by the tensiometers at various ports for this construction/deconstruction, it was determined that MEXE
test trial. As shown in these figures, suction only fluctuates cone testing must be completed closer to the time of traffic
by +/- 2 - 5 kPa. loading to increase the accuracy in CBR measurements.
To allow for MEXE cone testing closer to load cycling, 8 the excessive rutting during the control lane, which for this
PVC ports were installed parallel to the wheel path as trial was in lane C, immediately adjacent and cyclically
shown back in Figure 2. One upper and lower port is tested loaded to significant rutting immediately prior to
per side of the exterior box on the apparatus prior to and commencing Lane D with GG3. Future work will be
directly after the test is concluded. The results from the required to eliminate this effect.
through-wall MEXE cone testing can be seen below in
Figure 13. Like the top-down results, the values also
display an overall increase in CBR with penetration depth
and traffic loading.

Figure 15. Average Rut Depth Over Load Cycling

Figure 13. Through Wall MEXE Cone Testing.

2.3 Light Weight Deflectometer (LWD)
2.2 Rutting and Linear Variable Differential Transducer
(LVDT) The light weight deflectometer was used to measure the
changes in dynamic deformation modulus as loading
Deformations were measured using two LVDTs mounted progressed. Road structures must endure heavy loading
onto the hubs which connects the axle to the pneumatic on a regular basis. Shakedown will eventually wear down
cylinders. Data is collected at 0.5 second intervals to the structure until rehabilitation is required. The use of a
measure the loaded deflection across the entire trafficking geogrid, or appropriate geosynthetics, can reduce the rate
pass. The maximum differential value at each load cycle is of stiffness degradation by laterally spreading the applied
averaged between the two LVDTs to produce the loaded load and restraining particle movement. The dynamic
deflection for every pass. Below in Figure 14 is the data deformation modulus is therefore specifically important for
produced from the LVDTs showing the loaded quantifying the impact of the use of a geosynthetic on how
deformations per cycle. it may improve the soil-geosynthetic composite.

Control Geogrid 1 Geogrid 2 Geogrid 3 To evaluate the performance of each test section,
0
a grid of test locations were selected to measure the
10
change in dynamic deformation modulus. Changes in
20
dynamic deformation modulus are measured outside and
Deformation (mm)

30 inside the wheel path. Points within the wheel path were
40 collected at all intermediate load cycles. Points outside the
50 wheel path were collected prior to and following traffic
60 loading due to height restrictions within the apparatus.
70 Additional testing was conducted in subsequent lanes both
80 at some intermediate points and after testing was
0 1000 2000 3000 4000
Load Cycles
5000 6000 7000 8000
concluded in the tested lane. This additional study was
done to evaluate any changes in stiffness from loosening
Figure 14. Loaded Deformations Measured from LVDTs. of the surficial material.

Rut measurements are averaged over both wheel Figure 16 shows the distribution of dynamic
paths at various locations along the trafficked length. The deformation modulus with load cycles for the last trial.
results from this trial are shown in Figure 15. Both Figure Similar to the trends of the deformation results, geogrids 1
14 and Figure 15 show that geogrids 1 and 2 perform the and 2 provide the longest duration of added support as the
best, with similar performance in plastic and loaded dynamic deformation modulus was maintained for the
deformations. Geogrid 3 and the control section surpassed duration of testing after a brief decrease in the earliest
the tolerance of 75 mm for rut depth prior to reaching the stages of testing. Geogrid 3 and the control section show a
long-term load cycle count. Due to these test sections dramatic decrease in modulus in the early stages of testing.
surpassing the prescribed threshold, load cycling was However, the results stabilise in Geogrid 3 and the control
concluded. The reason for the poorer-than expected section after 500 cycles.
performance of GG3 in this trial is believed to result from
some loosening of the aggregate in this lane because of
 the control section which does not include any
geosynthetics.

5 ACKNOWLEDGEMENTS

The work described in this paper was supported by Tensar

International Corporation, a division of CMC. Special
thanks to Adam Hammerlindl for his knowledge of
construction and design of the wheel loader. Another
special thanks to Matthew Brunette, Justin Salandaman,
and Dominic Froese for their support throughout the
construction process.

Figure 16. Change in Dynamic deformation Modulus with 6 REFERENCES

Load Cycling.
ASTM. 2017. D4318-17e1: Standard Test Methods for
3 FUTURE WORK Liquid Limit, Plastic Limit, and Plasticity Index of
Soils. Standard, ASTM International, West
Further interpretation of this data will include a statistical Conshohocken, PA, USA.
analysis of stiffness change. This statistical analysis will ASTM. 2019. D8121/D8121M − 19: Test Method for
evaluate the significance of stiffness loss with increasing Approximating the Shear Strength of Cohesive
number of load cycles and the calculated recovered strain. Soils by the Handheld Vane Shear Device. ASTM
Single camera photogrammetry has been collected to International.
evaluate strains and deformation of the geogrid-subgrade ASTM. 2021. D698-12: Standard Test Methods for
surface. This evaluation will help determine the failure Laboratory Compaction Characteristics of Soil
mechanisms for each test section. Using Standard Effort (12,400 ft-lbf/ft3 (600 kN-
Total stress cell data has been collected which will m/m3)). ASTM International, West
enable development of a two-layer Boussinesq stress- Conshohocken, PA, USA.
distribution for varying levels of strain of the geogrid and Giroud, J., and Han, J. 2016. Mechanisms governing the
road surface. Determining the Boussinesq stress performance of unpaved roads incorporating
distribution at different strain levels will provide a better geosynthetics. Geosynthetics, 34(1).
indication of the lateral load distribution provided by the Houston, S.L., and Walsh, K.D. 1993. Comparison of Rock
geogrid. Correction Methods for Compaction of Clayey
Work is required to confirm apparent effect of Soils. Journal of Geotechnical Engineering,
significant rutting on the performance of the subsequent 119(4): 763–778. doi:10.1061/(ASCE)0733-
lane, i.e., Lane D performed unexpectedly poorly, perhaps 9410(1993)119:4(763).
because it was adjacent to Lane C, the Control in which Landry, E., Hammerlindl, A., Fleming, I., and Soliman, H.
significant rutting developed. If this is confirmed, 2022. Design and Construction of a Full-Scale
modifications to the system may be required to eliminate Traffic Simulating Apparatus Over a Soil-
this as a source of bias in the results. Geosynthetic Composite. Calgary, AB.
This apparatus will be used in the future to evaluate Wayne, M., Fraser, I., Reall, B., and Kwon, J. 2013.
changes to the soil-geosynthetic composite by varying any Performance verification of a geogrid
component of aggregate, subgrade, or geosynthetic. mechanically stabilised layer. In 18th International
Future research will also investigate the impact of freeze- Conference on Soil Mechanics and Geotechnical
thaw cycling of the soil-geosynthetic composite. Engineering. Paris. pp. 1381–1384.
Yesnik, C., Morozov, I., Fleming, I.R., Soliman, H., Landry,
4 CONCLUSIONS E., Hammerlindl, A., Wayne, M.H., Lees, A.S.,
and Kawalec, J. 2023. Using non-destructive
A new apparatus has been constructed for the purpose of testing to evaluate geogrid-stabilised aggregates
evaluating the performance of soil-geosynthetic subject to accelerated traffic loading. In 12th
composites subject to simulated traffic loading. This International Conference on Geosynthetics.
research studies the impact of geosynthetic stabilisation by
applying a channelized load with a weak subgrade soil. To
evaluate the performance of the SGC, deformation data is
collected by using two LVDTs and a rut wedge. The
dynamic deformation modulus is also measured to
determine the road structure rebound following applied
load. The results presented in this paper indicate that
Geogrids 1 and 2 perform the best in both areas, showing
the most vertical restraint and sustained stiffness over
traffic loads. It is also evident from this data that the
inclusion of a geogrid provides added support compared to
An investigation of tyre contact area on
unbound aggregate
Lees AS
Tensar International, Nicosia, Cyprus
Robinson WJ
USACE Engineer Research and Development Center, Vicksburg,
MS, USA
Wayne M
Tensar Corporation, Alpharetta, GA, USA
Han J
University of Kansas, Lawrence, KS, USA

ABSTRACT
An important input parameter to the design of unpaved roads is the wheel load from vehicle traffic. Some design methods
consider only the axle load regardless of tyre size whereas more advanced methods consider a tyre contact pressure and
contact area at the unbound aggregate surface. They assume the contact pressure equals the tyre inflation pressure and
the contact area is calculated accordingly – usually assumed to be a circular contact area. However, truck tyre pressures
are often higher than typical aggregate bearing capacity meaning that the unbound aggregate must yield somewhat under
such pressures while a proportion of tyre load must be transferred to the ground through the tyre carcass. A full-scale
physical trial was undertaken by measuring the contact area between a dual tyre of different loads and tyre inflation
pressures on a compacted unbound aggregate layer overlying a clay subgrade as well as on a concrete slab. The results
confirmed that contact areas were significantly higher, and hence average contact pressures lower, than predicted by the
assumption of tyre contact pressure equals the tyre inflation pressure. The error was as high as about 100% which has
serious implications for unpaved roads design methods that adopt this assumption. An alternative method of estimating
tyre contact pressure using Hertzian elastic contact theory was proposed. A low stiffness was adopted for the unbound
aggregate to take account of yield while the tyre stiffness was adjusted for tyre inflation pressure. The proposed method
provided a much improved prediction of the tyre contact areas and shapes in this physical study as well as other physical
and numerical studies. If adopted in road design, it should improve the accuracy of designs across a wide range of axle
loads and tyre sizes.

RÉSUMÉ
Un paramètre d'entrée important pour la conception des routes non revêtues est la charge des roues due à la circulation
des véhicules. Certaines méthodes de conception ne prennent en compte que la charge par essieu, quelle que soit la taille
du pneu, tandis que des méthodes plus avancées prennent en compte une pression de contact du pneu et une zone de
contact à la surface de l'agrégat non lié. Ils supposent que la pression de contact est égale à la pression de gonflage des
pneus et la zone de contact est calculée en conséquence - généralement supposée être une zone de contact circulaire.
Cependant, les pressions des pneus de camion sont souvent supérieures à la capacité portante typique des agrégats, ce
qui signifie que l'agrégat non lié doit céder quelque peu sous de telles pressions tandis qu'une proportion de la charge des
pneus doit être transférée au sol à travers la carcasse du pneu. Un essai physique à grande échelle a été entrepris en
mesurant la surface de contact entre un pneu jumelé de différentes charges et pressions de gonflage des pneus sur une
couche de granulats compactés non liés recouvrant une sous-couche d'argile ainsi que sur une dalle de béton. Les
résultats ont confirmé que les zones de contact étaient significativement plus élevées, et donc les pressions de contact
moyennes inférieures, que prévu par l'hypothèse selon laquelle la pression de contact des pneus est égale à la pression
de gonflage des pneus. L'erreur était aussi élevée qu'environ 100 %, ce qui a de sérieuses implications pour les méthodes
de conception de routes non revêtues qui adoptent cette hypothèse. Une méthode alternative d'estimation de la pression
de contact des pneus à l'aide de la théorie de contact élastique hertzienne a été proposée. Une faible rigidité a été adoptée
pour l'agrégat non lié afin de tenir compte du rendement tandis que la rigidité du pneu a été ajustée en fonction de la
pression de gonflage du pneu. La méthode proposée a fourni une prédiction bien améliorée des zones de contact et des
formes des pneus dans cette étude physique ainsi que dans d'autres études physiques et numériques. S'il est adopté
dans la conception de routes, il devrait améliorer la précision des conceptions sur une large gamme de charges par essieu
et de tailles de pneus.

1 INTRODUCTION aggregate layer is trafficked on directly without a durable

surface course. These instances include low-traffic public
Roads need an aggregate base course where the and private roads, heavy haul roads and temporary
subgrade soil has insufficient strength and stiffness to construction roads. The temporary construction roads
support traffic loads. In many instances, the unbound include cases where surface courses will be added later,
thus the temporary unpaved phase becomes an important design methods where the tyre inflation pressure has a
design case. significant effect on predicted performance.
Unpaved road design involves determining the required Fervers (2004) performed a tyre-soil interaction study
base course thickness to distribute tyre loads to the using 2D plane strain finite element analysis (FEA) of an
subgrade such that the road remains serviceable air-filled tyre in rolling, frictional contact with an elasto-
throughout its design life. Serviceability is usually defined plastic soil continuum. The soil types considered included
in terms of rut depth. The most important input parameters a dense sand with strength (φ′=40°, c′=5 kPa) similar to but
to the thickness calculation are the subgrade strength – slightly lower than a typical base course aggregate. The
often expressed as California bearing ratio (CBR) – and the analysis was performed with a typical tyre inflation
vehicle loading. pressure of 750 kPa and then again with a very low
Fundamental to the distribution of vehicle loads to the pressure of 150 kPa. The outputs of predicted deformation
subgrade is the vehicle’s distribution of its self-weight to the at true scale and mean principal stress from the two
aggregate surface. Yet, some empirical design methods analyses are compared in Figure 1.
(Giroud and Noiray, 1981; Powell et al, 1984) consider only Also added to Figure 1 are the contact lengths
the axle load regardless of tyre size or pressure which estimated by scaling from the diagrams the lengths where
restricts their application to the specific cases used to the tyres and ground are in contact and where mean
derive the design equations. Some, more analytical, design principal stress was at least 20 kPa. From these contact
methods, such as Giroud and Han (2004), Thompson et al lengths the average contact stresses were calculated and
(2018) and Gonzalez (2015), consider tyre contact areas are shown in Figure 1.
and pressures.
In such cases, the wheel load is assumed distributed to
the road surface as a uniformly distributed vertical stress 14 R20 MIL type
equal to the tyre inflation pressure (Huang, 1993; Yoder & p=750kPa Wheel load 30kN p=150kPa
Witczack, 1975). The contact area is assumed to be
circular, even for dual tyres, with the radius determined
such that the resultant force of the distributed stress equals Avge. contact stress
551mm
the total wheel load. Adoption of a circular area allows 368mm 228kPa 152kPa
simplified stress analysis down through the road layers
using an axisymmetric geometrical assumption. The tyre
inflation pressure is an important input parameter to these
design methods that affects load distribution to the
subgrade, calculated rut depths and hence the computed Mean principal stress (kPa)
base course thickness. 0 20 40 60 80 100 120 140 160 180 200 220 240
This approach works reasonably well to determine the
average contact stress on paved surfaces although the Figure 1. Effect of tyre inflation pressure on contact area
contact area is not circular and the contact stress non- and pressure (Fervers, 2004)
uniform (De Beer et al, 1997). Measured contact stresses
tend to be lower due to a proportion of the wheel load being
transferred to the ground directly through the tyre carcass At the typical truck tyre inflation pressure of 750 kPa,
(van Vuuren, 1974). However, on unpaved roads, tyre the average tyre contact pressure was about 228 kPa
inflation pressures typically exceed the yield stress of which is less than one-third of the inflation pressure as
unbound aggregates. Added to this, typical inflation assumed in some design methods. The average tyre
pressures on truck tyres have increased in recent decades contact pressure for the lower inflation pressure was
due to improvements in tyre technology. In surveys of truck approximately equal to the inflation pressure.
tyre pressures in South Africa, De Beer et al (2012) noted These observations suggest that in current unpaved
average values had increased from around 633 kPa in road design methods adopting the traditional tyre inflation
1974 to about 800 to 900 kPa in 2012, with values as high pressure contact stress assumption results in the contact
as 1 MPa being measured. Considering that the surface stress being over-estimated and the contact area
bearing capacity of an aggregate with a shear strength underestimated for typical truck tyres and inflation
typical of road bases of φ′=45° under typical truck tyre pressures. This could cause subgrade stresses to be
dimensions is only about 500 kPa, it is likely that unbound under-estimated (since smaller contact areas have a
aggregate materials will yield, resulting in increased shallower depth of influence), leading to inadequate
contact areas and lower contact pressures than predicted designs and excessive subgrade rutting that is difficult to
by the simple tyre inflation pressure approach. repair and which leads to further degradation of the
Tyre pressures are rarely varied during physical tests subgrade. Even though the tyre inflation pressure is an
on unpaved roads but Little (1992), after experiencing less important input parameter to these design methods, it
than expected rutting after 250 passes of an 86 kN axle would appear that it has less influence on the contact area
load and tyre pressure of 490 kPa, increased the tyre and stress than hitherto assumed.
pressure by 29% to 630 kPa for a further 750 passes in an Interactions between pneumatic tyres and rigid
attempt to increase rutting. He noted that the increase in surfaces have benefited from extensive physical (e.g. De
tyre pressure had little effect. This is in contrast to those Beer et al (2012), van Vuuren (1974)) and numerical study
(e.g. Roque et al (2000)). Similarly, interactions between
pneumatic tyres and terrain or soil have received extensive (254 mm) thick layer of the material was placed over a
study in the field of terramechanics for off-road vehicle and prepared 2% CBR clay subgrade in one lift and compacted
agricultural applications (Fevers (2004), Taheri et al as described by Robinson (2022), yielding an average CBR
(2015), Xia (2010)). However, interactions between value of approximately 50%. The concrete surface was
pneumatic tyres and unpaved, unbound but high-quality rigid relative to the pneumatic tyres and comprised a
aggregates have suffered from a paucity of scientific study, nearby hardstanding of flat, undamaged concrete slabs.
leading to apparently unrealistic assumptions in unpaved Two methods of measuring the tyre contact area at
road design. each load and inflation pressure combination were
The aim of the physical study presented in this paper employed. In Phase 1, the dual wheel was driven onto a
was to help fill the knowledge gap concerning tyre contact large sheet of thick paper overlaid on a rubber sheet (to
pressures and areas between hard (paved road) and soft prevent tearing of the paper) overlaid on the unbound
(subgrade soil) surfaces. It was also intended to help verify aggregate surface or concrete. Paint was sprayed around
an alternative calculation method. the base of the wheels to leave an uncoated tyre contact
area on the paper – an example of which is shown in Figure
2 PHYSICAL STUDY DESCRIPTION 3. Once the paint had dried, the contact area was
determined by measuring average dimensions of the
A full-scale physical study of tyre contact stress and rectangular contact areas. No attempt was made to
contact area on unbound aggregate was undertaken at the measure the precise contact area of the tyre tread since
US Army Engineer Research and Development Center only the average contact stress was of interest.
(ERDC) in Vicksburg, Mississippi. Tyre loads were applied
to an unbound aggregate layer and a comparatively rigid
concrete surface using the single-axle dual wheel tyre load
cart shown in Figure 2. Lead weights were added and
removed to create wheel loads of 45.8 and 29.3 kN
respectively. The tyre inflation pressure was also set at
either 689 or 827 kPa thereby creating four combinations
of wheel load and tyre inflation pressure. The tyres were
Samson GL283A 255/70R22.5.

Figure 3. Phase 1 tyre contact area measurement by spray

paint

The Phase 1 testing showed that the contact widths

remained constant. As the tyre load increased, only the
contact length increased. Amid concerns that the
compressibility of the rubber layer resulted in longer
contact areas, an alternative contact area measurement
method was adopted in a second phase of testing.
In Phase 2 the tyres remained in direct contact with the
unbound aggregate or concrete surface and the contact
lengths were measured by placing angle rules at each end
of the contact length of one tyre and measuring the contact
Figure 2. Dual-tyre load cart length directly as shown in Figure 4. Since the Phase 1
testing showed that the contact widths remained constant
in all cases, only the contact lengths were measured in
The aggregate base course comprised a crushed Phase 2. The thickness of the rulers still caused the contact
limestone aggregate as described by Robinson (2022) and lengths to be over-estimated – visible as a wedge-shaped
classified as A-1-a according to AASHTO (2012). A 10”
void at each end of the contact – so a nominal 40 mm was
subtracted from all the measured lengths.
While more precise technology exists to measure tyre 800
contact pressure, such as flexible pressure sensors, the
aim of this physical study was to measure total contact area Predicted by Hertzian contact theory

Average contact stress (kPa)

600
and average contact stress as used in design methods
rather than stress distribution within the contact area. For 400
these purposes, the rudimentary measurement methods
worked very well. In Phase 2, the measurement method
200
could have been improved by using thinner metal plates,
such as feeler gauges, to locate the edges of the tyre
contact lengths more precisely. 0
p=827kPa p=827kPa p=689kPa p=689kPa p=827kPa
P=29.3kN P=45.8kN P=29.3kN P=45.8kN P=45.8kN
(conc.)

(b) Average contact stress

Figure 5. Phase 1 test results and predictions (both tyres)

500
Predicted by Hertzian contact theory
400

Contact area (cm2)

300

200

Predicted by traditional method

100
based on tyre inflation pressure
0
827kPa 827kPa 689kPa 689kPa 827kPa 689kPa 827kPa 689kPa
13.2kN 20.6kN 13.2kN 20.6kN 20.6kN 20.6kN 13.2kN 13.2kN
(conc.) (conc.) (conc.) (conc.)
Figure 4. Phase 2 tyre contact length measurement
(a) Tyre contact area

3 PHYSICAL STUDY RESULTS 800

700 Predicted by Hertzian contact theory

Average contact stress (kPa)

The measured contact areas and corresponding calculated 600

average contact stress from Phase 1 are shown by the 500
vertical bars in Figure 5. The tyre inflation pressure p and 400
dual wheel load P are denoted for each case, while the 300
single case on concrete is denoted “(conc.)”. The contact 200
area on unbound aggregate appears to be more dependent
100
on the wheel load than the tyre inflation pressure. The
0
average contact pressures were much lower than the tyre 827kPa 827kPa 689kPa 689kPa 827kPa 689kPa 827kPa 689kPa
inflation pressures and the corresponding contact areas 13.2kN 20.6kN 13.2kN 20.6kN 20.6kN 20.6kN 13.2kN 13.2kN
much larger than predicted by the traditional method as (conc.) (conc.) (conc.) (conc.)
shown in Figure 5. The contact area shapes shown in
Figure 3 confirm that the contact areas were more (b) Average contact stress
rectangular under each individual tyre than circular. Figure 6. Phase 2 test results and predictions (left-hand
dual tyre)
1000
Predicted by Hertzian contact theory
800
In a similar way, the data from Phase 2 are shown in
Contact area (cm2)

600
Figure 6. Only the left tyre (in the direction of travel) of the
dual wheel pair was measured in this case, hence the lower
400 wheel load and smaller contact area. The spray paint
Predicted by traditional method method in Phase 1 detected a consistently larger contact
200 based on tyre inflation pressure area beneath the right-hand tyre of about a 55:45 ratio
indicative of a higher load in the right-hand wheel of about
0
p=827kPa p=827kPa p=689kPa p=689kPa p=827kPa
the same ratio. This imbalance was probably caused by a
P=29.3kN P=45.8kN P=29.3kN P=45.8kN P=45.8kN
slight tilt in the load cart towards the outrigger that prevents
(conc.) overturning. This imbalance was taken into account in the
(a) Tyre contact area derivation of the wheel load and average contact stress
indicated in Figure 6. The average contact stresses
obtained in both phases of testing were about the same on with
1
=
1
+
1

unbound aggregate which indicated that the 𝑅 𝑅1 𝑅2

measurements were not overly sensitive to the two

measurement methods. The one concrete case in Phase 1 where a=contact radius and R=radius of cylinders (see
had a lower average contact stress than Phase 2, perhaps Figure 7).
due to the use of a rubber sheet beneath the brown paper Equations 1 and 2 were adapted to provide an
which had a bigger impact on the high stiffness concrete approximate prediction of the contact area dimensions
surface than the unbound aggregate. between a tyre and different surfaces. The tyre width or,
The smaller contact area and corresponding higher more specifically, the tyre tread width was represented by
average contact stress on the concrete surface compared the cylinder length L and the tyre load by P. On a flat road,
with the unbound aggregate is clearly apparent for all R2 would be very large meaning that R≈R1 which is the tyre
combinations of wheel load and tyre inflation pressure. radius. The tyre contact width B is equivalent to 2a and
Again, prediction of the contact area based on tyre rearranging Equations 1 and 2 gives Equation 3 for its
inflation pressure alone resulted in a significant determination.
underestimation on the unbound aggregate as shown in
Figure 6(a). However, the method appeared to work rather 𝐵=√
16𝑅𝑃
[3]
well for the cases on concrete. This confirmed the need for 𝜋𝐸𝑟 𝐿

an improved calculation method on unpaved road surfaces.

1 1−𝜈𝑡2 1−𝜈𝑠2
with = +
𝐸𝑟 𝐸𝑡 𝐸𝑠

4 DATA INTERPRETATION
Using Equation 3 requires estimation of the relative
stiffness Er between the tyre and the road surface. Tyre
Hertzian contact mechanics (Hertz, 1881) provides
stiffness will depend on a multitude of factors including the
equations for the linear elastic deformation of solids in
tyre type (bias-ply or radial), tyre structure (geometry,
frictionless contact. The equations for parallel cylinders in
rubber and reinforcement), temperature, wheel load,
contact (Figure 7) are given below as Equations 1 and 2.
inflation pressure and rolling condition (free rolling,
accelerating, braking, cornering) (Wang et al, 2012). A
simplified equivalent tyre Young’s modulus Et will be
derived from a back-analysis of physical tests in this
section. Paved surfaces are orders of magnitude stiffer
than tyres so Er≈Et on paved roads. Unbound aggregate
surfaces and tyres have a similar order of magnitude
stiffness, so both Et and Es (road surface Young’s modulus)
values are important.
The physical test results described in the previous
section were analysed using Equation 3 and Et and Es
adjusted to achieve the best fit with the data as shown in
Figures 5 and 6. L was taken as the tyre tread width 187
mm and tyre radius R as 464 mm. A particularly good fit
with the unbound aggregate cases was achieved by taking
Es as 12 MPa which is particularly low for a compacted,
high-quality aggregate.
Similarly, the two cases from the Fervers (2004) study
shown in Figure 1 and Table 1 were back-analysed using
Equation 3. A good fit was obtained by taking a lower Es
value of only 2 MPa for the sand surface of that study.
Figure 7. Contact between parallel cylinders Being a numerical analysis, the stiffness was well-defined
(E=40 MPa, ν=0.2) but a much lower stiffness was required
using Equation 3 to take account of the yielding that
occurred in the elasto-plastic model of that study and which
𝜋
𝑃 ≈ 𝐸𝑟 𝐿𝑑 [1] would also explain the low stiffness needed to obtain a
4
good fit with the physical test results reported in this paper.
1 1−𝜈12 1−𝜈22 The Et values are shown in Table 1 for the various tyre
with = + inflation pressures that achieved the best fit with the
𝐸𝑟 𝐸1 𝐸2
physical and numerical analysis data for the unbound
where P=contact force, L=cylinder length, aggregate, sand and concrete surfaces.
d=indentation depth, E=Young’s modulus of cylinders and Also shown in Table 1 are the test cases from the van
ν=Poisson’s ratio of cylinders. Vuuren (1974) study. He tested a range of commercial
truck tyres for their contact area under different loads and
inflation pressures. Not all the data was presented but
𝑎 = √𝑅𝑑 [2] regression equations between tyre load, tyre inflation
pressure and contact area against a rigid surface were The Et values and corresponding tyre inflation
provided and, from these, Es values were determined for pressures p shown in Table 1 are plotted in Figure 9. It is
the best fit between Equation 3 and his physical test apparent that for the range of truck tyres considered across
results. The Es values are shown in Table 1 and the fit typical tyre inflation pressures, the linear relationship in
between the measured and Equation-3-predicted values of Equation 4 provides a reasonable fit (R2=0.87) to the data
tyre contact stress are shown in Figure 8. Figure 8 also and may be used to estimate appropriate Et values for
includes the physical study reported in this paper and the similar truck tyres when estimating tyre contact areas using
Fervers (2004) outputs shown in Figure 1. Across all three Equation 3.
data sources, a coefficient of determination R2 of 0.99 was .
achieved.
8000

Table 1. Parametric study tyre data

Tyre Young's modulus Et (kPa)

Tyre L(m) R(m) Et (MPa) Et (MPa) 6000

Physical study described in this paper

Dual Samson 1871 464 6.02 7.03
GL283A 255/70 4000
R22.5
Fervers (2004) numerical study
14 R20 MIL type 356 629 1.14 7.55
2000
Van Vuuren (1974) physical tests
7.50-15 Michelin 0.191 0.385 3.8 6.5
8.25-20 x 10 0.209 0.487 4.3 6.6
Firestone 0
0 200 400 600 800 1000
9.00-20 x 10 0.229 0.500 4.5 7.0
Firestone Tyre inflation pressure p (kPa)
10.00-20 x 14 0.254 0.525 4.6 7.3 Figure 9. Relationship between tyre inflation pressure and
Goodyear
equivalent tyre Young’s modulus Et for the cases studied.
11.00-20 x 14 0.279 0.547 5.1 7.4
General SDT
11.00-22 x 14 0.279 0.568 4.8 7.0
General 𝐸𝑡 = 7.4𝑝 + 16𝑝𝑎 [4]
1Treadwidth, 2at p=689kPa, 3at p=827kPa, 4at p=150kPa,
5p=750kPa, 6at p=400kPa, 7at p=700kPa.
where pa=atmospheric pressure at sea level so that the
equation can be applied with any unit system.

800
5 TYRE CONTACT AREA CALCULATION EXAMPLE

This section contains a worked calculation to illustrate how

From Equation 3 predicted contact area

600 Equations 3 and 4 may be used to estimate the tyre contact

area and stress on an unbound aggregate.

Tyre input data: inflation pressure p=800 kPa, load P=13

400 kN, width§ L=255 mm, radius R=450 mm, νt=0.5.

Unbound aggregate of the same characteristics as the

physical study reported in this paper: Es=12 MPa, νs=0.3.
200
Et=7.4*800+1600 = 7.5 MPa (Equation 4)

1 1−0.52 1−0.32
= + , ∴ 𝐸𝑟 = 5.7 MPa (Equation 3)
0 𝐸𝑟 7.5 12
0 200 400 600 800
From measured contact area 16×0.45×13
𝐵=√ = 0.143 m (Equation 3)
3.14×5700×0.255
Figure 8. Comparison between measured and predicted
(with Equation 3) average tyre contact stresses (kPa). Contact area = 0.143 m x 0.255 m = 0.0365 m2.
Average contact stress = 13 kN / 0.0365 m2 = 357 kPa 7 REFERENCES

Notes: AASHTO (Association of American Highway and

§ - In the case of dual wheels, each wheel is Transportation Officials) 2012. Standard Specification
considered individually as in this example. It would for Classification of Soils and Soil-Aggregate Mixtures
normally be assumed that half the wheel load or quarter the for Highway Construction Purposes. AASHTO M 145.
axle load represents an individual tyre load in the absence Washington, DC: AASHTO.
of more precise information. The tyre width printed on the De Beer, M., Fisher, C. and Jooste, F.J. 1997.
side wall is used to represent approximately the contact Determination of pneumatic tyre/pavement interface
width since the tyre tread width is usually unknown. contact stresses under moving loads and some effects
The calculation approach outlined above provides a on pavements with thin asphalt surfacing layers. Eighth
rectangular contact area more representative of true tyre International Conference on Asphalt Pavements,
contact area shapes measured both in the physical study August 10-14, Seattle, Washington, USA. 1: 179-227.
presented here and elsewhere on rigid surfaces (e.g. De ISBN 8790145356.
Beer et al (2012)). De Beer, M., Maina, J.W., van Rensburg, Y. and Greben,
J.M. 2012. Toward using tire-road contact stresses in
pavement design and analysis, Tire Science and
6 CONCLUSIONS AND RECOMMENDED FURTHER Technology 40(4): 246–271.
WORK Fervers, C.W. 2004. Improved FEM simulation model for
tire-soil interaction, Journal of Terramechanics, 41: 87-
A new method to estimate the tyre contact pressure and 100.
contact area on unbound aggregate layers has been Giroud, J-P and Han, J. 2004. Design method for geogrid-
proposed. This could form the basis of an improved reinforced unpaved roads. I. development of design
unpaved roads design method that takes better account of method, Journal of Geotechnical and
the distribution of traffic loads to the aggregate base course Geoenvironmental Engineering, 130(8): 775–786.
and underlying subgrade. Giroud, J-P and Noiray, L 1981. Geotextile-reinforced
The method uses the Hertzian contact stress equation unpaved road design, Journal of Geotechnical
for the frictionless contact of linear elastic, parallel Engineering 107(9): 1233–1254.
cylinders. The input parameters are the tyre width and Gonzales, C.R. 2015. Development and validation of a
radius, wheel load and tyre and road surface stiffness stress-based procedure for the design of military
properties. A low stiffness was adopted for unbound flexible pavements, PhD thesis, University of Illinois.
aggregate to include the effects of yield under the high tyre Hertz, H 1881, Über die berührung fester elastischer
pressures. The tyre stiffness was found to correlate with Körper, Journal für die reine und angewandte
tyre inflation pressure and a linear relationship was derived Mathematik 92: 156-171.
to estimate tyre stiffness for typical truck tyres and inflation Huang, YH 1993. Pavement Analysis and Design,
pressures. Prentice-Hall International, Englewood Cliffs, NJ.
This approach provided a better match with the Little, P.H. 1992. The design of unsurfaced roads using
behaviour predicted by numerical studies of the problem as geosynthetics, PhD thesis, University of Nottingham.
well as with the results of the physical study described in Powell, W.D., Potter, J.F., Mayhew, H.C. and Nunn, M.E.
this paper. 1984. The structural design of bituminous roads, TRRL
The physical study involved positioning the dual-tyre of Laboratory Report 1132, Transport and Road Research
a special test truck on an unbound aggregate layer with two Laboratory, Crowthorne, UK.
different loads and two different tyre inflation pressures. Robinson, W.J. 2022. Full-Scale Evaluation of Multi-axial
The tyre contact areas were measured by two different Geogrids in Road Applications ERDC/GSL-TR-22-4.
methods that provided similar results. Vicksburg, MS: U.S. Army Engineer Research and
The physical study confirmed that average tyre contact Development Center.
pressures can be significantly less than tyre inflation Roque, R., Myers, L and Ruth, B. 2000. Evaluating
pressure, resulting in larger contact areas than estimated measured tire contact stresses to predict pavement
based on tyre inflation pressure and tyre load alone. response and performance, Transportation Research
Recommendations for further study include: Record, 1716: 73–81.
Taheri, S., Sandu, C., Taheri, S., Pinto, E. and Gorsich, D.
• Measurement of the tyre contact length by the 2015. A technical survey on Terramechanics models
Phase 2 method could be improved by using thinner for tire–terrain interaction used in modeling and
straight edges (such as feeler gauges) to locate simulation of wheeled vehicles, Journal of
more precisely the start and end of the tyre contact Terramechanics, 57: 1-22.
length Thompson, R., Peroni, R. and Visser, A. 2018. Mining Haul
• Further physical studies could be undertaken with Roads Theory and Practice. CRC Press, London.
other tyre sizes and aggregate types to verify the van Vuuren, D.J. 1974. Relationship between tire inflation
approach for a wider range of situations and to pressure and mean tire contact pressure,
improve the accuracy of the tyre and road surface Transportation Research Record, 523: 76-87.
stiffness values. Wang, H., Al-Qadi, I.L. and Stanciulescu, I. 2012.
Simulation of tyre–pavement interaction for predicting
 contact stresses at static and various rolling conditions,
International Journal of Pavement Engineering, 13(4):
310-321.
Xia, K. 2011. Finite element modeling of tire/terrain
interaction: Application to predicting soil compaction
and tire mobility, Journal of Terramechanics, 48: 113-
123.
Yoder, E.J. and Witczack, K.M.W. 1975. Principles of
Pavement Design, 2nd ed., Wiley Interscience, New
York.
Vertical Deformation of Multi-Axial Geogrid
Stabilized Working Platform Design Using
Conventional Techniques and Back-
Analyses Using Composite Approach
Lois G. Schwarz, Ph.D. and Mark H. Wayne, Ph.D. P.E.
Tensar, A Division of CMC, Alpharetta, GA, USA

ABSTRACT
Presented is a comparison of the estimated vertical deformation using conventional techniques for the design of a multi-
axial geogrid stabilized working platform to the results of back-analysis using actual deformation measurements during
operation. Satisfying very stringent criteria for allowable load and differential settlement was the major challenge in
designing this working platform for a large ringer crane. Platform design and construction was 1.8 m thick and composed
of five layers of multi-axial geogrid and aggregate material. The vertical deformation back-analyses of the platform
incorporated a hybrid adaptation of a composite material model for the geogrid and aggregate layers. Composite material
modeling considers the interaction of geogrid and aggregate which results in enhancement of the strength of the composite
material, and this strength enhancement is not considered in typical models for stabilized aggregate layer design. Although
the back-analyses were not rigorous finite element calculations, the hybrid adaptation of the geogrid/aggregate composite
material used with a conventional deformation analysis approach provided results significantly less conservative and more
closely predicted the actual deformation measurements. The comparison of vertical deformation modeling and back-
analyses to initial pre-construction design provides important insight into material modeling and numerical techniques.

RÉSUMÉ
On présente une comparaison de la déformation verticale estimée à l'aide des méthodes classiques pour la conception
d'une plate-forme pour travaux temporaire stabilisée par géogrille multiaxiale. Les résultats de la rétro-analyse ont été
comparés aux mesures de déformation obtenues à partir du projet. Le défi principal consistait à satisfaire aux exigences
strictes en matière de la déformation de la plate-forme et au charge admissible par la grande grue à anneaux. La plate-
forme consistait une épaisseur totale de 1,8 m avec 5 couches de géogrille placées dans un remblai granulaire compacté.
Pour l'analyse, les deux matériaux ont été combinés en un seul matériel composite avec une résistance supérieure à celle
de l'agrégat seul. Cela concentre l'analyse sur l'interaction des deux matériaux entre eux, contrairement aux approches
typiques pour ce type de problème. Bien que les analyses ne soient pas des simulations numériques rigoureuses,
l'utilisation de l'approche des matériaux composites avec l'analyse conventionnelle de la déformation a fourni des résultats
moins conservateurs et correspondant étroitement aux mesures de déformation observées. La comparaison des
déformations prévues et mesurées avec la conception initiale de pré-construction peut donner un aperçu du comportement
des matériaux et des méthodes de modélisation appropriées.

1 INTRODUCTION geogrid, resulting in confinement of the aggregate

particles. In working platforms, mechanically stabilized
The design of a supportive foundation system for a heavy layers provide and increase in bearing capacity and reduce
lift crane includes consideration of the underlying soil deformation which has been verified by full-scale load tests
bearing capacity and the tolerable vertical deformation at (Lees et al., 2019. 2020).
maximum applied load. Geogrid stabilized granular layers Typically, applications with multi-axial geogrid tend to
are often used in working platforms to improve load spread model the soil plus a tensile membrane to simulate the
and bearing capacity on soft soils by confining lateral geogrid. This can lead to significant under-prediction of
deformation and reducing vertical deformation. This paper geogrid performance since only the tension membrane
presents deformation analyses performed to design a functions of the geogrid are considered and parametrically
multi-axial geogrid stabilized working platform using defined as in-air properties without consideration of
conventional techniques and results of deformation back- geogrid/aggregate interaction.
analyses using a hybrid adaptation of geogrid/aggregate
composite material modeling approach. Details of the
working platform design and construction have been 2 PROJECT BACKGROUND
presented by Schwarz and Wayne (2020, 2021).
The project involved the development and construction of
1.1 Aggregate Stabilization with Geogrid a petrochemical facility in southwest Louisiana for Lotte
Chemical. Due to limited available operating space and
The combination of multi-axial geogrid with aggregate extremely heavy lifts for construction of the facility, a PTC-
provides mechanical stabilization to an aggregate layer 200 DS ringer crane was needed.
when aggregate particles interlock with the apertures of the
 The crane footprint travels around on the ring track parameters representative of the soil profile was a rather
system, the footprints of the upper carrier and the wheel daunting task. Based on the information provided by the
positions are illustrated in Figure 1. Operational variances in blow counts, moisture contents, boring
requirements for the ringer crane include stringent observation with depth, and solicitation of local
differential settlement criteria over the circular track geotechnical engineering expertise, assumptions were
constituents. made regarding preconsolidation pressure of the upper
native soils and the overconsolidation ratio for soils below
12 m corresponding to the second layer of CH soil.
Upper carrier footprint Wheel footprint Numerous analyses were conducted to assess the
response various preconsolidation stresses had on
settlement with depth, as well as overconsolidation ratio for
the soils. With concurrence of the professional engineers
on the project, a preconsolidation pressure of 167.6 kPa
was assumed for the native soil to depth 12 m, followed by
constant OCR of 1.8 to estimate settlement for an applied
bearing pressure of 191.6 kPa due to the crane loads.
Collaboration on representative soil property values was
also conducted. To model load redistribution, bearing
Figure 1. PTC-200 DS ringer crane footprint. pressure of 191.5 kPa acted on the geogrid stabilized
platform to depth of 1.8 m (platform thickness) and then
105.3 kPa bearing pressure below 1.8 m depth.
To conventionally model the geometrically irregular
2.1 Soils and Materials
areas for wheel position loading on the track as shown
In the proposed working area, only 1 SPT boring was made previously in Figure 1, a pseudo rectangular area 11.58 m
by 24.38 m was assumed for the crane footprint on the
and indicated problematic soil conditions consisting of fat
loadspreader track ring. Illustrated in Figure 2 is the actual
clays and occasional sandy silt lenses and pockets; more
details have previously been presented (Schwarz and loaded area that revolves on the track footprint and the
Wayne, 2020, 2021). The geogrid stabilized working pseudo rectangular area with the four positions used for
estimating settlement. Three positions were used to
platform was 1.8 m thick with five layers of TX160 multi-
estimate differential settlement compared to the center
axial geogrid and crushed angular graded aggregate
compacted to 98% Modified Proctor. The center of the point of the pseudo rectangular area; these positions were
circular track was unexcavated and used to provide power designated as mid-point short side (MPSS), mid-point long
side (MPLS), and corner.
and drainage for crane operation.

3 ESTIMATED SETTLEMENTS USED FOR DESIGN

Primary vertical deformation or settlement of the

working platform was estimated using two different
software programs due to having variances in their
conventional methods of calculations and inherent
limitations. Dimension Solution software or DSS (Tensar
International) was used to estimate primary settlement
below the centerline of the geogrid stabilized area. SetCalc
software (Yang and Duncan) was used to estimate primary
differential settlement because of having the capability to
provide settlement estimates below any given point for a
distributed applied load and allowing calculation of Figure 2. Pseudo rectangular area and points on model.
differential settlement from point to point. Both software
programs have capabilities for use of Westergaard and Results from DSS analyses provided estimated
Boussinesq methodologies, or a combination thereof. settlement under the centerline of the pseudo rectangular
Consideration of the time constraint for construction start- area for both the geogrid stabilized and unstabilized
up and the three-dimensional geometrical considerations conditions. Results from SetCalc analyses provided
of the ringer crane, finite element analysis was not estimated settlement under the four different points on the
performed. pseudo rectangular area. Settlement analyses were
representative of total primary settlement; estimates of
3.1 Conventional Settlement Analyses consolidation settlement were not performed due to
insufficient data from the single SPT boring within the
A geogrid stabilized platform and an unstabilized platform designated crane platform area and considering the
were initially analyzed to realize the benefit provided by the relatively brief time the crane would be in position and
geogrid. Due to only one boring log and no triaxial testing operation. In addition, assumed consolidation parameters
or consolidation testing results available, selection of soil could compound inaccuracies of settlement estimates.
Consequently, settlement estimates were expected to be expected, it is seen that geometrical influence is evident for
somewhat conservative, realizing actual settlement should comparisons of differential settlement at the various points
be less than estimated and without consideration of on the model area.
construction compaction/traffic. It is noteworthy that neither
DSS nor SetCalc software incorporates FE methodology or
parametric input values directly for modeling stress-strain Table 2. Estimated total and differential settlement
material behavior. Regardless of the software program, all
analyses used identical soil properties, vertical profile and Location SetCalc DSS Average2
applied loads. Shown in Figure 3 are graphical results from
Total ΔSettle Total ΔSettle 1
Total ΔSettle
DSS and SetCalc settlement analyses.
(mm) (mm) (mm) (mm) (mm) (mm)
Center 125.0 - 85.3 - 105.2 -
MPLS 91.4 33.6 - 22.8 76.9 28.3
MPSS 79.3 45.7 - 31.2 66.7 38.5
Corner 51.8 73.2 - 50 43.5 61.7
1
calculated proportion based on SetCalc software results at points
2
averages of SetCalc and DSS software results

3.2 Actual Settlement and Monitoring

The crane manufacturer Mammoet carefully monitored the

various elevational levelness criteria of the platform to
assess any differential settlement throughout the timeline
the ringer crane was in operation. Monitoring included
elevational changes for levelness of the crane track before,
Figure 3. Estimated total settlement of working platform. during and after each lift, and after repositioning the
counterweights with any changes to rigging. A summary of
The curves for MPLS, MPSS, and corner estimated results from differential settlement monitoring is shown in
settlement are the results of SetCalc analyses and based Table 3 and no concerns of levelness occurred.
on the total estimated settlement of 125 mm at the center
point of the model area. Not surprisingly, the geogrid
stabilized platform showed less estimated settlement than Table 3. Differential settlement monitoring1 results
the unstabilized platform for the cases analyzed.
Levelness Tolerable Actual
As expected, predicted results using different methods Criteria Maximum Measured
of analyses can vary due to assumptions and inherent Settlement Settlement
limitations of different methodologies and numerical (mm) (mm)
solvers used in software calculations. Shown in Table 1 is Overall 56 5
a summary of the calculated settlement results at the four (front to rear)
different points on the pseudo rectangular model.
Micro 10 4
(rail girder to loadspreaders)
Table 1. Estimated settlement for crane platform Meso 25 2
(bogies and upperstructure)
Location SetCalc Software DSS Software Macro 53 4
(upper structure and boom systems)
Total Percent Total Percent
Settlement of Center Settlement of Center conducted by Mammoet
1

(mm) (%) (mm)

Center 125 100 85.3 85.3
4 SETTLEMENT BACK-ANALYSES
MPLS 91.4 73.2 - 62.5
MPSS 79.3 63.4 - 54.1 Having actual differential settlement measurements, back-
Corner 51.8 41.5 - 35.3 analyses were performed to determine if composite
material modeling and parameterization would be able to
Summarized in Table 2 are estimated results of closely predict the measured values. Factors including
differential and average settlement from DSS and SetCalc lack of soil information and the geometrical components
analyses. Differential settlement calculations were based not being conducive to 2D analysis albeit plane strain or
on assuming differential settlement is relative to the axisymmetric conditions, back-analyses with 2D FEA were
amount of total settlement estimated, without consideration not undertaken and 3D FEA software was not an available
of time-dependent consolidation behavior within the resource for this project. Alternatively, conventional
timeframe of crane operation which was 6 months. As software capable of 3D visualizations was used and
appropriate composite material parameters for the geogrid modulus, to account for particle confinement. Due to
stabilized platform were implemented for settlement back- Settle3 software limitations and not being capable of robust
analyses. FE modeling and calculations, only the modulus of the
Settle3 software was selected to perform settlement geogrid-stone composite material could be appropriately
back-analyses, defining geometries and analyses in 2D enhanced for the back-calculation of settlement.
with extrusion and visualization capabilities in 3D. Parameters for the clayey and sandy soils used in analyses
Westergaard methodology was selected to perform back- and shown in Table 4 were based on consideration of the
analyses and believed to reasonably represent soft and limited soil boring information available, consideration of
strong stratified soils. Inputs included Young’s modulus compaction effects, and supplemented by literature (Lees
and Poisson’s ratio of the interlayered soil parameters et al., 2020; Bowles, 1968; Plaxis, 2021; Settle3, 2022).
estimated from literature on material testing (Bowles, 1968;
Plaxis, 2021; Settle3, 2022).
The geogrid stabilized platform was modeled using a Table 4. Material parameters used in back-analyses
hybrid adaptation of the Tensar Stabilized Soil Model
(TSSM) by Lees et al. (2019). Parameters for the Value Platform CH CL- CH- CL SM
composite material of geogrid and stone were selected with M CH CL with S dense
from a library developed based on FEA results for large-
γ 21.5 17 16 16 17 17
scale triaxial testing (0.5 m diameter x 1 m high) on stone
samples with geogrid placed at mid-height in the specimen (kN/m3)
(Lees et al., 2019). Shown in Figure 4 is the basis of TSSM γsat 22 18 18 18 18 18
used specifically in FEA modeling, the stabilized soil (kN/m3)
parameters within the influence zone of the geogrid show Ko 1 1 1 1 1 1
the most enhancement in strength and lateral confinement ѵ 0.1 0.16 0.15 0.2 0.17 0.12
at the plane of the geogrid, and transition to partial
Es 150E3 - - - - 40E3
confinement with lesser strength enhancement as the
distance from the geogrid plane increases. This zone of (kPa)
strength enhancement and lateral confinement typically Cc - 0.15 0.066 0.11 0.1 -
ranges from about 200 mm to 400 mm depending on the Cr - 0.015 0.006 0.01 0.01 -
geogrid type and aggregate. eo - 1.2 1.5 1.1 1.5 -
OCR - 1.8 2.5 1 2.5 -
Cv - 1 1 1 1 -
(m2/y2)
Cvr - 1 1 1 1 -
(m2/y2)

To achieve the back-calculated values of settlement,

the modulus value was somewhat higher than typically
thought and reflect heavier compaction and pre-loading by
construction traffic during preparation of the platform. The
compactive effort to achieve 98% Modified Proctor density
at the elevation of the platform excavation prior to
construction of the geogrid stabilized platform would
transfer to increase underlying soil strength and stiffness to
some extent.
The maximum total allowable settlement designated by
Mammoet for crane operation was 100 mm and the actual
Geogrid Plane measured total overall (front to rear of crane) settlement
was 5 mm. The estimated maximum design total
settlement was 85 mm by DSS analysis. Back-analyses of
total maximum settlement was 12.4 mm by SetCalc
software using geogrid stabilized platform composite
material parameters; output is illustrated in Figure 5.
Visualization of back-analyses revealed most settlement
Figure 4. Tensar Stabilized Soil Model (TSSM) occurred within depths of -1.8 m to about -6 m, lying below
representation for geogrid stabilized composite materials. the geogrid stabilized platform in the upper SM soil layer
directly below the water table and into the underlying CL-
CH soil layer.
The TSSM model of a geogrid and aggregate composite
material enhances the strength parameters of the
aggregate, namely the friction angle, cohesion, and
Figure 5. Total settlement at end of crane operation.

Back-analyses of differential settlement along the key

locations of the crane structure corresponding to the crane Figure 7. Micro levelness settlement criteria of rail girder
levelness criteria were determined by numeric queries and to loadspreader.
using associated points on the model footprint as depicted
in Figures 6-9. A summary of results for differential
settlement back-analyses are presented in Table 5 and
show close agreement to the actual measured values.

Figure 8. Meso levelness settlement criteria of bogies and

upperstructure.

Figure 6. Overall settlement criteria of crane front to rear

on foundation track.

Figure 9. Macro levelness settlement criteria of

upperstructure and boom systems.
 modeling approaches can be taken to simulate and predict
Table 5. Differential settlement back-analyses performance of platforms constructed with geogrid
mechanically stabilized aggregate layers when robust FEA
Levelness Back- Actual is not performed.
Criteria Analyses Measured
Maximum Maximum
Settlement Settlement 6 REFERENCES
(mm) (mm)
Bowles, J.E. 1968. Foundation Analysis and Design,
Overall 5.2 5
McGraw Hill, New York, NY, USA.
(front to rear)
Brandes, D. 2018. Personal correspondences, Mammoet
Micro 4.4 4 Fabrication B.V.
(rail girder to loadspreaders) Dimension Solution software (DSS). 2013. Tensar,
Meso 3.3 2 Alpharetta, GA, USA.
(bogies and upperstructure) Lees, A. and Clausen, J. 2019. Strength Envelope of
Macro 4.8 4 Granular Soil Stabilized by Multi-Axial Geogrid in Large
(upper structure and boom systems) Triaxial Tests, Can. Geotech J., Art: cgi-2019-0036,
NCR Research Press.
Lees, A., Dobie, M., Bhawanin, M. 2020. Back-analysis of
5 CONCLUSIONS the Vung Tau Full-scale Trial Using FEA Simulation of
Geogrid Stabilization and the Geocell Mattress,
Due to inherent limitations of software capabilities, a Geotechnics for Sustainable Infrastructure
somewhat unconventional approach to interpreting Development, Lecture Notes in Civil Engineering 62,
software analyses and estimating differential settlement Spring Nature, Singapore, 1119-1126.
was used for geogrid stabilized platform design. The Plaxis. 2021. Plaxis Manuals. Bentley.
geometrically irregular extremities were selected to Schwarz, L. and Wayne, M. 2020. Geogrid Stabilized
estimate differential settlement relative to the loaded Working Platform for Ringer Crane in Southern US,
centerline of the model. Throughout crane operation, Geosynthetics 2020, IFAI, Charleston, SC, USA.
continual monitoring of the elevational levelness of the Schwarz, L. and Wayne, M. 2021. Specialized Multi-axial
crane platform was conducted to determine any differential Geogrid Stabilized Crane Platform Resilience to
settlement and operational concerns. The geogrid Hurricane Harvey, GeoExtreme 2021, Savannah, GA,
stabilized platform showed zero settlement issues. The USA.
conservatism in differential settlement estimates can be SetCalc 1.0 software. 2002. Yang, B. and Duncan, J.M.
due to some combination of estimation of soil parameters, Center for Geotechnical Practice and Research,
a short timeline of crane operation, and software Department of Civil and Environmental Engineering,
limitations. Notwithstanding the numerous assumptions Virginia Polytechnical Institute and State University,
made with juxtaposing the capabilities of the conventional VA, USA.
software used, the modeling method provided a Settle3 software. 2022. Rocscience Inc.
satisfactory design for the multi-axial geogrid stabilized
platform.
Comparison of back-analyses settlement calculations to
measurements of actual maximum settlement using 2D
software with 3D visualization provided results that
corresponded well with the measured total and differential
settlement values. Back-analyses did indicate that soil
moduli parameters were slightly higher than originally
interpreted and may be attributed to the impact of
excavation and compaction of the base soil and
construction traffic. Accordingly, consideration of staged
construction and compaction effects on material
parameterization help lessen conservatism of deformation
estimates and are typically performed in FEA calculations.
However, with less robust traditional deformation analyses,
caution must be exercised in material parameter selection
to restrain analyses from becoming unconservative along
with the risks involved with minimal soil data and software
capabilities. Defining the geogrid stabilized platform as a
composite material and using a hybrid adaptation of TSSM
stabilized aggregate parameters developed by and for use
with FEA was advantageous to improving traditional
calculation predictions while maintaining confidence in not
becoming unconservative. Similar composite material
Monday, October 2, 2023

TAILINGS AND
LIQUEFACTION
Soil-Simulator scripts: Bridging the gap for
beginners in soil testing simulations
Barrero A. R, Cueto I., Lino E., Ledesma O. N., & Lizcano A.
SRK Consulting, Vancouver, BC, Canada
Reyes D
SRK Consulting, Reno, NV, United States

ABSTRACT
Soil testing is an essential component of geotechnical engineering, as it helps engineers determine soil properties and
assess its behavior under different loading conditions. In recent years, the use of numerical simulations has become a key
component, as it allows engineers to predict the behavior of soil in complex situations without the need for expensive
physical testing. One widely-used numerical platform in the geotechnical community and industry is the finite difference
numerical software FLAC3D. However, using numerical platforms can be challenging, especially for beginners who are
not familiar with the programming languages. To address this issue, the Soil-simulator scripts were developed by SRK
Consulting to provide an open source and convenient solution for simulating basic and advanced soil lab tests using a
single element. These scripts do not require any prior knowledge of Python or FISH, the programming languages of
numerical platform, making it accessible to both beginners and advanced users. To demonstrate the capabilities of the
Soil-Simulator scripts, a series of examples were conducted using a cube element subjected to monotonic and cyclic,
drained and undrained compression using conventional triaxial and direct simple shear configurations. Numerical
simulations results were compared with available experimental laboratory results. Overall, the paper provides a clear and
concise introduction to the use of Soil-Simulator scripts to simulate the soil responses at the element test level under
various loading and drainage conditions, allowing for the assessment of characteristic properties of soils.

RÉSUMÉ
L'analyse du sol est un élément essentiel du génie géotechnique, car elle aide les ingénieurs à déterminer les propriétés
du sol et à évaluer son comportement dans différentes conditions de charge. Ces dernières années, l'utilisation de
simulations numériques est devenue un élément clé, car elle permet aux ingénieurs de prédire le comportement du sol
dans des situations complexes sans avoir besoin de tests physiques coûteux. Une plate-forme numérique largement
utilisée dans la communauté géotechnique et l'industrie est le logiciel numérique à différences finies FLAC3D. Cependant,
l'utilisation de plateformes numériques peut être difficile, surtout pour les débutants qui ne sont pas familiers avec les
langages de programmation. Pour remédier à ce problème, les scripts du simulateur de sol ont été développés par SRK
Consulting pour fournir une solution open source et pratique pour simuler des tests de laboratoire de base et avancés sur
le sol en utilisant un seul élément. Ces scripts ne nécessitent aucune connaissance préalable de Python ou de FISH, les
langages de programmation de la plateforme numérique, ce qui les rend accessibles aux débutants et aux utilisateurs
avancés. Pour démontrer les capacités des scripts du simulateur de sol, une série d'exemples a été réalisée en utilisant
un élément cubique soumis à une compression monotone et cyclique, drainée et non drainée, à l'aide de configurations
conventionnelles de cisaillement triaxial et de cisaillement simple direct. Les résultats des simulations numériques ont été
comparés aux résultats expérimentaux de laboratoire disponibles. Globalement, l'article fournit une introduction claire et
concise à l'utilisation des scripts du simulateur de sol pour simuler les réponses du sol au niveau des essais sur élément
dans diverses conditions de charge et de drainage, permettant l'évaluation des propriétés caractéristiques des sols.

1 INTRODUCTION models have been implemented in numerical platforms.

Nowadays there are numerous numerical platforms, that in
In recent years, the fast-paced evolution of computers has overall they share the same basic loop. Which consist of 1)
allowed numerical methods to achieve notable results in invoking the equations of motion to derive with the
solving complex problems in geotechnical engineering. A appropriate numerical method (e.g., FEM, FDM, MPM), 2)
wide range of problems are analyzed using the finite the new set of velocities and displacements from stresses
difference method (FDM), the finite element method (FEM), and forces, 3) the constitutive model takes the new set of
and more recently, the material point method (MPM). A key velocities as the input to solve the new state of stresses, 4)
component in the application of these methods is the that are updated and constitute the new input to solve the
constitutive relationship between strains and stresses. equilibrium equation.
Currently, there is no universally applicable constitutive The previous basic loop scheme provides a description
equation to reproduce the soil behavior under any type of in which the results obtained from a practical analysis are
loading. Instead, a large number of constitutive models are mainly driven by the selected constitutive model.
available tailored for specific type of loads, distinct ranges, Therefore, it is key for users to understand deeply the
and various soil materials. Naturally, several constitutive constitutive model response under different types of
loading, drainage conditions, initial densities, or this can be modified inside the code to meet the modelers
confinement. This type of task requires a numerical setup needs.
where the test can be easily simulated on the element test
level, and depending on the numerical platform, this can be 2.1 Triaxial Monotonic Compression
done using a single Gauss point or one single element.
Then, the main idea was developing a series of friendly In the simulation of a conventional triaxial tests, the axial
user soil simulator scripts to simulate the majority of the and radial directions of the cylindrical specimen are
laboratory soil tests. Such that advance and especially new assumed to be principal axes of the stresses and strains.
modelers can have a complete understanding of coding The axial strain (𝜀𝑎 = 𝜀𝑧 ) as well as the radial stress (𝜎𝑟 =
concepts and assess the numerical model response. The 𝜎𝑥 = 𝜎𝑦 ) are used to control the simulation test. The
developed scripts also allow users to compare the effective confining stress, 𝜎𝑟 , and the coefficient of earth
simulation results with laboratory data, if available. The pressure at rest, 𝐾0 , are the main inputs to initialize the
series of developed scripts are described and discussed in stress tensor. Notice that 𝐾0 = 1 represents a sample
this paper to narrow the gap between numerical platforms isotopically consolidated and 𝐾0 = 0.5 represents a 𝐾0
and new modelers. consolidation. The input of the maximum axial strain
percentage depicts the desired level of strain that is
2 STRUCTURE OF THE SOIL SIMULATOR SCRIPTS reached in shearing process.
To set up the drainage condition, the user has the
In overall, the soil simulator scripts were aimed to simulate possibility to choose from three different options. The
element tests of triaxial monotonic and cyclic compression, ‘Drained’ option is suitable for a type of loading with no
monotonic and cyclic direct simple shear (DSS), and excess of pore water pressure and no presence of the
oedometer compression. Each test can be performed for water fluid. The ‘undrained’ option enables the water fluid
both drainage conditions: drained or undrained. Each type presence, and it is allowed to generate excess of pore
of test requires a set of inputs, beyond the model water pressure. In this option, the bulk modulus of the
parameters, that should be define by the user to complete water is set to 2.0e6 kPa. The drainage option ‘constant
the test. The scripts were developed within the three- volume’ is meant to reduce the computing time that it takes
dimensional non-linear finite difference platform Fast to run an ‘undrained’ drainage simulation. In this option
Lagrangian Analysis of Continua (FLAC3D) by Itasca there is no presence of water fluid so the excess of pore
(2018), that uses a cubic sample to simulate element tests. pressure cannot be directly calculated. Instead, the soil is
The faces of the element test have been restricted to move sheared under an isochoric configuration, that is when the
via strain controlled in the principal strain and shear volumetric strain remains null throughout the shearing
components. This allows modelers to simulate the majority process. The results obtained using the constant volume
types of tests, e.g., conventional triaxial compression, option are equivalent to those obtained with the undrained
direct simple shear tests and oedometer tests. The option.
availability of the soil constitutive models is dependent on
the numerical platform. In this case, it includes most typical 2.2 Triaxial cyclic compression
soil models used in the industry. To facilitate the post
processing at the end of each computation, the soil The triaxial cyclic compression keeps the same model
simulator scripts automatically print each component of the configuration as in the monotonic compression, keeping
stress and strain tensor and pore pressure, if available. the strains and stress in the principal axes. Beyond, the
Moreover, a series of figures are plotted and saved in the previous inputs 𝜎𝑟 , 𝐾0 and drainage conditions. The user
output directory to quickly check the simulation results. must specify if the test is carried out using stress-controlled
In particular, the scripts also offer the possibility to set limits or strain-controlled limits. The stress-controlled
up multiple simulations that can be performed under defined by the cyclic shear stress ratio (CSR) is suitable for
different sets of initial state parameters, different types of tests that require to control the level of applied deviatoric
drainage conditions and different loading configurations e. stress. Analogously, the strain-controlled, defined by 𝜀amp ,
g. triaxial compression followed by an oedometer controls the amplitude of the applied axial strain. For either
consolidation. Once the computations are completed, the both, stress and strain controlled, the user must specify the
results can be displayed together with the experimental number of cycles, 𝑁𝑐𝑦𝑐 .
laboratory results. This framework can be a useful tool to
validate the simulation results and the laboratory test
2.3 DSS monotonic and cyclic compression
results. A very common process during the model
parameters calibration. The soil simulator scripts presented
The direct simple shear test loading conditions are closely
in this study can be downloaded in the home page
to a plane strain and allow for the principal axes of stresses
(https://cdn-web-
and strains to rotate. In simple shear, the shear strain 𝛾𝑦𝑧 ,
content.srk.com/upload/user/image/PyScript20221019161
618562.zip?_ga=2.237785728.1478049183.1686087123- the vertical effective stress 𝜎v′ , and 𝐾0 , are the main inputs
224853027.1659637604). used to control the simulation test. The simulation can be
In the following sections a detailed description for each performed drained, undrained or under constant volume,
type of simulated laboratory test is given followed by an accordingly to the selected drainage option. Similar to the
applied example. Is important to mention the default triaxial cyclic compression, the cyclic DSS compression
configuration system of units for stress is kPa. However, can be also performed with stress or strain controlled, for
the designated number of cycles, 𝑁𝑐𝑦𝑐 .
 al. 2015). The 12 primary parameters capture the soil
2.4 Oedometer compression response for a wide range of confinement pressure and soil
density. The model parameters shown in Table 2 are
The one-dimensional consolidation test, also named validated for a series of laboratory triaxial tests performed
oedometer compression, is controlled defining the on sand tailings material from Fundão (Morgenstern et al.
sequence of incremental loads 𝜎v′ , including unloading 2016).
sequences.
Table 2 NorSand model parameters
3 APPLICATION EXAMPLES
Parameter Unit Value
The soils simulator scripts offer a wide range of laboratory
Shear constant, 𝐺0 kPa 20000
tests that can be used in multiple ways. In this section an
example is presented for each laboratory compression. Elastic exponent, 𝑚 - 0.4
The selected constitutive models are the hardening soil Volumetric coupling factor, 𝑁 - 0.38
model (Schanz et al. 1999) and NorSand model (Jefferies Poisson ratio, 𝜈 - 0.3
1993). Stress ratio at critical state, 𝑀 - 1.33
Critical state parameter, 𝐶1 - 0.8
3.1 Hardening soil (HS)
Critical state parameter, C2 - 0.03
The hardening soil and its extended version with small Critical state parameter, C3 - 0.65
strains (Benz 2006) constitutive model is an effective stress Dilatancy factor, 𝜒𝑡𝑐 - 7.3
hardening plasticity model. That can capture the behavior Plastic hardening, 𝐻0 - 156
of geomaterials undergoing plastic compression,
Second plastic hardening, 𝐻𝑦 - 756
consolidation, and monotonic shear. The model is capable
of reproducing: increase in stiffness and shear strength OCR - 1
(undrained and drained) with confining pressure; small-
strain elasticity; prefailure hardening plasticity with
hyperbolic stress-strain response; hysteretic damping (no 3.3 HS triaxial monotonic compression
need to resource to Rayleigh damping); and compression
plasticity allowing for the simulation of primary A triaxial undrained simulation is performed using the
consolidation and compression stress-paths. The model model parameters in Table 1. The sample is isotopically
has 7 primary parameters that allows the applicability to consolidated using a confinement pressure of 100 kPa and
model the behavior of both fine-grained ad coarse grained. 𝐾0 = 1. The test is sheared until the level of strain reaches
The model parameters were assumed and summarized in 20% of axial strain. Once the computations are completed
Table 1. the soil scripts automatically will create two files with the
model outputs and figures. The simulation results of the
Table 1 Hardening soil model parameters stress and strain tensors are compiled and summarized in
the model output. The results plotted in Figure 1, shows the
Parameter Unit Value deviatoric stress (𝑞 = 𝜎𝑧 − 𝜎𝑟 ) and axial strain (𝜀𝑎 ), stress
path (𝑞 − 𝑝′), volumetric strain (𝜀𝑣 = 𝜀1 + 𝜀2 + 𝜀3 ) and
OCR - 1
axial strain (𝜀𝑎 ), and excess of pore water pressure (Δu)
Unloading-reloading stiffness, kPa 90000
𝑟𝑒𝑓 and axial strain (𝜀𝑎 ). The black dashed line in the stress
𝐸𝑢𝑟 path, represents the critical state line in triaxial
Secant stiffness, 𝐸50
𝑟𝑒𝑓 kPa 30000 compression with a slope of 𝑀 = 1.2.
Tangent stiffness,
𝑟𝑒𝑓
𝐸0 kPa 30000
Elastic exponent, 𝑚 - 0.5
Friction angle, 𝜙  30
Reference pressure, 𝑝𝑟𝑒𝑓 kPa 100

3.2 NorSand

Norsand it is an elastoplastic model that was developed

within the critical state soil framework. The model develops
a yield surface that evolves with the associate or non-
associate plastic flow. The model is suitable reproducing
the behavior of clean sands subjected to plastic
compression, consolidation, and monotonic shear under a
triaxial configuration. Recently, the softening effect of
principal stress rotation was incorporated into the model as
an effort to improve predictions of cyclic DSS (Jefferies et
 3.5 NorSand triaxial cyclic compression

A cyclic triaxial undrained compression under the strain

limits of 𝜀𝑎𝑚𝑝 = 1% is sheared for 5 cycles (𝑁𝑐𝑦𝑐 = 5) using
NorSand. The capability of this model to reproduce soil
softening is key to a have better understanding of the cyclic
triaxial results. The sample isotopically consolidated (𝐾0 =
1) to a confinement pressure of 𝜎𝑟 = 100 kPa, is sheared
using the model parameters shown in Table 2. The initial
void ratio is assumed 0.8. The model outputs remain the
same to those previously mentioned in Section 3.3. The
black dashed line in the stress path, represents the critical
state line in triaxial compression with a slope of 𝑀 = 1.33,
and 𝑀𝑒𝑥𝑡 = 0.92 in triaxial extension.

Figure 1 Simulation results of a triaxial compression using

hardening soil

3.4 HS DSS monotonic compression

Using the same model parameters in Table 1, a DSS

undrained monotonic compression is simulated using a
vertical pressure of 100 kPa and consolidated with 𝐾0 =
0.5. The test is monotonically sheared until the sample
deformation reaches 20% of shear strain. Upon
completion, the model outputs and figures are saved in the
output folder. The simulation results for the DSS are shown
in figure 1 in terms of the shear stress (𝜏𝑦𝑧 ) and shear
strain (𝛾𝑦𝑧 ), stress path (𝜏𝑦𝑧 − 𝜎′v ), volumetric strain and
shear strain (𝜀v − 𝛾𝑦𝑧 ) and excess of pore water pressure
and shear strain (Δ𝑢 − 𝛾𝑦𝑧 ).
Figure 3 Simulation results of a cyclic triaxial compression
using NorSand

3.6 NorSand DSS cyclic compression

Keeping the same model parameters in Table 2, a cyclic

DSS undrained compression is carried out using a strain-
controlled amplitude of 𝜀𝑎𝑚𝑝 = 1%, to shear the soil during
5 cycles. Prior to the cyclic shearing, the soil is initially
consolidated using a vertical pressure of 100 kPa, 𝐾0 = 0.5
and the initial void ratio of 𝑒0 =0.8. The model figure
outputs for cyclic DSS remain the same to those previously
described in Section 3.4

Figure 2 Simulation results of a DSS compression using

hardening soil
 Often engineers are encounter with the iterative task to
calibrate model parameters. In this process, laboratory
tests are the main input from which the modeler has to best
fit the constitutive model parameters to match the
laboratory test results. The soil simulator scripts have
embedded a framework to help the modelers in the
calibration exercise. In this example, the model NorSand
has been calibrated for the available laboratory triaxial
compression tests from the report on the immediate causes
of the failure of the Fundão dam (Morgenstern et al. 2016).
A combined series of 4 undrained triaxial compression and
9 drained triaxial compression were simulated using the
model parameters in Table 2. The experimental triaxial test
were uploaded in the data folder. A summary table of the
test ID, drainage condition, initial density and confinement
pressure is shown in Table 3.

Table 3 Triaxial tests summary from Fundão sand tailing

Test Drainage Initial void ratio, Confinement

ID condition e0 pressure, p (kPa)
Figure 4 Simulation results of a cyclic DSS compression
TX-01 Drained 0.808 300
using NorSand
TX-02 Undrained 0.838 200
TX-04 Undrained 0.3796 600
3.7 HS Oedometer compression TX-06 Undrained 0.804 300
TX-08 Drained 0.728 300
The Oedometer compression the soil is consolidated under TX-10 Drained 0.595 600
𝐾0 conditions for the established loading steps, 𝜎v′ = {10,
TX-11 Drained 0.611 200
100, 200, 400, 800, 400, 200, 400, 800, 1600, 3200, 1600,
800} (kPa). Noticed there are two unloading-reloading TX-12 Drained 0.69 400
instances. The compression is simulated under drained TX-13 Undrained 0.689 600
conditions and using the model parameter of Table 1. The TX-18 Drained 0.809 100
figure of the volumetric strain against the vertical effective TX-19 Drained 0.882 50
stress (𝜀𝑣 − 𝜎v′ ) is shown in Figure 5. TX-29 Drained 0.741 400
TX-30 Drained 0.748 200

The numerical tests were sheared until 30% of axial strain.

The simulations results and figures are saved individually
for each test ID. Figure 6 shows the comparison between
the numerical simulations and experimental results for the
selected tests ID’s 04, 11, and 13.

Figure 5 Simulation results of an oedometer compression

using HS

3.8 Calibration Set Up environment - Fundão

 Benz, T. 2006. Small-strain stiffness of soils and its
numerical consequences. Ph.D. Thesis. Institut für
Geotechnik, Universität Stuttgart.
Itasca Consulting Group, Inc. 2018. FLAC: Fast
Lagrangian Analysis of Continua, Version 7.0.,
Minneapolis, Minnesota.
Jefferies, M. G. 1993. Norsand: a simple critical state
model for sand. Geotechnique. 43: 91-103.
Jefferies, M. G., Shuttle, D. A. and Been, K. 2015. Principal
stress rotation as cause of cyclic mobility. Geotechnical
research, 2(2): 66-96.
Morgenstern, N. R., Vick, S. G., Viotti, C. B., and Watts, B.
D. 2016. Fundão tailings dam review panel: Report in
the immediate causes of the failure of the Fundão Dam.
New York: Cleary Gottlieb Steen & Hamilton LLP.
Schanz, T., Vermeer, P. A., and Bonnier, P. G. 1999.
Formulation and verification of the hardening-soil
model. Beyond 2000 in computation. R. B. J.
Brinkgreve, ed., Balkema Rotterdam, Netherlands:
281-290.

Figure 6 Fundão sand tailings triaxial compression.

Experimental data in dotted lines from Morgenstern et al.
(2016) and corresponding numerical simulations using
NorSand.

4 CONCLUSIONS

The soil simulator scripts for simulation of common

laboratory soil tests have been presented.
This paper describes idealized laboratory soils tests,
e.g., triaxial monotonic and cyclic compression, DSS
monotonic and cyclic compression, and oedometer
compression. Each test can be performed either drained,
undrained, or under constant volume, depending on the
modeler needs. Furthermore, the scripts offer an advanced
opportunity to concatenate multiple tests, i.e., performed
re-liquefaction instances that requires multiple stages of
undrained cyclic shearing followed by a drained
consolidation. The soil simulator scripts simulation data
results are automatically save for future consultation or
easy shearing. The conventional plots for post-processing
offer a quick opportunity to check the simulation results.
The soil simulator scripts also offer efficient framework
design specifically to calibrate constitutive parameters.
This includes a simple way to upload the laboratory tests
that later can be programmed to simulate as many tests as
needed. The scripts facilitate the iterative process of
comparing the numerical and laboratory results for model
validation. The soil simulator scripts would, indeed, be an
important step to facilitate the numerical modelling process
of conventional soils tests, which, hopefully, also can
induce new users to perceive the benefits of using
advanced computational tools.

5 REFERENCES

Terzaghi, K. and Peck, R.B. 1987. Soil Mechanics in

Engineering Practice, 2nd ed., McGraw Hill, New York,
NY, USA.
Magnitude scaling relationships for
liquefaction triggering procedures
considering large magnitude subduction
zone earthquakes
Yassein Keshty
Graduate student researcher, Department of Civil Engineering
University of British Columbia, Vancouver, BC, Canada,
[email protected]

Trevor J. Carey
Assistant Professor, Department of Civil Engineering
University of British Columbia, Vancouver, BC, Canada,
[email protected]

ABSTRACT
Along the coast of North America, including British Columbia, the Cascadia Subduction Zone is likely to produce a large-
magnitude earthquake ranging from Mw =8-9.1, which is expected to trigger soil liquefaction, resulting in the loss of stiffness
and strength and lead to ground failures of settlement or lateral movement. Within the liquefaction triggering procedures,
the Magnitude Scaling Factor (MSF) is the relationship that accounts for the additional number of loading cycles for larger
magnitude earthquakes. Current MSF relationships were developed for earthquake magnitudes consistent with crustal
faulting (Mw<8.0). Hence, it is not fully understood if these relationships can confidently predict intra- and inter-plate
earthquakes produced by Cascadia. In this study, a suite of ground motions from the NGA-subduction zone database is
used to assess if the current magnitude versus the number of cycle relationships (Mw-Ncyc) correctly predicts the significant
number of loading cycles for intra- and inter-plate earthquakes. The current Mw-Ncyc relationships overpredicted the number
of cycles for subduction zone inter-plate events at midrange magnitudes, while intra-plate events produce more significant
loading cycles than crustal or inter-plate subduction zone relationships predicted at the same magnitude. The revised Mw-
Ncyc relationships are recommended to be used in design practice when dealing with subduction zone tectonic
environments both locally and abroad.

RÉSUMÉ
Le long de la côte de l'Amérique du Nord, y compris la Colombie-Britannique, la zone de subduction de Cascadia est
susceptible de produire un tremblement de terre de grande magnitude allant de M w = 8 à 9,1, qui devrait déclencher la
liquéfaction du sol, entraînant une perte de rigidité et de résistance et de plomb. aux défaillances du sol de tassement ou
de mouvement latéral. Dans les procédures de déclenchement de la liquéfaction, le facteur d'échelle de magnitude (MSF)
est la relation qui tient compte du nombre supplémentaire de cycles de chargement pour les tremblements de terre de
plus grande magnitude. Les relations MSF actuelles ont été développées pour des magnitudes de séisme compatibles
avec des failles crustales (Mw<8,0). Par conséquent, il n'est pas entièrement compris si ces relations peuvent prédire avec
confiance les tremblements de terre intra- et inter-plaques produits par Cascadia. Dans cette étude, une suite de
mouvements du sol de la base de données de la zone de subduction NGA est utilisée pour évaluer si l'amplitude du
courant par rapport au nombre de relations cycliques M w-Ncyc prédit correctement le nombre important de cycles de
chargement. Il a été constaté que les relations M w-Ncyc actuelles surestiment le nombre de cycles pour les événements
inter-plaques de la zone de subduction à des magnitudes moyennes et que les événements intra-plaque pourraient
produire des cycles de chargement plus importants que les relations crustales ou inter-plaques à la même magnitude. Il
est recommandé d'utiliser les relations M w-Ncyc révisées dans la pratique de conception lorsqu'il s'agit d'environnements
tectoniques de zone de subduction à la fois localement et à l'étranger.

1 INTRODUCTION history data for lower-magnitude earthquakes because

they occur more frequently. Few studies have evaluated
The Cascadia Subduction Zone (CSZ), along the liquefaction in soils from large-magnitude earthquakes.
Pacific coast of North America, is likely to produce a Due to the longer shaking duration associated with large-
large-magnitude earthquake measuring Mw=8-9.1. Soil magnitude earthquakes, liquefaction may be triggered in
liquefaction is anticipated, triggering loss of stiffness and deposits not typically considered at risk.
strength, and this may cause ground failures of In this paper, we compare the current magnitude
settlement or lateral movement. Since the beginning of versus the number of equivalent uniform loading cycle
the 20th century, there have been 95 earthquakes relationships (Mw-Ncyc) using the NGA-subduction
between Mw=8-8.9 globally and only five earthquakes database, which consists of recorded earthquake motions
Mw=9 or larger. The engineering community’s knowledge with magnitudes that are expected from Cascadia. The
base for liquefaction was primarily developed using case NGA-sub database includes over 214,020 individual
ground motions records for seven global regions (Alaska, (2019) refined the functional dependency MSFs have on
Cascadia, Central America and Mexico, Japan, New soil types by analyzing weighting curves for a range of
Zealand, South America, and Taiwan) of both Intra- and fine-grained soils to obtain representative estimates of
inter-plate earthquakes (Mazzoni 2022). The Mw-Ncyc the exponent slope of the cyclic strength curve (i.e., the b
relationships account for the duration of cyclic loading value).
through the conversion of irregular earthquake-induced Stuedlein et al. (2023) developed MSFs, specifically
ground motions to an equivalent number of harmonic for subduction zone earthquakes, after analyzing 429
stress cycles (Ncyc), which is an essential component in two-component ground motion records. Their study
liquefaction triggering procedures. Existing models were utilized a larger dataset of motions recorded in different
developed predominantly using ground motions from geographical regions to examine the relationship
crustal tectonic conditions with magnitudes often lower between Ncyc and Mw. However, their study did not
than 8.0. Furthermore, more recent Mw-Ncyc relationships differentiate between inter-plate and intra-plate
do not differentiate between inter-plate or intra-plate subduction zone earthquakes.
subduction zone events. We conclude this study by
comparing the findings with other Mw-Ncyc relationships, 3 NGA-SUBDUCTION DATABASE
which have considered varying degrees of subduction
zone earthquakes in the formulation. The Next Generation Attenuation Subduction (NGA-
sub) project is a comprehensive database of processed
2 BACKGROUND ground motions from subduction zone earthquakes. It
includes 71,340 three-component ground motion records
Accounting for the variation of Ncyc from different from seven global regions (Alaska, Cascadia, Central
earthquake magnitudes was introduced in the liquefaction America and Mexico, Japan, New Zealand, South
triggering procedures by Seed and Idriss (1971). Seed America, and Taiwan) from 1,880 unique earthquake
and Idriss (1982) later refined the Mw-Ncyc relationship events. The database is a continuation of the NGA
and normalized earthquake loading to an Mw =7.5 event research programs started in 2003 (NGA-West1, NGA-
when proposing their updated triggering relationship that West2, and NGA-East) with the primary goal of
used standard penetration test (SPT) blow counts and developing ergodic Ground Motion Models (GMMs).
cyclic stress ratio (CSR) (i.e., cyclic stress normalized by
initial vertical effective stress) for liquefaction 3.1 Ground motion selection criteria
assessment. Liquefaction case history data was
supplemented with laboratory element testing data to The NGA-sub database contains ground motion
establish the relationship between CSR and the number records from a broad range of conditions (e.g.,
of cycles of loading and derive the first magnitude scaling earthquake magnitudes, site conditions, and PGAs).
factors (MSFs). MSFs account for different earthquake Therefore, it includes earthquake records that are not of
magnitudes and the number of loading cycles in interest or may bias study findings. 724 two-component
engineering practice (Seed and Idriss 1982 & Idriss (vertical component neglected) ground motion records,
1999). representing 1,448 individual records, were considered
Alternatively, several researchers proposed for this study. The ground motion selection criteria are
deriving MSF relationships solely from case histories described below:
(Ambraseys 1985, Andrus and Stokoe 1997). Their
findings indicated that MSFs derived from laboratory data • Time history records with Peak Ground
are low when earthquake Mw < 7.5 and high when Mw > Accelerations (PGAs) lower than 0.11g were not
7.5. To address this discrepancy, Liu et al. (2001) considered for consistency with established MSF
developed an empirical regression model to estimate Ncyc formulations (Boulanger and Idriss 2015) and to
that has functional dependency on moment magnitude exclude records with low intensity shaking that is
(Mw) and rupture distance (Rrup). Their analysis included unlikely to trigger liquefaction even at long shaking
107 unique earthquakes, all of which had magnitudes durations.
less than 7.6. Liu et al. (2001) showed Ncyc has a strong • Earthquake time histories were from recordings
dependence on Mw and a secondary dependence on Rrup. made at soil sites with time-averaged shear wave
Kishida and Tsai (2014) analyzed more than 3500 ground velocities in the top 30 meters (Vs30) ranging from
motions to develop a model for Ncyc that incorporates into 180 m/s < Vs30 < 360 m/s or site class D.
probabilistic seismic hazard analysis (PSHA). The • No selection criteria were enforced on rupture
regression model developed during their study has distances (Rrup) for two reasons: (1) to allow for
functional dependency on Mw and Rrup. studying the effects of Rrup on the Mw-Ncyc
Boulanger and Idriss (2015) examined the fatigue relationship, (2) the CSZ is approximately 1000 km
characteristics of different soil types from cyclic testing long and source-to-site distance will vary during a
and updated the MSF relationship for SPT- and CPT- large magnitude earthquake. However, distant
based liquefaction triggering procedures. They ground motion recordings will generally be
highlighted the importance of verifying current MSF attenuated and not meet the 0.11 PGA threshold.
models for large-magnitude earthquakes, and they
examined potential dependency on other parameters
such as source distance and site conditions. Verma et al.
3.2 Earthquake types Figure 2 are from the Japan and South America regions,
as records from the other regions did not meet the
The CSZ (Cascadia Subduction Zone) has the selection criteria or did not have earthquake recordings
potential to produce two types of earthquakes that with magnitudes between 8.2-9.2. Figure 3 provides the
contribute to the seismic hazard in British Columbia response spectra for the intra-plate events for the
(Halchuk et al. 2019). Inter-plate (Interface) events tend magnitude ranges of 6.5 to 7.0 and 7.0 to 7.5,
to be higher in magnitude and are responsible for the representing 19 and 15 different earthquakes,
largest earthquakes globally (e.g., 2010 M8.8 Maula, respectively. Ground motions from the Alaska, Cascadia,
Chile and 2011 M9.1 Tohoku, Japan earthquakes). Intra- Japan, New Zealand, and Taiwan regions are
plate (slab) events, while generally of lower magnitude, represented in Figure 3. The range of earthquake
exhibit a higher recurrence rate and contribute magnitudes in Figures 2 and 3 brackets the predicted
significantly to the overall seismic hazard. Figure 1 magnitude for an intra- and inter-plate from the CSZ
compares ground motion time histories from a typical (Halchuk et al. 2019).
crustal, subduction intra-plate, and subduction inter-plate
earthquake. The inter-plate earthquake has the longest 4 COMPUTING NCYC FROM TIME HISTORIES
shaking duration of the three motions in Figure 1. The Mw-
Ncyc was examined for each type of subduction zone Methods for converting the complex frequencies of
earthquake and compared with established relationships earthquake time histories to uniform equivalent cycles
for crustal earthquakes (Kishida and Tsai 2014, have been studied by other researchers. Three aspects
Boulanger and Idriss 2015). Subdividing M w-Ncyc are typically considered for these procedures: (1) The
relationships by earthquake types reduces aleatory methods for counting peak cycles, (2) the selection of a
variability from using one function to describe all reference stress level, and (3) the application of cut-off
earthquake types. criteria for low-amplitude cycles. In this study, the
maximum amplitude between each crossing of zero
acceleration is a half-cycle. Prior studies on cycle
counting have shown little to no effect on the N cyc with
variations on how half and peak cycles are counted. N cyc
is computed using Eqn. [1],

1 𝐶𝑆𝑅𝑖 −1
𝑁𝑐𝑦𝑐 = ∑𝑚
𝑖=1(𝐶𝑆𝑅 )𝑏 [1]
2 𝑟𝑒𝑓

where Ncyc is the number of equivalent cycles, m is the

total number of cycles, CSRref is the reference stress
level, CSRi is the CSR of cycle i, and b is the slope of the
cyclic strength curve. CSRref was taken as 65% the
maximum acceleration (0.65amax), consistent with the
work of other researchers, the current state of practice,
and cyclic strength relationships in the liquefaction
triggering procedures. The b value is the slope of the
power-law cycle strength curve and typically ranges from
0.04 to 0.4. Values of b vary with soil type, relative
density, plasticity, and stress history and are correlated to
penetration resistance obtained from in-situ testing
(Boulanger and Idriss, 2015). In this study, Ncyc
calculations were repeated for b values ranging from 0.04
to 0.4 to represent a range of soil types. Eqn. [1] does not
include a cut-off criterion to remove low amplitude cycles
because it had minimal effect on the Ncyc, with this finding
supported by other studies (Kishida and Tsai 2014).

4 RESULTS
Figure 1. Comparison of typical ground motions from The following section summarizes the relationship
crustal and subduction earthquakes. between Ncyc and Mw for inter- and intra-plate subduction
zone earthquakes.
The response spectra for a subset of the intra- and
inter-plate motions used in this study are shown in 4.1 Ncyc for Inter-plate events
Figures 2 and 3. Figure 2 provides the response spectra
for the inter-plate event for a magnitude range of 8.2 to Summarized in Figure 4 is the Ncyc for the b values of
9.0 and 9.0 to 9.2, representing 8 and 1 unique 0.04-0.4 for the two magnitude bins shown in Figure 2. In
earthquakes, respectively. The ground motion records in Figure 4a, the heavy red line represents the geometric
mean Ncyc for the bin where 8.2 < Mw < 9.0, accompanied Figures 5b and 5c – is significant for both magnitude bins
by the Ncyc data points for each ground motion record at compared to the inter- plate events in Figure 4. The
the corresponding b values. At the b values of 0.14 and variation observed in Figure 5 for different regions can be
0.34, the geometric mean Ncyc are 34 and 22, attributed, in part, to a higher representation of regions
respectively. The disaggregation of the individual within the database at the lower magnitudes. For the 6.5
geographical regions is given in Figure 4b, and the Ncyc < Mw < 7.0 bin in Figure 5b, the variation between the
curves are similar over the complete range of b values for geometric means at a b value of 0.4 is a factor of 2.2 for
the two regions. In Figure 4c, the geometric mean is Ncyc, the different regions, with the difference increasing to 3 in
given along with the individual data points for the Mw = the 7.0 < Mw < 7.5 bin in Figure 5d.
9.0-9.2 bin, with Ncyc of 43 and 35 at the b values of 0.14
and 0.34. The data in Figures 4c and 4d is only from the
2011 Tohoku earthquake, eliminating variability between
regions or different earthquakes.
For both magnitude bins in Figure 4, the geometric
mean values remain mostly constant for b values ranging
from 0.20-0.34, with an exponential increase for b values
less than 0.20. Ncyc varies considerably at any b value due
to the inherent variability of the earthquake ground
motions and local site effects at the ground motion
recording stations.

Figure 3. Response spectra for selected intra-plate

motions for magnitude bins of a.) 6.5 < M w < 7.0 and b.)
7.0 < Mw < 7.5.

5 COMPARISON AND DISCUSSION

Figures 4 and 5 show that the number of uniform

cycles is strongly correlated with earthquake type and Mw.
Figure 6 presents the Mw-Ncyc curves encompassing the
Figure 2. Response spectra for selected inter-plate complete range of magnitudes examined in this study.
motions for magnitude bins of a.) 8.2 < Mw < 9.0 and b.) These curves pertain to both subduction zone earthquake
9.0 < Mw < 9.2. types and incorporate the relationships suggested by
Stuedlein et al. (2023), Kishida and Tsai (2014), and
4.2 Ncyc for Intra-plate events Idriss (1999). The curves in Figure 6 are an exponential
function fit to the data with a b value of 0.34, which is the
Figure 5 is an analogous set of figures as Figure 4 typical value for sand-like soils recommended by Idriss
but for the intra-plate events. For the mean magnitude 6.8 and Boulanger (2008). The Ncyc curve for the intra-plate
in Figure 5a at b values of 0.34 and 0.14 the Ncyc are 17 earthquake (Figure 5) is larger than the Ncyc curves from
and 32, respectively. Ncyc for the mean magnitude 7.2 in inter-plate and crustal earthquakes for Mw > 6.25. At the
Figure 5c are 26 and 39 cycles for the same b values. The typical normalization magnitude of Mw = 7.5 the predicted
variability introduced by different regions – shown in Ncyc from crustal (from Kishida and Tsai 2014),
Figure 4. Equivalent number of cycles Ncyc for the inter-plate events shown for the a.) geometric mean for the
magnitude bin of 8.2 < Mw < 9.0, b.) geometric means for the different regions for bin of 8.2 < M w < 9.0, c.) geometric
mean for the magnitude bin of 9.0 < M w < 9.2, and d.) geometric means for the different regions for bin of 9.0 < M w <
9.2.

Figure 5. Equivalent number of cycles Ncyc for the intra-plate events shown for the a.) geometric mean for the
magnitude bin of 6.5 < Mw < 7.0, b.) geometric means for the different regions for bin of 6.5 < M w < 7.0, c.) geometric
mean for the magnitude bin of 7.0 < M w < 7.5, and d.) geometric means for the different regions for bin of 7.0 < M w <
7.5.
subduction zone inter-plate and subduction zone intra- The inter-plate curve developed in this study
plate earthquakes are 15, 18, and 27 cycles, respectively. overpredicts the Kishida and Tsai (2014) and Idriss
This difference indicates that intra-plate earthquakes (1999) relationships for Mw less than 7.5-7.75 but
have distinct ground motion characteristics that cannot be underpredicts Ncyc for Mw > 8.0. These differences may be
accurately represented by crustal or inter-plate motions. due to variations in the suite of ground motion used in
Furthermore, intra-plate events have a larger overall their studies. Idriss (1999) had a limited number of
contribution to seismic hazard in the lower mainland of earthquakes with Mw > 8.0, and at higher magnitudes, the
British Columbia at the low period range (Halchuk et al. relationship is an extrapolation. The study by Kishida and
2019), which is the seismic demand used in the Tsai (2014) included some subduction zone ground
liquefaction triggering procedures (Idriss and Boulanger motions, but their study relied primarily on NGAWest2
2008). Given the hazard from intra-plate motions, future crustal motions with magnitudes less than 8.0.
studies should evaluate why these motions are unique Nevertheless, the number of cycles from subduction zone
and if the cyclic counting procedures are applicable. ground motions in their study showed similar results to
These studies should include an evaluation of liquefaction the computed number of cycles, as shown in Figure 6.
case history data from intra-plate earthquakes to evaluate
if liquefaction triggering was unique from crustal case 6 SUMMARY AND CONCLUSIONS
histories.
The inter-plate correlation derived from this The CSZ along the Pacific coast of North America
investigation (depicted as a blue line in Figure 6) agrees is likely to produce a major earthquake (M W=8-9.1),
with the curve by Stuedlein et al. (2023) for magnitudes causing soil liquefaction and potential ground failures.
greater than 9.0. This agreement is partly due to the end Historically, lower-magnitude earthquakes have been the
of the correlation being by a single event (e.g., M9.1 Great focus of liquefaction research, but limited studies have
Tohoku Earthquake) and no variability introduced by explored the effects of large-magnitude subduction zone
magnitude binning procedures or earthquake records earthquakes. Longer shaking durations in such
from different regions. For 6.25 < Mw < 9.0, the earthquakes may trigger liquefaction in previously
relationship by Stuedlein et al. (2023) plots between the considered low-risk areas.
inter- and intra-plate curves developed in this study. This The objective of this study was to determine if
dissimilarity arises from differences in conceptual current Mw-Ncyc relationships are suitable for earthquakes
approaches in the cyclic counting investigations. produced in subduction zone environments. 1,448
Stuedlein et al. (2023) did not impose a minimum PGA individual ground motion records were used to compute
threshold or separate inter-plate and intra-plate ground the number of equivalent uniform cycles (Ncyc) for
motion types in their analysis. Both assumptions will earthquake magnitudes from 5.0 to 9.1. The variation of
increase the Ncyc for a given earthquake magnitude. Ncyc with the soil property exponent b value was
examined, and the results were compared with the
current state-of-practice Mw-Ncyc relationships available in
the literature. The dependency of Ncyc on exponent b is
evident, especially in the fine-grained soil range (b <
0.20), supported by multiple other studies (e.g., Kishida
and Tsai 2014; Boulanger and Idriss 2015; Verma et al.
2019; Stuedlein et al. 2023).
Ncyc is dependent on earthquake type (Inter-plate or
Intra-plate). Except for Stuedlein et al. (2023), the existing
Mw-Ncyc correlations underestimate the cycle count for
lower magnitudes subduction zone earthquakes.
Stuedlein et al. (2023) did not consider dependency on
earthquake type, nor was a minimum PGA threshold
applied, resulting in larger Ncyc estimates for inter-plate
events. Geometric mean values for individual magnitude
bins with Mw>8.0 (inter-plate events) presented in Kishida
and Tsai (2014) were found to be in good agreement with
the findings of this study.
The findings emphasize the significance of
distinguishing between intra-plate, inter-plate, and crustal
motions. Ground motion characteristics of intra-plate
earthquakes were observed to differ significantly from
those of crustal earthquakes. In addition, intra-plate
earthquakes have a greater contribution to seismic
Figure 6. Comparison between computed Ncyc for Inter- hazard in the lower mainland of British Columbia. Future
and Intra-plate events for b = 0.34. Relationships by studies should explore their distinct characteristics and
Stuedlein et al. (2023), Kishida and Tsai (2014), and evaluate the suitability of cyclic counting procedures.
Idriss (1999) plotted for comparison. There is considerable variability in the number of
cycles due to different geographical regions, which may
require further investigation. Due to the sensitivity of Ncyc Liu, A.H., Stewart, J.P., Abrahamson, N.A., and Moriwaki,
to the exponent b, refining in-situ penetration resistance Y. 2001. Equivalent number of uniform stress cycles
relationships for local soils may be another important for soil liquefaction analysis, Journal of Geotechnical
aspect to consider. and Geoenvironmental Engineering, 127(12): 1017–
1026.
ACKNOWLEDGMENTS Mazzoni, S. 2022. NGA-Subduction Portal: Ground-
Motion Record Selection and Download. The B. John
The Natural Sciences and Engineering Research Council Garrick Institute for the Risk Sciences. Interactive
of Canada (NSERC) provided the funding for this work Resource.
under grant RGPIN-2023-05520. The authors would also Seed, H.B., and Idriss, I.M. 1971. Simplified procedure for
like to thank researchers at the John Garrick Institute for evaluating soil liquefaction potential, Journal of the
the careful development of the NGA-sub database. Soil Mechanics and Foundation Divisions, ASCE,
97(9): 1249–1273.
REFERENCES Seed, H.B., Idriss, I.M., Makdisi, F., and Banerjee, N.
1975. Representation of irregular stress time histories
Ambraseys, N. 1985. Intensity‐attenuation and by equivalent uniform stress series in liquefaction
magnitude‐intensity relationships for northwest analyses, Report No. EERC 75-29, Earthquake
European earthquakes, Earthquake engineering & Engineering Research Center, University of
structural dynamics, 13(6), 733-778. California, Berkeley, Calif.
Arango, I. 1996. Magnitude scaling factors for soil Stuedlein, A. W., Dadashiserej, A., and Jana, A. (2023).
liquefaction evaluations, Journal of geotechnical Models for the Cyclic Resistance of Silts and
engineering, 122(11), 929-936. Evaluation of Cyclic Failure during Subduction Zone
Boulanger, R.W., and Idriss, I.M. 2014. CPT and SPT Earthquakes, PEER Report 2023/01, Pacific
based liquefaction triggering procedures, Report No. Earthquake Engineering Research Center, University
UCD/CGM-14/01, Center for Geotechnical Modeling, of California, Berkeley, CA.
Department of Civil & Environmental Engineering, Verma, P., Seidalinova, A., & Wijewickreme, D. 2019.
University of California, Davis, Calif., USA. Equivalent number of uniform cycles versus
Boulanger, R.W., and Idriss, I.M. 2015. Magnitude scaling earthquake magnitude relationships for fine-grained
factors in liquefaction triggering procedures, Soil soils, Canadian Geotechnical Journal, 56(11), 1596-
Dynamics and Earthquake Engineering, 79: 296–303. 1608.
Cetin, K.O., Seed, R.B., Der Kiureghian, A., Tokimatsu,
K., Harder, L.F., Kayen, R.E., and Moss, R.E.S. 2004.
Standard penetration test-based probabilistic and
deterministic assessment of seismic soil liquefaction
potential, Journal of Geotechnical and
Geoenvironmental Engineering, 130(12): 1314–1340.
Green, R.A., and Terri, G.A. 2005. Number of equivalent
cycles concept for liquefaction evaluations – revisited,
Journal of Geotechnical and Geoenvironmental
Engineering, 131(4): 477–488.
Halchuk, S., Adams, J., Kolaj, M., and Allen. T 2019.
Deaggregation of NBCC 2015 Seismic Hazard for
Selected Canadian Cities, In Proceedings of the 12th
Canadian Conference on Earthquake Engineering,
CCEE, Quebec, QC, Canada.
Idriss, I.M. 1999. An update of the Seed-Idriss simplified
procedure for evaluating liquefaction potential, In
Proceedings of the Transportation Research Board
Workshop on New Approaches to Liquefaction,
Publication No. FHWA-RD- 99-165, Federal Highway
Administration, Washington, D.C.
Idriss, I.M., and Boulanger, R.W. 2008. Soil liquefaction
during earthquakes, Monograph MNO-12,
Earthquake Engineering Research Institute, Oakland,
Calif.
Kishida, T., and Tsai, C.C. 2014. Seismic demand of the
liquefaction potential with equivalent number of cycles
for probabilistic seismic hazard analysis, Journal of
Geotechnical and Geoenvironmental Engineering,
140(3): 04013023.
Influence of geotechnical parameters on the
stability of progressively raised tailings
impoundments with waste rock inclusions
Abtin Jahanbakhshzadeh & Michel Aubertin
Polytechnique de Montréal,
P.O. Box 6079, Station Centre-ville, Montréal, QC, Canada

ABSTRACT
Waste rock and tailings are the two primary types of solid waste generated by mining operations. Coarse-grained waste
rocks are usually disposed in waste rock piles, while fine-grained tailings are generally deposited hydraulically in surface
impoundments. Waste rock can also be placed strategically within the impoundment to create waste rock inclusions (WRI)
aimed at improving its geotechnical response. Such WRI can greatly increase the impoundment stability by accelerating
drainage and consolidation of the tailings and by providing reinforcement against displacement and failure. In this article,
new analysis results are presented to illustrate the influence of different tailings properties, loading conditions, and WRI
configurations on the static stability of an impoundment raised progressively with upstream dikes. The calculations show
how the WRI affects the dissipation of excess pore water pressures (PWPs) generated during sequential filling of the
impoundment while increasing its stability due to the added reinforcing components.

RÉSUMÉ
Les roches stériles et les rejets de concentrateur sont les deux principaux types de rejets solides générés par les opérations
minières. Les roches stériles à granulométrie grossière sont généralement entreposées dans des haldes à stériles, tandis
que les rejets à grains fins sont usuellement déposés hydrauliquement dans des parcs à résidus en surface. Les roches
stériles peuvent également être placées stratégiquement dans le parc à résidus pour créer des inclusions de roches
stériles (IRS) qui visent à améliorer son comportement géotechnique. Les IRS peuvent augmenter considérablement la
stabilité des ouvrages en accélérant le drainage et la consolidation des résidus et en ajoutant un renforcement contre les
déplacements et la rupture. Dans cet article, de nouveaux résultats d'analyse sont présentés pour illustrer l'influence des
propriétés des résidus, des conditions de chargement et de la configuration des IRS sur l’évolution de la stabilité statique
d'un parc à résidus construit progressivement avec des digues rehaussées vers l’amont. Les calculs montrent comment
les IRS affectent la dissipation des pressions interstitielles en excès générées lors du remplissage séquentiel du bassin et
augmentent la stabilité géotechnique en raison du renforcement ajouté.

1 INTRODUCTION management. For instance, this approach has been used

to create cells that help with slurry deposition at the large
The mining industry continues to face significant Canadian Malartic gold mine (QC, Canada), which has
challenges to ensure the geotechnical stability of tailings successfully applied this technique for over a decade
impoundments, as shown by major failures reported in (James et al. 2013; 2017; Aubertin et al. 2021).
recent years (e.g., Davies 2002; Aubertin et al. 2002a; Recent calculations have been conducted to assess more
Azam and Li, 2010; Roche et al. 2017; Robertson et al. specifically the impact of adding waste rock inclusions on
2019; Strachan and Caldwell 2018; Santamarina et al. the static stability of tailings impoundments constructed
2019; Franks et al. 2021; Planet Team 2022). Such sequentially with upstream dikes. These stability
breaches often result in uncontrolled tailings flow that can calculations consider various configurations for the WRI
lead to devastating impacts on local communities and and external dikes, as well as different material properties,
ecosystems. based on effective stress analyses applied to progressively
An effective way to address stability concerns is to add raised impoundments.
reinforcement with waste rock inclusions (WRI) in the This article presents the stability evaluation procedure,
tailings impoundment. WRI are much stronger, stiffer, and with the material properties and model characteristics. The
more permeable than tailings, and have been shown to results demonstrate how the effect of WRI can be
help the geotechnical and environmental performance of assessed at various stages of the impoundment
impoundments. Extensive research conducted in recent construction, taking into account the variation of porewater
years has demonstrated that WRI can accelerate water pressures (PWP) generated by layered tailings deposition.
drainage and tailings consolidation, while providing The significant geotechnical benefits of incorporating WRI
structural reinforcement against both static and seismic are emphasized with the presented results. These results
loadings (Aubertin et al. 2002b, 2011, 2019, 2021; James are part of a research program aimed at developing an
and Aubertin 2010, 2012; L. Bolduc and Aubertin 2014; optimization strategy for the use of WRI in tailings
Jahanbakhshzadeh et al. 2019; Zafarani et al. 2020; impoundments.
Contreras et al. 2020).
In addition to enhancing geotechnical stability, strategic
placement of waste rock inclusions within a tailings
disposal facility also offers increased flexibility for tailings
2 SLOPE STABILITY ANALYSES

2.1 Calculation procedures and material properties

15m
The static stability of the tailings retaining dikes raised to 2m
create the impoundment is influenced by the distribution of 2.5
total and effective stresses and related material strength. 1
As the impoundment is gradually filled hydraulically with 10m
saturated tailings, consolidation takes place with
dissipation of the excess pore water pressures (PWP). This Bedrock
process can be greatly accelerated by adding WRI in the
impoundment. Figure 2. Cross section of the external tailings retaining
In the field, waste rock inclusions are gradually raised dikes of the impoundment after 8 stages (starter dike and
during the impoundment filling process, in parallel with 7 raises) with the upstream dike construction method.
external dikes construction. These WRI generally remain
above the level of saturated tailings to allow movement of The material properties used in the FEM simulations and
the machinery. Analyzing the stability of impoundments stability analyses are given in Table 1, for the values of E,
with (and without) inclusions must consider the different  and γ, along with the internal friction angle angle 𝜙′, and
construction stages and related loading conditions, with the dilation angle ’ (set to 0 with the non-associated flow rule).
progressive evolution of the total and effective stresses, The total height H of the model is 50 m.
and pore water pressures (Vick, 1990; Aubertin et al.
2002b, 2011, 2021). Table 1. Material properties used in the calculations, taken
Figure 1 presents the basic conceptual model employed to from experimental results in Peregoedova (2012),
analyze the behavior of the tailings impoundment with Boudrias (2018), Essayad (2021), and Jahanbakhshzadeh
upstream dikes (adapted from previous investigations and Aubertin (2022).
presented by Jahanbakhshzadeh et al. 2019; Tailings characteristics
Jahanbakhshzadeh and Aubertin 2020, 2021; Zafarani et Internal friction angle ϕ′ (°) 30-35 (variable)
al. 2020). The starter dike, made of compacted waste rock, E, Young’s modulus (kPa) 300
has a height of 10 m, and each subsequent dike raise is 2
ɣdry , dry unit weight (kN/m3) 17
m high, with a crest width of 10 m and local side slopes of
2.5H:1V. The global side slope of this impoundment is  , Poisson’s ratio 0.29
8H:1V. The subsurface conditions beneath the ksat, saturated hydraulic 5×10-4-5×10-6 (variable)
impoundment consist of a 4 m thick layer of stiff bedrock, conductivity (cm/s)
The phreatic level is maintained at the surface. Waste rock (starter dike and raised dikes)
The initial step of the analysis establishes the stress state Internal friction angle ϕ′ (°) 40
in the model using finite element simulations conducted ɣdry , dry unit weight (kN/m3) 20
with SIGMA/W (GeoSlope Inc.). In these simulations, ksat hydraulic conductivity 5×10-3
horizontal displacements are restricted on both sides (cm/s)
(external boundaries) of the model, and vertical Waste rock in the inclusions
displacements are blocked at the base. The model
Internal friction angle ϕ′ (°) 40
boundaries are impermeable, except at the top where
vertical drainage allows excess pore water pressure ɣdry ,dry unit weight (kN/m3) 19
dissipation. Simulations have been conducted with and k, hydraulic conductivity 2×10-2
without WRI in the modelled impoundment to compare (cm/s)
results and evaluate the effects of the inclusions for two
configurations. Figure 2 illustrates the cross section of models, with a
The (quasi-) static condition simulations apply the Mohr- global external slope of 8H:1V for a conventional tailings
Coulomb elasto-plastic model available within the impoundment (without inclusion, Figure 2a) and for an
SIGMA/W software. The resulting distributions of pore impoundment with two WRI (Figure 2b). Each
water pressures and stresses (total and effective) obtained impoundment model is filled to a final height of 50 m
from these simulations are subsequently used as input for (including the 4 m foundation layer). The distance from the
the slope stability analyses conducted using SLOPE/W starter dike to the left side of the model is 416 m, which
(Geoslope International Ltd; version 2019). These gives ample space to avoid boundary effects. One or two
calculations determine the critical slip surface location and waste rock inclusions are strategically incorporated in
related minimum factor of safety (FS). The determination these models. More details on the numerical procedures
of FS is carried out with the Morgenstern-Price (MP) are given in Jahanbakhshzadeh and Aubertin. 2020, 2021).
method and a stress-based calculation procedure (FE)
(Krahn 2003, 2007).
 8 The value of FS can however be much lower when the
1
50 m

Y
analysis is conducted shortly after the addition of a new
X tailings layer in the impoundement, due to the generation
of excess PWP. For instance, Figure 4 shows the stability
a) Waste rock conditions for an impoundment in which the hydraulic
Inclusions conductivity ksat varies along the thickness of the tailings;
the value of ksat goes from 5x10-5 (cm/s) at base to 1x10-3
Y
(cm/s) near the top. Figure 4a indicates that FS is 1.69
shortly after deposition the 10th layers. Figure 4b shows
that adding (one or) two WRIs (width W of 25 m) increases

10 m
X 416 m
the value of FS to 2.98 for the same slip surface. The most
b) critical slip surface for the case with 2 inclusions
Figure 2. The 2D model used to analyze the stability of the correspond to a value of FS of 2.95 (Figure 4c). Hence, the
tailings impoundment with upstream dikes : a) conventional short term stability of the impoundment, based on effective
impoundment (without inclusion) b) reinforced stress analyses, is significantly reduced when compared to
impoundment (with two WRI). the steady-condition due to the effect of excess PWP
(compare Fig. 3a and 4a).This reduction of FS is much less
pronounced when WRI are added, as these provide
2.2 SIMULATION RESULTS reinforcement in the impoundment.
Previous analyses have considered the long-term static
stability of the impoundment, under steady-state FS = 1.69
conditions, showing that the value of FS can be satisfactory X=116.35 m
even without inclusion (Jahanbakhshzadeh and Aubertin Y=70.52 m
R = 17.32 m 8
2020). For example, Figure 3 presents the critical slip Y 1
surface and the minimum factor of safety (FS), obtained
X
using the Morgenstern-Price method in SLOPE/W, for the
conventional tailings impoundment with an 8H:1V slope. a)
After 26 years, following the deposition of 10 layers and the FS = 2.98
dissipation of all excess pore water pressures (i.e. drained Y
conditions with equilibrium PWP), the value of FS is
X
evaluated as 3.71 (Figure 3a). Figure 3b shows that
incorporating a single waste rock inclusion (width, W, of 25 b) FS = 2.95
m) near the central portion of the upstream dike increases X=175.77 m
FS to 4.22 for the same slip surface. In this case however, Y=223.88 m
R = 56.51 m
the most critical slip surface would be located downstream Y

of the inclusion, yielding an FS of 4.18 (Figure 3c); the

X
potentially unstable area is then significantly reduced.
c)
FS = 3.71 Figure 4. Critical slip surfaces, position of the centers of
X=268.4 m rotation (X, Y, R), and FS values for cases where the
Y=402.6 m
8 tailings hydraulic conductivity varies with depth; the results
R = 393.9 m
Y 1 shown here correspond to the full thickness of the
impoundment after the addition of the top tailings layer (i.e.
a) X
10th layer) at t = 22 years a) no inclusion; b) with WRI and
the same slip surface; c) critical slip surface with WRI.
FS = 4.22

Y
Inclusion Another factor that affects the value of FS is the tailings
shear strength. Calculations were conducted to assess the
b) X
25 m
influence of varying the effective internal friction angle of
FS = 4.18
X=383.48 m
the tailings, which tends to increase over time with
Y=315.35 m consolidation and densification. In these calculations, a
R = 307.5 m zone of consolidated tailings (layer at bottom) has a higher
Y
friction angle ϕ' of 35°; this value is decreased in steps for
every 6 m, so the loose tailings recently added has a lower
c)
X

strength with ϕ' = 30°. Figure 5 presents the critical slip

Figure 3. Critical slip surfaces, position of the centers of surface and the corresponding FS for an impoundment with
rotation (X, Y, R), and FS values obtained with SLOPE/W varying values of the internal friction angle. The critical slip
(Morgenstern-Price method with stress-based surface shortly after addition of the top tailings layer, under
calculations) for the impoundment after 26 years a) no very short-term condition, is located near the crest of the
inclusion; b) with one inclusion and the same slip surface dikes (Figure 5a) and corresponds to a factor of safety FS
as in (a); c) critical slip surface with one inclusion.
=1.40. Figure 5b shows that the minimum FS is increased
to 2.84 for the tailings impoundment with two WRI. Figure 7 shows the most critical slip surfaces and the
corresponding values of FS for the conventional
impoundment, with tailings having different hydraulic
FS = 1.40 conductivity. The results have been obtained at t = 22
X=108.88 m years, just after the addition of the last saturated tailings
Y=78.02 m
R = 16.94 m
8 layer (4 m thick). The layer of loose (slurry) tailings is given
1
Y
a reduced shear strength with ϕ' = 30° (for the slurry) The
X results in this figure can be compared with those in Figure
a) 3. Figure 7a indicates that the factor of safety become very
FS = 2.84
X=175.77 m low, i.e. FS = 0.71, when the smallest hydraulic conductivity
Y=23.88 m of 5x10-6 (cm/s) is imposed (due to the remaining excess
Y R = 56.52 m PWP, see Figure 6). The short term value of FS is
X
increased to 1.42 when the value of k is 5x10-5 (cm/s)
(Figure 7b) and to 1.82 (Figure 7c) for k = 5x10-4 (cm/s). It
b) is seen also that these slip surfaces cover only a small
Figure 5. Critical slip surfaces, centers of rotation (X, Y, R), portion of the slope (near the top of the dikes), but this initial
and values of FS immediately after the addition of the top zone of instability could well lead to a more widespread
tailings layer (i.e. 10th layer) at t = 22 years with ksat=5x10-5 breach.
(cm/s); effective stress analyses conducted with a variation
of the internal friction angle of the tailings; ϕ’ goes from 30° FS = 0.71
near the surface to 35°at depth; a) no inclusion; b) critical X=111.32 m
Y=71.52 m
slip surface with two WRIs. R = 17.34 m
8
Y 1
The hydraulic conductivity of the saturated tailings may
also affect the response and stability of the impoundment. X

Figure 6 shows the variation of PWP at an elevation of 1 m

above the base of the tailings and 60 m horizontally from a)
FS = 1.42
the boundary condition at the left side within the X=108.88 m
conventional impoundment (no inclusion) for yearly Y=78.02 m
8
additions of tailings (with different hydraulic conductivity). It Y
R = 16.12 m 1
is seen in this figure that the PWP increases upon the
X
addition of each layer. The excess PWP than dissipate
over time. This dissipation is more rapid during the first few b) FS = 1.82
X=116.35 m
years and then slows down as the tailings thickness Y=70.51 m
increases; full dissipation of excess PWP takes longer later 8
Y
R = 16.43 m 1
in the mine life when the total tailings thickness is larger.
PWP dissipated more rapidly in tailings with a higher X

hydraulic conductivity. In the case analysed here, the lower

values of ksat lead to accumulation of excess PWP over c)
time, which becomes detrimental for the impoundment Figure 7. Critical slip surfaces with the position of the
stability (as shown in the next figure). rotation center (X, Y, R) and the corresponding FS values
obtained with the Morgenstern-Price (MP) method (and
effective stress-based calculation) for the conventional
impoundment (without inclusion) shortly after the addition
of the top tailings layer (i.e. 10th layer) at t = 22 years a) k
= 5x10-6 (cm/s) b) k = 5x10-5 (cm/s) c) k = 5x10-4(cm/s).

Figure 8 shows the value of FS of the impoundment

with two inclusions after the addition of the top layer, for
different hydraulic conductivity of the tailings (as in Figure
7). The much larger values of FS (near 3 or more) are due
to the reinforcement provided by the WRI and to the
quicker dissipation of excess PWP in the presence of the
inclusions. These results illustrate some of the benefits
from adding WRI on the short term (critical) static stability
of a progressively filled tailings impoundment.
Figure 6. Simulated porewater pressure (PWP, kPa) at h =
1 m above the base of the tailings impoundment and 356
m from the starter dike (60 m horizontally from the
boundary condition at the left side) for cases with different
hydraulic conductivity of tailings for t = 0 to 26 years.
 FS = 2.97
X=186.05 m
Y=154.03 m
R = 48.875 m
8
Y 1

a)
FS = 3.5
X=147.9 m
Y=116.88 m
R = 89.62 m
8
Y 1

b)
FS = 3.76
X=183.8 m
Y=157.53 m
R = 46.998 m
8
Y 1

X Figure 9. Evolution of the factor of safety over time after the

addition of the 10th tailings layer, for the conventional (no
c) WRI) and reinforced (2 WRI) impoundments for tailings
Figure 8. Critical slip surfaces with the position of the with r 3 different hydraulic conductivities.
rotation center (X, Y, R) and the corresponding FS values
for the impoundment with two inclusions, immediately after
adding the top tailings layer (i.e. 10th layer) at t = 22 years 3 DISCUSSION AND FINAL REMARKS
a) k = 5x10-6 (cm/s) b) k = 5x10-5 (cm/s) c) k = 5x10-4 (cm/s).
Stability analyses of tailings impoundments are often
Results of stability analyses conducted to assess the performed using a total stress approach or consider only
effect of ksat are summarized in Figure 9. This figure shows the effective stress behavior under steady state loading
the evolution of FS over time after adding the top (10th) conditions (with equilibrium PWP). Such types of analysis
layer. It is recalled that this last layer of saturated tailings thus neglect the changes of PWP and effective stresses
has an effective internal friction angle (ϕ') of 30°, while the during staged construction. However, effective stress
remaining tailings are stronger with ϕ' = 35°. These results stability analyses conducted at various stages during
indicate that FS for the conventional impoundment with the construction of the impoundment show that progressive
lowest hydraulic conductivity, k = 5x10-6 (cm/s), is the most filling with addition of new tailings layers lead to the
critical just after adding the layer, with FS = 0.71 (at t = 22 creation of excess pore water pressures (PWP) after the
on the graph). The initial value of FS, which is larger for placement of each layer. Such analyses indicate that the
higher ksat values, then increases progressively during the most critical stability condition, with the minimum value of
following months up to the maximum value attained upon FS, corresponds to the short term behavior, just after
reaching the equilibrium PWP. The results in Figure 9 also adding a new layer. The factor of safety then tends to
show that adding reinforcement with two WRI in the increase over time for each depositional step, up its
impoundment raises the minimum (initial) and final values maximum value when PWP reach their equilibrium values.
of FS for the three different hydraulic conductivity, in the Slope stability calculations have been conducted here with
short term and over time. Sigma/W and Slope/W (Geoslope International Ltd 2018)
to assess the response of tailings impoundments as a
function of the deposition conditions, tailings properties and
presence of waste rock inclusions.
The results illustrate how WRI can significantly improve
the static stability of tailings impoundments constructed
with upstream dikes by increasing the value of FS for the
most critical conditions, which occur shortly after the
deposition of new layers. The results also demonstrate how
WRI can be used to increase the most critical value of FS.
The calculation results also show the effects of the
hydraulic conductivity of the tailings on the accumulation of
excess PWP and on the impoundment stability. The results
further demonstrate how WRI improve stability by
accelerating dissipation of excess PWP, even when the
tailings have a low hydraulic conductivity. This construction
method has the potential to significantly reduce the risks
associated with impoundment constructed with upstream
dikes.
Additional results, not presented here, show that the
minimum FS can be increased by adjusting the number
and locations of the WRI in the impoundment, similarly to considering the effect of reinforcement
dynamic stability analyses (Ferdosi et al. 2015, inclusions. Paper presented at the 73rd CGS
Jahanbakhshzadeh et al. 2019; Zafarani, 2022; Conference - GeoVirtual 2020 - Resilience and
Contreras). innovation.
Other factors of influence have not been considered Davies, M. P. 2002. Tailings impoundment failures: Are
here, including the effect of the tailings deposition geotechnical engineers listening? Geotechnical News,
sequence and layer thickness, tailings segregation and BiTech Publisher, Vancouver, 20(3), pp. 31-36.
layering, settlements, and position of the phreatic level. n. Essayad, K. 2021. Migration des particules fines des
These factors and others are part of the ongoing study roches stériles et résidus à travers les inclusions et effet
aimed at understanding the effects of WRI to optimize their sur leurs conductivité hydraulique et capacité drainante
application and to maximize their effects on the stability (text in French), Ph.D thésis, Polytechnique Montréal.
and performance of tailings impoundments. The results Fell, R., MacGregor, P. Stapleton, D., Bell, G. 2005.
presented here can nonetheless be useful for the analysis Geotechnical engineering of dams. A.A. Balkema-
and design of WRI in tailings impoundment. Taylor & Francis.
Ferdosi, B., James, M., Aubertin, M. 2015. Effect of waste
ACKNOWLEDGEMENTS rock inclusions on the seismic stability of an upstream
raised tailings impoundment: a numerical investigation.
The authors acknowledge the financial support from Canadian Geotechnical Journal, 52(12): 1930-1944.
NSERC, FQRNT, and partners of the Research Institute on GEO-SLOPE., 2018. Stress-deformation modeling with
Mines and the Environment, RIME UQAT Polytechnique SIGMA/W 2018 and slope stability with Slope/W 2018.
(http://rime-irme.ca/en). GEO-SLOPE International Ltd.
Franks, D.M., Stringer, M., Torres-Cruz, L.A. et al. 2021.
Tailings facility disclosures reveal stability risks. Sci
REFERENCES Rep 11, 5353 (2021). https://doi.org/10.1038/s41598-
021-84897-0Jahanbakhshzadeh, A., Aubertin, M.,
Aubertin, M, Bussière, B, Bernier, B. 2002a. Yniesta, S., Zafarani, A. 2019. On the seismic response
Environnement et gestion des rejets miniers, Manuel of tailings dikes constructed with the upstream and
sur cédérom, Presses internationales Polytechnique center line methods. Geosciences for Sustainability,
Aubertin M, Mbonimpa M, Jolette D, Bussière B, Chapuis Proceedings of the 72th CGS Conference, St. John’s,
RP, James M, Riffon O. 2002b. Stabilité géotechnique Canada.
des ouvrages de retenue pour les résidus miniers: Jahanbakhshzadeh, A., Aubertin, M. 2020. The effect of
problèmes persistants et méthodes de contrôle. Défis waste rock inclusions on the slope stability of tailings
& Perspectives : Symposium sur l'environnement et les dikes. Geosciences for Sustainability, Proceedings of
mines, Rouyn-Noranda, Développement Économique the 73th CGS Conference, Geovirtual. pp. 1-7.
Canada/Ministère des Ressources Naturelles du Jahanbakhshzadeh, A., Aubertin, M. 2021. Effect of the
Québec/CIM. Comptes-Rendus sur CD-ROM. configuration of waste rock inclusions on the static
Aubertin, M., Bussière, B., James, M., Jaouhar, E.M., stability of tailings impoundments. Geosciences for
Martin, V., Pépin, N., Mbonimpa, M., and Chapuis, R.P. Sustainability, Proceedings of the 74th CGS
2011. Vers une mise à jour des critères de stabilité Conference, GeoNiagara. 7 pages.
géotechnique pour la conception des ouvrages de Jahanbakhshzadeh, A., Aubertin, M. 2022. Main results
retenue de résidus miniers. Symposium sur les Mines from an extensive geotechnical characterization of hard
et l’Environnement, Rouyn-Noranda, Québec, 6-9 rock tailings from an open pit gold mine. Geosciences
novembre 2011 for Sustainability, Proceedings of the 75th CGS
Aubertin, M. Jahanbakhshzadeh, A., Yniesta, S. 2019. The Conference, GeoCalgary. 8 pages.
effect of waste rock inclusions on the seismic stability James M., and Aubertin, A. 2010. On the dynamic
of a tailings impoundment. Proceedings of response of tailings and the stability of tailings
7th International Conference on Earthquake impoundments for hard rock mines. Geotechnical
Geotechnical Engineering News, 23(3), 39-43.
Aubertin, M., Jahanbakhshzadeh, A., Saleh Mbemba, F., James, M., and Aubertin, M. 2012. The use of waste rock
Yniesta, S., Pednault C. 2021. General guidelines for inclusions to improve the seismic stability of tailings
the design and construction of waste rock inclusions in impoundments. GeoCongress 2012, Okland, 22-25
tailings impoundment. Proceedings of the 74th CGS March 2012. American Society of Civil Engineers, pp.
Conference, GeoNiagara. 4166-4175.
Azam, S., Li, Q. 2010. Talings dam failures : A review of James, M., Aubertin, M., and Bussière, B. 2013. On the use
the last one hundred years. Geotechnical News. Vol. of waste rock inclusions to improve the performance of
28, no 4, pp. 50-54. tailings impoundments. 18th International Conference
Boudrias, G. 2018. Évaluation numérique et expérimentale on Soil Mechanics and Geotechnical Engineering,
du drainage et de la consolidation de résidus miniers à Paris, France, pp. 735-738.
proximité d'une inclusion de roches stériles. Master James, M., Aubertin, M., Bussière, B., Pednault, C., Pépin,
thesis, Ecole Polytechnique de Montreal. N., and Limoges, M. 2017. A research project on the
Contreras, C., Yniesta, S. & Aubertin, M. 2020. Seismic use of waste rock inclusions to improve the
and post-seismic stability of tailings impoundments,
 performance of tailings impoundments. Geo Ottawa,
Ottawa, ON. p. 8.
Krahn, J., 2003. The 2001 R.M. Hardy Lecture: The Limits
of Limit Equilibrium Analyses. Canadian Geotechnical
Journal, Vol. 40. pp.643-660.
Krahn, J., 2007. Limit Equilibrium, Strength Summation
and Strength Reduction Methods for Assessing Slope
Stability. Communication presented at Proceeding, 1st
Canada-US Rock Mechanics Symposium, Vancouver,
BC May (p. 27-31).
L. Bolduc, F. and Aubertin, M. 2014. A numerical
investigation of the influence of waste rock inclusions
on tailings consolidation, Canadian Geotechnical
Journal, 51(9): 1021-1032.
Peregoedova, A. 2012. Étude expérimentale des
propriétés hydrogéologiques des roches stériles à une
échelle intermédiaire de laboratoire. Mémoire de
Maitrise, École Polytechnique de Montréal.
Planet Team 2022. The causes of the 30 December 2020
Gjerdrum landslide in Norway. Planet Application
Program Interface: In Space for Life on Earth. San
Francisco, CA. https://www.planet.com/
Roche, C., Thygesen, K., and Baker, E. (Eds.). 2017. Mine
Tailings Storage: Safety Is No Accident, A UNEP Rapid
Response Assessment. United Nations Environment
Programme and GRID-Arendal, Nairobi and Arendal.
Robertson, P.K., de Melo, L., Williams, D.J., and Wilson,
G.W. 2019. Report of the expert panel on the technical
causes of the failure of Feijão Dam I. Available
from http://www.b1technicalinvestigation.com/ [access
ed December 2019].
Santamarina, J.C., Torres-Cruz, L.A., Bachus, R.C. 2019.
Why coal ash and tailings dam disasters occur. Science
364 (6440), 526-528. DOI: 10.1126/Science.axx 1927
Strachan, C. and Caldwell, J. 2018. Tailings facility
performance:2017 and 2018. Presentation for Tailings
& Mine Waste 2018, Session 2B, Design and operation
1.
Vick, S. G. 1990. Planning, design and analysis of tailings
dams. BiTech Publishers Ltd.
Zafarani, A., Yniesta, S., Aubertin, M. 2020. Effect of height
and downstream slope on the seismic behavior of
tailings impoundments reinforced with waste rock
inclusions. Proc. 73rd CGS Conference, GeoVirtual
2020.
Investigation of strain localization in sand
under triaxial compression: insights from
internal pore-water pressure measurements
Yawu Liang, Nicholas Beier & D.C. Sego
Geotechnical Center, Department of Civil & Environmental
Engineering, University of Alberta,
Edmonton, AB, T6G 2G7, Canada

ABSTRACT
The mechanism of strain localization in granular material is associated with the development of progressive failure of
slopes and soil liquefaction. A series of consolidated, undrained triaxial tests were conducted on saturated, dense and
loose sand with internal pore-water pressure measurements to investigate the correlation between the internal pore-water
pressure response and the evolution of strain localization. An incremental pore-water pressure coefficients, a and 1/a,
were introduced to evaluate the change of shear stress on a short-time scale. A spike in coefficient a before the peak shear
strength in dense sand is interpreted as the initiation of strain localization. In dense and loose sand, a substantial spike in
coefficient a around peak stress is associated with the onset of persistent shear band development or strain softening.
Also, the progression of persistent shear banding or the concentration of strain localization in the post-peak stage can be
characterized by the variation of coefficient 1/a.

RÉSUMÉ

Le mécanisme de localisation de la déformation dans un matériau granulaire est associé au développement de la rupture
progressive des pentes et à la liquéfaction du sol. Une série de tests triaxiaux consolidés, non drainés ont été réalisés sur
du sable saturé, dense et lâche avec des mesures de pression interne de l'eau de pore afin d'étudier la corrélation entre
la réponse de pression interne de l'eau de pore et l'évolution de la localisation de la déformation. Des coefficients de
pression interne de l'eau de pore, a et 1/a, ont été introduits pour évaluer le changement de contrainte de cisaillement à
court terme. Une pointe de coefficient a avant la résistance au cisaillement maximale dans le sable dense est interprétée
comme l'initiation de la localisation de la déformation. Dans le sable dense et lâche, une pointe substantielle de coefficient
a autour du pic de contrainte est associée au début du développement de la bande de cisaillement persistante ou à un
adoucissement de la déformation. De plus, la progression de la bande de cisaillement persistante ou la concentration de
la localisation de la déformation dans la phase post-pic peut être caractérisée par la variation du coefficient 1/a.

1 INTRODUCTION result, plain strain tests have become a preference for

researchers. In dense sand, the onset of strain localization
The phenomenon of strain localization, where shear has commonly been reported to occur before reaching the
deformation concentrates in a narrow zone known as a peak stress (i.e., during the hardening regime), with the
shear band, is commonly observed in granular materials. shear band becoming visible during the shear band
This process is associated with the progressive failure of softening phase when the stress drops abruptly (Desrues
slopes and soil liquefaction. et al., 1985; Hamza et al., 2009). This behavior has also
From a theoretical standpoint, strain localization is been observed in true triaxial tests (Wang and Lade, 2001).
often interpreted as a bifurcation problem, and constitutive However, Mokni and Desrues (1999) noted that strain
models have been developed since the early 1960s to localization and cavitation of the pore fluid occurred
predict its initiation and properties, such as thickness and simultaneously. For loose specimens, strain localization
inclination. Rudnicki and Rice (1975) predicted the has often been observed to initiate at or just prior to the
possibility of onset strain localization during the hardening peak effective stress ratio (Finno et al., 1996; Desrues et
regime, and their results showed that it strongly depended al., 2004).
on the parameters used in the constitutive model. In In contrast to the simple behavior observed in the
contrast, Vardoulakis et al. (1978) reported the initiation of biaxial tests, granular material under conventional triaxial
shear bands occurring at the peak in both initially loose and test conditions exhibits a more complex behavior and
dense samples under theoretical and experimental failure mode. With the introduction of advanced
investigation. However, due to the absence of experimental experimental techniques such as X-ray, CT, and image
evidence at the time, the onset of strain localization analysis, researchers have been able to directly observe
remained a controversial issue. the progression of strain localization and visualize it. For
In recent decades, extensive experimental studies specimens that exhibit the shear plane failure mode, it is
have been undertaken to understand the phenomenon of widely accepted that strain localization coalesces from a
strain localization. Researchers have observed that well- wide and diffuse zone before the peak stress to a narrow
defined shear bands often develop more readily in biaxial shear zone after the peak stress, and grain rotations can
tests compared to those under triaxial conditions. As a serve as an indicator of onset of localization
(Rechenmacher, 2006; Hall et al., 2009; Tagliaferri et al., comparison. Additional details regarding the apparatus can
2011; Andò et al., 2012; Tordesillas et al., 2013; Alikarami be found in Liang et al. (2022).
et al., 2015; Takano et al., 2015). On the other hand,
Alshibli et al. (2003) conducted a series of tests on
specimens with bulging failure mode under very low-
confining pressures, and they found no evidence of single
or multiple shear bands on the surface of specimens.
However, after analyzing the internal structure using CT
techniques, they discovered that multiple secondary
symmetrical shear bands developed outward in the radial
and axial directions during the post-peak regime. It should
be noted that the tests were often limited to a small size (22
mm in height and 11 mm in diameter), dry sample. The
evolution of strain localization for moist specimens (dense
or loose) with pore-water pressure (pwp) development
under triaxial conditions requires further investigation.
In a traditional triaxial test, the pore-water pressure
(PWP) is often measured at the base or top of the
specimen, which is referred to as the global PWP with Figure 1. Schematic of apparatus by Liang et al. (2022)
respect to the global stress. However, the global
measurements will dampen any instantaneous localized 2.3 Test Procedures
response in PWP and stress. Therefore, measurements of
internal PWP adjacent to the shear band are necessary to Triaxial tests were performed on 100 mm diameter and 196
investigate the propagation of the shear band. To address mm high specimens, which were prepared using two
this issue, Liang et al. (2022) developed a new method for different reconstitution techniques as described by Liang et
measuring the internal PWP of soil based on the filterless al. (2022). For loose specimens, the moist tamping
rigid piezometer (FRP) designed by Kia (2012). It is procedure was followed, which involves mixing dry sand
capable of tracking the pwp response at specific locations with a moisture content of 5% to obtain a very loose
within the soil. With this method, the internal PWP at structure through capillary cohesion. The material is then
different distances from the shear band can be monitored compacted layer by layer (4 cm thickness for each layer)
during the formation of shear band. The objective of this into a membrane-lined split mold. Dense specimens were
study, therefore, is to investigate the relationship between prepared using the slurry deposition method, which
the internal PWP response and the development of strain involves preparing five equal, mixed slurries and then
localization by performing consolidated, undrained triaxial transferring them to the split mold layer by layer with gentle
tests on saturated dense and loose sand and explore the vibration.
applicability of using incremental PWP coefficient a to Prior to the saturation stage, loose specimens were
manifest this mechanism. upwardly flushed with carbon dioxide for approximately 60
minutes, followed by percolation with about 2000 ml of de-
2 MATERIAL AND METHODS aired water in the same direction. In contrast, dense
specimens were percolated with de-aired water only. Back
2.1 Material pressure increments were then applied to reach maximal
saturation, while monitoring the degree of saturation using
The tests were carried out on a uniformly graded fine sand, Skempton's B-value. An effective confining pressure of
mainly comprising quartz grains. This sample material is approximately 30 kPa was maintained throughout the
the same material used by Liang et al. (2023). It has a saturation process. The back pressure required to achieve
mean diameter (D50) of 0.28 mm, and the specific gravity is a B-value higher than 0.97 was around 400 kPa.
2.64. The grain sizes vary from 0.15 mm and 0.85 mm Specimens were isotropically consolidated to the target
(Cu=1.5). effective stress, as presented in Table 1. Finally, the
specimens were sheared at a constant strain rate of
2.2 Test Apparatus 1%/min.

A modified, strain rate controlled triaxial apparatus Table 1. Summary of the testing program
developed by Liang et al. (2022) for measuring internal
PWP is illustrated in Figure 1. To study the relationship Initial Water Confining
Test Initial Void
Content Pressure
between the development of strain localization and the ID Ratio
(%) (kPa)
internal PWP response, it is necessary to measure PWP at
D-1 0.599 23 100
various distances from the shear band. For this purpose, D-2 0.599 23 200
three 0.53 mm inner diameter PeekSil tubes were placed D-3 0.597 23 400
at 8 cm, 10 cm, and 12 cm heights within the specimen and L-1 0.834 32 100
connected to fiber optic pressure sensors to measure the L-2 0.834 32 200
PWP above, near, and below the shear plane. A L-3 0.832 31 400
conventional pressure transducer at the base was used for Note: D, dense; L, loose.
3 RESULTS AND DISCUSSION 3.2 PWP Response

3.1 Calibration Figure 3 shows the deviator stress and PWP responses
from dense specimens at effective confining pressures of
Before the tests, all PWP sensors were calibrated in a 100, 200, and 400 kPa. All specimens exhibited dilative
water-filled triaxial cell by adjusting the cell pressure behavior and reached peak stress at around 6% axial
(loading and unloading). Liang et al. (2022) reported that strain. The FOPs showed a similar trend to the base
the FRPs showed a faster response for pressures above transducer, with an exception of an abrupt PWP drop
100 kPa during loading, but a slower response during observed in some FOPs between 3 and 4 % axial strain.
unloading below 130 kPa compared to the base This drop was observed in FOP-12 and 8 cm of specimen
transducer. Because of the cyclic variation of building D-1, FOP-12 cm of D-2, and FOP-12 and 10 cm of D-3.
pressure in the lab, the sensors were also calibrated under The drop may be attributed to shear deformation of PeekSil
constant loading conditions, as shown in Figure 2a. It was tubes during the test, as tube distortion was observed after
found that the period and amplitude of the pressure (cell the test (see Figure 4), which affected the continuity of the
pressure) were 8 minutes and 1 kPa, respectively. As a piezometer fluid and caused a delay in PWP response.
result, the PWP sensors followed the same pattern but had However, the FOP-10 cm of specimen D-1 was also
a lower amplitude (around 0.5 kPa). In addition, the sheared but did not record the sudden PWP drop.
response of FOPs is not as smooth as that of the Moreover, the tubes likely kinked after peak stress, when a
transducer. In other words, the FOPs are more sensitive well-developed shear band traversed the specimen.
than the transducer. Hence, the incremental change in Therefore, this phenomenon does not result from tube
pressure is compared in Figure 2b. The incremental deformation, but reflects local stress and internal structural
change in pressure measured by the transducer is changes. When a dilative shear band starts to form, the
relatively constant, while it varies from -0.05 to 0.05 kPa in local density increases (Tagliaferri et al., 2011; Tordesillas
FOPs at 10 and 12 cm. The FOP at 8 cm may have a et al., 2013), and this leads to a decrease in PWP. The tip
different factory calibration, and it shows a larger variation, of tubes was probably located in these localized zones at
between -0.15 and 0.15 kPa. the moment, and hence these PWP responses prior to
peak stress indicated the initiation of local shear band or
strain localization.

(a)

(a)

(b)
Figure 2. Results of Calibration under constant
unloading conditions: (a) variation in pressure; (b)
incremental change in pressure. (b)
 (c) (b)
Figure 3. PWP response of dense specimens: (a) 100
kPa; (b) 200 kPa; (c) 400 kPa.

(c)
Figure 5. PWP response of loose specimens: (a) 100
kPa; (b) 200 kPa; (c) 400 kPa.
(a) (b)
Figure 4. Images of specimen after failure: (a) shear
plane; (b) distortion of tubes. 3.3 PWP Coefficients A and a

In contrast, all loose specimens exhibited contractive The Skempton’s pore pressure coefficients are important
behavior and reached peak stress at around 1 % axial parameters to predict the change in pore-water pressure
strain (Figure 5). The PWP responses of FOPs were quite and undrained shear strength. The determination of
smooth, similar to the base transducer. The tip of the tube coefficient A during the application of a deviator stress is
may be temporarily compressed as the axial strain given by Skempton (1954):
increases, causing some spikes in the internal PWP
measurements from the FOP-10 cm of specimen L-1. ∆𝑈
𝐴= [1]
𝐵(∆𝜎1 − ∆𝜎3 )

Where
B is another PWP coefficient indicating the degree of
saturation;
∆𝑈 is the PWP change; and
(∆𝜎1 ,∆𝜎3 ) are principal stress changes.
In traditional triaxial tests, coefficient A is obtained by
the following total form:
𝑢𝑡 − 𝑢0
𝐴𝑡 = [2]
𝐵(𝑞𝑡 − 𝑞0 )

Where
(a) 𝐴𝑡 is the PWP coefficient at current axial strain;
𝑢𝑡 and 𝑢0 are the PWP at the current axial strain and
the beginning; and
 𝑞𝑡 and 𝑞0 are the deviator stress at the current axial an incremental form of coefficient a is required. The
strain and the beginning. incremental change of PWP and deviator stress can be
The obtained coefficient A of dense specimens is expressed as:
shown in Figure 6. The results indicate that the coefficients
A for higher FOPs (e.g., 12 and 10 cm) are typically larger ∆𝑢 = 𝑢𝑡 − 𝑢𝑡−1 [3]
than those for lower FOPs or the transducer during the ∆𝑞 = 𝑞𝑡 − 𝑞𝑡−1 [4]
initial contraction stage. A larger coefficient A suggests that
the higher FOPs experienced a more significant change in Where ∆𝑢 and ∆𝑞 represent the incremental change of
PWP from the beginning of the test. This phenomenon was PWP and deviator stress; 𝑢𝑡−1 and 𝑞𝑡−1 are the PWP and
not observed in the FOPs within the loose specimens, so deviator stress at the last step of axial strain. Here, the test
the results are not presented here. strain rate is 1%/min, and the data acquisition frequency is
1 Hz, so the length of each step is taken as 0.017% axial
strain.
The incremental form of PWP coefficient a is then given
by:

∆𝑢
𝒂= [5]
𝐵∆q

(a)

(b)

(c)
Figure 7. Coefficient a and incremental changes of PWP
Figure 6. Coefficient A of dense specimens: (a) 100 kPa;
and deviator stress of sample D-3: (a) coefficient a; (b)
(b) 200 kPa; (c) 400 kPa.
∆𝑢 & ∆𝑞; (c) coefficient 1/a.
However, the analysis of coefficient A is conducted over
The coefficient a and incremental changes of PWP and
widely spaced strain change, fails to discern the variation
deviator stress of sample D-3 are displayed in Figure 7.
of PWP and shear stress on a short-time scale. Therefore,
The results of other samples are not presented in this
paper since they demonstrated similar trends. At the peak
stress, a significant spike in coefficient a appeared from all
sensors (Figure 7a). While the stress continued to increase
before the peak, the PWP continuously decreased after the
initial contraction stage, resulting in a positive ∆q but a
negative ∆u and thus a negative coefficient a. However,
after the peak stress, there was a sudden drop in the
deviator stress (Figure 7b), leading to a negative ∆q but a
positive coefficient a. The shear deformations are known to
start to concentrate in a narrow zone soon after the peak
stress, and eventually, a persistent shear band will traverse
through the specimen (Rechenmacher, 2006). Thus, this
coefficient a is associated with the onset of the persistent
shear band. This type of spike was also observed in loose
specimens, for example L-2 in Figure. 8a, at the peak
stress or onset of strain softening. Furthermore, it was
observed that FOP-12 and 10 cm underwent a strong
variation in ∆u before the peak stress, and magnitude of
variation was larger than that recorded during the
calibration test (Figure 2b), which may reflect a change in
local shear stress. Moreover, there was a significant drop
in ∆q and ∆u from FOP-12 and 10 cm at around 4.1 % axial
strain, causing another slight spike in coefficient a. This
corresponds to the abrupt PWP drop mentioned earlier in
Figure 3 and demonstrates the initiation of a local shear
band once again. Such an important feature cannot be
reflected via the global PWP from the base transducer.
As the specimen enters the softening stage following
the peak stress, the shear deformation starts to converge
into a line of band which becomes a persistent shear band
later (Hall et al., 2009), accompanied by a reduction of
deviator stress. At the same time, the decrease in ∆q
becomes more and more evident, eventually stabilizing
around 0 kPa until the residual stress of the sample is
reached and a well-defined shear band is developed. This
feature can also be captured by coefficient 1/a, which
manifests ∆q when ∆u remains relatively constant. Thus,
the variation of coefficient 1/a becomes increasingly
significant after the peak stress until the residual state is
reached (Figure 7c), indicating the formation of the
persistent shear band or the concentration of strain
localization.
For specimen L-2 (Figure 8), the initial spike in ∆q and
Figure 8. Coefficient a and incremental changes of PWP
∆u suggests the rapid increase in deviator stress and
and deviator stress of sample L-2: (a) coefficient a; (b)
excess PWP before liquefaction. Subsequently, ∆q drops
slightly below 0 kPa and gradually stabilizes at 0 kPa. The ∆𝑢 & ∆𝑞; (c) coefficient 1/a.
subtle variation in ∆q can also be manifested by the
coefficient 1/a of FOPs, which shows a continuous
increase in fluctuation up to residual state, similar to the 4 CONCLUSION
dense specimens.
A study was conducted by performing a series of
consolidated, undrained triaxial tests on saturated, dense,
and loose sand to explore the relationship between the
development of strain localization and the internal PWP
response. The author utilized the incremental change of
PWP and shear stress and PWP coefficient a to capture
the instantaneous response. The internal PWP
measurements in dense sand revealed significant
variations, which were probably located near or inside the
local shear zones before peak stress. The spike in
coefficient a corresponded to the abrupt PWP drop
between 3% and 4% strain and indicated the initiation of
local shear band or strain localization. A significant spike in Hall, S., Lenoir, N., Viggiani, G., Desrues, J., and Bésuelle,
coefficient a around peak stress was associated with the P. 2009. Strain localization in sand under triaxial
onset of the persistent shear band or strain softening. loading: characterization by x-ray micro tomography
Additionally, the coefficient 1/a was used to manifest the and 3D digital image correlation. Proceedings of the 1st
incremental change in shear stress, which showed a Int. Symp. On Computational Geomechanics (ComGeo
continuous increase in variation after the peak stress up to 1),
the residual state in both loose and dense sand, indicating
the formation of the persistent shear band or the Hamza, M., Shahien, M., and El-Mossallamy, Y. 2009.
concentration of strain localization. A future interest will be Proceedings of the 17th International Conference on
the investigation of the applicability of coefficient a for fine- Soil Mechanics and Geotechnical Engineering: The
grained soil. academia and practice of geotechnical engineering,
Alexandria, Egypt, 5-9 October 2009.
5. ACKNOWLEDGEMENT
Kia, M. 2012. Measuring Pore-water Pressure in Partially
The authors would like to acknowledge technological staff Frozen Soils. Ph.D. thesis, Faculty of Graduate Studies
Lucas Duerksen and Christine Hereygers, from UofA and Research, University of Alberta, Edmonton, AB.
Geotechnical Center, Department of Civil and
Environmental Engineering, University of Alberta, for Liang Y, Beier N, and Sego DC. 2022. New Method for
assistance in design and construction of the apparatus. Internal Pore-Water Pressure Measurements.
The financial support of Natural Sciences and Engineering Geotechnical Testing Journal, 45(2): 490-502.
Research Council of Canada (NSERC) Discovery Grant
Program (RGPIN-2019-04573) is gratefully acknowledged. Liang Y, Beier N, and Sego DC. 2023. Strength of Partially
Frozen Sand Under Triaxial Compression. Canadian
Geotechnical Journal, 60(9):1277-1288.
6 REFERENCES
Mokni, M., and Desrues, J. 1999. Strain localization
Alikarami, R., Andò, E., Gkiousas-Kapnisis, M., Torabi, A., measurements in undrained plane‐strain biaxial tests
and Viggiani, G. 2015. Strain localization and grain on Hostun RF sand. Mechanics of Cohesive‐frictional
breakage in sand under shearing at high mean stress: Materials: An International Journal on Experiments,
insights from in situ X-ray tomography. Acta Modelling and Computation of Materials and
Geotechnica, 10: 15-30. Structures, 4(4): 419-441.

Alshibli, K. A., Batiste, S. N., and Sture, S. 2003. Strain Rechenmacher, A. L. 2006. Grain-scale processes
localization in sand: plane strain versus triaxial governing shear band initiation and evolution in sands.
compression. Journal of geotechnical and geo- Journal of the Mechanics and Physics of Solids, 54(1):
environmental engineering, 129(6): 483-494. 22-45.

Andò, E., Hall, S. A., Viggiani, G., Desrues, J., and Rudnicki, J. W., and Rice, J. 1975. Conditions for the
Bésuelle, P. 2012. Experimental micromechanics: localization of deformation in pressure-sensitive
grain-scale observation of sand deformation. dilatant materials. Journal of the Mechanics and
Géotechnique Letters, 2(3): 107-112. Physics of Solids, 23(6): 371-394.

Desrues, J., Lanier, J., and Stutz, P. 1985. Localization of Skempton, A. 1954. The pore-pressure coefficients A and
the deformation in tests on sand sample. Engineering B. Geotechnique, 4(4): 143-147.
fracture mechanics, 21(4): 909-921.
Tagliaferri, F., Waller, J., Andò, E., Hall, S. A., Viggiani, G.,
Desrues, J., and Viggiani, G. 2004. Strain localization in Bésuelle, P., and DeJong, J. T. 2011. Observing strain
sand: an overview of the experimental results obtained localization processes in bio-cemented sand using x-
in Grenoble using stereophotogrammetry. International ray imaging. Granular Matter, 13: 247-250.
journal for numerical and analytical methods in
geomechanics, 28(4): 279-321. Takano, D., Lenoir, N., Otani, J., and Hall, S. A. 2015.
Localised deformation in a wide-grained sand under
Finno, R. J., Harris, W. W., Mooney, M. A., and Viggiani, triaxial compression revealed by X-ray tomography and
G. 1996a. Strain localization and undrained steady state digital image correlation. Soils and Foundations, 55(4):
of sand. Journal of Geotechnical Engineering, 122(6): 906-915.
462-473.
Tordesillas, A., Walker, D. M., Andò, E., and Viggiani, G.
Finno, R. J., Harris, W. W., Mooney, M. A., and Viggiani, 2013. Revisiting localized deformation in sand with
G. 1996b. Strain localization and undrained steady complex systems. Proceedings of the Royal Society A:
state of sand. Journal of Geotechnical Engineering, Mathematical, Physical and Engineering Sciences,
122(6): 462-473. 469(2152), 20120606.
Vardoulakis, I., Goldscheider, M., and Gudehus, G. 1978.
Formation of shear bands in sand bodies as a
bifurcation problem. International Journal for numerical
and analytical methods in Geomechanics, 2(2): 99-128.

Wang, Q., and Lade, P. V. 2001. Shear banding in true

triaxial tests and its effect on failure in sand. Journal of
engineering mechanics, 127(8); 754-761.
Extending dilatancy-based constitutive
models to hydromechanical response of
unsaturated gold mine tailings

Mina Mofrad, Mehdi Pouragha, & Paul H. Simms

Department of Civil & Environmental Engineering – Carleton University, Ottawa, ON, Canada

ABSTRACT

Hard rock tailings predominantly contain mixtures of low plasticity sand and silt with wide particle size distributions ranging
from sub-millimetres to few micrometers. This induces a wide water retention curve spanning over multiple decades of
matric suction with potential of significant hydraulic hysteresis during drying and wetting cycles. In some climates and for
some deposition schemes, tailings may undergo complex hydro-mechanical stress paths including repeated cycles of
drying and wetting. It is now commonly understood that the influence of past desiccation of tailing shear behaviour can be
substantive. The partially saturated nature of tailings makes it necessary to adopt proper constitutive models capable of
capturing the coupled hydro-mechanical phenomena associated with such regimes. Most commonly used unsaturated
constitutive models, such as Barcelona Basic Model and Glasgow Coupled Model (GCM), are developed based on the
Cam-Clay theory which is applicable mostly to clayey materials. Hard rock tailings, however, consist of frictional particles
whose volumetric response requires a more careful treatment of plastic dilatancy. The current study applies the general
framework developed in GCM to a dilatancy-based constitutive model, Norsand, to better capture the behaviour of tailings
subjected to desiccation. The model is extended by considering two additional hydraulic yield surfaces and defining
appropriate coupling between mechanical and hydraulic processes. The model has been calibrated and used to reproduce
triaxial and simple shear tests carried on an unsaturated gold mine tailing. In particular, the accuracy of the model in
capturing the main characteristics of the tailings’ shear behaviour is investigated.

RÉSUMÉ

Les résidus de roche dure contiennent principalement des mélanges de sable et de limon à faible plasticité avec de larges
distributions granulométriques allant de sous-millimètres à quelques micromètres. Cela induit une large courbe de rétention
d'eau s'étendant sur plusieurs décennies d'aspiration matricielle avec un potentiel d'hystérésis hydraulique important
pendant les cycles de séchage et de mouillage. Dans certains climats et pour certains schémas de dépôt, les résidus
peuvent subir des trajectoires de contraintes hydromécaniques complexes, notamment des cycles répétés de séchage et
d'humidification. Il est maintenant communément admis que l'influence de la dessiccation passée sur le comportement de
cisaillement des résidus peut être substantielle. La nature partiellement saturée des résidus rend nécessaire l'adoption de
modèles de comportement appropriés capables de capter les phénomènes hydromécaniques couplés associés à de tels
régimes. Les modèles constitutifs insaturés les plus couramment utilisés, tels que le modèle de base de Barcelone et le
modèle couplé de Glasgow (GCM), sont développés sur la base de la théorie Cam-Clay qui s'applique principalement aux
matériaux argileux. Cependant, les résidus de roche dure sont constitués de particules de frottement dont la réponse
volumétrique nécessite un traitement plus minutieux de la dilatance plastique. La présente étude applique le cadre général
développé dans GCM à un modèle constitutif basé sur la dilatance, Norsand, pour mieux capturer le comportement des
résidus soumis à la dessiccation. Le modèle est étendu en considérant deux surfaces hydrauliques supplémentaires et en
définissant un couplage approprié entre les processus mécaniques et hydrauliques. Le modèle a été calibré et utilisé pour
reproduire des essais de cisaillement triaxiaux et simples effectués sur un résidu de mine d'or non saturé. En particulier,
la précision du modèle dans la capture des principales caractéristiques du comportement de cisaillement des résidus est
étudiée.

1 INTRODUCTION followed by desiccation, which is then followed by another

cycle of rewetting due to the placement of new layers(Qi et
Hard rock tailings are residual materials that are left al., 2020; Simms, 2017). Past desiccation can have a
over after the extraction of minerals from rock deposits, strong positive impact on the strength behaviour of even
which are conventionally deposited as slurries, the water the fully re-saturated tailings (Daliri et al., 2014, 2016).
originating from mineral processing requirements. In During the desiccation phase, the tailings enter an
certain climates and for certain deposition practices (e.g., unsaturated state while experiencing significant
Filtered or thickened), tailings may undergo complex irreversible structural change, and as a result, their hydro-
hydro-mechanical loading history due to consecutive mechanical behaviour upon re-wetting will be different. The
desiccation and saturation cycles. Such stress histories in behaviour of these unsaturated materials naturally
multilayer deposition involve self-weight consolidation, transcends the conventional soil mechanics that is
established for soils in dry or saturated conditions.
Contemporary unsaturated soil mechanic extends the
𝜎𝑖𝑖 3
traditional theories to include new physics such as matric 𝑝= , 𝑞 = √ 𝑠𝑖𝑗 𝑠𝑖𝑗 , 𝑠𝑖𝑗 = 𝜎𝑖𝑗 − 𝑝𝛿𝑖𝑗 [1]
suction, coupling of hydraulic and mechanical behaviours, 3 2
and hysteretic effects during cycles of drying and wetting
(Fredlund, 1993; Lu & Likos, 2004).
In order to predict accurately the hydro-mechanical 2 1
𝜀𝑣 = 𝜀𝑖𝑖 , 𝜀𝑠 = √ 𝑒𝑖𝑗 𝑒𝑖𝑗 , 𝑒𝑖𝑗 = 𝜀𝑖𝑗 − 𝜀𝑣 𝛿𝑖𝑗 [2]
behaviour of tailings that have experienced desiccation, 3 3
proper constitutive models are required that are able to
capture and predict their behaviour in their partially with 𝜹 being the Kronecker delta and repeated indices
saturated state. In particular, the concept of effective stress denoting summation. The over dot (𝑋̇) represents
has been extended to include additional variables such as incremental rate.
matric suction (Bishop, 1959). However, unlike the
saturated case, the behaviour of unsaturated soils has 2.1 Nor-sand Model
been shown not to be reducible to a single stress state
variable (Duriez et al., 2018). As such, additional state The Nor-sand model (Jefferies, 1993) is a generalized
parameters related to matric suction, as well as other Cambridge-type constitutive model developed from the
proper coupling terms, are also incorporated into fundamental assumptions of critical state theory. Initially it
constitutive modelling frameworks to properly address the was primarily used for modeling the mechanical behaviour
unsaturated soil behaviours. This led to a new branch of of sands. However, more recently, it is applied to soils with
constitutive modelling in soil mechanics for partially particle size range from silt to sand and aims at capturing
saturated soil. their unique characteristics. The model considers the
Amongst the first such constitutive models are the effective stress presented by Terzaghi (1943) as stress
Barcelona Basic Model (BBM) (Alonso et al., 1990) and the state variable which is defined as:
Glasgow Coupled Model (GCM) (Lloret-Cabot et al., 2013)
which extends the Modified Cam Clay Model (MCC) 𝜎′𝑖𝑗 = 𝜎𝑖𝑗 − 𝑢𝑤 𝛿𝑖𝑗 [3]
(Roscoe & Burland, 1968) to unsaturated soils. However,
such an effort has not been made for frictional granular
materials mainly because of their lower sensitivity to where 𝝈 is the total stress tensor, 𝑢𝑤 is the water pressure.
suction that is manifested in their relatively narrow soil- A nonlinear isotropic elasticity is adopted in terms of bulk
water characteristic curve (SWCC) compared to sensitive modulus, 𝐾, shear modulus, 𝐺, in which the shear modulus
clayey materials. is function of current mean effective stress, 𝑝′:
Hard rock tailings, on the other hand, contain a 𝜎̇ ′𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙 𝜀̇𝑘𝑙
relatively wide range of particle sizes from silts to sands. [4]
Although the particles are mostly cohesionless, the wide
particle size distribution leads to a correspondingly wide 2 1
SWCC as well as a more pronounced hydraulic hysteresis 𝐶𝑖𝑗𝑘𝑙 = (𝐾 − 𝐺) 𝛿𝑖𝑗 𝛿𝑘𝑙 + 2𝐺 [ (𝛿𝑖𝑘 𝛿𝑗𝑙 + 𝛿𝑖𝑙 𝛿𝑗𝑘 )] [5]
3 2
during cycles of drying and wetting. Moreover, the frictional
nature of these materials requires constitutive models that
go beyond the classical Cam-clay theory to realistically 2(1 + 𝜈) [6]
𝐾= 𝐺
capture the dilatancy response of tailings. 3(1 − 2𝜈)
The current study adopts the hydromechanical coupling
framework developed in GCM in order for extending a 𝐺 = 𝐼𝑟 𝑝′ [7]
dilatancy-based constitutive model to partially saturated
conditions. The framework is applied to a commonly used where 𝜈 is Poisson ratio and 𝐼𝑟 is material constant.
constitutive model, Norsand (Jefferies, 1993), that aims at Nor-sand introduces a state parameter 𝜓 that controls
better capturing the dilatancy behaviour of granular the dilatancy and is defined as the difference between
materials. Modifications introduced recently in (Mofrad et current void ratio (𝑒) and critical void ratio (𝑒𝑐 ) at the same
al., 2023) help in better capturing the hydraulic hysteresis. mean effective stress (𝑝′).
The extended modelled is employed for predicting the
behaviour of saturated gold mine tailings and its 𝜓 = 𝑒 − 𝑒𝑐
performance is compared with the GCM. The comparisons [8]
show that the extended model is capable of capturing the 𝑒𝑐 = 𝛤 − 𝜆 𝑙𝑛 𝑝′
pertinent trends when the tailing material is subjected to
hydromechanical loadings. in which the Γ and 𝜆 are material parameters. The model
has a yield surface that is defined by a bullet-like shape
2 CONSTITUTIVE MODEL similar to the Cam Clay, as expressed in Eq. [9, which is
cropped by a lower cap to better capture the hysteresis
In this section, we briefly review the most relevant parts during cyclic loading.
of the Norsand model. In the following formulations, the
hydrostatic and the deviatoric components of stress
tensors, and their conjugate strains are defined as follows:
 Figure 1. Yield surface of the Nor-sand model in stress ratio space, a) loose soil b) dense soils.

𝑞 𝑝′ 2.2 Extended Nor-sand Model

𝐹= ′ + 𝑀𝑖 (𝑙𝑛 ( ′ ) − 1) = 0
𝑝 𝑝𝑖 [9]
In its original form, the Nor-sand model is applicable to
𝑝′ 𝑖 −𝜒𝑖 𝜓𝑖
dry and saturated materials. Here we propose an extension
for ( ′ ) < 𝑒𝑥𝑝 ( ) for extending such a model to unsaturated conditions by
𝑝 𝑀𝑖,𝑡𝑐
adopting the framework used in Mofrad et al. (2023) which
itself is based on the GCM framework. The procedure
The variables 𝑀𝑖 , 𝜓𝑖 and 𝜒𝑖 depend on material constants involves (1) adopting the suitable state variables for
and are defined as below: unsaturated materials, (2) including new constitutive
relations that describe the evolution of the newly added
𝑁𝜒𝑖 |𝜓𝑖 | state variable, in this case, a suction-saturation relation,
𝑀𝑖 = 𝑀 (1 − ) [10]
𝑀𝑡𝑐 and (3) introducing proper coupling terms in the evolution
laws, such as hardening law.
𝜒𝑡𝑐 Following the thermodynamic framework by Houlsby
𝜒𝑖 = [11]
𝜒 𝜆 (1997) the bishop’s effective stress, 𝜎 ∗ , and modified
1 − 𝑡𝑐
𝑀𝑡𝑐 matric suction, 𝑠 ∗, are adopted as stress state variables
with total strain, 𝜀, and saturation ratio, 𝑆𝑟 , being their
𝑝′ [12] energy conjugates, respectively.
𝜓𝑖 = 𝜓 − 𝜆 ln ( ′𝑖 )
𝑝
𝜎𝑖𝑗∗ = 𝜎𝑖𝑗 − (𝑆𝑟 𝑢𝑤 + (1 − 𝑆𝑟 ) 𝑢𝑎 )𝛿𝑖𝑗
where 𝑁 is a materials constant and 𝑀 is the critical state [16]
slope in 𝑞 − 𝑝′ space which depends on the Lode angle 𝜃: 𝑠 ∗ = 𝑛 (𝑢𝑎 − 𝑢𝑤 )

𝑀𝑡𝑐 2 3 𝜋 where 𝝈 is the total stress tensor, 𝑢𝑎 and 𝑢𝑤 are the air and
𝑀 = 𝑀𝑡𝑐 −
3 + 𝑀𝑡𝑐
𝑐𝑜𝑠 ( 𝜃 + )
2 4 [13] water pressures, 𝑆𝑟 is the degree of saturation and 𝑛 is the
porosity. Explicitly considering 𝑠 ∗ and 𝑆𝑟 as energy
with 𝑀𝑡𝑐 being the slope of critical state line along triaxial conjugates allows Extended Nor-sand to capture more
compression path, and 𝜒𝑡𝑐 being a material parameter. consistently the transition between saturated and
The yield surface is shown in Figure 1 for initially loose unsaturated states, as well as to better include the
and dense samples. The size of yield surface evolves with influence of retention hysteresis.
the image hardening parameter, 𝑝′𝑖 , related to the change Similar to the GCM, the yield surface of the Extended
𝑝
of plastic deviatoric strain increments, 𝜀̇𝑞 : Nor-sand model is extended in the third dimension of 𝑠 ∗
and two additional caps describing the hydraulic yield
surfaces were added as follows:
𝑀𝑖 𝑝′ 2 𝜒𝑖 𝜓𝑖 𝑝′ 𝑝
𝑝′̇𝑖 = 𝑝′ 𝑖 𝐻 ( ′ ) [exp (− ) − ′𝑖 ] 𝜀̇𝑞 [14]
𝑀𝑖,𝑡𝑐 𝑝 𝑖 𝑀𝑖,𝑡𝑐 𝑝 𝐹𝑆𝐼 = 𝑠 ∗ − 𝑠𝐼∗ = 0
Nor-sand adopts a non-associated flow rule through a [17]
𝐹𝑆𝐷 = 𝑠𝐷∗ − 𝑠 ∗ = 0
stress–dilatancy relationship expressed as follows:
𝑞 where 𝑠𝐼∗ and 𝑠𝐷∗ are parameters whose evolution (hydraulic
𝐷𝑝 = 𝑀𝑖 − hardening) laws will be specified later.
𝑝′ [15] The additional constitutive relation for the newly added
hydraulic state variables is given in terms a suction-
saturation expression, or SWCC, which in this case, is
assumed to be described by a family of S-shape curves 𝑠̇𝑥∗
that follow the general form of the commonly used Van 𝑀𝑖 𝑝∗
2
−𝜒𝑖 𝜓𝑖 𝑝′ 𝑝
Genuchten expression (Van Genuchten, 1980). We = 𝑠𝑥∗ [𝑘2 𝐻 ( ′ ) (𝑒𝑥𝑝( ) − ∗𝑖 )(𝜀̇𝑞 )
𝑀𝑖,𝑡𝑐 𝑝 𝑖 𝑀𝑖,𝑡𝑐 𝑝 [20]
assume that the hardening of hydraulic yield surfaces
happens such that the modified suction and saturation (1 + (𝑎𝑠𝑥∗ )𝑏 )1+𝑚 𝑝
evolve according to the following relation: + (−𝑆𝑟̇ )] , 𝑥 = 𝐼 𝑜𝑟 𝐷
𝑏𝑚(𝑎𝑠𝑥∗ )𝑏

𝑆𝑟 = (1 + (𝑎𝑠 ∗ )𝑏 )−𝑚 where 𝑘1 and 𝑘2 are model parameters. The evolution of

[18]
the yield surface is described by the hardening parameters,
where 𝑎 and 𝑏 are model constants and 𝑚 is a parameter 𝑝̇ ′𝑖 , 𝑠̇𝐼∗ and 𝑠̇𝐷∗ , expressed in terms of plastic deviatoric strain
that is calculated based on the current state of material at 𝑝
increments, 𝜀̇𝑞 , and plastic increment of the degree of
beginning of hydraulic yielding. Upon yielding on the 𝑝
saturation, 𝑆𝑟̇ . Similar to the original NorSand, the flow rule
hydraulic yield surfaces, the hydraulic state of the soil related to frictional yield surface remains non-associated,
evolves along a curve described by Eq. [18 while during while the yielding on hydraulic yield surfaces is assumed to
reversible hydraulic processes, the hydraulic state be associated.
rebounds, similar to the GCM, on a semi-logarithmic line
with a constant slope of 𝜅𝑠 until the next hydraulic surface 3 RESULTS
is reached. At first instance of yielding, the parameter 𝑚 is
updated such that the current state of the material falls on The Extended Norsand model has been coded using
a curve belonging to family of curves described by Eq.[18. explicit numerical integration method. The model is then
The value of 𝑚 remains fixed during continuous yielding. calibrated and used in reproducing the behaviour of
Error! Reference source not found. shows schematically unsaturated tailings. The experimental dataset has been
the such scanning curves together with the log-linear extracted from Daliri et al. (2014, 2016) with the aim of
hydraulic rebound curves. investigating the effect of hydraulic stress history on the
mechanical behaviour of unsaturated gold mine tailings.
Soil specimens were first prepared in a slurry form and
deposited in two layers. The bottom layer experienced the
initial self-weight settling followed by desiccation to three
different gravimetric water contents, 𝑊𝑑 = 23%, 17%, 12%,
and then rewetted to saturated condition once the second
layer was placed. Soil samples taken from the bottom
layer, which all are saturated with history of desiccation to
different water content, were tested in a simple shear
apparatus in undrained condition subjected to vertical
stress of 50 𝑘𝑃𝑎.
The testing results during the desiccation/rewetting and
Figure 2. Example of scanning curves during drying and the shearing process were used to calibrate the constitutive
wetting paths. Starting at point A and assuming the state to model. In particular, the behaviour along rewetting and at
be on 𝑠𝐼 surface, the sample is desiccated to point B. Upon the initial states of the shearing were used to calibrate the
rewetting, the suction increases with a constant (semi-log) elastic properties, while the results at the end of the
slope of 𝜅𝑠 until the lower yielding suction, 𝑠𝐷 is reached at shearing test were used to calculate parameters
C, after which the hydraulic state will move along the van associated with the characterizing the critical state. The
Genuchten curve containing the current state. If the sample calibrated parameters are given below.
is dried again (at D), the state will move on the rebound
curve until the value of 𝑠𝐼 is reached at E. Afterwards, the Extended Norsand Model (14 parameters): 𝜈 = 0.3, 𝑁 =
van Genuchten curve is followed. 0.5, 𝐻 = 15.8, 𝛤 = 1.1, 𝑀𝑡𝑐 = 1.01, 𝑥𝑡𝑐 = 2.8, 𝐼𝑟 = 42, 𝜆 =
0.085, 𝑘1 = 0.73, 𝑘2 = 0.25, 𝑎 = 0.045, 𝑏 = 0.8, 𝑚 =
The hardening laws for the parameters describing the 0.4, 𝜅𝑠 = 0.001.
size of the yield surfaces are modified to introduce the
coupling terms similar to the GCM: Figure 3 and Figure 4 compare the predictions by the
Extended Norsand model with the experimental results
shown as square symbols. For a better comparison, the
𝑀𝑖 𝑝∗
2
−𝜒𝑖 𝜓𝑖 𝑝′ 𝑝
results from Modified Glasgow coupled model (MGCM)
𝑝̇ ′𝑖 = 𝑝′ 𝑖 𝐻 ( ′ ) (𝑒𝑥𝑝( ) − ∗𝑖 )(𝜀̇𝑞 ) developed in Mofrad et al. (2023) has also been provided.
𝑀𝑖,𝑡𝑐 𝑝 𝑖 𝑀𝑖,𝑡𝑐 𝑝
The figures include the change in void ratio vs. suction
[ stress and net normal stress for two samples (Figure 3),
1
1 𝑚+1 [19] and shear stress vs. shear strain together with the effective
( )
𝑘1 𝑆𝑟 𝑝 stress path (Figure 4). For visualization purposes, the
+ (−𝑆̇𝑟 )]
𝑏𝑚 1 evolution of the yield surfaces of Extended Norsand model
1 𝑚 in 𝑝∗ − 𝑠 ∗ and 𝑝∗ − 𝑞 spaces are presented in Figure 5 for
( ) −1
𝑆𝑟 ]
sample with history of desiccation to 𝑤𝑑 = 12%.
Figure 3. Comparison between the experimental results
and the model predictions for volume change. Samples are
desiccated to 𝑊𝑑 = 12% (a, c), or 𝑊𝑑 = 23% (b, d), then
rewetted to fully saturated state, and consolidated to
50 𝑘𝑃𝑎.

Figure 5. The evolution of the yield surfaces for Extended

Norsand model for the sample with 𝑤𝑑 = 12%. Figures (a)
shows the hydromechanical yield surface in 𝑝∗ − 𝑠 ∗ space,
and (b) shows the mechanical yield surface in 𝑝∗ − 𝑞
space.

4 DISCUSSION

The performance of the Extended Norsand constitutive

model is compared with MGCM in predicting the shear
response of desiccated/rewetted gold tailings. Focusing on
the void ratio changes during the desiccation, rewetting,
and consolidation, we observe that both models provide
realistic trends. The realistic representation of void ratio
Figure 4. Comparison between the experimental results changes in Figure 4 is associated with the appropriate
and the model predictions; (a, c) shear stress vs strain, and hydro-mechanical coupling and proper relationship
(b, d) simple shear effective stress path. between suction and saturation ratio proposed in the
extended model. As depicted in Figure 5, during the initial
drying stage, the soil is positioned on the SI hydraulic yield
surface and exhibits yielding as the water content
decreases. This leads to an expansion of the elastic zone
in both hydraulic and mechanical domains. Subsequently,
upon rewetting, the sample's behavior remains elastic until
it reaches the SD yield surface. Beyond this point, the
sample begins to yield, and the coupling between hydraulic
and mechanical behaviors causes the mechanical yield
surface to contract. The contraction of the mechanical yield
surface during rewetting results in the sample yielding
during consolidation loading, effectively capturing the
experimental observations of void ratio change during
consolidation within the proposed model.
Turning to the shearing results, comparing the results
of Extended Norsand model and MGCM shows that the
Extended Norsand predictions are more realistic due to its 6 REFERENCES
non-associated plastic flow rule and the hardening with
plastic shear strain (Eq. [19) that allows an asymptotic Alonso, E. E., Gens, A., & Josa, A. (1990). A constitutive
approach to the critical state line. This is also facilitated by model for partially saturated soils. Géotechnique,
the fact that, unlike models such as GCM, the critical state 40(3), 405–430.
values for stress and void ratio are allowed to be reached Bishop, A. W. (1959). The principle of effective stress.
separately. Therefore, the state of the material can Teknisk Ukeblad, 39, 859–863.
approach its critical state slope in 𝑞 − 𝑝 space while the Daliri, F., Kim, H., Simms, P., & Sivathayalan, S. (2014).
void ratio has not reached its critical state value. As a Impact of desiccation on monotonic and cyclic shear
result, a critical state is not realized, and stresses will strength of thickened gold tailings. Journal of
evolve asymptotic to the critical state line in the 𝑞 − 𝑝 Geotechnical and Geoenvironmental Engineering,
space. 140(9), 4014048.
The Extended Norsand model is able to capture the Daliri, F., Simms, P., & Sivathayalan, S. (2016). Shear and
general trend of increase in shear strength as desiccation dewatering behaviour of densified gold tailings in a
intensifies. However, its accuracy decreases for the lower laboratory simulation of multi-layer deposition.
degree of desiccation sample (𝑤𝑑 = 23 %) especially in Canadian Geotechnical Journal, 53(8), 1246–1257.
predicting the effective normal stress at which the failure Duriez, J., Wan, R., Pouragha, M., & Darve, F. (2018).
stress is reached (end of solid lines). Revisiting the existence of an effective stress for wet
The shearing results (Figure 4) show that the Extended granular soils with micromechanics. International
model is also able to capture the increase of shear Journal for Numerical and Analytical Methods in
hardening behaviour with the increase of level of Geomechanics, 42(8), 959–978.
desiccation. Fredlund, D. G. (1993). Soil mechanics for unsaturated
Finally, the comparison between the fictional and soils. John Wiley &amp; Sons.
cohesive models should be understood only in the context Houlsby, G. T. (1997). The work input to an unsaturated
of their complexity and the purpose of use. The Norsand granular material. Géotechnique, 47(1), 193–196.
model is a complex model in terms of its structure and Jefferies, M. G. (1993). Nor-Sand: a simle critical state
number of models parameters. It seems possible that model for sand. Géotechnique, 43(1), 91–103.
similar results could be achieved via simpler models such Lloret-Cabot, M., Sánchez, M., & Wheeler, S. J. (2013).
as Wan and Guo model (1998). The authors are exploring Formulation of a three-dimensional constitutive
extending this and other models to the unsaturated state. model for unsaturated soils incorporating
mechanical–water retention couplings. International
5 CONCLUSIONS Journal for Numerical and Analytical Methods in
Geomechanics, 37(17), 3008–3035.
Gold mine tailings consist mostly of granular Lu, N., & Likos, W. J. (2004). Unsaturated Soil Mechanics.
frictional materials such as silt and sand whose dilatancy J. Wiley.
behaviour is known to go beyond simple associated Mofrad, M., Pouragha, M., & Simms, P. (2023).
plasticity assumptions adopted in Cam-clay theory. In this Constitutive modelling of unsaturated gold tailings
study, we follow the framework developed in the GCM for subjected to drying and wetting cycles. International
extending constitutive model developed for dry and Journal for Numerical and Analytical Methods in
saturated soils to unsaturated conditions. The framework is Geomechanics.
applied to a commonly used model for sands and silts Qi, S., Simms, P., Daliri, F., & Vanapalli, S. (2020).
(Norsand) by extending the yield surface in the third Coupling elasto-plastic behaviour of unsaturated
direction dedicated to hydraulic state. Appropriate coupling soils with piecewise linear large-strain consolidation.
terms are also included in the hardening laws of both Géotechnique, 70(6), 518–537.
mechanical and hydraulic yield surfaces. Roscoe, K., & Burland, J. B. (1968). On the generalized
The wide particle size distribution of hard rock tailings stress-strain behaviour of wet clay.
induces an associated wide SWCC. Therefore, the suction- Simms, P. (2017). 2013 Colloquium of the Canadian
saturation can also exhibit considerable hysteretic Geotechnical Society: Geotechnical and
response to drying/wetting cycles. Therefore, the geoenvironmental behaviour of high-density tailings.
formulation in the GCM is extended to better capture the Canadian Geotechnical Journal, 54(4), 455–468.
evolution of saturation with suction. Terzaghi, K. (1943). Theoretical Soil Mechanics John Wiley
The developed model is calibrated based on simple and Sons Inc. New York, 314.
shear test carried out on gold mine tailings desiccated to Van Genuchten, M. T. (1980). A closed-form equation for
different levels of water content and re-saturated, to mimic predicting the hydraulic conductivity of unsaturated
the layered deposition, and then sheared under undrained soils. Soil Science Society of America Journal, 44(5),
conditions. The extended model is shown to capture the 892–898.
dependency of the shearing response to the initial Wan, R. G., & Guo, P. J. (1998). A simple constitutive
wetting/drying cycles with acceptable accuracy. A similar model for granular soils: modified stress-dilatancy
approach can be applied to other existing constitutive approach. Computers and Geotechnics, 22(2), 109–
models in order to coherently extend them to be applicable 133.
to partially saturated geomaterials.
On the numerical issues of the NorSand
model in the simulation of undrained
behavior of granular soils
Seyyed Kazem Razavi & Samuel Yniesta
Department of Civil, Geological, and Mining Engineering,
Polytechnique Montreal, Montreal, QC, Canada

ABSTRACT
The NorSand model is one of the most used soil models to simulate the undrained and drained response of granular
materials. The model requires the following parameters: two elastic material properties, three to four parameters to define
the critical state in e-p and q-p space, two parameters to define the image stress ratio and limiting dilatancy, and two
hardening parameters. These few material properties and a straightforward procedure to determine them are the main
reasons for the considerable growth of the NorSand application in the simulation of the stress-strain behavior of granular
materials. However, the material properties are obtained mainly based on drained test results conducted on dense
samples. Therefore, the model uses some assumptions and mathematical techniques to modify its response to simulate
the undrained behavior of soil states looser than the critical state. This paper explains how the NorSand model employs
this mathematical technique and discusses the effectiveness of this approach.

RÉSUMÉ
La loi de comportement NorSand est un des modèles de sol les plus utilisés pour simuler le comportement non-drainé et
drainé des matériaux granulaires. Le modèle demande la définition des paramètres d’entrée suivant : deux propriétés
élastiques, trois ou quatre paramètres pour définir l’état critique dans l’espace e-p et q-p, deux paramètres pour définir le
ratio des contraintes image et la dilatance maximum, et deux paramètres d’écrouissage. Le faible nombre de paramètres
et une procédure de calibration simple sont les principales raisons pour l’augmentation de l’usage de NorSand dans la
simulation du comportement contrainte-déformation des matériaux granulaires. Cependant, ces propriétés sont obtenues
principalement sur la base de tests drainés, réalisés sur des échantillons denses. Par la suite, le modèle repose sur des
hypothèses et des techniques mathématiques pour modifier sa réponse pour simuler le comportement non-drainé des sols
lâches. Cet article présente comment NorSand effectue cette modification, et discute de l’efficacité de cette approche.

1 INTRODUCTION 1999, Dafalias and Manzari, 2004, Lu and Huang, 2014),

resulting in the development of models well capable of
It is essential to study in depth the performance of a simulating sand’s behavior.
constitutive soil model in order to understand its principles, Likewise, understanding the performance of the NorSand
underlying concepts, and limitations, and be able to model in simulating the drained and undrained behavior of
develop novel ideas to improve said model. Jefferies granular soils is necessary to identify its limitations. The
(1993), for instance, conducted a thorough analysis of principal drawback of the original NorSand model
traditional Cambridge-type models (Roscoe et al. 1963) to (Jefferies, 1993) is its inability to accurately simulate the
reveal why they fail to simulate accurately the behavior of undrained behavior of loose states, such as flow
granular soils. His research showed that the attachment of liquefaction (Chang and Jefferies, 2020, Jefferies and
the yield surface to the critical state was the root cause of Been, 2015). To overcome this limitation, Jefferies and
the issue, resulting in a single normally consolidated line. Been (2015) proposed modifying the hardening rule by
This, however, is inconsistent with the behavior of granular adding a new term known as the cap-softening term, where
soils, which exhibit infinite normally consolidated lines. To the user is only capable to accept or reject its presence.
address this problem, Jefferies (1993) proposed a new This paper delves into the reasons behind the need for this
model, the NorSand model, in which the yield surface is modification and evaluates the effectiveness of the
separated from the critical state line and the model only proposed approach.
reaches it at large strains. This necessitated that the Note: All the mean confining pressures (p) in this research
location and size of the NorSand yield surface be controlled are effective stresses.
by the state parameter (Been and Jefferies, 1985). This
added feature enhanced the capability of Cambridge-type 2 NORSAND YIELD SURFACE
models to simulate the behavior of granular soils. The
influence of the state parameter on the size and location of The formulation of the NorSand model presented by
the yield surface and the use of an associated flow rule Jefferies and Been (2015) is considered here. NorSand
implies that the dilatancy is also a state parameter- uses an associated flow rule to define its yield surface by
dependent quantity. The concept of state-dependent integrating the stress-dilatancy rule (Equation 1).
dilatancy later has been recognized thoroughly and is used Depending on this rule, the NorSand yield surface
in models with non-associated flow rules (e.g. Li et al. (Equation 2) takes a bullet-type shape, similar in shape to
the yield surface of the original Cam-Clay model (Roscoe
et al. 1963). 𝑒𝑐 = Γ − 𝜆𝑙 ln(𝑝) [7]

𝐷 = 𝑀𝑖 − 𝜂 [1] 𝜒𝑖 = 𝜒𝑡𝑐 ⁄(1 − 𝜆𝑙

𝜒𝑡𝑐
) [8]
𝑝 𝑀𝑡𝑐
𝜂 = 𝑀𝑖 (1 − 𝑙𝑛 ( )) [2]
𝑝𝑖

The point on the yield surface that exhibits zero dilatancy

is referred to as the image state, which is characterized by
three parameters: image stress ratio (𝑀𝑖 ), image confining
pressure (𝑝𝑖 ), and image state parameter (𝜓𝑖 ). In order to
determine the image stress ratio, the NorSand model
combines various concepts and assumptions, including the
State-Dilatancy law, Nova's rule, and the Stress-Dilatancy
rule. The State-Dilatancy law describes the relationship
between limiting dilatancy (𝐷𝐿 ) and its corresponding state
parameter (𝜓𝐿 ).
During a drained triaxial compression (DTC) test, there is
a specific point on the stress-strain path where the soil
experiences minimal dilatancy (𝐷𝑡𝑐−𝑚𝑖𝑛 ) and maximum
stress ratio (𝜂𝑡𝑐−𝑚𝑎𝑥 ). At this point, the state parameter can
be defined as 𝜓𝑡𝑐−𝑚𝑖𝑛 . NorSand considers 𝐷𝑡𝑐−𝑚𝑖𝑛 and
𝜓𝑡𝑐−𝑚𝑖𝑛 as limiting dilatancy (𝐷𝐿 ) and its corresponding
state parameter (𝜓𝐿 ), respectively. Moreover, the stress
ratio at the limiting dilatancy (𝜂𝐿 ) is equal to 𝜂𝑡𝑐−𝑚𝑎𝑥 . In
most cases, the following linear equation fits well with
experimental data:

𝐷𝑡𝑐−𝑚𝑖𝑛 = 𝜒𝑡𝑐 ∙ 𝜓𝑡𝑐−𝑚𝑖𝑛 [3]

Where, 𝜒𝑡𝑐 is a material constant called the state-dilatancy

parameter. For a general drained path conducted on dense
samples, state-dilatancy law has been modified as follows:
𝑀
𝐷𝐿 = 𝜒𝑡𝑐 𝜓𝐿 [4] Figure 1 a) Relationship between limiting Dilatancy (𝐷𝐿 )
𝑀𝑡𝑐
and its corresponding state parameter, b) the concept of
Where 𝑀𝑡𝑐 is the critical stress ratio for triaxial compression limiting Dilatancy on the NorSand yield surface
test, and M is the critical stress ratio depends on the Lode
angle (𝜃). In Equation (7), Γ is the critical void ratio at p=1 kPa, and
𝜆𝑙 is a material constant used to define the linear CSL. As
𝑀𝑡𝑐 2 3 𝜋 mentioned above, for a drained triaxial compression (DTC)
𝑀 = 𝑀𝑡𝑐 − 𝑐𝑜𝑠 ( 𝜃 + ) [5]
3+𝑀𝑡𝑐 2 4 test conducted on dense states, there is a point where the
soil experiences a minimum dilatancy (𝐷𝑡𝑐−𝑚𝑖𝑛 ) with
Lode angle represents the shear mode from triaxial maximum stress ratio (𝜂𝑡𝑐−𝑚𝑎𝑥 ). Nova’s rule relates 𝐷𝑡𝑐−𝑚𝑖𝑛
compression to triaxial extension where the shearing takes to 𝜂𝑡𝑐−𝑚𝑎𝑥 through the following formulation:
place. 𝐷𝐿 is used to control the NorSand yield surface
expansion. It must be noted that for the loose states under 𝐷𝑡𝑐−𝑚𝑖𝑛 = (𝑀𝑡𝑐 − 𝜂𝑡𝑐−𝑚𝑎𝑥 )⁄(1 − 𝑁) [9]
drained condition, there is no measurable quantity as
minimum dilatancy. Therefore, NorSand assumes that Where, 𝑁 is a material constant called shear-volume
Equation (4) can be extended in its linear format for states coupling factor. The original NorSand formulation assumes
looser than critical (see Figure 1). that it can be applied for the whole stress path as follows:

According to Equation (4), 𝜓𝐿 is required to calculate 𝐷𝐿 . 𝐷𝐿 = (𝑀 − 𝜂)⁄(1 − 𝑁) [10]

Therefore, NorSand presents the following relationship to
calculate 𝜓𝐿 from the image state parameter 𝜓𝑖 . By combining Equations (10), (6) and (1), the image stress
ratio can be obtained:
𝑀 𝑀
𝐷𝐿 = 𝜒𝑡𝑐 𝜓𝐿 = 𝜒𝑖 𝜓𝑖 [6]
𝑀𝑡𝑐 𝑀𝑡𝑐 𝜒𝑖 𝜓𝑖
𝑀𝑖 = 𝑀 (1 + 𝑁 ) 𝑓𝑜𝑟 𝜓𝑖 < 0 [11]
𝑀𝑡𝑐
Where, 𝜒𝑖 is called the NorSand state-dilatancy parameter For loose states (𝜓𝑖 ), however, Equation (11) is not valid,
and is a positive value. For linear critical state line shown because in that case, 𝑀𝑖 becomes greater than 𝑀. The
in Equation (7), 𝜒𝑖 can be easily obtained as Equation (8). developers of the NorSand model assume that a symmetric
version of Equation (11) can be used for loose states. deviatoric stress, should also reach 𝑞𝑐 , the deviatoric stress
Subsequently, the image stress ratio can be written in its at critical state.
general format as follows:
𝑝 → 𝑝𝑖 → 𝑝𝑐 𝑜𝑟 𝜓 → 𝜓𝑖 → 0
𝜒𝑖 |𝜓𝑖 | 𝜀𝑞 𝑝 → ∞ ⟹ { [15]
𝑀𝑖 = 𝑀 (1 − 𝑁 ) [12] 𝑞 → 𝑞𝑐 𝑜𝑟 𝜂 → 𝑀
𝑀𝑡𝑐
The second requirement of CST forces that at large strain
For a nonlinear CSL defined with Equation (13), Equation 𝑑𝜓𝑖 = 0. This means that at large strains the state
(8) is no longer valid. parameter after reaching critical state should not change.
𝜉
𝑒𝑐 = 𝛤 − 𝜆 (
𝑝
) [13] 𝜓𝑖 = 𝑒 − (Γ − 𝜆𝑙 ln(𝑝𝑖 )) [16]
𝑝𝑟𝑒𝑓
𝑑𝑝𝑖
𝑑𝜓𝑖 𝑑𝑒
In Equation (13), Γ is the critical void ratio at p=1 kPa, and [17]
𝑝𝑖
= + 𝜆𝑙 =0
𝑑𝜀𝑞 𝑝 𝑑 𝜀𝑞 𝑝 𝑑𝜀𝑞 𝑝
𝜆 and 𝜉 are material constants used to define a nonlinear
CSL. For nonlinear CSL, it can be proved that in Equation
𝑑𝑒 at critical state is zero, therefore the second term in the
(8), 𝜆𝑙 should be calculated from the following equation to
above equation must reach zero at critical state. The
use a nonlinear CSL in the NorSand formulation.
second term can be used as a hardening rule to control the
𝜉 size of yield surface as follows:
𝑝
𝜆𝑙 = 𝜆𝜉 ( ) [14]
𝑝𝑟𝑒𝑓 𝑑𝑝𝑖 𝑝𝑖 𝑝
= 𝐻. 𝑝. ( − 𝑖) [18]
𝑑𝜀𝑞 𝑝 𝑝𝐿 𝑝
The formulation of the NorSand model considering the
CSL, stress-dilatancy rule, yield surface, and image stress Therefore, whenever 𝑝𝐿 < 𝑝, the size of the yield surface
ratio is summarized in Table 1. will increase, otherwise, the yield surface will shrink. For
𝜓𝑖 ≥ 0, 𝐷𝐿 is positive and 𝑝𝐿 ≥ 𝑝𝑖 , and for 𝜓𝑖 ≤ 0 the
Table 1 general formulation used to obtain the yield surface opposite condition will apply. These conditions along with
of NorSand model the NorSand hardening concept are shown schematically
Characteristic Formulation in Figure 2, for (𝜓𝑖 < 0), 𝜂𝐿 = 𝜂𝑚𝑎𝑥 (maximum stress ratio
CSL 𝑒𝑐 = Γ − 𝜆𝑙 ln(𝑝) that can be developed) and 𝐷𝐿 = 𝐷𝑚𝑖𝑛 (minimum dilatancy)
observed under drained tests.
Stress – dilatancy rule 𝐷 = 𝑀𝑖 − 𝜂
𝑝
Yield surface 𝜂 = 𝑀𝑖 (1 − 𝑙𝑛 ( ))
𝑝𝑖
𝜒𝑖 𝑁|𝜓𝑖 |
𝑀𝑖 = 𝑀 (1 − )
𝑀𝑡𝑐
Image Stress ratio 𝜒𝑡𝑐
𝜒𝑖 = 𝜒
1 − 𝜆𝑙 𝑡𝑐
𝑀𝑖,𝑡𝑐
𝑀 − 𝜂𝐿
Nova’s rule 𝐷𝐿 =
1−𝑁
𝑀
State-Dilatancy law 𝐷𝐿 = 𝜒 𝜓
𝑀𝑡𝑐 𝑡𝑐 𝐿

3 NORSAND HARDENING RULE IN THE

FRAMEWORK OF CRITICAL STATE THEORY
(CST)

The classic CST framework posits that critical state can be

achieved at large strains only when two conditions are met:
(i) plastic volumetric strain rate (𝑑𝜀𝑣 𝑝 ) or dilatancy (𝐷)
should be zero, and (ii) there should be continuous zero
dilatancy under continuous shearing at large strains
(𝑑(𝑑𝜀𝑣 𝑝 ) = 0). In the case of the NorSand model with an
associated flow rule, zero dilatancy occurs when the
current confining pressure (𝑝) reaches the image confining
pressure (𝑝𝑖 ). Therefore, to be at critical state at large
strains, both 𝑝 and 𝑝𝑖 should reach 𝑝𝑐 , the confining Figure 2 Hardening rule used in the NorSand model.
pressure at the critical state. Additionally, 𝑞, the current
4 RANGE OF APPLICABILITY OF THE NORSAND emerges at a particular stress ratio (𝜂𝐹𝐿 ) in zone 2, ranging
MODEL from 𝜂𝐿 to 𝑀𝑖 .

From Equation (12), the image stress ratio is a linear

function of the image state parameter. It is obvious that for
𝑀
|𝜓𝑖 | ≥ , 𝑀𝑖 ≤ 0, and the model loses its applicability. It
𝑁.𝜒𝑖
must be noted that near to zero image stress ratio also
affects the volumetric response of the NorSand model, and
it is recommended that 𝑀𝑖 should not be lower than a
certain value:
𝑀
≤ 𝑀𝑖 [19]
1+𝑁

By using this condition in Equation (12), the range of state

parameter where the NorSand model is applicable can be
obtained:
𝑀 Figure 3 Schematic view of the location where static
|𝜓𝑖 | ≤ [20]
(1+𝑁)𝜒𝑖 liquefaction is initiated with the NorSand model

Beyond this range, the NorSand model reaches critical

state at higher mean confining pressure because of lower Based on the above discussion, the classic NorSand model
image stress ratio. can simulate flow liquefaction. However, the major
limitation is the accuracy of the simulation results in
5 CLASSIC NORSAND MODEL AND UNDRAINED comparison with measured data. By using the stress
LOADING dilatancy rule and applying the limiting dilatancy, the
following expression can be derived:
In this remainder of this article the formulation of the
NorSand model presented in the previous sections is 𝜂𝐿 = 𝑀𝑖 + 𝐷𝐿 = 𝑀𝑖 +
𝑀
𝜒𝑖 𝜓𝑖 [21]
referred to as the classic NorSand model. This version has 𝑀𝑡𝑐
two main limitations in the simulation of granular soils
under undrained loading: 𝜒𝑖 is the material property that is obtained from tests
conducted on dense states (𝜓𝑖 < 0); and then used to
• Inability of simulating flow liquefaction properly (In simulate the mechanical behaviour of loose states. This
order to enhance the simulation results, assumption produces 𝜂𝐿 that is slightly lower than 𝑀𝑖 ≅ 𝑀
adjustments were required either in the hardening for loose states. As explained earlier, flow liquefaction
happens somewhere between 𝜂𝐿 and M (see Figure 3).
rule or the bulk modulus, as noted by Jefferies
meaning that 𝜂𝐹𝐿 takes place at high stress ratio close to
and Been (2015) and Ng et al. (2023))
M. This is not in agreement with the observed loose soil
behaviour, where measured 𝜂𝐹𝐿 is much lower. Because of
• Inability of simulating limited flow feature these reasons, researchers such as Shuttle and Jefferies
(Jefferies, 1993) (2016), Reid and Smith (2021), and Ng et al. (2023) have
attempted to manipulate the bulk modulus to enable
5.1 Classic NorSand Model and Flow Liquefaction NorSand to achieve a better 𝜂𝐹𝐿 that aligns with
experimental data.
Flow liquefaction is a type of cohesionless soil behaviour
that occurs at states initially looser than critical state (𝜓𝑖 > 5.2 Classic NorSand Model and Limited Flow Behavior
0). According to the explanations presented in Section 3,
and visually depicted in Figure 2b, a positive image state The classic NorSand model cannot simulate the limited
parameter yields 𝑝𝑖 < 𝑝𝐿 and 𝑀𝑖 < 𝜂𝐿 . At the test's flow behaviour due to its hardening rule. Equation (18) is
inception, 𝑝𝐿 < 𝑝, hence h > 0, and the yield surface only capable to simulate flow liquefaction as shown in
expands. Given that 𝑝𝑖 < 𝑝, normality rules mandate a Figure 3. To simulate the limited flow behaviour, the
reduction in mean confining pressure during the yield hardening rule should be modified in such a way that
surface expansion, this indicates that p and 𝑝𝑖 converge produces the desired stress-path drawn schematically in
(with 𝑝𝑖 increasing due to hardening and p declining due to Figure 4. As it can be seen the path contains three zones.
normality). Zone 1 in Figure 3 shows this part of the stress Zone 1, presenting an increase in q and a decrease in p,
path. Once 𝑝𝐿 = 𝑝, the yield surface expansion ceases. Zone 2, having both p and q decreasing, and Zone 3, where
Subsequently, when 𝑝 < 𝑝𝐿 , normality, and hardening rules increases in q and p are observed. The new modified
work in tandem, leading to a reduction in 𝑝 and 𝑝𝑖 towards hardening rule must be able to simulate strain hardening
the critical state. This section of the stress path is (zone 1), strain softening (zone 2), and strain hardening
represented by zone 2 in Figure 3. Flow liquefaction (zone 3). However, the hardening rule in the classic
NorSand model only allows for the simulation of zone 1 and
zone 2 (strain hardening followed by strain softening before Therefore, in the new hardening rule, the location of 𝑝 in
reaching critical state) for loose states. comparison with 𝑝𝐿−𝑐𝑎𝑝 , controls the hardening-softening
behaviour of the NorSand yield surface. Figure 5 shows
6 NORSAND MODEL WITH CAP-SOFTENING schematically the effect of cap-softening term on the
FEATURE location of the limiting hardening quantity. It must be
mentioned that, at the beginning of any simulation with
To solve the classic NorSand limitations under undrained 𝜂
initial isotropic condition, 𝑝𝐿−𝑐𝑎𝑝 = 𝑝𝐿 due to the term =
conditions, developers of the NorSand model modified the 𝜂𝐿

hardening rule by adding a new term called the cap- 0. At critical state, 𝑝𝐿−𝑐𝑎𝑝 → 𝑝𝐿 , since 𝐷 → 0.
softening term (Jefferies and Been, 2015). This additional
term is only used for the undrained response and is not 6.1 NorSand Model with Cap-Softening and Flow
considered for the drained response. Although, this kind of Liquefaction
modification improves the model’s response, it implies that
the model is unable to reach a unique formulation for both As shown schematically in Figure 5b, for loose states, the
drained and undrained loading conditions. The modified presence of cap-softening results in shifting of the limiting
hardening rule for undrained loading condition is written as hardness quantity from 𝑝𝐿 to 𝑝𝐿−𝑐𝑎𝑝 . Due to the shape of
follows: the yield surface, 𝜂𝐿−𝑐𝑎𝑝 < 𝜂𝐿 . Based on the new hardening
rule, flow liquefaction (𝜂𝐹𝐿 ) must occur between 𝜂𝐿−𝑐𝑎𝑝 and
𝑀 instead of 𝜂𝐿 and 𝑀. As a result, the new hardening rule
𝑝
𝑖
𝑑𝑝𝑖 𝑝𝑖 𝜂 𝑑(𝑝𝑝𝐿)
ℎ𝑢𝑛𝑑𝑟𝑎𝑖𝑛𝑒𝑑 = = 𝐻 (𝑝 − 𝑝𝑖 ) + 𝑆 [22] reduces the simulated 𝜂𝐹𝐿 , and improves the agreement of
𝑑𝜀𝑞 𝑝 𝑝𝐿 𝜂𝐿 𝑑𝜀𝑞 𝑝
the simulations with real data.

Figure 4 Ideal simulation of limited flow feature through the

concept of hardening rule and normality rule in Cambridge
type models

Where S is a switch, 0 for drained loading and 1 for

undrained loading. By expanding the cap-softening term in
Equation (22), the undrained hardening rule becomes:
𝑝𝑖 𝜂 𝑝
𝑑𝑝𝑖 𝐻(𝑝 −𝑝𝑖 )−S ( 𝑖 )𝐾𝐷
[23]
𝑝𝐿 𝜂𝐿 𝑝 𝐿
ℎ𝑢𝑛𝑑𝑟𝑎𝑖𝑛𝑒𝑑 = = 𝑝
𝑑𝜀𝑞 𝑝 𝜂
𝑑(𝑝 𝑖 )
𝐿
1+p
𝜂𝐿 𝑑𝜓𝑖

Where 𝐾 is the elastic bulk modulus. In the new hardening

rule, 𝑝𝐿 and everything associated with it, including 𝐷𝐿 and
𝜂𝐿 , lose their meaning as a limiting quantity. This fact can
be proven by Equation (23), where at 𝑝 = 𝑝𝐿 , it is not
necessary for ℎ𝑢𝑛𝑑𝑟𝑎𝑖𝑛𝑒𝑑 to be zero. In other words, the
cap-softening term shifts the point where the hardening
regime reverses the expansion of the yield surface to
shrinkage. The new point (𝑝 = 𝑝𝐿−𝑐𝑎𝑝 ) can be obtained by
finding the root of Equation (23).
𝜂 𝐾
𝑝𝐿−𝑐𝑎𝑝 = 𝑝𝐿 + 𝑆 𝐷 [24]
𝜂𝐿 𝐻
Figure 5 Effect of cap-softening term on the location of
limiting dilatancy by shifting it to a new location
6.2 NorSand Model with Cap-Softening and Limited The material properties and initial state condition (𝑝0 , 𝑒0 )
Flow Feature used in the simulation for Figure 6 are shown in Table 2.
In the figure, 𝑆 = 0 corresponds to the classic NorSand
As explained in Section 5.2, the limited flow behavior model where the cap-softening has not been applied, 𝑆 =
requires a strain hardening-softening-hardening feature. 1 is with the cap-softening term suggested by the
This fact means that for NorSand to model limited flow developers of the NorSand model. An example of
behavior, it should experience at least two changes of simulation with 𝑆 = 1.5 is presented to illustrate the rigidity
hardening regime before reaching critical state. The of the formulation, although the value is outside of the
undrained hardening regime change takes place when 𝑝 = recommended range. As it can be seen, by increasing S
𝑝𝐿−𝑐𝑎𝑝 and as a result 𝑑𝑝𝑖 = 0. This condition is reached outside of the recommended value, the model starts
several times at 𝑝 = 𝑝𝑖 = 𝑝𝐿−𝑐𝑎𝑝 ≠ 𝑝𝑐 or at 𝜓 = 𝜓𝑖 = bouncing around the critical state, gets instable and unable
𝜓𝐿−𝑐𝑎𝑝 ≠ 0. However, at large strains, the NorSand to simulate the soil behaviour properly, not giving the
reaches the critical state when 𝑑𝑝𝑖 and 𝜓𝑖 both reach zero. necessary flexibility to adjust the behavior of the model.

The developers of the NorSand model restricted the value Table 2 Material property of Fraser River Sand used to
of S in Equation (24) to be 0 or 1, which limits the number simulate the undrained behaviour of a loose sample with
of times where undrained hardening regime change would the NorSand model.
occur. Figure 6, presents examples of simulations using parameters Value
the NorSand model, where the value of S varies. 𝑀 1.47
Γ 1.01
𝜆 0.087
𝜉 0.38
𝜒𝑡𝑐 5
𝑁 0.4
𝐻0 120
𝐻𝑦 400
𝑝0 700e3 Pa
𝑒0 0.9

6.3 Outcomes of the Cap-Softening Feature

It appears that the NorSand model, in its current

formulation may lack the flexibility to effectively reproduce
a wide range of undrained response in loose states,
instead, it replicates an undrained behavior pattern. To
study the capability of their model to match the behavior of
different soils, the developers of the NorSand model have
compared its response to soil data using a wide range of
material properties. For instance, in Figure 7, the
normalized undrained peak shear strength of loose
samples (𝑠𝑢 ⁄𝑝0) is illustrated alongside their initial state
parameters (𝜓0 ) for various soils. The results obtained from
the NorSand model for different values of (𝐻 ⁄𝐼𝑟 ) are also
displayed in the figure. Here, 𝐼𝑟 denotes the shear rigidity
and can be derived using the following equation:
𝐺
𝐼𝑟 = [25]
𝑝

Where G is the shear modulus of the soil. The graph

reveals that for initially loose samples having high initial
state parameters, 𝜓0 , the model's efficacy in accurately
predicting undrained peak responses weakens.
Additionally, the model may require different (𝐻 ⁄𝐼𝑟 ) values
to match the behavior of a specific soil under different
loading conditions and initial state parameter. This means
Figure 6 Effect of the cap-softening term on the undrained that manipulation on the elastic material properties may be
behaviour of loose states a) deviatoric stress-axial strain, required to get more accurate results. The H value selected
b) deviatoric stress-confining pressure
must be applicable not only for the undrained condition but undrained behavior of granular materials, has several
also for the drained response. limitations to capture some of the specific features of the
behavior of granular materials. First, it can only simulate
soil response within a particular range of initial state
parameters. Secondly, to capture the flow behavior of
loose material, the model employs a cap-softening term to
modify its undrained hardening rule that does not provide
flexibility in its response. This term is chosen to satisfy the
NorSand approach to the critical state and avoid adding
additional parameters to the model. By altering the limiting
hardening point, the cap-softening term enhances
NorSand's ability to simulate flow liquefaction and limited
flow behavior in the undrained response. However, the
value of the cap-softening term is either 1 or 0, which limits
the capacity of a user to calibrate the model to match a
target behavior.

Figure 7 Effect of (𝐻 ⁄𝐼𝑟 ) on the normalized

undrained peak shear strength simulated through the
NorSand model in comparison with measured data
(Modified after Jefferies and Been, 2015)

7 IMPROVING THE NORSAND MODEL

By going back to Sections 2 and 5, it is evident that the

underlying issue comes from the extrapolation of the state-
dilatancy law from dense states to loose states. To address
this problem and enhance the model's performance, one
possible solution is to introduce a second state-dilatancy
law with a distinct formulation for loose states. The simplest
form can be linear like Equation (4) with a new material
property (𝜒𝑡𝑐−𝑙𝑜𝑜𝑠𝑒 ).
𝑀
𝐷𝐿 = 𝜒𝑡𝑐−𝑙𝑜𝑜𝑠𝑒 𝜓𝐿 [26]
𝑀𝑡𝑐

𝜒𝑡𝑐−𝑙𝑜𝑜𝑠𝑒 must be equal or greater than 𝜒𝑡𝑐 . However,

according to Section 4, higher 𝜒𝑡𝑐−𝑙𝑜𝑜𝑠𝑒 further limits the
range of state parameters where the NorSand model is
applicable. Choosing another form of state-dilatancy law
for loose states adds complexity to the model and may Figure 8 Influence of undrained dilatancy-state
require further changes to the model’s formulation. parameter used in a new state-dilatancy law for loose state,
Nevertheless, by incorporating an appropriate state- on the flow behaviour of loose states, a) stress-strain
dilatancy law and its associated material parameters and behaviour, b) stress path
formulations, the transition point from hardening to
softening under undrained loading can be more accurately 9 REFERENCES
controlled. The authors have tried to introduce new
formulations of state-dilatancy law with a new parameter to Been, K. and Jefferies, M.G., 1985. A state parameter for
simulate static liquefaction more precisely. Figure 8 sands. Géotechnique, 35(2), pp.99-112.
illustrates the impact of this new parameter (called
undrained dilatancy-state parameter) on controlling static Cheng, Z. and Jefferies, M., 2020, February.
liquefaction without modifying any other material properties Implementation and Verification of NorSand Model in
in the model. A comprehensive presentation of the General 3D Framework. In Geo-Congress 2020:
modified model, named PolySand, will be provided in a Geotechnical Earthquake Engineering and Special
future publication (Razavi, 2023). Topics (pp. 10-19). Reston, VA: American Society of Civil
Engineers.
8 CONCLUSION
Dafalias, Y.F. and Manzari, M.T., 2004. Simple plasticity
According to the above discussion, the NorSand sand model accounting for fabric change effects. Journal
model, while widely used for simulating the drained and of Engineering mechanics, 130(6), pp.622-634.
Jefferies, M. and Been, K., 2015. Soil liquefaction: a critical
state approach. CRC press.

Jefferies, M.G., 1993. Nor-Sand: a simple critical state

model for sand. Géotechnique, 43(1), pp.91-103.

Li, X.S., Dafalias, Y.F. and Wang, Z.L., 1999. State-

dependant dilatancy in critical-state constitutive modelling
of sand. Canadian Geotechnical Journal, 36(4), pp.599-
611.

Lu, X. and Huang, M., 2015. Static liquefaction of sands

under isotropically and K0-consolidated undrained triaxial
conditions. Journal of Geotechnical and Geoenvironmental
Engineering, 141(1), p.04014087.

Ng, C.W., Crous, P.A. and Jacobsz, S.W., 2023. Centrifuge

and Numerical Modeling of Liquefied Flow and
Nonliquefied Slide Failures of Tailings Dams. Journal of
Geotechnical and Geoenvironmental Engineering, 149(9),
p.04023075.

Razavi, S.K., 2023. Constitutive modeling of static

liquefaction of tailings (Doctoral dissertation, in
Preparation), Polytechnique Montreal, Montreal, QC.

Reid, D. and Smith, K., 2021. Interpretation of state

parameter in partially drained tailings: a case history
examination. Géotechnique Letters, 11(4), pp.276-280.

Roscoe, K.H., Schofield, A. and Thurairajah, A., 1963.

Yielding of clays in states wetter than
critical. Geotechnique, 13(3), pp.211-240.

Shuttle, D. and Jefferies, M., 2016. Determining silt state

from CPTu. Geotechnical Research, 3(3), pp.90-118.
Preliminary evaluation of Speswhite kaolin
as a physical analogue material for
unsaturated oil sands tailings
Tony Zheng & Nicholas Beier
University of Alberta, Edmonton, Alberta, Canada

ABSTRACT
Thin-lift deposition of treated oil sands tailings is currently investigated as a potential solution for creating a trafficable cap layer
for future reclamation activities. To determine the optimal thickness of the cap layer, physical modeling experiments using the
50g-ton geotechnical beam centrifuge at University of Alberta were proposed. Prior to the deposition of fresh layers, the cap
layer is often unsaturated due to freeze-thaw and atmospheric drying. In this paper, Speswhite kaolin is investigated as the
candidate for a physical analogue of unsaturated oil sands tailings. Due to its high brightness and repeatability, Speswhite
kaolin has been widely used as a modeling clay in geotechnical centrifuge experiments. However, its suitability of simulating
unsaturated tailings with high clay content was rarely considered in the past. Two different sample preparation methods using
Tempe cells and air-drying to create blocks of unsaturated Speswhite kaolin are evaluated. Undrained shear strength was
measured by the miniature vane shear device and unconfined compression test while drained shear strength was determined
by the direct shear device. Test results showed that the two materials behaved similarly in drained conditions while Speswhite
kaolin exhibited a higher shear sensitivity in undrained conditions. Test results also showed different drying behaviour between
the two materials: Speswhite kaolin dries faster, more uniformly and to a higher void ratio than treated oil sands tailings.
Potential causes behind these observations are discussed in the paper.

RÉSUMÉ

Le dépôt en couches minces de résidus de sables bitumineux traités est actuellement étudié comme solution potentielle pour
créer une couche de recouvrement praticable pour les futures activités de remise en état. Pour déterminer l'épaisseur optimale
de la couche de couverture, des expériences de modélisation physique utilisant la centrifugeuse à faisceau géotechnique de
50 g de l'Université de l'Alberta ont été proposées. Avant le dépôt des couches fraîches, la couche de couverture est souvent
insaturée en raison du gel-dégel et du séchage atmosphérique. Dans cet article, le kaolin Speswhite est étudié comme candidat
pour un analogue physique des résidus de sables bitumineux insaturés. En raison de sa grande luminosité et de sa répétabilité,
le kaolin Speswhite a été largement utilisé comme argile à modeler dans les expériences de centrifugation géotechnique.
Cependant, son aptitude à simuler des résidus insaturés à forte teneur en argile a rarement été envisagée dans le passé. Deux
méthodes différentes de préparation d'échantillons utilisant des cellules Tempe et le séchage à l'air pour créer des blocs de
kaolin Speswhite insaturé sont évaluées. La résistance au cisaillement non drainé a été mesurée par le dispositif de
cisaillement miniature à palettes et le test de compression non confiné tandis que la résistance au cisaillement drainé a été
déterminée par le dispositif de cisaillement direct. Les résultats des tests ont montré que les deux matériaux se comportaient
de manière similaire dans des conditions drainées, tandis que le kaolin Speswhite présentait une sensibilité au cisaillement
plus élevée dans des conditions non drainées. Les résultats des tests ont également montré un comportement de séchage
différent entre les deux matériaux : le kaolin Speswhite sèche plus rapidement, plus uniformément et à un taux de vide plus
élevé que les résidus de sables bitumineux traités. Les causes potentielles derrière ces observations sont discutées dans le
document.

1 INTRODUCTION (McKenna et al 2016). Desiccation during summer and

freezing/thawing during winter and spring are known
In Alberta’s oil sands mining operation, there are currently methods to further strengthen treated tailings (Kolstad et al
220km2 of tailings that are mandated by Directive 085 to be 2016). Figure 1 shows a field trial where the mine operator
reclaimed to a “self-sustaining, maintenance-free” landscape placed treated tailings in thin layers over a large area allowing
(Alberta Energy Regulator 2021). Meeting the requirement of the current layer to sufficiently dry or freeze/thaw before
Directive 085 has eluded experts for over a decade. One of placing the next layer (Song et al 2011, COSIA 2022).
the reasons why it has remained elusive is because of the Despite the inevitable rewetting, the thin-lift deposit is still
geological variability in the clay minerals of the McMurray stronger than tailings placed continuously in thick layers
formation in Alberta, which affects the composition of tailings. (Daliri et al 2016).
Additionally, operators add various chemicals to strengthen The overarching objective of this research program is
the highly variable oil sands tailings, creating unique using scaled physical models to investigate the optimal
properties in each tailings deposit. thickness of the stiff layer created by thin-layer placement. In
After mine operators add polymers (known as flocculants) physical modelling experiments, this stiff layer can be
and in some cases coagulants to the “fluid-like” oil sands idealized as a single block of soil (see Figure 2). Using
tailings, the resultant product has a “yogurt-like” consistency tailings samples collected in the field, while valuable, can
make interpretation difficult due to inherent geological (Cv) is derived from oedometer tests on desiccated rewetted
variability, sample disturbance and logistical challenges. centrifuge cake.
Synthetic soils are often used in physical modelling Speswhite Kaolin is a naturally occurring clay deposit
experiments. This paper evaluates Speswhite Kaolin as a quarried near Cornwall in Southwest England. Due to the
candidate for simulating unsaturated tailings. Previous works extensive use in physical modelling experiments, its
on unsaturated kaolin are concentrated on statically geotechnical properties are well documented in the literature.
compacted material (Sivakumar 1993, Tarantoni and Index properties from the supplier (Imerys 2008), Jia (2021)
Tombolato 2005, Tarantino and De Col 2008). This paper and unsaturated Cv measured by Sivakumar (1993) are
aims to characterize Speswhite Kaolin subjected to listed in Table 1
desiccation, mimicking the stress path in the field.
Table 1. Material properties of centrifuge cake and Speswhite
Kaolin

Material Properties Centrifuge Cake Speswhite Kaolin

Liquid Limit (%) 57 65
Plastic Limit (%) 26 35
Shrinkage Limit (%) 21 32
Specific Gravity (Gs) 2.48 2.61
Clay Fraction (%) 52 76 to 83
MBI (meq/100g) 4.80 1.79*
Specific Surface Area (m2/g) 20 14
Cv (m2/yr) 0.20 2.27
Figure 1. Field trials of thin-lift deposition of flocculated oil *Back-calculated from specific surface area using equations
sands tailings with instrumentation installed in the docks from Hang and Brindley 1970
(Matthews et al 2011).

3 EXPERIMENTAL METHODS

3.1 Sample Preparation

Centrifuge cake is collected shortly after deposition while

Speswhite Kaolin is shipped in bagged powder form. The
initial water content for the centrifuge cake samples ranges
from 52% to 70% while the Speswhite Kaolin samples are
pre-mixed with de-ionized water at 1.5 times the liquid limit.
Prior to desaturation, both materials are homogenized by a
hand-held electrical mixer for a minimum of two minutes.
Unsaturated samples are prepared by ambient air-drying
Figure 2. Idealization of field conditions in a geotechnical and dewatering inside Tempe cells. Ambient air-drying is
centrifuge performed under controlled laboratory environment (i.e. room
temperature at 20 Celcius). Air-dried samples, denoted by
“AD”, are slowly desiccated inside a Shelby ring with a
2 MATERIAL PROPERTIES diameter of 63.5mm and height of 32mm for the shrinkage
tests and inside a standard 20L pail for the vane shear test.
Unsaturated centrifuged tailings (known as centrifuge cake) Tempe-dried samples, denoted by “TD”, are desaturated
and Speswhite Kaolin are compared in this paper. Centrifuge by applying a positive air pressure to the sample sitting on
cake is currently produced commercially at several oil sands top of a high air-entry ceramic porous stone (1-bar and 5-bar)
mining operations. The centrifugation process involves which allows the pore water to drain through the water outlet
dilution and flocculation of FFT with a flocculating agent while keeping the air in the cell (see Figure 3). Tempe cells
(typically high molecular weight anionic polyacrylamide) and use the axis-translation technique to achieve target matric
coagulant (typically gypsum or lime), followed by dewatering suction in the soil sample (Fredlund et al. 2012). The tempe
in a bowl centrifuge press prior to deposition. Flocculant cell is weighed daily to calculate how much pore water is
injection often occurs close to the bowl centrifuge feed tube expelled. Equilibrium is reached when the weight difference
to avoid over-shearing. The end-product typically has a solids after 24 hours is less than 0.5g, and then the specimen can
content of more than 50%. (Spence et al. 2017). be extracted for further testing. Air pressures of 25, 50,100,
Geotechnical index properties of centrifuge cake in Table 1 200 and 400kPa are applied to achieve matric suctions in a
are based on previous work from Rima and Beier (2022) and single load step, using the technique developed by Kabwe et
Schafer and Beier (2020). The coefficient of consolidation al. 2023. Multiple Tempe cells are used to speed up the
sample preparation process. Moisture distribution of TD 3.3 Undrained Test (Unconfined Compression, Vane
samples at the end of the test is highly uniform without any Shear)
discernible cracking (see Figure 4). The walls of the Shelby
rings and Tempe cells are well-greased in petroleum jelly to Unconfined compression strength (UCS) tests are conducted
minimize side friction. according to the ASTM (2016) standard at a constant
displacement rate of 1%/min. Specimen for the UCS test are
trimmed from TD samples by a Shelby tube with a diameter
of 28mm. Due to volume shrinkage and insufficient height of
the dried sample inside the Tempe cell, a length-to-diameter
ratio of 1 is adopted instead. While this is a deviation from the
length-to-diameter ratio of 2 as per the ASTM standard, it is
considered acceptable here since the objective is to compare
the load-displacement responses under identical boundary
conditions.
For vane shear tests, a vane blade with 12.7mm height
and 12.7mm width is attached to a motorized vane shear
device which rotates at a speed of 60 degrees per minute.
The vane shear test is conducted according to ASTM D4648.
For remolded strength, the vane blade is rapidly rotated 5
times before strength is re-evaluated. This remolding
procedure is repeated until there are no further decreases in
the remolded strength.

3.4 Drying Test (Air-Drying and Tempe Cell)

Slurried samples are placed in a series of Shelby rings on top

Figure 3. Tempe cell setup for preparing Tempe-dried of a petri dish (see Figure 5). The bottom of the glass dish is
samples (Kabwe et al 2023) covered by well-greased wax paper to avoid leakages from
the side. Dimensions of the AD samples are measured by a
caliper at different stages of air-drying to calculate the void
ratio and the shrinkage curve. Gravimetric water contents
from the top and bottom layer are taken at different stages of
drying.
Soil water retention properties are measured using the
procedure from Fredlund and Houston (2009). The Tempe
cell procedure is similar to Section 3.1 except that air
pressures are step loaded instead. Samples are subjected to
increasing air pressure in the Tempe cells until reaching a
maximum matric suction of 400kPa.

Figure 4. Centrifuge cake (left) and Speswhite Kaolin (right)

at the end of the Tempe cell test under an applied matric
suction of 400kPa

3.2 Drained Test (Direct Shear)

All samples in the direct shear box have been Tempe-dried

under an applied matric suction of 400kPa. At equilibrium, the
sample is removed from the Tempe cell and trimmed by a
Shelby ring. The trimmed sample is then transferred into the
direct shear box.
Once the direct shear box is attached to the direct shear Figure 5. Air-dried sample preparation setup
machine, de-ionized water[RRR1][TZ2] starts rewetting the sample and a
vertical load of 15kPa is immediately applied. Subsequent
load steps are applied at twice the magnitude of the previous
load step until the target normal stress is reached. The
sample is sheared at a rate of 0.008 mm/min.
4 RESULTS AND DISCUSSION

4.1 Drained Behaviour (Direct Shear)

Based on best-fit lines of the data points in Figure 6, the

frictional angle in direct shear is similar between Tempe-dried
centrifuge cake (18 degrees) and Tempe-dried Speswhite
Kaolin (19.9 degrees). The similarity in frictional angle can be
explained, in part, by the identical stress path the two
materials experienced which produced similar fabric prior to
shearing.
The over-consolidation ratio by desiccation (OCR-d) can
quantify stress history during the desaturation process in
Tempe cells. OCR-d is defined by the ratio of the maximum
suction stress during desiccation and the maximum effective
consolidation stress achieved during the consolidation stage
of the direct shear tests. The maximum suction stress is
calculated as the applied matric suction multiplied by the
degree of saturation based on Lu and Likos (2006) and Akin
and Likos (2020). The degree of saturation is derived from
the shrinkage curves and the Tempe cell tests.
In Figure 7, the horizontal strain at which peak shear
occurs converges at lower OCR-d. At higher OCD-d,
Figure 7. Variation of mobilized horizontal strain
Speswhite Kaolin exhibited a stiffer response. Figure 8 is an
[RRR3][TZ4]during direct shear tests at peak shear stresses under
alternative way to look at the stress-strain behaviour. The
different desiccation over-consolidation ratio (OCR-d)
peak shear stresses of both materials were mobilized at a
greater horizontal strain with increasing normal stresses. The
absence of flocculant/coagulant treatment and centrifugation
process in Speswhite Kaolin does not seem to create a
stress-strain behavior different from the centrifuge cake
provided that both materials undergo the same desiccation
stress path.

Figure 8. Comparison of mobilized horizontal strain at peak

shear stresses in Tempe-dried samples under direct shear

4.2 Undrained Behaviour

As shown in Figure 9, both materials showed similar trends

Figure 6. Comparison of peak shear stresses from direct in undrained shear strength regardless of the sample
shear testing of Tempe-dried (TD) samples preparation method. Figure 10 shows that Speswhite Kaolin
is three to five times more sensitive than centrifuge cake as
liquidity index decreases toward the plastic limit. Unlike the
Speswhite Kaolin, shear sensitivity in centrifuge cake
increases only slightly near the plastic limit, possibly due to
aggregation and bonding from the flocculation, coagulation
and centrifugation treatment during the slurry forming stage.
The load-displacement response curves under
unconfined compression in Figure 11 showed similar stress-
strain behaviours in Tempe-dried samples at 400kPa applied
matric suction near the plastic limit. At 100kPa applied matric
suction, centrifuge cake showed a higher shear strength at a
larger strain. The final water content of the failed specimen
decreases less than 1% compared to the initial water content
of the TD samples, indicating minimum moisture loss during
UCS.

Figure 10. Comparison of shear sensitivity measured by vane

shear.

Figure 9. Comparison of peak undrained shear strength

measured by vane shear.

Figure 11. Comparison of load vs displacement curve under

unconfined compression loading of Tempe-dried samples.

4.3 Drying Behaviour

As shown in Figure 14, moisture distribution in air-dried

centrifuge cake is non-uniform with the greatest difference
observed near the liquid limit. As air-drying progresses
toward the plastic limit, differences between the top and
bottom water content are reduced. This non-uniformity is
confirmed by visual observations after the Shelby ring is
removed during the air-drying process. The top stiffer section
of the centrifuge cake remains intact while the bottom softer
section slumps laterally (see Figure 12).
On the other hand, moisture distribution in Speswhite
Kaolin is highly uniform: most data points lie on the 1:1 line in
Figure 14 confirmed by visual observations throughout the
air-drying process (see Figure 13).
Single-step and step-loaded Tempe cell tests produced
similar liquidity indices at a given matric suction (see Figure
15). Liquidity indices of Tempe-dried Speswhite Kaolin is
consistent with Tripathy et al (2014) using a similar method.
At a given matric suction, centrifuge cake dewaters to a lower
liquidity index and a lower void ratio. (Figure 16).
The larger coefficient of consolidation (Cv) in Speswhite
Kaolin contributes to the faster and more uniform drying
Figure 13. Profile view of air-dried Speswhite kaolin with
phenomenon compared to centrifuge cake. Tempe-dried
uniform distribution of water content over depth.
centrifuge cake takes approximately two weeks to reach
equilibrium while Tempe-dried Speswhite Kaolin takes only
one day. It is hypothesized that microstructures and clay-
water interaction mechanisms due to flocculation and
coagulation, clay mineralogy, and pore water chemistry may
have contributed to the different drying behaviours. For
Speswhite Kaolin, moisture re-distribution in the bottom
layers occurs quickly as water in the top layer evaporates. In
contrast, pore water in the centrifuge cake contains elevated
salt concentrations at a basic pH (>8) because of the
geological environment and the caustic extraction process.
The precipitation of salts at the surface suppress evaporation
due to increased osmotic suction and reduced hydraulic
conductivity (Simms et al 2019). The presence of illite inter-
mixed with smectite in centrifuge cake which has a higher
specific surface area than pure Kaolinite mineral may also
contribute to the slower drying.
Understanding the differences in drying behaviour can
help optimize thin-lift placement. Further studies in the
microstructure and physio-chemical forces behind the drying
behaviour are needed. However, this is outside the scope of
this paper.

Figure 14. Comparison of top and bottom gravimetric water

content during various stages of air-drying

Figure 12. Profile view of centrifuge cake during ambient air-

drying with un-even distribution of water content over depth.
 5. CONCLUSION

A suite of geotechnical tests is conducted on unsaturated

Speswhite Kaolin and centrifuge cake. The two materials
behave similarly in drained conditions and at lower water
content in undrained conditions except that Speswhite Kaolin
is more sensitive in the vane shearing mode. The two
materials behave differently during air-drying: Speswhite
Kaolin dries more quickly and uniformly with a higher void
ratio at the shrinkage limit. The suitability of Speswhite Kaolin
as a physical analogue of unsaturated centrifuge cake
depends on the purpose of the experiment. Speswhite Kaolin
is a suitable candidate if the objective of the physical
modelling experiment is to assess bearing capacities and
long-term stability of the closure landform. If the objective is
to investigate the impact of thin-lift deposition and climate, an
alternative recipe needs to be developed for a better match
of the drying and rewetting behaviour.

6. ACKNOWLEDGEMENT

The authors would like to thank the Natural Sciences and

Engineering Research Council of Canada Industrial
Research Chair (NSERC IRC) program in Oil Sands Tailings
Figure 15. Comparison of liquidity indices achieved by the
Geotechnique and Canada’s Oil Sands Innovation Alliance
step-loaded and one-step Tempe cell method.
(COSIA) for their funding support. The authors would also like
to acknowledge Dr Louis Kabwe for his guidance in Tempe
cell tests.

REFERENCES

Akin, I.D., Likos, W.J., 2020. Suction Stress of Clay Over a

Wide Range of Saturation. Geotechnical and
Geological Engineering 38, 283–296.
https://doi.org/10.1007/s10706-019-01016-7
Alberta Energy Regulator, 2021. State of Fluid Tailings
Management for Mineable Oil Sands, 2020. AER,
Calgary, Alberta, Canada.
ASTM, 2016a. ASTM D2116 Standard Test Method for
Unconfined Compressive Strength of Cohesive Soil
(Standard No. D2116). American Society for Testing
and Materials, West Conshohocken, PA.
ASTM, 2016b. ASTM D4648/D4648M-16 Standard Test
Methods for Laboratory Miniature Vane Shear Test for
Saturated Fine-Grained Clayey Soil. American Society
of Testing Material, West Conshohocken, PA.
COSIA, 2022. Deep Deposit Design Guide for Oil Sands
Tailings. Canadian Oil Sands Innovation Alliance,
Calgary, Alberta, Canada.
Figure 16. Comparison of shrinkage curves between Daliri, F., Simms, P., Sivathayalan, S., 2016. Shear and
centrifuge cake and Speswhite Kaolin dewatering behaviour of densified gold tailings in a
laboratory simulation of multi-layer deposition.
Canadian Geotechnical Journal. 53, 1246–1257.
https://doi.org/10.1139/cgj-2014-0411
Fredlund, D.G., Houston, S.L., 2009. Protocol for the
assessment of unsaturated soil properties in
geotechnical engineering practice. Canadian
Geotechnical Journal 46, 694–707.
https://doi.org/10.1139/T09-010
Fredlund, D.G., Rahardjo, H., Fredlund, M.D., 2012. Beersing, A. (Eds.), . Presented at the Mine Closure
Unsaturated Soil Mechanics in Engineering Practice. 2011: Sixth International Conference on Mine Closure,
John Wiley & Sons, Inc. Australian Centre for Geomechanics, Perth, Australia,
Hang, P.T., Brindley, G.W., 1970. Methylene Blue Absorption pp. 373–382.
by Clay Minerals. Determination of Surface Areas and Spence, J., Bara, B., Lorentz, J., Mikula, R.J., Lee, J., Lahaie,
Cation Exchange Capacities (Clay-Organic Studies R.D., 2017. Centrifuge process for dewatering oil
XVIII). Clays and Clay Minerals 18, 203–212. sands tailings. US9656189B2.
https://doi.org/10.1346/CCMN.1970.0180404 Tarantino, A., De Col, E., 2008. Compaction behaviour of
Imerys, 2008. Speswhite Kaolin TDS. clay. Géotechnique 58, 199–213.
Jia, S., 2021. Beam Centrifuge Modelling of Caprock Integrity https://doi.org/10.1680/geot.2008.58.3.199
(PhD Thesis). University of Alberta, Edmonton, Alberta, Tarantino, A., Tombolato, S., 2005. Coupling of hydraulic and
Canada. mechanical behaviour in unsaturated compacted clay.
Kabwe, L.K., Wilson, G.W., Beier, N.A., Barsi, D., 2023. Géotechnique 55, 307–317.
Application of Tempe Cell to Measure Soil Water https://doi.org/10.1680/geot.2005.55.4.307
Characteristic Curve along with Geotechnical Tripathy, S., Tadza, M.Y.M., Thomas, H.R., 2014. Soil-water
Properties of Oil Sands Tailings. Geosciences 13. characteristic curves of clays. Canadian Geotechnical
https://doi.org/10.3390/geosciences13020036 Journal 51, 869–883. https://doi.org/10.1139/cgj-2013-
Kolstad, D., Borree, B., Song, J., Mahood, R., 2016. Field 0089
pilot performance results for flocculated fluid fine
tailings under three depositional variations. Presented
at the International Oil Sands Tailings Conference,
University of Alberta Geotechnical Centre, Lake
Louise, Alberta, Canada, pp. 51–59.
Lu, N., Likos, W.J., 2006. Suction Stress Characteristic Curve
for Unsaturated Soil. Journal of Geotechnical and
Geoenvironmental Engineering 132, 131–142.
https://doi.org/10.1061/(ASCE)1090-
0241(2006)132:2(131)
Matthews, J., Dhadli, N., House, P., Simms, P., 2011. Field
trials of thin-lift deposition of amended mature fine
tailings at the Muskeg River Mine in Northern Alberta,
in: Jewell, R., Fourie, A. (Eds.), Paste 2011:
Proceedings of the 14th International Seminar on
Paste and Thickened Tailings. Australian Centre for
Geomechanics, pp. 271–280.
https://doi.org/10.36487/ACG_rep/1104_25_Matthews
McKenna, G., Mann, V., Fisseha, B., Beier, N., Olmedo, N.,
2016. The Geotechnical Vane Shear Strength of Soft
Tailings Compared to Soft Foods. Presented at the
International Oil Sands Tailings Conference, University
of Alberta Geotechnical Centre, Lake Louise, Alberta,
Canada, pp. 329–339.
Rima, U.S., Beier, N.A., 2022. Effects of seasonal weathering
on dewatering and strength of an oil sands tailings
deposit. Canadian Geotechnical Journal 59, 447–457.
https://doi.org/10.1139/cgj-2020-0533
Schafer, H., Beier, N., 2020. Estimating soil-water
characteristic curve from soil-freezing characteristic
curve for mine waste tailings using time domain
reflectometry. Canadian Geotechnical Journal 57, 73–
84. https://doi.org/10.1139/cgj-2018-0145
Simms, P., Soleimani, S., Mizani, S., Daliri, F., Dunmola, A.,
Rozina, E., Innocent-Bernard, T., 2019. Cracking,
salinity and evaporation in mesoscale experiments on
three types of tailings. Environmental Geotechnics 6,
3–17. https://doi.org/10.1680/jenge.16.00026
Sivakumar, V., 1993. A Critical State Framework for
Unsaturated Soil (PhD Thesis). University of Sheffield.
Song, J., O’Kane, M., Dhadli, N., Matthews, J., 2011.
Deposition thickness and evaporative drying for oil
sands tailings in northern Alberta, in: Fourie, A.,
Tibbett, M., Beersing, A., Fourie, A., Tibbett, M.,
Monday, October 2, 2023

COVER SYSTEMS
Assessment of the effectiveness of a low
permeability layer made of sand amended
with superabsorbent polymers as a barrier
to oxygen and water
Fatoumata B. K. Diallo, Tikou Belem, Mamert Mbonimpa, &
Abdelkabir Maqsoud
Research Institute of Mines and Environment (RIME) – Université
du Québec en Abitibi-Témiscamingue (UQAT), Rouyn-Noranda,
Quebec, Canada

ABSTRACT
This study aims to evaluate the feasibility and effectiveness of a low saturated hydraulic conductivity (LSHC) layer
composed of natural sand modified with sodium polyacrylate (Na-PA) superabsorbent polymers (SAPs) as an oxygen
and/or water barrier in fine tailings impoundment cover systems. A specific type of fine-grained SAP, SAG-M85F, was
chosen to amend the sand layer. Saturated hydraulic conductivity (ksat) tests were conducted using varying dosages of
SAP in the sand (0.5%, 0.6%, 0.7%, 0.8%, 0.9%, and 1.0%), along with infiltration tests using instrumented columns. The
permeability tests revealed that the k sat of the LSHC layer (sand-SAP mixture) decreased with increasing proportions of
SAP, ranging between 10-7 and 10-8 m/s, compared to the ksat measured for the sand alone (control) at 10 -5 m/s. This
represents a reduction of 2 to 3 orders of magnitude. Furthermore, the instrumented column tests demonstrated that,
depending on the SAP dosage, the LSHC layer (sand-SAP mixture) became impermeable to water after the SAP
transformed into a gel during rinsing. In addition, The LSHC layer, composed of the fine sand-SAP mixture, effectively acts
as a moisture retention layer, achieving the targeted objective of limiting the oxygen flux to less than 1 mol/m²/year. These
findings highlight the potential of utilizing a sand-SAP mixture as an LSHC layer in tailings cover systems, offering an
effective barrier against oxygen and water infiltration.

RÉSUMÉ
Afin de limiter les risques liés au drainage minier acide (DMA) et à son impact à long terme sur l’environnement, les rejets
miniers sulfureux doivent être gérés de manière adéquate. Parmi les techniques existantes figurent les systèmes de
recouvrement jouant le rôle de barrières à l’oxygène ou à l’eau pour limiter la production de DMA. Pour ce faire, des
matériaux granulaires géologiques proches des sites, des géocomposites ou des géomembranes sont utilisés à cette fin.
Cette étude vise à évaluer la faisabilité et l’efficacité d’une couche de faible perméabilité (CFP) faite de sable naturel traité
avec des polymères superabsorbants (PSA) de type polyacrylate de sodium (PA-Na) comme barrière à l’oxygène et/ou à
l’eau dans les systèmes de recouvrement des parcs à résidus miniers fins. Un type de PSA à grains fins (SAG-M85F) a
été sélectionné pour amender la couche de sable. Des essais de conductivité hydraulique saturée (k sat) ont été réalisés à
différents dosages de PSA dans le sable (0.5, 0.6 et 0.7%) ainsi que des essais d’infiltration dans des colonnes
instrumentées. Les résultats des essais de perméabilité ont montré que le ksat de la CFP (mélange sable et PSA) diminuait
avec l’augmentation de la proportion de PSA, soit entre 10 -7 et 10-8 m/s, comparé au ksat mesurée pour le sable seul
(témoin) qui était de 10-5 m/s, soit une réduction de 2 à 3 ordres de grandeur. Les essais en colonnes instrumentées ont
également montré que, selon le dosage en PSA, la CFP (sable plus PSA) devenait imperméable à l’eau après que le PSA
s’est transformé en gel à la suite du rinçage. Ce qui suggère qu’il serait possible de créer une CFP efficace en amendant
du sable avec une très faible quantité ( 1% de la masse sèche de sable) PSA de type PA-Na.

1 INTRODUCTION to the leaching of heavy metals, posing a threat to the

surrounding ecosystem (Bouzahzah et al., 2014; Bussière
Sulfide mine tailings, generated during the extraction and et al., 2005; Blowes et al., 1998). Furthermore, oxygen
processing of polymetallic ore deposits, are commonly transport within the tailings primarily occurs through
stored in tailings storage facilities (TSF). In many cases, diffusion within pore spaces (Aubertin et al., 2000).
the containment dams surrounding the TSF are built using Consequently, preventing and managing AMD resulting
the tailings themselves (Aubertin et al., 2002). However, from sulfide mineral oxidation is a significant
when these tailings are exposed to environmental environmental challenge faced by the Canadian mining
conditions such as precipitation and oxygen in the air, the industry, particularly in Quebec. Numerous studies in the
sulfides present within them react with the oxygen, literature have proposed preventive techniques for the
resulting in the formation of acid mine drainage (AMD) with effective management of sulfide mine tailings (Bussière,
a pH of 3 or lower. This acidic water can subsequently lead 2007; Bussière et al., 2005), including water covers
(Awoh, 2012) and single or multilayer soil covers (Aachib multi-layer systems as a drainage layer and/or capillary
et al., 1994; Aubertin et al., 2002; Demers et al., 2008). break layer within covers that exhibit capillary barrier
The underlying principle is to limit the diffusion of oxygen effects (CCBE). It is worth mentioning that the sand was
and infiltration of water into the sulfide tailings by using fine oven-dried to facilitate the mixing process with the
geomaterials (such as silt), geosynthetic materials (such superabsorbent polymers.
as geomembranes, geosynthetic clay liners-GCLs),
desulfurized tailings, or combination of synthetic and
geomaterials or composite barriers (Aubertin et al., 2002)
to construct a moisture retention layer (MRL) with a high
degree of saturation ( 85%). This MRL serves as a barrier
against oxygen and water in the cover system, thereby
mitigating the risk of acid mine drainage.
Sodium polyacrylate (Na-PA) superabsorbent
polymers (SAPs) [CH2-CH(COONa)-]n are synthetic
macromolecular materials with a three-dimensional
structure renowned for their exceptional fluid absorption
capacity. They can absorb fluids, water, and mining Figure 1. Picture of SAP powder
effluents hundreds of times their dry mass (Chen and
Zhao, 1999). SAPs find applications in various fields such The particle size distributions (PSD) were determined
as agriculture and horticulture, where they serve as water using a Mastersizer 3000 laser diffraction particle size
or moisture retention materials (Casquilho et al., 2013; analyser from Malvern Panalytical and equipped with a dry
Huettermann et al., 2009; Hüttermann et al., 1999). They dispersion unit. The PSD curves (see Figure 2) revealed
are also utilized in civil engineering for cementitious that the SAG-A03P type exhibited a coarse-sized particle
material hardening (Jensen and Hansen, 2001, 2002; distribution, whereas the SAG-M06L and SAG-M06K
Snoeck et al., 2012). types demonstrated a medium-sized particle distribution.
In the hydrocarbon sector, SAPs offer numerous On the other hand, the SAG-M85F type displayed a very
advantages (Farkish and Fall, 2013; Peer and Venter, fine-grained particle distribution characterized by a
2003; Roshani et al., 2017; Wang et al., 2010). Within the coefficient uniformity (CU) value lower than 3 (see Table
mining industry, they have shown promise in tailings 1).
dewatering (Belem et al., 2019; Kabamba, 2019), and
acting as lining materials for acid mine drainage Table 1. Particle gradation parameters of sand and the
management (Gain et al., 2005; Gao, 2003; Haug and different types of SAP
Pauls, 2002; Zhou et al., 1993). When exposed to water,
these fine-particle polymers swell and form a gel-like D10 D30 D60
Material CU
substance that impedes the infiltration of water into (m) (m) (m)
underlying layers. This phenomenon is known as gel- Sand 150 251 421 2.80
blocking. Consequently, when suitable materials such as Goldex tailings 3.7 16.8 62.5 17.05
silt or desulfurized tailings are not readily available, sand SAG-M85F 139 187 259 1.87
mixed with SAPs can be utilized to create low saturated SAG-M06K 254 346 476 1.87
hydraulic conductivity (LSHC) barriers that can oxygen SAG-A03P 261 350 477 1.83
diffusion and limit water infiltration. Indeed, LSHC barriers
SAG-M06L 522 625 750 1.44
usually include a low saturated hydraulic conductivity (k sat)
layer made of clay, geomembrane or geosynthetic clay
liner (GCL) with a ksat less than 10-9 m/s (Aubertin et al., 100
SAG-M06K
2015; Maqsoud et al., 2021). 90 SAG-A03P
80
Cumulative %passant

SAG-M85F
2 MATERIALS AND METHODS 70 SAG-M06L
60 Sand
50 Goldex tailings
2.1 Materials
40
30
In this study, various materials were utilized, including 20
commercial normal concrete sand and four types of 10
sodium polyacrylate superabsorbent polymers (Figure 1) 0
provided by Recyc PHP Inc. from Drummondville, 0.01 0.1 1 10 100 1000 10000
Quebec, Canada (SAG-M85F, SAG-M06K, SAG-A03P, Particle diameter (µm)
SAG-M06L). The superabsorbent polymers were used as
sand amendment materials to create low saturated Figure 2. Particle size distribution curves for base
hydraulic conductivity (LSHC) barriers, with the objective materials
of restricting or preventing water infiltration and the
diffusion of oxygen towards the acidic tailings layer. The
choice of sand, characterized by a saturated hydraulic
conductivity (ksat) ranging between 10-5 and 10-3 m/s, was
based on its inert nature and its frequent application in the
restoration of mine tailings impoundments, particularly in
2.2 Methods in the study consisted of regular sand without any SAP
content.
2.2.1 Assessment of the SAP absorption capacity
2.2.4 Permeability tests
The study commences with the evaluation of the
absorption rate of the four types of superabsorbent The saturated hydraulic conductivity (k sat) is a crucial
polymers (SAP), namely SAG-M85F, SAG-M06L, SAG- parameter that characterizes the permeability of porous
M06K, and SAG-A03P. This assessment aims to materials, indicating their ability to allow fluid flow. Prior to
determine their water retention capacity. To achieve this, conducting the tests, achieving 100% saturation of the
1 gram of each dry SAP type is immersed in 350 ml of materials is strongly emphasized. The laboratory
deionized water for a fixed duration of one hour. Once permeability tests were carried out using the constant
equilibrium is reached, the swollen SAP or hydrogel is head method with the rigid wall permeameter for the
filtered using tea bags and air to eliminate any remaining regular sand (control sample) and in accordance with
water trapped between the swollen SAP particles. ASTM D2434-06. The permeability tests on the sand-SAP
Subsequently, the hydrogel is weighed to determine the mixtures containing 0.5%, 0.6%, and 0.7% SAP were
mass of the swollen SAP at equilibrium, enabling the performed using the falling head method with the rigid wall
calculation of absorbency using the following equations: permeameter following ASTM D5856-95. However, direct
permeability testing on swollen SAP (hydrogel) was not
𝐾𝑆𝐴𝑃 =
𝑀
[1] performed due to challenges associated with achieving
𝑀0
complete saturation within the permeameter, as well as
M−M0
the gelatinous nature of the swollen SAP (Figure 3).
𝑄𝑒𝑞 = = 𝐾𝑆𝐴𝑃 – 1 [2] Limitations in the laboratory equipment and techniques
M0
(such as Tempe cells) prevented the determination of
where KSAP is the absorption capacity or absorbency, Qeq water retention curves (WRC) for the Sand-SAP mixture
is the equilibrium absorption rate, M and M0 represent the samples.
mass of the water-swollen SAP grains and the mass of the
dry SAP grains, respectively. As a reminder, the (a) (b)
absorption capacity represents the maximum amount of
water or solution that can be absorbed by the SAP
samples at equilibrium. Subsequently, the SAP samples
in their swollen state are exposed to the laboratory
ambient air for the desorption process.

2.2.2 Assessment of the SAP desorption capacity

Desorption experiments were conducted to evaluate the

water storage rate of swollen SAPs in their bulk state
under ambient conditions. The initial water absorption
capacity of SAPs after a residence time of at least 24
hours (KSAP-gel(t0)) was determined by dividing the mass of Figure 3. Rigid permeameters with SAP modified sand
the swollen SAP gel (MSAP-gel(t0)) by the mass of the dry (0.3% to 0.7% of SAP): a) the saturation phase, and b) the
SAP (MSAP-dry). To assess the desorption rate or water swelling of the material after completion of saturation
storage capacity, the SAP gels were exposed to air and
weighed at various time intervals (2, 5, 9, 13, 16, 23, 30, 2.2.5 Column seepage experiments
and 37 days). By considering the mass of the SAP gel at
a given time (MSAP-gel(t)), the desorption rate or water The column seepage experiments involved the installation
retention/holding capacity (%RC) can be calculated using of instrumented columns equipped with volumetric water
the following equation: content sensors (EC-5 type) and matric suction sensors.
Two distinct column configurations were employed: the
𝐾𝑆𝐴𝑃−𝑔𝑒𝑙 (𝑡) 𝑀𝑆𝐴𝑃−𝑔𝑒𝑙 (𝑡)
%𝑅𝐶 = = [3] first configuration (Figure 4a) represented solely the LSHC
𝐾𝑆𝐴𝑃−𝑔𝑒𝑙 (𝑡0 ) 𝑀𝑆𝐴𝑃−𝑔𝑒𝑙 (𝑡0 )
barrier or layer, which consisted of a Sand-SAP mixture
containing 0.5% SAP (emulating a monolayer cover). The
where KSAP-gel(t) represents the residual water absorbency second configuration (Figure 4b) included a capillary
of SAP gels after desorption under ambient air exposure break, with the LSHC layer sandwiched between two
for a specific time period (t). layers of sand (emulating a cover with capillary barrier
effect). For the second configuration (Figure 4b), Sand-
2.2.3 Sand-Superabsorbent polymer mixtures SAP mixtures with proportions of 0.7% and 1% SAP were
utilized, and the thickness of the sand layers was 10 cm at
To create low saturated hydraulic conductivity (LSHC) the bottom of the column and 15 cm above the LSHC
barriers, a series of Sand-SAP mixtures were prepared layer. This configuration was chosen in order to counteract
using varying proportions (0%, 0.5%, 0.6%, 0.7%, and the swelling of the SAP modified sand, considering the
1%) of the SAP type SAG-M85F, known for its fine grains loading effect exerted by the sand layer placed above the
and superior gel-blocking ability. The control material used LSHC layer.
 0 cm 0 cm

atermar soil ater

15 cm 60
tension sensor
5
C 5 volumetric
ater content sensor
50
0 cm Column void space
0
0 Sand SAP mixture layer or
0 cm S C layer
10 Sand layer (protection or
10 cm 5 drainage)

Figure 4. Schematic diagrams of the instrumented column assembly and configurations

2.2.6 Determination of oxygen flux 3 RESULTS

Assessing the oxygen flux through the LSHC layer is 3.1 Absorbed water retention capacity
crucial for evaluating its effectiveness as a barrier against
oxygen and/or water, thereby preventing or significantly The objective of the desorption tests was to identify the
reducing the oxidation of underlying sulfide tailings. SAP type that exhibits the highest long-term water
Various experimental methods can be employed to retention capacity (%RC). Such a capacity is crucial for
estimate the oxygen flux, including the sulfate depletion maintaining the LSHC layer in a saturated state, which
method, oxygen consumption method, and gradient effectively reduces the diffusion of water and oxygen. The
method. In this study, the gradient method in combination results revealed that the finest-grained SAP (SAG-M85F)
with gas chromatography will be utilized. To deem the exhibited the highest water retention capacity, retaining
LSHC layer effective, it is important that the annual oxygen 90% of absorbed water even after 5 weeks of desorption.
diffusion falls within the range of 20 to 40 g of O2/m2/year In contrast, the coarser-grained SAPs (SAG-M06L and
(Bussière et al., 2009; Dagenais, 2005). SAG-A03P) maintained this performance only for the initial
The gradient method entails measuring oxygen two weeks of exposure to ambient air. Figure 5
concentrations at multiple elevation levels within the LSHC demonstrates that after 5 weeks of desorption, the
layer material. These measurements can be performed in retention capacity decreased to a minimum of 70%,
the laboratory using instrumented columns or in situ on indicating a release of up to 30% of the SAP water for the
existing cover layers. This practical method is applicable coarser-sized SAPs (SAG-M06L and SAG-A03P). These
to both inert and reactive materials under steady-state significant findings from the desorption tests provide
conditions. By determining or estimating the effective encouraging evidence for the use of SAP in the
diffusion coefficient (De) and obtaining oxygen preparation of Sand-SAP mixtures as LSHC barrier
concentration measurements, Fick's first law is applied to materials.
calculate the oxygen flux through the LSHC layer.
In the experimental procedure outlined in this study, a 3.2 Saturated hydraulic conductivity
3 to 5 ml air sample is extracted using a five-milliliter
syringe from the gas sampling ports installed on the It should first be pointed out that for Sand-SAP mixtures
columns. Subsequently, the collected air is introduced into containing 0.9% and 1% SAP, the flexible-wall
the Agilent microGC gas chromatograph, which operates permeameter (triaxial cell) was used with the falling head
using the Agilent Cerity Data System software. The gas is test method in accordance with ASTM D5856-95. In
then analyzed using oxygen and nitrogen detectors, as addition, the results of tests with Goldex mine tailings will
well as CO2 detection coils. not be presented in this paper for lack of space.
Subsequently, the calculation of oxygen fluxes using The results of the saturated hydraulic conductivity (k sat)
the oxygen gradient method was carried out at three measurements of sand and Sand-SAP type M85F
locations within the LSHC layer: top, center, and bottom. mixtures using the rigid and flexible wall permeameters
The calculation considered an initial oxygen concentration are presented in Table 1.
of 8.65 mol/m3 (corresponding to 276.7 g/m3) in order to
determine the oxygen gradient.
 100% SAG M 5 F

5%

ater retention /holding capacity %

0%

5%

0%

75%
SAG M06
M5F
70%
M06
SAG A0 P M06
65%
A0 P
60%
0 10 0 0 0
lapsed time ( day) SAG M06
Figure 5. Desorption curves showing the retention capacity of the absorbed water and the textural aspect of the four types
of SAP gel after 5 weeks of desorption at ambient air

Table 2. Measured hydraulic conductivity (ksat) of sand and It is therefore difficult, if not impossible at this stage, to
Sand-SAP type M85F mixtures compare the results of this study with those in the
literature.
ksat measured Percolation rate
Sand-SAP mixture
(m/s) (mm/yr)
Sand (control) 2.2110-5 696,946
Sand + 0.5% SAP 4.3410-7 13,687 1.00 0
Sand + 0.6%SAP 3.4810-7 10,975
Sand + 0.7% SAP 3.6510-7 11,511
Measured sat values (m/s)

Sand + 0.8% SAP 3.5110-7 11,069

Sand + 0.9% SAP¶ 2.9410-7 9,272 1.00 05
Sand + 1% SAP¶ 2.0710-8 653
¶
Obtained with the flexible wall permeameter

1.00 06
The measured saturated hydraulic conductivity (k sat)
values of the sand-SAP mixtures range from 10-5 m/s to
10-8 m/s. The obtained results clearly demonstrate that
incorporating a small quantity of SAP type M85F (0.3% to 1.00 07
1% of the dry mass of sand) into the polymer-modified
sand led to a substantial reduction in k sat values by two
orders of magnitude, depending on the proportion of SAP
type M85F in the mixture. It is important to note that, in the 1.00 0
available literature, there is a lack of permeability test
results specifically conducted on mixtures of geomaterials
containing sodium polyacrylates SAP; instead, the focus
has been on polyacrylamide SAP.
Figure 6 clearly illustrates the substantial impact of
incorporating a small proportion of SAP in the sand-SAP
mixture on the saturated hydraulic conductivity (k sat)
values. As the proportion of SAP increases, ksat Figure 6. Variation in ksat of the sand-SAP mixtures as a
decreases. However, a slight variation in k sat is observed function of the proportion of SAP added
among the sand-SAP mixtures with 0.5% to 0.9% SAP
addition. This marginal difference can be attributed to the
presence of coarse particles in the sand, as shown in
Figure 7. Therefore, it can be hypothesized that if the sand
was pre-sieved to remove these coarse particles, a more
significant reduction in saturated hydraulic conductivity
could be achieved, potentially reaching values on the
order of 10-9 m/s. Notably, the sand-SAP mixture with 1%
SAP demonstrated good consolidation even after
complete drying, as depicted in Figure 7.
 (a) (b) functioned as a pore-filling material, effectively filling the
pores of the sand-SAP mixture in its gel form.

Supernatant
f fect of the ater
presence of
coarse particles S ollen Protection layer
(Sand SAP hydrogel
mixture ith igh density dried
1% SAP) sand SAP mixture S C ayer
ith 1% SAP (Sand SAP)
specimen
Drainage layer
Figure 7. Representative samples of sand-SAP mixture
(1% SAP) after drying following permeability tests: a) Figure 8. Diagram illustrating the observations following
specimen with presence of coarse particles, which may column infiltration tests: a) monolayer configuration with
affect the permeability characteristics, b) specimen with LSHC containing 0.5% SAP of type M85F, b) multilayer
dense and consolidated appearance, indicating a potential configuration with LSHC containing 0.7% or 1% SAP of
reduction in permeability due to the presence of SAP type M85F
3.3 Measurement of percolation through the LSHC
layer Furthermore, it is worth noting that the water content by
volume of the mixture containing 0.7% SAP, influenced by
Figure 8a illustrates the results of water percolation tests the gel blocking phenomenon, reached only 0.4. This low
conducted on columns containing a LSHC layer with 0.5% water content reflects the significant impact of the gel's
SAP of type M85F (monolayer configuration). Notably, it is presence within the mixture. Notably, throughout the entire
important to highlight that a significant amount of time, at rinsing period, not a single drop of water escaped from the
least 2 weeks, was required to moisten a depth of less columns, emphasizing the effectiveness of the gel in
than 2 cm within the LSHC layer. This outcome is preventing any water loss.
intriguing considering the inherent high permeability of the These observations highlight the transformative effect
sand alone. Remarkably, the SAP located in the upper of the gel formation, which enhances the water retention
portion of the column exhibited rapid swelling, ascending capacity of the sand-SAP mixture while effectively sealing
several centimeters towards the top. Subsequently, a the pores. This gel-based system shows promise for
phenomenon known as "gel blocking" occurred, which applications where water retention and control are critical
impeded further water penetration and resulted in water factors.
retention at the surface. The observed volume increase
towards the upper section of the column, due to the 3.4 Determination of oxygen flow through the LSHC
swelling of SAP, can be attributed to the absence of any layer
load exerted on the SAP at the surface, which allowed
unrestricted swelling. It is worth noting that these findings The measurements of volumetric water content (VWC)
emphasize the intricate interaction between the along with the porosity (n) values obtained from material
permeability of the sand and the behavior of SAP within placement in the columns allowed for the calculation of
the system. effective oxygen diffusion coefficients (De) in each LSHC
Figure 8b presents the results for columns containing layer using the equations developed by Aachib et al.
the LSHC layer with 0.7% and 1% SAP of type M85F. It is (2004).
important to reiterate that in this configuration (multilayers After approximately one week of rinsing for the upper
system), the layer of sand above the LSHC layer serves part of the LSHC layer and around ten days for the lower
as a protective barrier, exerting a certain geostatic part of the LSHC layer, the oxygen diffusion coefficient
pressure. Indeed, this sand layer prevented significant becomes nearly zero, and this remains the case for the
swelling of the SAP while facilitating improved water entire duration of the test. This confirms the oxygen barrier
percolation towards the underlying LSHC layer. role of the LSHC layer.
Consequently, this multilayer configuration led to the
wetting of the sand-SAP mixture within the initial 2 to 3 Table 3. Calculated effective oxygen diffusion coefficient
weeks of rinsing, in contrast to the monolayer (De) at different sampling points in the polymer modified
configuration comprising the sand-SAP mixture with 0.5% sand with 0.7% type M85F SAP
SAP of type M85F (Figure 8a).
After achieving complete wetting, the LSHC layer
exhibited a remarkable absence of water drainage during Cycle C/Z (mol/m4) De (m2/s)
subsequent rinsing cycles. At this stage, the SAPs present monitored Bottom Top Bottom Top
in the sand-SAP mixture underwent complete 1 8.2E-01 9.3E-01 1.06E-05 3.17E-08
transformation, forming a gel that occupied all the pores
2 6.6E-01 6.6E-01 2.55E-10 5.22E-10
within the mixture. This gel played two significant roles:
firstly, it acted as a water retention material, effectively 3 4.8E-01 4.8E-01 4.00E-10 2.26E-10
retaining water within the system, and secondly, it
Table 4. Calculated effective oxygen diffusion coefficient further studies are necessary, it can be concluded that the
(De) at different sampling points in the polymer modified SAPs modified sand layers have demonstrated their
sand with 1% type M85F SAP effectiveness as low-permeability barriers, successfully
achieving the targeted objectives in terms of oxygen flux,
C/Z (mol/m4) De (m2/s) which should be below 1 mol/m²/year (the threshold for
Cycle
barrier system effectiveness).
monitored Bottom Top Bottom Top
1 1.04E+00 1.21E+00 1.74E-07 3.22E-09
2 0.41E-01 0.71E-01 2.26E-10 2.26E-10 4 CONCLUDING REMARKS
3 0.38E-01 0.66E-01 2.26E-10 2.26E-10
This paper focuses on determining the saturated hydraulic
conductivity and gas diffusion characteristics of materials
The calculation results show that the diffusion coefficients
composed of sand mixed with different proportions
(De) range from 10-5 to 10-8 for the sand below the LSHC
(ranging from 0.5% to 1.0%) of fine-grained SAP (type
layer (capillary break) and up to 10 -10 for the LSHC layer M85F). Based on the obtained results and observations,
with 0.7% SAP. the following key findings emerge:
In this study, the oxygen (O2) gradient method was
▪ Saturated hydraulic conductivity (ksat): The
employed to determine the O2 flux. It is important to note
introduction of a small quantity (0.5% to 1%) of SAP
that the accuracy of the oxygen concentration significantly reduces the ksat of the sand alone (10-5
measurements obtained from the gas chromatograph and m/s) by 2 to 3 orders of magnitude. The k sat of the
the volumetric water content (VWC) measurements
sand-SAP mixture drops to a range between 10-7 and
significantly influence the reliability and precision of the
10-8 m/s, rendering it virtually impermeable.
gradient method. ▪ LSHC layer and water percolation: The LSHC layer,
The oxygen flux values obtained from the three rinses comprised of sand modified with SAP (sodium
(or cycles) of the tested LSHC layer configurations are
polyacrylate, Na-PA), slows down water percolation
presented in Tables 5 and 6. These results indicate that,
and infiltration primarily due to the gel blocking
except for cycle 1, the O2 fluxes remain constant and phenomenon. The formed gel fills the pores within the
comparable between the top and bottom of both LSHC LSHC layer, derived from the sand-SAP mixture,
layer configurations (0.7% and 1% SAP type M85F) and
resulting in virtual impermeability.
are remarkably low. This behavior can be attributed to the
▪ Gas diffusion and oxygen migration: Calculated
presence of SAPs, which form a gel that fills the voids and oxygen flows demonstrate that employing an LSHC
hinders the infiltration of water and O2. It is worth noting layer made of a fine sand-SAP mixture, at a low
that these fluxes are below 1 mol/m²/year, which is the
proportion (0.7% to 1.0% of the dry mass of sand),
target value for effective cover systems.
significantly reduces the migration of oxygen towards
underlying reactive residues. The LSHC layer,
Table 5. Oxygen fluxes calculated by the O2 gradient composed of the fine sand-SAP mixture, effectively
method at different sampling points in the polymer acts as a moisture retention layer, achieving the
modified sand with 0.7% SAP type M85F
targeted objective of limiting the oxygen flux to less
than 1 mol/m²/year.
Cycle C/Z (mol/m4) O2 flux (mol/m2/yr)
monitored Bottom Top Bottom Top These findings highlight the potential of utilizing a capping
system with an LSHC layer composed of a fine sand-SAP
1 8.2E-01 9.3E-01 3.0E+02 1.0E-02
mixture for industrial applications. However, further
2 6.6E-01 6.6E-01 5.3E-03 1.0E-02 studies are necessary to determine the applicability and
3 4.8E-01 4.8E-01 6.0E-03 3.0E-03 scalability of these results at the field scale.

5 REFERENCES
Table 6. Oxygen fluxes calculated by the O2 gradient
method at different sampling points in the polymer Aachib, M., Aubertin, M. and Chapuis, R. P. 1994. Column
modified sand with 1% SAP type M85F tests investigation of milling wastes properties used to
build cover systems. Proceedings of 3rd International
O2 flux (mol/m2/yr)
Conference on the Abatement of Acidic Drainage,
Cycle C/Z (mol/m4)
Pittsburgh.
monitored Bottom Top Bottom Top Aubertin, M., Pabst, T., Bussière, B., James, M.,
1 1.04E+00 1.21E+00 3.90E+00 7.2E-02 Mbonimpa, M., Benzaazoua, M. et Maqsoud, A. 2015.
2 0.41E-01 0.71E-01 5.08E-03 2.0E-03 Revue des meilleures pratiques de restauration des
sites d’entreposage de rejets miniers générateurs de
3 0.38E-01 0.66E-01 5.00E-03 3.0E-03
DMA. In Proceedings of the Symposium 2015 sur
l’Environnement et les Mines, Rouyn-Noranda,
The preliminary findings from the oxygen flux calculations Québec, Canada; 14 - 17 June 2015.
suggest that a cover system employing an LSHC layer Aubertin, M., Aachib, M. et Authier, K. 2000. Evaluation of
composed of a sand-SAP mixture (with a small proportion diffusive gas flux through covers with a GCL.
of SAP, ranging from 0.7% to 1% of the dry mass of sand) Geotextiles and Geomembranes, 18(2-4), 215-233.
in a cover system would significantly mitigate the migration
of oxygen towards the reactive tailings beneath. Although
Aubertin, M., Bussière, B., Aachib, M., Chapuis, R. et Gao, D. 2003. Superabsorbent polymer composite
Crespo, J. 1996. Une modélisation numérique des (SAPC) materials and their industrial and high-tech
écoulements non saturés dans des couvertures applications.
multicouches en sols. Hydrogéologie (Orléans), (1), 3- Haug M.D., and Pauls G. 2002. A review of non-traditional
14. dry covers. MEND Project Report 2.21.3b, 142 p.
Aubertin, M., Bussière, B., Bernier, L., Chapuis, R., Julien, Huettermann, A., Orikiriza, L.J. and Agaba, H. 2009.
M., Belem, T., Simon, R., Mbonimpa, M., Benzaazoua, Application of superabsorbent polymers for improving
M. et Li, L. 2002. La gestion des rejets miniers dans un the ecological chemistry of degraded or polluted lands.
contexte de développement durable et de protection Clean–Soil, Air, Water, 37(7), 517-526.
de l’environnement. Congrès annuel de la Société Hüttermann, A., Zommorodi, M. and Reise, K. 1999.
canadienne de génie civil, 5-8. Addition of hydrogels to soil for prolonging the survival
Awoh, A.S. 2012. Étude expérimentale du comportement of Pinus halepensis seedlings subjected to drought.
géochimique de résidus miniers hautement sulfureux Soil and tillage research, 50(3-4), 295-304.
sous un recouvrement en eau [Université du Québec Jensen, O.M. and Hansen, P.F. 2001. Water-entrained
en Abitibi-Témiscamingue]. cement-based materials: I. Principles and theoretical
Belem, T., El Mahboub K., Kabamba C.N., Poudama L.É., background. Cement and concrete research, 31(4),
Addi N., Mbonimpa M. et A., M. 2019. Évaluation de la 647-654.
performance des polymères super-absorbants (PSA) Jensen, O. M. and Hansen, P.F. 2002. Water-entrained
comme alternative de déshumidification des résidus cement-based materials: II. Experimental
miniers en pulpe avant leur entreposage en surface ou observations. Cement and concrete research, 32(6),
sous terre. Rapport synthèse FRQNT, 12 pages. 973-978.
Blowes, D.W., Jambor, J. L., Hanton-Fong, C.J., Lortie, L. Kabamba, C. 2019. Optimisation du dosage des
and Gould, W.D. 1998. Geochemical, mineralogical polymères superabsorbants (PSAs) dans le processus
and microbiological characterization of a sulphide- de densification des résidus miniers fins en pulpe:
bearing carbonate-rich gold-mine tailings application au cas de la mine Canadian Malartic. Ecole
impoundment, Joutel, Québec. Applied Geochemistry, Polytechnique, Montreal (Canada).
13(6), 687-705. Maqsoud, A., Bussière, B. and Mbonimpa, M. 2021. Low
Bouzahzah, H., Benzaazoua, M., Bussiere, B. and Plante, Saturated Hydraulic Conductivity Covers. In Hard
B. 2014. Prediction of acid mine drainage: importance Rock Mine Reclamation: From Prediction to
of mineralogy and the test protocols for static and Management of Acid Mine Drainage, Bussière, B.,
kinetic tests. Mine Water and the Environment, 33(1), Guittonny, M., Eds, CRC Press, Boca Raton, FL, USA,
54-65. 2021, p. 93-113.
Bussière, B. 2007. Colloquium 2004: Hydrogeotechnical Peer, F. and Venter, T. 2003. Dewatering of coal fines
properties of hard rock tailings from metal mines and using a super absorbent polymer. Journal of the
emerging geoenvironmental disposal approaches. Southern African Institute of Mining and Metallurgy,
Canadian Geotechnical Journal, 44(9), 1019-1052. 103(6), 403-409.
Bussière, B., Aubertin, M., Zagury, G.J., Potvin, R. et Roshani, A., Fall, M. and Kennedy, K. 2017. Drying
Benzaazoua, M. 2005. Principaux défis et pistes de behavior of mature fine tailings pre-dewatered with
solution pour la restauration des aires d’entreposage super-absorbent polymer (SAP): column experiments.
de rejets miniers abandonnées. Symposium 2005 sur Geotechnical Testing Journal, 40(2), 210-220.
l’environnement et les mines, Rouyn-Noranda, Snoeck, D., Van Tittelboom, K., De Belie, N., Steuperaert,
Québec, Canada. S. and Dubruel, P. 2012. The use of superabsorbent
Casquilho, M., Rodrigues, A. and Rosa, F. 2013. polymers as a crack sealing and crack healing
Superabsorbent polymer for water management in mechanism in cementitious materials. Proceedings of
forestry. Agricultural Sciences, 4(05), 57-60. the 3rd International conference on concrete repair,
Chen, J. and Zhao, Y. 1999. An efficient preparation rehabilitation and retrofitting, Cape Town.
method for superabsorbent polymers. Journal of Wang, X., Feng, X., Xu, Z. and Masliyah, J.H. 2010.
applied polymer science, 74(1), 119-124. Polymer aids for settling and filtration of oil sands
Demers, I., Bussière, B., Benzaazoua, M., Mbonimpa, M. tailings. The Canadian Journal of Chemical
and Blier, A. 2008. Column test investigation on the Engineering, 88(3), 403-410.
performance of monolayer covers made of Zhou, Z, Wiwchar, B., Gunter, W.D. and Gao, D. 1993.
desulphurized tailings to prevent acid mine drainage. Polymer-modified clay as impermeable barriers for
Minerals Engineering, 21(4), 317-329. acid mining tailings. MEND Project Report 6.2, 31 p.
Farkish, A. and Fall, M. 2013. Rapid dewatering of oil sand
mature fine tailings using super absorbent polymer
(SAP). Minerals Engineering, 50, 38-47.
Gain, O., Espuche, E., Pollet, E., Alexandre, M. and
Dubois, P. 2005. Gas barrier properties of poly (ε‐
caprolactone)/clay nanocomposites: Influence of the
morphology and polymer/clay interactions. Journal of
Polymer Science Part B: Polymer Physics, 43(2), 205-
214.
Experimental and numerical study of the
hydrogeological behaviour of filtered
tailings exposed to climatic conditions
M. Chai-Onn, T. Pabst
Department of Civil, Geological and Mining Engineering -
Research Institute on Mines and Environment (RIME)
Polytechnique Montréal, Montréal, Québec, Canada

ABSTRACT
Filtered tailings are an alternative approach to mine waste management that consists of reducing water content of tailings
to allow their disposal without the need for impoundments nor dams. However, filtered tailings storage facilities (FTSF) are
left exposed to climate conditions during the operations and the interactions between the tailings and the atmosphere may
lead to long-term modification of their hydrogeotechnical behaviour and the water table position. The objective of this study
was to determine the effects of climate conditions on the hydrogeological behaviour of filtered tailings. An experimental
column was constructed and monitored while various climatic conditions were simulated at the surface. These results were
used to calibrate a numerical model of the filtered tailings column. The calibrated results effectively matched global
laboratory trends. Preliminary results emphasized the need to account for climate conditions, particularly precipitation
intensity and initial water content, when assessing the long-term hydrogeological behavior of FTSF.

RÉSUMÉ
Les résidus filtrés sont une approche alternative de gestion des résidus miniers qui consiste à réduire leur teneur en eau,
pour permettre leur déposition sans avoir recours à des digues de rétention. Cependant, les parcs à résidus (PAR) filtrés
sont exposés aux conditions climatiques pendant les opérations de la mine et l’interaction entre les résidus et l'atmosphère
peuvent entraîner des modifications à long terme de leur comportement hydrogéotechnique et de la position de la nappe
phréatique. L'objectif de cette étude était d’évaluer les effets des conditions climatiques sur le comportement
hydrogéologique des résidus filtrés. Une colonne expérimentale a été construite et suivie pendant que différentes
conditions climatiques étaient simulées à la surface. Ces données ont été utilisées pour calibrer un modèle numérique de
la colonne expérimentale. Les résultats calibrés correspondent aux tendances globales observées en laboratoire. Les
résultats préliminaires ont souligné l'importance de prendre en compte les conditions climatiques, notamment l'intensité
des précipitations et la teneur en eau initiale, lors de l'évaluation du comportement hydrogéologique à long terme d’un
PAR filtrés.
Tailing dewatering is an alternative tailings
1 INTRODUCTION management approach that aims to reduce water content
of tailings before their transport and their deposition, to
One of the main challenges for the mining industry is to increase their shear resistance and lower the liquefaction
manage responsibly the large quantities of waste rocks and susceptibility, thus reducing risks of dam failure. Filtered
tailings that are produced during the operations (Spitz & tailings are typically dewatered to a water content < 20 %.
Trudinger, 2019). Mine tailings have the particle size of As filtered tailings behave like a soil rather than a liquid,
sand or silt, with a typically low percentage of solids and a they must be transported by truck or conveyor, compacted
high water content (Bussière, 2007; MEND, 2017), making (which also contributes to improve their geotechnical
them prone to geotechnical instabilities and liquefaction properties) and stacked without the need for dams (MEND,
(Hassani et al., 2008; James et al., 2011). Mine tailings are 2017).
conventionally disposed of as slurry in surface tailings The disposal of filtered tailings represents a
storage facilities (TSF) surrounded by embankment dams significant advancement in tailings management, because
(Aubertin et al., 2003; Bussière & Guittonny, 2020; MEND, it effectively increases the slope stability of the TSF (Li et
2017; Vick, 1990) which are typically progressively al., 2016), maximises water recovery (Gunson et al., 2012;
constructed as the TSF is filled (Lottermoser, 2007). Oldecop & Rodari, 2021) and increases the storage
Climate change (Bresson et al., 2022), the indefinite capacity by reducing the volume of tailings to manage.
lifetime of the impoundments, the large size of the TSF Despite the costs associated with filtered tailings
(Azam & Li, 2010; Lyu et al., 2019; Rico et al., 2008), and production and management being higher than those of
the particular properties of tailings all concur to increase conventional tailings (Carneiro & Fourie, 2020; Furnell et
the risk for geotechnical instabilities in the long-term. al., 2022; MEND, 2017), filtration is increasingly considered
Despite the improvement of technologies and methods for by mining compagnies.
the construction of impoundments, the volume of tailings As the performance of surface FTSF depends on
discharged in the environment and the number of human site-specific conditions, especially in humid climate with
lives lost due to dam failure has been steadily increasing positive water balance, climatic conditions and climate
over the past decades (Santamarina et al., 2019). changes raise questions about the long-term stability of
FTSF (Ulrich et al., 2019). During mining operations, FTSF
are directly exposed to climate conditions. In regions with
positive water balance, water can accumulate in the FTSF.
Particularly, precipitation events and snow melting can
contribute to progressively raise the water table in the
tailings. High water table increases the pore water
pressure, reducing the effective stress within the soil. This
leads to reduced shear strength, thus increasing the risk
for slope instabilities (Skempton, 1960). Hence the
importance of understanding the interaction of filtered
FTSF with the atmosphere.
This paper presents the preliminary results of wetting
and drying cycles applied on an experimental column to
evaluate the effects of climate conditions on the
hydrogeological behaviour of filtered tailings. The
experimental column was also simulated with a calibrated
numerical model.

2 METHODOLOGY
2.1 Tailings properties Figure 2: Water retention curve of tailings. The van
Genuchten model parameters fitted on the experimental
The filtered tailings used in this study were desulphurized points were: alpha = 0.0017, n = 1.9, AEV = 2 m. The van
tailings sampled from a partner mine site located in Genuchten model parameters of the calibrated WRC were:
Québec, Canada. alpha = 0.0049, n = 1.8, AEV = 1.1 m.
The average specific gravity of the tailings
determined from a total of 9 laboratory tests was 2.78 2.2 Experimental column setup
(ASTM-D854-14 (2023)). The particle size distribution
curve was measured according to ASTM-D6913-04 (2021) An experimental column of 25.5 cm in diameter and 150
and ASTM-D1140-14 (2017) (Figure 1) from 15 laboratory cm in height was setup in the laboratory to study the
tests. The average diameters corresponding to 10% and hydrogeological behaviour of filtered tailings exposed to
60% of the cumulative mass passing, are 0.0029 mm and climatic conditions. A sand layer was placed at the bottom
0.0271 mm and the average uniformity coefficient is 9.97. of the column to allow free drainage. The column was filled
The water retention curve (WRC) was measured using a with 120 cm of filtered tailings compacted to a dry density
pressure plate apparatus (ASTM-D6836-16 (2016)) and of 1585 kg/m3 and a void ratio of 0.78. The top of the
was described using the van Genuchten (1980) model column was exposed to the atmosphere and a fan was
(Figure 2). The tailings were compacted in the pressure installed to favor air circulation and increase evaporation.
plate rings to a void ratio of 0.77, similar to that measured Four Watermark model 200ss matric suction probes were
in the field (eaverage = 0.75 and e = [0.60 ; 0.96]). The field installed 10 cm, 20 cm, 60 cm, and 90 cm under the surface
saturated hydraulic conductivity was measured using a of the tailings and recorded suction automatically every 15
Guelph permeameter and was 7 x 10-7 m/s. Properties of minutes with a 1 kPa precision. The suctions measured by
tailings used in this study are typical of other gold mine the sensors were converted to degrees of saturation using
tailings (Bussière, 2007). the van Genuchten WRC fitted on the measured points in
the laboratory (see Figure 2). Temperature, relative
humidity, and potential evaporation in the laboratory were
measured for the duration of the tests.
The mass of the column was monitored
throughout the duration of the tests to assess the water
balance of the column. The actual evaporation (E) was
indirectly estimated by subtracting the measured drainage
(D) and the change in water storage (change in column
mass, ΔS) from the precipitation (mass of water added at
each cycle, P). There was no runoff because the surface of
the filtered tailings in the column was flat,

2.3 Wetting and drying cycles

The degree of saturation in the tailings was monitored for

160 days while the experimental column was exposed to a
total of 25 wetting and drying cycles under controlled
conditions. The laboratory temperature varied between
21.5 °C and 25 °C, the relative humidity was between 10%
and 62%, and the wind speed was deemed constant.
Figure 1: Grain size distribution of the tailings.
 The study was conducted in two phases: (1) Phase The degree of saturation increased rapidly when water was
1 includes cycles 1 to 4 (day 0 to 85) and corresponds to added (reaching a maximum between 94% and 100%),
high intensity precipitations at a low frequency and (2) corresponding to the infiltration front passage, and then
phase 2 includes cycles 8 to 24 (day 95 to 160) and decreased progressively due to the combined effect of
corresponds to moderate intensity precipitations at a high drainage and evaporation (Figure 3). The surface sensors
frequency (see Table 1). During phase 1, the volume of responded more quickly and strongly to infiltration than the
demineralized water added to the column at each cycle sensors located deeper, but the infiltration front penetrated
varied from 2L to 8L corresponding to extreme to at least 90 cm depth in the columns because a peak in
precipitations of 40 mm to 160 mm. The drying period the degree of saturation was observed at all depths after
lasted 14 to 31 days to allow the column to return to a each water addition. For example, the infiltration front
measured degree of saturation of approximately 75% at usually reached the first probe, located 10 cm below the
10 cm below the surface before the next wetting cycle. On surface, within 2 hours, while the infiltration front reached
the 3rd cycle, 0.5L were added twice a day during 2 days 90 cm below the surface up to 5 days after the addition of
for a total of 2L. During phase 2, for each cycle, 1L of water water (e.g., day 75 to 80). The degree of saturation
was added corresponding to a precipitation of 20 mm. The increased with depth and the amplitude of variation of
drying phase lasted from 1 to 7 days to reach a measured degree of saturation was greater closer to the surface
degree of saturation of approximately 90% at 10 cm below (10 cm below the surface), while the degree of saturation
the surface before the next cycle started. measured deeper varied less. For example, the degree of
saturation 10 cm below the surface varied from 100 % to
2.4 Numerical simulations and calibration 78% (22% difference), while the amplitude of variation of
degree of saturation at 20, 60 and 90 cm below the surface
SEEP/W 2021 (Geostudio Int.) was used to simulate the was 10%, 8% and 6%, respectively. The degree of
wetting and drainage cycles in the column. The column saturation increase rate and magnitude depended on the
was simulated in 1D as a 1.2 m high vertical domain. The quantity of water added.
mesh size was 0.01 m, the maximum number of iterations During phase 1, corresponding to the high
was 500 and the maximum pressure load difference was intensity cycles at low frequency (day 0 to 90), the evolution
0.005 m. Initial tailings properties were defined based on of the degree of saturation measured at 10 cm below the
laboratory characterization (see above). The initial surface showed three distinct phases and a distinct drying
conditions were a null total water head at the top of the slope. The drying slope at 10 cm below the surface was
tailings to impose an initial degree of saturation similar to characterized by two clear slopes created by the combined
that measured during column setup. The laboratory effect of drainage and evaporation (Fredlund et al., 2012;
conditions (temperature, relative humidity, solar radiation, Hillel, 1998; Miyazaki, 2006), while the drying slope was
and wind speed) and the volume of water added at each more continuous for deeper probes. Initially, the addition of
cycle (applied as a precipitation) were simulated as a land- water resulted in the degree of saturation reaching a
climate interaction (LCI) boundary conditions at the top of maximum value that was sustained for 1 to 5 days
the model. The duration of the precipitations was adjusted depending on the quantity of water added. Then, the
and varied from 1 hour for the 20 mm precipitations, to degree of saturation rapidly decreased during 3 to 4 days
2 days for the 80 mm precipitation, to avoid runoff and due to drainage and maximal evaporation rate. Finally, the
allow full infiltration. The albedo of dry tailings was set to transition to the third phase was marked by a change in
0.09. The software calculated the evaporation using the slope, resulting in a slower decrease in the degree of
Pennman-Wilson equation (Wilson et al., 1994). A free saturation where mainly evaporation occurred until the end
drainage boundary condition was imposed at the bottom of of the cycle. For example, during the third cycle, the degree
the column. of saturation at 10 cm below the surface initially peaked at
The numerical model was then calibrated by 100% from day 55 to day 57, then decreased rapidly from
comparing the measured and simulated degree of 100 % to 88% between day 57 and day 61, corresponding
saturation at 0.10, 0.20, 0.60 and 0.90 m depth (i.e., to a decrease of 3% per day. Following that, the degree of
location of sensors). The WRC and hydraulic conductivity saturation decreased at a slower rate from 88% to 78%,
function were adjusted iteratively to minimize the difference between day 61 and day 76, corresponding to a decrease
between the measured and simulated degree of saturation. of 0.7% per day. These three stages were also observed
The calibrated properties were very similar to measured on the 20 cm depth sensor, although it was not as
WRC (Figure 2) and saturated hydraulic conductivity. The pronounced. At depths greater than 20 cm, the drying rate
air entry value of the calibrated WRC was slightly smaller was relatively steady, and no marked change in slope were
than the measured one (1.1 m vs. 2.0 m for the measured observed. Therefore, the first 20 cm of depth seemed most
one) but the slope was nearly identical (ncalibrated = 1.8 vs. influenced by evaporation.
nmeasured = 1.9). The calibrated saturated hydraulic During phase 2, corresponding to the moderate
conductivity was also in the same order of magnitude intensity precipitations at high frequency (days 95 to 160),
(3 x 10-7 m/s vs. 8 x 10-7 m/s measured with Guelph the degree of saturation generally remained above 90 %.
permeameter). Tailings were saturated at 90 cm depth for the duration of
the cycles and only small variations (+/- 2%) were observed
3 RESULTS at 60 cm depth. The effect of evaporation stages was only
3.1 Evolution of the degree of saturation noticeable when the drying phase was longer than 4 days.
Figure 3: Evolution of the measured and simulated degree of saturation with time and the quantities of water added. Phase
1, from day 0 to 90, corresponds to high intensity precipitations at a low frequency and phase 2, from day 95 to 160,
corresponds to moderate intensity precipitations at a high frequency.

For example, a slight slope break is noticeable in the drying saturation at 20 and 40 cm depth were very similar (usually
phase of the cycles starting at 100 and 125 days. a 2% difference was measured), but up to 8% difference
The degree of saturation increased with depth. was noted in the model (e.g., at day 75, 82% modeled at
However, during phase 1 (day 0 to 90), there was a 20 cm depth vs. 90% at 60 cm depth). The model generally
significant difference between the measured degree of underestimated degree of saturation at 20 cm depth by up
saturation at 10 and 20 cm depth for a given time. For to 12% (day 32) and showed greater amplitude of variation
example, the increase of degree of saturation with depth than measured.
between 10 and 20 cm was 12% at day 75, but the increase Simulations during phase 2, (moderate intensity
was only 4% between 20 cm and 90 cm depth. During the cycles at high frequency, from day 95 to 160) were
phase 2 (day 95 to 160, moderate intensity cycles at high particularly close to measurements following both trends
frequency), the increase of degree of saturation with depth and amplitudes (< 3% difference). During this phase, the
was more gradual than for phase 1. model successfully reproduced the measured degree of
saturation at 20 cm below the surface with an error smaller
3.2 Numerical simulation calibration than 2%.

Laboratory column test results were used to calibrate 3.3 Water balance
numerical simulations conducted with Seep/W (Figure 3).
The numerical model simulated relatively well the general The main components of the water balance (precipitation,
hydrogeological behaviour measured in the experimental drainage, evaporation, and water storage) were monitored
columns with a good correlation between the measured throughout the duration of the tests (Figure 4 and Table 1).
and simulated degrees of saturation. Peaks in water storage were observed following every
The maximal and minimal degrees of saturation were well cycle (precipitation events), and the water storage
reproduced, and the difference between measurements progressively decreased during the drying periods. The
and simulations were usually less than 10%. The model highest water storage (698 mm) was measured after the
successfully simulated the drying period, particularly at 60 first cycle corresponding to a precipitation of 160 mm (day
and 90 cm depth during phase 1. The two distinctive drying 0). The lowest water storage (452 mm) was measured at
slopes, for both the drainage and the evaporation, at 10 cm the end of the drying period following the fourth cycle (day
depth were also well simulated. The arrival time of the 85). The maximal amplitude of variation of the water
infiltration front was generally well reproduced at all depths. storage was 246 mm difference between the lowest and
The amplitude of variation of the degree of saturation highest values.
between the top and bottom sensors for a given time was At the end of phase 1 (days 0 to 90, Table 1), the
well simulated (e.g., < 4% difference at day 89). The net infiltration from precipitation (312.18 mm) was entirely
vertical increase of the degree of saturation with depth lost, showing a negative difference in water storage change
simulated with the model was more gradual than the from day 0 to day 90 of -87.60 mm. The water was mostly
degree of saturation profiles measured in the columns. For released by evaporation (354.27 mm) although 45.51 mm
example, the difference between the degree of saturation were released by drainage during the first cycle. No
measured at 10 cm and 20 cm depth was usually 2% in the drainage water was collected for cycles 2,3 and 4.
model during phase 1, but up to 12% difference was During phase 2 (days 95 to 160), the net
measured in the laboratory. Also, the measured degrees of infiltration (338.16 mm) was also entirely lost to drainage
and evaporation, ending with a negative final water storage
change of -6.07 mm. During this phase, 74% of the
precipitations were released by evaporation, and 28% were
released by drainage. During this second phase of
moderate intensity cycles at high frequency, the 90 cm
depth sensor was saturated, which explains the increased
drainage water collected during that phase (93.00 mm vs.
43.74 mm during phase 1).
During phase 1 (day 0 to 90), the different stages
of evaporation can be observed in the cumulative
evaporation trend (Figure 4). The evaporation increased to
a relative maximal evaporation rate during the first days
following the cycle and decreased to a slower evaporation
rate. For example, during the third cycle, the evaporation
rate is high (approximately 6.7 mm/d) from day 54 to day
60, then from day 60 to day 75 the evaporation rate drops
to approximately 2.0 mm/d. During phase 2, water was Figure 4: Daily water balance of experimental column.
added regularly, so more water was available to evaporate, Phase 1 corresponded to high intensity precipitations at a
and the evaporation rate was relatively constant low frequency (day 0 to 90) and phase 2 corresponded to
(approximately 3.80 mm/d) from day 95 to day 160. moderate intensity precipitations at a high frequency (day
95 to 160).

Table 1 : Description of the duration of the cycle and the water balance components for each cycle. The water balance
components were calculated at the end of each cycle. The positive values of water storage change correspond to an
increase in water storage in the column and negative values correspond to a loss of water storage (by evaporation or
drainage).

Duration of
Phase Cycle Day P (mm) D (mm) ΔS (mm) E (mm)
cycle
1 0.00 31.33 152.22 43.74 -73.94 182.41
1. High intensity
2 32.76 22.04 36.81 0.00 -2.39 39.20
precipitations at
3 54.80 21.21 75.97 0.00 4.19 71.78
a low frequency
4 76.00 14.00 35.05 0.00 -12.06 47.11
case
Total phase 1 300.06 43.74 -84.20 340.51
5 90.04 1.96 37.60 0.00 28.67 8.93
6 92.00 1.13 37.63 0.00 30.59 7.05
7 93.13 2.62 37.87 0.25 20.13 17.49
8 95.76 1.04 18.45 2.13 8.73 7.59
9 96.80 3.34 18.37 9.54 -10.46 19.29
10 100.08 4.83 18.99 3.55 -9.20 24.65
11 104.97 2.13 18.52 0.16 9.09 9.27
12 107.08 3.92 18.64 5.08 2.19 11.37
13 111.00 3.00 19.11 4.66 1.21 13.24
14 114.00 4.04 18.88 5.36 -1.45 14.97
2. Moderate 15 118.04 2.92 18.84 4.25 3.09 11.50
intensity 16 120.93 3.04 18.68 5.44 3.41 9.83
precipitations at 17 123.97 6.87 18.80 5.52 -14.29 27.57
a high frequency 18 130.88 3.08 19.50 0.00 7.71 11.79
case 19 133.93 4.00 18.88 3.10 2.70 13.07
20 137.93 3.13 18.64 5.45 3.33 9.86
21 141.04 3.75 19.07 6.56 0.43 12.08
22 144.84 3.21 18.80 7.75 1.21 9.83
23 148.04 4.00 18.80 8.23 -1.33 11.90
24 152.00 2.00 18.52 5.92 5.21 7.40
25 154.00 7.88 18.68 10.30 -17.66 26.04
Total phase 2 338.16 93.00 -6.07 251.23
Total 751.31 136.99 -10.89 625.21
Note : P, precipitation; D, drainage; ΔS, water storage change; E, actual evaporation.
4 RESULTS ANALYSIS AND DISCUSSION Hanging Column, Pressure Extractor, Chilled Mirror
Hygrometer, or Centrifuge.
The hydrogeological behaviour of the column depended on https://compass.astm.org/document/?contentCode=ASTM
the initial water content. During phase 1, characterized by %7CD6836-16%7Cen-US
extreme precipitations with long drying periods and lower ASTM. (2017). ASTM-D1140-14 Determining the
initial water contents (approximately 80% of degree of Amount of Material Finer than 75-μm (No. 200) Sieve in
saturation) the total precipitations (300.06 mm) were Soils by Washing.
completely released by evaporation (340.51 mm) and very https://compass.astm.org/document/?contentCode=ASTM
little drainage (43.74 mm). Phase 1 conditions could be %7CD1140-14%7Cen-US
compared to drought periods during summer characterized ASTM. (2021). ASTM-D6913-04 Particle Size
by longer drying periods, increased evaporation and Distribution of Soils Using Sieve Analysis. Dans ASTM:
decrease in water storage. Vol. 04.09. https://www.astm.org/d6913-04r09.html
During phase 2, characterized by moderate ASTM. (2023). ASTM-D854-14 Specific Gravity of Soil
precipitations with very short drying periods and high initial Solids by Water Pycnometer.
water contents (approximately 90% of degree of https://compass.astm.org/document/?contentCode=ASTM
saturation), water balance is significantly different, %7CD0854-14%7Cen-US
although the cumulative precipitation is similar to phase 1 Aubertin, M., Bussière, B., & Bernier, L. (2003).
(300.06 mm for phase 1 vs. 338.16 mm for phase 2). 74% Environnement et gestion des rejets miniers. Presses
of the precipitations were released by evaporation and 28% internationales Polytechnique.
were released by drainage. Such a high initial water Azam, S., & Li, Q. (2010, juillet 20). Tailings Dam
content could be compared to the conditions at the end of Failures : A Review of the Last One Hundred Years.
winter when snow melts and increases degree of saturation https://www.semanticscholar.org/paper/Tailings-Dam-
as well as water table level. Then, a hydrogeological Failures%3A-A-Review-of-the-Last-One-Azam-
behaviour like the one observed during the second phase Li/e57ebdac0a801b412cefd42017c2dded29cafd41
could be expected: high water storage and increased water Bresson, É., Bussière, B., Pabst, T., Demers, I.,
drainage. Charron, P., & Roy, P. (2022). Climate change risks and
vulnerabilities during mining exploration, operations, and
5 CONLUSION reclamation : A regional approach for the mining sector in
Québec, Canada. CIM Journal, 13(2), 77‑96.
An experimental column was exposed to wetting and https://doi.org/10.1080/19236026.2022.2055706
drying cycles during 160 days to assess the Bussière, B. (2007). Colloquium 2004 :
hydrogeological behaviour of filtered tailings in controlled Hydrogeotechnical properties of hard rock tailings from
parameters. The monitored degree of saturation was used metal mines and emerging geoenvironmental disposal
to calibrate a 1D numerical model. approaches. Canadian Geotechnical Journal, 44(9),
The main conclusions of these preliminary results indicate: 1019‑1052.
• Calibrated results reproduced well the global Bussière, B., & Guittonny, M. (2020). Hard Rock Mine
laboratory trends. The model was particularly Reclamation : From Prediction to Management of Acid
close to measurements for the surface sensor (10 Mine Drainage. CRC Press.
cm depth, < 5% difference) and for the deepest Carneiro, A., & Fourie, A. (2020). Assessing the
sensor (90 cm depth, < 4 % difference.). impacts of uncertain future closure costs when evaluating
• Evaporation plays a key role in water storage strategies for tailings management. Journal of Cleaner
change and is thereby a critical parameter to Production, 247, 119173.
accurately simulate hydrogeological behaviour. https://doi.org/10.1016/j.jclepro.2019.119173
• Precipitation intensity and initial water content Fredlund, D. G., Rahardjo, H., & Fredlund, M. D.
affected the hydrogeological response of the (2012). Unsaturated soil mechanics in engineering practice
column, hence the importance of considering (1‑1 online resource). John Wiley & Sons. https://cornell-
climate conditions to evaluate long-term library.skillport.com/skillportfe/main.action?assetid=49783
geotechnical stability of a filtered TSF. Furnell, E., Bilaniuk, K., Goldbaum, M., Shoaib, M.,
Future works include the validation of the numerical Wani, O., Tian, X., Chen, Z., Boucher, D., & Bobicki, E. R.
calibration using experimental field data to eventually (2022). Dewatered and Stacked Mine Tailings : A Review.
develop a numerical model of the filtered TSF. ACS ES&T Engineering, 2(5), 728‑745.
https://doi.org/10.1021/acsestengg.1c00480
6 ACKNOWLEDGEMENTS Gunson, A. J., Klein, B., Veiga, M., & Dunbar, S. (2012).
Reducing mine water requirements. Journal of Cleaner
The authors acknowledge the financial support from the Production, 21(1), 71‑82.
partners of the Research Institute on Mines and the https://doi.org/10.1016/j.jclepro.2011.08.020
Environment. Hassani, F., Mantegh, A., Mohamed, A.-M., & Ouellet,
J. (2008). Evaluation of flow liquefaction in fine grained
7 REFERENCES tailings using elastic wave velocities. International Journal
of Geotechnical Engineering, 2(3), 169‑178.
ASTM. (2016). ASTM-D6836-16 Determination of the https://doi.org/10.3328/IJGE.2008.02.03.169-178
Soil Water Characteristic Curve for Desorption Using
 Hillel, D. (1998). Environmental soil physics. Academic
press, san diego. Environmental soil physics. Academic
Press, San Diego.
James, M., Aubertin, M., Wijewickreme, D., & Wilson,
G. W. (2011). A laboratory investigation of the dynamic
properties of tailings. Canadian Geotechnical Journal,
48(11), 1587‑1600. https://doi.org/10.1139/t11-060
Kossoff, D., Dubbin, W. E., Alfredsson, M., Edwards, S.
J., Macklin, M. G., & Hudson-Edwards, K. A. (2014). Mine
tailings dams : Characteristics, failure, environmental
impacts, and remediation. Applied Geochemistry, 51,
229‑245.
https://doi.org/10.1016/j.apgeochem.2014.09.010
Li, S., Chen, Q., & Wang, X. (2016). Superiority of
Filtered Tailings Storage Facility to Conventional Tailings
Impoundment in Southern Rainy Regions of China.
Sustainability, 8(11), Article 11.
https://doi.org/10.3390/su8111130
Lottermoser, B. (2007). Mine Wastes :
Characterization, Treatment and Environmental Impacts.
Springer.
Lyu, Z., Chai, J., Xu, Z., Qin, Y., & Cao, J. (2019). A
Comprehensive Review on Reasons for Tailings Dam
Failures Based on Case History. Advances in Civil
Engineering, 2019, e4159306.
https://doi.org/10.1155/2019/4159306
MEND. (2017). Study of Tailings Management
Technologies : MEND Report 2.50.1 (24th Annual BC
MEND Metal Leaching /Acid Rock Drainage Workshop).
MEND (Mine Environment Neutral Drainage); MAC
(Mineralogical Association Of Canada). http://mend-
nedem.org/mend-report/study-of-tailings-management-
technologies-2017/
Miyazaki, T. (2006). Water flow in soils (2nd ed, 1‑1
online resource (418 pages) : illustrations.). Taylor &
Francis. http://www.crcnetbase.com/isbn/9780824753252
Oldecop, L., & Rodari, G. (2021). Unsaturated mine
tailings disposal. Soils and Rocks, 44, 1‑12.
https://doi.org/10.28927/SR.2021.067421
Rico, M., Benito, G., & Díez-Herrero, A. (2008). Floods
from tailings dam failures. Journal of hazardous materials,
154, 79‑87. https://doi.org/10.1016/j.jhazmat.2007.09.110
Santamarina, J. C., Torres-Cruz, L. A., & Bachus, R. C.
(2019). Why coal ash and tailings dam disasters occur.
Science. https://doi.org/10.1126/science.aax1927
Skempton, A. W. (1960). The Pore-Pressure
Coefficient in Saturated Soils. Géotechnique, 10(4),
186‑187. https://doi.org/10.1680/geot.1960.10.4.186
Spitz, K., & Trudinger, J. (2019). Mining and the
Environment : From Ore to Metal (2e éd.). CRC Press.
https://doi.org/10.1201/9781351183666
Ulrich, B., Ajc, P., Fourie, A. B., & Reid, D. (2019).
Practical thoughts regarding filtered tailings (p. 71‑79).
Australian Centre for Geomechanics.
https://doi.org/10.36487/ACG_rep/1910_01_Ulrich
Vick, S. G. (1990). Planning, Design, and Analysis of
Tailings Dams. BiTech Publishers Ltd.
https://doi.org/10.14288/1.0394902
Wilson, G. W., Fredlund, D. G., & Barbour, S. L. (1994).
Coupled soil-atmosphere modelling for soil evaporation.
Canadian Geotechnical Journal, 31(2), 151‑161.
https://doi.org/10.1139/t94-021
Large-scale laboratory testing to evaluate
leakage through defects in a geomembrane
cover system
Erika L Erlandson & Ian R Fleming
Department of Civil, Geological, and Environmental Engineering –
University of Saskatchewan, Saskatoon, SK, Canada

ABSTRACT
This paper presents the findings of a series of large-scale laboratory trials with the purpose of evaluating the leakage rate
through a defect in a geomembrane used as a cover for a mine waste rock pile. Preliminary results were presented at
GeoCalgary 2022 for leakage rates through a defect in an exposed geomembrane cover. This paper provides follow-up
data for leakage rates through a defect in a geomembrane cover system, which included a drainage geocomposite and a
cover soil overlying the geomembrane. The cover system was tested in a large-scale hillslope apparatus that allowed for
control of slope angle and precipitation rate. The leakage rate was assessed for different sizes, shapes, and locations of
defects.

RÉSUMÉ
Cet article présente les résultats d'une série d'essais en laboratoire à grande échelle dans le but d'évaluer le taux de fuite
à travers un défaut dans une géomembrane utilisée comme couverture pour une halde de stériles miniers. Les résultats
préliminaires ont été présentés à GeoCalgary 2022 pour les taux de fuite à travers un défaut dans une couverture de
géomembrane exposée. Cet article fournit des données de supplémentaire pour les taux de fuite à travers un défaut dans
un système de couverture de géomembrane, qui comprenait un géocomposite de drainage et un sol de couverture
recouvrant la géomembrane. Le système de couverture a été testé dans un appareil de pente à grande échelle qui
permettait de contrôler l'angle de la pente et le taux de précipitation. Le taux de fuite a été évalué pour différentes tailles,
formes et emplacements de défauts.

1 INTRODUCTION This knowledge gap is partly responsible for a slower

adoption of geomembrane covers in the mining industry.
Geomembranes are commonly used to control fluid flow in The goal of this research was to begin to fill this
waste containment applications, due to their very low knowledge gap in order to provide engineers and
permeability. When free of defects, a geomembrane is communities with greater confidence in the effectiveness
effectively impermeable to water, and thus also of geomembrane covers over time. Preliminary results
impermeable to any contaminants carried through were presented at GeoCalgary 2022 (Erlandson and
advective flow. However, geomembranes are very rarely Fleming, 2022), focusing on an exposed geomembrane
completely free of defects in practice. Defects can occur cover. This paper expands on those results with additional
during manufacturing, transport, handling, and particularly data from a geomembrane and cover soil system.
during installation. Quality control measures greatly help to
limit the number of defects in an installed geomembrane
but often cannot eliminate them entirely. The goal of defect 2 BACKGROUND
leakage research is to quantify the amount of leakage that
could be expected through a geomembrane with a given While the purpose of a waste liner is to keep contaminants
number or type of defects. from flowing out of the containment system, a cover is
This type of research has been extensively conducted intended to keep water from flowing into the waste pile.
for geomembranes used as bottom liners, such as in Water flowing through tailings or waste rock piles will often
landfills or lagoons. In these cases, quantifying the leakage leach contaminants, such as heavy metals, out of the
is a fairly straightforward problem of fluid flow through a waste material and carry them into the environment.
defect on a flat surface under hydrostatic and saturated Another common problem in Canadian mines, and around
conditions. This can be assessed using Bernoulli flow or the world, is the oxidation of sulfide minerals when exposed
similar flow equations. When a geomembrane is used as a to water and oxygen. This creates an acidic condition
cover, several more variables are introduced. Cover known as acid rock drainage (ARD) (INAP, 2014).
defects are often on a sloped surface and subject to highly These environmental hazards are typically mitigated
transient flow conditions under the influence of precipitation during mine closure by capping the waste with a barrier to
and runoff. The soil below the defect is also usually keep out water and/or oxygen. This barrier might be a
unsaturated. These conditions create a complex and time- cover with a very low hydraulic conductivity, a
dependent flow regime over and through a defect in the geomembrane or other geosynthetic, or a soil-geosynthetic
geomembrane cover. Thus, the case of cover leakage has combination.
not been as widely studied as its base liner counterpart.
 Giroud and Bonaparte (1989a) found that an HDPE
geomembrane without any defects can have a hydraulic
conductivity as low as 10-15 m/s, compared to the 10-9 m/s
typically achieved for compacted clay liners (Benson et al.,
1994). Using a geomembrane in combination with a clay
liner can increase the liner effectiveness while decreasing
the amount of clay required. Sourcing and hauling clay is a
major cost of liner construction and can be cost prohibitive
if there is no suitable material available nearby.
However, it is difficult to ensure a geomembrane is
installed without any defects occurring during
manufacturing, transport, placement, seam welding, or
placement of cover material. Nosko and Touze-Foltz
(2000) surveyed over 300 sites of geomembrane failures
and found that 71.17% of geomembrane defects were
caused by stones within the protection layer, with a further
15.59% caused by heavy equipment. Giroud and
Bonaparte (1989a) focused on seam defects, as the most
frequent defect found by forensic analyses at the time.
Thanks to improved welding methods and better quality
control, McQuade and Needham (1999) found that the
frequency of defects at seams had greatly decreased. The
researchers found an average frequency of 4.2 holes per
hectare in 111 surveys, while those surveys with a Figure 1. Cover system profiles with a) exposed
thorough quality assurance program often found no holes geomembrane and b) cover soil, geomembrane, and
at all. drainage geocomposite
The expected leakage rate through defects in a
geomembrane liner has been extensively studied and
many equations have been proposed. Some of these The exposed geomembrane cover was chosen for
equations are purely empirical, while others are based on initial testing so that the geomembrane could be accessed
a combination of experimental results and analytical or easily, and so that the defect could be modified without the
numerical models. A selection of these equations are need to dig it up. The cover soil system was designed as a
summarized in Table 1. scaled-down version of a cover system design under
Many of these equations were developed to address consideration for use at an existing mine. Assessing the
the effects of the shape of the defect, the permeability of leakage rates for this design was an additional objective of
the material above and below the defect, or the contact the research.
conditions between the geomembrane and the surrounding The waste rock that was used was sourced from a
soil. A common thread through all of the equations in Table Canadian mine in order to better represent field conditions.
1 is the presence of an h or hw term – the head of liquid It was a variable material consisting of shale particles
above the defect. This is one of the time-dependent factors ranging from <1mm to >200mm. These particles were
that is difficult to define for a geomembrane cover, as there brittle and friable, breaking down easily when wetted. The
is generally not a constant head of liquid above the defect. material also contained harder limestone and other
Instead, the head is affected by the rainfall rate and the materials in pieces up to 0.6m in diameter. The same waste
microtopography of the geomembrane surface. rock was used as the cover material in the trials with a
Another factor that is not considered in these equations cover soil.
is the effect of a defect on a slope. Where slope is The geomembrane used for all of the trials was a
considered, it is only for the purpose of calculating the 1.5mm thick LLDPE material, textured on both sides, and
hydraulic head above the defect (Giroud et al, 1992). with a texture asperity of 16mil. A 1.5mm textured HDPE
To more accurately quantify the leakage through a geomembrane was also considered for use; however, the
defect in a geomembrane cover, the following research geomembrane material and thickness were not considered
attempts to consider as many of these factors as possible, the focus of these trials. Specific defects were cut into the
including defect size, shape, and location, slope angle, geomembrane covers using a rotary cutting tool.
rainfall rate, and surface microtopography. The drainage geocomposite used in the second set of
trials was manufactured by Afitex-Texel. It was made up of
perforated 25mm diameter minidrains placed at 0.5m
3 MATERIALS intervals between two nonwoven, needle-punched
geotextile sheets. The geotextile was intended to provide
Two cover profiles were considered: one exposed some protection to the geomembrane when the cover soil
geomembrane cover and one cover system with a was placed on top, as well as to help with drainage by
geomembrane, a drainage geocomposite, and a cover soil. carrying water to the minidrains.
Both of the covers were built in a laboratory apparatus as
described in section 4.1. The two cover systems are shown
in Figure 1.
Table 1. Existing equations for flow through a defect in a geomembrane

Application Equation Notes Source

𝜋𝜌ℎ𝑤 𝑑 4 Pipe flow through pinhole
Poiseuille’s 𝑄=
128η𝑇𝑠 leaks (hole diameter < GM
equation Giroud and
thickness)
Bonaparte,
𝑄 = 𝐶𝐵 𝛼√2𝑔ℎ𝑤 Orifice flow – maximum
Bernoulli’s 1989a
possible flow through a circular
equation
defect
𝑄 = 4𝑟0 ℎ𝑑 𝑘 Minimum possible flow,
Forchheimer’s Rowe and
assumes perfect contact with
equation Booker, 1998
soil
𝑄 = 0.976𝐶𝑞0 𝑑0.2 ℎ0.9 𝑘𝑠0.74 Circular defect
h < liner thickness Giroud et al,
𝑄 = 𝐶𝑞∞ 𝑏0.1 ℎ0.45 𝑘𝑠0.74 Defect of infinite length
1989 & Giroud,
𝑄 = 𝐶𝑞0 𝑖𝑎𝑣𝑔0 𝑎0.1 ℎ0.9 𝑘𝑠0.74 Circular defect
h > liner thickness 1997
𝑄 = 𝐶𝑞∞ 𝑖𝑎𝑣𝑔∞ 𝑏0.1 ℎ0.45 𝑘𝑠0.87 Defect of infinite length
Excellent contact 𝑄 = 0.7ℎ𝑤 𝑎0.1 𝑘𝑠0.87 Circular defect
ℎ𝑤 0.95 Giroud et al,
0.9 0.1 0.74
𝑄 = 1.15ℎ𝑤 𝑎 𝑘𝑠 [1 + 0.1 ( ) ] Circular defect
𝐻𝑠 1992 & Giroud,
Poor contact
ℎ𝑤 0.95 1997
𝑄 = 1.22ℎ𝑤0.45 0.1 0.87
𝑏 𝑘𝑠 [1 + 0.2 ( ) ] Defect of infinite length
𝐻𝑠
ℎ𝑤 + 𝐻𝑠
2
𝑄 = 𝜋𝑟0 𝑘𝑠 − 2𝜋𝑟0 𝜃𝛼[𝐴𝐼1 (𝛼𝑟0 ) − 𝐵𝐾1 (𝛼𝑟0 )] Circular defect
Analytical 𝐻𝑠 Touze-Foltz et
solutions ℎ𝑤 + 𝐻𝑠 2 ℎ𝑤 + 𝐻𝑠 al, 1999
𝑄𝐿 = 𝑘𝑠 {𝑏 + 𝑡𝑎𝑛ℎ [𝑐𝑜𝑠ℎ−1 ( )]} Two-dimensional defect
𝐻𝑠 𝛼 𝐻𝑠

4 METHODOLOGY 4.2 Design and Instrumentation

4.1 Large-Scale Laboratory Testing 4.2.1 Exposed Geomembrane Cover

Testing for this project took place in the Multi-purpOse The slopes within the test bed were designed and
Slope Testing (MOST) facility at the University of instrumented as a water balance problem:
Saskatchewan, which was equipped with the large scale
hillslope apparatus that was used for this project. The 𝑄𝑖𝑛 = 𝑄𝑜𝑢𝑡 + ∆𝑆𝑡𝑜𝑟𝑎𝑔𝑒 [1]
large-scale test bed consisted of a tri-axle dump trailer,
modified for slope testing by Pratt and McDonnell (2017). where Qin and Qout are water flow rates.
A full description of the modifications and advantages of The flow into the system came from precipitation
the trailer system is given in their paper. controlled by the rainfall simulator. This rate was measured
The slope angle for testing could be adjusted using the by two or three tipping bucket rain gauges placed on the
trailer’s hydraulic scissor lift. The interior dimensions of the slope surface.
trailer were 4.53m length, 2.08m width, and 1.1m depth. Flow out of the system was divided into surface flow
The test bed size was decreased for the second set of tests and baseflow for the exposed geomembrane. Surface flow
by lining the trailer walls with EPS foam. This was done to off of the exposed geomembrane was measured using an
reduce the total weight on the lifting mechanism. The orifice bucket measurement device. The orifice bucket
reduced dimensions were 4.04 m length and 1.88 m width. used the principles of orifice flow to determine the flow rate
This size allowed for testing at a larger scale than is at any head of liquid.
generally practicable in a laboratory setting while also
maintaining control of laboratory conditions. 𝑄 = 𝐶𝑑 𝐴√2𝑔ℎ [2]
To simulate precipitation on the slope, the facility was
equipped with two peristaltic pumps that supplied water to where A is the area of the orifice, h is the head above the
a rainfall simulator manifold suspended over the test bed. orifice, and Cd is the coefficient of discharge. The orifice
bucket was constructed with four different orifices, with
The manifold contained rows of hypodermic needles, each area sized to the expected runoff from a
spaced at approximately 75mm, that formed water predetermined rainfall intensity. The head in the orifice
droplets. The water supplied to the simulator was purified bucket was measured with an ultrasonic level sensor
using reverse osmosis to remove any minerals that may positioned above the tank.
precipitate and clog the manifold pipes. Baseflow was allowed to drain from the bottom layer of
waste rock. It was collected through an opening at the base
of the slope, as shown in Figure 2, and measured with a 4.2.2 Cover Soil System
tipping bucket.
The cover soil trials used the same water balance
approach, with a few modifications. The cover soil reduced
the surface flow rate from >90% of the precipitation rate to
less than 20%. The decreased surface flow was low
enough to be measured with a tipping bucket instead of the
orifice bucket.
Water that flowed through the cover soil and over the
top of the geomembrane was collected through the
drainage geocomposite. Flow was collected separately
from each minidrain and measured with tipping buckets.
The change in storage, measured by the weight change
of the trailer, was divided between the base soil and cover
soil. Six more 5TM moisture sensors were installed in the
cover soil to differentiate the moisture content changes
between the two layers.

4.3 Defect Trials

Figure 2. Instrumentation for exposed geomembrane cover The trials conducted on the hillslope were intended to test
the effects of defect size, shape, and location, slope angle,
and rainfall intensity on the resulting leakage rate.

a)

b)

Figure 3. Instrumentation for system with cover soil c)

The change in storage within the waste rock was

measured using four low-profile load cells. The trailer was
raised off of the ground and supported by the load cells,
allowing the weight of the trailer to be measured to within
approximately 1.5 kg, or 1.5L of water added (0.03% of
10,000lb capacity). The amount of water that accumulated d)
in the waste rock was verified using six 5TM moisture and
temperature sensors, placed in a 2x3 grid within the waste
rock.
Together, all of these measured flows were used to
confirm the water balance of the cover system. The
leakage rates for these trials were primarily calculated from
the measured change in storage as water flowed through Figure 4. a) 150mm horizontal defect, b) 300mm horizontal
the defect. The baseflow was also added to this rate to defect, c) 500mm horizontal defect, and d) 20mm and
account for the water leaving the mine waste rock at the 40mm circular defects
base.
a) b) c)

Figure 5. Effects of defect size and shape for a) horizontal tears, b) circular holes, and c) vertical tears

a) b) c)

Figure 6. Effect of location for a) 150mm, b) 300mm, and c) 500mm tears

a) b)

Figure 7. Effect of slope angle for a) 3:1 slope and b) 10:1 slope
Figure 8. Leakage rates for horizontal defects with cover Figure 9. Leakage rates for circular and vertical defects
soil and drainage geocomposite with cover soil and drainage geocomposite

Three defect shapes were tested: a horizontal tear, a The leakage rate was assessed at two different slope
vertical tear, and a circular hole. Three sizes of horizontal angles of 10:1 and 3.5:1 for the exposed geomembrane.
tear were tested – 150mm, 300mm, and 500mm – and two The 10:1 slope was selected to be nearly flat, while still
sizes of circular hole – 20mm and 40mm diameters. The allowing flow down the slope. The 3.5:1 slope was chosen
vertical tear was 300mm in length, for a total of six defect to reflect the maximum slope a mine waste pile could
types. Examples of the defect types are shown in Figure 4. typically be built at in the field. An intermediate 6:1 slope
Defects were tested in two locations for the exposed was also tested for the cover soil system trials.
geomembrane cover: upstream and downstream on the The final variable tested was rainfall intensity. The
slope. The upstream defects were located approximately rainfall rates and durations that were used were
1m from the top of the hillslope and the downstream approximately 2mm/hr for 16 hours, 10mm/hr for 4 hours,
defects were between 3m and 3.5m from the top of the and 20mm/hr for 2 hours. For some of the earlier tests, only
slope. a high (~15mm/hr) and low (~2mm/hr) rainfall rate were
The microtopographic location was also considered for used. The rainfall intensities and test durations for the
the downstream defects. The waste rock surface was exposed geomembrane were based on the limitations of
smoothed and compacted with multiple passes of a hand- the rainfall simulator and a typical intensity-duration-
pushed vibrating plate compactor before the cover was frequency curve. The same rainfall intensities were used
placed. Nevertheless, the geomembrane surface was for the cover soil system, but the tests were run until a
found to exhibit microtopographic highs and lows. The steady state was reached, rather than for a predetermined
microtopographic elevation differences were typically less duration.
than 1cm and occurred over rocks or larger particles that
protruded from the soil surface. Different locations were 5 RESULTS AND DISCUSSION
tested over this microtopography to determine how
significantly it affected the surface flow regime and 5.1 Exposed Geomembrane Cover
resulting leakage rates. The three defect locations that
were tested were downstream microtopographic high, Leakage rates for the exposed geomembrane cover are
downstream microtopographic low, and upstream shown in Figures 5 to 7. The results were discussed in
microtopographic low. detail in the paper presented at GeoCalgary 2022
Defect location and microtopography were not (Erlandson and Fleming, 2022). The main observations are
assessed for the cover soil system due to the difficulty of summarized as follows:
creating new defects once the cover soil was in place.
 • Larger defects received more leakage. velocity, increasing the contact time between the flow and
• The three defect shapes received approximately the defects, and creating the same effect as lowering the
the same amount of leakage. slope in the exposed geomembrane trials.
• Downstream defects received more leakage than
the same defect located upstream, due to
increased accumulation of surface runoff. 6 CONCLUSION
• Defects in microtopographic lows received
noticeably more leakage than those located in Large-scale laboratory testing was used to quantify the
microtopographic highs. leakage received through defects in a geomembrane mine
• Defects on lower slopes (10:1) received more waste cover. The geomembrane was tested as an exposed
leakage than those on steeper slopes (3.5:1) due cover, and as part of a cover soil and drainage
to more contact time between the flow and the geocomposite system. The leakage rate through a defect
defect. was assessed based on defect size and shape, defect
• Higher rainfall intensities resulted in higher location, surface microtopography, slope angle, and rainfall
leakage rates, with a possible saturation point intensity. All of these factors influenced the leakage rate to
being reached where a given defect could not some degree. The microtopography had a significant effect
take on any more leakage, even with increased on the exposed geomembrane, while the cover soil
rainfall. reduced the effects of slope for the cover soil system. In
both cases, the flow paths, or lack thereof, were identified
Flow paths, or the paths taken by surface flow as it as a significant factor in quantifying the leakage rates.
traveled down the slope, were one of most significant
indicators of whether a given defect would receive more or
less leakage. Flow paths on the 10:1 slope generally
followed the microtopographic low points on the 7 ACKNOWLEDGEMENTS
geomembrane surface. These paths were winding, the flow
velocity was low, and there was often pooling in any The authors would like to acknowledge the financial
particularly low points. assistance and materials provided by Solmax as well as
The effect of microtopography was reduced on the assistance from the team at the MOST Facility. This
3.5:1 slope because the flow paths tended to travel straight research was partially funded through a NSERC
down the slope at a higher velocity without winding around Collaborative Research and Development Grant.
the microtopography. These straight flow paths are also the
reason for the very high leakage rate in some of the 500mm
long horizontal defects. With a 500mm defect taking up 8 REFERENCES
25% of the slope width, it was almost guaranteed to
intersect with at least one flow path, regardless of whether Benson, Craig, Huaming Zhai, and Xiaodong Wang. 1994.
it was intended to be in a microtopographic high or low. Estimating Hydraulic Conductivity of Compacted Clay
Liners. Journal of Geotechnical Engineering; 120:2
5.2 Cover Soil System (February).
Erlandson, Erika L. and Ian R Fleming. 2022. Large-scale
Leakage rates for the cover soil system trials are shown in laboratory testing of the performance of geosynthetic
Figures 8 and 9. Some of the trends observed in the mine waste covers. GeoClagary 2022. Canadian
exposed geomembrane cover remained consistent for the Geotechnical Society. Calgary, Canada.
cover soil system. Larger defects and increased rainfall Giroud, J. P., and R. Bonaparte. 1989a. Leakage through
resulted in higher leakage rates, with a similar saturation Liners Constructed with Geomembranes—Part I.
point suggested by a decreasing slope in some of the trials. Geomembrane Liners. Geotextiles and
Other trends were not as evident in the cover soil system Geomembranes 8 (1): 27–67.
trials, such as the effect of slope. Some of the trials showed Giroud, J., A. Khatami, and K Badu-Tweneboah. 1989.
decreased flow on a steeper slope, while others showed Evaluation of the Rate of Leakage through Composite
the opposite. Liners. Geotextiles and Geomembranes, 8(4), 337-340.
Many of the changes in the leakage rate trends may be Giroud, J., K. Badu-Tweneboah, and R. Bonaparte. 1992.
explained by the lack of flow paths, or rather by the cover Rate of Leakage Through a Composite Liner due to
soil acting as one large flow path. Rainfall trials were run Geomembrane Defects. Geotextiles and
until a steady state was reached, resulting in the flow Geomembranes, 11(1), 1-28.
moving at a constant rate through the cover soil. This rate Giroud, J.P. 1997. Equations for Calculating the Rate of
would have been largely dependent on the cover soil’s Liquid Migration Through Composite Liners Due to
characteristics, rather than the geometry of the slope or the Geomembrane Defects. Geosynthetics International 4
defect. As the cover soil became saturated, it may have (3–4): 335–48.
delivered the same amount of flow to a defect, regardless INAP. 2014. Global Acid Rock Drainage Guide.
of where it was located on the slope. International Network for Acid Prevention. Retrieved
The cover soil system also resulted in greater overall September 29, 2020 from http://gardguide.com.
leakage rates compared to the exposed geomembrane.
This may be due to the cover soil slowing down the flow
McQuade, S. J., and A. D. Needham. 1999. Geomembrane
Liner Defects—Causes, Frequency and Avoidance.
Proceedings of the Institution of Civil Engineers -
Geotechnical Engineering 137 (4): 203–13.
Nosko, Vladimir, and Nathalie Touze-Foltz. 2000.
Geomembrane Liner Failure: Modelling of Its Influence
on Contaminant Transfer. EuroGeo 2. Italian
Geotechnical Society. Bologna, Italy.
Rowe, R. K., and J. R. Booker. 1998. Theoretical Solutions
for Calculating Leakage through Composite Liner
Systems. In Geotechnical Research Centre Report,
739–88. Kluwer Academic Publishers.
Touze-Foltz, N., R.K. Rowe, and C. Duquennoi. 1999.
Liquid Flow Through Composite Liners Due to
Geomembrane Defects: Analytical Solutions for Axi-
Symmetric and Two-Dimensional Problems.
Geosynthetics International 6 (6): 455–79.
Revisiting Large Scale Physical Models for
Biogeochemical Studies
Louis K. Kabwe & G. Ward Wilson,
Department of Civil and Environmental Engineering, University of
Alberta, Edmonton, AB, Canada

ABSTRACT
Large-scale physical models also referred to as mesocosms, such as the 2.4 m diameter x 4.6 m high cylinder at Environment
Canada’s Innovation Place in Saskatoon, are of sufficient size that relevant chemical, biogeochemical processes are active
and thus permitted field behavior under controlled conditions. The 4.6 m high cylinder was filled with 65 t of C-horizon sand
excavated at a field site located 10 km south of Saskatoon and was monitored for seven years. Due to a lack of maintenance
and application of water to the surface, the top 2 m of the soil was desiccated. Today, several institutions have set up their own
large-scale physical models, and thus the 4.6 m high cylinder could have served as the basis to evaluate the performance of
newer large-scale physical models. This paper revisits the CO2 fluxes measured from the 4.6 m high cylinder from day 2254
to day 2604 and a waste rock pile at the Key Lake uranium mine from 2000 to 2002, using the dynamic closed chamber system
DCC.

RÉSUMÉ
Les modèles physiques à grande échelle, également appelés mésocosmes, comme le cylindre de 2,4 m de diamètre et de
4,6 m de hauteur à la Place d’Innovation d'Environnement Canada à Saskatoon, sont d'une taille suffisante pour que les
processus chimiques et biogéochimiques pertinents soient actifs et permettent ainsi un comportement de terrain dans des
conditions contrôlées. Le cylindre de 4,6 m de haut a été rempli de 65 t de sable de l'horizon C extrait sur un site situé à 10 km
au sud de Saskatoon et a été surveillé pendant sept ans. En raison d'un manque d'entretien et d'application d'eau à la surface,
les 2 mètres supérieurs du sol ont été desséchés. Aujourd'hui, plusieurs institutions ont mis en place leurs propres modèles
physiques à grande échelle, et ainsi le cylindre de 4,6 m de haut aurait pu servir de base pour évaluer les performances des
nouveaux modèles physiques à grande échelle. Cet article revisite les flux de CO 2 mesurés à partir du cylindre de 4,6 m de
haut du jour 2254 au jour 2604 et d'une pile de stériles à la mine d'uranium de Key Lake de 2000 à 2002, en utilisant le système
dynamique à chambre fermée DCC.

1 INTRODUCTION to verify the CO2 flux measurements using the dynamic

closed chamber system DCC (Kabwe et al. 2000; 2002 and
Large-scale physical models also referred to as 2005). The results of the moisture content measurements
mesocosms, have been widely used in aquatic and marine conducted in the mesocosm on day 3894 indicated that the
research (Caquet et al. 2000; Carboni et al. 2021; Sharma upper 2 m of the soil was desiccated, thus, no further
et al. 2021), in the study of carbon cycling and waste studies were undertaken.
disposal in the subsurface (Lawrence et al. 1993; Hendry
et al. 2001; Hendry et al. 2002; Kabwe et al. 2002).
Mesocosm experiments can bridge the gap between 2 MATERIAL AND METHODS
laboratory experiments and field experiments. The
mesocosm, such as the 2.4 m diameter x 4.6 m high The mesocosm used was a 2.4 m dia x 4.6 m high cylinder
cylinder at Environment Canada’s Innovation Place in filled with 65 t of C-horizon sand (Figure 1) from
Saskatoon, is of sufficient size that relevant physical, Environment Canada’s Innovation Place in Saskatoon.
chemical, and biogeochemical processes are active and The methods of excavating the sand and filling the
thus permit field behavior under controlled conditions mesocosm as well as the installation of the
(Hendry et al., 2001; Kabwe et al. 2002). This 4.6 m high instrumentation, are described in Hendry et al. (2001). The
mesocosm was at steady-state with respect to gases O2 water table in the mesocosm was lowered from the ground
and CO2 concentrations and moisture content for over surface to a depth of 3.2 m below surface on day 1
2600 days (7.1 years) (Kabwe et al. 2002), and thus (December 28, 1992). The temperature of the mesocosm
provided the opportunity to verify a newly developed was held at room temperature (18-23 oC) from Day 1. A
technique to measure CO2 gas fluxes over the range uniform, controlled volume of distilled water was applied to
reported for field conditions (50–750 mg m2/h) (Kabwe the surface on a weekly basis (30 l/week), 6.63 mm/week)
2001);). This paper re-visits the research study conducted using a rain simulator system. Gravimetric water content
on the 4.6 m high cylinder over a period of 283 days (day measurements were made through a neutron access tube
2225 to day 2535) and on Deilmann south waste rock (0.05 m dia.Schedule 40 PVC pipe) installed in the center
(DSWR) pile at the Key Lake uranium mine (2000–2002) of the mesocosm to a depth of 3.7 m prior to lowering the
water table (Hendry et al. 2001). Measurements were bundles of gas ports (west and east profiles) installed at
made using a neutron moisture probe Model 3322 Troxler depths of 0.2, 0.47, 0.97, 1.47, 1.99, and 2.52 m below the
(Electronic Labs Inc. (Kabwe et al. 2002). Gravimetric surface in the mesocosm. All gas samples were analyzed
water content was converted to volumetric water contents for CO2 and O2 on a Carle Special Series S model 311
using dry density measurements and a calibration curve. analytical gas chromatograph (Kabwe et al. 2002).
Mesocosm gas samples were collected from duplicate

Figure 1. Photograph and diagram of the 4.6 m high mesocosm, Environment Canada’s Innovation Place in Saskatoon

CO2 gas fluxes were measured in a chamber of Plexiglas

collar (0.28 m x 0.07 m high, and 0.005 m thick) and a lid
using a ADC 2250 differential infrared CO2 gas analyzer
(ADC BioScientic Ltd) (Kabwe et al. 2002) (Figure 2). This
technique, termed the dynamic closed chamber DCC
ADC2250 method, is based on a direct measurement of the change
in CO2 concentration with time in the headspace of a
DCC chamber installed on the ground surface.

Alkali trap

Figure 2. ADC 2250 CO2 gas analyzer, DCC chamber

and alkali traps on top of the mesocosm ground surface
3 RESULTS AND DISCUSSIONS studies performed between day 1000 and day 2254. The
presence of CO2 gas concentrations in the mesocosm was
3.1 Steady-State CO2 Concentration Profiles in the attributed to microbial activity in the C-horizon sand
Mesocosm (Hendry et al., 2001). Results show that from day 1 to day
200 the CO2 concentrations greatly exceeded field
Figures 3 and 4 present the results of temporal variations conditions. This was due to the disturbance of soil
in CO2 gas concentration analyses on samples from the introduced during the filling of the mesocosm. The CO2 gas
gas ports from day 1 to day 2600 in the mesocosm (Hendry concentrations remained about the same (near steady-
et al. 2001 and Kabwe et al. 2002). The first study was state) after day 200 for the entire duration of the studies.
carried out between day 1 and day 1000 (Hendry et al. The observed CO2 concentration variations were attributed
2001), and the third study was carried out by Kabwe et al. to seasonal temperature changes.
(2002) between day 2254 and day 2600. There were no

3.5
CO2 Eastside
3.0
Study 3
2.5

2.0
% CO2

1.5

1.0

0.5

0.0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Days (from start)
0.22 m. 0.48 m. 0.98 m. 1.48 m. 2.00 m. 2.53 m.

Figure 3. Temporal variation of CO2 gas concentrations in the 4.6 high mesocosm, east side

3.5 CO2 Westside

3.0
Study 3
2.5

2.0
% CO2

1.5

1.0

0.5

0.0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600
Days (from start)

0.30 m 0.56 m 1.06 m 1.56 m 2.08 m 2.61 m

Figure 4. Temporal variation of CO2 gas concentrations in the 4.6 high mesocosm, west side

The CO2 concentrations from the second study (day 2254 during the 281 days of the study period. Statistical analysis
to day 2535), from the east and west sides, ranged from of the CO2 data (n=64) from the duplicate sets of gas ports
0.04 % (0.22 m) to 2% (2.53m) and from 0.084 % (0.30 m) in the mesocosm (1.5 m apart in an east-west direction)
to 2.26 % (3 m), respectively. Standard deviations in CO2 showed that there was no lateral variability in CO2
concentrations from the east and west sides ranged from concentrations. Figure 5 presents the averages of CO2
0.013 to 0.12% and from 0.018 to 0.21%, respectively. The concentrations profiles in the mesocosm from day 2254 to
small standard deviations suggest that the CO2 2535. The CO2 concentrations increased with depth,
concentrations were at or very near steady-state conditions reaching the greatest concentrations at the capillary fringe.
The average CO2 concentrations ranged from 0.04 (0.22) Vol. moisture content (%)
to about 2% (2.5 m).
0 5 10 15 20 25 30 35 40
0.0
Avg %CO2 2289
0.0 0.5 1.0 1.5 2.0 0.5 2326
0.0 2358
1.0 2414
2442
0.5 2473
1.5
2510
1.0 2.0 2546

Depth (m)
2573
1.5 West side 2.5 2600
2632
Depth (m)

East side 3.0 2657

2.0
2689
3.5 3894
2.5
4.0
3.0
4.5
3.5
Figure 6. Moisture content profiles in the 4.6 m high
4.0 mesocosm from day 2289 to day 3894

3.3 Temperature Profiles in the Mesocosm

Figure 5. Averages CO2 concentrations profiles in the
4.6 m high mesocosm Results of the measurements of the temperature profiles in
the mesocosm from day 2289 to day 2689 are presented
in Figure 7. The mesocosm was kept at room temperature
between 20 and 22 oC. The temperature decreases with
3.2 Steady-State Moisture Content Profiles in the increasing depth in the upper 1.5 m of the mesocosm and
Mesocosm remain constant at depths deeper than 2 m. The
fluctuations in temperature in the upper zone of the
The results of the moisture content profiles are presented mesocosm were attributed to seasonal temperature
in Figure 6. The moisture content increased with increasing changes.
depth except from 1 to 2 m depth. In this zone, the water
content remained consistently high, which was attributed Temperature oC
to heterogeneity introduced during filling. The same trend 12 14 16 18 20 22 24 26 28
was observed by Hendry et al. 2001. This strong similarity 0.0
between these data sets showed that the mesocosm was day 2289
at a steady state with respect to moisture content for over 0.5 2326
2689 days. Figure 6 also shows the results of the moisture
content profile measured on day 3894 (10.7 years). 1.0 2358
Results of the moisture content showed that the upper 2 m
of the soil was desiccated (i.e., 0 % moisture content) for a 2442
1.5
lack of water application to the surface of the soil, and thus, 2473
no further studies were undertaken.
Depth (m)

2.0 2510
2.5 2546
2573
3.0
2600
3.5
2632
4.0 2657

4.5 2689

Figure 7. Temperature profiles in the 4.6 m high mesocosm

from day 2289 to day 2689
3.4 CO2 Fluxes Under Steady-State CO2 Gas mg/m2/h; n=21) (Figure 8). The DCC and concentration
Concentrations in the Mesocosm gradient methods (results not shown here) (Kabwe et al.
2002) yielded comparable CO2 fluxes of 296 and 312 mg
Soil respiration rates from the mesocosm under steady- CO2/m2/h, respectively. The static chamber (alkali traps)
state CO2 gas concentrations were determined using the yielded a relatively high flux value of 269 mgCO2/m2/h for
optimized DCC (Kabwe 2001; Kabwe et al. 2000) that an adsorption time of 1 h (Figure 10). This value is similar
yielded a mean CO2 flux value of 296 mg CO2/m2/h (sd. 20 to those obtained with the other two methods.

400
alkali
350 DCCS1
DCCS2
300 DCCS3
Flux (mg CO2/m2/h)

250

200 DCCS

150 Alk.

100

50

0
0.1 1 10 100 1000 10000
Log time (min)

Figure 8. CO2 fluxes measured from the 4.6 high mesocosm using the DCC and alkali trap methods

The CO2 fluxes determined using alkali traps were the DSWR pile are presented in Birkham et al. (2003). The
sensitive to adsorption times, exhibiting a power decrease DCC provided the opportunity to quantify temporal and
with time (Figure 8). Fluxes from the mesocosm decreased spatial CO2 fluxes under field conditions and, to compare
from 267 mg CO2/m2/h for adsorption times of 1 h to these flux measurements to those obtained from the eddy
minimum values of about 24 mg CO2/m2/h for an covariance (EC) method (Kabwe et al. 2005). Figure 9
adsorption time of 110 h. presents the depth profiles for DSWR: (a) geologic profile;
(b) mean volumetric moisture contents; (c) O2; and (d) CO2
3.5 CO2 Fluxes Under Steady-State Gas Concentrations values (Birkham et al. 2003). In panels b-d, the symbols
at Deilman South Waste Rock Pile represent the mean measured values, and the horizontal
lines represent one standard deviation. O2 and CO2
The laboratory-verified DCC technique was used to concentrations were measured seasonally between 1997
measure the CO2 efflux at a field site, where conditions and 1999 using a portable gas analyzer (GasTech CO2
were similar to those used in the mesocosm tests. The series). The accuracy of the measurements was 10 vol %
study site consisted of a large, flat-topped, Deilmann south for CO2. The trends in O2 and CO2-depth profiles were
waste rock pile (DSWR) at the Key Lake uranium mine, in stable over time (Figure 9 c-d), suggesting near steady-
northern Saskatchewan (Kabwe et al. 2005). The pile was state conditions with respect to gas concentrations and,
constructed of uniform, fine to medium-grain sand and was thus, reaction rates (Birkham et al. (2003).
devoid of a soil zone and vegetative cover. Full details of
Figure 9. (a) Geologic profile, (b) moisture content, (c) O2, and (d) CO2 concentrations profiles at DSWR pile (Birkham et al.
2003)

The DCC technique (Figure 12) was used to characterize of 171 mg CO2 m2/h measured between July and
the spatial variability associated with the CO2 flux; whereas September 2000 and a time-averaged CO2 flux of about
the EC method (Figure 12) was used to characterize 150 mg CO2 m2/h between July and August 2002 (Kabwe
temporal variability on DSWR. Results showed that the et al. 2005). Spatial variability associated with the flux was
DCC method (Figure 10) yielded a time-averaged CO2 flux relatively small.

250
DCC - DSWR
200
Jul-00 Aug-00
Flux, mgCO2/m2/h (avg.)

150
Sep-00 Jul-02
100

50 Aug-02

0
2000-09-01
2000-07-01

2000-11-01

2001-01-01

2001-03-01

2001-05-01

2001-07-01

2001-09-01

2001-11-01

2002-01-01

2002-03-01

2002-05-01

2002-07-01

2002-09-01

Date
Figure 10. Spatial and temporal averages CO2 fluxes measured with DCC in 2000 and 2002 at the DSWR

Fluxes measured using the EC method (Figure 11) tended temporal resolution provided by the EC method
to underestimate those measured using the DCC method, demonstrated that the CO2 flux exhibited a distinct diurnal
with a time-averaged CO2 flux (150 mg CO2 m2/ h) that was pattern, but that long-term variations in the flux were
about 12% lower than the average flux calculated from the relatively small (CVZ23%).
DDC data in 2002 (Kabwe et al. 2005). The greater
 250

200

CO2 flux, mg/m2/h

150

100

50

0
6-30 7-3 7-6 7-9 7-12 7-15 7-18 7-21 7-24 7-27 7-30

EC DSWR July 2002 Date July, 2002

Figure 11. Temporal CO2 fluxes measured using EC method in July 2002 at the DSWR

Figure 12. CO2 flux measurements at DSWR using dynamic close chamber DDC and eddy covariance EC methods
5 SUMMARY AND CONCLUSIONS Carboni, A., Slomberg, D.L, Nassar, M., Santaella, C.,
Masion, A., Rose, J., Auffan, M. (2021). Aquatic
The steady-state mesocosm with respect to CO2 Mesocosm Strategies for the Environmental Fate and
concentrations and moisture content provided the Risk Assessment of Engineered Nanomaterials.
opportunity to verify the DDC method over the range Environ Sci Technol. 2021 Dec 21;55(24):16270-
reported for field conditions (50–750 mg CO2 m2/h). This 16282. doi: 10.1021/acs.est.1c02221. Epub 2021 Dec
large-scale mesocosm displayed active relevant chemical 2 .PMID: 34854667.
and biogeochemical processes and thus permitted field Hendry, M.J., Mendoza, C.A., Kirkland, R., and Lawrence,
behavior under controlled conditions. The DCC technique J.R. (2001) An assessment of a mesocosm approach
was used to quantify temporal and spatial CO2 fluxes under to the study of microbial respiration in a sandy
field conditions at DSWR and, at the same time, compare unsaturated zone. Ground Water 39, 391–400.
these flux measurements to those obtained from the EC
method. Results of CO2 measurements from the Hendry, M.J., Wassenaar, L.I., Birkham, T.K. (2002).
mesocosm indicated that the DCC and concentration
Microbial respiration and diffusive transport of O2,
gradient methods yielded comparable CO2 fluxes of 296
16O2, and 18O16O in unsaturated soils: a mesocosm
and 312 mg CO2/m2/h, respectively. The static chambers
(alkali traps) yielded a comparable flux value of 269 experiment, Geochimica et Cosmochimica Acta,
mgCO2/m2/h for an adsorption time of 1 h. Field 10.1016/S0016-7037(02)00949-3, 66, 19, (3367-
measurements results for 2000 and 2002 showed that the 3374).
DCC method yielded time-averaged CO2 fluxes of 171 mg Kabwe, L.K., Farrell, R.E., Carey, S.K., Hendry, M.J., and
CO2 m2/h and 150 mg CO2/m2/h, respectively. The EC Wilson, G.W. (2005). Characterizing Spatial and
method yielded a time-averaged CO2 flux of about 150 mg Temporal Variations in CO2 Fluxes From Ground
CO2/m2/h. Fluxes measured using the EC method tended Surface Using Three Complimentary Measurement
to underestimate those measured using the DCC method, Techniques. J. Hydrol., 311: 80-90.
that was about 12% lower than the average flux measured Kabwe, K.L., Hendry, M.J., Wilson, G.W., Lawrence, J.R.,
with the DCC method. Based on the laboratory and field (2002). Quantifying CO2 fluxes from soil surfaces to the
tests reported in this paper, the DCC method is suitable for atmosphere. J. Hydrol. 260, 1–14.
research to characterize the temporal and spatial dynamics
Kabwe L. K. (2001) Design and application of a dynamic
of soil CO2 fluxes. The DCC method can be used to
measure the CO2 fluxes in situ at the same locations using closed chamber system for measuring CO2 fluxes from
the same chambers with minimum disturbance of the soil. unsaturated C-horizon soils and waste-rock piles to the
When used together, the DCC and EC methods provide atmosphere. M.Sc. thesis. University of Saskatchewan,
information of a complementary nature that can greatly Canada
improve our understanding of the processes at work in Kabwe, L.K., Hendry, M.J., Wilson, G.W., and Kirkland,
unsaturated media. These fluxes will provide needed upper R.A. (2000); Quantification of the CO2 fluxes from the
boundary conditions for mathematical modeling of CO2 gas surface of a minicosm to the atmosphere. Proceedings
production and redistribution in the piles. of the 53rd Canadian Geotechnical Conference,
Montreal, Vol. 1: 473-479.
. Lawrence J.R., Zanyk B.N., Wolfaardt, G.M., Hendry,
Acknowledgements M.J., Robarts, R. D., and Caldwell D. E. (1993). Design
and evaluation of a mesoscale model vadose zone and
Cogema Resources Ltd., Cameco Corporation and the
ground water system. Ground Water 31, 446–455.
Natural Sciences and Engineering Research Council of
Canada (NSERC) funded this work through an NSERC- Sharma KV, Sarvalingam BK, Marigoudar SR. 2021. A
CRD. review of mesocosm experiments on heavy metals in
Portions of this data were obtained as part of L.K. Kabwe’s marine environment and related issues of emerging
M.Sc and PhD research. Eddy Covariance EC data were concerns..Environ Sci Pollut Res Int. 2021
provided by S.K. Carey. M.J. Hendry provided technical Jan;28(2):1304-1316. doi: 10.1007/s11356-020-
assistance. Environment Canada is also acknowledged for 11121-3. Epub 2020 Oct 20.
facility support.

References

Birkham, T.K., Hendry, M.J., Wassenaar, L.I., Mendoza,

C.A., Lee, E.S., 2003. Characterizing geochemical
reactions in unsaturated mine waste-rock piles using
gaseous O2, CO2, 12CO2, and 13CO2. Environ. Sci.
Technol. 37, 496–501.
Caquet, T., Lagadic, L., Sheffield, S.R. (2000). Mesocosms
in ecotoxicology (1): Outdoor aquatic systems. Rev
Environ Contam Toxicol.;165:1-38. doi: 10.1007/978-1-
4612-1172-3_1.
Improving reactive waste rock valorization
potential using mechanical separation
Karine Sylvain & Thomas Pabst
Department of Civil, Geological and Mining Engineering -
Research Institute on Mines and Environment (RIME)
Polytechnique Montreal, Montreal, Qc, Canada

Isabelle Demers
Université du Québec en Abitibi-Témiscamingue, Rouyn-Noranda, Québec, Canada

ABSTRACT
Recent studies have shown, based on laboratory experiments, field testing and numerical modelling, that non-acid
generating mine wastes could be reused in reclamation of tailings storage facilities (TSF), therefore allowing for practical
applications of circular economy approaches on mine sites. However, the quantities of non-acid generating waste rock on
site are often limited. This study evaluated the potential of valorization of potentially acid generating (PAG) waste by
controlling its reactivity using sieving. Waste rocks were sampled at three partner mine sites and the physical and chemical
properties of 7 different fractions were characterized in the laboratory. A total of 21 column tests were carried for various
fractions to assess their geochemical behaviour. Results obtained in this study showed that fine fractions were controlling
reactivity. Recommendations regarding material sieving were proposed and are presented and discussed in this paper.

RÉSUMÉ
Des études récentes, basées sur des expériences en laboratoire, des tests in situ sur le terrain et la modélisation
numérique, ont montré que les stériles miniers non générateurs d'acide pourraient être réutilisés pour la restauration des
parcs à résidus miniers (PÀR), permettant ainsi des applications pratiques d'approches d'économie circulaire. Cependant,
les quantités de stériles non générateurs d'acidité sont souvent limitées. Cette étude a évalué le potentiel de valorisation
des stériles potentiellement générateurs d'acide (PGA) en contrôlant leur réactivité par tamisage. Les roches stériles ont
été échantillonnés à trois sites miniers partenaires et les propriétés physico-chimiques de 7 fractions différentes ont été
caractérisées en laboratoire. Un total de 21 essais en colonne a été réalisé pour évaluer le comportement géochimique
des matériaux. Les résultats obtenus dans cette étude ont montré que les fractions fines contrôlaient la réactivité. Des
recommandations concernant le tamisage des matériaux ont été proposées et sont présentées et discutées dans cet
article.

1 INTRODUCTION of the sulfides minerals which is dependent on the transport

of gas and water (Bao et al. 2020; Chi et al. 2013), climatic
The mining industry generates significant volumes of waste conditions (Neuner et al. 2013), mineral composition
rock during ore recovery, which are typically stored in (Elghali et al. 2018), particle size (Erguler and Erguler
stockpiles on the land surface (Vriens et al. 2020a). 2015), and microbiological activities (Blackmore et al.
Stockpiles can contain hundreds of millions of cubic meters 2018; Nordstrom et al. 2015). AMD can be prevented by
of waste rock and can reach several dozens of meters in eliminating one of the components of the oxidation
heigh and excess hundred hectares in surfaces (Amos et reactions (i.e., water, oxygen or sulfides).
al. 2015).Typically, waste rock particle size is ranging from Waste rock has traditionally been stored on-site
micrometer to meters and their physicochemical properties indefinitely, leading to substantial costs for reclamation due
can vary widely depending on the blasting methods (i.e., to the lack of technical ability to revalorize or reuse (Vriens
firing diagram) and the minerals hardness (Aubertin 2013; et al. 2020a). An innovative strategy aims to manage mine
Lapakko et al. 2006). Stockpiles are therefore waste by integrating economic efficiency and
heterogeneous structures controlled by construction environmental sustainability through the reuse of mine
method (e.g., end-dumping, push-dumping), the mine waste for specific purposes (Tayebi-Khorami et al. 2019).
sequencing and the spatial variation of the This strategy offers several benefits, including the
physicochemical composition of the waste rock (Lahmira et reduction of the stockpiles volume as well as the costs
al. 2017). This internal heterogeneity poses significant associated with material extraction and transportation
challenges for waste rock management, particularly (Aznar-Sánchez et al. 2018). While non-acid generating
regarding the generation of acid mine drainage (AMD). waste rock has found extensive applications in the
AMD is produced when sulfide minerals are exposed to construction of mine hauls (Laverdière et al. 2023), dam
atmospheric conditions (water and oxygen) under structures, drainage inclusion (Saleh-Mbemba et al. 2019),
unsaturated conditions such as encountered in the large backfilling (Feng et al. 2022) and covers system for the
stockpiles (Molson et al. 2005). The oxidation of sulfides reclamation of tailings storage facilities (Kalonji-Kabambi et
minerals can result in the mobilization of trace elements, al. 2020; Pabst et al. 2017), its availability on-site is often
acidic pH, high concentration of sulfates and metal(oid)s limited. The reuse of reactive waste rock can offer an
(Amos et al., 2015). AMD is controlled by the oxidation rate
opportunity to mitigate environmental impacts associated This implies that their particle size range is wide with
with their disposal. significant spread between d60 and d10 which is commonly
One of promising approach in the control of the observed in typical waste rock samples. C-pre-oxidized
reactivity of the waste rock involves segregating and waste rock differed significantly, with over 50 % of particles
sieving (removing) the fine particles, a strategy that has having size finer than sand (d50=1.6 mm), a d10 finer than
been successfully demonstrated for a non-potential of acid 0.1 mm, and presented a lower coefficient of uniformity
generation (NPAG) waste rock (Elghali et al. 2019). Fine (Cu=70), showing a narrower range of particle sizes.
fractions generally generate higher contaminants release
rate due to their elevated specific surface area and have
larger influence on the overall reaction rate (Erguler and
Erguler 2015; Lapakko et al. 2006). Particle size influences
the drainage pH but also the rate of acid production and
acid neutralization.
The reuse of acid generating waste rock by the control
of the reactivity through sieving method has received
limited attention and there are significant challenges that
must be addressed to improve their potential for reuse.
Waste rock reactivity is site specific and can present high
physicochemical variability (St-Arnault et al. 2020), even
within a mine site or a specific lithology. This can result in
different mineralogical compositions, particle size
distribution (Smith et al. 2013) and hydrogeological
properties (Azam et al. 2007), which can affect their reuse.
Furthermore, waste rock may have been exposed to prior
oxidation over several years and even decades which can
increase the complexity for waste rock valorization.
In this research, laboratory-scale tests were conducted
to evaluate the potential of reactive waste rock valorization Figure 1 : Particle size distribution of the three waste rock
by mechanical sieving. Three different types of waste rock types fitted with Fredlund model (2005).
were extensively characterized by fractions and a total of
21 weathering columns tests were performed to assess the The specific gravity of the particles was measured for the
geochemical behavior of the studied fractions. The waste 7 columns fractions following D854 ASTM (2014) for the
rocks lithologies and fractions studied were classified as particle diameter smaller than 4.75 mm and C127 ASTM
uncertain to potentially acid generating. The results (2015) for the particle diameter larger than 4.75 mm. The
obtained in this study showed the importance of the specific gravity of the total sample of A was 2.70, 2.92 for
distribution of sulfidic minerals and neutralizing minerals B and C was 2.76.
with particle size on the potential of the reuse of the
reactive waste rock. 2.2 GEOCHEMICAL CHARACTERIZATION

2 MATERIAL AND METHODS The total sulfur (S) and carbon (C) of the 7 fractions for
2.1 SAMPLING each type of waste rock were analyzed using induction
furnace (ELTRA CS-2000).
Three types of waste rock from gold mines sites located in The neutralization potential (NP) was calculated based
Abitibi-Témiscamingue region, in Canada, were sampled: on the total carbon content by Lawrence et Scheske (1997)
A (open-pit gold mine), B (underground mine), C-pre- method. The acid-generating potential (AP) was calculated
oxidized (underground mine). All samples were taken by Sobek method (1978) which consists of converting the
directly at the pile of the mine sites (≈ 450 kg/type) and total S content to maximum AP, assuming negligible
stored in barrels. Waste rock samples were dried using a sulfates concentrations. The neutralization potential ratio
fan to a water content of less than 3 % in the laboratory (NPR=NP/AP) was evaluated to classify all the fractions:
prior to the tests. Waste rock samples were truncated at potentially acid generating (PAG) if NPR<1, non-acid
the laboratory to a maximum particle size of 37.5 mm (to generating (NPAG) if NPR>2 or uncertain for NPR between
respect the ratio of the laboratory scale). 1 and 2.
Waste rock samples were separated into 14 fractions in Mineralogical semi-quantification of the major mineral
the laboratory using mechanical sieving (37.5, 25, 19, 14, phases (±1 % precision and limit of detection of 1 %) was
9.5, 4.75, 3.35, 2.36, 2.0, 1.4, 1.0, 0.85, 0.425, 0.25, 0.075 determined by X-ray diffraction (XRD) using Rietveld
mm). The particle size distribution (PSD) was determined method for the specific fractions in the weathering columns
according to ASTM D6913 (2021) and the distribution of (0-1 mm, 1-2.36 mm, 2.36-3.35 mm, 3.35-4.75 mm,
the particles smaller than 0.075 mm was determined using 4.75 -37.5 mm,0-37.5 mm, 10-37.5 mm). Mineralogy was
hydrometer test following ASTM D7928 (2017). A and B reconciled using bulk chemical composition by inductively
were characterized by similar PSD, and both had a particle coupled plasma atomic emission spectrometry (ICP-AES)
diameter at 10 % passing, d10, of 0.1 mm and a coefficient following a four-acid digestion with HNO3/Br2/HF/HCl.
of uniformity (Cu = d60/d10) of approximately 130 (Figure 1).
2.3 EXPERIMENTAL SET-UP 50 and 70 mg/L solution prepared by using a standard
solution of 100±1 mg/L.
A total of 12 small weathering columns (i.e., 4 for each
type) and 9 larger columns (i.e., 3 per type) were set up in 3 RESULTS
the laboratory (Figure 2). Each column contained 3.1 Geochemical characterization
complementary fractions: C1(0-1 mm), C2(1-2.36 mm),
C3(2.36-3.35 mm), C4(3.35-4.75 mm), C5(4.75-37.5 mm), All three waste rocks showed a significantly greater sulfur
the coarse fraction C6(10-37.5 mm) and the total fraction content than carbon content in all the 7 fractions (Figure 3).
C7(0-37.5 mm). The larger columns were set up to ensure Therefore, the three waste rock were all characterized by
a ratio of 8 between the column inner diameter and the a low neutralization potential, NP, with an overall arithmetic
maximum particle size diameter. A quantity of 1 kg of waste mean of 47 kg CaCO3/t for A, 28 kg CaCO3/t for B and
rocks was placed dry and loose in the small columns at a 10 kg CaCO3/t for C.
porosity around 0.4. For the large columns, a total of Sulfur content in A waste rock tended to be relatively
55.3 kg was placed dry and loose at a porosity varying constant and around 2% in all fractions (Figure 3a). Sulfur
between 0.25 to 0.41. The columns were left open and in B waste rock was predominantly contained in the coarse
exposed to the laboratory conditions. Multiple wetting- particle range C5(4.75-37.5 mmm) comprising
drying cycles were applied to all the columns. Cycles approximately 67 % of the total mass of sulfur (Figure 1).
duration varied with the size of the columns and the Carbon content was relatively constant for fractions greater
geochemical behavior. 10 to 20 cycles for a duration of 70 than 1.00 mm (%C=0.5) but was slightly greater for fraction
to 140 days were applied to the small columns, while the C1(0-1.00 mm) with %C=0.9, suggesting higher
larger column tests were conducted for 15 cycles and neutralization potential in the fines. However, considering
lasted 465 days. the particle size distribution (Figure 1), most of the carbon
was in fact associated with the coarse fraction > 4.75 mm
(57%). NP ranged from 41 kg CaCO3/t to 64 kg CaCO3/t,
while AP ranged from 36 kg CaCO3/t to 59 kg CaCO3/t,
depending on the fraction size. More specifically, NP were
slightly higher for fractions C1(0-1.00 mm) and
C2(1.00-2.36 mm) than for coarser fractions, which
resulted in an uncertain classification as NPR>1. However,
all the other fractions C3 to C7 of waste rock A were
classified as PAG.
Sulfur in waste rock B was preferentially (i.e.,
approximately 49%) distributed in fractions smaller than
4.75 mm and sulfur content tended to increase with
decreasing particle size (Figure 3b). For example, sulfur
content was greater than 6% in the fine fractions
C1(0-1.00 mm) and less than 3% in the fraction greater
than 4.75 mm. No significant variation of carbon content
was observed in the different fraction of B waste rock
Figure 2 : Configuration of the weathering column tests. (0.2<%C<0.4). Among all the fractions, the fine fraction
a) Small columns tests: 4 columns tests were set up for C1(0-1.00 mm), had the highest NP. Most of the carbon
each type of waste rock (A, B, C) b) Large columns kinetic content (55 %) was in the coarse particles >4.75 mm
test: 3 columns per type of waste rock was set up with a considering particle size distribution. NP of B waste rock
total of 9 large columns. was lower than the AP for all the fractions. Consequently,
all the fractions of B waste rock were classified as PAG.
At the beginning of each wetting-drying cycle, distilled Finally, sulfur in pre-oxidized C waste rock was limited
water (0.4 L for the small columns - i.e., ratio solid/liquid of (%S<0.5) for fraction <4.75 mm but tended to increase in
0.4 - and 19 L for the larger columns – i.e., ratio solid/liquid the coarser fraction (>1.1 %S in fraction 4.75 mm +; Figure
of 0.35), was added on the top of the columns and left in 4c). Only 38.5 % of the total mass of sulfur was associated
contact with the samples for 4 hours. Leachates were with the fine fraction <4.75 mm, which represented 62.3 %
analyzed for pH, redox potential (ORP), electrical of the mass of the samples. The carbon content was
conductivity (EC) and temperature using Hanna HI5222 smaller than 0.5 % for all fractions, with most of the carbon
Benchtop Meter. Alkalinity and acidity were determined by (76 %) associated with the fine particles <1.00 mm. NP and
titration using 0.02 N H2SO4 and 0.02 N NaOH. The AP were low because of the limited S and C content
composition of the leachate (Al, As, Ba, Be, Bi, Ca, Cd, Co, (<50 kg CaCO3/t). All the fractions of waste rock C were
Cr, Cu, Fe, K, Li, Mg, Mn, Mo, Na, Ni, Pb, S, Se, Si, Sr, Te, classified as PAG.
Ti and Zn) was analyzed by ICP-AES (Perkin Elmer Optima The main gangue chemical composition of the three
3100 RL) after filtration (0.45 microns) and acidification types of waste rock included Al, Na, Ca, K and Mg. S and
(50 % volume/volume HNO3). Sulfates concentrations the XRD quantification results revealed that pyrite was the
were determined by turbidimetric method (Clesceri et al., only primary sulfide mineral.
1998) using Hach DR3900 spectrophotometer. Sulfates Pyrite concentration concentrations were similar for all
calibration was performed using a working set of 0, 10, 30, samples in A. No discernible trend was observed between
TYPE A

TYPE B

TYPE C

Figure 3: Sulfur and carbon content for each fraction (left), AP and NP (middle) and NPR (right) of type A, B and C waste
rock.
pyrite content and particle size distribution in A waste rock. No sulfide mineral or pyrite was detected by XRD
Calcite (CaCO3) was identified as the primary carbonate analysis in the fraction C1(0-1.00 mm) to
mineral and was slightly less present in the coarse fraction C4(3.35-4.75 mm) of the pre-oxidized waste rock C and
>4.75 mm (2.3 %) than the finer samples C1 (2.9 %), C2 was only detected in samples C5(4.75-37.5 mm),
(2.9 %), C3 (2.9 %) and C4 (2.7 %). C6(10-37.5 mm) and C7(0-37.5 mm). No neutralizing
Concentration of pyrite in B waste rock increased with mineral was detected by XRD analysis in the fraction C1 to
decreasing particle size, where the highest concentration C4. Calcite was detected in fractions C5, C6 and C7.
observed was in sample C1 (7.2%) followed by C2 Secondary minerals were detected by XRD analysis in C
(6.8 %)>C3 (2.9 %)>C4 (2.7 %)=C5=C6. Calcite content (gypsum and jarosite), suggesting a prior oxidation.
was slightly greater in the sample C1 (2.8 %) than the The geochemical characterization results indicated
+4.75 mm sample (2.3%). Elemental concentrations of Fe, distinct distribution of sulfide minerals and neutralizing
Cu and Zn varied significantly with particle size in B waste minerals among the different lithologies of waste rock (A, B
rock, following a decreased as the particle size increases and C). However, for all types of waste rock, the highest
(C6<C5<C4<C3<C2<C1). The presence of Zn and Cu neutralization potential was observed in the small fraction
suggested sulfide minerals in traces such as chalcopyrite C1(0-1.00 mm). Consequently, while sieving finer particles
and sphalerite. Secondary minerals were detected by XRD can effectively reduce the production of acidity, it may also
analysis in B (gypsum and jarosite), suggesting a prior decrease the capacity of the material to neutralize acidity.
oxidation. Kinetic columns tests were conducted to validate the
geochemical characterization findings and provide a more
comprehensive understanding of the consequences oxidized waste rock. For example, iron concentration
associated with the sieving of the small particles in reactive decreased from 100 mg/L to 1 mg/L in column C1 after 60
waste rock. days. Similar behavior was observed for Zn concentrations.
Concentrations started from >1 mg/L to ~ 0.5 mg/L in all
3.2 Small columns results the columns, except from C4 where it dropped to 0.3 mg/L
after 60 days.
The leachate pH from the small columns containing A Calcium and magnesium concentration were following
waste rock were initially slightly alkaline (Figure 4a). In a rapid decrease trend and were generally greater in
column C1, the pH became neutral after 75 days, while the columns C1 and C2, which contained the finest particles
leachate remained alkaline in columns C2, C3 and C4. For (< 2.36 mm). There was therefore a greater dissolution of
B waste rock, the pH remained consistently neutral neutralizing minerals in these columns compared to the
throughout the test (175 days; Figure 4b). Only C waste coarser fractions. The presence of small particles likely
rock showed acidic leachate (pH=3) throughout the entire contributed to increased specific surface area, thus leading
testing period (Figure 4c). The leachate from the columns to enhanced mineral dissolution and subsequent release of
containing the fine particles for all waste rock (A, B, C) calcium and magnesium ions.
exhibited the lowest pH values.
Sulfate concentrations rapidly decreased after the 3.3 Large scale columns results
fourth cycle for all 12 columns. For A waste rock, sulfate
concentrations remained stable and relatively low (mean of The pH in waste rock A tended to decrease over time
15 mg/L) in columns C3(2.36-3.35 mm) and (Figure 5). For example, in column C7 which contained the
C4(3.35-4.75 mm) compared to the column containing the whole fraction, the pH decreased to a slightly more acidic
small particles C1 (mean of 365 mg/L; Figure 4e). For B level (pH=6.3) than in C5 (pH=6.7) and C6 (pH=6.8). Initial
waste rock, columns C2, C3 and C4 exhibited sulfate pH in waste rock B was around neutrality but progressively
concentrations lower than 100 mg/L, while column C1 decreased to slightly acidic values in all the columns (~6.1).
showed sulfate concentration exceeding 1000 mg/L The lowest pH (5.7) was measured in column C5(4.75-37.5
(Figure 4f). In the case of C waste rock, sulfates remained mm). Leachate from the columns containing C waste rock
consistently high (>500 mg/L) in all columns, and particle showed a decreasing acidic pH ranging from 3.5 to 3
size had limited influence (Figure 4e). across the columns containing the coarse fractions (C5 and
C7). However, the pH increased from 2.7 to 4.1 for column
Concentrations of metals, such as iron and zinc, were C7(0-37.5 mm).
relatively small (< 0.5 mg/L) and even decreased below the Sulfate concentrations were the highest in the columns
detection limits after 70 days for the columns with waste containing the fine particle (C7) for all types of waste rock.
rocks A and B. The neutral pH likely contributed to the Initially, sulfates concentrations rapidly decreased during
lower solubility and leaching of metals. For C waste rock, the first 4 cycles and then stabilized until the end of the
as the pH remained acidic, higher concentration of iron tests (461 days). For type A waste rock, a secondary
were detected in the first leachate (>50 mg/L). Iron decrease was also observed after 250 days. The sulfates
concentration was decreasing in all columns for C pre- concentrations in type A waste rock exhibited a decreasing

a) b) c)

d) e) f)

Figure 4 : Evolution of pH and sulfates concentrations in the small columns on material A, B and C.
Figure 5 : Evolution of the leachate chemistry in the large column tests.
trend as the particle size increased. Concentrations were
one order of magnitude higher in column 3.4 Effect of the particle size on waste rock reactivity
C7(0-37.5 mm; 100 mg/L) compared to column
C5(4.75-37.5 mm; 10 mg/L). Sulfates concentration in Normalized sulfate release rates were calculated after
column C7 was two orders of magnitude greater than in the fourth cycle for each type of waste rock (
column C6(10-37.5 mm; 1 mg/L). In B waste rock, the
sulfates concentrations demonstrated the same Table 1) in the small columns. The highest rate was
decreasing trend as the particle size increased. observed in the columns C1(0-1.00 mm) containing the
Specifically, column C7(0-37.5 mm) had the highest smallest fractions and for all types of waste rock. For
sulfates concentration of ~180 mg/L, followed by column example, for waste rock A, sulfate release rate was
C5(4.75-37.5 mm) with a concentration of ~60 mg/L. The 8 mg/kg/d for C1, 1 mg/kg/d for C2 and 0.4 mg/kg/d for C3
sulfates concentration was the lowest in column and C4. The release rate tended to decrease rapidly for
C6 (10-37.5 mm), recording ~15 mg/L. In the increasing particle size. For B, the sulfate release rate
pre-oxidized C waste rock, the sulfates concentration in decreased significantly with particle size. For example, the
column C7(0-37.5 mm; 300 mg/L) was like those observed sulfate release rate for C1 was 44.6 mg/kg/d and 1.1
in C5(4.75-37.5 mm; 200 mg/L) and C6(10-37.5 mm; mg/kg/d for C2. This decrease is, however, also explained
200 mg/L). by the significant decrease of pyrite content in coarser
Metals concentration such as iron and zinc in A waste fractions. The rate exhibited only minor variation as the
rock remained low during the tested period (respectively, < particle size increase for C2, C3 and C4 indicating no
0.5 mg/L and < 0.3 mg/L). In B waste rock, iron and zinc significant decrease of the reactivity between the selected
concentrations were the highest in the columns fractions. In the case of C pre-oxidized material, the rate
C5(4.75-37.5 mm; Fe≈1 mg/L and Zn≈2 mg/L) compared remained high across all the columns, even if pyrite was
to C7(0-37.5 mm; Fe≈0.1 mg/L and Zn≈1 mg/L). not detected by XRD analysis and the low sulfur
Furthermore, iron and zinc concentrations were increasing concentration.
during the tested period for all columns of B waste rock as
the pH was decreasing. In C waste rock, iron Table 1: Sulfate release rates in the small columns
concentrations were greater in the columns C5 and C6 Fractions Sulfates release rates
than in the column containing the fines C7. (mg/kg/d)
Calcium concentrations observed in all the leachates A B C
were greater in the columns containing the fine particles
C7, which is consistent with the geochemical behavior C1(0-1.00 mm) 8 44.6 43
observed in the small columns. Magnesium concentrations C2(1.00-2.36 mm) 1 1.1 17
followed a similar pattern, except for C waste rock, where C3(2.36-3.35 mm) 0.4 0.9 18
magnesium concentrations remained similar regardless of C4(3.35-4.75 mm) 0.4 0.8 21
the particle size (~8 mg/L).
 3.5.1 Effect of the pre-oxidation

3.5 Reactivity reduction by mechanical sieving Waste rock C was classified as acid generating
independently of the particle size with little neutralizing
The results of the normalized sulfate release rates, potential observed in the static test. Despite the low
calculated after the fourth cycle for each type of waste rock, concentration of pyrite (less than 0.3%) in fraction smaller
are compared between the total sample C7(0-37.5 mm) than 4.75 mm (C1 to C4), the pH remained acidic for the
and the sieved material at C6 and C5 ( duration of the tests.
Table 2). First, the sulfate release rates were the greater in C waste rock was left exposed to oxidation for several
the column containing the fine for all materials. The rate decades. Secondary minerals such as hydroxysulfates
(i.e., jarosite, schwertmannite), oxyhydroxides (i.e.
Fractions Sulfates release rates ferrihydrite, goethite), or gypsum, can precipitate during
(mg/kg/d) oxidation and neutralization reactions (Blowes et al. 2014).
Secondary minerals have the capacity to store acidity,
A B C
metal(loid)s and sulfates and are only stable for certain pH
C5(4.75-37.5 mm) 0.1 0.6 3.0 and redox conditions (Kaasalainen et al. 2019). The
presence of jarosite was confirmed in C samples and its
C6(10-37.5 mm) 0.03 0.2 2.3 dissolution during the experiment may have contributed to
C7(0-37.5 mm) 1 1.6 4.5 acid production. Those secondary minerals can therefore
increase the complexity of the management of waste rock.
significantly reduced as the particle size increased for A The potential of reuse as an exposed material (water and
and B waste rock. For example, the rate was reduced by 2 oxygen) is limited in the present study.
orders of magnitude for A. The static test showed greater
acidification potential in the coarser fractions C5 and C6 4 LIMITATIONS
than in the total fraction. These results suggest that the
particle size was controlling the reactivity for A. The The potential of valorization and the control of the reactivity
columns C5 and C6 of A waste rock were showing water of PAG waste rock was demonstrated only at the laboratory
quality respecting the local regulation for the duration of the scale for the duration of the test. The time required to
test which suggests a potential of valorization for both produce acidity (i.e., the lag time) may take between weeks
fractions. and months. The duration of the test may not have been
The rate for B was reduced by 1 order of magnitude able to capture the geochemical behavior of the material.
between the total fraction and the coarser fraction (C6). The particle size of the fractions can further increase the
The sulfate release rate did reduce as particle size was lag time of the material before generating acidity (Erguler
increasing as observed between column C5 and the total et al. 2014). The weathering columns were operated under
sample C7 in B. However, iron and zinc concentration were controlled and simplified conditions which may not fully
showing the highest concentration in the coarser fraction capture and represent the complexity and the
C5, as the pH was decreasing. The static test showed a heterogeneity of the field conditions (Vriens et al. 2020b).
higher acidification potential in the total fraction C7, due to
the high concentration of pyrite in the fine fraction (C1) but 5 CONCLUSIONS
also a higher neutralization potential in the fine fraction.
The results in the coarser fraction C6(10-37.5 mm) showed This research showed that there was a great potential for
an overall better water quality. Therefore, the coarser reducing waste rock reactivity by increasing its particle size
fraction C6(10-37.5 mm) showed a greater potential of and removing finer particles by sieving. However, the
valorization than all the other fractions. results indicate that higher concentrations of metals and
The pre-oxidized C waste rock sulfates release rate lower pH could be generated by the coarse fraction due to
remained similar, regardless of the fine content and the the neutralization availability, which highlights the
particle size of the fractions. The acidic pH and the overall complexity of sieving method to reduce reactivity. The
water quality of the leachate from the pre-oxidized C waste static results by fractions showed the implications of the
rock showed little valorization potential. distribution of sulfide minerals and the neutralizing minerals
within the fractions. Finally, the pre-oxidized material C
Table 2 : Sulfate release rates in the large columns showed little potential for reuse as the water remained
acidic regardless of the particle sizes and the sieving
process.
Fractions Sulfates release rates The reduction of reactivity in waste rock can indeed
(mg/kg/d) facilitate waste rock management and present a significant
A B C
opportunity for generating materials for construction
purposes such as mine hauls, covers for mine closure and
C5(4.75-37.5 mm) 0.1 0.6 3.0 dams. Conducting larger-scale field tests over extended
periods would be necessary to fully assess the potential of
C6(10-37.5 mm) 0.03 0.2 2.3 utilizing reactive waste rock.
C7(0-37.5 mm) 1 1.6 4.5
6 REFERENCES Erguler, Z.A., and Erguler, G.K. 2015. The effect of particle
size on acid mine drainage generation: Kinetic
Amos, R.T., Blowes, D.W., Bailey, B.L., Sego, D.C., Smith, column tests. Minerals Engineering, 76: 154-167.
L., and Ritchie, A.I.M. 2015. Waste-rock Feng, J., Zhang, K., Zhan, H., and Chao, L. 2022. Improved
hydrogeology and geochemistry. Applied Soil Evaporation Remote Sensing Retrieval
Geochemistry, 57: 140-156. Algorithms and Associated Uncertainty Analysis
Aubertin, M. 2013. Waste rock disposal to improve the on the Tibetan Plateau. Hydrology and Earth
geotechnical and geochemical stability of piles. System Sciences Discussions: 1-28.
Proc., 23rd World Mining Congress. Kaasalainen, H., Lundberg, P., Aiglsperger, T., and
Azam, S., Wilson, G.W., Herasymuik, G., Nichol, C., and Alakangas, L. 2019. Impact of declining oxygen
Barbour, L.S. 2007. Hydrogeological behaviour of conditions on metal (loid) release from partially
an unsaturated waste rock pile: a case study at oxidized waste rock. Environmental Science and
the Golden Sunlight Mine, Montana, USA [journal Pollution Research, 26(20): 20712-20730.
article]. Bulletin of Engineering Geology and the Kalonji-Kabambi, A., Bussière, B., and Demers, I. 2020.
Environment, 66(3): 259-268. Hydrogeochemical behavior of reclaimed highly
doi:10.1007/s10064-006-0077-7. reactive tailings, Part 2: Laboratory and field
Aznar-Sánchez, J.A., García-Gómez, J.J., Velasco- results of covers made with mine waste materials.
Muñoz, J.F., and Carretero-Gómez, A. 2018. Minerals, 10(7): 589.
Mining Waste and Its Sustainable Management: Lahmira, B., Lefebvre, R., Aubertin, M., and Bussière, B.
Advances in Worldwide Research. Minerals, 8(7): 2017. Effect of material variability and compacted
284. Available from https://www.mdpi.com/2075- layers on transfer processes in heterogeneous
163X/8/7/284 [accessed. waste rock piles. Journal of contaminant
Bao, Z., Blowes, D.W., Ptacek, C.J., Bain, J., Holland, S.P., hydrology, 204: 66-78.
Wilson, D., Wilson, W., and MacKenzie, P. 2020. Lapakko, K.A., Engstrom, J.N., and Antonson, D.A. Effects
Faro Waste Rock Project: Characterizing Variably of particle size on drainage quality from three
Saturated Flow Behavior Through Full-Scale lithologies. In Poster paper presented at the 7th
Waste-Rock Dumps in the Continental Subarctic International Conference on Acid Rock Drainage
Region of Northern Canada Using Field (ICARD). 2006. pp. 1026-1050.
Measurements and Stable Isotopes of Water. Laverdière, A., Hao, S., Pabst, T., and Courcelles, B. 2023.
Water resources research, 56(3): Effect of gradation, compaction and water content
e2019WR026374. doi:10.1029/2019wr026374. on crushed waste rocks strength. Road Materials
Blackmore, S., Vriens, B., Sorensen, M., Power, I.M., and Pavement Design, 24(3): 761-775.
Smith, L., Hallam, S.J., Mayer, K.U., and Beckie, Molson, J., Fala, O., Aubertin, M., and Bussière, B. 2005.
R.D. 2018. Microbial and geochemical controls on Numerical simulations of pyrite oxidation and acid
waste rock weathering and drainage quality. mine drainage in unsaturated waste rock piles.
Science of the total environment, 640: 1004-1014. Journal of contaminant hydrology, 78(4): 343-
Blowes, D., Ptacek, C., Jambor, J., Weisener, C., Paktunc, 371.
D., Gould, W., and Johnson, D. 2014. The Neuner, M., Smith, L., Blowes, D.W., Sego, D.C., Smith,
geochemistry of acid mine drainage. L.J., Fretz, N., and Gupton, M. 2013. The Diavik
Chi, X., Amos, R.T., Stastna, M., Blowes, D.W., Sego, waste rock project: Water flow through mine
D.C., and Smith, L. 2013. The Diavik Waste Rock waste rock in a permafrost terrain. Applied
Project: Implications of wind-induced gas Geochemistry, 36: 222-233.
transport. Applied Geochemistry, 36: 246-255. Nordstrom, D.K., Blowes, D.W., and Ptacek, C.J. 2015.
doi:https://doi.org/10.1016/j.apgeochem.2012.10. Hydrogeochemistry and microbiology of mine
015. drainage: an update. Applied Geochemistry, 57:
Elghali, A., Benzaazoua, M., Bussière, B., and Bouzahzah, 3-16.
H. 2019. Determination of the available acid- Pabst, T., Aubertin, M., Bussière, B., and Molson, J. 2017.
generating potential of waste rock, part II: Waste Experimental and numerical evaluation of single-
management involvement. Applied layer covers placed on acid-generating tailings.
Geochemistry, 100: 316-325. Geotechnical and Geological Engineering, 35(4):
Elghali, A., Benzaazoua, M., Bouzahzah, H., Bussière, B., 1421-1438.
and Villarraga-Gómez, H. 2018. Determination of Saleh-Mbemba, F., Aubertin, M., and Boudrias, G. 2019.
the available acid-generating potential of waste Drainage and consolidation of mine tailings near
rock, part I: Mineralogical approach. Applied waste rock inclusions. In Sustainable and Safe
Geochemistry, 99: 31-41. Dams Around the World/Un monde de barrages
Erguler, G.K., Erguler, Z.A., Akcakoca, H., and Ucar, A. durables et sécuritaires. CRC Press. pp. 3296-
2014. The effect of column dimensions and 3305.
particle size on the results of kinetic column test Smith, L.J., Blowes, D.W., Jambor, J.L., Smith, L., Sego,
used for acid mine drainage (AMD) prediction. D.C., and Neuner, M. 2013. The Diavik Waste
Minerals Engineering, 55: 18-29. Rock Project: Particle size distribution and sulfur
characteristics of low-sulfide waste rock. Applied
Geochemistry, 36: 200-209.
St-Arnault, M., Vriens, B., Blaskovich, R., Aranda, C., Klein,
B., Mayer, K.U., and Beckie, R.D. 2020.
Geochemical and mineralogical assessment of
reactivity in a full-scale heterogeneous waste-
rock pile. Minerals Engineering, 145: 106089.
Tayebi-Khorami, M., Edraki, M., Corder, G., and Golev, A.
2019. Re-thinking mining waste through an
integrative approach led by circular economy
aspirations. Minerals, 9(5): 286.
Vriens, B., Plante, B., Seigneur, N., and Jamieson, H.
2020a. Mine waste rock: Insights for sustainable
hydrogeochemical management. Minerals, 10(9):
728.
Vriens, B., Seigneur, N., Mayer, K.U., and Beckie, R.D.
2020b. Scale dependence of effective
geochemical rates in weathering mine waste rock.
Journal of contaminant hydrology, 234: 103699.
Applying threshold analysis to the selection
of tailings and waste rock closure covers
Holly Williams
SRK Consulting, Vancouver, BC, Canada

ABSTRACT
This paper describes a framework that was used to guide the selection of
tailings and waste rock covers for a mine remediation project in Northern
Canada. Previous analyses had shown that no remediation option would eliminate the need for long-term water treatment;
however, cover construction could reduce the volume of water that required treatment. A trade-off study was completed
comparing the higher cost of more complex (lower infiltration) covers to the associated savings in long-term water treatment
costs to identify cases where the construction of lower infiltration covers could potentially “pay for themselves.” Threshold
analyses were completed to test the robustness of the base case conclusions to changes in input assumptions and identify
priority areas for further studies. The results for the waste rock dumps were robust and showed clear patterns for
constructing very low infiltration (geosynthetic) covers on higher-strength waste rock dumps. The results for the tailings
were less clear, and the optimal cover selection was dependent on many factors including reasonable changes in
construction cost assumptions, cover performance (infiltration), tailings geochemistry, and timing of future water quality
changes.

1 INTRODUCTION have been proposed, ranging from isolation soil covers to

very low infiltration covers using geosynthetics.
This paper presents a case study that showcases a Constructing lower infiltration covers has the advantage
framework used to guide the selection of tailings and waste of reducing the amount of net percolation through the
rock covers for a mine remediation project in Northern cover. This, in turn, leads to decreased seepage through
Canada. The study was conducted during the early design the tailings and waste dumps, resulting in lower long-term
stage when multiple investigations were underway. The water treatment costs. However, lower infiltration covers
final cover design depends on several factors including the are also more costly to construct. By comparing cover
quantity and quality of available borrow material, construction costs to water treatment savings, it is possible
hydrology, water quality, seepage estimates, geochemical to identify when lower infiltration covers “pay for
characterization of the underlying material, and water themselves,” i.e., to identify the cases where the
balance estimates. These factors were all subjects of construction of lower infiltration covers can be justified
ongoing studies and had high uncertainty. solely on the basis that doing so reduces total costs.
This case study used an economic trade-off study to To guide the selection of cover designs early in the
compare the higher cost of more complex (lower infiltration) remediation planning stage, this case study compared the
covers to the associated savings in long-term water higher cost of more complex cover systems to the
treatment costs to identify cases where the construction of associated savings in long-term water treatment costs.
lower infiltration covers could potentially “pay for It is important to note that the full design process should
themselves” Threshold analyses were completed to test consider additional benefits (both financial and
the robustness of the base case conclusions to changes in nonfinancial) associated with each cover type.
input assumptions and identify priority areas for further
studies. 2 METHODOLOGY
This case study serves as an illustration of how one can
approach the broader challenge of decision-making under 2.1 Cover Options
uncertainty in geoenvironmental engineering projects. It
also provides a framework for making early design The five cover types analysed are listed in Table 1. It
decisions amidst ongoing studies and evolving should be noted that the cover types are broad categories
uncertainties. rather than specific prescriptions, and many details will
1.1 Case Study Background need to be evaluated in further design.
All remediation options considered for the mine in this case
study involve the construction of covers for the tailings and
waste rock dump areas. Extensive analyses have
demonstrated that none of these options can entirely
eliminate the need for long-term water treatment. However,
the construction of covers can significantly reduce the
volume of water requiring treatment. Several cover designs
2.2 Cost Inputs and Assumptions in loading and (2) reduction in loading proportional to the
reduction in percolation.
2.2.1 Cover Construction Costs The analysis assumed a baseline flow rate to the water
treatment plant and then varied the flowrate depending on
Cover construction cost estimates included: infiltration through the covers. Comparing that range to a
• Supply and installation of geogrid / geofabric for cost-capacity-curve developed elsewhere showed the
trafficability (tailings areas only) better covers would allow a capital cost savings of $6 per
• Slope regrading (dump areas only) each cubic metre of annual flow reduction. This is the
• Liner supply and supply and installation (for very capital cost used in the analysis.
low infiltration covers only) Power, labour, maintenance costs, and sustaining
• Loading, hauling, and placing cover materials capital costs would be minimally affected by the changes
• Compacting of barrier layers and loosening of in flow and load predicted over the ranges tested in this
growth medium layers (as required) study (up to a 15% decrease in total design flows).
However, lime costs will be significantly influenced by the
Earthmoving unit rates were based on calculated differences in loading. Therefore, changes in lime demand
productivities that followed standard methods, as used by (in $/t) represent the operational cost in the analysis.
earthwork contractors. Haul distances were calculated by Therefore, total water treatment savings were
grouping the waste dumps, tailings areas, and potential calculated as the capital savings per each cubic metre of
borrow areas into representative areas and measuring annual flow reduction plus the operational savings of lime
grades and distances from topographic plans, following demand from the differences in sulphate loadings.
existing road alignments wherever possible.
2.2.3 Analysis
2.2.2 Water Treatment Costs
The capital costs of cover construction are compared to
Water treatment capital and operating costs were reductions in water treatment costs (benefits) that have
estimated for each waste dump and tailings area based on been converted to a net present value (NPV) applying an
preliminary water quality predictions. interest rate proportional to inflation. The analysis also
Many of the waste dumps and tailings areas have an included the following variants mentioned earlier:
acid generation potential that exceeds their acid • Acidity loading varying from no change after cover
neutralization capability, meaning there will be a installation to a reduction proportional to the
degradation of water quality over time. Future water quality infiltration reduction
estimates have been made for each dump and tailings area • Change from current to future water quality
to reflect the degradation of water quality. However, when immediately (after cover construction) or 30 years
the water quality will degrade cannot be estimated after covers are constructed.
precisely. To account for that uncertainty, the analyses
were run with two variants: (1) water quality would change
very soon (by the end of cover construction) or (2) water
quality would change in the more distant future (30 years
post-cover construction). Both assumptions are
represented as a step-change in the model.
Soil covers can have a range of effects on acidity
loadings. Two extremes were considered: (1) no reduction

Table 1. Cover types

Tailings Waste Rock

Cover Type Percolation Percolation
Description Description
(% of MAP) (% of MAP)
No Cover Hypothetical case with no cover and no pond 20% Hypothetical case with no 40%
(10-30%) cover (30-60%)
Rock / Trafficability Rockfill trafficability layer to limit dust exposure 40% N/A N/A
Layer Cover (30-60%)
Isolation Cover Rockfill trafficability layer + Uncompacted soil 20% Uncompacted soil layer to 20%
layer to support vegetation (15-25%) support vegetation (15-25%)
Low Infiltration Cover Rockfill trafficability layer + Compacted soil 10% Slopes regraded to 2.5H:1V 20%
layer to act as an infiltration barrier + (5-12%) (for long-term stability) (15-25%)
Uncompacted soil layer to support vegetation
Very Low Infiltration Geomembrane liner + Rockfill drainage layer + 2% Compacted soil to act as an 2%
Cover Uncompacted soil layer to support vegetation infiltration barrier +
Uncompacted soil to support
vegetation
 The analysis assumes that some sort of cover is
70000
needed as a minimum requirement, and the results look at

Sulphate Concentration (mg/L)

the net costs associated with upgrading from one cover 60000

type to another. 50000

40000
3 RESULTS
30000

The result tables below show the net comparison of 20000

incremental treatment savings to incremental construction 10000

costs from upgrading to more complex, lower infiltration 0
cover types. Positive values are bolded and shaded in grey

2-Fine
1-Coarse

1-Fine

2-Coarse

4-Coarse

4-Fine
3A-Fine

3B-Fine

5A-Coarse

5B-Coarse

5B-Fine
and indicate that the cover upgrade “pays for itself” (the
benefits outweigh the additional construction costs).
Table 2 shows the results for the tailings areas; Table 3 for
the waste dumps. Figure 1. Minimum acidity concentrations required to
upgrade to low infiltration covers (red) compared to base
4 DISCUSSION case acidity concentration inputs (blue)

4.1 Base Case Figure 2 shows the minimum sulphate concentrations

that would need to be assumed to make very low infiltration
4.1.1 Tailings Areas covers “pay for themselves.”. For many areas, the sulphate
concentration estimates for the coarse tailings areas are
The base case results for the tailings indicate isolation quite close to the threshold values, and uncertainties in the
covers (20% infiltration) will always “pay for themselves” geochemistry of these areas could make constructing very
when compared against rock covers (40% infiltration). low infiltration covers more favourable.
Similarly, Table 2 shows upgrading to a low infiltration
cover also always reduces total costs when compared to Uncertainties in the geochemistry of
an isolation cover. Finally, very low infiltration covers only these areas could be an important
reduce total costs for two coarse, high-strength sub-areas, decision factor on whether to construct
provided acid onset occurs earlier.
100000
90000
Sulphate Concentration (mg/L)

4.1.2 Waste Dumps 80000

70000
60000
50000
The base case results for the waste dumps show low 40000
infiltration covers start to reduce total costs for the higher- 30000
20000
strength sources (>30,000 mg/L), provided loading 10000
reduces proportionally to infiltration. Net savings increase 0
1-Coarse

2-Coarse

4-Coarse
1-Fine

2-Fine

4-Fine
3A-Fine

3B-Fine

5A-Coarse

5B-Coarse

5B-Fine
when acid onset occurs earlier. For medium-strength
sources (>5,000 mg/L to <30,000 mg/L), low infiltration
covers only reduce net costs for two of the dumps when
acid onset occurs earlier. Very low infiltration covers
reduce total costs only for very high strength sources
(~100,000 mg/L). Figure 2. Minimum acidity concentrations required to
upgrade to very low infiltration covers (red) compared to
4.2 Threshold Sensitivities base case acidity concentration inputs (blue)
Threshold analyses were completed for each input This process was completed for each input parameter,
parameter. This involved calculating the minimum or for both the tailings areas and waste dumps.
maximum value that swings a decision. For example,
Figure 1 below shows (in red) the minimum sulphate 4.2.1 Tailings Areas
concentrations that would need to be assumed to make low
infiltration covers “pay for themselves” and compares them The results of the sensitivity analysis show that the base
to the sulphate concentrations estimated from the case pattern for low infiltration covers is robust. Upgrading
geochemistry review (in blue). For all tailings areas, the to low infiltration covers always reduces total costs even
estimated sulphate concentrations greatly exceed the with very large swings in cost, Mean annual precipitation
minimum requirement. (MAP), infiltration, and NPV assumptions.
The net savings associated with constructing very low
infiltration covers on tailings sub-areas is sensitive to a
Table 2. Tailings area cost-savings results for upgrading from one cover type to another

Tailings Future Load From Rockfill to an Isolation From Isolation to Low From Low Infiltration to Very
Sub-Area Concentration1 Cover Infiltration Cover Low Infiltration Cover
None Proportional to None Proportional to None Proportional to
Assumed loading reduction
infiltration infiltration infiltration
Water quality change timing No Change Early Late No Change Early Late No Change Early Late
1-Fine High 0.375 3.38 3.14 -0.415 1.09 0.97 -3.265 -2.06 -2.16
1-Coarse Very High 0.195 3.18 2.84 -0.205 1.29 1.12 -1.62 -0.43 -0.56
2-Fine High 0.555 4.96 4.61 -0.615 1.59 1.41 -4.79 -3.03 -3.17
2-Coarse Very High 0.235 3.96 3.54 -0.26 1.6 1.39 -2.025 -0.53 -0.7
3A-Fine High 0.235 2.52 2.14 -0.26 0.88 0.69 -2.005 -1.09 -1.24
3B-Fine High 0.245 2.61 2.21 -0.27 0.91 0.71 -2.065 -1.12 -1.28
4-Coarse Very High 0.13 2.48 1.92 -0.135 1.04 0.76 -1.035 -0.1 -0.32
4-Fine Medium 0.745 4.32 4.5 -0.835 0.95 1.04 -6.49 -5.06 -4.99
5A-Coarse Very High 0.155 3.57 2.5 -0.17 1.54 1 -1.285 0.08 -0.35
5B-Fine Medium 0.44 2.52 2.63 -0.49 0.55 0.61 -3.79 -2.95 -2.91
5B-Coarse Very High 0.23 5.52 3.85 -0.265 2.38 1.55 -1.99 0.13 -0.54
1
Medium: >3,000 and <10,000 mg/L; High: >10,000 and <50,000 mg/L; Very High: >50,000 mg/L

number of inputs. Small changes to construction costs, 4.3 Comparison to Traditional Sensitivity Analyses
infiltration assumptions, lime costs, MAP, or discount rates
could make it more likely that constructing very low Traditional approaches to sensitivity analyses could
infiltration covers would pay for themselves. include simple approaches like testing upper and lower
ranges for each parameter to more complex approaches
4.2.2 Waste Dumps that require defining probability distributions for each
parameter and running simulations. The challenge with the
The waste dump threshold analysis indicated the base latter is there often is not enough data to generate
case conclusions for the low and very high sulphate meaningful distributions. The challenge with the former is
concentration waste dumps are robust for a range of cover the assumed ranges can be too narrow and risk missing
infiltration rates and large swings in cover construction, the true answer. This is a known issue commonly referred
liner, lime, and water treatment costs: to overconfidence bias.
• Isolation covers are appropriate for dumps with low The alternative threshold analysis approach
sulphate concentrations (<2,500 mg/L) demonstrated in this case study offers distinct advantages.
• Very low infiltration covers are appropriate for It does not require any definitions of ranges or distributions,
dumps with very high sulphate concentrations of and instead shows how “wrong” our initial assumptions
approximately ~100,000 mg/L need to be to change the original conclusions. This also
helps identify which uncertainties should be prioritized in
The base case cover selection for most of the further study.
remaining dumps is also robust, with the following For example, in the case of the waste dumps, the base
exceptions: case patterns were robust, indicating it is not necessary to
• Low infiltration covers can reduce total costs for meticulously reassess assumptions as studies progress
many of the medium-strength dumps with and minor changes in inputs may occur. However, the
sulphate concentrations for reasonable swings in tailings area was sensitive to several assumptions, many
cover construction costs, lime costs, MAP, and of which were also highly uncertain at the time of the
infiltration reductions. analysis. These include the assumptions regarding cover
• Very low infiltration covers can reduce total costs performance (such as the availability of sufficient materials
for many of the higher strength dumps (with to achieve the assumed infiltration rates) and the
sulphate concentrations between 30,000 mg/L delineations of tailings areas and assumptions around
and 40,000 mg/L when construction costs are differences in geochemistry and percolation rates between
decreased by approximately 15% or when NPV “fine” and “coarse” areas. This highlighted the need for
discount rates are below 2.1%. The cover revisiting the screening-level conclusions as studies
selection is also sensitive to lime costs, MAP progressed.
assumptions, and infiltration assumptions.
Table 3. Waste dump cost-savings results for upgrading from one cover type to another

Waste Future load From Isolation to Low Infiltration Cover From Low Infiltration to Very Low Infiltration
Dump concentration1 Cover
None Proportional to None Proportional to infiltration
Assumed loading reduction
infiltration
Water quality change timing No Change Early Late No Change Early Late
1 Very Low -0.32 -0.24 -0.24 -4.38 -4.32 -4.32
2 Very Low -0.15 -0.13 -0.13 -1.26 -1.24 -1.24
3 Very Low -0.05 -0.03 -0.03 -0.34 -0.32 -0.33
4 Very Low -0.29 -0.23 -0.23 -2.18 -2.12 -2.13
5 Very Low -5.74 -5.7 -5.71 -32.32 -32.17 -32.18
6 Low -0.37 -0.25 -0.28 -2.47 -2.37 -2.39
7 Low -1.38 -1.11 -1.18 -7.99 -7.75 -7.82
8 Low -0.37 -0.27 -0.27 -1.75 -1.67 -1.67
9 Low -1.65 -1.11 -1.31 -9.89 -9.44 -9.61
10 Low -0.28 -0.17 -0.21 -1.6 -1.51 -1.54
11 Low -3.09 -2.53 -2.74 -22.52 -22.02 -22.21
12 Low -0.16 -0.07 -0.1 -1.07 -1 -1.02
13 Medium -0.58 -0.25 -0.39 -3.62 -3.35 -3.46
14 Medium -1.26 -0.9 -1.06 -8.05 -7.71 -7.85
15 Medium -0.13 0 -0.06 -1.01 -0.9 -0.95
16 Medium -0.77 -0.18 -0.45 -4.8 -4.29 -4.52
17 Medium -0.38 -0.02 -0.18 -2.42 -2.13 -2.26
18 Medium -1.43 -1.02 -1.2 -15.15 -14.8 -14.96
19 Medium -1.33 -0.43 -0.84 -6.57 -5.84 -6.17
20 Medium -2.02 -0.37 -1.16 -11.92 -10.45 -11.16
21 Medium -2.13 0.1 -1 -10.03 -8.25 -9.14
22 High -0.04 0.13 0.13 -0.27 -0.14 -0.14
23 High -0.43 2.59 1.4 -2.95 -0.54 -1.49
24 High -0.44 2.73 1.47 -3.1 -0.56 -1.57
25 High -0.2 0.73 0.36 -0.94 -0.2 -0.49
26 High -0.06 0.49 0.27 -0.54 -0.1 -0.28
27 High -0.12 0.77 0.42 -0.87 -0.16 -0.44
28 High -0.65 1.46 0.62 -2.98 -1.3 -1.97
29 Very High -0.06 1.31 1.31 -0.51 0.59 0.59
30 Very High -0.03 0.61 0.61 -0.23 0.29 0.29
31 Very High -0.1 1.95 1.95 -0.74 0.9 0.9
32 Very High -0.13 2.4 2.4 -0.91 1.11 1.11
33 Very High -0.12 2.44 2.44 -0.93 1.11 1.11
Very Low: < 1,000 mg/L; Low: >1,000 and <3,000 mg/L; Medium: >3,000 and <10,000 mg/L; High: >10,000 and <50,000 mg/L; Very
1

High: >50,000 mg/L

rock dumps were robust and showed clear patterns for

5 SUMMARY AND CONCLUSIONS constructing very low infiltration (geosynthetic) covers on
higher-strength waste rock dumps. The results for the
This paper presented a case study where an economic tailings were less clear, and the optimal cover selection
trade-off was used to guide preliminary cover design was dependent on many factors including reasonable
decisions. It compared the increased construction costs of changes in construction cost assumptions, cover
upgrading from one cover type to another to the associated performance (infiltration), tailings geochemistry, and timing
potential savings in water treatment costs. Many of the of future water quality changes. It should also be noted that
multidisciplinary input assumptions were highly uncertain the analysis focused solely on costs and did not consider
at the time of the study and threshold analysis was used to other potential benefits of covers, such as enhanced
test the sensitivity of each input. The results for the waste potential for vegetation success and alignment with final
land use objectives. Incorporating these additional factors for further study. While the case study provides a specific
is likely to result in cover recommendations that are example of how the framework is applied to help guide the
variants or improvements of the broad cover categories in selection of tailings and waste rock covers, the overall
this study. approach provides a universal framework for dealing with
Threshold analysis is a very useful technique to uncertainty that can be extended to many other
account for the many uncertainties in input parameters applications.
early in the design stage when it is not possible or practical
to run a thorough sensitivity analysis with explicitly defined
ranges and/or probability distributions for each input. It can
show whether base case results and decisions are robust
to conceivable changes in inputs and is also able to identify
which inputs are most sensitive and should be prioritized

Table 4. Results of threshold analysis to determine change required to change base-case conclusions
Parameter Base Case Upgrade from Isolation to Low Upgrade from Low Infiltration to Very Low Infiltration Covers
Assumption Infiltration Covers
Cover Unit varies with haul Relative construction costs can Construction costs need to decrease by 5-25% in order to
construction distance increase by a factor of 2 or pay for themselves in all coarse areas, and by 50-75% for
costs higher. fine areas.
Operating $632/t, delivered These covers always pay for These covers start to pay for themselves in coarse areas
costs (lime themselves when lime costs are when lime costs exceed $570/t and always pay for
costs) above $206/t. themselves when lime costs exceed $830/t. Lime costs need
to increase to over $1,500 / t to ever pay for themselves for
fine areas.
Water $6/m3 annual flow This cover selection is These covers pay for themselves when capital costs are
treatment reduction insensitive to water treatment increased to $40/m3 in coarse areas. In fine areas, these
capital costs capital costs since the operating covers only pay for themselves when capital costs are
costs alone always reduce total increased to unrealistically high levels of more than $150 -
costs. $500/m3.
Mean annual 369 mm/year These covers always pay for For coarse areas, these covers start to pay for themselves at
precipitation themselves when MAP is above 468 mm/year and always pay for themselves at 676 mm/year.
(MAP) 180 mm/year. For fine areas, only when increased to unrealistically high
levels of 1,200 to 5,300 mm/year.
Change in From isolation to low These covers always pay for For coarse areas, these covers start to pay for themselves
cover infiltration=10% themselves when the change in when the change in infiltration is 7% and always pay for
infiltration From low infiltration infiltration is greater than 4%. themselves in fine areas when the change in infiltration is 7%.
to very low Infiltration Fine areas require a change in infiltration greater than 17%.
= 8%
Discount rate 2.5% These covers always pay for For coarse areas, these covers pay for themselves when
themselves when the discount discount rate is 2.0 – 2.7%.
rate is 6% or less.
 Monday, October 2, 2023

GEOENVIRONMENTAL II
Investigating Shear Behaviour of
Bituminous Geomembrane (BGM)-Overliner
Interface in Heap Leach Applications
Muhammad Asim
GeoEngineering Centre at
Queen’s-RMC, Queen’s University, Kingston, ON, Canada,
WSP Canada

Rahul Solanki
Geotechnical Design Coordinator, Kiewit Engineering Group Canada ULC, Oakville, ON, Canada

Fady B. Abdelaal
Assistant Professor in Geotechnical and Geoenvironmental Engineering, GeoEngineering Centre at
Queen’s-RMC, Queen’s University, Kingston, ON, Canada

Sam Bhat
VP, Global Business Development & Chief Technical Officer, Geosynthetics
Titan Environmental Containment, Canada

ABSTRACT
Bituminous Geomembranes (BGMs) are currently used in different mining applications including heap leach pads. These
pads are typically constructed on slopes to allow the collection of the pregnant leach solution or constructed with slopes to
minimize the footprint of the mining application. However, there is a paucity of research examining the shear properties of
geomembrane-soil interface when considering BGM as the primary liner material. Thus, a Laboratory investigation is
conducted using the large-scale Direct Shear Apparatus to study the interface shear strength mobilized between a BGM
and a coarse-grained typically in contact with geomembrane liners in heap leach pads. A customized base plate is used
with the apparatus for all the testing to avoid the slippage of the BGM during the interface tests. It is shown that with an
increase in normal stress from 50 kPa to 1000 kPa, the observed interface shearing is a function of the BGM thickness,
internal structure and the overliner characteristics. Additionally, the interface shear parameters obtained are compared to
those obtained for the textured and smooth polyethylene geomembranes.

RÉSUMÉ
Les géomembranes bitumineuses (BGM) sont actuellement utilisées dans différentes applications minières, y compris les
tampons de lixiviation en tas. Ces plates-formes sont généralement construites sur des pentes pour permettre la collecte
de la solution de lixiviation enceinte ou construites avec des pentes pour minimiser l'empreinte de l'application minière.
Cependant, il existe peu de recherches examinant les propriétés de cisaillement de l'interface géomembrane-sol lorsque
l'on considère le BGM comme matériau de revêtement principal. Ainsi, une enquête en laboratoire est menée à l'aide de
l'appareil de cisaillement direct à grande échelle pour étudier la résistance au cisaillement de l'interface mobilisée entre
un BGM et un grain grossier généralement en contact avec des revêtements de géomembrane dans des coussins de
lixiviation en tas. Une plaque de base personnalisée est utilisée avec l'appareil pour tous les tests afin d'éviter le glissement
du BGM lors des tests d'interface. Il est montré qu'avec une augmentation de la contrainte normale de 50 kPa à 1000 kPa,
le cisaillement d'interface observé est fonction de l'épaisseur du BGM, de la structure interne et des caractéristiques du
revêtement. De plus, les paramètres de cisaillement d'interface obtenus sont comparés à ceux obtenus pour les
géomembranes en polyéthylène texturées et lisses.

1 INTRODUCTION in a variety of mining applications. BGM is used in the

construction of tailings dams, water supply dams, canals,
Bituminous geomembranes (BGM) are manufactured by stockpile covers, heap leach pads, and solid waste
impregnating a polyester geotextile and a glass fleece layer landfills. BGMs are introduced as a liner material in these
with an elastomeric bitumen mix (Daly et al. 2018). This applications application due to their high mechanical
reinforcement layer essentially provides the mechanical properties, relatively high density, and low coefficient of
properties, while the bitumen provides the waterproofing thermal expansion (Samea and Abdelaal 2019).
properties of the BGM (Samea and Abdelaal 2019). Since In heap leach pad applications, the geosynthetic liner is
their development in 1974, BGMs have been widely used subjected to a combination of shear and normal stresses.
The behaviour of BGMs under such conditions and their 2.1 Testing Materials
performance relative to polyethylene (PE) geomembranes
(GMB) requires investigation and is the objective of this The soil used as the overliner for testing is referred to as
study. UL1. This soil represents one of the liner beddings used in
The interface shear strength of GMB relies on heap leach pad construction. UL1 is well graded gravel,
numerous factors, such as the type and texture of the constituting angular crushed limestone with a maximum
GMB, the nature and characterization of the overliner, the particle size of 50 mm (Brachman et al., 2014;).The
magnitude of normal stress and the displacement achieved properties of the overliner can be viewed in Table 1.
during the testing (Markou and Evangelou 2018).
Interface shear strength is one of the important Table 1. Characteristics of tested soils
parameters that contribute to the effectiveness of the GMB Soil D10 D30 D60 D85 Cu Cc
as a barrier system in mining and other applications. This Type (mm) (mm) (mm) (mm)
is because the GMB shall resist shear forces generated by UL1 2.2 8.9 18.2 30.3 8.3 2.0
the waste material, seismic events, wind, and other
external factors. If the GMB does not possess sufficient For geosynthetics used in testing, a smooth and a
interface shear strength, it may fail or tear, allowing textured PE GMBs are examined in addition to the 4.1 mm
contaminants to migrate into the environment (Punetha et BGM. The properties of the materials can be viewed in
al 2020). Table 2.
BGMs and PE GMBs have distinct strengths and are
commonly used in different applications based on their
respective characteristics. The specific choice between Table 2. Characteristics of the BGMs and GMB used.
BGMs and PE GMBs will depend on the requirements of
Material Manufacturer GMB Type Thickness
the mining project, the specific site conditions, and the (mm)
desired performance characteristics.The interface shear
strength of BGMs can vary depending on several factors, MxA 1 Smooth PE 2.0 ± 0.01
including the composition and thickness of the BGM, the MyTB 2 Double 2.0 ± 0.01
nature and properties of the interfacing medium, and the sided
installation techniques used. As such, it is important to textured PE
conduct rigorous testing and analysis to ensure that the BGMxB- 3 Elastomeric 4.1 ± 0.163
BGM will perform as expected in each application. 4.1 bituminous

2.2 Testing Procedure

2 TEST SETUP
As per ASTM D5321, the testing process typically involves
The shear strength of BGM-soil interface is measured the following steps:
using a large-scale shear test (Figure 1), in accordance
with the ASTM D5321 test method (ASTM, 2021a). This • Construction of the test specimen: A full-scale
type of test involves subjecting a 300x300 mm BGM section of the geosynthetic is installed in the lower
sample to shear forces that simulate the stresses that will box of the large-scale direct shear apparatus, with
be present at the interface with the surrounding soil in the an overlying soil with a thickness of 120 mm
field. placed above the geosynthetic in the upper box.
• Preparation of the Overliner: Before testing, the
overliner is compacted and the surface levelled to
attain uniform load distribution. UL1 is compacted
at standard proctor density of 1580 kg/m3.
• Normal Stress Application: After the assembly of
the shear box, the desired normal stress is
applied at a rate of 100 kPa/min with a seating
period of 15 minutes.
• Application of Shear Forces: Shear forces are
applied to the specimen by hydraulically moving
the upper box while keeping the lower box
stationary, at the standard specified rate of 1
mm/minute. The force required to cause the
geosynthetic to slip or slide along the interface is
measured as the maximum shear force.
• After each test, the tilting of the loading plate is
measured and checked against the limit outlined
in ASTM D 5321. The outputs from the GDS Data
Acquisition System are used to plot the shear
Figure 1. Large Scale Direct Shear Testing Apparatus strength envelopes and attain the required shear
used for the interface testing of BGMs. strength parameters.
2.3 BGM Fixation in the test setup

A recurrent issue with the testing procedure for BGMs is

the BGM slippage under different normal stresses. A
number of trials were conducted by Solanki (2020), using
multiple approaches, such as clamping, gluing (acrylic
plate coated with structural adhesive underlying the BGM
sample), and heat application for bonding, to ensure no
free movement occurred due to slippage between the BGM
and the base material. However, the displacement of the
BGMs could not be prevented, as shown in Figure 2 with
the direction of shearing depicted.

Figure 4. BGM sample after the interface shear test

showing no slippage

In addition to the pictorial observation, interface test

results comparing past attempts by Solanki (2020) in
Figure 5 and with the utilization of the custom plate in
Figure 6 establish the requirement of the plate for all further
BGM testing.
0 10 20 30 40
300 300
250 KPa
Horizontal Stress (KPa)

400 KPa
Figure 2. BGM slippage during initial testing 750 KPa
200 200
This was followed by a number of varied iterative
measures, which culminated in the development of a
customized base plate with adjustable teeth to conform to
100 100
BGMs of varied thicknesses. This plate is modified from the
Pyramid Tooth Gripping Plate developed by Allen and Fox
(2007). The plan view of the schematic for the plate design
can be seen in Figure 3. 0 0
0 10 20 30 40
Horizontal Displacement (mm)

Figure 5. BGM interface attempts by Solanki (2020)

0 10 20 30 40
50 (kPa)
500 100 (kPa) 500
200 (kPa)
Horizontal Stress (kPa)

250 (kPa)
400 500 (kPa) 400
700 (kPa)
1000 (kPa)
300 300

200 200

100 100
Figure 3. Custom plate design for BGM testing
0 0
The use of this plate provided consistent prevention of 0 10 20 30 40
free slippage of the BGM samples, hence allowing the Horizontal Dislacement (mm)
testing to progress as per ASTM D5321. As viewed in
Figure 4, the BGM sample after the test can be seen to be Figure 6. BGM interface tests with the custom plate
in position on the plate with no observable slippage in the
shear direction indicated.
3 RESULTS 0 10 20 30 40
700 700
Results obtained are modelled using the Mohr-Coulomb 600 600

Horizontal Stress (KPa)

envelope in order to obtain the interface friction angle ‘𝛿’
and the adhesion ’𝑐𝑎 ’. The interface friction angle 500 500
pertaining to the maximum shear stress is termed 𝛿𝑃 , while 400 400
the friction angle at the termination of the test (at 45 mm
shear displacement) is termed 𝛿𝐿𝐷 . Since the overliners 300 700 KPa 300
considered in this study are granular, with minimal 800 KPa
200 200
adhesion reported at the interface, the interface friction 900 KPa
angle will be treated as the criteria of comparison. In 100 1000 KPa 100
addition, since post peak response is not particularly
evident in the testing conducted up to 45 mm, only 𝛿𝑃 is 0 0
used for the basis of comparison. 0 10 20 30 40
Figures 7 to 12 show the interface testing performance Horizontal Displacement (mm)
of BGMxB-4.1, MxA 2mm and MyTB 2mm. It can be Figure 9. MyTB 2mm (Textured GMB) interface shear
observed that in the application of high stress conditions, performance with UL1
the smooth GMB mobilizes the highest shear stress,
followed by the textured GMB and the BGM. The post-test
pictures of the specimen samples in the lower box as
observed in Figures 8, 10 and 12, show the surface
damage to the GMB due to the shearing force applied by
the gravel.
0 10 20 30 40
800 800
700 700
Horizontal Stress (KPa)

600 600
500 500
400 400
300 300
700 KPa
200 800 KPa 200
900 KPa
100 100
1000 KPa
0 0 Figure 10. Surface wear recorded after MyTB 2mm testing
0 10 20 30 40 with the direction of shear at 1000 kPa.
Horizontal Displacement (mm)

Figure 7. MxA 2mm (Smooth GMB) interface shear 0 10 20 30 40

performance with UL1 500
250 KPa
500
500 KPa
Horizontal Stress (kPa)

700 KPa
400 1000 KPa 400

300 300

200 200

100 100

0 0
0 10 20 30 40
Horizontal Dislacement (mm)

Figure 11. BGMxB-4.1 interface shear performance with

UL1.

Figure 8. Surface wear recorded after MxA 2mm testing

with the direction of shear at 1000 kPa.
 This raises the intriguing question of the reason for the
low interface frictional performance of the BGM at high
stresses. As observed in Figures 14 and 15, without any
slippage of the BGM specimens in the lower box during the
interface tests, the upper coat of bitumen exhibited
horizontal (shear) deformation at high stresses. This
occurrence can have an impact on the development of
shearing resistance on the BGM surface and can result in
the reduction of the interface friction angle, which would
otherwise be expected to range between the textured and
smooth GMB.

Figure 12. Surface wear recorded after BGMxB-4.1 testing

with the direction of shear at 1000 kPa.

Based on the Mohr Coulomb failure envelopes

compared in Figure 13 and per the comparison of interface
friction angles at peak in Table 3, the envelopes for the
smooth and textured GMB are found to agree with past
results reported as per Abdelaal and Solanki (2021). At
high stresses, the BGM is found to have lower interface
friction comparatively.

Figure 14. Surface shear deformation in BGM at 1000 KPa

0 200 400 600 800 1000 (Plan View)
800 800
700 BGMxB-4.1 700
MxA 2mm
600 600
Shear Stress (KPa)

MyTB 2mm
500 500
400 400
300 300
200 200
100 100
0 0
0 200 400 600 800 1000
Normal Stress (KPa)

Figure 13. Mohr Coulomb Shear Strength Envelope at Figure 15. Surface shear deformation in BGM with UL1 at
Peak 1000 KPa (Sectional View)

This surface shear displacement was only evidenced

Table 3. Interface Friction Angles observed at Peak for for the BGM since neither the textured GMB ( Figure 16)
GMB-UL1 testing nor the smooth GMB (Figure 17) had any signs of
horizontal surface deformation.
Material (GMB-UL1) Interface Friction
Angle 𝜹𝒑 (°)
BGMxB-4.1 24.7
MxA 2mm 31.8
MyTB 2mm 33.8
 parameters that could accelerate or affect this surficial
bitumen movement.

5 ACKNOWLEDGEMENTS

The research presented in this paper was funded by

NSERC Alliance Grant (ALLRP) with Mitacs and Titan
Environmental Containment Ltd. to Dr. F. B. Abdelaal
(ALLRP 571820–21 and IT29902). The authors wish to
thank Titan Environmental Containment Ltd. for providing
the BGM material examined in this study. Special thanks to
Rahul Solanki for his contributions to the assessment
attempts for BGM testing.

6 REFERENCES

Figure 16. MyTB 2mm sample after testing with UL1 at Abdelaal, Fady B., and Rahul Solanki. "Shear strength of
1000 KPa the geomembrane–subgrade interface in heap leaching
applications." Environmental Geotechnics 40, (2021): 1-17.

ASTM (2021a) D 5321: Standard test method for

determining the shear strength of soil–geosynthetic and
geosynthetic–geosynthetic interfaces by direct shear.
ASTM International, West Conshohocken, PA, USA.

Daly, N., Aguirre, T., Breul, B., Breul, B. and Barfett, B.,
2018. Bituminous geomembranes (BGM) for heap leach
pads and dumps for solid wastes in mine construction. In
Proceedings of the 71st Canadian Geotechnical
Conference, Edmonton, AB, Canada.

Samea, A. and Abdelaal, F.B., 2019. Chemical durability of

bituminous geomembranes in heap leaching applications
at 55oC. In Proceedings of Geosynthetics 2019
Conference, Houston, Texas, USA (pp. 298-308).
Industrial Fabrics Association International (IFAI) USA.

Markou, I.N. and Evangelou, E.D., 2018. Shear resistance

Figure 17. MxA 2mm sample after testing with UL1 at 1000 characteristics of soil–geomembrane interfaces.
KPa International Journal of Geosynthetics and Ground
Engineering, 4, pp.1-16.
4 CONCLUSIONS
Brachman RWI, Rowe RK and Irfan H (2014) Short-term
BGM interface testing and comparison with PE smooth and local tensile strains in HDPE heap leach geomembranes
textured GMBs has highlighted a number of observations: from coarse overliner materials. Journal of Geotechnical
Due to the lateral translation of BGM samples during and Geoenvironmental Engineering 144(5): 1–8,
interface shear tests, low interface shear strength was https://doi.org/10.1061/(ASCE)GT.1943-5606.0001120.
initially obtained. The development of the custom base
plate has led to the resolution of this experimental issue at Punetha, Piyush, and Manojit Samanta. "Modelling of
relatively higher normal stresses. shear behaviour of interfaces involving smooth
The BGM-UL1 interface shearing strength was found to geomembrane and nonwoven geotextile under static and
be lower than interfaces involving smooth and textured PE dynamic loading conditions." Geotechnical and Geological
GMBs investigated at high stresses. This has been Engineering 38, no. 6 (2020): 6313-6327.
attributed to the surface shear deformation of the BGM due
to the movement of the upper coat of bitumen relative to its Solanki, R., 2020. Interface shear behaviour of different
lower surface. Since this was not the case with the geomembranes with rough surfaces in Heap Leaching
interafaces involving the PE GMB specimens, the BGM- applications (Doctoral dissertation).
UL1 had lower interface shear resistance at high stresses.
The findings of this paper that highlight that the
mechanism of surface shear deformation during interface
testing needs to be further investigated and quantified. In
addition, further studies are needed to examine all
Effect of heating and depressurization on
the geomechanical behaviour of hydrate
bearing sands

Derrick Ayebazibwe*, Jeffrey A. Priest and Jocelyn L. Hayley

Department of Civil Engineering, Schulich School of Engineering,
University of Calgary, Calgary, AB, Canada

ABSTRACT
The study examines methane hydrate-bearing sand (HBS) deposits, a potential future energy resource found in deep-
water marine sediments or below permafrost. Extracting methane involves dissociating the hydrate, leading to changes in
mechanical behaviour of the HBS. Current constitutive models applied to simulate reservoir response have been calibrated
neglecting the effects of hydrate dissociation, relying on post-formation laboratory tests. This research investigates how
temperature and pressure changes influence hydrate dissociation and the resulting strength and stiffness of HBS. Tests
involving hydrate dissociation to different degrees were conducted on synthesized HBS specimens, that showed reductions
in stiffness and peak strength. These changes began to occur within the bulk hydrate stability boundary, likely due to pore-
scale effects. Further studies aim to explore the contributing factors to these observations.

RÉSUMÉ
Cette étude porte sur les dépôts de sable contenant des hydrates de méthane (HBS), une source d'énergie potentielle que
l'on trouve dans les fonds marins et le pergélisol. L'extraction du méthane implique la dissociation des hydrates, ce qui a
un impact sur la résistance des HBS. Les modèles actuels ne tiennent pas compte de cette dissociation, ni des effets de
la température et de la pression. La recherche analyse comment ces facteurs influencent la dissociation des hydrates et
le comportement de l'HBS. Des tests sur des échantillons d'HBS synthétisés ont montré une diminution de la rigidité et de
la résistance due à la dissociation, principalement liée aux effets d'échelle des pores. Des recherches supplémentaires
sont nécessaires pour expliquer ces résultats.

1 INTRODUCTION instability (Sultan et al., 2004) and sand production

(Yamamoto et al., 2022), necessitating the premature
Rising global population and industrial growth have termination of field tests. Furthermore, natural hydrate
spurred an upsurge in energy consumption, predominantly dissociation resulting from ongoing climate change could
met through increased reliance on fossil fuels, particularly potentially trigger sub-sea slope instability (Westbrook et al.,
coal and oil. However, mounting environmental concerns 2009; Paull et al., 2011).
linked to greenhouse gas emissions and climate change have Extensive research has been conducted on the
prompted a shift towards natural gas (methane) due to its mechanical behavior of hydrate-bearing sands (HBS),
lower carbon dioxide emissions, aligning with aspirations for primarily focused on experimental investigations that explore
a more sustainable 'green energy' future. Methane gas the influence of hydrate formation on soil properties. Such
hydrate, a crystalline ice-like compound found beneath the studies as summarized by Priest and Hayley (2019) and
seafloor and under permafrost, represents a vast reserve of Waite et al. (2009) have revealed that hydrate formation
methane gas, ranging from 105 to 109 trillion cubic feet within the sand leads to enhanced strength (Winters et al.,
(Collett, 1992). These hydrates are present in both coarse- 2004; 2007; Hyodo et al., 2005; 2007; Masui et al., 2005) and
and fine-grained sediments, but the high hydrate saturation in stiffness (Kingston et al., 2008; Priest et al., 2009; Clayton et
coarse sediments (up to 90% of pore space) and their higher al., 2010), alongside diminished permeability and
permeability make them an appealing prospect for energy compressibility (Yoneda et al., 2019). Consequently, most
extraction, considering current exploration and production constitutive models assume that changes in physical
technologies (Boswell and Collett, 2006). properties during dissociation can be directly related to the
Extraction of methane gas from hydrate-bearing coarse- changes observed during hydrate formation, considering
grained sediments has been demonstrated through various variations in hydrate saturation as the governing factor.
field tests, utilizing dissociation methods like thermal Sultaniya et al. (2015) showed that significant reductions in
stimulation, depressurization, or their combinations. small-strain stiffness occur with relatively minor changes in
However, these dissociation processes can induce hydrate saturation, contrasting the stiffness enhancement
geomechanical alterations within the sediment, leading to observed during hydrate formation for equivalent changes in
reductions in strength and potential sediment failure (Hovland saturation. The study by Sultaniya et al. (2015) considered
& Gudmestad, 2001; Nixon & Grozic, 2007). These changes hydrate dissociation to commence after the ambient Pressure
have caused production-related issues such as wellbore (P) and Temperature (T) conditions were beyond the bulk

1
Hydrate Stability Boundary (HSB), potentially explaining the ‘local’ linear variable displacement transducers (LVDTs)
rapid stiffness reduction. Previous literature (Wright et al., attached to the middle portion of the specimen, while an
1998) provides evidence of a shift in the bulk HSB caused by external LVDT measures global axial strains.
capillary pressure effects in a porous medium, however, the A schematic diagram illustrating the triaxial cell components
impact of these changes in hydrate-stable P-T conditions on and specimen showing the locally attached measurement
the strength and stiffness of HBS are still not fully understood. devices of the apparatus is presented in Figure 1.
As hydrate dissociation is a time-dependent process
hinged on heat and mass transfer, the long-term stability of
hydrate-bearing reservoirs, reliant on the strength and
stiffness of HBS, will evolve alongside hydrate dissociation.
As such, models must incorporate changes in strength and
stiffness that manifest during dissociation. To bridge this
knowledge gap, this paper reports on an experimental study
that was conducted to investigate the geomechanical
properties of HBS as the ambient P-T conditions are altered
after complete hydrate formation/incubation to dissociation.
The findings from this study are expected to foster the
development of more robust models that better describe the
impact of gas hydrate dissociation on the behavior of coarse-
grained HBS.

2 MATERIALS AND METHODS

2.1 Experimental Set-up

Laboratory experiments were conducted on artificially

synthesized HBS specimens using an Environmental Triaxial
Automated System (ETAS) manufactured by GDS
Instruments. The system comprises a load frame integrated
with a Hardin type Resonant Column (RC) Drive head for
measuring torsional small strain stiffness. A triaxial cell,
capable of applying confining stresses of up to 20 MPa,
allows tests to be conducted under hydrate stability
conditions. Temperature control is accomplished using a
Julabo cooling unit that circulates a cooling fluid through a Figure 1. Schematic of triaxial apparatus with specimen
tube coiled along the inner wall of the cell. The temperature set-up highlighting measurement devices.
of the cell is monitored through a Pt 100 thermocouple, which
is positioned on the outside of a neoprene membrane 2.2 Selection of Host sand
enclosing the specimen. Furthermore, the temperature of the
HBS specimen itself is directly measured using a The sand specimens utilized were prepared from angular
thermocouple that extends from the base platen into the silica sand sourced from Lane Mountain quarry (Washington,
specimen. A schematic of the triaxial cell during a typical test USA). The particle size distribution of the sand was chosen
along-with some associated internal and external to approximate that of natural HBS (Priest et al. 2018), whilst
components is shown in Figure 1 ensuring uniform water distribution and water retention
During a test the cell is filled with silicon oil up to the properties, which are essential for uniform hydrate formation
height of top cap of the specimen, with confining pressure within the specimen (Minagawa et al. 2009). Previous
applied via injection of nitrogen (N2) gas to the remaining cell investigations have shown negligible fluid movements during
space, which allows the effective use of the RC drive head to hydrate formation and subsequent shearing (Pandey et al.
measure the small strain stiffness of the specimen. Cell 2022). A summary of the key properties of the host sand
pressure is regulated using a computer-controlled pneumatic employed is presented in Table 1.
controller.
Hydrates were formed using the 'excess gas' method, Table 1: Characteristic Properties of host sand.
where a known volume of water is added to dry sand which
then forms hydrate by introducing methane gas into the pore Characteristic Property Value (units)
space of the sand specimen under hydrate stability
Mean particle size diameter, 𝒅𝟓𝟎 0.46 (mm)
conditions. A pneumatic controller is initially used to inject the
methane gas until the target hydrate formation pressure is Effective particle size, 𝒅𝟏𝟎 0.12(mm)
attained (typically 6 MPa), after which point, a high-precision Coefficient of Uniformity, 𝑪𝑼 5.09
pressure volume controller (PVC) is used to regulate
Coefficient of Curvature 𝑪𝑪 0.82
methane gas pressure and measure gas volume consumed
during hydrate formation. Range of void ratio 𝒆𝒎𝒊𝒏 to 𝒆𝒎𝒂𝒙 0.44 to 0.82
Axial load is applied through a loading ram connected to Target/Average Porosity, 𝑛 0.36
an internal load cell. Strain measurements are obtained using

2
2.3 Specimen Preparation

Host sand specimens were prepared by combining a

predetermined mass of water (approximately 7% of the sand
mass) with the dry sand. The sand mixture was then
compacted in 10 equal layers within a 50mm diameter
cylindrical mold positioned on the base pedestal to form a
specimen of height =100 mm. Once formed, the mold, sand
specimen and base pedestal, were weighed to determine
exact mass of unsaturated sand specimen. Subsequently, the
base pedestal was securely fixed to the base of the triaxial
cell, and a nominal seating load applied via the axial ram. A
suction of 45 kPa was applied to facilitate mold extraction and
attachment of local axial and radial LVDTs as shown in Figure
1. The triaxial cell was then lowered and filled with silicone oil
before a nominal 500 kPa, N2 gas cell pressure was applied.
Methane gas up to the target pore pressure of 6 MPa was
then injected into the specimen pore space whilst
simultaneously raising cell pressure to maintain the 500 kPa
effective stress on the specimen. Once at target pressures, a
resonant column test is performed to determine the baseline
stiffness of the sand without hydrates.
Figure 2: P-T conditions for bulk hydrate stability and
2.4 Hydrate Formation
pathways for hydrate formation and dissociation
As stated above the 'excess gas' method, known for its During thermal stimulation (path B→C1→D1 in Figure 2),
expedited formation rate and superior control over hydrate
distribution, was used for hydrate synthesis, drawing from the PVC was maintained at constant volume as the specimen
prior studies (Waite et al., 2004; Priest et al., 2005; Kneafsey temperature was incrementally elevated from B to C1 (beyond
et al., 2007). Hydrate formation can be achieved either by bulk hydrate stability). Consequently, pore pressure
altering the temperature or by elevating pore pressure to get increases from C1 to D1 (Figure 2) owing to hydrate
the specimen into the hydrate stability region (refer to Figure dissociation which causes restoration of hydrate stability.
2). In the conducted tests, the temperature-driven pathway Target effective stress on the specimen was maintained by
was selected to ensure consistency, considering the adjusting the cell pressure to counteract any pore pressure
significant impact of the P-T pathway on the mechanical increase.
behavior of formed HBS (Pandey et al., 2022). Hence, after In the depressurization method (path B→C2→D2 in
subjecting the specimen to the target pore pressure (O A),
the system's temperature was lowered (A B) into the hydrate Figure 2) the methane back pressure was reduced, using the
stability region (cross-hatched area). Throughout the hydrate PVC controller, to a value just outside the hydrate stability
formation phase, the methane gas pore pressure was boundary (from B to C2, Figure 2) to initiate dissociation. At
regulated using PVC (detailed in Section 2.1) to uphold this point, the PVC volume was kept constant and any
relatively constant pressure and monitor CH4 gas volume changes in pore pressure observed. The reduction in
consumption while initiating hydrate formation through pressure and subsequent hydrate dissociation process leads
temperature reduction. Hydrate formation was considered to a reduction in specimen temperature (see Figure 2), due to
complete 24 hours after initiation, demonstrated by a distinct Gay-Lussac's law (Temperature vs gas pressure relationship)
surge in methane gas consumption and subsequent stiffness and the endothermic nature of hydrate dissociation. As the
increase as observed in RC testing (Priest et al., 2005). system temperature is kept constant, the specimen
temperature returns to its initial value, accompanied by a
2.5 Hydrate Dissociation gradual rise in pore pressure (C2→D2) as hydrate stability
equilibrium conditions are re-established and no further
Following hydrate formation, the specimens were increase in pore pressure is observed. During the
subjected to stepwise dissociation to obtain HBS specimens depressurization stage, the cell pressure is adjusted to
with varying hydrate saturation remaining after dissociation. maintain a constant effective stress on the specimen.

Two primary methods were employed to initiate dissociation 2.6 Determining the Hydrate Saturation
( Figure 2):-
The computation of hydrate volume (𝑉𝐻 ) in the specimen
(i) Thermal stimulation (along pathway B→C1→D1) and;
is crucial: (i) after the formation of hydrate and (ii) after
(ii) Depressurization (along pathway B→C2→D2). hydrate dissociation at specific temperature or pressure.

3
Hydrate formation: In this stage the degree of hydrate The Resonant Column (RC) testing setup features a
saturation (Sh) is calculated based on the total mass of water magnetic drive head connected to the top cap (Figure 1).
(H2O) added to the specimen assuming it is completely Applying a sinusoidal voltage to the drive head generates
torsional motion at the specimen's top, detected by an
converted to CH4 hydrate (Priest et al., 2005) under the molar accelerometer attached to the drive head. As the frequency of
ratio (CH4 : H2O) of 1 : 5.75 (Sloan & Koh, 2007), and the applied voltage is altered the resonant frequency of the
computed using the equation: drive head and specimen can be determined when the phase
of the applied torque matches that of the system. The shear
modulus (G) is subsequently determined by employing ASTM
(Mw ∗ Mhydrate )
Sh = ( ∗ 100) [1] D-4015-15-1 guidelines, utilizing the resonant frequency and
(5.75 ∗ MH2 O ∗ Vv ) specimen dimensions. Given the low induced shear strains (<
Where: 10-5), the computed shear modulus is defined as the maximum
shear modulus (Gmax), related to the specimen's Elastic
MW is mass of added water,
modulus (E). The specimen's damping ratio, indicative of
Mhydrate is the molar mass of hydrate (119.63 g/mol), energy dissipation capacity under cyclic loads, is calculated by
ρhydrate is the density of methane hydrate (0.917g/cc), evaluating vibration amplitude decay given by the
accelerometer after power shutdown to the drive head and
MH2 O is the molar mass of water (18.015 g/mol). when the system is vibrating at its resonant frequency (Hardin
VH and Black, 1966).
VV is volume of the void spaces & Sh = ⁄V
V

Hydrate dissociation: In this stage the change in methane 2.7.2 Shear Compression Tests
pore pressure, under constant PVC volume, is used to
determine the number of moles of CH4 gas released during Drained shear compression tests were conducted to
assess the peak deviatoric stress, large strain stiffness, and the
the dissociation process. The moles of free methane gas in relationship between volumetric strain and axial strain of the
the pore space (ndis ) at different stages of hydrate dissociation specimens. All shear tests were conducted at a constant strain
can be computed using the Peng-Robinson equation of state rate of 0.1% per minute. The local LVDTs were used to
(Peng & Robinson, 1976), given by measure the axial and circumferential deformation during the
initial stages of the shear compression test. These local LVDTs
RT aα offer enhanced accuracy in capturing shear deformation
P= ( )−( 2 ) [2] behavior before the occurrence of specimen bifurcation
Vm − b Vm + 2bVm − b 2
Where: compared to the global LVDT (mounted on the loading ram)
which is used to provide comparatively more accurate
Vvg measurements of axial deformations after this point.
(Vm = ⁄n ), Vvg is volume of gas,
dis

R is the universal ideal gas constant (8.314472 Pa/K/mole) 3 RESULTS AND DISCUSSION
a, b and α are Peng and Robinson coefficients;
for methane gas; a = 0.24663, b = 2.64804 × 10 −5 , All sand specimens tested were prepared targeting a
2
final hydrate saturation in pore space, of approximately 40%
T (initial water content ~7%) and average specimen porosity of
and α = 1.391264925 × (1 − √ )
190.6 36% prior to pressurizing the specimen and formation of
hydrates. Table 2 summarizes the tests carried out on both
The degree of hydrate dissociation, 𝐻𝑑𝑖𝑠 (%), can be defined base sand and hydrate-bearing sand specimens along with
as the ratio of change in moles of CH4 gas at any given time the conditions imposed to initiate partial hydrate dissociation.
during dissociation (ndis - nstart ) with total moles available
from the hydrate (nend − nstart ) given by : Table 2: Test-type, method of dissociation initiation, test
labels, and detailed description.
ndis − nstart
Hdis (%) = [( ) × 100] [3] Test -type Test Label Description
nend − nstart
HFS – 0.5 Base / Host Sand
Baseline Tests
HBS – F – 0.5 HBS Sheared after Hydrate Formation
2.7 Geomechanical Testing HBS Sheared after
HBS – TI – 7°C Temp. Increase (4% dissociation)
HBS with
Throughout the hydrate formation and hydrate Thermal Induced HBS – TI – 9°C
HBS Shear after
Hydrate Temp. Increase (55% dissociation)
dissociation stages, RC tests, which are nondestructive Dissociation HBS Shear after
HBS – TI – 10°C
(Richart et al., 1970), were carried out. After the dissociation Temp. Increase (76% dissociation)
stage, triaxial shear compression tests were conducted. After HBS – DI – 4.0 MPa
HBS Shear after
Depressurization (7% dissociation)
the shear test the specimen temperature is increased to allow
HBS with HBS Shear after
complete dissociation of the hydrate. Depressurization
HBS – DI – 3.5 MPa
Depressurization (23% dissociation)
Induced Hydrate HBS Shear after
Dissociation HBS – DI – 3.0 MPa
Depressurization (56% dissociation)
2.7.1 RC Testing
HBS Shear after
HBS – DI – 2.5 MPa
Depressurization (73% dissociation)

4
3.1 Hydrate Formation to ~5.2°C. Beyond this temperature, a significant rise in the
rate of evolved gas pressure is observed, despite the ambient
For a pore pressure of 6 MPa the temperature for hydrate conditions of the specimen being within the region where bulk
stability condition is ~8.0°C (See Figure 2). Therefore, the hydrate stability is expected.
specimen temperature is first reduced to 8.0°C overnight,
before being lowered to 2.0°C the next morning so that RC
measurements could be conducted as hydrate is forming.
Figure 3 shows methane consumed as temperature of the
system is reduced over time, both the temperature inside the
specimen, and on outside of the specimen (at the membrane-
silicon oil interface – See Figure 1) are shown for comparison.
It can be seen that as the bulk hydrate stability condition is
crossed (~15 hours since start of test) with no significant
increase in methane gas consumption until a temperature of
~6.5°C is achieved. The temperature difference between the
bulk hydrate stability line and the initiation of formation is
known as sub-cooling. As hydrate formation occurs (given by
the rapid increase in methane gas consumption) a spike in
the specimen temperature is observed (~2 hrs from start of
the hydrate formation stage) due to the exothermic process
of hydrate formation. After ~28 hrs in the hydrate formation
Figure 4: Measured PVC pressure with temperature during
stage, the PVC volume is kept constant (undrained) and
thermal induced dissociation for specimen: HBS-TI-10 °C.
temperature is lowered further to ~ -1.6°C, to guarantee that
all free water in the specimen had been converted to hydrate
3.2.2 Dissociation by pressure reduction
and examine the temperature induced changes in pore
pressure without any alteration to hydrate saturation.
Depressurization was carried out by increasing the PVC
volume resulting in a decrease in the pore pressure exerted
on the specimen (Figure 5) triggering a notable temperature
decrease due to adiabatic cooling (Gay Lussac’s law). This
temperature reduction is compounded by the endothermic
nature of the dissociation process, as shown by the red line
in Figure 5. At the target pore pressure, the PVC volume is
kept constant, resulting in a pore pressure increase up to the
bulk hydrate stability boundary and a recovery in the
specimen temperature up to the system temperature (~ 2°C).
The anticipated pressure reduction based on PVC volume
expansion computed using the Peng-Robinson equation
(Equation 2) is also shown in Figure 5. As the CH4 gas within
the specimen is at a lower localized temperature compared
to the PVC (which is at room temperature), an expansion of
the extracted gas (increase in pore pressure) along with
Figure 3: Temperature reduction and methane gas volume expected dissociation are anticipated but not accounted for in
consumption over time before and during the hydrate the computation.
formation stage for HBS-TI-10 °C

3.2 Hydrate dissociation

Two distinct methods were used for hydrate dissociation

with each one producing distinctive mechanical behavior.

3.2.1 Dissociation by thermal stimulation

To guarantee that all free water in the specimen had

been converted to hydrate and examine the temperature
induced changes in pore pressure without altering hydrate
saturation, the initial step involved lowering the cell
temperature to -1.6°C before gradually raising it. Figure 4
shows the measured changes in pore pressure given by the
PVC during the temperature excursion. It can be seen that
the slope of the pressure vs temperature trend (shown in red) Figure 5: Pressure and temperature response during
between ~ -1.6°C to ~2.0°C is similar to the trend from 1.7°C depressurization resulting from increasing PVC volume for
specimen HBS-DI-2.5MPa.

5
3.3 RC Tests hydrate dissociation had commenced before the bulk hydrate
stability temperature and pressure conditions have been met.
In order to understand how the formation and
dissociation of hydrate influences the small-strain stiffness
and damping within a sand, a series of resonant column tests
were conducted throughout these stages.

3.3.1 Evolution of stiffness during hydrate formation

Figure 6 shows the evolution of stiffness (Gmax), damping

ratio (D,%), the degree of hydrate formation based on the
volume of CH4 gas consumed, over time from the beginning
of the hydrate formation stage for specimen HBS-DI-10°C. It
can be observed that a rapid increase in small-strain stiffness
occurs after more than 50% of hydrates have been formed,
consistent with previous literature findings (Pandey et al.,
2022; Sultaniya et al., 2017). In the pre-hydrate formation
stage, damping ratio is low. As hydrate crystallization begins,
damping ratio sharply rises, implying possible squirt flow
among sediment grains and the forming hydrate crystals.
Figure 7: Changes in stiffness, damping ratio and
Once hydrates have fully formed, and stiffness has reached
temperature during thermal stimulation as a function of time
its highest value, damping ratio reverts to the initial value.
for HBS-TI-10°C.
Furthermore, subsequent temperature reduction after the 24-
hour hydrate incubation period yields no substantial increase
Depressurization induced dissociation
in small-strain stiffness neither are there any changes in
As depressurization can cause rapid dissociation and
damping ratio, signifying complete hydrate formation.
reduction in stiffness (Sultaniya et al., 2017), a stepwise
reduction in pressure was conducted for the HBS–DI–3.0
MPa specimen. Figure 8 presents the measured stiffness
at each pressure reduction step. A reduction in stiffness and
corresponding increase in damping ratio is observed as
soon as depressurization is first carried out (at ~ 5.7 MPa
and ~ 2.6°C) with a change in rate of stiffness reduction
alongside damping ratio increment observed after ~ 4.2
MPa and ~ 2.5°C.

Figure 6: Evolution of Gmax during the hydrate formation

and subsequent temperature reduction stages for HBS-TI-
10°C.

3.3.2 Changes in stiffness during hydrate dissociation

Thermal induced dissociation

It can be seen in Figure 7, that at the start of this stage Figure 8: Changes in pore pressure and small-strain
when the specimen temperature was raised from ~ -1.6°C to stiffness / shear modulus with time during depressurization.
~4.0°C there was no appreciable impact of temperature
increase on the measured stiffness or damping ratio, 3.3.3 Evolution of stiffness with hydrate saturation
reinforcing the notion that there was no available free water
in the specimen. As the temperature was slowly increased Figure 9 presents the reduction in stiffness as a function of
from 4.0°C to ~6.6°C a gradual reduction in stiffness and hydrate saturation for specimens dissociated through
corresponding increase in damping ratio was observed, with Thermal Stimulation (HBS-TI-10°C) and Depressurization
the reduction in stiffness, and increase in damping, occurring (HBS-DI-3.0MPa). During thermal stimulation, the shear
once the temperature was >4.8°C suggesting that minor modulus reduced to that of the host sand after less than 30%
of the hydrate volume had been lost. In contrast, when

6
depressurization is used, the specimen exhibits a higher in peak shear strength although the pressure-temperature
stiffness (>4.0 GPa) after the same amount of hydrate is conditions are within the bulk hydrate stability region.
dissociated The peak stress reduces from ~5.9 MPa to ~4 MPa and
occurs at a slightly higher axial strain (Ԑa(peak)~ 0.8%)
. compared to the ~0.4% for HBS sheared after complete
hydrate formation. For only a minor loss of hydrate (Hdis =
4%), a ~30% reduction in strength occurs, which is
comparable with the loss in stiffness (Figure 7).

Figure 9: Changes in small-strain stiffness with hydrate

saturation during dissociation for HBS; Thermal stimulation
(red circles) and Depressurization (green diamond shapes)

3.4 Triaxial Shear Strength Tests

As highlighted in Section 3.3.1, the formation of hydrate

within coarse-grained soil using the “excess-gas” method
leads to significant increase in small-strain stiffness (Gmax)
while dissociation by depressurization or thermal stimulation
causes a rapid reduction in Gmax. To consider the impact of
these conditions on the shear strength of sand specimens,
triaxial shear tests were conducted on specimens before and
after hydrate formation, as well as after defined changes in
hydrate saturation due to hydrate dissociation.

3.4.1 Tests on HBS after Thermal dissociation

Figures 10 highlights the stress-strain response of a Figure 10. (a) Stress-strain response of HBS specimens after
sand before and after hydrate formation, as well as after thermal simulation to different temperatures: a) Deviatoric
subsequent dissociation of hydrate through thermal Stress vs. Axial Strain, (b) Volumetric strain vs Axial Strain.
stimulation. It can be seen that the formation of hydrates
within the specimen (solid-blue line) leads to a significant Increasing temperatures results in greater hydrate
increase in peak deviatoric stress when compared to the base dissociation with a corresponding loss in strength. Reduced
sand (Figure 10a, solid black line). In addition, this peak peak deviatoric stress occurs at increasingly higher axial
stress (shear strength) occurs at axial strains of ~0.5%, after strains for higher degrees of hydrate dissociation (i.e.; 𝐻𝑑𝑖𝑠 =
which a large degree of strain softening occurs. The onset of 55%, Ԑa(peak)~ 2%, & 𝐻𝑑𝑖𝑠 = 76%, Ԑa(peak)~ 4.5%), with
strain softening coincides with rapid dilation (Figure 10b),
with the specimen reaching failure (local radial LVDT increasing volumetric compression occurring before the onset
reaching its limit) at around 1.3% axial strain. In contrast, the of dilation (Figure 10b). It is interesting to note that as
base sand exhibits more strain hardening behavior with a temperature increases the reduction in stiffness (Figure 7)
minor peak shear stress identified at an axial strain of ~ 5% becomes much greater than strength (Figure 10a), although
(Figure 10a) with significant compressional volumetric strain the strength does approach the behaviour of the base sand
(Figure 10b) occurring before dilation which occurs well with increasing degree of hydrate dissociation.
before the peak stress. This behavior is similar to the stress-
strain response obtained from many studies on HBS (Priest 3.4.2 HBS after dissociation by depressurization
and Hayley, 2019). Figure 10 also presents the stress-strain
behaviour for HBS specimens subjected to different degrees Figure 11 (a) and (b) present the stress-strain behaviour
of thermal induced hydrate dissociation. Raising the of HBS after undergoing hydrate dissociation to different
temperature of an HBS specimen to 7C leads to a reduction degrees by depressurization, along with the results from the

7
base sand and HBS sand without dissociation for sands. At the onset of hydrate formation, a sudden increase
comparison. Similar to the thermal stimulation, the reduction in damping ratio along with a rapid uptake of methane gas is
in strength is observed to be higher for HBS specimens that indicative of the start of hydrate formation. The small strain
were depressurised to 3.5MPa than 4.0 MPa, and at the same stiffness lags the methane consumption and damping,
temperature of ~ 2.5°C, all within the hydrate stability region. indicating that significant hydrate growth is required before
hydrate impacts stiffness. Stiffness continues to evolve
throughout the hydrate incubation period, along with the
damping ratio gradually reducing even though changes in
methane consumption are minor, suggesting a ripening of the
hydrate.
Increasing specimen temperature to initiate dissociation
leads to a reduction in stiffness even when the specimen
pressure and temperature are still within the bulk hydrate
stability conditions. The reductions in stiffness correspond
with a rapid increase in damping, similar to that observed
during initial hydrate formation. The higher damping ratios
remained when the temperature was kept constant above the
hydrate incubation temperature, suggesting that minor
hydrate dissociation within the bulk hydrate stability region
had occurred, releasing water for enhanced squirt flow. This
rise in damping ratio was also observed during
depressurization to induce hydrate dissociation even when
the specimens remained within the bulk hydrate stability
region. Once outside the bulk hydrate stability region, hydrate
dissociation was more evident, resulting in a significant
increase in pore pressure under constant PVC volume
(undrained conditions).
Improved understanding of the behavior of HBS during
dissociation, both within and outside the hydrate stability
boundary, is crucial for accurate predictions of hydrate
reservoir behavior during long-term methane production.
Future research in this area will contribute to improved
modeling and management of hydrate reservoirs.

5 ACKNOWLEDGEMENT

We would like to thank the University of Calgary, Natural

Sciences and Engineering Research Council of Canada
(NSERC), the Canadian Research Chairs Program and the
Figure 11. (a) Stress-strain response of HBS specimens after IBET PhD Program for their financial support.
depressurization to different back pressures: a) Deviatoric
Stress vs. Axial Strain, (b) Volumetric strain vs Axial Strain. 6 REFERENCES
Further reduction in pressure, i.e.; as observed for HBS-DI- Boswell, R., & Collett, T. (2006). The gas hydrates resource
3.0MPa and HBS-DI-2.5MPa (higher degrees of dissociation) pyramid, Fire in the Ice, US Department of Energy, Office
led to lower peak shear strength occurring at increasingly of Fossil Energy, National Energy Technology Laboratory,
higher axial strains. Minor differences in peak stress, and the Vol. 6, Issue 3, pp. 5 – 7
axial strains at which these occur, are observed for the Collett, T., S. (1992). Well log evaluation of natural gas
specimens undergoing depressurization when compared to hydrates. United States of America (USA)
the specimens undergoing thermal stimulation (Figure 10). https://doi.org/10.2172/6824342
However, these differences may arise due to inherent Hovland, M., & Gudmestad, O.T.(2001). Potential Influence
differences between specimens rather than an effect of the of Gas Hydrates on Seabed Installations. In Natural Gas
dissociation method. Further studies are required to confirm Hydrates: Occurences, Distribution and Detection (eds
or authenticate these conclusions. C.K.Paull and W.P. Dillon).
Hyodo, M. Li, Y., Yoneda, J., Nakata, Y., Yoshimoto, N., &
4 CONCLUSIONS Nishimura, A. (2014). Effects of dissociation on the shear
strength and deformation behavior of methane hydrate-
This paper reports on a series of tests that provide bearing sediments. Journal of Marine and Petroleum
valuable insights into the mechanical behavior of hydrate- Geology, Vol. 51, Elsevier, Amsterdam, Netherlands, pp.
bearing sands (HBS) when subjected to thermal stimulation 52 – 62. https://doi.org/10.1016/j.marpetgeo.2013.11.015
and depressurization. Hyodo, M., Li, Y., Yoneda, J., Nakata, Y., Yoshimoto, N.,
Similar to previous research findings, the formation of Nishimura, A., & Song, Y. (2013). Mechanical behavior of
hydrate has a significant influence on small strain stiffness, gas-saturated methane hydrate-bearing sediments.
damping ratio, and the shear strength of hydrate bearing

8
 Journal of Geophysical Research: Solid Earth, Vol. 118, Stiffness of a Hydrate Bearing Sand, Canadian
Issue 10, pp. 5185 – 5194 Geotechnical Journal, Vol. 55, pp. 988 – 998.
Kneafsey, T. J., Tomutsa, L., Moridis, G. J., Seol, Y., Freifeld, Waite, W. F., Santamarina, J. C., Cortes, D. D., Dugan, B.,
B. M., Taylor, C. E., & Gupta, A. (2007). Methane hydrate Espinoza, D. N., Germaine, J., Jang, J., Jung, J. W.,
formation and dissociation in a partially saturated core- Kneafsey, T. J., Shin, H., Soga, K., Winters, W. J., & Yun,
scale sand sample. Journal of Petroleum Science And T. S. (2009). Physical properties of hydrate-bearing
Engineering, Vol. 56, Issues 1 – 3, pp. 108–126. sediments. Reviews of Geophysics, Vol. 47, Issue 4
Minagawa, H., Ohmura, R., Kamata, Y., Nagao, J., Ebinuma, [RG4003].
T., Narita, H. & Masuda, Y. (2009). Water permeability of Wesbrook, G. K., Thatcher, K. E., Rohling, E. J., Piotrowski,
porous media containing methane hydrate as controlled by A. M., Pälike, H., Osborne, A. N., Nisbet, E. G., Minshull, T.
the methane-hydrate growth process. AAPG Memoir. Vol. A., Lanoisellé, M., James, R. H., Hühnerbach, V., Green,
89, pp 734-739. D., Fisher, R. E., Crocker, A. J., Chabert, A., Bolton, C.,
Nixon, M. F., & Grozic, J. L. H. (2007). Submarine slope Beszczynska-Möller, A., Berndt, C., & Aquilina, A. (2009).
failure due to gas hydrate dissociation: A preliminary Escape of methane gas from the seabed along the West
quantification. Canadian Geotechnical Journal, Vol. 44, Spitsbergen continental margin, Journal of Geophysical
Issue 3, pp. 314 – 325. Research Letters, Vol. 36, Issue 15, pp. L15608/1 – 5
Pandey M.R., Priest, J.A., & Hayley, J.L., (2022). Comparison Wright, J.F., Chuvilin, E.M., Dillimore, S.R. Yakushev, V.S.
of small-strain stiffness of laboratory synthesized hydrate and Nixon, F.M. (1998) Methane hydrate formation and
bearing sands for different formation methods In dissociation in fine sands at temperatures near 0 °C, In
Proceedings of the 75th Canadian Geotechnical Proceedings of the 7th International Conference on
Conference, Calgary, Canada. Permafrost Yellowknife (Canada) Collection Nordicana No.
Paull, C. K., Ussler, I., & Holbrook, W. S. (2011). Tracking the 55.
decomposition of submarine permafrost and gas hydrate Yamamoto, K., Boswell, R., Collett, T.S., Dallimore, S. R. &
under the shelf and slope of the Beaufort Sea. In Lu, H. (2022). Review of Past Gas Production Attempts
Proceedings of the 7th International Conference on Gas from Subsurface Gas Hydrate Deposits and Necessity of
Hydrates (ICGH 2011), Edinburgh, Scotland. Long-Term Production Testing, Energy & Fuels, Vol. 36,
Peng, D., & Robinson, D. (1976). New Two-Constant Issue 10, pp. 5047 – 5062.
Equation of State. Industrial & Engineering Chemistry https://doi.org/10.1021/acs.energyfuels.1c04119
Fundamentals. Vol. 15. Yoneda, J., Oshima, M., Kida, M., Kato, A., Konno, Y., Jin, Y.,
https://doi.org/10.1021/i160057a011. Jang, J., Waite, W. F., Kumar, P., & Tenma, N. (2019c).
Priest, J. A., & Hayley, J. L. (2019). Strength of laboratory Permeability variation and anisotropy of gas hydrate-
synthesized hydrate-bearing sands and their relationship to bearing pressure-core sediments recovered from the
natural hydrate-bearing sediments. Journal of Geophysical Krishna–Godavari basin, offshore India. Journal of Marine
Research: Solid Earth, Vol. 124, Issue 12, pp. 12556 – and Petroleum Geology, Vol. 108, pp. 524 – 536.
12575
Priest, J.A.; Hayley, J.L.; Smith,W.E.; Schultheiss, P.;
Roberts, J. (2018) PCATS triaxial testing: Geomechanical
properties of sediments from pressure cores recovered
from the Bay of Bengal during expedition NGHP-02. Marine
and Petroleum Geology Journal, Vol. 108, pp. 424–438.
Priest, J. A., Rees, E. V. L., & Clayton, C. R. I. (2009).
Influence of gas hydrate morphology on the seismic
velocities of sands. Journal of Geophysical Research
Atmospheres, Vol. 114, Issue B11, pp. B11205/1 – 13
Priest, J. A., Best, A. I., & Clayton, C. R. I. (2005). A laboratory
investigation into the seismic velocities of methane gas
hydrate-bearing sand. Journal of Geophysical Research:
Solid Earth, Vol. 110, Issue B4, pp. B04102/1-13.
Sloan, E.D. Jr., Koh, C.A., & Koh, C.A. (2007). Clathrate
Hydrates of Natural Gases (3rd ed.). CRC Press.
https://doi.org/10.1201/9781420008494
Sultan, N., Foucher, J.P., Cochonat, P., Tonnerre, T.,
Bourillet, J.F., Ondreas, H., Cauquil, E., & Grauls, D.
(2004). Dynamics of gas hydrate: Case of the Congo
continental slope, Marine Geology, Volume 206, Issues 1
–4
Sultaniya, A. K., Priest, J. A., & Clayton, C. R. I. (2015).
Measurements of the changing wave velocities of sand
during the formation and dissociation of disseminated
methane hydrate, Journal of Geophysical Research: Solid
Earth, Vol. 120, Issue 2, pp. 778 – 789.
Sultaniya, A. K., Priest, J. A., & Clayton, C. R. I. (2017).
Impact of formation and Dissociation Conditions on

9
Heavy metals and organic contaminants
impact on smectite behaviors at a similar
double layer thickness
Ouhadi, V.R.
Civil Eng. Department, Bu-Ali Sina Univ., Hamedan, Iran
Yaghoubi, M.
Civil Eng. Department, University College of Omran_Toseeh,
Hamedan, Iran

ABSTRACT
Despite extensive research about soil contaminant interaction, there needs to be more research on the geo-
environmental behavior of smectite with different contaminants while the double layer thickness is remained similar. The
main objective of this paper is to study the mechano-chemical behavior of smectite samples contaminated with different
organic components at an equal double-layer thickness. The experimental study includes a series of unconfined
compression tests, Atterberg limits, and XRD tests to address contaminated smectite's microstructural changes and
engineering properties. The results suggest that even though, according to the double layer theory, the dielectric
constant is the primary factor that controls the soil-organic interaction process, the influence of other characteristics of
organic compounds, such as molecule weight is the dominant controlling factor. This indicates some of the significant
limitations of the double layer theory for clay-organic contaminant interaction.

RÉSUMÉ
Malgré des recherches approfondies sur l'interaction des contaminants du sol, il existe peu de recherches sur le
comportement géo-environnemental de la smectite dans le processus d'interaction avec différents contaminants alors
que l'épaisseur de la couche double est considéré similaire. L'objectif principal de cet article est d'étudier le
comportement mécano-chimique d'échantillons de smectite qui sont contaminés par différents composants organiques à
une épaisseur de la couche double similaire. L'étude expérimentale considère une série d'essais de compression simple,
de limites d'Atterberg, de XRD et d'essais de pH pour traiter les changements microstructuraux et les propriétés
mécaniques des échantillons de smectite contaminés. Les résultats suggèrent que même si, selon la théorie de la
couche double, la constante diélectrique est la principale facteur qui contrôle le processus d'interaction sol-organique,
l'influence d'autres caractéristiques des composés organiques telles que le poids de la molécule est le facteur de
contrôle dominant. Cela indique une partie de la limitation majeure de la théorie de la couche double pour l'interaction
argile-contaminant organique.

1 INTRODUCTION Koupai et al., 2020). Solute transport in compacted

cohesive and clay liners is receiving particular attention in
Soil contamination is one of the common and the contexts of waste storage and the design of materials
controversial environmental issues all over the world with physical properties since the interaction of solute and
(Nazari et al., 2020). The response of soil to the clay particles can significantly affect soil properties (Dabat
contaminants depends on the type of soil and nature of et al., 2020). In other words, pore fluid chemistry is a
organic contaminants (Yong, 2000). Leachate of significant factor in how soil behaves (Muththalib &
contaminant through soil layers may significantly affect Baudet, 2019). Soil pore water quality would affect the
the geotechnical and geo-environmental performance of properties of soils even for a short period of time and
soil (Sunil et al. 2009; Rosli et al. 2019; Wang et al., consequently have noticeable impact upon solute and
2020). Organic chemicals are produced in different contaminant transport in soil (Oztoprak and Pisirici 2011;
industrial projects such as fuel refining, petrochemical Wang et al., 2018; Yan et al., 2018; Shen et al., 2019;
complexes, plastics manufacturing, and detergent Koupai et al., 2020).
industry. The improper use of organic chemicals is Among organic contaminant, methanol and
another source of environmental contamination dichloromethane are important basic industrial chemicals
(Estabragh et al., 2020). (Ouhadi et al., 2017; Liu et al., 2020). Methanol is used as
Cohesive soils are commonly used in geotechnical raw material or solvent in the production of pesticide,
and geo-environmental projects. Therefore, engineers medicine, dye, perfume, coating and synthetic materials.
require a fundamental understanding of cohesive soil In addition to synthetic ammonia, methanol is the only
behaviour as well as qualitative and quantitative large-scale chemical synthesized by coal gasification and
prediction methods of their behavior (Kyokawa et al., natural gas reforming (Xia, 2020). In addition,
2020). Clays are the most important components of dichloromethane (DCM) is a toxic volatile compound
cohesive soils, which are classified as fine-grained soils in which is found in the ground waters and wastewaters of
geotechnical engineering (Yukselen-Aksoy et al., 2008;
the pharmaceutical, chemical, textile, metal-working and aggregated structure due to the soil solute interactions. In
petroleum industries (Shestakova and Sillanpää, 2013). such a case the soil behaves like silty–sandy soils when
To prevent contaminant transport in soils, cohesive the dielectric constant of the pore fluid is considerably
clays rich in montmorillonite are used due to their lower than that of water (Kaya & Fang, 2005; Moavenian
contaminant retention, self-healing properties, and low & Yasrebi, 2008).
permeability (Kyokawa et al., 2020). Smectite is widely The addition of heavy metals and salt to pore water
used as an appropriate material for contaminant retention can affect the behaviour of the soil by influencing the
in clay liners (Ouhadi et al., 2021). electrochemical forces exist between the solid, liquid and
It is well known that electrochemical phenomena on dissolved phases (Ike, 2020).
the surface of clay mineral crystals considerably affect From mechanical point of view, researches have
their macro-structural behavior (Kyokawa et al., 2020) and shown that the unconfined compressive strength
soil properties are sensitive to preferred orientations of increases as the NaCl and CaCl2 salt concentration
the microscopic plate-like clay particles (Wensrich et al., increase. Furthermore, it is reported that the failure strain
2018). increases in NaCl added specimens, thus becoming more
The DDL is an ionic phenomenon that describes the ductile (Vanda, 2014). However, the amount of bentonite
variation of electric potential near a charged surface, such expansion can be significantly reduced with an increase in
as clay (Yong et al., 1992; Koupai et al., 2020). Generally, the calcium concentration in the clay double layer
there are two factors controlling the volume change (Mitchell and Soga, 2005; Lu et al., 2018; Rowe et al.,
characteristics of clays, namely shearing resistance and 2019; Chai & Prongmanee, 2020).
DDL repulsive forces, in which, the DDL repulsive forces According to the Gouy-Chapman theory (1910-1913),
play an important role in explaining the compressibility at specific concentration of organic and salt
and plasticity behaviour, especially for smectite mineral concentrations the achievement of similar thickness of
(Yong and Mulligan, 2004; Thanh Duong & Van Hao, double layer is possible. However, the question which
2020). The amount of physicochemical interaction in soil remains to answer is as follows: If smectite samples with
can be explained by the diffuse double layer theory. The different soil pore fluids have similar thickness of double
charged surface and the distributed charge in the layer, do they show similar geo-environmental behaviour?
adjacent phase are together termed the diffuse double It should be emphasized that according to the Gouy-
layer. The thickness of double layer, 1/K, (equation 1) Chapman theory the dielectric constant is the only
which is the distance from the surface of a clay particle to parameters that governs the clay-organic interaction. In
the centroid of the area of the diffuse layer, represents the other words, other properties of organic electrolytes such
expected soil structure of clay particles due to the soil as molecule size, molecule weight, and viscosity are not
solute interaction (Mitchell & Soga, 2005). considered in this theory. According to the above
discussion, the main objective of this study is to
1 𝜀° 𝐷𝑘𝑇 0.5 investigate on the some of the limitations of double layer
=( ) [1]
𝐾 2𝑛° 𝑒 2 𝜈 2 theory in addressing the process of smectite-different
organic contaminants interaction at similar thickness of
In the above equation, 𝑛° : concentration of ions, k: double layer. For this purposes, impact of two different
Boltzmann constant, T: temperature (K), 𝑒: electronic organic contaminants on the plasticity behaviour,
charge, 𝜀° : static permittivity of the medium, D: dielectric microstructural changes, and unconfined compressive
constant, and 𝜈 is the valence of cations of soil pore fluid. strength of smectite at similar thickness of double layer
The above equation shows that the thickness of double are investigated and discussed.
layer varies inversely with the valence and the square root
of the concentration. In addition, it is directly varying with
the square root of the dielectric constant and temperature,
2. MATERIALS AND METHODS
assuming other factors remaining constant.
2.1. Materials
Generally, reduction or shrinkage of the double layer
thickness produces a flocculated structure while an
Bentonite soil sample was used in this study which is
increase in its thickness results in a dispersed structure
taken from Iran Barit Company. A part of geotechnical
(Ouhadi et al., 2021). Furthermore, interaction of clay
and geo-environmental properties of bentonite sample is
minerals with organic chemicals with dielectric constants
shown in Table 1.
lower than water will result in the development of thinner
According to the results of Table 1, sodium ions takes
interlayer spacing because of the contraction of the soil-
around 61% of exchangeable sites of clay fraction of
electrolyte system (Yong, 2000). This causes a reduction
bentonite. In other words, the initial structure of soil is
in liquid limit (LL) due to soil contamination (Yong et al.,
relatively oriented structure due to the existence of
1992). It is shown that this reduction is dependent on the
sodium ions as the major exchangeable cation of double
value of dielectric constant of pore fluid; decrease in
layer.
dielectric constant causes reduction in the value of liquid
In order to address the mechano-chemical behaviour
limit (Estabragh et al., 2020). Decreasing the dielectric
of bentonite at the presence of different organic and salt
constant and increasing the organic fluid/water ratio
concentrations three different electrolyte was used.
decreased the liquid limit and plasticity index (Evangeline
Methanol and dichloromethane was used as organic
& John, 2010; Olgun & Yildiz, 2010). The research studies
contaminants. Sodium chloride was used as an alkaline
have shown that the clay soil structure changes to the
metal to preserve similar double layer thickness of
contaminated smectite samples with methanol and sample has very large CEC and low soluble electrolyte
dichloromethane. concentration. Since for sample preparation the applied
electrolyte concentrations were very high, therefore the
influence of initial soluble salt concentration of soil upon
Table 1. Some of the physical and geo-environmental its double layer behaviour was ignorable. Secondly, since
properties of bentonite sample. the bentonite sample for all of the experiments was
similar, therefore the background of soil samples in terms
Soil Properties Quantity Soil Properties Quantity of the initial concentration of soluble salts was similar.
Liquid Limit 330 CEC 68.03 Therefore, the impact of initial concentration of soluble
(cmol/kg-soil) salts for all samples upon thickness of double layer was
the same. In other words, the influence of initial interactive
Plastic Limit 80 Exchangeable 41.93 forces of clay particles was similar for all samples. In
sodium sample preparation process, in order to have different
(cmol/kg-soil) ranges of dielectric constant, as the Table 2 shows, in
addition to the pure organic, mixtures of 30% organic
Plasticity Index 250 Exchangeable 24.10 materials and 70% of distilled water are used.
calcium According to the previous researches (Yong et al.,
(cmol/kg-soil) 1992), the plasticity index variations of soil samples are a
very significant indicator which shows the extent of
Carbonate 12.5% Exchangeable 1 interaction of clay-electrolyte systems. For this reason, a
Content Magnesium series of the Atterberg limits tests were performed on
(cmol/kg-soil) different mixtures of soil samples and electrolyte. For this
purpose, soil samples were initially mixed with the
pH (in 1:30 9.8 Exchangeable 1 calculated electrolyte quantity whish had a specific
soil: water potassium concentration of organic chemical for similar thickness of
ratio) (cmol/kg-soil) double layer. In sample preparation stage, soil
suspensions were mixed for two hours on a mechanical
Electrical 0.1 Soil CH shaker in every 24 hours for a period of 96 hours. After
conductivity Classification this so called equilibrium condition, soil suspensions were
(µZ/cm) dried in 60 degrees centigrade in oven for 24 hours. After
drying, soil samples were grinded and passed through
sieve number #200. Then, these organic contaminated
Table 2 shows the general characteristics of water and bentonite samples were exposed to 23 and 72 coml/kg-
organic materials of this study. The data for distilled water soil of sodium chloride. Once again, the previously
are presented for comparison purposes. By the use of mentioned equilibrium concentration process was applied
distilled water and these two organic materials a wide to soil suspension samples. After another drying process,
range of dielectric constant from 9.1 to 80.1 is achieved. Atterberg limits tests were performed on soil samples
based on ASTM4318-00.
To address the behaviour of soil samples from
Table 2. Some of the characteristics of distilled water and microstructural point of view, a series of XRD experiments
organic materials at 20 degrees centigrade were performed upon soil sample. By the use of achieved
(Albright & Gosting, 1946; Wang and Anderko, 2001). XRD spectra, the position and intensity for the major
basal spacing of minerals are reported. XRD analysis was
Material Name Dielectric
constant
Viscosity
(mPa/s)
Bulk
density
Molecule
weight
Molecule
diameter
carried out based on the method proposed in the study
(g/cm3) (g/mol) (nm) stated by other researches (Moore and Reynolds, 1989;
Distilled water 80.1 1.002 0.998 18.02 0.32 Ouhadi and Yong, 2003; Nikolic et al., 2018). For sample
preparation, in each case prepared samples were dried,
100% Methanol 33.1 0.597 0.79 32.04 0.44
pulverized, and passed through sieve #200. Then, 0.2 g
30% Methanol 72.8 1.539 - - - of contaminated dry sample was mixed with distilled water
in a 10 ml volumetric flask. After achieving equilibrium
100% 9.1 0.43 1.33 85 0.92 condition, 10 drops of soil suspension were poured on a
Dichloromethane 3x3 cm glass slide by the use of precise pipet. The XRD
spectra were obtained after the glass slides were air
30% 23 - - - - dried. Those slides were scanned in the 2θ range 4 to 60
Dichloromethane degrees by the use of Siemens-Diffractometer D8
Advance with Cu-Kα radiation. The X-pert High Score
Plus software based on PDF-2/ICDD (2011 released) was
2.2. Sample Preparation and Experimental Methods used to evaluate the data. In order to comprehensively
investigate the impact of microstructure of smectite-
To calculate the thickness of double layer, the equation 1 contaminant mixtures, series of scanning electron
is used. For calculation of thickness of double layer, the microscope photographs were prepared. The SEM
initial electrolyte concentration of soil sample is not taken photographs were taken from prepared cubic
into account. This is due two reasons. First, the bentonite
contaminated samples with 1 cm dimension. Samples In Table 3, BMS-72 and BDS-23 samples have similar
were covered with thin layer of gold and then the SEM thickness of double layer. For these two samples the
photographs were taken by VEGA//TESCAN-LMU in calculated thickness of double layer was 18.5 angstrom.
Tehran Razi Research Center. For a comparison purposes, the plasticity index for
Finally, to investigate the soil-contaminant from bentonite samples exposed to each of these electrolyte
mechanical point of view, a series of unconfined also is presented in Table 3. For the above mentioned two
compression test were performed upon different soil samples, the bentonite sample was initially exposed to
samples. Experiments were performed based on ASTM diluted organic electrolytes in accordance to the
D2166-87. In sample preparation process, dried concentrations presented in Table 3, and then samples
contaminated bentonite samples were mixed with 30% were exposed to different sodium chloride concentrations
distilled water. After complete mixing, the prepared soil according to the concentration shown in the Table 3 for
samples were passed through sieve #10. Then the BMS-72 and BDS-23 samples. Giving to the results of
homogenized sample was compacted into the Table 3, all of these electrolytes at their individual
compression mold of 33 mm diameter and 60 mm height concentrations have caused a noticeable decrease in
by the use of Harvard Miniature Apparatus in three layers, plasticity index of bentonite. For organic contaminants,
subjected to 30 blows per layer. By series of try and error, these effects mainly come from impact of low dielectric
it was found that this condition of soil compaction is constant of electrolyte upon liquid limit of samples,
required to achieve a dry density of 1.5 g/cm 3. The soil whereas the plastic limits of samples have relatively small
samples were extracted from unconfined compression variations. According to the results of Table 3, the addition
mold with the use of extractor of Harvard Miniature of 30% methanol and 30% dichloromethane has
apparatus. Then, the extracted soil samples were air dried decreased liquid limits 51% and 18% respectively. In
for 7 days. To prevent soil cracking during drying period, other words, these two electrolytes at similar
samples were dried in a closed container to be dried concentration have significant difference upon soil
gradually. Finally the unconfined compressive strength of plasticity index. The results show that methanol has more
samples were measured based on ASTM standard impact on soil structure than that of dichloromethane. This
(ASTM D 2166-00). reduction of plasticity index can be attributed to the
formation of flocculated structure at the presence of
electrolytes with low dielectric constant. Even though
3. DISCUSSIONS methanol has caused more flocculated structure in
bentonite, according to Gouy-Chapman theory it is
To investigate the impact of organic electrolyte and normally expected to have lower plasticity index for
alkaline metals upon plasticity behaviour of bentonite at mixture of dichloromethane and bentonite due to the
similar thickness of double layer, series of Atterberg limit smaller dielectric constant of dichloromethane. However,
tests was performed upon bentonite samples at different the mixture of bentonite and methanol has shown lower
concentration of different electrolyte. As was previously plasticity index at similar electrolyte concentration with
addressed, sodium chloride, Methanol, and dichloromethane. In other words, according to the
Dichloromethane was used to achieve similar thickness of achieved results of Table 3, for BMS-72 and BDS-23
double layer at different concentrations of these samples which have similar thickness of double layer, the
electrolytes. Table 3 shows the sample combinations, combination of methanol and sodium has shown more
sample names and impact of different electrolyte impact upon reduction of plasticity index of bentonite. In
combination upon plasticity behaviour of bentonite fact, while the presence of 30% methanol and 72 cmol/kg-
samples. soil of sodium chloride has caused a 62% reduction in
bentonite plasticity index, the electrolyte concentration of
30% dichloromethane and 23 cmol/kg-soil of sodium
Table 3. Atterberg limits of bentonite samples at the chloride has caused only 42% reduction in bentonite
presence of organic and alkaline metals contaminants. plasticity index. These results are in agreement with
results of plasticity index for BM and BD samples.
Sample
Description
Sample
Name
Sodium
concentration
Weight
percentage
Liquid
Limit
Plastic
Limit
Plasticity
Index One of the reasons that the above mentioned results
(cmol/kg-soil) of organic
material
are not suited with double layer theory can be attributed to
Bentonite B - - 330 80 250 the different viscosity of methanol and dichloromethane.
Bentonite+ BS-72 72 - 229 82 147 Other researcher have shown that at the presence of
Sodium
ethanol, methanol, and acetic acid at the concentration
Bentonite+
Methanol
BM - 30 161 80 81
range of 10% to 40%, the viscosity and dielectric constant
Bentonite+ BMS-72 72 30 176 81 95
have contradictory effects (Ouhadi et al., 2017). In one
Methanol+ aspect, a smaller dielectric constant (i.e. in case of
Sodium
dichloromethane) causes much more reduction in double
Bentonite+
Sodium
BS-23 23 - 250 98 152
layer thickness which reduces the water adsorption of
Bentonite+ BD - 30 269 102 167
clay particles. Therefore, a part of retained water behaves
Dichloromethane as free water; consequently the liquidity behaviour of soil
Bentonite+ BDS-23 23 30 230 85 145 happens in lower water content. However, in another
Dichloromethane+
Sodium aspect, the organic material with larger molecule weight
(i.e. in case of dichloromethane) causes a resistance
against liquidity behaviour of soil, consequently the soil This over again proves that molecules of ethanol have
having organic material with larger molecule weight tends shown more interaction with montmorillonite particles.
to show greater liquid limit. According to the results of
Table 2 and Table 3, the second phenomenon controls
the liquid limit variations of dichloromethane contaminated
bentonite.
Commonly, organic contaminants can form hydrogen
bonds with clay particles. These are electrostatic or ionic
bonds. In general, organic contaminants can interact with
clay particles through three different mechanisms: 1)
adsorption to the clay surface through hydrogen bonding
or exchangeable ions. 2) the adsorption of large
molecules by van der Waals forces and entrance between
silicate layers. 3) influence of weight of organic molecules
in interaction with clay particles (Yong, 2000). In addition,
the retention of organic contaminants may happen
through the entrance of organic molecules between clay
layers or by retention of organic molecules in edges of
clay particles (Lagaly, 1981). In acidic conditions the
hydroxyl groups have more access to the H+ ions which
causes an increase in positive charge of clay particles in
their edges. This causes a reduction in repulsive forces.
Consequently clay particles get closer to each other.
Consequently a reduction in water adsorption happens
and the liquid limit reduces (Tajudin, 2016). The initial pH
of methanol and dichloromethane are 7.9 and 9.2,
respectively. This difference provides a more oriented
structure for dichloromethane contaminated smectite
prompting larger liquid limit.
On the other hand it is reported that the large
molecules of organic contaminant cannot easily enter to
the clay double layer. Therefore, if every condition
remains similar, dichloromethane with large molecule
weight in comparison to methanol (around 2.5 times) has
lower impact upon double layer. The larger molecule
diameter of dichloromethane (2 times in comparison with
methanol diameter) contributes to this phenomenon as
well. Therefore in spite of similar thickness of double layer
which achieves from calculation, ethanol has more impact
upon smectite double layer. This could be the main
reason for achieving lower liquid limit and plasticity index
for bentonite sample containing methanol in comparison
to mixture of bentonite and dichloromethane, assuming Figure 1. The SEM photograph of BM (Top) and BD
similar electrolyte consideration and despite of lower (Bottom) samples.
dielectric constant of dichloromethane.
To prove this discussion, scanning electron
microscope photograph was taken from methanol To make more precise investigation on this interaction
contaminated bentonite and dichloromethane process, the x-ray diffraction spectrum of BMS-72 and
contaminated bentonite at similar concentration (Figure 1) BDS-23 samples were examined. Figures 3 and 4 show
and at similar thickness of double layer (Figure 2). As the the XRD results of BDS-23 and BMS-72 and samples,
SEM photographs of BM and BD samples show, in spite respectively. According to the results of Figures 3 and 4,
of lower dielectric constant for BD sample in comparison the interaction of smectite and pore fluid electrolyte has
to BM sample, the BD sample shows a more oriented caused a noticeable decrease in the peak intensity of
structure which verifies low interaction of dichloromethane smectite basal spacing.
with smectite particles. Furthermore, Figure 2 presents In other words, the reduction in the thickness of
the SEM photographs of mixture of bentonite with 30% double layer causes a reduction on the repulsive forces of
dichloromethane and 23 cmol/kg-soil of sodium chloride clay particles. Consequently, clay particles get closer to
(i.e. BDS-23 sample) and mixture of bentonite with 30% each other and a face to edge and edge to edge structure
methanol and 72 cmol/kg-soil of sodium chloride (i.e. forms. In BDS-23 sample (Figure 3), since a flocculated
BMS-72 sample). Even though in these two samples the structure shows less reflection of XRD radiation,
thicknesses of double layer are equal, the sample that therefore, around 63% reduction of the intensity of major
contains methanol shows a more flocculated structure. basal spacing of montmorillonite is observed.
 2000
Bentonite

1600 Bentonite + 30% Dichlromethane + 23

cmol/kg-soil NaCl

Peak Intensity (CpS)

1200

800

400

0
4 14 24 34 44 54
2-Theta-Scale
Figure 3. The XRD diffraction spectrum of BDS-23
sample.

2000
Bentonite
1800
1600
1400
Peak Intensity (CpS)

1200
1000
800
600
400
200
0
4 14 24 34 44 54
2-Theta-Scale
Figure 4. The XRD diffraction spectrum of BMS-72
sample.

As the results of Table 4 shows the bentonite and

mixture of bentonite and dichloromethane has shown
Figure 2. The SEM photograph of BMS-72 (Top) and
relatively similar strength (0.5 and 0.6 MPa, respectively).
BDS-23 (Bottom) samples.
On the other hand a mixture of bentonite and methanol
has shown much larger strength (i.e. 1.14 MPa).
Furthermore, for BDS-23 sample the achieved strength is
However when the x-ray diffraction spectrum of BMS-
0.79 MPa while the measured strength for BMS-23
72 sample was measured (Figure 4), in spite of similar
sample is 1.06 MPa. In one aspect since the
thickness of double layer, much more reduction for the
dichloromethane has lower dielectric constant than that of
intensity of basal spacing of montmorillonite (83%) was
methanol sample, according to the double layer theory the
observed. This again supports the above mentioned
thickness of double layer is smaller for BD sample in
points in which the sample which contains methanol
comparison with BM sample. In fact, due to the lower
shows more flocculated structure.
dielectric constant of dichloromethane in comparison to
In order to investigate the impact of the above
dielectric constant of methanol, one expects to achieve a
mentioned microstructural changes between those two
lower cohesion and higher internal friction angle for BD
samples upon the mechanical behaviour of them, series
sample. However, the SEM photographs (Figure 1)
of unconfined compression tests were performed. Table 4
showed a more oriented structure for BD sample in
presents the results of unconfined compression tests of
comparison with BM sample. This proves the above
bentonite and mixtures of bentonite and different
mentioned discussion that in spite of lower dielectric
electrolytes which are used in this study.
constant for dichloromethane, the more oriented soil
structure of BD sample can be attributed to the role of similar double layer thickness proves that the dielectric
larger molecules of dichloromethane which have not been constant cannot entirely interpret the process of clay –
able to enter to the diffuse double layer of smectite organic contaminant interaction. From the results of this
particles as much as molecules of methanol. study in spite of lower dielectric constant for
dichloromethane contaminated bentonite (BD) than that of
methanol contaminated bentonite (BM), the BD sample
Table 4. Unconfined compressive strength of bentonite shows a more oriented structure which verifies the lower
and contaminated bentonite with different electrolytes. interaction of dichloromethane with smectite particles.

Sample Description Sample Sodium Weight% Unconfined

Name concentration of compressive
(cmol/kg-soil) organic strength 5. References
material (MPa)
Bentonite B - - 0.50
Albright, P. S.and Gosting, L. J. 1946. Dielectric constants
Bentonite+ Sodium BS-72 72 - 0.70 of the methanol-water system from 5 to 55, Journal of
the American Chemical Society, 68(6), 1061-1063.
Bentonite+ BM - 30 1.14 American Society for Testing and Materials, 2016. Annual
Methanol
Book of ASTM Standards, Soil and Rock; Building
Bentonite+ BMS-72 72 30 1.06 Stones. Philadelphia.
Methanol+ Sodium Chai, J. C. and Prongmanee, N. 2020. Barrier properties
Bentonite+ Sodium BS-23 23 - 0.67
of a geo-synthetic clay liner using polymerized sodium
bentonite. Geotextiles and Geo-membranes.
Bentonite+ BD - 30 0.60 Dabat, T. Porion, P. Hubert, F. Paineau, E. Dazas, B.
Dichloromethane Grégoire, B. ... & Ferrage, E. 2020. Influence of
Bentonite+ BDS-23 23 30 0.79 preferred orientation of clay particles on the diffusion
Dichloromethane+ of water in kaolinite porous media at constant
Sodium porosity. Applied Clay Science, 184, 105354.
Estabragh, A. R. Afsari, E. Javadi, A. A. & Babalar, M.
2020. Effect of two organic chemical fluids on the
According to the results of Table 4, the larger mechanical properties of an expansive clay
unconfined compressive strength of BMS-72 sample in soil. Journal of Testing and Evaluation, 48(5).
comparison with BDS-23 sample (1.06 MPa and 0.79 Evangeline, Y. S. & John, R. A. 2010. Effect of Leachate
MPa, respectively), proves that in spite of similar on the Engineering properties of different bentonites.
thickness of double layer for these two samples, due to In Indian Geotechnical Conference, GEOtrendz, IGS
the much more reduction in repulsive forces and Mumbai Chapter & IIT Bombay, Mumbai.
achievement of larger internal friction angle for BMS-23 Ike, E. 2020. Effect of ionic concentrations and ph on the
sample, a larger unconfined compressive strength is Atterberg limit of cohesive soil. Global Journal of Pure
achieved. and Applied Sciences, 26(1), 73-85.
Kaya, A. and Fang, H.Y. 2005. Experimental evidence of
reduction in attractive and repulsive forces between
4. CONCLUDING REMARKS clay particles permeated with organic liquids.
Canadian Geotechnical Journal, 42(2), 632-640.
The following conclusions can be addressed from the Koupai, J. A. Fatahizadeh, M. and Mosaddeghi, M. R.
experimental results of this study: 2020. Effect of pore water pH on mechanical
The results of Atterberg limit tests, SEM, XRD, and properties of clay soil. Bulletin of Engineering Geology
unconfined compressive strength of organic contaminated and the Environment, 79(3), 1461-1469.
bentonite samples are in highly regarded agreement Kyokawa, H. Ohno, S. & Kobayashi, I. 2020. A method for
which show a correlation between micro-structural and extending a general constitutive model to consider the
macro-structural response of organic contaminated electro‐chemo‐mechanical phenomena of mineral
bentonite. crystals in expansive soils. International Journal for
According to the diffuse double layer theory the Numerical and Analytical Methods in Geo-
organic contaminant with lower dielectric constant should mechanics, 44(6), 749-771.
have more impact upon achieving a flocculated structure. Lagaly, G. 1981. Characterization of clays by organic
However, the results of this study show that the methanol compounds. Clay Miner., 16(1), 1.
contaminated bentonite with larger dielectric constant Liu, Y. Li, G. Chen, Z. Shen, Y. Zhang, H. Wang, S. ... &
than dichloromethane appears a more flocculated Gao, J. 2020. Comprehensive analysis of
structure. This conflicting result between expectation from environmental impacts and energy consumption of
double layer theory and experimental study can be biomass-to-methanol and coal-to-methanol via life
attributed to the larger molecule weight and diameter of cycle assessment. Energy, 117961.
dichloromethane which cause lower interaction between Lu, Y. Abuel-Naga, H. Leong, E. C. Bouazza, A. & Lock,
electrolyte and montmorillonite particles. P. 2018. Effect of water salinity on the water retention
Even though the double layer theory significantly helps curve of geo-synthetic clay liners. Geotextiles and
to evaluate the soil-contaminant interaction, impact of Geo-membranes, 46(6), 707-714.
different organic contaminants on smectite behaviors at
Mitchell, J. K. & Soga, K. 2005. Fundamentals of soil Shestakova, M. and Sillanpää, M. 2013. Removal of
behavior, New York: John Wiley & Sons. dichloromethane from ground and wastewater: A
Moavenian, M. H. and Yasrobi, S. S. 2008. Volume review, Chemosphere, Vol. 93, Issue 3, pp. 1258-
change behavior of compacted clay due to organic 1267.
liquids as permeant. Applied Clay Science, 39(1-2), Sunil, B. M. Shrihari, S. and Nayak, S. 2009. Shear
60-71. strength characteristics and chemical characteristics
Moore, D. M. and Reynolds, R. C. 1989. X-ray Diffraction of leachate-contaminated lateritic soil. Engineering
and the Identification and Analysis of Clay Minerals, Geology, 106(1-2), 20-25.
332: Oxford University Press New York. Tajudin, S. A. A. Azmi, M. A. M. Shahidan, S. Abidin, M.
Muththalib, A. and Baudet, B. A. 2019. Effect of heavy H. Z. and Madun, A. 2016. Relationship of physical
metal contamination on the plasticity of kaolin- parameters in Pb-contaminated by
bentonite clay mixtures and an illite-smectite rich stabilization/solidification method. MATEC Web of
natural clay. In E3S Web of Conferences, Vol. 92, p. Conferences.
10005. EDP Sciences. Thanh Duong, N. and Van Hao, D. 2020. Consolidation
Nazari Heris, M. Aghajani, S. Hajialilue-Bonab, M. and Characteristics of Artificially Structured Kaolin-
Vafaei Molamahmood, H. 2020. Effects of Lead and Bentonite Mixtures with Different Pore
Gasoline Contamination on Geotechnical Properties of Fluids. Advances in Civil Engineering.
Clayey Soils. Soil and Sediment Contamination: An Vanda, Y. 2014. Effect of Pore Water Chemistry on
International Journal, 29(3), 340-354. Hydro-Mechanical Behavior of Compacted Expansive
Nikolic, V. Komljenović, M. Džunuzović, N. Miladinović, Z. Clay. Eastern Mediterranean University (EMU)-Doğu
2018. The influence of Pb addition on the properties of Akdeniz Üniversitesi (DAÜ).
fly ash-based geopolymers. Journal of Hazardous Wang, P. and Anderko, A. 2001. Computation of dielectric
Materials 350, 98-107. constants of solvent mixtures and electrolyte solutions.
Olgun, M. and Yıldız, M. 2010. Effect of organic fluids on Fluid Phase Equilibria, 186(1-2), 103-122.
the geotechnical behavior of a highly plastic clayey Wang, X. Huang, L. Yan, C. & Lian, B. 2018. HKCV
soil. Applied Clay Science, 48(4), 615-621. rheological constitutive model of mudstone under dry
Ouhadi, V.R. Fakhimjoo, M.S. and Omid Naeini, S.T. and saturated conditions. Advances in Civil
2017. The Comparison of Plastic and Permeability Engineering.
Behavior of Bentonite in the Presence of Organic and Wang, Q. Chen, J. Liu, J. Yu, M. Geng, W. Wang, P. and
Heavy Metal Contaminants. Journal of Civil and Wu, Z. 2020. Relationships between shear strength
Environmental Engineering Volume 46, Issue 4, pp. parameters and microstructure of alkaline-
25-36. contaminated red clay. Environmental Science and
Ouhadi, V. & Yong, R. 2003. Experimental and theoretical Pollution Research, 1-15.
evaluation of impact of clay microstructure on the Wensrich, C. Pineda, J. Luzin, V. Suwal, L. Kisi, E. and
quantitative mineral evaluation by XRD analysis. Allameh-Haery, H. 2018. Deformation and Fabric in
Applied Clay Science, 23(1-4), 141-148. Compacted Clay Soils. Physical Review Applied, 9(5),
Ouhadi, V. R. Yong, R. N. and Deiranlou, M. 2021. 054003.
Enhancement of Cement‐Based Xia, W. 2020. Comparison of Several Methods for
Solidification/Stabilization cement-based Industrial Methanol Production. Journal of New Energy
solidification/stabilization of a lead-contaminated Developments, 1(1), 12-15.
smectite clay. Journal of Hazardous Materials, 403, Yan, C. Xu, X. & Huang, L. 2018. Identifying the impact
123969. factors of the dynamic strength of mudded
Oztoprak, S. and Pisirici, B. 2011. Effects of micro intercalations during cyclic loading. Advances in Civil
structure changes on the macro behaviour of Istanbul Engineering
(Turkey) clays exposed to landfill Yong, R. N. 2000. Geoenvironmental engineering:
leachate. Engineering Geology, 121(3-4), 110-122. Contaminated soils, pollutant fate, and mitigation:
Rosli, R. N. Selamat, M. R. and Ramli, M. H. 2019. Shear CRC press.
strength and permeability properties of lateritic soils Yong, R. N. Mohamed, A. M. O. Warkentin, B. P.
from northwest Malaysia due to extended Principles of Contaminant Transport in Soils, Elsevier,
compaction. Materials Today: Proceedings, 17, 630- 1992,.
639. Yong, R.N. and Mulligan, C. 2004. Natural Attenuation of
Rowe, R. K. Garcia, J. D. D. Brachman, R. W. I. & Contaminants in Soils, Lewis Publishers, CRC Press,
Hosney, M. S. 2019. Chemical interaction and Florida, 319 p.
hydraulic performance of geo-synthetic clay liners Yukselen-Aksoy, Y. Kaya, A. and Ören, A. H. 2008.
isothermally hydrated from silty sand Seawater effect on consistency limits and
subgrade. Geotextiles and Geo-membranes, 47(6), compressibility characteristics of clays. Engineering
740-754. Geology, 102(1-2), 54-61.
Shen, P. Tang, H. Huang, L. & Wang, D. 2019.
Experimental study of slaking properties of red-bed
mudstones from the Three Gorges Reservoir
area. Marine Geo-resources & Geo-technology, 37(8),
891-901.
Ageing effects on consolidation behaviour
of soft soils
Arazooben Patel, Narges Gheisari & Paul H. Simms
Department of Civil and Environmental Engineering,
Carleton University, Ottawa, Ontario, Canada

ABSTRACT
The management of young materials such as dredged sediments or clayey
tailings is dominated by their consolidation properties, which govern the magnitude and rate of settlement of such soft
deposits. Conventionally, the consolidation behaviour of these materials is designed employing large strain consolidation
theory. Due to prolong consolidation time associated with young materials, however, the time-dependent effects such as
creep and structuration (ageing) may substantively affect the consolidation behaviour. Ageing, in particular, by increasing
the pre-consolidation, can potentially decrease the time required before reclamation can start but can also reduce the
average residual shear strength of the deposit. The ageing process, however, is not well understood, nor easy to predict
from any kind of standard geotechnical measurements. This study aims to understand the ageing effects and develop
correlations to estimate the magnitude and rate of ageing expected under different deposition scenarios. Subsequently,
the experimental findings are incorporated into a large strain consolidation analysis to account for the time-dependent
effects.

RÉSUMÉ
La gestion des matériaux jeunes tels que les sédiments de dragage ou les résidus argileux est dominée par leurs propriétés
de consolidation, qui régissent l'ampleur et la vitesse de tassement de ces dépôts mous. Classiquement, le comportement
de consolidation de ces matériaux est conçu en utilisant la théorie de la consolidation des grandes déformations.
Cependant, en raison du temps de consolidation prolongé associé aux matériaux jeunes, les effets dépendant du temps
tels que le fluage et la structuration peuvent affecter considérablement le comportement de consolidation. Le vieillissement,
en particulier, en augmentant la pré-consolidation, peut potentiellement diminuer le temps nécessaire avant que la remise
en état puisse commencer, mais peut également réduire la résistance résiduelle moyenne au cisaillement du dépôt. Le
processus de vieillissement, cependant, n'est pas bien compris, ni facile à prédire à partir de n'importe quel type de
mesures géotechniques standard. Cette étude vise à comprendre les effets du vieillissement et à développer des
corrélations pour estimer l'ampleur et le taux de vieillissement attendus dans différents scénarios de dépôt. Par la suite,
les résultats expérimentaux sont incorporés dans une analyse de consolidation des grandes déformations pour tenir
compte des effets dépendant du temps.

1 INTRODUCTION (2020), Salam et al. (2023), Patel et al., (2021), Zeng et al.,
(2015), Shahriar et al., (2018), Alam et al., (2021), Ren et
In the thick deposits of remoulded or dredged soils and al., (2022). In addition to laboratory scale studies, signs of
clayey mine tailings, their low hydraulic conductivity can ageing have also been observed for larger scales of
lead to consolidation times from years to decades for studies e.g., in pilot studies and in real field deposits
practical deposit depths. In such circumstances the (Zhang et al., 2022, Dunmola et al., 2022).
consolidation of young material may be substantially Additionally, the magnitude of ageing has been
affected by time-dependent effects such as creep (Watabe observed to depend on certain factors such as water
et al., 2008 & Jeeravipoolvarn et al., 2009) and ageing. The content (Skempton and Northey, 1952; Banas, 1991; Seng
ageing phenomenon can induce an increase in peak and Tanaka, 2012; Shahriar and Jadid, 2018), pore water
strength while reducing compressibility. The conventional chemistry (Miller, 2010; Perret et al., 1996), in tailings -
practice designs deposition plans of young material flocculation protocols (Aladeef and Simms, 2020; Patel et
employing large strain-consolidation theory, which al., 2021) and, polymer dosage (Salam et al.,2023),
assumes a constant compressibility function. Therefore, plasticity index (Yang and Anderson,2016; Shahriar et al.,
the influence of ageing on the consolidation process could 2018; Alam et al.,2021), mineralogy and activity (Skempton
be substantial and bear on efforts to predict settlement and and Northey, 1952; Shahriar et al., 2018), particle size
strength development. (Miller, 2010; Patel et al., 2020), etc. Recently, Ren et al.
Ageing is a phenomenon that occurs in a wide variety (2021) reviewed numerous studies to understand the
of materials ranging from rocks to ideal colloidal thixotropic behavior in soft clays and attempted to develop
suspensions. It has also been extensively studied in natural a correlation between thixotropic behavior and some index
soils by researchers including Skempton and Northey properties of soft soils. However, no well-established
(1952), Leonard and Altschaeffl (1964), Schmertman relationships were proposed.
(1991), Mitchell and Soga (2005) as well as for different Moreover, the rate of aging is also important for freshly
kinds of tailings and dredged sediments by Banas (1991), deposited tailings, dredged sediments, or similar materials
Suthaker and Scott (1997); Jeeravipoolvarn (2005), as they are “young” materials by nature and yet to age.
Jeeravipoolvarn et al., (2009), Miller (2010), Igbinedion Therefore, information on their ageing behaviour may be
required to correctly estimate their long-term consolidation 2.2 Sample preparation
behaviour. Researchers have proposed empirical
relationships to quantify the rate of thixotropic strength gain Samples were prepared with different initial gravimetric
in remoulded natural soils (Zhang et al., 2017; Tang et water contents (GWC) (higher than its liquid limit) by
al.,2020). However, those empirical relationships require adding a predetermined amount of deionized water to the
information on sensitivity and/or the final ageing strength natural Leda clay samples. Natural undisturbed Leda clay
which is not readily available for the young materials. was placed on a metal dish and disturbed through
Recently, Salam et al. (2023) adapted concepts kneading for 50-55 minutes to obtain a soft plastic
developed to understand the ageing behaviour in ideal remoulded state and then a pre-determined amount of
colloidal suspensions and proposed the concept of the dual deionized water was added to reach the targeted water
mode of ageing for soft soils. They suggested that it may contents and hand-mixed again for another 50-55 minutes.
be possible to predict the maximum likely magnitude of Samples were then stored in airtight containers for 48-72
ageing in soft soils based on the strength measured at hours to achieve a homogenous distribution of water. After
Early Equilibrium Strength (EES). Based on these findings, that samples were again hand-mixed for 15-30 minutes to
Patel and Simms (2022) proposed a relationship to quantify eliminate the effects of consolidation during the storage
the rate of ageing and applied it to various soils, including period and bring the sample to the well-reproducible initial
remoulded natural soils and oil sands tailings. They state. This process ensured that the samples were in a
observed that the rate of ageing is fairly constrained for well-defined, reproducible state before being deposited
young materials. Furthermore, studies conducted on well- into containers of different sizes and heights depending on
studied remoulded Leda clay material demonstrated that the experimental setup as shown in Table 2.
the rate of ageing remains nearly constant in self-weight
consolidated samples regardless of their initial water
contents (Patel and Simms, 2022), similar to the finding Table 2 Experimental details for ageing tests
reported for ideal silica gels.
Hence, this study builds on the previous study (Patel Initial Experimental Sample Ageing Ageing
and Simms, 2022) conducted on well-studied Leda clay GWC1 setup size time consolidat
material. It aims to further validate the independence of the (height (days) on stress
rate of ageing under different loading conditions. To *diameter) (kPa)
achieve this, soft Leda clay samples were subjected to (cm2)
different consolidation stresses using two different test set
ups, and ageing was tracked by monitoring the evolution of 1.88 LL Modified 77.61 80 0.55
elastic shear modulus and apparent pre-consolidation consolidation
pressure. Based on experimental findings on Leda clay, the cell
large strain consolidation model was modified to 1.53 LL Modified 57.61 73 1.63
incorporate ageing effects. consolidation
cell
1.95 LL Conventional 0.25 50 6.87
2 MATERIALS AND METHODS Oedometer
1.52 LL Conventional 0.25 51&35 27.5
2.1 Leda clay
Oedometer
The Leda clay used in this study was collected from the 1.51 LL Conventional 0.25 N/A N/A
Navan Landfill in Ottawa (Aladeef and Rayhani, 2017).. Oedometer
The properties of the Leda clay are summarized in Table 1 1
water content is expressed in terms of Liquid Limit of Leda clay.

In this study, a modified consolidation cell was

Table 1 Characteristics of Leda clay developed to study the effects of different sample heights
and consolidation stresses on the ageing behaviour of soft
Characteristics (%) Values Leda clay samples. The modified consolidation cell was
made of acrylic material with a height of 15 cm and a
Natural water content 64 diameter of 7.61 cm. It was equipped with both an LVDT to
Natural solids content 56 measure deformation over time and a TEROS 31 pore
water pressure sensor at 2 cm above the base to monitor
Liquid limit 55
the dissipation of pore water pressure over time (Figure 1).
Plastic limit 25 Both the LVDT and TEROS 31 were connected to a data
Specific gravity 2.7 acquisition system, which collected the data every minute
during testing. The samples were loaded from the top using
Clay content 71 plunger and dead loads. The plunger acted as a porous
D90, D60, D10 (μm) 6,1, <1 stone with a thickness of approximately 3 mm and several
holes of 1 mm diameter in it. The bottom of the plunger was
Clay mineralogy Illite (83%); Kaolinite
covered with a very fine screen to facilitate the dissipation
(11%); Chlorite (6%)
of pore water while preventing the migration of clay
particles during consolidation. During testing, after controlled by 24 hours of loading time. For the test “without
assembling all the accessories of the test setup, the ageing” or EOP test, loading steps commenced with 1.72
dissipation of pore water pressure was monitored kPa and continued until 879 kPa following a LIR of 2. To
continuously. When the pore water pressure reached close prevent the sample from squeezing, the value of the first-
to the expected hydrostatic pressure or nearing the end of loading step was adjusted depending on the initial water
the primary consolidation state, elastic shear modulus, G’, content of the sample. For instance, the starting load step
was measured using an oscillatory rheometer at different was set to about 1.72 kPa for 1.52 LL sample, while it was
time intervals. Following the G’ measurement, the sample 0.21 kPa for 2LL sample. The test with ageing is carried
was reassembled with the load and LVDT. out in the same manner as the one without ageing, except

Figure 1 Schematic of Modified consolidation test set up.

To examine small-strain stiffness (G’), stress-controlled for the elapsed time of ageing under the specifically
amplitude sweep tests were performed on Leda clay assigned effective stress. Upon completion of ageing,
samples using HAAKE viscometer iQ with a vane rotor (FL small load increments were applied in steps until the
22 4B/SS). The important feature of the device is that it destructuration is observed. Details on the initial conditions
allows running the test on the sample in its original of the samples, test setup, ageing time and consolidation
container directly and the vane rotor can be inserted into stresses are available in Table 2. Table 3 shows the
the sample with very minimal sample disturbance. The loading path for the consolidation tests conducted using
vane rotor consisted of four thin blades attached to a conventional oedometers.
central shaft, having a radius of 11 mm and a length of 16
mm (Figure 1). The vane fixture was inserted into the
undisturbed sample up to 2 cm after removing the bleed Table 3 Loading path for the consolidation tests.
water and then oscillated at a constant frequency of 1.6 Hz
while gradually increasing the applied stress. The resulting Sample Loading path
deformation of the sample was used to determine the name
elastic shear modulus (G') and viscous modulus (G"). G'
was determined from the recoverable strain during each 1.95 LL 0.21→ 0.43→ 0.86→ 1.72→ 3.43→ 6.87→
oscillation, while G" was calculated from the measured 8.24→ 9.62→ 10.9→ 12.4→ 13.7→ 15.1→
irrecoverable strain. In this study, samples were only 27.5→ 55→ 109→ 220→ 440→ 879
sheared within the true elastic region, typically half of the 1.52 LL 1.72→ 3.43→ 6.87→13.7→ 27.5→ 32.9→
true elastic range to avoid sample disturbance.1.88 LL and 38.5→ 43.9→ 49.5→ 55→ 109→ 220 → 440→
1.53 LL samples were sheared only up to 10 and 25 Pa 879
respectively to avoid sample disturbance. Replicate
1.51 LL 1.72→ 3.43→ 6.87→ 13.7→ 27.5→ 55→
samples were prepared for the 1.53 LL sample tested
under a modified consolidation test setup. 109→ 220→ 440→ 879
Since the modified consolidation cell has top loading
conditions, it cannot be used to load the sample at
relatively higher consolidation stresses. Conventional 3 RESULTS
oedometers, therefore, were used to load soft Leda clay
samples at relatively higher consolidation stresses. Four 3.1 Ageing under modified condition cell
different consolidation tests were conducted under two
different conditions i.e., with ageing and without ageing. Figures 2 and 3 illustrate the pore water pressure (PWP)
The duration of consolidation during each loading step was dissipation and elastic shear modulus(G’) development in
1.88 LL sample loaded at consolidation stress of 0.55 kPa The conventional oedometers were used to study the
and 1.53 LL sample loaded at consolidation stress of 1.63 ageing behavior under relatively higher consolidation
kPa, respectively. The pore water pressure reached the stresses. Figure 4 shows the compression curves in terms
expected hydrostatic pressure (dotted line) around 18 and of void ratio versus the logarithm of effective stress for the
17 days in 1.88 LL and 1.53 LL samples, respectively. Leda clay samples with and without ageing. Samples aged
However, the elastic shear modulus continued to increase under specifically assigned effective stresses behaved
even after the end of the primary consolidation (EOP). This differently compared to those without ageing. It can also be
observation is consistent with previous studies conducted observed that the void ratio decreased from that of the
by Afifi and Richart (1973), Kokusho et al. (1982), and corresponding virgin compression curve when ageing is
Athanasopoulos G.A. (1993). These studies indicated that allowed under specific effective stress (Figure 4, graph is
the increase in G’ after EOP cannot be explained by void zoomed in between effective stress of 1 and 100 kPa to
ratio change alone as observed by Anderson and Wood better data representation). Such change in void ratio at a
(1975). constant effective stress is attributed to the time-dependent
effect-creep (Leonards and Ramiah,1959; Perret,1995;
Leroueil et al.,1996). After a certain period of ageing at
1.6 120 constant effective stress, when loaded, the compression
Pore Water Pressure (kPa) curves with ageing did not follow the corresponding EOP
Elastic shear modulus (kPa) compression curves but crossed over and lied well on the
1.4 100 right side of the virgin compression curve. This observation
Pore Water Pressue (kPa)

for aged clay is consistent with previous studies reported

1.2 80 Elastic Shear Modulus (kPa) by Leonards and Altschaeffl (1964), Perret,1995; Leroueil
et al., 1996 for different types of clays, and Zeng et al.,
(2015) for dredged sediments. They attributed it to
1 60 structural resistance due to bonding or increase in friction
between clay particles and aggregates. However, beyond
0.8 40 a certain effective stress or vertical yield stress or apparent
preconsolidation pressure, the samples demonstrated a
noticeable decrease in void ratio. This illustrated the
0.6 20 phenomenon of destructuration, causing the samples to fall
close to the virgin compression curve. The effective stress
0.4 0 at the point where the e-log ’v curve changes curvature
0 20 40 60 80 was used to determine the value of the apparent
Time (days) preconsolidation pressure. Table 4 reports the value of
preconsolidation developed in Leda clay samples. It can be
Figure 2 PWP and G’ measurements in 1.88 LL sample observed that the ratio of apparent preconsolidation
under consolidation stress of 0.55 kPa. pressure to ageing consolidation stress increased as the
age of the sample increase, regardless of initial water
contents of the sample. This observation is consistent with
the previous study conducted on self-weight consolidated
2.5 400 Leda clays samples by Patel and Simms (2022). The 1.88
Pore Water Pressure (kPa) LL sample loaded at 0.55 kPa using modified consolidation
Elastic shear modulus (kPa) cell was reloaded after 80 days of ageing period and it
Elastic Shear Modulus (kPa)
Pore water Pressure (kPa)

2.0 reported distinct increase in pore water pressure and

300
reduction in void ratio at effective stress about 1.25 kPa.
1.5
200 Table 4 Development of preconsolidation pressure under
1.0 ageing consolidation stress

100 Initial Ageing Ageing Apparent pre- ′𝑝

0.5 GWC time consolidation consolidation ′𝑣
(days) stress (kPa) pressure
0.0 0 ’ (kPa) ’p
0 20 40 60 80 1.95 LL 50 6.87 12.4 1.80
Time (days) 1.52 LL 35 27.48 43.9 1.59
1.52 LL 51 27.48 49.5 1.80
Figure 3 PWP and G’ measurements in 1.53 LL sample
under consolidation stress of 1.63 kPa. 1.88 LL 80 0.55 1.25 2.27

3.2 Ageing under conventional oedometer

4 DISCUSSION Previous studies by Leroueil et al. (1996) and Perret (1995)
have also shown that the elastic shear modulus is directly
As mentioned earlier, this study expands on the preliminary linked to the preconsolidation pressure of the clay and can
study conducted on self-weight consolidated soft Leda clay be used to evaluate the structuration in clays.
samples by Patel and Simms (2022). Their study found a
dual mode of ageing for self-weight consolidated Leda clay 𝐺′ ′𝑝 [1]
samples, similar to Salam et al. (2023)’s findings. After =
′
𝐺𝐸𝐸𝑆 ′𝑣
reaching the “Early Equilibrium State (EES)”, the rate of
ageing in Leda clay samples was nearly constant The findings from this study also align well with data
regardless of their initial water contents. These findings reported in the literature for two different kinds of materials:
raised two important questions. 1) How does the EES soft soil and colloidal gel. Perret (1995) examined the
correlate to the consolidation process? and 2) Is the rate of development of G' in remolded Jonquière clay samples
ageing still constant under different loading conditions? (initial GWC of 2 LL) under effective stresses ranging from
Salam et al. (2023) primarily reported that the EES 0.9 kPa to 50 kPa over a 120-day period. The comparison
occurs close to the end of the primary consolidation, which of normalized G’ for all effective stresses revealed a nearly
is consistent with the observation made in this study. As constant rate of G’ development in Jonquière clay.
shown in Figures 2 and 3, the G’ reached some kind of Similarly, Manley et al. (2005) also reported a nearly
plateau close to the end of the primary consolidation and constant rate of ageing in colloidal gels prepared at
started to increase again afterward. This confirms that the different packing fractions.
EES occurs near the EOP or/and after reaching a certain
critical strain rate. To address the second question, Figure
5 presents the normalized G’ values for the loaded 1.88 LL
and 1.53 LL Leda clay samples, along with the normalized 4.5
values of G’ for self-weight consolidated Leda clay samples 2LL
1.51 LL
reported by Patel and Simms (2022).
4.0 1.55 LL
1.22 LL
1.88 LL@ 0.55 kPa
2.5 3.5 1.53 LL @ 1.63 kPa
Rate of ageing equation
Normalized G' (G'aged /G'EES)

3.0
2
2.5
void ratio

1.5 2.0

1.5
1.95 LL; 6.87 kPa-51 days
1 1.52 LL; 27.48 kPa-50 days 1.0
1.52 LL; 27.48 kPa-35 days
1.51 LL ; EOP
0.5
0.5 1 10 100
1 10 100 Time (days)
Effective stress (kPa)

Figure 4 Compression behaviour of Leda clay samples with Figure 5 Rate of ageing in Leda clay samples under
ageing and without ageing (EOP) different loading conditions (G’ data for self-weight
consolidated Leda clay samples (Black data points) was
Figure 5 shows that the rate of ageing remains nearly retrieved from Patel and Simms (2022)).
constant when samples aged under relatively smaller
consolidation stresses in the modified consolidation cell.
Notably, the ratio of normalized G’ for a given ageing time 5 NUMERICAL MODELLING
also appeared to be roughly similar to the ratio of
preconsolidation pressure to ageing consolidation stress, To analyze these experiments a code was implemented to
as reported in Table 4. This suggests that the rate of ageing simulate LSC based on the piece-wise linear framework
is also nearly constant at relatively higher consolidation developed by Fox and Berles (1997) which was also used
stresses. It also shows the linear relationship between in the existing UNSATCON model (Qi et al., 2017a; Qi et
elastic shear modulus and preconsolidation (Eq.1).
al., 2020). Elements were added to the code to account for For Leda clay samples, the NCL (or λ line) was
creep, structuration, and de-structuration in compressibility obtained from a conventional oedometer test, so two
behavior. In essence, compressibility is defined as a components of linear modulus, multiplier (a) and power (b)
function of the current void ratio and magnitude of were calibrated using NCL (Figure 7), where there is no
structure, in such a way that compressibility decreases by structure (𝜋 = 1) so compressibility is high, however by
decreasing void ratio (which can be decreased due to developing the structure compressibility decreases, and a
consolidation or creep deformation or both), and and b will change too. So, a and b must be also obtained
decreases/increases by increasing/decreasing magnitude by fitting to the recompression line (κ line) where the
of the structure. compressibility is low, and the magnitude of the structure
The parameter 𝜋 is introduced to quantify the is fully developed (𝜋 = 𝜋𝑚𝑎𝑥 ). The recompression index (κ)
magnitude of structure, which ranges from 1 for remolded is assumed equal to 0.05 for Leda clay (Figure 7).
soils (with no structure), to a maximum value (𝜋𝑚𝑎𝑥 ) when The measured elastic shear modulus over time for
the structure is fully developed. The 𝜋 is determined using different Leda clay samples with different initial conditions
𝑑𝜋 ′
Equations shown below, in which its rate ( ) is assumed emphasized that the ratio of
𝐺𝐴𝑔𝑒𝑑
can be reached to 3
𝑑𝑡 ′
𝐺𝐸𝐸𝑆
to be dependent on the current strain rate and itself. (Figure 5) so 𝜋𝑚𝑎𝑥 is assumed to be equal to 3. Then, to
integrate the effect of the magnitude of structure into
𝑑𝜋 𝑒̇ for 𝑒̇ > 𝑒̇𝑟 [ 2] compressibility behaviour, a linear function (a =
= 𝑓(𝜋, 𝑒̇ ) = −𝑛. 𝜋. l n ( )
𝑑𝑡 𝑒̇𝑟 123197(𝜋) − 119592) is adopted which relates the
calibrated a to the magnitude of the structure.
𝑑𝜋 𝑒̇
= 𝑓(𝜋, 𝑒̇ ) = 𝑚. (𝜋 − 𝜋𝑚𝑎𝑥 ). l n ( ) for 𝑒̇ < 𝑒̇𝑟 [ 3]
𝑑𝑡 𝑒̇𝑟

𝑑𝜋 for 𝑒̇ = 𝑒̇𝑟 [ 4]
= 𝑓(𝜋, 𝑒̇ ) = 0 2.5
𝑑𝑡
𝐸 = a. 𝑒 −𝑏 = 250000 (𝑒 −6 )
𝑑𝑒
where 𝑒̇ (= ) is strain rate, 𝑒̇𝑟 is defined as the
𝑑𝑡 2
reference strain rate (or the strain rate at EES or strain rate
on NCL), and m and n are constants that control the rate of (𝜅 = 0.05)
structuration and de-structuration respectively. For strain 1.5 (𝜆 = 0.2)
𝑑𝜋
Void ratio

rate above the reference strain rate, is negative which

𝑑𝑡 𝐸 = a. 𝑒 −𝑏 = 3605 (𝑒 −6 )
indicates de-structuration (structuration is destroyed), and
𝑑𝜋 1
for strain rates below the reference strain rate, would be
𝑑𝑡
positive which means structuration is developed. For strain Fitted linear compressibility (E line)
rates equal to the reference strain rate the magnitude of 𝜋 Logarithmic compressibility (𝜅 line)
0.5
𝑑𝜋
remains constant with no change in its rate ( = 0). Fitted linear compressibility (E line)
𝑑𝑡
The void ratio-effective stress relationship is described Logarithmic compressibility (𝜆 line)
using a linear modulus (E) which itself is a function of the 0
void ratio. A power function is adopted to define this linear 0.1 1 10 100 1000
modulus with two constants, “a” the multiplier, and “b” the Effective stress (kPa)
power that both are obtained by fitting using the normal
consolidation line. Therefore, any change in the void ratio
which is either associated with the consolidation or creep
Figure 7 Calibrating components of linear compressibility
deformation can be accounted for compressibility
modulus using compression line (λ line) and
behaviour. Figure 6 shows the schematic plots describing
recompression line (κ line).
the two approaches to relate void ratio and effective stress.
As it is illustrated, the λ relates the void ratio and effective
To integrate the effect of creep into LSC, the
stress logarithmically while E relates them linearly.
compressibility function needs to be modified. The creep
deformation delays the dissipation of excess PWP which
means for a certain amount of change in the void ratio, a
smaller increment of effective stress is developed
compared to a situation that ignores creep. Therefore, a
new parameter was added to the compressibility function
which modifies the increment of effective stress (𝑑𝜎𝑣′ )
corresponding to the increment of change in the void ratio
(de). The creep parameter is defined as a function of the
current magnitude of the structure, so once the structure is
small the value of creep is higher and more creep
deformations should be expected while for greater
Figure 6 Two different approaches to describe the void structure, the creep is smaller and, as soon as the structure
ratio-effective stress relationship. reaches to the maximum value it becomes zero. Eq. 5 is
the adopted function for the creep in time increment of dt Structuration rate 2.5E-8
which is then added to the compressibility function: De-structuration rate 8E-4

(𝜋𝑚𝑎𝑥 − 𝜋) [ 5] Creep Constant for creep 1E-7

𝐶𝑟𝑒𝑒𝑝 = . 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡. 𝑑𝑡 parameters adjustment (kPa)
𝜋𝑚𝑎𝑥

where the shifting of compressibility curves due to

creep strains is affected by the value of a constant. The
2.4 1.2
measured elastic shear modulus under self-weight
consolidation of Leda clay samples (1.22 LL, 1.51 LL and
2LL) (Figure 8) in 13 cm column test were then used to

PWP (kPa) in 1.53 LL @1.63 kPa

2 1

PWP (kPa) in 1.88 LL @0.55 kPa

calibrate the reference strain rate, structuration rate, and
constant. The calibrated parameters were then used for
simulating consolidation behaviour in the modified and 1.6 0.8
conventional oedometer tests. (Figure 9 to 11).
1.2 0.6
4.0
2LL _Measured
0.8 0.4
2LL_Modeled
3.5 1.51 LL_Measured
1.51 LL_Modeled 0.4 0.2
3.0 1.25 LL_Measured
Normalized G' (G'aged /G'EES)

1.25 LL_Modeled
2.5 0 0
0 20 40 60 80
Time (days)
2.0
1.53 LL@ 1.63 kPa _Measured
1.53 LL@ 1.63 kPa_Modeled
1.5 1.88 LL@ 0.55 kPa_Measured
1.88 LL@ 0.55 kPa_Modeled
1.0
Figure 9 Modeled and measured PWP in sample tested
under modified consolidation cell.
0.5
1 10 100
Time (days) 4
1.53 [email protected] kPa_Measured
Figure 8 Modeled and measured normalized G’ (Patel and 3.5 1.53 [email protected] kPa_Modeled
Normalized G' (G' /G'EES)

Simms,2022) in self-weight consolidated Leda clay 1.88 [email protected] kPa_Measured

samples. 3
1.88 [email protected] kPa_Modeled
2.5

2
Table 5 Input parameters for the aging model
1.5
Parameters
Compressibility Compression index (𝜆) 0.2 1
parameters Recompression index 0.05 0.5
(𝜅) 0 50 100
𝑎𝜆 3605
Time (days)
𝑎𝜅 250000
𝑏 -6
Figure 10 Measured Modeled values of normalized G’ for
Structuration/de- Initial magnitude of 1 samples tested under modified consolidation cell.
structuration structure (𝜋0 )
parameters Maximum magnitude of 3
structure (𝜋𝑚𝑎𝑥 ) 6 DISCUSSION
Reference strain rate 2E-7
The experimental study was conducted to expand on the
(1/s) previous findings by Patel and Simms (2022) and Salam et
al. (2023) by investigating the ageing behaviour of Leda 8 REFRENNCES
clay samples under different initial conditions. The major
finding from this study includes: Afifi, S.S., Richart, F.E.1973. Stress history effects on
shear modulus of soils. Soils and Foundations 13 (1),
1. The elastic shear modulus increased during both 77–95.
primary and secondary consolidation. However, Alam, M. K., Shahriar, A. R., Islam, M. S., Islam, N.,&
the increase in G’ during secondary compression Abedin, M. Z. 2021. Experimental investigation on the
cannot be explained due to void ratio change but it strength and deformation aspects of thixotropic aging in
is attributed to stronger physical-chemical bonding reconstituted clays. Geotechnical and Geological
between the clay particles due to ageing effects. Engineering, 39(3), 2471-2486.
2. The rate of ageing appeared to be nearly constant Aldaeef, A.A. 2015. Hydraulic performance of compacted
for the Leda clay samples tested under different clay liners (CCL) under simulated landfill conditions.
loading conditions, suggesting that the rate of Master’s thesis, Carleton University, Ottawa, ON
ageing could be independent of its initial state for Aldaeef, A.A., & Simms, P.H. 2019. Influence of pipeline
a given material. transport on sedimentation and consolidation of
3. For this Leda clay sample, the maximum likely flocculated Fluid Fine Tailings. Proceedings of
magnitude of ageing appeared to be up to three GeoSt.John’s 2019, 72nd Canadian Geotechnical
times the strength that is measured at EES. The Conference, Newfoundland, Canada.
EES was also found to be strongly related to the Anderson, D.G., Woods, R.D.1976.Time-dependent
end of the primary consolidation state. It should be increase in shear modulus of clay. Journal of the
mentioned that Patel and Simms (2022) identified Geotechnical Engineering Division, ASCE 102 (5),
“5” as a good average ratio between EES and 525–537.
maximum strengths for a wide range of soils. Athanasopoulos, G.A. 1993. Effects of ageing and
4. A linear relationship between elastic shear overconsolidation on the elastic stiffness of a
modulus and preconsolidation pressure was remoulded clay. Geotechnical and Geological
observed. Engineering.
5. The conventional large stain consolidation theory Banas, L. C. (1991). Thixotropic behaviour of oil sands
was modified to incorporate the time-dependent tailings sludge (master’s thesis). The University of
effects, and the ageing model was able to capture Alberta, Edmonton, Alberta.
the observed behavior quite well. Dunmola, A., Werneiwski, R.A., McGowan, D., Shaw, B.,
and Carrier, D. 2022. Geotechnical performance of fine
tailings in an oil sands pit lake. Canadian Geotechnical
Journal.60:595-610.
2.5
1.95 LL;6.87 kPa_51 days_Measured Fox, P.J. and Berles, J.D., 1997. CS2: a piecewise‐linear
1.95 LL;6.87 kPa_51 days_Modeled model for large strain consolidation. International
1.52 LL;27.48 kPa _50 days_Measured Journal for Numerical and Analytical Methods in
1.52 LL;27.48 kPa_50 days_Modeled Geomechanics, 21(7), pp.453-475.
2 Igbinedion, D. 2020. Ageing and Large-scale consolidation
in centrifuged Cake oil sands tailings (Masters Thesis),
Void ratio

Carleton University, Ottawa, ON.

Jeeravipoolvarn, S. 2005. Compression behaviour of
1.5 thixotropic oil sands tailings (Master’s thesis). The
University of Alberta, Edmonton, Alberta.
Jeeravipoolvarn, S., Scott, J. D., & Chalaturnyk, R. J. 2009.
10 m standpipe tests on oil sands tailings: long-term
1 experimental results and prediction. Canadian
5 50 Geotechnical Journal, 46(8), 875-888.
Effective stress (kPa) Kokusho, T., Yoshida, Y., Esashi, Y. 1982. Dynamic
properties of soft clay for wide strain range. Soils and
Foundations, 22 (4), 1–18.
Figure 11 Modeled and measured compressibility curves Leonards, G. A., & Altschaeffl, A. G. 1964. Compressibility
for sample tested under conventional oedometer. of clay. Journal of the Soil Mechanics and Foundations
Division, 90(5), 133-156.
Leroueil, S., Perret, D., and Locat, J. 1996. Strain rate and
7 ACKNOWLEDGEMENTS structure effects on the compressibility of young clay.
ASCE,61, 137-149.
The authors are grateful to Dr. Mohammad Rayhani, Manley, S., Davidovitch, B., Davies,R.N., Cippelletti, L.,
Associate Professor at Carleton University for supplying Bailey,A.E., Christianson, R.J.,Gasser, U., Parasad,V.,
Leda clay samples. This work is funded by an NSERC Serge, P.N., Doherty, M.P.,Sankaran, S., Jankovsky,
Collaborative Research and Development grant supported A.L., Shiley,B., Bowen, J.,Eggers, J., Kurta, C., Lorik,
by Canada ‘s Oil Sands Innovation Alliance (COSIA). T., and Weitz, D.A. 2005. Time-dependent strength of
colloidal gels. Physical review letters, PRL95,048302.
Miller, W. G. 2010. Comparison of geoenvironmental Zeng, L. L., Hong, Z. S., & Cui, Y. J. 2016. Time-dependent
properties of caustic and noncaustic oil sand fine compression behaviour of dredged clays at high water
tailings. (PhD Thesis, the University of Alberta). contents in China. Applied Clay Science, 123, 320-328.
Patel, A., Salam, A.M., Aldaeef, A.A., Simms, P.H. (2020). Zhang, X.W., Kong, L.W., Yang, A.W., Sayem, H.M. 2017.
Sample preparation and ageing in soft soils. Thixotropic mechanism of clay: a microstructural
Proceedings of GeoNiagara 2020, 73rd Canadian investigation. Soils Found. 57 (1), 23–35.
Geotechnical Conference, Ontario, Canada. Zhang, Y., Abusaid, A.H.,Pollock, G.W., and Uffen,
Patel, A., Aldaeef, A.A., Simms, P.H. 2021. Ageing in soft D.2022. Characterization study of oil sands tailings at
soils and clayey tailings. Proceedings of the Twenty- Suncor’s Pond 5 tailings storage factility. Proceedings
Fifth International Conference on Tailings and Mine of the seventh International Conference on Oil Sands
waste, 7-10 November 2021, Banff, Alberta, Canada. Tailings, 4-7 December 2022, Edmonton, Alberta,
Page no: 813-823. Canada. Page no: 16-23.
Patel, A., and Simms, P.H. 2022. On the magnitude and
rate of ageing in both soft soils and oil sands tailings.
Proceedings of the seventh International Conference
on Oil Sands Tailings, 4-7 December 2022, Edmenton,
Alberta, Canada. Page no: 263-275.
Perret, D. (1995). Diagénèse mécanique précoce des
sédiments fins du Fjord du Saguenay. Ph.D. Thesis,
Université Laval, Ste-Foy, Québec, Canada.
Qi, S., Simms, P. and Vanapalli, S. 2017. Piecewise-linear
formulation of coupled large-strain consolidation and
unsaturated flow. I: Model development and
implementation. Journal of Geotechnical and
Geoenvironmental Engineering, 143(7), p.04017018.
Qi, S. and Simms, P. 2020. Robust methods to estimate
large-strain consolidation parameters from column
experiments. Canadian Geotechnical Journal, 57(5),
pp.683-705.
Ren, Y., Yang, S., Anderson, K., Yang, Q. and Wang, Y.
2021. Thixotropy of soft clay: A review. Engineering
Geology,287,106097.
Salam, M.A., Simms, P.H., and Ormeci, B. 2023.
Ageing/thixotropy effects on strength and
compressibility in Leda clay and polymer-amended
clayey tailings. Engineering Geology, 317,1007088.
Schmertmann, J.H. 1991. The mechanical ageing of soils.
Journal of Geotechnical Engineering, V.117, P.26151
Shahriar, A. R., & Jadid, R. 2018. An experimental
investigation on the effect of thixotropic aging on
primary and secondary compression of reconstituted
dredged clays. Applied Clay Science, 162, 524-533.
Shahriar, A. R., Abedin, M. Z., & Jadid, R. 2018.
Thixotropic aging and its effect on 1-D compression
behavior of soft reconstituted clays. Applied Clay
Science, 153, 217-227.
Skempton, A.W. and Northey, R.D. 1952. The sensitivity of
clays. Geotechnique, 3(1), 30-53.
Suthaker, N. N., & Scott, J. D. 1997. Thixotropic strength
measurement of oil sand fine tailings. Canadian
Geotechnical Journal, 34(6), 974-984.
Tang, B., Zhou, B., Xie, L., Yin, J., Zhao, S., and Wang, Z.
2021. Strength recovery model of clay during
Thixotropy. Advances in Civil Engineering, Volume
2021, Article ID8825107.
Watabe, Y., Udaka, K., Nakatani, Y. and Leroueil, S. 2012.
Long-term consolidation behavior interpreted with
isotache concept for worldwide clays. Soils and
Foundations, 52(3), 449-464.
Yang, S., and Anderson, K.H. 2016. Thixotropy of marine
clays. Geotechnical Testing Journal, 39(2), 331-339.
Creation and segmentation of synthetic images
of granular materials
Harold Simo Tenekam, François Duhaime, Jean-Sébastien Dubé
Department of Construction Engineering – École de technologie
supérieure
Matthew Toews
Department of Systems Engineering– École de technologie supérieure,
Montréal, Québec, Canada

ABSTRACT
Software packages based on image analysis techniques may be used to determine the particle size distribution (PSD) of
coarse materials in the field based on photographs. Image segmentation is currently the main approach available to
determine the PSD. Segmentation is used to delimit each particle to distinguish the superimposed particles from the
isolated particles and the fine areas. The earliest approaches for image segmentation relied on applying Gaussian filters
to an image, while more recent techniques use deep neural networks. A major bottleneck in neural network-based
approaches is the availability of training data. This article proposes to create synthetic training photographs of granular
materials and their corresponding segmented images, without the need for any manual adjustment. We used the Unity
game engine to generate a large number of hyper-realistic training images, including their associated PSDs, segmentation,
and photographic depth of field, given parameters such as the desired PSD and the sample weight.

RÉSUMÉ
Les techniques d'analyse d'images peuvent être utilisés pour déterminer la distribution granulométrique (D.G.P.) des
matériaux grossiers sur le terrain à partir de photographies. La segmentation des images est actuellement la principale
approche disponible pour déterminer la DSP. La segmentation est utilisée pour délimiter chaque particule afin de distinguer
les particules superposées des particules isolées et des zones fines. Les premières approches de segmentation d'images
reposaient sur l'application de filtres gaussiens à une image, tandis que les techniques plus récentes utilisent des réseaux
neuronaux profonds. La disponibilité des données d'entraînement constitue un obstacle majeur pour les approches basées
sur les réseaux neuronaux. Cet article propose de créer des photographies synthétiques d'entraînement de matériaux
granulaires et leurs images segmentées correspondantes, sans qu'aucun ajustement manuel ne soit nécessaire. Nous
avons utilisé le moteur de jeu Unity pour générer un grand nombre d'images d'entraînement hyperréalistes, y compris les
DSP associées, la segmentation et la profondeur de champ photographique, en fonction de paramètres tels que la DSP
souhaitée et le poids de l'échantillon.

1 INTRODUCTION in numerous characterization projects (e.g. Bouchard

(2016) and Waterman (2014)). Although such software
Image segmentation is a method commonly used in produces satisfactory results, it is very often necessary to
computer vision that allows the classification of all the manually adjust the edges of the raw image to remove
pixels of an image by assigning a label to each one. The irrelevant backgrounds from the image. This is one of the
literature provides information on several segmentation reasons why Nurzynska and Iwaszenko (2020) explored
methods, including watershed segmentation. This method the use of image textures in different neural networks to
is commonly used to determine the contours of particles on automatically segment the contours of rock grains. They
a photographed soil sample. The purpose is to exploit local were able to perfectly segment 79% of the particle contours
maxima (points of interest per region) and contact zones present in a granular sample photograph using an artificial
between seeds to distinguish non-overlapping particles neural network. The distribution of different particle sizes in
from superimposed particles and delimit fine zones present a soil sample can be associated with the statistical
in the image. Bamford et al. (2021) and Campbell and distribution of segmented areas (Yaghoobi et al. 2019).
Thurley (2017) used this technique to assess particle size More robust machine learning models (e.g., convolutional
distribution (grain size) on images of granular materials. neural network) can be used to better differentiate image
Even though 95% of particles are generally well backgrounds from distinct categories of objects in the
segmented, manual adjustment of particle contours is image.
necessary in some images. This adjustment is possible There are CNN models trained to automatically
using simple tools such as Photoshop or the GrabCut segment images, including SegNet, a segmentation model
program. This step is important since the size of a particle composed of encoding and decoding networks and a per-
is related its equivalent diameter of each particle obtained pixel classification layer. In rock mechanics, poor
from its segmented area. segmentation will always involve a manual adjustment of
Some commercial optical grading systems such as the image before its use (Bamford et al., 2021). This
Wipfrag (Maerz et al. 2018) use this segmentation principle motivated Ramezani et al. (2018) and Yamada and Di
to characterize images of granular media. Wipfrag is used Santo (2022) to develop deep neural network (DNN)
models capable of accurately and automatically delineate varied sizes depending on the granular characterization
granular materials. Convolutional neural network (CNN) is required. Blender is the most suitable software for
one of the subdomains of DNNs that has demonstrated generating realistic 3D models of sand, rock, brick, asphalt
excellent performance in object classification and etc. In this study, 200 different grain models designed on
detection. He et al. (2018) developed Mask R-CNN Blender were used to generate synthetic images. Some
(Regional Convolutional Neural Network Mask) which examples are shown in Figure 1. Csharp scripts were then
extends FasterR-CNN, an instance segmentation created to specify parameters such as the number of
framework that has been able to correctly detect rock particles to be generated per sample, the creation time for
fragments on UAV images of granular environments to each sample or even the method for destroying each
predict their size using neural networks (Schek et al. 2019). sample after it has been created.
Training this type of neural network is usually very time
consuming (several days, depending on data size).
Transfer learning can be used to adjust a pre-trained model
to optimize it on a specific database in a more reasonable
amount of time. Although this method can be used on
various datasets, training and fitting a CNN model still
requires thousands of images.
A potential solution to quickly overcome data collection
is the use of computer graphics. Temimi (2020) and Pirnia
et al. (2018) respectively developed program on Unity and
YADE (discrete element code) that generate synthetic
images of granular materials to train artificial neural
networks to predict the grain size of a photographed soil
sample. A complementary step to this process would be to
a)
segment the images before incorporating them into the
neural network, to facilitate the extraction of particle-size-
specific features from the sample. These optimistic
methods have a number of limitations, not least the manual
labeling of the data used to train the neural networks.
Automating the process of annotating the data using
synthetic images could make it easier to estimate grain size
from their segmentation. In this article we will focus on the
creation of a program to generate realistic images of
granular materials, and then develop an automatic and
accurate segmentation model of these synthetic images. b)

2 METHODS
Figure 1. a) Examples of 3D models of soil particles, b)
scene of the assembly developed on Unity
2.1 Dataset
2.1.2 Functional and non-functional requirements of the
2.1.1 Unity Engine
program
Unity is a well-known development platform for video game
The purpose of the program presented herein is to produce
design. It was used to create the 3D model of granular
samples whose PSD is provided via a '.txt' file. The file
materials presented in this paper. A similar approach was
contains the percentage by mass of particles retained on
developed by Temimi (2020) for creating 3003 images with
each sieve according to the size of the different control
individual particle sizes ranging from 75 μm to 1180 μm.
sieves of the BNQ 2501-025 standard, which describes
Moreover, the 3D environment provided by Unity can be
particle size analysis by sieving. These values will be used
manipulated in different ways. A gravitational field can be
to randomly generate a particle size according to the
applied to all particle models which can influence the
proportion given. For example, for a sample of which 30%
velocity at which a particle falls into the container (3D bin
of the mass passes through the 5 mm sieve and none
used to store our synthetic samples), depending on its size
through the 2.5 mm sieve, the program will have to
and density. It is also possible to model 3D objects and
generate an equivalent number of particles with diameters
control the luminosity in the container. Unity’s compiler
randomly selected in the interval [2.5 mm, 5 mm [, so that
considers Csharp scripts (developed on Visual Studio
the mass of all the particles generated corresponds to 30%
Code) for the animation and the behavior to be applied on
of the sample.
a scene (a space in which all elements are placed).
For this, the program goes through several steps to
Several elements are required to generate images of
produce the images. First, it assigns a mass and a diameter
granular materials using Unity. First, a set of 3D grains of
to each particle. It is important to know the volume of each
different morphologies (e.g., angular, sub-angular,
particle in order to estimate its equivalent mass. In the
rounded, sub-rounded) is required, which can take on
present case, the morphological irregularity of each 3D
particle model forces approximations in the estimation of where 𝑁𝑖𝑗 represents the number of particles to be
the volume of each particle. The 3D models have been generated for each particle size and mineralogical class,
considered as spheres with an equivalent diameter D. This mT is the total mass, xi is the proportion of the total mass
is done without having a major impact on the overall occupied by size fraction i, 𝑦𝑖𝑗 is the proportion of size
characterization of the sample (Cordary, 1994). fraction i occupied by mineralogical fraction j, di is the
Then the program fixes the total mass of the sample
according to the largest particle size present in it. The
weight of synthetic samples will therefore be 1000 g if the
diameter of the largest particle in the sample is less than
5 mm. Otherwise, the sample will have a mass of 3000 g.
For each particle size range, these masses correspond to
the quantities of sample required to fill a 160x250 mm
container, the average drop height of each particle is
10 cm. Finally, the program takes photographs of the top
and bottom of the scene to account for segregation inside
the container. The added value of bottom photographs is
demonstrated in
Figure 2. Fine particles tend to accumulate at the
bottom of the container, and coarser particles stay on top.
All particles are generated in a set of random positions,
all within a 50x130x230 mm cube placed 100 mm above
the container. Particles are generated over time shifts of
0.05 s. Each particle is then equipped with a 'collider' made
up of a set of polygons that allow the complete surface of
all the particles to be discretized, despite their irregularities
(Figure 2). Because of this, Unity is able to detect all
collisions taking place in the container and, in turn, simulate
the segregation of particles in the scene. The photographs
have a resolution of 3840 x 2160 pixels. This program is Top PSD
completely autonomous once executed and does not Bottom PSD
require any manipulation by the user. PSD request
Each particle is surrounded by a collider to detect
contact with other particles in the container. An
approximate mass was assigned to each particle based on
its approximate volume and density. A time offset of 0.05 s
was set for each frame during program execution. Events
will occur after 0.05 s in the scene to give the processor
time to capture the collisions of the thousands of particles
generated. Once executed, the program will take around
10 minutes to generate a sample containing an average of
15,000 particles. The program requires a multitude of
recommended features for proper operation. A Core i7
2.8GHZ x8CPUs processor, 16.0 GB unified random-
access memory (RAM), a 4 GB dedicated graphics card
and a screen with 4k resolution.
Figure 2. a) Top and bottom sampling in our container, b)
2.1.3 Synthetic images Particle sizes corresponding to each sub-sample

The sizes of all our 3D objects have been dimensioned in

mm on Blender. An image is generated according to the
proposed particle size. For each proportion of particle size
classes, the equivalent number of particles to be generated
is calculated by a simple quotient between the total mass
for this class and the average mass of a rock grain of this
same class, as given by equations 1 and 2.

𝑚𝑇 𝑥𝑖 𝑦𝑖𝑗 𝑚𝑇 𝑥𝑖 𝑦𝑖𝑗
𝑁𝑖𝑗 = = 4 3 [1]
𝑣𝑖 𝜌𝑖𝑗 𝜋𝑟 𝜌
3 𝑖 𝑖𝑗
Figure 3. Digital modeling of a 3D model
𝑝
𝑁 = ∑𝑛𝑖=1 ∑𝑗=1 𝑁𝑖𝑗 [2]
average particle diameter in size fraction i and 𝜌𝑖𝑗 is the semantic segmentation of the image and even the
particle density of mineralogical fraction j in size fraction i. segmentation of instances of the camera rendering.
N is the total number of particles to be generated for a Semantic image segmentation involves assigning the
sample, p the number of mineralogical classes and n the same color label to all pixels in an object category present
number of particle size classes. Subsequently, all 𝑁𝑖𝑗 in the image. However, instance segmentation assigns a
values are converted to integers since they must represent distinct color label to all objects of the same category in the
particle quantities. Furthermore, although approximations same image.
are used to determine the mass and volume of each The advantage of this method is really the precision of
particle, once the sample has been generated the program the different segments of all the particles. There are 8
recovers the true particle size distribution at the end of each different camera views in Unity from which the depth of the
synthetic image. Figure 5 shows a proposed particle size photographic field, the semantic segmentation, the
and the approximate particle size generated by the segmentation of instances is used herein. A unique
program as well as the top and bottom images of the identifier is assigned to each clone of the 3D grain model
container. when it is generated. Once synthetic images equipped with
An approach was used to check the visual similarity of our their perfectly segmented images are obtained, it is
synthetic data with real samples. The aim here is to possible to generate a dataset that can be used to create a
generate synthetic images using the particle size neural network model capable of segmenting real images
characteristics of real samples available in the laboratory. of granular materials. Yaghoobi et al. (2019) associated the
Considering a specific real sample, after sieving it PSD from a photographed sample with the statistical
according to BNQ standards, we then use its particle size distribution of colors on its segmented image.
curve as an input parameter for our program. Finally, we
simply compare the actual photographs of the sample
taken in the laboratory with the synthetic images generated 3 RESULTS AND DISCUSSIONS
from its characteristics. We used the Xception neural
model to extract the characteristics of each image. The In our program, fine particle size samples present a real
characteristics of these images were then simplified using challenge, given the considerable number of particles to be
PCA (principal component analysis), a useful method for generated (millions of particles). In this paper. the focus is
reducing data size and facilitating analysis. Figure 4 shows on coarser particles (diameter greater than 2.5 mm), but
calibration process of our synthesis images. the program is still capable of generating samples
composed of finer particles if larger computing capacities
2.2 Image segmentation are available. Figure 6 shows the overall output of the
granular material generator. Figure 6a shows the raw
There are several approaches to segmenting an image., synthetic images initially generated under the particle size
Neural networks may have the capacity of segmenting curves shown in Figure 6d. Figures 6c and 6d show
images of granular materials perfectly and automatically respectively the segmented images as well as the depths
(Sylvester 2023). However, this requires a good database of each image that was deemed relevant to generate since
of soil photographs on the one hand, and their the superimposed particles will have a lower gray level than
segmentations on the other. One of the advantages of the non-superimposed particles. The creation of these
synthetic data is that it is possible to perfectly segment all synthetic images can enrich databases to train neural
the images generated. Unity has several screens that networks to predict various geotechnical characteristics. In
represent different display modes. The default mode doing so, the photographic depth of field may have an
presents the real rendering of what the camera visualizes. adjustment interest like the Xc and N adjustment factors
It is also possible to visualize the depth of the image, the used in Wipfrag to correct predicted fragmentation curves
(Maerz et al. 1996). Including more data can only improve pixels per mm) in a container of dimension 250x160 mm in
the performance of data networks. In environmental a Unity scene. An identifier was assigned to each particle
applications, the program can be used to create images of to exploit the Unity instance segmentation display mode.
soil contaminated with waste particles such as wood, brick, Finally, the rendering of the scene is captured to generate
concrete and asphalt residues. Figure 7a shows a synthetic the synthetic images as well as the segmented images and
image of a sample containing 30% asphalt by mass. the corresponding depths for each image.
Figures 7b and 7c show a perfect superposition between This program has the advantage of being able to
the original image and the segmented image. correctly label finer particles down to 0.080 mm. The
particle size of a soil sample could be associated with the
statistical distribution of the segmented areas. It would be
meaningful to incorporate our synthetic images and their
corresponding segmentations into pre-trained
convolutional neural networks such as SegNet or Segment
Anything Model (SAM), to automate the segmentation of
real soil photographs.

Figure 5. a) synthetic image of the top, b) synthetic image

of the bottom, c) particle size analysis

4 CONCLUSION

This article presents the creation and segmentation of

synthetic images of granular materials using the Unity
game engine. A program was developed to automatically Figure 7. a) image of a sample contaminated with 30%
generate different grain sizes referred to the BNQ 2501-
asphalt residue, b-concatenation of the synthetic image
025 control sieve size. Two hundred (200) 3D particle with its segmented image, c) perfect superposition of the 2
models were used to generate the different grain
images
morphologies in the program. These grain sizes were used
to create the corresponding synthetic images (scale 11. 26
5 REFERENCES fragmented rock size distribution using textural feature
extraction of images. Powder Technology, 342: 630-
Bamford, T., Esmaeili, K. and Schoellig, A. P. 2021. Deep 641.
learning approach for rock fragmentation analysis. Yamada, T. and Di Santo, S. 2022. Instance Segmentation
International Journal of Rock Mechanics and Mining of Piled Rock Particles Based on Mask R-CNN. IEEE
Sciences, 145: article 104839. International Geoscience and Remote Sensing
Bouchard, J. 2016. Caractérisation de l'enrochement du Symposium, pp. 163-166.
barrage Romaine-2. Master’s thesis. Université Laval.
Bureau de normalisation du Québec, B. 2013. BNQ 2501-
025. Sols–Analyse granulométrique des sols
inorganiques.
Campbell, A. D. and Thurley, M. J. 2017. Application of
laser scanning to measure fragmentation in
underground mines. Mining Technology, 126: 240-247.
Cordary, D. 1994. Mécanique des sols. Lavoisier-Tec et
Doc, Paris, France.
Manashti, J., Duhaime, F., Toews, M. and Pirnia, P. 2021.
Predicting the Particle Size Distributions of Spherical
Particle Sets from Synthetic Images: A Comparison of
9 Classic Image Features.
https://doi.org/10.31224/osf.io/phjkm.
He, K., Gkioxari, G., Dollár, P. and Girshick, R. 2018. Mask
R-CNN. https://arxiv.org/abs/1703.06870v3.
Lang, N., Irniger, A., Rozniak, A., Hunziker, R., Wegner, J.
D. and Schindler, K. 2021. GRAINet: mapping grain
size distributions in riverbeds from UAV images with
convolutional neural networks. Hydrology and Earth
System Sciences, 25(5): 2567-2597.
Maerz, N. H., Palangio, T. C. and Franklin, J. A. 2018.
WipFrag image based granulometry system.
Measurement of Blast Fragmentation, Routledge, pp.
91-99.
Manashti, M. 2022. Grain size analyses of granular
material based on image analysis techniques and
machine learning. Ph.D. thesis. École de technologie
supérieure.
Nurzynska, K. and Iwaszenko, S. 2020. Application of
texture features and machine learning methods to
grains segmentation in rock material images. Image
Analysis & Stereology, 39(2): 73-90.
Pirnia, P., Duhaime, F., and Manashti, J. 2018. Machine
learning algorithms for applications in geotechnical
engineering. Proceedings of the 71st Canadian
Geotechnical Conference. Edmonton AB: paper 339.
Schenk, F., Tscharf, A., Mayer, G. and Fraundorfer, F.
2019. Automatic muck pile characterization from UAV
images. ISPRS annals of the photogrammetry, Remote
Sensing and Spatial Information Sciences, 4(W5): 163-
170.
Sylvester, Z. 2023. segmenteverygrain: A SAM-based
model for instance segmentation of images of grains.
https://github.com/zsylvester/segmenteverygrain.
Temimi, M. 2020. Création d’images de matériaux
granulaires avec la plateforme de développement de
jeux Unity et prédiction de la granulométrie à l’aide de
textures et de réseaux de neurones. École de
technologie supérieure.
Waterman, T. (2014). The application of an underground
optical fragmentation analysis method at the Lisheen
Mine.
Yaghoobi, H., Mansouri, H., Ebrahimi Farsangi, M. A. and
Nezamabadi-Pour, H. 2019. Determining the
Evaluation of the Effect of Nanofibrillated
Cellulose on the Drying and Dewatering
Behaviour of Oil Sands Tailings
Yunhai Zhang, & Nicholas Beier
Department of Civil and Environment Engineering – University of
Alberta, Edmonton, AB, Canada
Keith Gourlay & Gurminder Minhas
Performance BioFilaments Inc., Vancouver, BC, Canada

ABSTRACT
As a sustainable material, Nanofibrillated Cellulose (NFC) is increasingly used in reinforcing earth-based materials due to
its unique mechanical and physical properties. It has been used in enhancing the geotechnical properties of oil sands
tailings for the first time. Preliminary study has found that NFC can improve the shear strength of oil sand fluid fine tailings
(FFT). However, the use of NFC would potentially change the dewatering behavior of FFT in the meantime. This current
study aims to evaluate the influence of NFC addition on the geotechnical properties of FFT, including drying behavior,
shear strength gain, soil-water characteristics, and dewaterability. A favorable treatment method for the optimal dose was
determined using 2 different sources of FFT based on the near-surface shear strength gain subject to atmospheric drying
in the lab. Capillary suction time (CST) and pressure filtration tests were carried out to evaluate the influence of NFC on
the immediate dewatering and filterability of FFT. In addition, laboratory experiments using Tempe cell, Pressure plate
extractor, and WP4 Dewpoint Potentiometer were conducted to assess the water retention behavior of the FFT samples,
which demonstrate the influence of NFC on the desaturation of FFT. Consequently, it was noted that adding NFC would
potentially affect the evaporation rate, CST, specific resistance to filtration (SRF), and the desaturation point of FFT.
However, the impact on these properties varies between FFT with different sand to fine rate (SFR) or fines content.

RÉSUMÉ
En tant que matériau durable, la Cellulose Nanofibrillée (NFC) est de plus en plus utilisée pour renforcer les matériaux
géotechniques à base de terre en raison de ses propriétés mécaniques et physiques uniques. Elle a été employée pour la
première fois afin d'améliorer les propriétés géotechniques des résidus de sables bitumineux. Une étude préliminaire a
révélé que la NFC peut accroître la résistance au cisaillement des résidus fins fluides de sables bitumineux (FFT).
Cependant, l'utilisation de la NFC pourrait potentiellement modifier le comportement de déshydratation des FFT
simultanément. Cette étude actuelle vise à évaluer l'influence de l'ajout de la NFC sur les propriétés géotechniques des
FFT, notamment le comportement de séchage, le gain de résistance au cisaillement, les caractéristiques sol-eau, et la
capacité de déshydratation. Une méthode de traitement favorable pour la dose optimale a été déterminée en utilisant deux
sources différentes de FFT, en se basant sur le gain de résistance au cisaillement en surface soumis à un séchage
atmosphérique en laboratoire. Des essais de Temps de Succion Capillaire (CST) et de filtration sous pression ont été
réalisés pour évaluer l'influence de la NFC sur la déshydratation immédiate et la filtrabilité des FFT. De plus, des
expériences en laboratoire au moyen d'une cellule Tempe, d'un extracteur à plaque de pression, et d'un Potentiomètre
WP4 pour le Point de Rosée ont été conduites afin d'appréhender le comportement de rétention d'eau des échantillons de
FFT, démontrant ainsi l'impact de la NFC sur la désaturation des FFT. Par conséquent, il a été noté que l'ajout de NFC
pourrait potentiellement affecter le taux d'évaporation, le CST, la résistance spécifique à la filtration (SRF), ainsi que le
point de désaturation des FFT. Toutefois, l'incidence sur ces propriétés varie en fonction du taux de sable par rapport aux
fines (SFR) ou de la teneur en fines des FFT.

1 INTRODUCTION on dewatering and increasing the strength of FFT deposits.

The application of additives like coagulants and flocculants
In northern Alberta, the footprint of tailings storage have been extensively evaluated for these purposes
facilities is continuously expanding. Fluid fine tailings (FFT) (Jeeravipoolvarn et al. 2009, Beier et al. 2013, Mizani et al.
contains mainly silts and dispersed clay with poor 2017, Afacan et al. 2019). Flocculants contribute to
consolidation properties and low strength, which has led to bonding tailings particles, forming non-segregating
a huge challenge in the management, reclamation, and deposits because of their bridging effect, thus, enhancing
closure of tailings storage facilities. To address this the dewatering, consolidation, and strength gain of tailings
growing issue, tailings treatment technologies including (Vajihinejad et al. 2018). Nanocellulose was found as a
inline flocculation, thin lift drying stack, and tailings cross-linking agent, which is competent in bringing
thickening strategies such as centrifuge and filtration have particles together in some scenarios (Martin et al. 2017).
been developed and effectively utilized in the oil sand Due to their unique properties such as high strength and
industry. The main focus of these technologies is to reduce stiffness, wood-derived cellulosic nanomaterials have been
the volume of FFT and accelerate the reclamation and recognized as strength agents to improve the mechanical
closure. For terrestrial closure, technologies should focus strength of paper, film, coating, etc. (Osong et al. 2015).
 Typically, nanofibrillated cellulose (NFC) can be fine content was obtained based on the results of PSD
fibrillated down to 10-10 nm in width and up to several while bitumen was excluded. The sand to fine ratio (SFR)
micrometers in length depending on the source of was determined as 0.04 and 0.25 of FFT(a) and FFT(c),
cellulose. Besbe et al. (2011) studied the reinforcing respectively. The methods for determining the properties of
potential of one type of NFC isolated from alfa with a width FFT are listed below, more detailed information could be
ranging from only 5 to 20 nm. Undoubtedly, with such a found in the reference documents:
large aspect ratio, and high strength, NFC can form a • Water content and solid content was determined
highly porous mesh with high tensile strength to strengthen using the geotechnical definition by laboratory oven
earth structures. Correia et al. (2018) discovered that NFC dry for 24 hours under 105 ℃ (ASTM D2216-19).
could be used as particle packaging agent to improve the • Bitumen content was determined using Dean stark
adhesion of cement particles with the property of high analysis (Dean & Stark 1920).
specific surface area, then further decrease the growth of • Clay content was calculated using equation (1)
microcracks in cement-based material, therefore improving (Kaminsky 2014) with the Methylene Blue Index
its integrity and strength. Stanislas et al. (2022) (MBI) determined by the titration end-point method.
demonstrated that adding NFC into earth-based
composites increased the density and flexural strength and 𝑀𝐵𝐼(
𝑚𝑒𝑞
)+0.04
reduced the apparent void volume of the material. %𝑐𝑙𝑎𝑦 = 100𝑔
[1]
0.14
Referring to the capacity of bridging particles combined
with the eco-friendly nature of NFC which can be obtained
from most vegetation, it may be considered a potential • Atterberg Limit, determined according to ASTM
reinforcing agent to improve the geotechnical properties of D4318-17.
oil sands tailings. Zhang et al. (2022) applied NFC to oil • Specific gravity was determined using vacuum
sands tailings and found that NFC can lead to an increase water pycnometer (ASTM D854-14).
in the shear strength of FFT during atmospheric drying. • PSD was determined using laser diffraction.
Since NFC is derived from wood, it has been reported
as a hydrophilic material which holds water theoretically
(Shaker et al. 2018, Durand et al. 2021). However, Salama Table 1. Characteristic properties of the FFT samples
et al. (2021) investigated the adsorption properties of NFC
in water wastewater treatment, this nanoflocculant induces
Characteristics FFT(a) FFT(c)
the bridging of the suspended particles in contaminated
wastewater by neutralizing the surface charge of the Water content (%)
1
152 224.6
particles. It is known that FFT is fine-abundant material, Solid content (%) 42.8 30.8
and the negatively charged clays are the main problem in Bitumen content (%) 5.75 2.68
the dewatering of FFT in particular. Thus, despite the MBI (meq/100g) 6.6 6.4
extraordinary performance in enhancing the shear strength
Clay content (%) 47.7 46.1
of soil-based materials and bringing suspended particles in
wastewater, the impact of NFC on the dewatering of oil Fine content (%) 96 80
sands tailings should be further evaluated. Capillary Liquid limit (%) 49 45
suction time (CST) test as an effective method has been Plastic limit (%) 34 22
largely used for evaluating the immediate dewatering of Plastic index (%) 15 23
tailings, a lower CST correlates with higher dewaterability
Specific gravity 2.08 2.25
(Yao 2016, Oliveira et al. 2018, Aldaeef & Simms 2019).
The benchtop pressure filtration tests were carried out to
1
water content was calculated using geotechnical definition, the
determine the filterability which in turn indicates the ratio of solids to mass of water.
dewatering behavior of oil sands tailings (Xu et al. 2008,
Loerke et al. 2017, Qureshi et al. 2021). In this study, the
impact of NFC on the dewatering behavior of FFT was
investigated utilizing CST and benchtop pressure filtration.

2 MATERIAL AND METHODS

2.1 Materials

In this study, 2 different sources of FFT are used. The

initial characteristic properties of the FFT samples are
tabulated in Table 1, including water content, solid content,
bitumen content, clay content, fine content, Atterberg limit,
and specific gravity. Figure 1 shows the particle size
distribution (PSD). The definition of fines used in this study
is particles with a diameter smaller than 0.044 mm, as clay
are particles with a diameter smaller than 0.002 mm. The
Figure 1. Particle size distribution of the FFT samples
Figure 2. Schematic description of the manufacturing NFC product (image provided by BioFilament Inc.)

NFC is a product derived from natural wood. Figure 2 Brookfield DV3T rheometer was used to measure the near-
illustrates the process of disintegrating NFC from wood surface shear strength which is more accurate for
stem to fibers with width of 10-100 nm and up to several measuring shear strength up to 8 kPa. Subsequently, a
micrometers in length using mechanical shearing with laboratory geotechnical vane shear apparatus was
various pre-treatment strategies. employed to measure shear strength over 8 kPa.
A pressure filtration test and CST test were also utilized
2.2 Testing Methods for evaluating the impact of NFC on the water release
behavior of FFT. In this trial, only a FFT control sample and
The experiments were conducted in 4 phases including the NFC-FFT dose with better performance in enhancing
samples preparation and 2 different types of laboratory the shear strength gain of each FFT determined in the
tests including atmospheric drying test, pressure filtration atmospheric drying tests were tested. For each pressure
test and CST test, and the SWCC test. Figure 3 illustrates filtration test, 300 mL of samples were added into Millipore
the experiments for each phase. Sigma filtration system which can accommodate up to 500
The initial mass fraction of the NFC dilution as received g of samples with a diameter of 142 mm. A Millipore Sigma
was 9.6 %, which was then diluted to 4 % for the following glass fiber filter paper was placed at the bottom of the unit.
tests based on Zhang et al. 2022 findings. The criteria for The pore size of the filter paper is 0.7 µm with a thickness
using 4 % NFC dilution and the doses were concluded from of 0.475 mm and a flow rate of 6 mL/min /cm2. The samples
the previous study, more information could be found in the were filtered with a compressed air supply of 400 kPa over
paper (Zhang et al. 2022.) To prepare 4 % dilution, the around 3.5 hours. The filtered water was collected in a
initial NFC sample was mixed in a beaker with distilled beaker placed on a scale, and the mass was recorded.
water using a benchtop mixer at 200 rpm for 1 hour. To After each test, the filtered samples were removed, and
prepare NFC-amended FFT samples (NFC-FFT), the 4 % oven dried to determine the final solid content. The ratio of
NFC dilution was added into 500 mL homogenized FFT time over the volume of filtrates (t/V) was tabulated in the
samples at the required dose and mixed at 600 rpm for 360 t/V vs. V plot. The specific resistance was then calculated
s in a beaker using a 4-blade impeller (Zhang et al 2022). using equation (2) (Qureshi et al. 2021). Additionally, the
The NFC: FFT doses used in this study are 0.05 %, 0.1 %, Triton Electronics type 319 Multi-CST apparatus with 5
0.15 %, and 0.2 % of each sample, which was calculated multi-radii testing heads was employed to measure the
based on the ratio of the dry mass of NFC to the dry mass capillary suction time of the samples using Whatman
of FFT solids. Grade 17 Chroma filter paper whose pore size is 8 µm with
Following sample preparation, atmospheric drying tests a thickness of 0.92 mm and a flow rate of 6.3 mL/min/cm2.
were undertaken for evaluating the impact of NFC on the
shear strength and drying behaviour of FFT. The prepared
NFC-FFT samples were placed in an open area on a 𝑆𝑅𝐹 =
2𝐴2 ∆𝑃
𝑏 [2]
laboratory workbench to confine the ambient conditions. A 𝜇𝑓 𝑐
water control sample and 2 untreated FFT control samples
with the same volume of 500 mL as treated samples in
identical containers were also included in this trial. The Where:
mass change of the samples was measured every day to 𝑆𝑅𝐹 = specific resistance to filtration (m/kg).
determine the solid content, while the shear strength was ∆𝑃 = pressure drop, a constant head of 400 kPa was
measured on a weekly basis. Meanwhile, the ambient used in this study.
temperature and the relative humidity were recorded. The 𝐴 = area of filter in m2.
actual evaporation (AE) of NFC-FFT samples and FFT 𝜇𝑓 = viscosity of filtrate.
control samples were then compared with the potential 𝑐 = mass of dry cake solids per filtrate volume in kg/m3.
evaporation (PE) derived from the water control sample to 𝑏 = slope of t/V versus V plot.
evaluate the influence of adding NFC on the evaporation
behavior of FFT. In the early stage of the drying test, a
Figure 3. Illustration of staged testing

Finally, SWCC tests and shrinkage tests for evaluating relationship between the void ratio and gravimetric water
the impact of NFC on the soil-water behavior of FFT were content of the samples (Fredlund et al. 2019). Combining
conducted. Similar to the pressure filtration test and CST SWCC tests and shrinkage tests, the SWCC of volumetric
test, the FFT control and doses with the best drying water content vs. suction, or saturation degree vs. suction
performance were tested. To obtain the SWCC of the was derived using curve fitting (Fredlund & Xing, 1994).
samples, 3 types of devices were utilized to determine the
suction in different ranges including a Tempe cell for
measuring the suction below 400 kPa, pressure plate 3 RESULTS AND DISCUSSION
extractor for measuring the suction between 500 kPa to
1500 kPa, and WP4 Dewpoint Potentiometer for measuring 3.1 Atmospheric Drying Test
the suction over 1500 kPa. In the Tempe cell and the
pressure plate tests, each load step was completed when Figure 4 illustrates the development of shear strength
there was no significant change in the mass of the samples vs. the solid content for control FFT samples and NFC-FFT
was observed. Then a relationship between gravimetric samples dosed at 0.05, 0.1, 0.15, and 0.2 wt.% subject to
water content and the suction matrix of the sample can be atmospheric drying. The lines in the figures are the results
obtained. A detailed test procedure for determining the over 14 days while the near-surface shear strength of the
SWCC can be found in ASTM D6836-16. The shrinkage samples was measured every 7 days. The first data point
test was completed using approximately 137 g samples in on each line was measured using a Brookfield DV3T
the Shelby ring with a diameter of 63 mm and 32 mm in rheometer immediately after the samples were mixed (Day
height. The bottom of the Shelby ring was sealed using 0). After being mixed with NFC dilution, all the samples
petroleum jelly to avoid leaking liquid samples. The change remain fluid, the initial yield stress of NFC-FFT samples is
of mass and the dimension of the samples were recorded around 12 Pa which is only 1 ~2 Pa higher than that of FFT
twice each day until no significant mass change was control samples. As the drying commenced, the solid
observed. Then the results were used to characterize the content and shear strength both increased over time.
However, neither the shear strength gain nor the solid Figure 5 illustrates the drying behavior of the samples
content was in a linear relationship with the dose of NFC- over 29 days through the ratio between the Actual
FFT sample. The highest near-surface shear strength evaporation (AE) and Potential evaporation (PE). In the
among FFT(a) samples on day 14 was found as 47 kPa of earlier drying stage, the evaporation is mainly controlled by
FFT(a) control. However, it is 45 kPa of NFC-FFT(a) dosed ambient conditions, the surface moisture is allowed to
at 0.15 wt.%, which is greater than the other NFC-amended evaporate freely as open water due to the saturation of the
FFT(a) samples. However, the difference in the near- FFT samples (Kabwe et al. 2022). However, a value of
surface shear strength between FFT(a) control and NFC- AE/PE lower than 1 is observed for all the samples at the
FFT(a) at optimum dose is not obvious when the difference beginning. The potential reason is that a thin film of
in solid content between these two samples is 4 %. It is bitumen formed at the surface of the samples due to
believed that FFT would achieve a significant strength gain mixing. As can be clearly observed in Figure 6, the black
when its solid content is above 70 % during drying substance on the surface of the samples is bitumen. The
(Sobkowicz and Morgenstern, 2009). The small difference actual evaporation increases as the roughness or
in the near-surface shear strength between FFT(a) control microcracks at the samples surface due to the menisci
and NFC-FFT(a) dosed at 0.15 wt.% indicates the between the particles enlarge the evaporation surface
reinforcement effect of NFC. In Figure 4-(2), due to the (Tollenaar, 2018). The AE/PE increases to between 0.7
higher initial water content of FFT(c), the solid content of and 1.2 subsequently before decreasing. The evaporation
the samples is relatively lower than samples prepared with rate started decreasing accordingly in the stage where the
FFT(a) on day 14. The highest near-surface shear strength evaporation is mainly controlled by the properties of the
is found at 4.6 kPa for NFC-FFT(c) dosed at 0.5 wt.% with samples, such as hydraulic conductivity (Kabwe et al.
a solid content of 63 %. 2018). In Figure 5-(1), the AE/PE of the FFT(a) control
samples start to drop earlier than NFC-FFT samples.
Whilst the AE/PE of NFC-FFT(c) samples start to decrease
before the untreated FFT(c) sample. Meanwhile, with a
higher initial solid content, FFT(a) samples dry faster than
FFT(c) samples. In addition, Figure 4 shows a significant

Figure 4. Shear strength vs. solid content of NFC-FFT and

FFT control samples under atmospheric dry over 14 days.
(1) samples of FFT(a); (2) samples of FFT(c)

Figure 5. Drying curves of NFC-FFT and FFT control

samples under atmospheric dry. (1) samples of FFT(a); (2)
samples of FFT(c)
variation in the solid content was observed of the samples
for each FFT on day 7 and day 14 since the atmospheric
drying test started. It indicates that the addition of NFC
impacts the drying behavior or the evaporation rate of FFT.
As can be seen, the solid content of FFT(a) control
continuously surpasses that of the NFC-FFT samples,
while the solid content of FFT(c) was lower than all the
NFC-FFT(c) samples. Therefore, NFC potentially
decreases the evaporation rate of FFT with higher SFR
when a reversed effect is observed for FFT with lower SFR.

Figure 7. CST results of FFT control samples and NFC-

FFT samples with the dose of better performance in
enhancing the shear strength gain

Figure 5. Sample imaged on day 3 shows the bitumen at

the sample surface Table 2 tabulates the SRF and the solid content of the
samples before and after the pressure filtration tests. The
SRF results of the control samples conform to those
3.2 Dewatering Behavior findings obtained in the studies (Xu et al. 2008, Loerke et
al. 2017). Similar to the CST results, increased SRF of
The impact of NFC on the dewatering behavior of FFT NFC-FFT(a) dosed at 0.15 wt.% is observed with an
was investigated in this study utilizing the constant head increment less than one order of magnitude compared to
pressure filtration method and CST. CST has been FFT samples. Corresponding to SRF, the solid content in
commonly used as an effective method to gain a sight on the filtered cake of NFC-FFT(a) dosed at 0.15 wt.% is
the immediate dewatering of FFT. It determines the time reduced which indicates more water was retained in the
used for the fluid to be pulled through two electrodes by the filtered sample. Meanwhile, the difference in SRF value
suction force of the filter paper. For example, a smaller CST and the solid content in filtered cake between NFC-FFT(c)
value indicates a higher dewaterability of the sample. dosed at 0.05 wt.% and FFT(c) control is inappreciable.
Meanwhile, constant pressure filtration tests were
performed to determine the specific resistance to filtration
(SRF) of the samples. As a laboratory parameter, SRF Table 2. Pressure filtration results of FFT control samples
reflects the hydraulic resistance of fluids flowing through and NFC-FFT samples
the sample (Agerbæk and Keiding, 1993). Similar to the
relationship between CST and dewaterability, the higher Solid
Initial solid
the SRF value is, the more difficult it is for the fluids to flow Sample content in SRF
content (%)
cake (%)
through the filtered sample.
Figure 7 illustrates the CST results of the FFT control FFT(a) 42.8 68.61 7.20E+13
samples and NFC-FFT samples with the dose of better NFC-FFT(a) @0.15 % 42.8 60.35 1.01E+14
performance in enhancing the near-surface shear strength FFT(c) 30.8 75.18 4.80E+13
gain determined in the atmospheric drying tests. For each NFC-FFT(c) @0.05 % 30.8 75.34 4.29E+13
sample, 5 tests were carried out, an averaged CST was
calculated and plotted with standard deviation. As can be
seen, the standard deviation range between 2 % and 5 %
of the average value was obtained which indicates the 3.3 SWCC
reliability of CST test. The results show that the average
CST of NFC-FFT(a) dosed at 0.15 wt.% is 10 % higher Figure 8 shows the SWCC for the same samples tested
compared to FFT(a) control. However, the variation of CST in the CST and pressure filtration tests. These curves were
between FFT(c) control and NFC-FFT (c) dosed at 0.05 typically plotted in the form of gravimetric water content vs.
wt.% was negligible, which could be potentially due to the suction, or volumetric water content vs. suction, or degree
lower NFC dose. The CTS results indicate that the impact of saturation vs. suction. The gravimetric SWCC curve can
of NFC on the dewatering of FFT with higher SFR is more be obtained directly after the SWCC tests. However, in
apparent. order to derive the other two types of SWCC curves, the
relationship between the gravimetric water content and
void ratio needs to be first understood by conducting
shrinkage tests. Subsequently, the SWCC for instance in
forms of saturation degree versus soil suction can be
derived use equation 3. The air entry value (AEV) Meanwhile, it is noted that the AEV value of NFC-FFT(a)
determined in this study is the matric suction, which dosed at 0.15 wt.% is increased from 597 kPa to 857 kPa.
indicates the soil suction when the air starts to enter the However, the AEV of NFC-FFT(c) dosed at 0.05 wt.% is
largest pores in the soil, where the desaturation of soil increased slightly. Due to the hydrophilicity of NFC fibers,
commences. It was estimated by targeting the intersection more water was trapped in the samples which would
of the horizontal tangent line through the initial point and require more energy or higher suction to drive the water out
the tangent line through the inflection point on the SWCC of the samples. This finding is consistent with the pressure
curve (Fredlund & Xing 1994). Table 3 summarizes the filtration tests.
changes in saturated volumetric water content and AEV of
the FFT samples before and after adding NFC. Generally,
the AEV increase with the plasticity of the soil (Fredlund & Table 3. AEV and Volumetric water content at saturation of
Xing, 1994), the findings from these tests comply with the FFT control samples and NFC-FFT samples estimated
conclusion as the AEV is 597 kPa for FFT(a) with a plastic from the SWCC
index of 15 % and 1122 kPa for FFT(c) with the plastic
index of 23 %. Sample AEV (kPa)
FFT(a) 597
𝑆𝑒 = 𝑤𝐺𝑠 [3] NFC-FFT(a) @0.15 % 857
FFT(c) 1122
NFC-FFT(c) @0.05 % 1204
Where:
𝑆 = saturation degree.
𝑒 = void ratio. 4 CONCLUSION
𝑤 = gravimetric water content.
𝐺𝑠 = the specific gravity of the solid in the samples. Overall, the results suggested that NFC has the
potential to change the drying pattern, shear strength gain,
dewatering behaviour, and soil water characteristics of
FFT. The shear strength of NFC-FFT under atmospheric
drying is enhanced. As the initial solid content of the FFT
increases, a higher dose of NFC: FFT is more effective in
improving the shear strength gain, while a lower dose was
determined better for FFT with less solids. Meanwhile, the
desaturation of FFT with smaller SFR or higher fine content
is delayed as evidenced by increased CST, SRF and AEV.
However, the FFT samples with greater SFR or a lower fine
content dries slightly faster with the addition of NFC when
the dewatering behavior and soil water characteristics were
not significantly impacted.
It is believed that clay is critical in influencing the
properties of FFT, especially, its water-holding capacity
dominates the behavior of FFT (Mikula & Omotoso 2006).
The effect of the impact on the drying and dewatering
behavior varies between different types of FFT tested in
this study. Therefore, it would be worth conducting further
testing to evaluate the interaction between NFC and clays.

ACKNOLEDGEMENT

The authors acknowledge the following sources of

financial support: Natural Sciences and Engineering
Research Council of Canada (NSERC), MITACS, and
Alberta Innovates. The authors also thank Performance
BioFilaments Inc. (PBI) for providing the nanocellulose
product and technical support for this project. The authors
extend their gratitude to the staff at Institute for Oil Sands
Innovation (IOSI) for providing training and access to their
lab equipment for the pressure filtration and CST testing.
Figure 8. SWCC of FFT control samples and NFC-FFT Lastly, the authors would like to express their appreciation
samples with degree of saturation vs. suction. (1) samples to Umme Rima for her great contribution to the laboratory
of FFT(a); (2) samples of FFT(c) work and reporting.
REFERENCE Near-Surface Shear Strength. PanAm Unsaturated Soils
2017, ASCE, pp. 279-288.
ASTM D854-14. Standard Test Methods for Specific Kaminsky, H. 2014. Demystifying the methylene blue
Gravity of Soil Solids by Water Pycnometer. index. International Oil Sands Tailings Conference 2014,
ASTM D6836-16. Standard Test Methods for Banff, Alberta, Canada, pp.221-229.
Determination of the Soil Water Characteristic Curve for Li, Y., Kaminsky, H., Gong, X. Sun, Y., Ghuzi, M.,
Desorption Using Hanging Column, Pressure Extractor, Sadighian, A. 2021. What Affects Dewatering Performance
Chilled Mirror Hygrometer, or Centrifuge. of High Density Slurry?, Minerals, 11(7).
ASTM D4318-17. Standard Test Methods for Liquid Limit, Loerke, R., X. Tan & Q. Liu 2017. Dewatering of Oil Sands
Plastic Limit, and Plasticity Index of Soils. Mature Fine Tailings by Dual Polymer Flocculation and
ASTM D2216-19. Standard Test Methods for Laboratory Pressure Plate Filtration. Energy & Fuels, 31(7): 6986-
Determination of Water (Moisture) Content of Soil and 6995.
Rock by Mass. Mikula, R. and O. Omotoso 2006. Roles of clay in
Afacan, C., R. Narain & J. B. P. Soares 2019. Water soluble controlling tailings behavior in oil sands processing. Clay
polymeric nanofibres for rapid flocculation and enhanced Science, 12 (Suppl.2): 177-182.
dewatering of mature fine tailings. The Canadian Journal Mizani, S., P. Simms & W. Wilson 2017. Rheology for
of Chemical Engineering, 98(1): 96-103. deposition control of polymer-amended oil sands tailings.
Agerbæk, M. L. & K. Keiding 1993. On the Origin of Specific Rheologica Acta, 56(7-8): 623-634.
Resistance to Filtration. Water Science and Technology, Qureshi, T., Y. Li, A. Sedgwick, H. Kaminsky & J. Ng 2021.
28(1): 159-168. Assessing oil sands tailings consolidation using a modified
Ali, A., K. Shaker, Y. Nawab, M. Jabbar, T. Hussain, J. benchtop filter press. Paste 2021, Australian Centre for
Militky & V. Baheti 2018. Hydrophobic treatment of natural Geomechanics, Perth, Australia.
fibers and their composites—A review. Journal of Industrial Salam, A. M., B. Örmeci & P. H. Simms 2021.
Textiles, 47(8): 2153-2183. Determination of optimum polymer dosage for dewatering
Beier, N., W. Wilson, A. Dunmola & D. Sego 2013. Impact of oil sands tailings using torque rheology. Journal of
of flocculation-based dewatering on the shear strength of Petroleum Science and Engineering, 197: 107986.
oil sands fine tailings. Canadian Geotechnical Journal Sobkowica, J.C., Morgenstern, N.R. 2009. A geotechnical
50(9): 1001-1007. perspective on oil sands tailings. Thirteenth International
Besbes, I., M. R. Vilar & S. Boufi 2011. Nanofibrillated Conference on Tailings and Mine Waste, Banff, AB,
cellulose from Alfa, Eucalyptus and Pine fibres: Canada.
Preparation, characteristics and reinforcing potential. Stanislas, T. T., G. C. Komadja, Y. R. Nafu, A. A. Mahamat,
Carbohydrate Polymers, 86(3): 1198-1206. P. W. H. Mejouyo, J. F. Tendo, E. Njeugna, P. A. Onwualu
Da Costa Correia, V., S. F. Santos, R. Soares Teixeira & and H. Savastano 2022. Potential of raffia nanofibrillated
H. Savastano Junior 2018. Nanofibrillated cellulose and cellulose as a reinforcement in extruded earth-based
cellulosic pulp for reinforcement of the extruded cement materials. Case Studies in Construction Materials,16.
based materials. Construction and Building Materials, 160: Tollenaar, R. N., L. A. Van Paassen & C. Jommi. 2018.
376-384. Small-scale evaporation tests on clay: influence of drying
Dean, E. W. and D. D. Stark 1920. A Convenient Method rate on clayey soil layer. Canadian Geotechnical Journal,
for the Determination of Water in Petroleum and Other 55(3): 437-445.
Organic Emulsions. Journal of Industrial & Engineering Vajihinejad, V., S. P. Gumfekar, B. Bazoubandi, Z. Rostami
Chemistry, 12(5): 486-490. Najafabadi & J. B. P. Soares. 2018. Water Soluble Polymer
Durand, S., B. E. Jackson, W. C. Fonteno & J.-C. Michel Flocculants: Synthesis, Characterization, and Performance
2021. The Use of Wood Fiber for Reducing Risks of Assessment. Macromolecular Materials and Engineering,
Hydrophobicity in Peat-Based Substrates. Agronomy, 304(2).
11(5): 907. Xu, Y., T. Dabros & J. Kan. 2008. Filterability of oil sands
Ferrer, A., L. Pal & M. Hubbe 2017. Nanocellulose in tailings. Process Safety and Environmental Protection,
packaging: Advances in barrier layer technologies. 86(4): 268-276.
Industrial Crops and Products, 95: 574-582. Zhang YH., Beier N., Gourlay K. & Minhas G. 2022.
Fredlund, D. G. & A. Xing 1994. Equations for the soil- Application of Nanocellullose Filament (NFC) in Enhancing
water characteristic curve. Canadian Geotechnical the Geotechnical Properties of Oil Sands Tailings.
Journal, 31(4): 521-532. International Oil Sands Tailings Conference 2022,
Fredlund, D. G. 2019. State of practice for use of the soil- Edmonton, Alberta, Canada.
water characteristic curve (SWCC) in geotechnical
engineering. Canadian Geotechnical Journal, 56(8): 1059-
1069.
Jeeravipoolvarn, S., J. D. Scott & R. J. Chalaturnyk 2009.
Geotechnical characteristics of laboratory in-line thickened
oil sands tailings. Tailings and Mine Waste 2009, Banff,
AB, Canada, pp. 813–828.
Kabwe, L. K., G. W. Wilson, N. A. Beier & J. D. Scott 2017.
Effects of Atmospheric Drying and Consolidation of
Flocculated Thickened Tailings and Centrifuged Cake on
Monday, October 2, 2023

GEOHAZARDS II
Geological Model Influence on Slope
Stability: Torrioni di Rialba Case Study
Roham Akbarian
Arup, Toronto, Ontario, Canada
Laura Longoni & Davide Brambilla
Department of Civil and Environmental Engineering –
Politecnico di Milano, Milan, Lombardy, Italy

ABSTRACT
Landslides are among the most destructive types of natural disasters that can cause significant loss of life and property
damage within their zone of impact. This paper presents the study on stability of Torrioni di Rialba (Rialba Towers), a 135-
m rock cliff located in Abbadia Lariana, northern Italy, during numerically simulated seismic events of different magnitudes.
Torrioni di Rialba is located near a narrow corridor, where important lifelines including rail, road and pipes run alongside
each other and connects the northern part of Lombardy to the rest of the country. The study adopted Discrete Element
Method (DEM), using Itasca 3DEC, to model the mechanics of the rock cliffs with complex joint orientations, during
significant seismic events (with return period of 475 years), recorded in northern Italy. Sensitivity analyses were conducted
on key input parameters. The results suggested that the dominant possible slope failure mechanism was sliding of the
towers along the dip direction of the slope towards Lake Como. The results show high sensitivity to joint properties while
remain less sensitive to intact rock properties.

RÉSUMÉ

Les glissements de terrain font partie des catastrophes naturelles les plus destructrices et peuvent entraîner des pertes
humaines et des dégâts matériels considérables dans leur zone d'impact. Cet article présente l'étude de la stabilité de la
falaise Torrioni di Rialba (Tours de Rialba), une falaise rocheuse de 135 m située à Abbadia Lariana, dans le nord de
l'Italie, lors d'événements sismiques de différentes magnitudes simulés numériquement. Torrioni di Rialba est situé à
proximité d'un corridor étroit qui relie la partie nord de la Lombardie au reste du pays et qui regroupe d'importants lignes
de vie réseaux, dont notamment des voies ferrées, des routes et des canalisations. L'étude a adopté la méthode des
éléments discrets (DEM) grâce au logiciel Itasca 3DEC, pour modéliser la mécanique de ces falaises rocheuses et les
orientations complexes des joints, pendant des événements sismiques significatifs (avec une période de retour de 475
ans), enregistrés dans le nord de l'Italie. Des analyses de sensibilité ont été menées sur les principaux paramètres d'entrée.
Les résultats suggèrent que le principal mécanisme de rupture possible est le glissement des tours le long de l'inclinaison
de la pente vers le lac de Côme. Les résultats montrent une grande sensibilité aux propriétés des joints, et une moindre
sensibilité aux propriétés de la roche intacte.

1 INTRODUCTION

Landslides are among the most destructive types of natural

disasters that can cause significant loss of life and property
damage within their zone of impact. Comprehensive
knowledge about landslide structure, formation and
triggering factors are essential for landslide hazard
assessment and risk analysis. Landslide's path and
distance-from-origin are also key considerations for
landslide risk mitigation.
The Rialba Towers (Torrioni di Rialba; called the towers
herein) are 135 m rock cliffs located in Abbadia Lariana,
northern Italy, near to the SS36 national road which
connects Lecco to the Valtellina. The risk of landslide is
potentially high due to the vicinity of the towers to a narrow Figure 1. Case study area
corridor where lifelines, a national road and train tracks
pass alongside each other which connects the northern integration for complex modelling purposes. However,
part of the area with 200,000 inhabitants to the rest of the complex geological structures and complicated formations
country (Figure 1). sometimes lead to more sophisticated problems and
Our understanding of the mechanics of landslides has require in-depth investigations in all aspects.
been improved with the development of new, innovative Numerical modelling techniques have been widely
data collection methods and the advancements in data used to solve complex slope problems, which otherwise,
could not have been possible using conventional Esino formation is made up of limestone, dolomitic
techniques. These advanced methods are used to model limestone from light gray to dark gray, sometimes pinkish,
slopes of different properties, from soft soils to hard rocks, and dolomite in color from gray to brown, with massive or
with geologies of different complexities. Rock slopes are thick layers. It is a carbonate platform with numerous sub-
composed of rock masses, usually with discontinuities environments. The unit is generally fossiliferous, rich in
(fractures), which can govern the mechanics of the slope. stromatolites, conchoids and vegetative bodies of algae.
The aim of this study is to evaluate the influence of the Rialba Conglomerate (conglomerate or rubble with a
geological model, developed based on the geological, wide range of clasts, of which up to 25% are of the gravel
structural and geophysical data, on landslide modelling and size) massive sharp-edged, very well cemented (lithified),
stability analysis. The data was integrated into a 3D weakly karstified; exclusively local pebbles, even with
geological model, which enables a better understanding of yellow and reddish sandstone. The conglomerate may
the structural layout and kinematics of a potential collapse, sometimes consist blocks of considerable dimensions (a
to be used for detailed numerical modelling. few meters). It is observed both at the base and the roof.
In the top part of the deposit, the layers have a finer
2 GEOLOGICAL STRUCTURE particle size, but always retain a high degree of
The case study area has been subject to numerous field cementation. The Rialba’s Conglomerate could be
investigations, geological mapping and geophysical interpreted as an ancient slope deposit. The degree of
surveys (Arosio et al., 2013; Brambilla et al.,2017; Taruselli cementation is usually linked to the age of the deposit, in
et al., 2018 and Taruselli et al., 2019), to better understand the sense that the oldest moraines are often cemented;
the geology of the slope. sometimes they have real compact conglomerates. The
The towers are located above the eastern shoreline of outcrop area of the Rialba’s conglomerate is rather small;
the east branch of lake Como, in the vicinity of Lecco fault, deposits of this type have been found only in the Rialba
alongside the fractures oriented in the direction of the lake region at Abbadia Lariana .
(N-S). Other important faults oriented in an orthogonal San Giovanni Bianco formation is evidence of the
direction (E-W), giving origin to a compressive tendency retreat of sea water, considerably worse at the end of
with the direction N-S that generated folds and thrust faults. Carnico with the subsequent formation of vast coastal
The towers are in south of Insubrica Line (Periadriatic lagoons, the bottom of which have accumulated fine
Seam), in the west-central portion of the Southern Alps sediments clayey; interspersed them are salts lenses,
Lombardo and precisely in Lariane Alps. Insubrica Line is determined by the evaporation of sea water. In the Lecco
the expression on the Earth's surface of the clash between area, the different zones are poorly differentiated, and the
the European and African plates and marks the boundary unit is predominantly represented by sandstones,
between the Eurasian plate and the Adriatic. The main siltstones and reddish argillites, well outcropping on the
tectonic units are (from north to south) Northern Grigna, southern slope of Mount Ooze.
Southern Grigna, Coltignone and Autoctono of Lecco. The formation was so named because it has area-type
The lithological and structural complexities of the area and section-type near San Giovanni Bianco and with
surrounding the towers are related to different depositional respect to the outcrop area. It has a discreet extension
episodes and deformations (connected Alpine orogenesis) between San Giovanni Bianco and Camerata Cornello in
and the resulting structure is composed of a series of the Brembana Valley. Elsewhere, for its lithological
tectonic units, variously faulted and overthrust, which characteristics and its rheological behavior, it is often
causes direct contact units and different stratigraphy covered or laminated along structural surfaces.
(Brambilla et al., 2017). In order to formulate hypotheses about the stratigraphy
There are three main formations that characterize the of the rock mass at the base of the towers, it was necessary
area of the towers: Esino formation, Rialba’s Conglomerate to conduced geophysical surveys. The outcomes were
(RC) and San Giovanni Bianco (SGB) formation (Figure 2). mainly used to understand what the tectonic evolution of
the layer of San Giovanni Bianco below the towers might
be. One seismic tomography survey and two electric
tomography surveys were conducted between 2010-2012
(Arosio et al., 2013).
The contact between the Rialba’s Conglomerate and
San Giovanni Bianco is not a stratigraphic contact
(natural), but tectonic. The presence of a so-called “Breach
of Friction” (breccia di frizione) or Cataclasite is a clear
indicator of the tectonic limit. In fact, it was generated as a
result of the friction of the two rock masses spread between
them. Brambilla et al (2017) provided detailed review of
geological mapping, geophysical surveys as well as major
joint sets of the rock mass (Figure 3).
From Figure 3, the joint F1 is the most upstream among
the main ones, at an altitude of about 570 m. It has a
Figure 2. Location of Torioni di Rialba and the composing
trapezoidal geometry with the maximum length of 5-6 m
geological formations (After Brambilla et al, 2017)
and minimum of 3.2 m. Joint F1 is evident and follows
continuously to the base (for an estimated height of about
 with unfavourable contribution from the water pressure
present in the fractures.
In the present case, it is considered that the constituent
rigid material of the towers (cemented conglomerate) and
the pressures exerted on the underlying plastic material
(shales), causes movements of extension with the
consequent formation of numerous cracks in the body of
the towers. Apart from the main fracture family, numerous
vicarious fractures were observed as well (Brambilla et al,
2017).
The Discrete Element Method (DEM) is a numerical
method for modelling the dynamics of solid particles, which
interact with each other at discrete contact points. Since it
was firstly proposed by Cundall and Strack (1979) to study
the mechanical behavior of rock at the microscopic level,
the DEM has been developed rapidly, with applications in
rock/soil mechanics. In DEM simulations, the properties of
Figure 3. Major dicontinuities a stressed assembly of rigid spherical particles (e.g.
position, velocity and contact forces) are updated at every
70 m), where it ends in the axial part of the watershed. F1 numerical iteration time step. Particle shapes are arbitrary:
has almost rectilinear, mean dip of 75-80° at an altitude of any particle may interact with any other particle, and there
about 565 m and it is the most closed one. Therefore, it is are no limits placed on particle displacements or rotations.
not easily possible to locate along the ridge of the towers. Due to this generality, a robust and rapid method must be
The trend roughly straight, tilt of 75-80°, dipping towards defined to identify pairs of particles that are touching, and
SW and estimated height of about 65 m. F2 ends at the to represent their geometric and physical characteristics.
base, in the right mountain slope of the watershed. The translational and rotational displacements of each
The joint F3 is located at an altitude of about 550 m and particle are obtained by explicitly integrating the governing
on the summit. It is wide and about 2m. It is a broad and differential equations based on the Newton’s second law of
deep fissure on the southern slope of the towers, following motion, while the contact forces between particles are
continuously to the base (for a height of about 70 m). Trend calculated using well defined force–displacement contact
has roughly straight, tilt around 80-85° and dipping into models (Hertz, 1882; Mindlin and Deresiewicz, 1953;
south and southwest. F3 ends in the axial part of the large Cundall and Strack, 1979).
watershed at an altitude of about 440 m. The joint F4 is The Mohr-coulomb model is a well-known model for
located at an altitude of approximately 537 m and it is the plasticity in soil and rock mechanics. The failure envelope
most downstream of the main joints. It is about 15-20 m in 3DEC consists of a Mohr-Coulomb criterion with tension
(measured directly on site on the top of the towers) from cut-off. There is a non-associated flow rule for shear failure;
the tip. It is a very deep gap (about 100 m) which tends to no flow rule for tension failure is considered in this model.
isolate, almost completely, a block of about 50,000 m 3 The basic joint constitutive model incorporated in 3DEC
(width about 25 m, depth of approximately 20 m and a is the generalization of the Coulomb friction law. This law
height of about 100 m). At the bottom of this block (at an works in a similar fashion both for sub-contacts between
altitude of about 420-425 m) cavitations were observed. rigid blocks and for sub-contacts between deformable
There is a water circulation inside the rock mass for blocks. Both shear and tensile failure are considered, and
fracturing and/or karst, with the emergence at the tectonic joint dilation is included. In the elastic range, the behavior
features (fault - overthrust). is governed by the joint normal and shear stiffness, Kn and
Ks, respectively.
3 Methodology For dynamic modelling, the physical stiffness of joints
can have a substantial influence on seismic wave
The risk scenarios associated with this site are related to propagation. Pyrak‐Nolte et al. (1990) presented field and
the instability or collapse of towers, due to either sliding, laboratory test results which demonstrate the effect of the
toppling or general failure of the slope. stiffness of dry natural fractures in rock on attenuation and
One key unknown input to the model appears to be the changes in travel time of the seismic waves.
extent of the San Giovanni Bianco formation layer Numerical distortion of the propagating wave can occur
(argillite/shale) beneath the towers. Different models can in a dynamic analysis, whether it is based on a continuum
be developed based on the change in the geometry of this or discontinuum program, as a function of the modeling
layer below the surface, which can differ remarkably in conditions. Both the frequency content of the input wave
behaviour. A slight change in the inclination of the and the wave speed characteristics of the system will affect
formation, while being compatible with the measurements, the numerical accuracy of wave transmission. Kuhlemeyer
can lead to a physical model that varies completely the and Kuhlemeyer & Lysmer (1973) show that for accurate
stability of the towers. representation of wave transmission through a model, the
The toppling mechanism occur with rotation around a spatial element size must be smaller than approximately
point of a rock block under the action of the force of gravity, one-tenth to one-eighth of the wavelength associated with
of the pressure exerted by the blocks of adjacent materials the highest frequency component of the input wave. This is
important to assure the block size after meshing is suitable
for dynamic modelling. Table 2. Joint properties

3.1 Geometry Joint

Properties
Kn [Pa/m] Ks [Pa/m] Jϕ [°]
LiDAR laser scanning surveys provided the digital surface Vertical joints 4.0E+09 2.0E+09 10
model (DSM). Since, the DSM cannot be directly imported
Sub-horizontal joint
to 3DEC, a plug-in called PGEN was used to generate the (between SGB and
model from the contour lines (Figure 4). 6.0E+08 3.0E+08 28
Rialba’s
conglomerate)

3.3 Dynamic Inputs

Dynamic analysis was conducted to evaluate the stability

of the towers during seismic events. EUCENTRE seismic
database allows obtaining the seismic inputs for nonlinear
dynamic analyses, for a return period of 475 years.
The seismic inputs were defined in term of a group of
seven real accelerograms, recorded on outcropping
bedrock, and compatible with Italian Standard NTC08
(2008). The magnitude range is from 5.3 Mw to 6.9 Mw
(Moment Magnitude). The seismic scenarios adopted for
stability analysis are listed in Table 3.

Table 3. Seismic scenarios

Epic.
Figure 4. Building model geometry from DSM Magnitude Scaling
No. Distance Source
[Mw] Factor
[Km]
3.2 Model Inputs
EQ1 5.3 37 0.98 ESD
Once the model is generated, material behaviour models EQ2 6.2 32 0.62 ESD
must be defined for all the blocks and discontinuities. The EQ3 6.9 94 0.61 NGA
towers are made of Rialba’s Conglomerate overlaying a EQ4 6.6 36 0.38 KNET1
layer of San Giovanni Bianco’s argillite. Figure 5 shows the EQ5 5.8 21 0.86 ITACA
EQ6 5.6 22 0.76 ITACA
material blocks and the main discontinuities separating the EQ7 5.4 11 0.62 ITACA
towers.
Among these seven earthquake scenarios, two of them are
critical for our studies: EQ1 has the highest scaling factor
while EQ3 has the highest magnitude.
SeismoSpect software was used to generate the
strong-motion parameters from the accelerograms. The
software is also capable of applying filters for baseline-
corrections, using up to third order polynomials. The tool is
Figure 5. Left: Rialba towers separated by main also capable of computing the mean spectral response of
discontinuities, Right: blocks materials (Conglomerate: a collection of accelerograms and to compare these results
blue, SGB: green) with a reference spectrum.
The accelerograms cannot be directly imported to
The material and joint properties used for this study are 3DEC as the boundary conditions in 3DEC are velocity-
listed in Table 1 and Table 2, respectively. based. Therefore, the velocity-time diagrams are derived
from the accelerograms and imported to 3DEC in the form
Table 1. Material properties of tables.
In order to understand the effect of the boundary
Properties conditions and wave attenuation, the seismic velocities
Material ρ E ν K G ϕ were imposed, once only at the bottom of the SGB layer
[g/
[Pa] - [Pa] [Pa] [°] and once to the whole SGB layer. Figure 6 shows the
cm3 ] nodes where the seismic velocities were imposed for each
RC 2.6 2.5E+10 0.3 2.1E+10 9.6E+9 42 scenario.

SGB 2.2 8.0E+08 0.4 1.3E+09 2.8E+8 30

Esino 2.4 2.5E+10 0.3 2.1E+10 9.6E+9 30

 reduced by half to meet the criterion. This had a significant
impact to the runtime due to the size of the model.
For all the earthquakes, the seismic waves imported
both in X and Y direction separately, in order to determine
the critical wave direction for instability. Figure 8 shows an
example of velocities recorded on control points on the
front tower during first earthquake scenario (EQ1).

Figure 6. Seismic boundaries imposed on: left - bottom of

SGB layer (higher attenuation) and right: entire SGN layer
(lower attenuation)

4 RESULTS

A set of control points were set beneath the towers (A to

E), to monitor the relative base displacement of the towers.
A set of vertical control points (BB, CC) were defined on
the front tower to evaluate the relative rotation of the first
tower during different imposed conditions. Figure 7 shows
the control points.

Figure 8. Recorded velocities on the front tower – EQ1

For the base model, the seismic waves were imposed

to the entire SGB layer beneath the towers. The SGB layer
has weaker strength properties in comparison to Rialba’s
conglomerate and this plays an important role in wave
attenuation. A sensitivity check was conducted separately
on impact of imposing the seismic waves to only the bottom
of the SGB layer, as well as on the thickness of the SGB
layer (see Section 4.1.1).
The calculated movements at control point A, located
Figure 7. Control points for movement monitoring at the toe of the front tower, as well as the stability checks
are provided in Table 5.
Kuhlemeyer and Lysmer (1973) check was conducted on
each of the earthquake scenarios. The results are provided
Table 5. Control point A results – low attenuation
in Table 4.
EQ in Y direction EQ in X direction
Table 4. Kuhlemeyer and Lysmer (1973) check for seismic
(towards the slope’s (towards the slope’s Stability
events
dip direction) strike) Check
EQ1 EQ2 EQ3 EQ4 EQ5 EQ6 EQ7 Y-disp. [m] Y-disp. [m]

Eq. 1 0.072 0.062 Stable

𝑉𝑠 [m/s] 357 357 357 357 357 357 357
Eq. 2 0.023 0.018 Stable
𝑓 [𝑠 −1 ] 0.5 2 1 1 2.2 1.9 4.1
Eq. 3 0.055 0.106 Stable
𝜆 [m] 179 714 357 357 785 678 1463 Eq. 4 0.029 0.017 Stable
1 Eq. 5 0.009 0.010 Stable
𝜆 [m] 22 89 45 45 98 85 183
8
Eq. 6 0.036 0.030 Stable
Selected
Eq. 7 0.014 0.019 Stable
Mesh 20 40 40 40 40 40 40
Size [m]
The results shows that although a maximum displacement
of 5.5 cm was recorded on control point A during EQ3, total
For all the earthquakes, we used the same mesh size
stability remained, and the system reached equilibrium
except for the first earthquake where the mesh size was
again. Moreover, no clear correlation was observed
between the earthquake wave direction and maximum
settlement.

4.1 Sensitivity Analysis

c)
A series of sensitivity checks were conducted to evaluate
the sensitivity of the results to different geometry, seismic
boundary, material and joint properties. The sensitivity
checks were only conducted to EQ1 and EQ3, as the main
earthquake scenarios.

4.1.1 Sensitivity to Seismic Boundary Conditions

d)
The base analysis was conducted by imposing seismic
velocities to the entire SGB layer. As the thickness of the
SGB was one of the key uncertainties of this study, the
impact of the thickness of the SGB layer was then
evaluated by defining three scenarios for SGB layer Figure 9. Y-displacement of control point A a) EQ1 in X
thickness: 80 m (default thickness), 40 m and 10 m. These direction, b) EQ1 in Y direction, c) EQ3 in X direction and
measurements are taken from the toe of the front tower to d) EQ3 in Y direction
the bottom of the SGB layer.
Figure 9 shows the Y-displacements calculated at By imposing earthquakes in two perpendicular directions,
control point A, for different SGB thicknesses during it was observed that the wave attenuation is independent
earthquakes EQ1 and EQ3 (imposed in both x and y from the magnitude and direction of the earthquake and in
directions). From the results, a negative correlation was all cases, larger displacements were recorded for thinner
observed between the thickness of the SGB and the SGB thicknesses.
calculated displacements. This implied that the presence
of the SGB layer, a less stiff material comparing to Rialba’s 4.1.2 Sensitivity to Joint Properties
conglomerate, increases the wave attenuation.
a considerable difference in displacement was Sensitivity analyses on joint properties were conducted
observed between the case of “no attenuation’ and the to evaluate the impact of uncertainties in the experimental
cases where there were wave attenuation happening values on the modelling outcomes. Earthquake 3 (EQ3)
through different SGB thicknesses, which shows the was selected as the seismic scenario, with wave direction
impact of the SGB thickness on wave attenuation. A being imposed in Y direction (towards the slope’s dip
behaviour that could be explained using the basic rules of direction) Y, and seismic velocities being imposed on the
wave transmission from a “less dense” medium to “dense” entire SGB layer (no attenuation).
medium. Figure 10 shows the results of the sensitivity analysis
on joint properties. Joint Material 1 was used to model the
vertical joints while Joint Material 2 was used to model the
sub-horizontal discontinuity beneath the towers. The
results showed sensitivity to both reduction in joint friction
and joint normal stiffness while remain insensitive to
a) increase. More importantly, it was concluded that the
stability of the towers is highly sensitive to the joint friction
angle of the sub-horizontal joint, where a 20% decrease in
joint friction will lead to the failure of the towers in sliding
(Figure 11).

b)
 5 CONCLUSION

The main purpose of this work was to evaluate the stability

of Rialba Towers during different potential seismic events,
considering the influence of the geological model. A
physical model has been built based on the data gathered
from previous geological mappings, and geophysical
surveys. These investigations illustrated that the Rialba
towers are formed from Rialba’s conglomerate and located
Figure 10. Sensitivity analysis on joint properties – top: on a layer of San Giovanni Bianco formation
Joint Material 1 (vertical discontinuities), bottom: Joint (argillite/shale). However, the extent and thickness of this
Material 2 (sub-horizontal discontinuity) SGB layer beneath the towers remained as the main
source of uncertainty for modelling.
Different seismic scenarios were established using the
EUCENTRE seismic database. All these earthquakes had
the same return period of 475 years. For numerical
modeling, the seismic waves were imported to the model
in form of velocities. Therefore, the velocity-time spectrum
of each earthquake extracted from the accelerogram, and
linear baseline corrections were performed. In order to
consider different uncertainties regarding the direction of
wave propagation, earthquakes were imposed in both X
(towards slope’s strike) and Y (towards slope’s dip
direction) directions.
The results of the base model showed that although a
maximum displacement of 5.5 cm was recorded on control
point A during EQ3, total stability remained and the system
reached equilibrium again.
As the thickness of the SGB was one of the key
uncertainties of this study, the impact of the thickness of
the SGB layer was then evaluated by defining three
scenarios for SGB layer thickness: 80 m (default
thickness), 40 m and 10 m. A considerable difference in
displacement was observed between the case of “no
Figure 11. Failure in sliding due to decrease in joint friction attenuation’ and the cases where there were wave
angle attenuation happening through different SGB thicknesses,
which shows the impact of the SGB thickness on wave
4.1.3 Sensitivity to Material Properties
attenuation. A behaviour that could be explained using the
basic rules of wave transmission from a “less dense”
The results showed sensitivity with respect to the stiffness
medium to “dense” medium. It was also concluded that the
of the SGB layer. The increase in the SGB stiffness will wave attenuation is independent from the magnitude and
decrease the wave attenuation leading to higher calculated
direction of the earthquakes.
displacements. Figure 12 shows that a 60% increase in
Sensitivity analyses on joint properties were conducted
SGB stiffness will increase the displacement of control
to evaluate the impact of uncertainties in the experimental
point A by 80%. values on the modelling outcomes. It was concluded that
the stability of the towers is highly sensitive to the joint
friction angle of the sub-horizontal joint where a 20%
decrease in joint friction will lead to towers instability
(sliding). It was also concluded that the outcomes are
sensitive to SGB stiffness, where a 60% increase in SGB
stiffness will increase the displacement of control point A
by 80%.

REFERENCES
Figure 12. Sensitivity analysis on block properties
Arosio D., Longoni L., Papini M., & Zanzi L. (2013). Seismic
In contrast, no remarkable sensitivity was observed on the characterization of an abandoned mine site. Acta
Material 1 (representing the Rialba’s conglomerate). Less Geophysica, 61, 611-623.
than 2% change in the displacement was observed over a
20% change in block stiffness.
Brambilla, D., Ivanov, V. I., Longoni, L., Arosio, D., &
Papini, M. (2017). Geological assessment and physical
model of complex landslides: integration of different
techniques. In Advancing Culture of Living with
Landslides: Volume 2 Advances in Landslide
Science (pp. 431-437). Springer International
Publishing.
Cundall, P.A. and Strack, O.D.L. (1979) A Discrete
Numerical Model for Granular Assemblies.
Geotechnique, 29, 47- 65.
Hertz, H. (1882). Ueber die Berührung fester
elastischer Körper.
Kuhlemeyer, R. L., & Lysmer, J. (1973). Finite
element method accuracy for wave propagation
problems. Journal of the Soil Mechanics and
Foundations Division, 99(5), 421-427.
Mindlin, R. D., & Deresiewicz, H. E. R. B. E. R. T.
(1953). Elastic spheres in contact under varying
oblique forces.
NTC (2008) Norme tecniche per le costruzioni. D.M.
Ministero Infrastrutture e Trasporti 14 gennaio 2008,
G.U.R.I. 4 Febbraio 2008, Roma (in Italian)
Pyrak‐Nolte, L. J., Myer, L. R., & Cook, N. G. (1990).
Transmission of seismic waves across single
natural fractures. Journal of Geophysical
Research: Solid Earth, 95(B6), 8617-8638.
Taruselli, M., Arosio, D., Longoni, L., Papini, M., & Zanzi,
L. (2019, September). Raspberry shake sensor field
tests for unstable rock monitoring. In 1st Conference on
Geophysics for Infrastructure Planning Monitoring and
BIM (Vol. 2019, No. 1, pp. 1-5). EAGE Publications BV.
Taruselli, M., Arosio, D., Longoni, L., Papini, M., Corsini,
A., & Zanzi, L. (2018, September). Rock stability as
detected by seismic noise recordings-Three case
studies. In 24th European Meeting of Environmental
and Engineering Geophysics (Vol. 2018, No. 1, pp. 1-
5). EAGE Publications BV.
Forensic Engineering Assessment of the
Failure of a Temporary Sheet Pile Wall
Mrinmoy Kanungo, M.E.Sc., P.Eng.
GHD, Mississauga, ON, Canada

ABSTRACT
A case history documenting the failure of a temporary sheet pile shoring wall at a wastewater treatment facility expansion
project is presented. The wall was designed to retain a height (H) of 6 m of soils overlain by 3.8 m high sloping ground
behind the shoring. The wall was designed as a cantilever system without secondary supports such as tiebacks and struts,
and with nominal embedment depth (D) yielding an H/D ratio of only 0.5. The subsurface conditions consisted of topsoil
and fill overlying loose to compact gravelly sand and silty sand, with groundwater levels at a depth of 1 to 1.5 m below
ground surface. The cause of failure can be attributed to inadequate embedment of the shoring wall leading to overturning
failure, lack of secondary lateral supports as well as inadequate flexural capacity of the sheet pile sections. The failure
reinforces the importance of adhering to fundamental principles of geotechnical engineering in the design of excavation
support systems and the necessity of geostructural expertise in shoring design for interpreting the results of design
analysis. The case history serves as another example of the general lack of attention to the design of temporary works in
the industry despite the risk of severe consequences associated with failures.

RÉSUMÉ
Un cas d'expérience documentant la défaillance d'un mur d'étaiement temporaire de palplanches dans le cadre d'un projet
d'agrandissement d'une installation de traitement des eaux usées est présenté. Le mur a été conçu pour retenir une
hauteur (H) de 6 m de sols recouverte d'un terrain en pente de 3.8 m de haut derrière l'étaiement. Le mur a été conçu
comme un système en porte-à-faux sans supports secondaires tels que des tirants et des entretoises, et avec une
profondeur d'ancrage nominale (D) donnant un rapport H/D de seulement 0.5. Les conditions du sous-sol consistaient en
une couche arable et un remblai recouvrant du sable graveleux meuble à compact et du sable limoneux, avec des niveaux
d'eau souterraine à une profondeur de 1 à 1.5 m sous la surface du sol. La cause de la rupture peut être attribuée à un
encastrement inadéquat du mur d'étayage entraînant une rupture par renversement, au manque de supports latéraux
secondaires ainsi qu'à une capacité de flexion inadéquate des sections de palplanches. L'échec renforce l'importance
d'adhérer aux principes fondamentaux de l'ingénierie géotechnique dans la conception des systèmes de support
d'excavation et la nécessité d'une expertise géostructurale dans la conception de l'étaiement pour interpréter les résultats
de l'analyse de conception. L'histoire de cas est un autre exemple du manque général d'attention à la conception des
travaux temporaires dans l'industrie malgré le risque de conséquences graves associées aux défaillances.

1 INTRODUCTION approximately 26 m by 26 m footprint located near the

southern boundary of the plant close to a creek (Figure 1).
A case history documenting the failure of a temporary Historically, the original creek alignment passed through
sheet pile shoring wall at a wastewater treatment facility the facility and near the northeast corner of the tertiary
expansion project is presented. The wall was designed to excavation. The original creek alignment was subsequently
retain a height (H) of 6 m of soils overlain by 3.8 m high backfilled and the creek was relocated to its current
sloping ground behind the shoring. The collapse occurred alignment through the south of the facility.
on a weekend and fortunately no contractor staff were
present on site. Following the failure, the general contractor
undertook emergency response measures to secure the N
site including fencing off the collapsed area and relocating
stockpiled materials away from the failure area.
This paper presents an engineering assessment of the
cause of the failure and highlights the importance of
adhering to the fundamental principles of geotechnical and
geostructural engineering in the design of excavation
support systems.

2 FAILURE FEATURES

The project included the construction of a new tertiary

treatment building and associated deep excavation of Figure 1. Plan view of shoring system and failure location
 During an investigative site visit five days after the
failure event, it was observed that the north wall of shoring
along with a portion of east wall had collapsed into the
excavation (Figure 2 and Figure 3). The retained soil
behind the shoring had slumped into the excavation,
exposing a failure ‘scarp’ at the ground surface and a
sloping failure surface along the north boundary of the
excavation.

Figure 4. View of failure (looking northwest)

Figure 2. View of failure (looking north)

Figure 5. Distorted soldier pile flanges

Site personnel reported that the sheet pile depth at the

failed area was up to 12.2 m. The final depth of excavation
Figure 3. View of failure (looking east) had reportedly been reached ‘weeks’ prior to the failure
event. Two days before the failure, an equipment access
The shoring system consisted of sheet piles as the ramp located at the northeast corner of the excavation had
primary vertical support element. Wide flange vertical steel been removed. The dewatering wells were damaged as a
piles (‘soldier piles’) were installed at an approximate result of the failure and the dewatering system pumps were
spacing of 5 m in front of the sheet piles. Continuous wide shut off following the failure.
flange beam sections (‘wales’) were attached along the
inside perimeter of the sheet piles at the top of the soldier 3 SITE GEOLOGY AND GEOTECHNICAL DATA
piles. An array of PVC riser pipes was located along the
inside perimeter of the shoring as part of the dewatering The ground surface topography at the site generally sloped
system for the excavation. down from northeast to the southwest. The regional
Unlike the north and the east walls of the shoring quaternary geology was characterized by glaciofluvial ice-
system, the shoring along the south wall was supported by contact deposits comprised of gravel, sand and till, in
an earth berm in front of the sheet piles (Figure 3). Along addition to esker, kame, end moraine, ice-marginal delta
the western boundary, the excavation was sloped and subaqueous fan deposits. The bedrock depth was at
(Figure 4) leading down to a short sheet pile shoring wall least 25 m below the ground surface. The majority of the
at the bottom of the slope. Near the northwest corner of the structures proposed as part of the expansion project,
failed north wall, signs of local buckling/distortion were including the tertiary treatment buildings, were planned in
evident on the outside (facing the excavation) flange of the a low-lying area near the southwest corner of the plant.
soldier piles (Figure 5). The geotechnical investigation reported that the
While the soldier piles were observed to be generally boreholes encountered topsoil at the ground surface,
abutting the sheet piles, the nature of connection (e.g. underlain by fill materials extending to a depth between 3.8
welding), if any, between the two structural elements could m and 4.5 m. The fill consisted of non-plastic soils such as
not be determined. Soil filled gaps existed between the sand and gravel, silty sand, silt with occasional presence
sheet pile and the adjoining flange of the soldier piles in of organic silt layers and wood pieces. The fill was
some areas. Buried tree roots and organic soils were underlain by loose to compact gravelly sand and sand
visible on the exposed failure surface behind the collapsed layers extending to depths of 7.6 to 10.5 m. The gravelly
north wall. sand to sand was underlain by loose to compact silty sand
soils to the termination depth of the boreholes. Dynamic 0.61 m from the top of the sheet piles, above the top of the
cone penetration testing was conducted below the retained soil.
termination depth of the deepest borehole, from 18.8 m
depth to 27 m depth. Generally loose soils were
encountered to approximately 20 m depth, transitioning to
denser soils below this depth.
The site was characterized by a shallow groundwater
level, within 1 m depth below the ground surface. The
hydraulic conductivity of the sandy ‘aquifer’ at the site was
approximately 4.32x10-2 to 4.48x10-2 cm/s.

3.1 Geotechnical Design Recommendations

The geotechnical investigation report for the project noted

that the excavations for most of the structures would
extend well below the measured groundwater level and
would need to be vertical or near vertical due to space
restrictions imposed by adjacent underground services,
roadways and structures. Consequently, the excavation
should be undertaken within the confines of an engineered
support system designed by a Professional Engineer to
mitigate ground water seepage into the excavation, The Figure 7. Shoring details and typical soil conditions
geotechnical report cited some typical excavation support
systems such as sheet piling, secant pile walls or slurry The bottom of the excavation (‘dredge line’) in the
walls as feasible shoring options. failure area was at a depth of 9.81 m below original ground
The geotechnical report provided design earth pressure surface. This was the deepest part of the excavation within
distributions for multiple and singly braced shoring walls the overall sheet pile enclosure. The ground surface profile
(Figure 6). The geotechnical report also provided as it approaches the shoring system was shown to slope
geotechnical design parameters related to soil unit weight, down at an inclination of 1 Vertical to 3 Horizontal (1V: 3H).
strength and lateral earth pressure coefficients for the The top of retained soil is shown at a depth of 3.81 m. The
purposes of shoring design. shoring drawing mentioned that the soldier pile
“embedment depth” would be 13.72 m. While the bottom of
sheet pile depth was shown as 12.19 m, the tip/bottom
elevation of the soldier piles was not illustrated on the
profile view. Based on an assumed soldier pile tip depth of
13.72 m, the soldier piles extended an additional 1.53 m
below the tip of the sheet piles. It was also not mentioned
whether the soldier piles were to be welded to the sheet
piles or only abut the sheet piles. The sequence of
installation of the structural elements was also not
mentioned on the design drawings.

Ph = 0.65KH (Multi-level Support) Ph = KH (Single 5 GROUNDWATER CONTROL SYSTEM

level Support
K = Lateral earth pressure coefficient The dewatering system consisted of an array of deep wells
 = Unit weight of soil installed along the inside perimeter of the sheet piles. The
Figure 6. Lateral earth pressure diagrams suggested in wells were 125 mm diameter PVC tubes fitted with
geotechnical report minimum 6 m length well screens at the bottom. Each well
was provided with 3 horsepower submersible pumps. The
4 REVIEW OF SHORING DESIGN tip of the well screens was nominally at about 7 m below
the tip of the sheet piles. The wells were connected to 200
The shoring design engineer was retained by the general mm diameter HDPE ‘header pipes’ at the ground surface.
contractor of the project. The shoring consisted of Z55 The collected groundwater was directed to a sediment tank
sheet piles extending to a depth of 12.19 m below the located at the ground surface and eventually to the creek
existing ground surface (Figure 7). The shoring drawing flowing along the southern boundary of the site.
indicated that vertical W12x53 ‘soldier’ piles were to be The dewatering contractor estimated the maximum
placed in front of the sheet piles. The soldier piles would be combined dewatering rate to be 14,412,000 litres/day. The
spaced at a maximum horizontal spacing of 4.57 m. dewatering contractor further stated that the interlocking
Perimeter beams (commonly known as ‘wales’ or steel sheet piling were to be designed and installed to
‘walers’) consisting of W12x53 sections were shown to be withstand residual exterior hydrostatic pressure as
attached to the inside and outside faces of the sheet piles. required in the project specifications.
These beams were located at a depth of approximately
6 REMEDIATED SHORING 7.1 Deficiencies in Shoring Design

During a 2nd site visit approximately 2.5 months following The original shoring design was based on removal of 3 m
the failure, it was observed that the collapse debris of the existing fill from the ground surface adjacent to
including the collapsed sheet piles along the north and east shoring and have the shoring retain about 6 m of soil
walls had been removed and a new line of sheet piles height. Groundwater pressures were ignored in the shoring
(thicker wall consisting of Z65 and Z75 sections) had been design relying on the dewatering system. The soldier piles
installed approximately 5 to 6 m north of the original Z55 were not accounted for in the analysis.
section north shoring wall, thereby enlarging the
excavation footprint towards the north (Figure 8 and The following soil parameters were assumed in the
Figure 9). An earth berm was present in front of the new analysis:
wall. Additional reinforcing measures had been
implemented such as corner braces and ‘Dywidag’ Table 1. Soil Parameters Adopted in Shoring Design
deadman anchors that connect to the top wale.
Soil Layer Depth (m) Angle of Cohesion Deformation
Internal (kPa) Modulus
New North Wall Friction (°) (kPa)
Fill 0-3.8 26 0 15
Gravelly 3.8-5.7 39.4 0 85
Sand, Top
Gravelly 5.7-9.4 34.6 0 85
Sand,
Bottom
Silty Sand 9.4-12.19 26 62.4 10

The selected input parameters related to the strength and

stiffness of the soils were too high and not supported by the
geotechnical investigation report. For comparison, the
geotechnical report recommended an angle of internal
Figure 8. Remediated shoring system, looking northeast friction of 30° for the loose to compact gravelly sand
deposit. Similarly, a deformation modulus of 85 MPa is also
too high for a loose to compact gravelly sand deposit.
Available correlations with SPT N values suggest a
deformation modulus value in the range of 10 MPa to 40
MPa (Kulhawy and Mayne 1990). Furthermore,
considering the non-plastic/cohesionless nature of the silty
sand deposit, assigning a cohesion value of 62.4 kPa for
this soil was unjustified. The level of precision of some of
the input parameters (e.g., soil friction angles of 39.4° and
34.6°) is also not justified by the available geotechnical
testing data.
The shoring design assumed a horizontal support
located 0.61 m below the top of the sheet piles and at a
horizontal spacing of 3.05 m. This support was assigned a
‘spring stiffness’ value of 136.4 kN/m. Considering that the
perimeter wales are installed at a depth of 0.6 m below the
Figure 9. Remediated shoring system, looking west top of sheet piles, it appears that the shoring design
assumed some form of lateral reaction/restraint offered by
7 ANALYSIS OF FAILURE CAUSES
the wales at this level. It is unclear how the stiffness and
spacing parameters for this support system were derived.
Following the failure, various root cause theories were
The original shoring design did not include any
presented by the different parties such as the rain event
construction surcharge loading or residual hydrostatic
during the weekend of the failure. Unforeseen conditions
pressures in the calculations, contrary to the geotechnical
encountered during construction such as greater depth of
recommendations and project specifications. While it can
fill (weaker ground) associated with backfilling of the former
be argued that the groundwater control system managed
creek alignment, and saturated soils behind the shoring
to lower the groundwater level outside the shoring wall
were also alleged as the cause of failure.
(depending on when the dewatering system became
The following sections present an analysis of the root
operational and the actual tip elevation of the dewatering
cause and potential contributory factors associated with the
wells), data from any groundwater monitoring well from
failure, including commentary on the various alleged
outside the excavation was not available to confirm this
causes proposed by others.
assumption.
 The displacement diagram produced by the finite 7.2 Required Shoring Embedment
element based geostructural analysis software used in
shoring design indicated that the wall would rotate and Unlike conventional shoring design practice for similar
displace outward towards the excavation with a maximum excavation heights, the sheet pile system for the tertiary
wall deflection of 1200 mm occurring at the top of the pile excavation was not provided with standard secondary
(Figure 10). The results also showed rotation occurring at support systems such as tiebacks, struts or rakers. The
the tip of pile located at a depth of 12.19 mbgs. The perimeter wale near the top of the sheet piles did not
maximum bending moment of about 157 kN-m/m was provide adequate support/reaction that can be considered
shown to occur below the dredge line. The location of comparable to the standard support systems mentioned
maximum lateral deflection and the maximum bending above. Given the lack of a conventional horizontal support,
moment along the height of the shoring are generally the shoring wall can be expected to essentially act as a
consistent with the behavior of a cantilever shoring system. cantilever system, as indicated by the original design
The shape of the deflection and bending moment diagrams outputs.
as well as the excessive deflection of the shoring show the Cantilever shoring systems are heavily reliant on an
inadequacy of the pile embedment as well as the adequate penetration depth below the base of the
ineffectiveness of the perceived secondary support system excavation and wall structural stiffness for maintaining
located 0.61 m below the top of the shoring. stability. In conventional analysis, cantilevered walls are
assumed to rotate about some point between the dredge
line and tip/bottom of the wall (Figure 11). The direction of
lateral earth pressures (‘active’ and ‘passive’ pressures)
and the dominant mode of loading acting on the wall
therefore changes along the embedded portion of the wall
below the dredge line. The idealized net pressure
distribution on the shoring wall looks like the shaded area
in Figure 11.

Figure 10. Shoring wall displacement and bending moment

predicted by design analysis

For comparison, the maximum displacement expected

from a properly designed and constructed shoring system
in granular soils is in the range of 0.2% to 0.3% of the Figure 11. Cantilever retaining wall behaviour and
excavation depth according to the Canadian Foundation idealized earth pressure distribution, adopted from CFEM
Engineering Manual (CFEM 2006). This translates to a 2006
maximum expected displacement value of between 20 mm
and 30 mm for the approximately 10 m deep tertiary In general, the minimum required penetration depth (D)
excavation. With an estimated 1200 mm deflection at the of cantilever sheet piles below the dredge line is equal to
top of the shoring, the top wale would likely fail structurally the height of the retained soil (H), dependent on soil and
and/or disconnect from the sheet piles, at which point the groundwater conditions. For this project, a penetration of at
shoring wall would go into fully cantilever mode, which may least 6 m was be required for a retained soil height of 6 m,
ultimately lead to overturning failure of the cantilever even without considering the presence of an additional 3.8
shoring system. m of soil height (surcharge loading) above this retained
Although the design software apparently indicated a height. Depending on the soil strength and to add a factor
stable system, a proper evaluation of the deflection shape of safety, penetration depths of up to 2H may be required
and magnitude would have shown that the system for cantilever shorings, as mentioned in the CFEM (2006).
response was unacceptable. It is also possible that had the Naturally, weaker soils will require more embedment/lateral
shoring design used realistic soil parameters as discussed support from a shoring wall compared to a stronger soil,
above, the software would have indicated instability of the and therefore greater embedment is required for structural
system. stability. Typically, cantilever systems are not used when
the height H is greater than 5 m, generally due to excessive
lateral displacements.
 As part of the forensic investigation, the required such as additional depth of loose fill materials (as alleged
penetration depth of the cantilever shoring wall was by others as cause of failure), whether saturated or
estimated based on first principles, i.e., the wall has to unsaturated, would not have any material impact on the
satisfy horizontal force and moment equilibriums, following failure.
the guidelines presented in the CFEM (2006). For the sake
of simplicity, homogeneous soil conditions with uniform soil 7.3 Shoring Flexural Resistance
unit weight and soil strength throughout the soil profile were
assumed. As a parametric study, the soil strength, Although the primary mode of failure appears to be
represented by the angle of internal friction, was varied overturning failure of the sheet piles, the structural
from 30° (loose soils) to 40° (relatively dense soil), keeping adequacy of the sheet piles was also analyzed. The
the soil unit weight constant at 20 kN/m3. Groundwater original shoring design estimated a peak bending moment
pressures were ignored assuming that the dewatering demand of about 157 kN-m/m (Figure 10). In comparison,
system lowered the groundwater level below the bottom of the Z55 sheet pile sections used in the original design and
the shoring. These assumptions represent the most construction have a bending moment capacity of
favourable conditions for the shoring in that the back- approximately 193 kN-m/m. Considering the
analyses ignore additional lateral pressure from aforementioned shortcomings of the original design
groundwater and competent soil conditions are assumed analysis and the likelihood of the actual peak bending
for the entire height of the retained soils, especially for the moment being greater than 157 kN-m/m, the Z55 shoring
analysis cases having soil friction angle values of 35° and wall likely did not have an adequate FoS (of 1.5) against
40°. Further, the back-analysis did not account for any bending failure. Furthermore, the lack of lateral bracing
temporary surcharge loading (e.g., equipment, stockpiles) along the height of the soldier piles and the resultant
on the ground surface adjacent to the shoring. exposed height of the soldier piles potentially made the
The required wall penetration depths for the various soldier piles susceptible to lateral torsional buckling.
back-analysis cases are presented in Table 2, below. For
loose to compact soil conditions, a D/H ratio ranging 7.4 Influence of Earth Berm
between 1.7 and 1.4 is needed to ensure a satisfactory
factor of safety (FoS) of 1.5 against overturning of the Considering the gross design deficiencies, the question
shoring. arises as to how the shoring system managed to stay
stable temporarily. Based on construction photographs
Table 2. Soil Parameters Adopted in Shoring Design (Figure 12), a temporary construction ramp was located at
the northeast corner of the tertiary excavation. It appears
Soil Friction Penetration Depth Required, Required that the shoring system was inadvertently reliant on this
Angle D (m) D/H* for berm for stability, similar to the berm that was subsequently
(Degrees)
For FoS = 1.0 for FoS=1.5
FoS=1.5 provided for additional support to the remediated shoring
wall (Figure 8). Once the earth berm was removed,
Loose (30) 9.0 10.4 1.73 reportedly two days before the failure, the shoring
collapsed shortly after.
Compact (35) 7.4 8.5 1.42
Dense (40) 6.0 7 1.17
* H=6 m, the top 3.8 m of sloped ground considered as surcharge loading

Even for a soil friction angle of 40°, which represents

relatively dense soil conditions requiring the least support
from a shoring wall and which is higher than the friction
angles recommended in the project geotechnical report, a
minimum D/H ratio of approximately 1.2 is required for a
satisfactory design with FoS = 1.5. Even for marginal
stability/incipient failure conditions (FoS=1.0), an
embedment ratio of 1.0 is required for the dense soil case.
In comparison, an embedment ratio of only 0.4 (based on
a penetration depth of 2.38 m, see Figure 7) was provided
in the shoring design, which was grossly inadequate and
rendered the shoring susceptible to overturning failure.
Even if one considers the additional embedment depth of
1.5 m for the solider piles placed in front of the sheet piles Figure 12. Construction access ramp at northeast corner
(although the nature of their connection to the sheet pile (photograph taken approximately 1.5 months before
wall, if any, is suspect, and the soldier piles were located failure)
approximately 4.6 m apart along the sheet pile wall), a
maximum embedment ratio of 0.65 is obtained which is
less than what is required for minimum stability.
Since the shoring system is unstable even under
competent soil conditions and without groundwater
pressure, an adverse change in anticipated site conditions
7.5 Other Contributory Factors the design of excavation support systems The case history
serves as another example of the general lack of attention
7.5.1 Lack of Shoring Design Review, Construction and to the design of temporary works in the industry despite the
Performance Inspection risk of severe consequences associated with failures.
Based on the lessons learned from this investigation, some
In accordance with contract requirements, settlement of the important aspects to be mindful of in future projects
monitoring targets were established on the face of the are as follows:
existing buildings and existing tanks at the project site.
These targets were intended to check for potential • Checking stability of shoring systems using
settlements resulting from the excavation dewatering fundamental equations of statics (e.g., force and
operation. However, the contract specifications did not call moment equilibrium) in combination with
for survey targets to be established on the shoring system analyzing the outputs from design analysis
itself or the ground immediately behind the shoring system. software programs
It appears that the shoring designer was also not afforded • Review of shoring design by the project
an opportunity to inspect the as-built shoring system during geotechnical engineer to ensure consistency with
construction. Additionally, it is not known whether the geotechnical recommendations
project geotechnical engineer was afforded the opportunity • Co-ordination among the project teams to ensure
to review the geotechnical aspects of the shoring design that shoring design submittals have accounted for
for consistency with their recommendations. recommendations by other
consultants/contractors (e.g., dewatering
7.5.2 Climatic Factors contractor)
• Displacement monitoring of the shoring
As part of the forensic investigation, the precipitation data system/adjacent ground to potentially serve as an
recorded at the nearest weather station was reviewed early warning system of design issues
(heavy rainfall was cited by others as cause of failure).
Approximately 17 mm of rain fell 10 days before the shoring 9 REFERENCES
collapse. Approximately 22 mm of rainfall was recorded the
day after the failure. For comparison, the three weeks Canadian Geotechnical Society. 2006. Canadian
before the failure experienced a total of 26 mm of rainfall, Foundation Engineering Manual, 4th Edition, BiTech
compared to a total of 74 mm of rainfall in the same period Publisher Ltd.
the previous year. Therefore, it can be concluded that no Kulhawy, F.H. and Mayne, P.W. 1990. Manual on
unusually heavy rainfall event was noticed leading up to the Estimating Soil Properties for Foundation Design,
failure. Electric Power Research Institute EL-6800, Project
Regarding the winter construction and possible effects 1493-6
of frost, there was some evidence of freezing and thawing
during the 3-week period leading up to the failure, with daily
mean temperatures varying between approximately -19°C
and 6°C. Fluctuating daily temperatures have the potential
to induce additional lateral stresses on shoring walls during
freezing and/or relaxation of restraining structural
elements.
Considering the major design deficiencies discussed
above, it is unlikely that climatic factors would have had any
dominant influence on the overall stability of the shoring.

8 CONCLUSIONS

The shoring system collapsed likely because of inadequate

embedment of the walls below the base of the excavation
in combination with a lack of sufficient horizontal support
systems in the forms of tiebacks, struts or rakers. While
overturning appears to be the dominant mode of failure, the
Z55 sheet pile section was also too light from a flexural
capacity point of view for a shoring wall that exhibited
cantilever-like response following excavation. The use of
overestimated soil strength and stiffness parameters and
inadequate evaluation of design software outputs such as
deflection magnitude and deflected shape of the shoring
wall were some of the contributory factors to the failure.
The failure case study presented in this paper
reinforces the importance of adhering to the fundamental
principles of geotechnical and geostructural engineering in
Generation of In-situ Stresses for Analysis
of Progressive Failure Mechanism in a
Sensitive Clay Slope
Andries Kirstein, Ariane Locat
Université Laval, Québec City, QC, Canada
Hans Petter Jostad
Norwegian Geotechnical Institute, Oslo, Norway
Gustav Grimstad
Norwegian University of Science and Technology, Trondheim, Norway

ABSTRACT
The in-situ stress situation in a sensitive clay slope plays an important role in determining whether the slope will remain
stable following a disturbance (such as toe erosion, earthquake, etc.). The paper compares the in-situ stress state
calculated using two soil models available in Plaxis 2D. These analyses are part of a larger project to understand the
mechanism of the initiation of spreads that is conducted using the UDSM “NGI-ADPsoft” in Plaxis. It is numerically
convenient to generate in-situ stresses using NGI-ADP as stress is guaranteed to be limited by the failure criterion of NGI-
ADPSoft. However, since NGI-ADP is a total stress model, the paper investigates the validity of using NGI-ADP to generate
in-situ stresses resulting from valley erosion in the drained condition, as compared to Hardening Soil Model. It is concluded
that it is reasonable to use NGI-ADP to generate in-situ stresses for the case study considered.

RÉSUMÉ
La situation de contrainte in situ dans un talus argileux sensible joue un rôle important pour déterminer si le talus restera
stable suite à une perturbation (telle que l'érosion du pied, un tremblement de terre, etc.). L'article compare l'état de
contrainte in situ calculé à l'aide de deux modèles de sol disponibles dans Plaxis 2D. Ces analyses font partie d'un projet
plus vaste visant à comprendre le mécanisme d'initiation des étalements qui est mené à l'aide de l'UDSM « NGI-ADPsoft »
dans Plaxis. Il est numériquement pratique de générer des contraintes in situ à l'aide de NGI-ADP car la contrainte est
garantie d'être limitée par le critère de rupture de NGI-ADPSoft. Cependant, étant donné que NGI-ADP est un modèle de
contrainte totale, l'article étudie la validité de l'utilisation de NGI-ADP pour générer des contraintes in situ résultant de
l'érosion de la vallée à l'état drainé, par rapport au modèle Hardening Soil. Il est conclu qu'il est raisonnable d'utiliser NGI-
ADP pour générer des contraintes in situ pour l'étude de cas considérée.

1 INTRODUCTION horizontal failure plane behind the toe of a sensitive clay

slope. An important aspect of the theory is that the horsts
Determination of in-situ stresses is an important (and and grabens that characterize spreads develop after strain-
sometimes underappreciated) step in modelling a softening has occurred along the horizontal failure plane.
geotechnical problem, as they can affect failure The study showed that higher in-situ shear stresses along
mechanisms obtained in later calculation phases when the the potential failure plane resulted in less disturbance being
material is state dependent. It is thus important to have required to initiate the progressive failure.
confidence in the procedure by which in-situ stresses are Recently, other researchers have obtained, using
generated. This paper presents a comparison of the in-situ advanced 2-dimensional numerical techniques, final
stresses calculated by three soil models for a valley slope topographies that resemble spreads. These techniques
in a deposit of normally consolidated sensitive marine clay include large deformation finite element analyses (e.g. Dey
located in the province of Québec that was subject to a et al. (2015), Wang et al. (2021, 2022)) and use of the
spread landslide in 1994. material point method (e.g. Tran & Solowski (2019),
The modelling is part of a larger project to understand Troncone et al. (2023)). In parametric studies, Dey et al.
the initiating mechanism of spreads in homogeneous (2015) concluded that a higher lateral earth pressure ratio
sensitive clay. A sensitive clay spread is a retrogressive increases the likelihood of initiating a spread rather than a
landslide characterized by intact blocks of clay (known as flowslide, and Wang et al. (2021) noted that the in-situ
horsts and grabens) that translate along a quasi-horizontal deviatoric strains differ depending on whether or not
plane on which the shear strength of the clay has been seepage forces were included in the calculation of in-situ
reduced to its large-deformation strength (Cruden & stresses, which could influence the initiation of failure. The
Varnes (1996), Locat et al. (2011)). The retrogression of aforementioned studies have shown that the initiation and
spreads is commonly understood to occur due to post-rupture behaviour of spreads can be successfully
progressive failure in the undrained condition (e.g. Locat et captured by numerical models. However, these studies
al. (2011)). generally provide limited discussion on the mechanism
Locat et al. (2013) outlined a theory of progressive resulting in the initiation of spreads, although it is generally
failure to explain the initiation and propagation of spreads, agreed that the in-situ stresses play an important role.
based on 1-dimensional analysis along an assumed
 The final goal is to investigate the spread mechanism earth pressure ratio at rest. Details on NGI-ADP are
using the NGI-ADPsoft soil model (description available in available in the Plaxis User Manual (Plaxis bv, 2021b).
(Grimstad et al., 2010)). In-situ stresses are generally first
calculated using a different soil model, followed by a 2.2.2 Hardening Soil Model
transition to NGI-ADPsoft. It is numerically advantageous
to generate in-situ stresses using NGI-ADP (Plaxis bv., Hardening Soil Model (HSM) is an isotropic, effective
2021b) as it guarantees that the initial stress state is at or stress soil model based on a Mohr-Coulomb failure
below the yield surface. However, since NGI-ADP is based criterion (𝑐′ and 𝜙′). A key feature of the model is stress
on total stresses, it must first be determined whether the in- dependent stiffness and strength defined by the reference
situ stresses calculated by NGI-ADP are reasonable. The 𝑟𝑒𝑓
stiffness 𝐸50 (which is the secant stiffness at 50% of the
primary purpose of the present paper is to evaluate the in- peak strength under axial loading at the reference stress
situ stresses calculated using NGI-ADP through 𝑝𝑟𝑒𝑓 ), the power 𝑚, and the Mohr-Coulomb strength
comparison with results from Hardening Soil Model (HSM). parameters. The unloading-reloading stiffness (𝐸𝑢𝑟 ) and
To gain further insight into the results, a simple Linear
oedometer stiffness (𝐸𝑜𝑒𝑑 ) are defined similarly.
Elastic Model is also included. The stresses along a
Additionally, the model incorporates parabolic stress-strain
horizontal cross section (see HXS in Figure 1) representing
behaviour as a function of mobilisation of the Mohr-
the potential failure surface are emphasized.
Coulomb failure criterion controlled by the stiffness inputs
The goals of the comparison of the soil models are: (i)
and the parameter 𝑅𝑓 . Details on HSM are also available
to evaluate whether it is appropriate to use NGI-ADP to
generate in-situ stresses for the case study of the spread in Plaxis User Manual (Plaxis bv, 2021b).
at Sainte-Monique-de-Nicolet and (ii) to understand the
parameters influencing these stresses. 2.2.3 Linear Isotropic Elastic Model

2 METHODOLOGY The linear isotropic elastic model, available in Plaxis, has

the option that Young’s modulus can be defined to increase
2.1 Case Study linearly with depth. The model is described in the Plaxis
User Manual (Plaxis bv, 2021b). This model was included
In 1994, a spread occurred in the village of Sainte- to compare the results from NGI-ADP and HSM to a soil
Monique-de-Nicolet, QC. Subsequently, an extensive model with no plasticity.
geotechnical investigation, including both field and
laboratory tests, was carried out at the landslide site, as 2.3 Slope Geometry and Boundary Conditions
reported by Locat et al. (2015). The deposit in which the
spread occurred was found to be normally consolidated, The geometry of the slope and applied boundary conditions
sensitive marine clay with an approximately linear increase are shown in Figure 1 and are similar to those used by
of undrained shear strength with depth. Locat et al. (2015). The slope is divided into an upper slope
The spread at Sainte-Monique-de-Nicolet is used in the and a lower slope, which is steeper. The slope has a height
current research to study the mechanism by which spreads of 16.6 m and length of 42.5 m. The stream bed at the right
are initiated in sensitive clay. This case study was selected boundary is modelled as horizontal. Behind the crest of the
due to the amount of data available for the site as well as slope, the ground is modelled as horizontal and extends a
the existence of previous studies modelling the spread for distance of 110 m such that stress gradients near the left
comparison. The software used for the study is Plaxis 2D domain boundary are negligible. The left and right
(Plaxis bv., 2021a). boundaries are both fixed in the horizontal direction only.
The horizontal cross-section behind the toe shown as a
2.2 Constitutive Models black dashed line (HXS) represents a potential failure
surface for a spread. Plane strain condition is assumed.
2.2.1 NGI-ADP Along the left and right domain boundaries, the soil
body is allowed to deform vertically, as represented by grey
NGI-ADP is a total stress soil model that is based on a circles in Figure 1. Along the bottom domain boundary, the
translated slightly approximated Tresca failure criterion. soil body is fixed both horizontally and vertically, as
Directional dependent undrained shear strengths are represented by grey triangles in the same figure.
defined directly by the user, with optional linear strength Hydraulic boundary conditions are shown as blue
increase with depth within a layer. To account for soil dashed lines in Figure 1. The phreatic surface generally
anisotropy, peak strengths and associated plastic shear follows the ground surface. The total head at the bottom
strains at these strengths are defined for three modes of boundary, representing the interface with an underlying till
layer, is 16.16 m, equivalent to 240 kPa. There is thus a
failure: active (where 𝑆𝑢𝐴 is the active peak shear strength),
downward hydraulic gradient to the left of the slope, which
direct simple shear (DSS, controlled by the input parameter
decreases toward and below the slope. These hydraulic
𝑆𝑢𝐷𝑆𝑆 /𝑆𝑢𝐴 ) and passive (controlled by the input parameter
boundary conditions are consistent with field observations
𝑆𝑢𝑃 /𝑆𝑢𝐴 ). The elastic soil stiffness is assumed isotropic and
as discussed by Locat et al. (2015). In the current paper,
is defined through the input parameter 𝐺/𝑆𝑢𝐴 , where 𝐺 is a stream water level of 1 m is assumed near the right
the shear modulus during initial loading and unloading. domain boundary.
Initial active shear stress is defined through the input
parameter 𝜏0 /𝑆𝑢𝐴 . This parameter also controls the lateral
Figure 1. Slope geometry and boundary conditions

2.4 Workflow strength increase with depth, 𝑆𝑢𝐴 𝑖𝑛𝑐 = 2.7 𝑘𝑃𝑎/𝑚, was
extrapolated from the undrained strength profile reported
In-situ stresses were generated following the approach by Locat et al. (2015), although the weathered crust was
used by Locat et al. (2015). In the initial phase, a perfectly neglected. As the deposit is normally consolidated, it is
flat ground surface is assumed, with the phreatic surface at reasonable to assume that the undrained strength
the ground elevation and the hydraulic boundary condition approaches zero at the ground surface; thus, a small value
at the bottom boundary as shown in Figure 1. The valley of 2 kPa was selected for 𝑆𝑢𝐴 𝑟𝑒𝑓 . The parameter 𝜏0 /𝑆𝑢𝐴 of
is then excavated in a single step in the drained condition,
0.9 was selected to satisfy the requirement 𝐾0 ≈ 0.5, as
applying the boundary conditions shown in Figure 1.
discussed later in the current paper.
2.5 Soil Parameters Used
Table 1. Parameters used in the NGI-ADP soil model
The input parameters used for NGI-ADP and HSM are
shown in Tables 1 and 2, respectively. For all soil models,
since the soil is assumed saturated throughout the domain, Parameter Value
a constant value of 𝛾 = 𝛾𝑠𝑎𝑡 = 16 𝑘𝑁/𝑚3 was used. The 𝐺𝑢𝑟 /𝑆𝑢𝐴 280
drained Poisson ratio with a value of 𝜈 ′ = 0.25 was used. 𝛾𝑓𝐶 1.275 %
2.5.1 NGI-ADP 𝛾𝑓𝐷𝑆𝑆 2.07 %
𝛾𝑓𝐸 3.5 %
In Table 1, the symbol 𝛾𝑓 represents failure shear strain,
and the superscripts A, DSS and P represent the active, 𝑆𝑢𝐴 𝑟𝑒𝑓 2 kPa
direct simple shear, and passive modes of failure, 𝑌𝑟𝑒𝑓 36.7 m
3
respectively. The parameters 𝐺𝑢𝑟 /𝑆𝑢𝐴 of 280, 𝛾𝑓𝐶 = 𝜖1𝐶 of 𝑆𝑢𝐴 𝑖𝑛𝑐 2.7 kPa/m
2
1.275%, 𝛾𝑓𝐷𝑆𝑆 of 2.07%, and 𝑆𝑢𝐷𝑆𝑆 /𝑆𝑢𝐴 of 1 were interpreted 𝑆𝑢𝐷𝑆𝑆 /𝑆𝑢𝐴 1
from DSS and triaxial laboratory test data from Locat et al. 𝑆𝑢𝑃 /𝑆𝑢𝐴 0.4
(2015) as well as additional internally available data for the 𝜏0 /𝑆𝑢𝐴 0.9
site. The value of 𝑆𝑢𝐷𝑆𝑆 /𝑆𝑢𝐴 of 1 was chosen to be consistent
with laboratory results; since the peak undrained strength
in the DSS failure mode is generally expected to be lower
than the active peak undrained strength (Plaxis bv, 2021b), 2.5.2 Hardening Soil Model
this value will be adjusted in future analyses as more data
becomes available. Locat et al. (2015)’s laboratory The values of the soil parameters used in HSM are shown
program did not include triaxial extension tests; thus, the in Table 2 and were adopted from (Locat et al., 2015). 𝐾0𝑁𝐶 ,
parameter 𝛾𝑓𝐸 of 3.5% was selected to be toward the lower the normally consolidated lateral earth pressure ratio at
rest, was defined using Jaky’s equation, 𝐾0 = 1 − sin 𝜙 ′ .
end of the recommended range of 3%-8% (Plaxis bv,
2021b) and the parameter 𝑆𝑢𝑃 /𝑆𝑢𝐴 of 0.4 was selected to fall
within a range of observed values, from a collection of data
from Scandinavian clays (D’Ignazio, 2016). These
parameters will also be refined in the future. The
parameter controlling
 Crest Toe
(a) Max
.

𝜃 = −45°

(b)

(c)
𝜎1 − 𝜎3 Max
𝜎𝑥𝑦 = sin 2𝜃 .
2

(d)

Far behind crest:

1 𝑑𝑃𝑥
𝑃𝑥 = 𝐾0𝑇𝑂𝑇 𝛾𝐻 2 𝜎𝑥𝑦 =
2 𝑑𝑋 Inflection Points

Figure 2. Distribution of selected stress parameters along horizontal cross-section behind toe (see HXS, Figure 1):
𝜎 −𝜎
(a) maximum shear stress ( 1 3); (b) principal stress direction (𝜃); (c) cartesian shear stress (𝜎𝑥𝑦 ); and (d) lateral
2
side force (𝑃𝑥 )
Table 2. Parameters used in Hardening Soil Model 3.1.1 Factor of Safety

Parameter Value The results from the analyses with the two different soil
models can broadly be compared by looking at the Factor
𝑟𝑒𝑓
𝐸50 10 000 kPa of Safety resulting from a c-phi reduction analysis. As
𝑟𝑒𝑓
𝐸𝑜𝑒𝑑 10 000 kPa shown in Table 3, both result in 𝐹𝑆 ≈ 1.1 using the input
parameters summarized in Tables 1 and 2.
𝑟𝑒𝑓
𝐸𝑢𝑟 30 000 kPa
𝑚 1
Table 3. Factors of Safety obtained from c-phi reduction for
𝑐′ 7.5 kPa
NGI-ADP and Hardening Soil Model (HSM)
𝜙′ 30⁰
𝑝𝑟𝑒𝑓 100 kPa Soil Model NGI-ADP Hardening Soil Model
𝐾0𝑛𝑐 0.5 Factor of Safety 1.12 1.08
𝑅𝑓 0.9

Although the calculated Factors of Safety are similar,

the shapes of the slip surfaces are different, as shown in
2.5.3 Linear Elastic Model
Figure 3. The slip surface obtained with NGI-ADP is
approximately circular and quite deep, which is
Young’s modulus was defined to increase linearly with
characteristic of an undrained slip surface. The slip surface
depth, having a value of 1143 𝑘𝑃𝑎 at the crest elevation
obtained with HSM is much shallower (slope circle) and is
and an increase of 472 𝑘𝑃𝑎/𝑚. These values were
characteristics of a drained slip surface. The result of a
extrapolated from the results from Hardening Soil Model
similar Factor of Safety between the two soil models is
that were calculated by Plaxis based on the inputs
consistent with the finding reported by (Leroueil, 2001, see
described in the previous section. The calculated stresses
Fig. 74) that, for a given slope, the difference between the
were unaffected by whether or not the stiffness profile was
short-term and long-term factors of safety is a function of
multiplied by a coefficient of 3, meant to represent E 50 vs
OCR and approaches -0.1 for OCR = 1.0, which can be
Eur. 𝐾0 = 0.5 was once again used.
considered negligible.
3 RESULTS AND DISCUSSION

The following sections will highlight results and provide

discussion that are relevant to the goals stated in the
introduction. First, only Hardening Soil Model (HSM) and
NGI-ADP will be compared; then, the results from the
Linear Elastic Model will also be discussed.

3.1 Comparison of NGI-ADP and Hardening Soil Model

The results from HSM and NGI-ADP are compared based

on several parameters. First, the Factor of Safety and slip
surface shape from a c-phi reduction analysis is discussed,
followed by three vertical profiles of the stress ratio 𝜎𝑥′ /𝜎𝑦′ .
The subsequent discussion focuses on the distribution of
various stress parameters both in 2D and along the
Figure 3. Slip surface shapes predicted by c-phi reduction
horizontal cross-section (HXS) defined in Figure 1.
for NGI-ADP and Hardening Soil Model
Figure 2 shows the distributions of the following
parameters along the HXS: (a) maximum shear stress
𝜎 −𝜎
( 1 3), (b) principal stress direction (𝜃), and (c) cartesian 3.1.2 Coefficient of Lateral Earth Pressure at Rest
2
shear stress (𝜎𝑥𝑦 ). From Mohr’s circle, these stress
parameters are related by Equation 1. Figure 2 (d) shows In the current paper, the symbol K0 is used to represent the
the distribution of the resultant horizontal total force (Px) on coefficient of lateral earth pressure at rest in terms of
vertical slices above the toe elevation. In Figure 2, the effective stresses, and the symbol K0TOT is used to
results from NGI-ADP are in black and those from HSM are represent the same coefficient in terms of total stresses. In
in red. HSM, K0 is a direct input.
In NGI-ADP, the coefficient of lateral earth pressure at
rest is controlled by the input parameter 𝜏0 /𝑆𝑢𝐴 , as seen
𝜎𝑥𝑦 =
𝜎1 −𝜎3
sin 2𝜃 (Eq. 1) from the relationship shown in Equation 2. When 𝑆𝑢𝐴 𝑟𝑒𝑓 >
2
0, 𝐾0 increases with depth and approaches a constant
value. In Equation 2, 𝜏0 /𝑆𝑢𝐴 represents the ratio of active
in-situ shear stress to the peak active undrained shear 𝐾0 = 0.5 reflects the direct input value. For NGI-ADP, the
strength, Y represents elevation, Yref is a reference stress ratio deviates from 0.5 near the ground surface, but
elevation (taken in the analysis as the crest elevation of converges to approximately this value below approximate
36.7 m), and 𝛾′ is effective unit weight. elevation 30 m, resulting from the input value for 𝜏0 /𝑆𝑢𝐴 as
per Equation 2.
Below midslope, the ratio 𝜎𝑥′ /𝜎𝑦′ at the bottom domain
1 ′ boundary is close to 0.5 for both soil models, and generally
𝜏0 (𝜎 − 𝜎𝑥′ )𝑖𝑛𝑖𝑡𝑖𝑎𝑙
= 2 𝑦 increases to approximately 4.9 for NGI-ADP and 2.4 for
𝑆𝑢𝐴 𝑆𝑢𝐴 𝑟𝑒𝑓 + 𝑆𝑢𝐴 𝑖𝑛𝑐 ∗ (𝑌𝑟𝑒𝑓 − 𝑌) HSM at the ground surface, which is represented in
(Eq. 2) Figure 4 by the dashed line labelled “Midslope Elevation”.
1 ′
𝛾 𝐻 1−𝐾0 The curve representing HSM shows a local decrease in
2
= 𝜎𝑥′ /𝜎𝑦′ around elevation 15 m that is not present in the curve
𝑆𝑢𝐴 𝑟𝑒𝑓 +𝑆𝑢𝐴 𝑖𝑛𝑐 ∗ 𝑌𝑟𝑒𝑓 −𝑌
representing NGI-ADP. However, the approximate
magnitude of 𝜎𝑥′ /𝜎𝑦′ is similar along this VXS for both soil
models.
For comparison of NGI-ADP and HSM, as well as to
Near the right domain boundary, which represents the
observe the influence of the slope, the distributions of the
valley centre, the ratio 𝜎𝑥′ /𝜎𝑦′ increases for both soil models
stress ratio 𝜎𝑥′ /𝜎𝑦′ along three vertical cross-sections (see
from close to 0.5 at the lower boundary to values of greater
VXS locations in Figure 1) are shown in Figure 4, with solid
than 15 for NGI-ADP and greater than 6 for HSM at the
lines representing NGI-ADP and dashed lines representing
ground surface, which is represented by the dashed line
HSM.
labelled “Toe Elevation”. The horizontal axis on Figure 4
stops at 𝜎𝑥′ /𝜎𝑦′ = 5 to show the shape of the distributions.
Here, the ratio 𝜎𝑥′ /𝜎𝑦′ results from a combination of vertical
unloading and an additional influence of the stresses
imposed by the adjacent slope.

𝜎1 −𝜎3
3.1.3 Maximum Shear Stress, 𝜏𝑚𝑎𝑥 =
2

The distribution of 𝜏𝑚𝑎𝑥 along the HXS is shown in Figure

2 (a). Far behind the crest, near the left boundary in
Figure 1, the influence of the slope is negligible and the
principal stress difference is controlled by K 0. The slight
differences obtained by NGI-ADP and HSM at this location
along the HXS thus result from the different ways K 0 is
defined in the two soil models.
Approaching the slope, 𝜏𝑚𝑎𝑥 increases to a maximum
value of 48.5 ± 0.1 𝑘𝑃𝑎 that occurs below the slope, then
decreases to a minimum value of 15.8 ± 2.5 𝑘𝑃𝑎 near the
toe. The initial increase results from the stress rotation that
occurs beginning approximately 20 m behind the crest, as
will be discussed in Section 3.1.4 for Figure 2 (b). The
subsequent decrease toward the toe occurs when the
influence of the decreasing slope height surpasses the
influence of the stress rotation. The distribution of 𝜏𝑚𝑎𝑥 is
similar for both NGI-ADP and HSM. For both soil models,
the maximum 𝜏𝑚𝑎𝑥 and maximum 𝜎𝑥𝑦 do not occur at the
same location, except when 𝜃 = 45° (as per Eq. 1).

3.1.4 Principal Stress Direction, 𝜃

Figure 4. Comparison between NGI-ADP and HSM for
The distributions of principal stress direction along the HXS
σ’x/σ’y profiles along three vertical cross-sections (VXS):
is shown in Figure 2 (b). An angle of 𝜃 = 0 corresponds to
1 m from Left Domain Boundary; Midslope; and 0.1 m from
a vertical major principal stress and +𝜃 increases in the
Right Domain Boundary (see Fig. 1 for locations)
clockwise direction (Plaxis bv, 2021a). Far behind the
crest, as the ground surface is horizontal, the influence of
the slope is minimal, and 𝐾0𝑇𝑂𝑇 < 1, we thus have 𝜃 = 0.
Near the left boundary, the ratio 𝜎𝑥′ /𝜎𝑦′ represents 𝐾0𝑁𝐶
Starting approximately 20 m behind the crest of the slope,
since the ground surface here is flat and the stress rotation a stress rotation begins to occur, resulting in increasing 𝜃.
is essentially zero. The elevation of the ground surface
The maximum value of 𝜃 = 71.2° ± 0.2° occurs at the toe.
behind the crest is represented in Figure 4 by a dashed line
This value is strongly influenced by the slope angle – since
labelled “Crest Elevation”. For HSM, the consistent result
𝜎3 = 0 along the slope face, the principal stress direction
along the ground surface must therefore be parallel to the 3.1.6 Resultant Horizontal Total Force for Slice Above
ground surface. The observed value of 𝜃 differs from the Toe Elevation, Px
expected value of 90° − 24.3° = 65.7° by about 5.5°,
possibly due to influence of corner effects. The distribution The distributions along the HXS of resultant horizontal total
of 𝜃 below the slope thus results from a combination of force Px that are shown in Figure 2 (d) were obtained by
slope geometry and soil response. specifying a series of vertical cross-sections from the toe
elevation to the ground surface.
3.1.5 Cartesian Shear Stress, 𝜎𝑥𝑦 From consideration of horizontal equilibrium within a
stable slope (e.g. as discussed by Locat et al. (2013);
Figure 5 presents the distributions of 𝜎𝑥𝑦 in the entire 2D Bernander et al. (2016)), any vertical slice of width Δ𝑥 with
domain from both NGI-ADP and HSM. For both models, a base at the toe elevation will have an average cartesian
near the left boundary, where the ground surface is shear stress at the base of the slice that is related to the
horizontal the influence of the slope is negligible. Cartesian resultant horizontal total forces on the sides of the slice (𝑃𝑥 )
shear stresses below the valley centre are also as per Equation 4.
approximately zero due to the modelled valley symmetry.
For both soil models, 𝜎𝑥𝑦 is higher below the slope, having
similar approximate magnitudes and increasing with depth. 𝜎𝑥𝑦 ∗ Δ𝑥 = 𝑃𝑥𝑖+1 − 𝑃𝑥𝑖 (Eq. 4)
However, the maximum value is lower for HSM (𝜎𝑥𝑦 =
62.1 𝑘𝑃𝑎) than for NGI-ADP (𝜎𝑥𝑦 = 75.6 𝑘𝑃𝑎).
Equation 4 can alternatively be represented as 𝜎𝑥𝑦 =
𝑑𝑃𝑥 /𝑑𝑋. This relationship is demonstrated by considering
the maximum cartesian shear stress (𝜎𝑥𝑦 𝑚𝑎𝑥 ) in Figure
2 (c), which coincides with the inflection point of the Px
distribution in Figure 2 (d), as indicated by vertical lines
labelled 𝜎𝑥𝑦 𝑚𝑎𝑥 .
Near the left boundary, where the ground surface is
horizontal and the influence of the slope is negligible, the
value of 𝑃𝑥 from both soil models is 1469 ± 1 𝑘𝑃𝑎, a value
that is consistent with Equation 5.

1
𝑃𝑥0 = 𝐾0𝑇𝑂𝑇 𝛾𝐻02 (Eq. 5)
2

Figure 5. Cartesian shear stress (σxy) resulting from (a)

NGI-ADP (b) HSM (note: colour scale applies to both The value for both soil models begins to decrease
figures) starting at some distance behind the crest and continues to
decrease to a minimum value near the toe. The overall
shapes of the distributions of Px are similar between
Figure 2 (c) highlights the distribution of 𝜎𝑥𝑦 on the NGI-ADP and HSM. This is expected, considering the
horizontal cross-section (HXS) defined in Figure 1. The similarity of the distributions of 𝜎𝑥𝑦 , as discussed
maximum value from both soil models is 𝜎𝑥𝑦 = 38.2 ± previously.
0.2 𝑘𝑃𝑎, occurring at similar locations approximately below
midslope as indicated by arrows in Figure 2 (c). Left of 3.2 Comparison to Linear Elastic Model
approximately 𝑋 = 10 𝑚, the distribution from NGI-ADP
drops more rapidly than that from HSM. This is Examination of Figure 2 (a), (b), and (c) shows that the
compensated for by HSM having slightly elevated values results obtained from the Linear Elastic Model, shown in
near the toe of the slope relative to NGI-ADP to respect the green, are quite different from those obtained by NGI-ADP
horizontal equilibrium demonstrated in Equation 3, which and HSM. At the toe, a very high value of 𝜏𝑚𝑎𝑥 =
requires that the summation of cartesian shear stresses 151.4 𝑘𝑃𝑎 was obtained from the Linear Elastic model
along the HXS be equal to the horizontal total force that (higher than the extent of the vertical scale used in
was removed during valley unloading (see Locat et al. Figure 2 (a)). It is expected that this value depends on
(2013)). Overall, the distributions of 𝜎𝑥𝑦 from NGI-ADP and element size and that 𝜏𝑚𝑎𝑥 tends toward infinity for this soil
HSM along the HXS are similar. model. A high value of 𝜎𝑥𝑦 was obtained for similar
reasons. In contrast, 𝜏𝑚𝑎𝑥 decreases to a minimum at the
toe for both NGI-ADP and HSM. Notably, 𝜏𝑚𝑎𝑥 near the
−∞
∫𝑇𝑜𝑒 𝜎𝑥𝑦 𝑑𝑋 = 𝑃𝑥 0 (Eq. 3) left boundary is nearly identical between HSM and the
Linear Elastic Model; this reflects the fact that the K0
procedure with a specified value of 0.5 was used to
calculate initial stresses for both soil models.
The result from the Linear Elastic Model that is most
like those from NGI-ADP and HSM is the principal stress
direction (𝜃). This results from the fact that the value of 𝜃 to generate in-situ stresses using NGI-ADP. The method
at both extreme ends of the HXS is controlled by the for comparison presented here is relevant for analyses to
geometry, as discussed previously. However, the be performed in future works investigating the mechanism
distribution of 𝜃 below the slope is visibly different. of the initiation of spreads in sensitive clay.
This comparison to the Linear Elastic Model suggests Additional field work is planned at the site at Sainte-
that the similarity in results between NGI-ADP and HSM Monique-de-Nicolet, including CPTu tests and a large-
results from both the geometry and the soil models. It also diameter borehole. The additional field and laboratory data
highlights the extent to which the results obtained using will be used to supplement the input data used in the
NGI-ADP and HSM are similar to each other. modelling. Ultimately, the purpose of the analysis
summarized in this paper was to generate in-situ stresses
4 CONCLUSION for subsequent calculation phases that will simulate the
initiation of sensitive clay spreads. The insights gained
The goals of the present paper were to evaluate whether it through the in-situ modelling will thus be used in the
is appropriate to use the soil model NGI-ADP to generate investigation of the spread landslide mechanism.
in-situ stresses for the case study presented in Section 2.1,
and to provide insight into the factors that influence these 5 ACKNOWLEDGEMENTS
stresses. In-situ stresses calculated using NGI-ADP were
compared to those obtained using Hardening Soil Model This work is being carried out using funding from Cadre
(HSM) and a Linear Elastic Model. Emphasis was pour la prevention des sinistres du gouvernement du
generally placed on parameters that are relevant to the Québec (CPS 21-22-20). The writers acknowledge the
analysis of spreads according to the theory of progressive valuable discussions with members of the Section des
failure, particularly in the 2D domain and along certain mouvements de terrain of the Ministère des Transports du
selected horizontal and vertical cross-sections. HSM was Québec, namely Denis Demers, Pascal Locat, and Julie
selected for comparison because it is a well-established Therrien.
soil model formulated in effective stresses, which is
appropriate for simulating valley excavation in the drained 6 REFERENCES
condition.
It was shown that the factor of safety obtained by both Bernander, S., Kullingsjo, A., Gylland, A.S., Bengtsson, P.-
NGI-ADP and HSM are approximately 1.1. The limit E., Knutsson, S., Pusch, R., Olofsson, J., and Elfgren,
equilibrium slip surface resulting from NGI-ADP was much L. 2016. Downhill progressive landslides in long natural
deeper than that from HSM, possibly due to the total stress slopes: triggering agents and landslide phases
formulation of NGI-ADP. modeled with a finite difference method, Canadian
It was also shown that, despite the very different Geotechnical Journal, NRC Research Press, 53: 1565-
formulations of the two soil models, similar values of lateral 1582.
earth pressure at rest below horizontal ground, as well as Cruden, D. and Varnes, D. 1996. Landslides types and
similar distributions of the stress ratio 𝜎𝑥′ /𝜎𝑦′ along vertical processes. In Landslides investigation and mitigation,
cross-sections, were obtained from the two soil models. Special Report 247, Transportation Research Board,
The distributions of maximum shear stress (𝜏𝑚𝑎𝑥 ), National Research Council. Edited by A.K. Turner and
principal stress direction (𝜃) and cartesian shear stress R.L. Schuster. National Academy Press, Washington,
(𝜎𝑥𝑦 ) were similar along a horizontal cross-section (HXS) D.C. pp. 37–75.
behind the toe that represents the potential failure surface Dey, R., Hawlader, B., Phillips, R., and Soga, K. 2015.
of a spread. Slight differences noted between the Large deformation finite-element modelling of
distributions from the two soil models were discussed. progressive failure leading to spread in sensitive clay
These three stress parameters are linked through slopes. Géotechnique, ICE, 65(8):657-668.
Equation 1, and it was demonstrated that it is not always D’Ignazio, M. 2016. Undrained shear strength of Finnish
evident which parameter controls the others except for clays for stability analyses of embankments, Tampere
some simple boundary cases. It was also shown that the University of Technology, Publication 1412 (PhD
distribution of resultant horizontal total force (𝑃𝑥 ) on vertical Thesis).
slices along the HXS is similar for the two soil models, and Grimstad, G., Jostad, H.P., and Andresen, L. 2010.
the relationship of this parameter to 𝜎𝑥𝑦 through Equation 4 Undrained capacity analyses of sensitive clays using
the nonlocal strain approach. In Proceedings of the 9 th
was discussed.
HSTAM International Congress on Mechanics,
Finally, it was shown that although similar stress
Vardoulakis Mini-Symposia, Limassol, Kypros, 12–14
distributions were obtained by NGI-ADP and HSM, the
July, pp. 153–160.
results obtained from the Linear Elastic Model were notably
Leroueil, S. 2001. Natural slopes and cuts: movement and
different. It was thus demonstrated that the similarity in the
failure mechanisms, Géotechnique, ICE, 51(3): 197-
results of NGI-ADP and HSM is a result not only of
243.
geometry but also of the soil response.
Locat, A., Leroueil, S., Bernander, S., Demers, D., Jostad,
Although the in-situ stresses calculated by NGI-ADP
H.P., and Ouehb, L. 2011. Progressive failures in
and HSM were not identical, they can generally be
eastern Canadian and Scandinavian sensitive clays,
considered similar for the parameters discussed. Based
Canadian Geotechnical Journal, NRC Research Press,
on these considerations, we conclude that, for the case of
48: 1696-1712.
the spread at Sainte-Monique-de-Nicolet, it is appropriate
Locat, A., Jostad, H.P., and Leroueil, S. 2013. Numerical
modeling of progressive failure and its implications for
spreads in sensitive clays, Canadian Geotechnical
Journal, NRC Research Press, 50: 961-978.
Locat, A., Leroueil, S., Fortin, A., Demers, D., and Jostad,
H.P. 1015. The 1994 landslide at Sainte-Monique,
Quebec: geotechnical investigation and application of
progressive failure analysis, Canadian Geotechnical
Journal, NRC Research Press, 52: 490-504.
Plaxis bv. 2021a. CONNECT Edition V21.01: Reference
Manual, Bentley Systems, Incorporated, The
Netherlands.
Plaxis bv. 2021b. CONNECT Edition V21.01: Material
Models Manual, Bentley Systems, Incorporated, The
Netherlands.
Tran, Q.-A. and Solowski, W. 2019. Generalized
Interpolation Material Point Method modelling of large
deformation problems including strain-rate effects –
Application to penetration and progressive failure
problems, Computers and Geotechnics, 106: 249-265.
Troncone, A., Pugliese, L., Parise, A. and Conte, E. 2023.
Analysis of a landslide in sensitive clays using the
material point method, Geotechnical Research, ICE.
Wang, C., Hawlader, B., Perret, D., Soga, K., and Chen, J.
2021. Modeling of Initial Stresses and Seepage for
Large Deformation Finite-Element Simulation of
Sensitive Clay Landslides, Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 147(11):
04021111.
Wang, C., Hawlader, B., Perret, D., and Soga, K. 2022.
Effects of geometry and soil properties on type and
retrogression of landslides in sensitive clays,
Géotechnique, ICE, 72(4): 322-336.
Experiences with Slope Instability in Lake
Saskatchewan Clays in the Saskatoon Area
Jon Osback, Wayne Clifton
Clifton Engineering Group, Saskatoon, SK, Canada

ABSTRACT
Glacial Lake Saskatchewan, formed during Wisconsinian deglaciation of southern Saskatchewan, deposited considerable
depths of deltaic and lacustrine sediments that exhibit an increase in clay content and plasticity with depth, typical of the
sedimentary sequence associated with an “on-lapping” delta. The geotechnical properties of these materials and their
associated impacts on the stability of the slopes of the South Saskatchewan River Valley have been studied extensively.
Of particular interest is the Beaver Creek Landslide near the southern limits of Saskatoon, located within the deeper
lacustrine clay sequences.

Natural processes have continued to precipitate slope instability within the Lake Saskatchewan plain as riverbank
development has progressed since the earlier studies. Choosing the appropriate shear strength for these clay sediments
is a perennial question for geotechnical slope stability analysts, since it is clear there are variances in the properties of the
clay sediments within the Lake Saskatchewan basin.

This paper will focus on addressing the strength of Lake Saskatchewan clays by presenting further analysis of the Beaver
Creek Landslide, supported by six additional case histories in which the strength of the clay was determined by both
laboratory testing and back-analysis. The results will be compared to published values for similar sediments, with
discussion on the application of this data to new slope stability analyses and remedial design.

RÉSUMÉ
Le lac glaciaire Saskatchewan, formé lors de la déglaciation du Wisconsinien dans le sud de la Saskatchewan, a déposé
des profondeurs considérables de sédiments deltaïques et lacustres qui présentent une augmentation de la teneur en
argile et de la plasticité avec la profondeur, typique de la séquence sédimentaire associée à un delta « à chevauchement
». Les propriétés de ces matériaux et l'impact associé sur la stabilité des pentes le long de la rivière Saskatchewan Sud
sont étudiés depuis au moins le siècle dernier. d'un intérêt particulier, le glissement de terrain de Beaver Creek près des
limites sud de Saskatoon, situé dans les séquences d'argile lacustre plus profondes.

Les processus naturels ont continué de précipiter l'instabilité des pentes dans la plaine lacustre du lac Saskatchewan au
fur et à mesure que le développement a progressé depuis les études précédentes. Le choix de la résistance au cisaillement
appropriée pour ces sédiments argileux est une question permanente pour les analystes géotechniques de la stabilité des
pentes, car il est clair qu'il existe des variations dans les propriétés des sédiments argileux dans le bassin du lac
Saskatchewan.

Cet article se concentrera sur la résistance des argiles du lac Saskatchewan en présentant une analyse plus approfondie
du glissement de terrain de Beaver Creek, étayée par quatre études de cas supplémentaires où la résistance de l'argile a
été déterminée à la fois par des méthodes de laboratoire et de rétro-analyse. De plus, les résultats seront comparés aux
valeurs héritées publiées pour des sédiments similaires, avec une discussion sur le comportement de ces sédiments en
ce qui concerne la stabilité des pentes, la conception corrective et les nouvelles constructions à Saskatoon.
1. INTRODUCTION The SSD were deposited through a variety of
Background processes; in the case of the LSC and upper SSD in
Slope instability and the associated impact to infrastructure Saskatoon, primarily glaciolacustrine, outwash, and fluvial
along the South Saskatchewan River Valley in Saskatoon, processes, during the Wisconsinan deglaciation of
SK has presented an ongoing challenge for city planners Saskatchewan. Glacial Lake Saskatchewan covered most
and engineers for at least a century (Clifton et al 1981). The of the province about 17,000 years ago. As the glaciers
City of Saskatoon has implemented various remedial continued to retreat and the lake drained, the Saskatoon
measures to counter these ever-occurring slope area emerged roughly 11,500 years ago, with SSD up to
movements throughout the city; the earliest being a 1913 100 m thick in some areas. Battleford till deposited by the
brick-and-mortar drainage conduit constructed parallel to glacier was veneered with highly plastic clay, accumulating
the river to intercept groundwater, with an outlet to the river. along the bottom of Lake Saskatchewan. As the glacier
Interestingly, the principle of effective stress had not yet retreated to the northeast, silt and sand carried by the
been introduced; however, the concept of increasing soil Saskatchewan rivers were deposited as expansive
strength by lowering the groundwater level appears to have prograding deltas over the LSC as the water flowed into
been understood by the engineers at the time. glacial Lake Saskatchewan, thereby forming the current
Other remedial measures used throughout Saskatoon upper stratigraphic sequences of the region.
include slope flattening, addition of toe berms, drainage Moving from south to north, the thickness of the SSD is
improvements, shear keys, light weight fill or, commonly, a about 36 m at the Beaver Creek site, thinning toward the
combination of these solutions. Further discussion on the 25th St. (University) bridge; north of the bridge, the SSD has
history of slope instabilities in Saskatoon is provided by been mostly eroded. The elevation of the SSD/till contact
Clifton et al (1981). is about 14 m below river level the contact at Beaver Creek.
Most slope failures within the Saskatoon River Valley Through the downtown area, the contact is located roughly
share a common characteristic: a shear plane originating 5 m to 10 m above river level on average. This stratigraphic
along the contact of upper Surficial Stratified Deposits progression is also apparent when looking at the behavior
(SSD) and the underlying glacial till, through a highly plastic of the South Saskatchewan River. South of the city, the
clay zone at the base of the SSD which will be referred to river meanders across the eroded deltaic SSD in a
as the Lake Saskatchewan Clay (LSC). floodplain up to about 6 km wide. The river valley through
Saskatoon is eroded into the till constraining the river flows
The aims of this paper were to: to a much narrower channel.
- Re-analyze the Beaver Creek landslide using
modern tools to compare the strength parameters 2 SITE LOCATIONS
of the LSC obtained through the previous work, The Beaver Creek Landslide lies just south of
while also considering more recent retrogression Saskatoon (Figure 1), outside the area discussed by Clifton
of the landslide which was not visible in the field in et al (1981); investigated in detail by Haug (1977) and
1985. Yoshida and Krahn (1985).
- Summarize the shear strength parameters of the
LSC at Beaver Creek and those used in the
analysis of six other landslides in Saskatoon.
- Compare the strength values of the LSC to
published correlations.
- Provide a general understanding of the shear
strength characteristics of the LSC after being
subjected to large strains (landslide movement)
based on back-analysis of landslides in
Saskatoon, and comment on the applicability of
this information to future work.

Geologic Setting

The geologic history and stratigraphy of the Saskatoon

area have been discussed in detail by Christiansen (1968,
1970,1979, 1992). Since the landslides being considered
in this paper occur only within the SSD and along the
surface of the till, lower stratigraphic units are not
discussed herein.
In ascending order, the local stratigraphy relevant to
this paper consists of glacial till of the Saskatoon Group
(Battleford formation overlying the Floral Formation)
overlain by variable SSD. The texture of the SSD tends to
become finer at depth. Sandier soils near surface grade
towards silty sands and silty clays, with the characteristic Figure 1 – Beaver Creek Landslide location
presence of highly plastic LSC at the SSD/till contact.
In addition to the re-analysis of the Beaver Creek landslide, Other than the Beaver Creek site, which was analyzed for
this paper considers the LSC strength values observed or research purposes, all other sites were analyzed to
measured at six other landslide sites in and around produce construction or remedial designs; however, the
Saskatoon. The details of some of the project sites must specifics of remediation and construction are outside the
remain confidential; permission was granted to use project scope of this paper.
data only. As such, only general locations and approximate
geometry of the sites are provided. Figure 2 shows the 3 LANDSLIDE FAILURE MECHANISMS IN
general areas of the project sites. SASKATOON
The majority of the landslides along the river valley in
Saskatoon share a common primary failure mechanism.
The upper, coarser SSD acts as an unconfined aquifer,
allowing runoff and precipitation to infiltrate, resulting in
saturation and progressive softening of the lower LSC,
leading to instability. Left unchecked, these failures
develop into classic translational retrogressive landslides.
In some cases, particularly at Beaver Creek, erosion along
the toe of the slope and subsequent destabilizing of the
river valley wall also contributes to a reduction in the Factor
of Safety (FoS). Similarly, adjustments to the topography
of the slope, or changes in vegetation (e.g., tree clearing)
can reduce the FoS. Variation in river level alone does not
typically impact the FoS significantly enough to initiate
failure; this was explored by Yoshida and Krahn (1985)
who found that variation of the piezometric level by up to 1
m resulted in only about a 4% change to the FoS at Beaver
Creek. Irrigation can play a significant role – the effect of
groundwater from various sources on slope stability FoS is
discussed in detail by a number of authors; Elwood and
Heilman (2023) provides a good summary and list of
references.
Figure 2 - General location of landslide sites
4 TYPICAL CHARACTERISTICS OF LAKE
For the purpose of this paper, the sites have been SASKATCHEWAN CLAY
numbered one through six moving south to north. The summary below considers laboratory testing data
Five of the six sites share similar stratigraphic and from the sites discussed in this paper as well as data
geometric conditions; SSD overlying till with the till surface discussed by other authors as summarized by Elwood and
located above the river elevation. Site 4 is in an area where Heilman (2023). In general, the LSC shows some variability
the till surface was previously eroded prior to SSD but falls within the range of values below.
deposition; the till surface dips below the river level, and The clay typically possesses a water content between
the landslide shear zone is located within an intermediary 30% to 40%; a Plastic Limit from 25 to 30, Liquid Limit
layer of highly plastic clay; however, Site 3 was included as between 50 to 80 with the corresponding range of Plasticity
the strength parameters for that site are similar to the other Index between 30 to 50, with an average closer to 40. The
sites. clay content is relatively high, usually 70% or more, up to
nearly 100 % in some cases.
Failure Geometries As would be expected, when subjected to large shear
A summary of the geometry of each landslide, focusing on strains, the structure of the LSC exhibits typical features of
the cross sections considered for stability analysis, is disturbance. Figure 3 shows the shear zone during shear
provided in Table 1. The dimensions provided are intended key construction at Site 1: the LSC was blocky with
to give a general understanding of the scale of each slide: extensive breccia and abundant polished slickensides.
height from top of head scarp to shear zone; overall length
from head scarp to toe, elevation of the SSD/till contact
with respect to river level, and the depth from the ground
surface at the head scarp to static measured groundwater
level. Illustrations from the back analysis at Sites 1 and 2,
showing typical landslide stratigraphy and geometry, are
shown in Figure 5.
The till surface throughout the area is typically gently
sloping towards the river. The ratio of length to height of the
failures in question is just over 3 on average, up to about 8
at Site 1.
 Retrogression since 1981 and Individual Blocks
Yoshida and Krahn (1985) discussed a potential new
headscarp based on SI measurements and anticipated
ongoing retrogression of the landslide; however, it was
noted that this additional scarp was not yet visible in the
field. Based on the previous landslide length of about 120
m. and considering more recent (2011 and onwards) aerial
imagery of the site (Google Earth), it is apparent that the
landslide has retrogressed approximately 40 m since the
previous assessment. The model geometry was updated
to show a new head scarp located about 40 m from the
original, consistent with recent aerial imagery. A vertical
displacement of 2 m was used to simulate the new
topography. Figure 4, 2 shows the assessment of recent
retrogression.
Previous researchers (Haug (1977), Yoshida and Krahn
(1985) Misfeldt et al (1991) and others) have discussed the
Figure 3 - Typical condition of LSC at the shear zone, Site behaviour of translational retrogressive landslides, where it
1. has been observed that individual blocks or the overall
landslide tend to exhibit similar values for shear strength.
Summary of Shear Strength Parameters Whether considering a single block, or the entire landslide,
Table 2 summarizes the drained shear strength the back calculated shear strengths for either scenario tend
parameters for the LSC obtained either through back to be similar since the exercise is simply an equal up or
analysis or laboratory testing, along with an estimate of the down scaling of the driving and resisting forces to satisfy
residual strength based on correlation with the soil’s limiting equilibrium, in this case controlled by the strength
plasticity index (Kanji (1974) and Bowles (1996)). of the LSC.

5 RE-ANALYSIS OF THE BEAVER CREEK Stability Analysis Results

LANDSLIDE The geometry of the slip surfaces generated using the
The 2D Limit Equilibrium Method (LEM) stability modeling 1981 site conditions varied slightly from those outlined in
by Yoshida and Krahn (1985) was revisited, for the Yoshida and Krahn (1985). Slight adjustments were made
following three main reasons: to calibrate the new model to correlate slip surfaces with a
1. Modern software allows efficient consideration of FoS of 1.00 to those shown by the 1985 studies. This
many thousands of slip surfaces; the intent was to required an adjustment to the piezometric surface, which
explore whether the previously determined was raised by 0.5 m (river elevation was kept constant); φ’
strength parameters or FoS changed significantly was reduced from 8.21 to 8.10 which resulted in a close
when doing so. match between entry and exit locations as well as the
2. The landslide has retrogressed approximately 40 geometry of the slip surfaces and elevation of
m since the early 1980’s. The new geometry was displacements measured in the SIs. The estimated FoS for
analysed to provide an updated FoS and strength the recalibrated model of the overall landslide was 1.00 as
parameters to compare to previous work. shown in Figure 4, 1.
3. The other sites considered in this paper were When considering the most recent retrogression, with a
assessed using modern SLOPE/W software – the new head scarp about 40 m upslope, the FoS appeared to
Beaver Creek landslide was reanalysed with the remain constant with φ’ = 8.1o unchanged (Figure 4, 2).
same software to allow comparison among sites.
Assessment of the lowest landslide block showed a FoS of
Back Analysis 1.00 with φ’ = 6.7o- see Figure 4, 3.
The revised back analysis was completed using Slope/W
software (version 11.2.2.23310) with the Morgenstern- Consideration was also given to a version of the model
Price method and half-sine interslice side force function. assuming a 2 m thick layer of residual (φ’ = 8.1o ) highly
Model geometry, piezometric conditions and initial strength plastic clay along the till surface, with the remainder of the
parameters were adopted from Yoshida and Krahn (1985). SSD assigned φ’ in the range of 13 degrees, similar to the
The entry and exit method of defining slip surface geometry assessment of the new head scarp by Yoshida and Krahn
was implemented to match the shear surfaces in the (1985) which considered different strengths between the
previous models, while also considering additional slip primary failure plane and the newly developed shear
surfaces that shared similar entry and exit locations and surface corresponding with the newest scarp. Results
correlated with shear elevations shown in the slope between these two versions of the model were similar, so
inclinometers. the figures included in this paper are from the original
revision only.
The updated calibrated back analysis is shown in Figure 4,
1.
 1

Figure 4 – 1: Back analysis. 2: Newest retrogression. 3: Assessment of lowest block.

Table 1 - Landslide Geometries

Location Height - Scarp to Length Ratio length/ SSD/Till Contact Water Table
shear zone (m) Height (height from river level) Depth from
(m) (m) top of Scarp
(m)

Beaver Creek 36 120 3.3 -14 9

Site 1 8.5 70 8.2 7.5 1
Site 2 15.5 40 2.6 6.5 12

Site 3 11 40 3.6 9 10
Site 4 7.5 20 2.7 -9 7
Site 5 13 50 3.9 12 4
Site 6 11 36 3.3 10 3
Table 2 - Summary of Shear Strength Parameters

Location Shear Zone Strength Method Correlations

(Kanji (1974) Bowles (1996)

’r, o c’, kPa

Beaver Creek ’81 to 6 to 8 0 2D LEM Back analysis
‘85 6.6 to 7.2 0 Direct Shear Test on Reconstituted Slurry
10.6 0 Direct Shear Test on Tube Sample
Beaver Creek ‘23 8.1 0 2D LEM Reassessment 9.3 8.7
Site 1 12 4.8 2D LEM Back Analysis
Site 2 13 5 2D LEM Back Analysis 8.2 7.6

Site 3 12 0 2D LEM Back Analysis 9.2 8.5

11.4 0 Direct Shear Test on Tube Sample
Site 4 11 5 2D LEM Back Analysis
Site 5 11 0 2D LEM Back Analysis 9.0 8.3
11.2 to 11.4 0 Direct Shear Test on Tube Sample
Site 6 12 0 2D LEM Back Analysis 9.0 8.3
11.3 8 Direct Shear Test on Tube Sample

Figure 5 – Examples of landslide stratigraphy and geometry

 Discussion of the new model Slope Inclinometers (SIs) were sheared off indicating at
Previous analyses showed values of φ’ from about 6o to least 50 mm to 70 mm of displacement at the shear zone.
8.5o were required to achieve a FoS of unity for individual Using a typical SI probe wheel spacing of 500 mm, this
slide blocks, or the overall landslide. Results from the would correlate to about 10% or greater strain, suggesting
updated model fell within this range. that the LSC has been sheared to its residual state for at
Values of φ’ obtained in the new SLOPE/W models least a portion of the shear zone at all sites except Site 4.
varied up to about 1.5% from those observed during
previous assessments, when considering specific slip It is recognized that the progressive nature of these
surfaces; this difference is attributed to the slight variation landslides inherently involves a variation in mobilized shear
in the LEM methods employed by the two models. strength across the slip surface. As such, the 2D LEM
The main differences between the early assessments method remains an approximation of average LSC strength
and the modeling completed for this paper are the number across a shear plane. Further analysis using nonlinear
of slip surfaces considered, and the variation between inter displacement methods along with more accurate
slice side force functions. SLOPE – II, the earlier version of measurements of strain would provide a more thorough
SLOPE/W, used during the 1980’s work, used a constant understanding of the behaviour of the LSC at varying states
side force function. While slight, the difference can result in of strain across the shear zones.
a variation between FoS compared to using the half – sine
function: a comparison between methods by Fredlund and The correlations to estimate φ’r suggested by Kanji (1974)
Krahn (1977) show a difference in FoS up to about 3% and Bowles (1996) fall within the available range of
depending on site geometry and piezometric conditions. laboratory test results for residual shear strength, but below
Given these slight differences between the models, the the back calculated strengths. For “new” landslides with
new calibrated back analysis and subsequent revised low to moderate displacement, use of these relationships
stability assessments were deemed reasonable. for back analysis or remedial design in Saskatoon will likely
result in over-conservative estimates of FoS or inaccurate
During assessment of the most recent retrogression, it estimates of slip surface geometry.
was not possible to match the shear elevation indicated in
the 1980’s SI’s (SI10); it is assumed that the movement 7 CONCLUSIONS
measured in SI10 was intermediate between the old and Seven landslides in the Saskatoon area with varying
new head scarps, and that additional instrumentation geometry and piezometric conditions, with shear planes
would be necessary to confirm the slip surface geometry originating in the LSC, were reviewed. With the exception
associated with the newest scarp. of the Beaver Creek landslide, the back-calculated and
laboratory measured shear strength of the LSC fell within
6 LSC DESIGN PARAMETERS FOR LANDSLIDES IN a relatively narrow range of about φ’ = 11o to 13o, and c’ =
SASKATOON 0 to 8 kPa. The Beaver Creek landslide was reassessed
with an updated 2D LEM stability model to investigate
The large displacements at Beaver Creek (3 m or more potential changes to LSC strength due to more recent
near the toe) appear to have sheared the LSC to its retrogression since the original investigations in the late
residual state along the majority of the shear plane, 1970’s to early 1980’s. Current estimates of shear strength
represented by values of φ’ = 6o to 8.2o and c’=0 kPa. This agreed with the previous work with back calculated
range correlates well with the residual strengths obtained strength in the range of φ’ = 6.7 o to 8.1o, and c’ = 0 kPa.
from testing of slurried samples by Yoshida and Krahn
(1985). With the Beaver Creek landslide having experienced
significantly greater strains than the others, the difference
The back calculated shear strengths at the other sites, in observed strengths is attributed to a greater portion of
which were assessed relatively soon after initial failure, the Beaver Creek landslide shear plane having been
was in the range of φ’ = 11o to 13 o, with c’ = 0 kPa in most sheared to its residual strength. At the other sites, the back-
cases but as high as 5 kPa occasionally. Direct shear tests calculated strengths from these six sitesappear to provide
on Shelby tube samples showed values within this range, a reasonable estimate of average strength along the shear
suggesting that the direct shear test may not be capable of plane.
shearing the LSC to its full residual state, but does provide
values appropriate for 2D LEM analysis of these failures. The impact to infrastructure due to landslides in Saskatoon
This strength range has been proven as a good can be severe, sometimes requiring emergent response
representation of the average shear strength along the without appropriate time for detailed field and laboratory
shear plane for “new” failures where large scale investigations. The above noted initial strength parameters
retrogression was not realized. These values have thus will provide some insight into LSC based landslide failures
shown to provide a reasonable understanding of the in Saskatoon and provide some initial guidance for back
behaviour of these landslides, and have been used for analysis, model calibration and remedial design. Armed
successful remediation at Sites 1, 2, 5 and 6 and new with this base knowledge of LSC strength, an initial
development at Site 4. scoping analysis of remedial options may be developed
soon after observed landslide movement.
No attempt has been made to correlate displacements with
apparent mobilized strengths. At all sites except Site 4,
8 ACKNOWLEDGEMENTS
The authors wish to thank the City of Saskatoon and the
other clients and landowners we have had the pleasure of
working with, both for the opportunity to assist on their
projects and for sharing of their project data and granting
permission to present this paper.

9 REFERENCES

Christiansen, E. A. (1968). Pleistocene stratigraphy

of the Saskatoon area, Saskatchewan, Canada.
Canadian Journal of Earth Sciences, 5(5), 1167-
1173.

Christiansen, E. A. (1970). Physical Environment of

Saskatoon, Canada.

Christiansen, E. A. (1979). The Wisconsinan

deglaciation, of southern Saskatchewan and
adjacent areas. Canadian Journal of Earth
Sciences, 16(4), 913-938.

Christiansen, E. A. (1992). Pleistocene stratigraphy

of the Saskatoon area, Saskatchewan, Canada:
an update. Canadian Journal of Earth
Sciences, 29(8), 1767-1778.

Clifton, A. W., Krahn, J., & Fredlund, D. G. (1981).

Riverbank instability and development control in
Saskatoon. Canadian Geotechnical Journal, 18(1),
95-105.

Elwood, D. and Heilman, V. (October 2023).

Saskatoon River Valley Geohazards – An Update
[Paper presentation]. Proc. GeoSaskatoon 2023:
Bridging Infrastructure and Resources. 76th Annual
Conference of the Canadian Geotechnical Society.

Fredlund, D. G., & Krahn, J. (1977). Comparison of

slope stability methods of analysis. Canadian
geotechnical journal, 14(3), 429-439.

Kanji, M. A. (1974). The relationship between drained

friction angles and Atterberg limits of natural
soils. Géotechnique, 24(4), 671-674.

Misfeldt, G. A., Sauer, E. K., & Christiansen, E. A.

(1991). The Hepburn landslide: an interactive slope-
stability and seepage analysis. Canadian
Geotechnical Journal, 28(4), 556-573.

Yoshida, R. T., & Krahn, J. (1985). Movement and

stability analysis of the Beaver Creek landslide,
Saskatchewan, Canada. Canadian Geotechnical
Journal, 22(3), 277-285.
Historical Background and GeoEngineering
Assessment of the Military Underground
Complex at Canadian Forces Base North
Bay
Vlachopoulos, N. and Moore, K.S.
Royal Military College of Canada, Kingston, Ontario, Canada

ABSTRACT
Canadian Forces Base (CFB) North Bay is a Royal Canadian Air Force base that is located in North-Eastern Ontario. CFB
North Bay was the site selected for the construction of an underground command and control center for NORAD. The
construction of the underground complex (UGC) took 3 years between 1959 and 1962. Excavations took place within
primarily gneiss rock which also featured intrusions of Kimberlite dykes. The facility remained in service until 2006. The
UGC features two main facilities and two access tunnels. The approximate depth of the UGC is 180 m below the surface.
The main operations section consists of a large cavern which houses a three-story building. Two vehicular access tunnels
make their way to the main operations centre and a utility cavern. This paper outlines the historical and current site
assessments of the ground conditions associated with the access tunnels of the UGC.

RÉSUMÉ
La Base des Forces canadiennes (BFC) North Bay est une base de l'Aviation royale canadienne située dans le nord-est
de l'Ontario. La BFC North Bay a été le site choisi pour la construction d'un centre de commandement et de contrôle
souterrain pour le NORAD. La construction du complexe souterrain (CST) a duré 3 ans entre 1959 et 1962. Les
excavations ont eu lieu principalement dans la roche de gneiss qui comportait également des intrusions de dykes de
kimberlite. L'installation est restée en service jusqu'en 2006. Le CST comprend deux installations principales et deux
tunnels d'accès. La profondeur approximative du CST est de 180 m sous la surface. La section principale des opérations
se compose d'une grande caverne qui abrite un bâtiment de trois étages. Deux tunnels d'accès pour les véhicules se
dirigent vers le centre d'opérations principal et une caverne utilitaire. Cet article décrit les évaluations de site historiques
et actuelles des conditions des roches associées aux tunnels d'accès du CST.

1 BACKGROUND OF CFB NORTH BAY The base was commissioned on September 1st, 1951
and its primary purpose was the provision of North
Canadian Forces Base (CFB) North Bay is a Royal American air defense during the Cold War. In 1957,
Canadian Air Force (RCAF) base, 22 Wing, that is located Canada and the United States of America decided to form
in North-Eastern Ontario as seen in Figure 1. The base a bi-national command, North American Aerospace
houses an underground complex (UGC). This site was Defense Command (NORAD) for the purpose of air
selected for the UGC due to existing military, transport- defense against the Soviet Union. As such, North Bay was
ation and telecommunication infrastructure in place as well the site selected for the construction of an underground
as the human and natural geography. command and control center for NORAD. Today it is also
known as Canadian Air Defence Sector (CADS) HQ and
the base has a critical role in North American air
sovereignty. The base is very well protected and guarded
24/7.

2 UNDERGROUND COMPLEX FACILITY

The construction of the underground complex began in

1959. Construction was completed three years later in
1962 and the facility opened in 1963. The facility remained
in service until 2006. The UGC features two main facilities
and two access tunnels. The approximate depth of the
UGC is 180 m below the surface. The UGC was designed
to withstand a nuclear bomb explosion in order to continue
operations.
As seen in Figures 2 and 3, the main operations section
Figure 1. Location of Canadian Forces Base (CFB) North consists of a large cavern measuring 131 m x 70.1 m x 16.5
Bay m which houses a three-story building. Two vehicular
access tunnels make their way to the main operations
center and a utility cavern with an excavation size of 122.2
m x 15.2 m x 8.2 m. These access tunnels are named the over 2.5 billion years. The rock formations in the area have
North Tunnel and the South Tunnel due to their orientation. been strongly metamorphosed, folded and subsequently
The North Tunnel is approximately 2 km long and the South intruded by igneous rocks. (Gartner, J.F., 1980). North
Tunnel is approximately 960m long. This is the focus of the Bay is located within a granitic gneiss rock formation within
investigation of the paper. i.e., the access tunnels and not the straight of the Central Gneiss Belt at the west end of
the utility cavern (DCC, 2012). In 2005, the UGC, including the Grenville Geological Province. Within this region, the
the tunnels, were classified as a Federal Heritage building. eldest and also most abundant rocks are metasediments
Since being decommissioned in 2006, the facility requires derived primarily from siliceous siltstones and sandstones.
regular monitoring, assessments, and maintenance; These metasediments consist of a variation of gneisses
something that the Department of National Defence (DND) that can be found throughout most of the area.
is being very proactive on. The heterogeneous gneisses of the area fall into two
categories: Archean and Lower Proterozoic migmatic
gneisses that were likely reworked units of the Southern
and Superior provinces and Middle Proterozoic meta-
sedimentary gneiss. “These rocks were intruded by 1.7 and
1.45 billion-year-old granitic plutonic rocks, both of which
are less deformed than the host rocks. Postdating this
intrusive event, the region underwent high-grade
metamorphism, experiencing temperatures of 650°C–
750°C and pressures of 8.0–8.5 kilobars”. (Kusky, 2010).
In terms of the surficial geology, the last, most recent
glaciation period had a profound influence. Glacial ice
advanced toward the southwest depositing ground
moraine till over the bedrock. This till is characterized by a
grey to red colour, a lack of carbonates, and a low
percentage of fine-grained material (i.e., silt size or
smaller). (Gartner, J.F., 1980).

3.2 Site-Specific Geology at the UGC Site

Figure 2. Schematic of NORAD Underground Complex in In terms of the geology associated with the UGC site itself,
North Bay (NORAD Complex, CFB North Bay, 2023). the excavations took place within primarily gneiss rock
which also featured intrusions of Kimberlite dykes which
are primarily oriented east-west. The Kimberlite dykes
occur in a variety of mineral compositions and widths. As
well, a network of regular joints or discontinuities exists in
the rock that are oriented primarily horizontally and
vertically. Due to the orientation of the joints, and the
foliation and dip of the rock, there is a concern for
delamination or the process of layers of relaxation and
unravelling which may lead to instability.

4 GEOLOGICAL ENGINEERING ASSESSMENT

An engineering geological assessment commissioned by

Natural Resources Canada was conducted in 2009. The
focus was on the existing ground and support conditions of
the access tunnels of the UGC with a view to
recommending short-term and long-term ground support
considerations for these tunnels. Their assessment
according to the NGI diagram of tunnel support is seen in
Figure 3, Large Blast Door of Bunker at UGC (Underground Figure 4 which takes into account the rock mass quality
NORAD complex, 2023). and recommended rock support solutions.
In terms of rock strength, typical values published in the
3 GEOLOGY literature for granitic gneiss were used within the NRCan
study (Doucette, C. and Labrie, D., 2009) and can be seen
3.1 General Geology of the North Bay Area in Table 1. As well, an assessment of the principal
stresses, their magnitude and orientation were estimated
The geology of the North Bay Region is characterized by from measurements done in the Canadian Shield and
Precambrian rocks of the Grenville Structural Province. published in literature (Arjang 2001). These values can be
The rocks have a long complex geological history covering seen in Table 2.
 load to the bolts. In terms of results from the standard
testing, the bolts tested were all able to support at least
90% of their elastic limit (approximately 90 kN) and no
slippage was detected. Of these tests, the maximum
applied load was 107 kN which had a displacement of
approximately 3 mm (Bahrani and Maloney, 2013). An
example of one of the pullout test results can be seen
below in Figure 5.

Figure 4. Diagram for tunnel support (NGI, 1997) ground

conditions and size of caverns observed at 22 Wing North
Bay Underground Complex are highlighted with the Green
box recreated after (Doucette, C. and Labrie, D., 2009). Figure 5. Example of pullout testing results for rock bolts
within the UGC. (after, Bahrani and Maloney, 2013).
Table 1. Typical values used for intact rock properties for
the North Bay 22 Wing tunnels (Doucette, C. and Labrie, 5 CONSTRUCTION OF THE UGC
D., 2009).
The underground tunnelling and cavern excavations were
mined using the well-established method of sequential drill-
and-blast. The blasting that was utilized was “smooth-
blasting”. This blasting technique that was new at the time
of the excavation ensured that the damage to the resulting
excavation peripheries (i.e., walls and roof) was limited.
The number of drill hole traces left on tunnel and chamber
walls is evidence of the high degree of precision achieved
by the excavation technique that was employed. This was
Table 2. Estimates of in situ stresses for the North Bay 22 also observed with the latest site visit conducted by the
Wing tunnels (Arjang 2001). authors in 2022 (Figure 6).

In 2012 and 2013, maintenance and repair construction

was conducted in order to address selected topics that
were discussed in the CANMET-MSSL report (Doucette, C.
and Labrie, D., 2009). Mostly, it included the requirement
for bolting and screening for more than half of the North
Tunnel (DCC, 2012). In addition to bolting and screening,
sprayed shotcrete cover was recommended at certain
locations. To deal with water intrusion, weeping holes were
placed in the shoulder of the drift and the bottom of the Figure 6. Precision blasting to achieve “clean” excavation
Kimberlite Dyke (Taylor, 2013). The work was done by profile with minimal damage to tunnel boundaries.
Technica and concluded in February of 2013. Non-
destructive pull testing was done by Mirarco. Testing was The primary support that was utilized within the excavation
conducted on 20 resin-grouted rebar rock bolts using a was the use of grouted rock bolts and heavy wire mesh at
hollow, cylindrical, hydraulic ram in order to apply the axial certain locations (as seen in Figure 7). The rock bolts were
tensioned at the plate and were intended as a permanent
reinforcement solution vice a temporary support measure. were also selected discontinuities at random orientations)
These measures were introduced in order to increase the at particular locations that yielded smaller blocks.
overall stability of the tunnel and cavern excavations with a
view to meeting the high-quality criteria that was specified
for this unique underground facility. Concrete liners and
arches were also introduced at locations where seepage
was extreme and pumped grout injection was not sufficient.
The same criteria of stability and safety were used for both
tunnel portals and other service openings with access to
the surface (Doucette, C. and Labrie, D., 2009).

Rock Bolts

Figure 8. Granitic gneiss host rock outcrop at UGC site.

It was observed that Kimberlite dykes of various

thicknesses were embedded within host rock (Figure 9).
The dyke formations we oriented mostly horizontally with
respect to the orientation of the tunnels. These dykes were
oriented east-west and intruded the gneiss rock mass at
regular intervals.

Figure 7. Kimberlite dyke reinforced with rock bolts and

Screen Mesh.

6 RECENT, INITIAL GEOLOGICAL ENGINEERING

Kimberlite
OBSERVATIONS AND ASSESSMENT

A site visit was conducted from 17 - 18 November 2022 by

members of the RMC Green Team, Canadian Military
GeoWorks Lab in order to observe and assess the current
state of the two access tunnels of the UGC. This section
summarizes the main observations from the site visit that
was conducted. There is information that has not been
included due to the sensitivity associated with military and Gneiss
strategic nature of such installations. As well, this site visit (Host Rock)
was an initial reconnaissance in order to familiarize the
team members with the site and to prepare for more
rigorous site assessments and the inclusion of potential
monitoring techniques in the near future. It is proposed that Figure 9. Kimberlite feature embedded within host rock.
the distributed optical technique that has been developed
by the RMC Green Team be implemented within the UGC As with previous assessments, it was observed that there
as a long-term monitoring technique. These can be were three primary sets of discontinuity orientations
employed within grouted rock bolts as stipulated by throughout the length of the North and South tunnels. One
Vlachopoulos et. al, 2017. of discontinuity set orientations is primarily horizontal in the
same alignment with the tunnels. This orientation is also
6.1 Geology parallel to the foliation of the granitic gneiss parent rock.
The other two major joint sets are vertical, one oriented in
The main rockmass that the UGC was excavated in was the northwest direction and the other one in the northeast
granitic gneiss. This host rock (Figure 8) has a network of direction. There are also other, fewer discontinuities that
discontinuities and joints that form large blocks with relation were observed that were in other orientations other than
to the diameter of the tunnels. The size of these larger these 3 primary sets. The authors are in agreement and
blocks increases the stability of the overall underground confirm the findings of CAMNET in terms of the Q (NGI,
structure. That being said, there was also a network of 1997 designation as seen in Figure 4. The coherent and to
joints and discontinuities (mostly horizontal and vertical to a large degree self-supporting host rock mass of the UGC
the longitudinal direction of the tunnel alignment. There
 Rock Bolts

Joint Set
Orientations

North
Portal

Figure 10. North Portal walls; Evidence of gneissic banding, discontinuities, and rock bolt support.

Steel Mesh

Rock
Kimberlite Bolts
Rock (Roof)
Bolts

Delamination
Potential

Arch
Support

Figure 11. Fracture feature, netting and evidence of Figure 12. South Portal. Arched Support.
delamination.

Figure 13. South Portal Entrance to UGC. Granitic Gneiss, host rock outcrops evident on either side of the tunnel opening.
 were observed to have been used at several locations.
These support measures were employed in order to
reinforce the tunnels at intersections with dykes, primarily
to try and limit the amount of water seepage at these
locations.

Figure 14. Evidence of minor water intrusion.

Figure 15. Rock bolts and steel mesh support.
can be classified as a rock mass of fair quality with a Q
index between 5 and 10. The joint sets were persistent,
long and consistent in their arrangement throughout the
tunnels. The blocks defined by the joint sets are large
blocks of gneiss that are many cubic meters in terms of
volume. Overall, a very good level of stability was
observed within the North and Southern tunnels. Selected
photos (Figures 10-14) depict the main geological features
observed within the access tunnels as well as at the tunnel Rock
Steel Mesh
portals. Bolts
Rock
6.2 Rock Support Measures Bolts

The North and South access tunnels of the UGC have

width to height dimensions of 3.7 m x 4.0 m and 4.9 m x
4.2 m respectively. In order to ensure such tunnel design
tolerances a variety of support measures were employed
within the tunnels. Rock support in terms of rock bolts, steel
meshes (Figures 15 and 16) and reinforced concrete
support were observed within the UGC access tunnels.
These support measures are conventional based on the
rock mass assessment and the NGI Q designation that was
determined. The bolting pattern that is observed is 1.5 m
x 1.5 m, with resin grouted steel rebar rock bolts. In other
locations spot bolting and bolting patterns of 1.8 m x 1.8 m
were also observed. Historical documentation (CAMNET)
denotes lengths of 2.4 m long for the rock bolt. In terms of
the steel mesh, a conventional No. 9 mesh has been used
for screening. Figure 16. North Tunnel. Rock bolts and steel mesh
In selected locations within the tunnels, grout injection support – roof and sidewalls.
and concrete encasement in the form of liners and arches
7 SUMMARY AND WAY FORWARD NORAD Complex, CFB North Bay,
https://www.facebook.com/NORADComplex/?paipv=0&
Overall, the Department of National Defence has been eav=AfbD9yGI8Sz8fAaxkQ_SjUcaoWQHrUtJXpuc4yjts
proactive in keeping this facility serviceable and safe until uuLbOIGXnBWTbmyEJGztRYDeCI&_rdr, Accessed
its eventual decommissioning. The condition of the North May 2023.
and South tunnels were seen to be in good, overall repair.
The granitic gneiss host rock provides much of the stability NGI, 1997. Engineering Geology -Practical use of the Q-
itself in terms of the structural integrity of the UGC. The method. Report 592046-4, Norwegian Geotechnical
support measures that have been implemented within the Institute (NGI), Oslo (Norway): 43 p.
facility will have to be monitored, assessed and maintained
on an ongoing basis in order to ensure their serviceability. Taylor, S. 2013. Final Site Inspection - Deb 18, 2013.
With this in mind, the authors will be visiting the UGC in the (March): 1–12.Underground NORAD Complex,
near future in order to conduct a more comprehensive site https://noradcomplex.ca/2021/08/29/norad-complex-put-
visit and assessment of the UGC. In terms of long-term on-endangered-places-list/ , accessed May 2023.
monitoring, it is proposed that a structural health monitoring
approach using instrumented grouted rock bolts outfitted Vlachopoulos, N., Cruz, D., and Forbes, B. 2017. Utilizing
with fiber optics (i.e., a distributed fiber optic sensing a Novel Fiber Optic Technology to Capture the Axial
technique that is able to determine the strain on a rock bolt Response of Fully Grouted Rock Bolts. Journal of Rock
at a spatial resolution of every 0.65 mm) be installed for this Mechanics and Geotechnical Engineering. Submitted 26
purpose as per Vlachopoulos et al., 2017. July 2017, JRMGE-2017-225, V10 N3 2018.

8 ACKNOWLEDGEMENTS

The author would like to acknowledge the support of the

Natural Sciences and Research Council of Canada
(NSERC) as well as the provision of associated material
and site access by CFB North Bay personnel as well as the
office of the Assistant Deputy Minister (Infrastructure and
Environment) within the Department of National Defence.
The authors would also like to thank current graduate
student, MASc, Captain Chesley Guo, for her participation
in the site visit that was conducted in 2022.

9 REFERENCES

Arjang, B., 2001. Database on Canadian in situ ground

stresses. CANMET-MMSL Division Report 01-029 (TR),
CANMET Mining and Mineral Sciences Laboratories,
Natural Resources Canada, Ottawa, Ontario: 33 p.

Bahrani, N. & Maloney, S. (2013). Non-destructive Pull

Testing on Resin-Grouted Rebar Rockbolts. MIRARCO
Geomechanics Research Centre.

DCC (Defence Construction Canada). 2012. Rock Bolting

in the North Tunnel of the Underground Complex at 22
Wg - Environmental Assessment Screening-Level
Report. Hornell Heights.

Doucette, C. and Labrie, D., 2009. 22 Wing North Bay

Tunnels Assessment. Investigation, Survey and Results.
CANMET-MSSL report 08-057(CR), Annex A.

Gartner, J.F., 1980, Engineering Geology Terrain Study

101, North Bay Area, Geological Survey of Ontario,
Northern Ontario, Ministry of Natural Resources, 1980.

Kustky, T.M. Encyclopedia of Earth and Space Science,

Library of Congress Cataloging-Publication Data. New
York, New, York. 2010.
S042 Rock Slope: Use of remote sensing
with rock slope rating systems for
Geotechnical Asset Management
Taylor Wollenberg-Barron, Renato Macciotta
University of Alberta, Edmonton, Alberta, Canada
Chris Gräpel
Klohn Crippen Berger, Edmonton, Alberta, Canada
Kristen Tappenden, Roger Skirrow
Alberta Transportation and Economic Corridors, Edmonton, Alberta, Canada

ABSTRACT
Alberta Transportation and Economic Corridors (TEC) is currently working towards the development of a formalized
geotechnical asset management (GAM) program and can benefit from research conducted to link the results of rockfall
geohazard condition assessment tools with performance monitoring techniques via remote sensing and the subsequent
change detection analyses. Integrating the use of remote sensing technologies with condition assessment tools may
provide transportation agencies with an additional monitoring method to aid in prioritizing capital expenditure on geohazard
sites. This paper presents Light Detection and Ranging (LiDAR) change detection results for a rockfall geohazard site
located near Canmore, Alberta, along with the results of five different rock slope and rock mass rating systems applied to
the slope. The results presented in this paper provide a preliminary assessment of how established rock mass
characterization tools can incorporate, and benefit, from performance monitoring through remote sensing. The findings will
enhance TEC’s GAM progress by standardizing the integration of initial condition assessment tools with performance
monitoring for rock slopes through remote sensing in a practicable and repeatable manner.

RÉSUMÉ
Alberta Transportation and Economic Corridors (TEC) travaille actuellement à l'élaboration d'un programme formalisé de
gestion des actifs géotechniques (GAM) et peut bénéficier des recherches menées pour lier les résultats des outils
d'évaluation de l'état des géorisques liés aux chutes de blocs aux techniques de surveillance des performances par
télédétection et les résultats ultérieurs analyses de détection de changement. L'intégration de l'utilisation des technologies
de télédétection aux outils d'évaluation de l'état peut fournir aux agences de transport une méthode de surveillance
supplémentaire pour aider à hiérarchiser les dépenses en capital sur les sites géorisques. Cet article présente les résultats
de la détection des changements par détection et télémétrie par la lumière (LiDAR) pour un site géorisque de chutes de
pierres situé près de Canmore, en Alberta, ainsi que les résultats de cinq différents systèmes d'évaluation des pentes
rocheuses et des masses rocheuses appliqués à la pente. Les résultats présentés dans cet article fournissent une
évaluation préliminaire de la façon dont les outils établis de caractérisation des massifs rocheux peuvent intégrer et
bénéficier de la surveillance des performances par télédétection. Les résultats amélioreront les progrès du GAM de TEC
en normalisant l'intégration des outils d'évaluation de l'état initial avec la surveillance des performances des pentes
rocheuses par télédétection d'une manière pratique et reproductible.

1 INTRODUCTION with an additional monitoring method to aid in prioritizing

capital expenditure on geohazard sites.
Alberta Transportation and Economic Corridors’ (TEC) The S042 Spray Lakes Rockfall site is an existing
Geohazard Risk Management Program (GRMP) includes rockfall geohazard within the GRMP where a LiDAR
approximately 250 actively monitored geohazard sites remote sensing and change detection analysis program
along the provincial highway network (31,000 kilometres of was started in 2018. The change detection analysis
roadway infrastructure). The management of the risk conducted on the S042 site is presented in this paper. Five
presented by these geohazard sites requires monitoring rock mass and rock slope rating systems were previously
through robust and cost-effective strategies to maintain a applied to the S042 site, along with other sites, to compare
safe and efficient highway network. In recent years, remote the applicability of each for use as condition assessment
sensing techniques have allowed an improved tools as part of a geotechnical asset management (GAM)
understanding of slope deformation processes and are program (Wollenberg-Barron et al. 2023). In addition to
now part of the state of practice in slope risk management. Alberta’s GRMP Risk Level (RL) rating for the S042 site,
One such technique is the use of a series of remote the additional rock slope and rock mass rating systems
sensing surveys using Light Detection and Ranging applied include: the Rockfall Hazard Rating System
(LiDAR) followed by change detection analysis. Integrating (RHRS) (Pierson 1992), the modified Colorado Rockfall
the use of remote sensing technologies with condition Hazard Rating System (CRHRS) (Russel et al. 2008), the
assessment tools may provide transportation agencies Q-Slope system (Barton and Bar 2015), the Geological
Strength Index (GSI) (Hoek 1994), and the Rock Mass detection methods, before measuring the change, point
Rating (RMR) system (Bieniawski 1993). clouds must first be aligned considering topographic
TEC is currently working towards the development of a changes that may have occurred over time.
formalized GAM program to enhance their ability to monitor Several methods of change detection are available,
the condition and deterioration of geotechnical assets, including DEM of difference (DoD), direct cloud-to-cloud
including rock slopes. This requires the assessment of comparison (C2C), cloud-to-mesh or cloud-to-model
available tools to effectively collect and manage data for distance (C2M), and multi-scale model-to-model cloud
evidence-based decision making (Tappenden and Skirrow comparison (M3C2) (Lague et al. 2013, DiFrancesco et al.
2020). The analysis presented in this paper was conducted 2020). The M3C2 change detection method was developed
to continue the enhancement TEC’s GAM program by by Lague et al. (2013) and works by first sub-sampling core
standardizing the integration of initial condition assessment points from the reference cloud. Then, normal vectors are
tools with performance monitoring for rock slopes through calculated for points which fall within a specified diameter
remote sensing in a practicable and repeatable manner. around each core point. Finally, an average distance
between the two clouds is calculated along the previously
1.1 Geotechnical Asset Management determined core point vectors, within a specified cylindrical
diameter. Some limitations arise with complex topography
GAM emerged from Transportation Asset Management and surface roughness where the distance can be
(TAM) due to the impact and interrelated performance of overestimated due to normal misorientation (Lague et al.
geotechnical assets on transportation infrastructure. The 2013). M3C2 has become a widely used method for
basis for a GAM program is generally defined by its change detection in a variety of fields, including rock slopes
governing agency’s strategic goals, including but not (Macciotta and Martin 2019, Rodriguez et al. 2019,
limited to the management of risk within a transportation DiFrancesco et al. 2020, Deane et al. 2020, Macciotta and
system. An assessment of each asset’s condition within an Hendry 2021).
asset inventory is necessary in determining how the
agency’s budget and resources will be used to the greatest 2 S042 SITE DESCRIPTION
benefit. Thorough documentation is essential to a
comprehensive management program; each asset The S042 site is located on Highway 742 (Spray Lakes
identified within a target area must be cataloged with its Road) approximately 5 km southwest of Canmore, Alberta.
physical characteristics but also some measure of its cost, The weighted annual average daily traffic (WAADT) at this
performance, and impact of failure (Bernhardt et al. 2003). location is approximately 1420 vpd (vehicles per day) (TEC
The development of an asset inventory is a continuous 2022). The average daily air temperature recorded near
process; following the initial assessment and identification S042 ranges from lows of -14.1 °C in the winter to highs of
of an unstable asset, continued monitoring is required to 20.7 °C in the summer. The average annual precipitation
maintain adequate allocation of resources. This gives rise of combined rainfall and snow near the S042 site is 384
to the need for robust analytical tools capable of measuring mm (ACIS 2023). The S042 site is part of the Mount Rundle
the performance of a geotechnical asset, forecasting future range which comprises Paleozoic limestones, dolomitic
condition, and advocating for the needed capital limestones, dolostones, and shales (Macciotta et al. 2020).
investment. The slope extends approximately 310 m along Highway
742 and reaches a maximum height of approximately 155
1.2 LiDAR m, above highway level.
A call-out inspection was performed at the S042 site in
LiDAR is a range-based imaging tool that uses the 2013 following a relatively large rockfall event with an
reflection of light to determine a target’s location (Lato estimated volume of 15 m3, which originated from the north
2010). The quality of LiDAR data depends on two major slope and narrowly missed a vehicle and group of people
attributes: density and accuracy (Lato 2010). Generally, travelling along Highway 742 (AMEC 2015). Since then,
these attributes are controlled by the equipment, skill of the the S042 site has been included in TEC’s GRMP annual
operator, and data processing experience and technique Southern Region geohazards inspection tour. Two risk
utilized. Point density is controlled by the equipment used level (RL) ratings were assigned to the S042 site to account
and highly dependent on the distance from the target. for small volume events (frequent), and large volume
Additionally, for static LiDAR platforms, surface roughness events (infrequent) to distinguish isolated rockfalls from
may result in occlusion patterns (i.e., missing data) block detachment and larger wedge failures. Large rockfall
depending on the viewpoint of the device (Lague et al. block volumes (denoted with L) were defined by KCB
2013) For this research a ground-based LiDAR device (2019) as greater than 15 m3 while small volumes (denoted
(Optech ILRIS-LR) was utilized. The Optech ILRIS-LR has with S) are defined for this analysis as less than 0.3 m3
a maximum range of approximately 3,000 m (under ideal following the observations and analysis presented in
conditions) and an accuracy of ± 7 mm. Macciotta et al. (2020) for the S042 site.
Due to the two opposing slope aspects at the S042 site,
1.3 Change Detection separate LiDAR viewpoints were required to capture the
entire slope. One for the north slope, with an eastern slope
Change detection is a process of detecting the variation aspect (S042-North), and another for the south slope, with
between two temporally independent topographic models
(Deane et al. 2020). In the case of LiDAR-based change-
 Figure 1. (a) Location of site relative to map of Alberta (Google Earth 2022). (b) S042-
South from LiDAR viewpoint. (c) S042-North from LiDAR viewpoint.

a southern slope aspect (S042-South). The rock mass cloud was measured in CloudCompare using the Compute
rating systems were individually applied to each slope Geometric Features tool. The resulting cloud densities
aspect which resulted in 4 overall ratings (S042-North-S, ranged from approximately 300 to 1400 points per m2
S042-North-L, S042-South-S, and S042-South-L). The which corresponds to a point spacing between 6.7 mm and
location of the S042 site along with photos of the two slope 1.5 mm, respectively. CloudCompare’s Statistical Outlier
aspects from their LiDAR viewpoints is presented in Figure Removal (SOR) filter may be utilized to remove outlier
1. Both slopes comprise a loose talus slope below the points and reduce cloud noise; however, the SOR filter was
intact rock slope which extends to the edge of the gravel found to exacerbate occlusion patterns present within the
highway (with no ditch). S042-South site point clouds due to the high degree of
surface roughness.
2.1 Change Detection Methodology Before change detection can be performed, the first
(temporally) recorded cloud for each slope was aligned
Following the collection of LiDAR data with the Optech with each subsequently recorded cloud using the Align
ILRIS-LR, the raw data was parsed using OPTECH’s (point pairs picking) tool, for rough alignment, and the Fine
parser software which processes the raw scanner data into Registration (ICP) tool for fine alignment. Prior to final
an output file that is compatible with the change detection alignment, any vegetation obscuring the slope was
software being used. Once parsed, the point cloud data removed so that it did not negatively impact the resulting
was loaded into CloudCompare V2.12 (CloudCompare alignment registration error. The M3C2 plugin was utilized
2022) software which was utilized for all the subsequent to generate change detection clouds for each pair of
analysis steps. Each point cloud was first reviewed in detail reference and aligned clouds. Selection of M3C2
to confirm adequate point density, cloud completeness, parameters was conducted via an iterative process as they
and presence of any abnormalities. The density of each
have a significant impact on the surface roughness and The change detection point clouds for the 2018 to 2022
slope detail of the resulting M3C2 point cloud. period are presented in Figure 2(a) and Figure 2(b) for
The limit of detection (LoD) for each analysis was S042-North and S042-South, respectively. Overall, the
estimated via two methods. The first was set as two volume of recorded movement is relatively low compared
standard deviations of the measured change across the to some other rockfall geohazard sites being currently
slope (Deane et al. 2020), excluding areas of identified monitored in Alberta (Rodriguez et al. 2020). More events
movement or rockfalls. The second used the equation and a greater total volume were detected originating from
presented in Lague et al. (2013). The registration error S042-North. Most of the volume, however, was attributed
between each cloud pair was estimated using RMS value to movements from the extensive talus slope. Only 6 of the
resulting after the final fine registration. The results for both 44 movements detected were greater than the small
methods were averaged across each of the cloud pairs and rockfall threshold, of 0.3 m3 (introduced in Section 2), and
the largest of 0.2 m, was selected as the overall LoD, which none of these were detachments from the slope. The
was applied to each analysis. largest rockfall from the slope at S042-North was
measured to be approximately 0.12 m3. Large rockfall were
3 CHANGE DETECTION RESULTS more prevalent from the S042-South slope with the two
largest volumes estimated as 2.10 m3 and 0.42 m3, both of
A total of five change detection analyses were completed which are highlighted in Figure 2(b). It should be noted that
for the S042 site, two for S042-North and three for S042- none of the events detected within the 4-year period were
South, utilizing point clouds collected in 2018 as reference close to the 15 m3, large rockfall threshold, but 4 events
clouds. Following the completion of the change detection were detected greater than 1.0 m3, indicating an estimated
analysis described in Section 2.1, each cloud was closely event frequency for 1.0 m3 events equal to 1 (one per year).
inspected to identify rockfall events across the slope and Macciotta et al. (2020) estimated failure frequencies at the
movements within the talus which were likely to have S042 site to be 0.1 (one every 10 years) and 0.01 (one
reached Highway 742. The volume of each isolated every 100 years) for volumes 1 to 10 m3 and 10 to 100 m3,
movement was estimated two different ways. The first was respectively. Considering the large rockfall event in 2013
the 2.5D volume tool built into the CloudCompare software. which originated from S042-North (highlighted in Figure 2a)
It works by rasterizing point pairs between the reference the current period of change detection results fails to
and aligned cloud and summing the product of the capture the likelihood and return period of large rockfall
columnal distances over a specified grid spacing. As noted events (greater than 15 m3) at the S042 site. While a
by DiFancesco et al. (2020) this method, while robust, is greater volume of movement is attributed to S042-North,
highly sensitive to the grid spacing and other input the change detection results provide a misleading
parameters. A relatively high degree of uncertainty was prediction that rockfall detachments from the slope are
observed in the volumes ascertained via the 2.5D volume more likely to occur from S042-South. This indicated that
calculations mainly attributed to the low point density at the the continuation of monitoring, through remote sensing, is
scale of isolated movements. prudent for the S042 site for TEC to effectively quantify the
The second method used to compute volumes utilized risk posed to the public at this location.
the M3C2 clouds; the change detected was averaged over
the area of each detected movement to estimate their 4 SLOPE CONDITION ASSESSMENT
volume. Table 1 presents each of the change detection
analyses completed with the corresponding rockfalls and A site visit to S042 and application of five rock slope and
talus slope movements identified from each analysis. While rock mass rating systems was completed on September
the change detection analysis presented in this paper 25, 2022. The rating system results available for the S042
includes events detected from both the rock slope and talus site include: the GRMP RL rating which is the product of a
slope. The influence of the talus slope on potential rockfalls probability factor (PF) and a consequence factor (CF) (TEC
reaching the highway has not been assessed. It is 2020), the RHRS, the CRHRS, Q-Slope and Q-Slope β
expected that some energy from upper slope rockfall angle, GSI, and the 1989 version of the RMR system which
events would be dissipated by the talus slope. was used to validate the selected GSI range (RMR89’)
(Hoek and Brown 1997). Details pertaining to each of the
rating systems and how they were implemented is
Table 1. Change detection analyses and estimated presented in Wollenberg-Barron et al. (2023). A summary
movement volumes. of the rating system results for the S042 site is presented
in Table 2. The results of each rating system were reviewed
Slope Change Total measured change Number to consider how each can benefit from performance
Detection (m3) of events monitoring via remote sensing techniques as part of a
Slope Talus Total formalized GAM program.
S042-North 2018-2021 0.24 4.23 4.47 21
2018-2022 0.13 2.87 3.00 23
S042-South 2018-2020 0.02 0.10 0.12 5
2018-2021 0.42 0.43 0.85 7
2018-2022 2.19 0.48 2.67 8
Table 2. Rock slope and rock mass rating system results such as the rockfall history and block size (RHRS
for site S042. categories) or the rockfall frequency and block
size/volumes (CRHRS categories) can be updated to
Rock Slope S042- S042- S042- S042- reflect the observed event frequencies. While the RHRS
Rating Systems North-S North-L South-S South-L scores were similar for both the north and south slopes
PF 13 7 13 7 (Table 2), the CRHRS scores deviated significantly from
north to south due to the additional scoring categories for
GRMP

CF 4 7 4 7
slope aspect. S042-South, with a southern slope aspect
RL 52 49 52 49 scored the maximum value in this category, corresponding
Slope angle (°)* 60 60 60 60 to the increased number of freeze thaw cycles the rock
RHRS 293 358 287 352 mass is likely to experience. The value of including a slope
CRHRS 747 777 819 849 aspect category is corroborated by the current period of
change detection results which yielded the greatest volume
Q-Slope 0.983 0.983 0.986 0.986
of rockfall detachments from S042-South. However, the
Q-Slope β (°) 64.8 64.8 64.9 64.9 large rockfall event in 2013 from S042-North indicates
GSI 55 – 70 55 – 70 50 – 65 50 – 65 clearly that slope aspect is not the only factor contributing
RMR89’ 69 69 66 66 to the occurrence of large rockfall events at this site. This
* Average slope angles approximated in the field and corroborated highlights a possible limitation when combining change
with remote sensing data. detection results with the RHRS and CRHRS. If the
historically observed rockfall occurrences are neglected
from the development of RHRS and CRHRS scores, the
4.1 GRMP Risk Level Rating risk may be underestimated. For example, at the S042 site,
the largest event volumes measured from the S042-South
The GRMP RL rating is TEC’s current risk assessment tool results in an equivalent RHRS block size category score to
for geohazard sites across Alberta. TEC has four tables of S042-South-L. For S042-North however, the RHRS block
probability and consequence factors to compare a variety size category score would be reduced by approximately 33
of geohazard sites including: earth sides and debris flows, points from what is presented for S042-North-L. Following
rockfalls, erosion, and dispersive (collapsible) soils. the same logic, the RHRS rockfall history would be
Integrating change detection results with the GRMP RL increased for S042-North-L by approximately 6 points,
ranking would improve TEC and their consultant’s ability to resulting in a deceptive reduction of the perceived risk.
make informed decisions when selecting an appropriate PF While combining change detection analysis with RHRS and
and CF based on a measured event frequency, in addition CHRHS ratings has benefits, they are limited by the period
to observations made by the highway maintenance of remote sensing data available. Therefore, historically
contractors. While subjective in nature, the two categories observed rockfall events must not be neglected.
which make up the RL, the PF and CF, have key
descriptive words which align well with rockfall frequency 4.3 Q-Slope
analysis results, indicating a range of event frequencies.
Performance monitoring metrics, in the form of a measured The Q-Slope results for site S042 are nearly the same for
rockfall frequency, allows practitioners to re-evaluate the both the north and south slopes (Table 2), deviating slightly
selected category from a desktop setting. For small events due to the joint orientation changes resulting from the
(S042-North-S and S042-South-S), several should be change in slope aspect. It is less clear how Q-Slope, as a
expected each year. This aligns with the assigned PF of 13 condition assessment tool, can benefit from performance
(on a scale from 1 to 20). With regards to large events monitoring through remote sensing since Q-Slope is not
(S042-North-L and S042-South-L), while the event in 2013 impacted by changes in rockfall frequency or volume. Q-
was upwards of 15 m3, no events close to this volume was Slope could benefit from other forms of performance
detected within the current 4-year change detection period. monitoring, such as subsurface exploration or structural
This suggests that while 15 m3 is an important historical mapping of the rock mass. Detailed structural mapping
rockfall volume, it may be too large as a threshold value for could be performed on remote sensing data collected from
the S042 site. A rockfall event of 2 m3 would result in a the S042 site to increase the confidence of the rock quality
similar consequence if it were to impact Highway 742, while designation (RQD) and Jn parameters. Subsurface
its corresponding PF would be greater than 7, possibly in exploration, however, is not practical for the S042 site due
the 9 to 11 range. From the analysis presented herein on to drill rig access limitations. From this research, Q-Slope
the S042 site, the GRMP RL rating can benefit from change is limited as an initial condition assessment tool to assist in
detection results to provide quantifiable event frequencies the prediction of likelihood of failure for rockfall
when selecting an appropriate PF and to verify the selected geohazards. This limitation is exacerbated by how
CF corresponds to the consequences of event volumes challenging it is to estimate Q-Slope parameters through
measured through change detection. visual inspection alone, which is the preferred method for
TEC when performing annual inspection of rockfall
4.2 RHRS and CRHRS geohazard sites across Alberta.

Both the RHRS and CRHRS can benefit from the

integration of remote sensing data. Scoring categories
 Figure 2. Change detection analysis results from 2018 and 2022 for (a) S042-North, highlighting the
approximate location of the 2013 rockfall event and (b) S042-South highlighting location of two largest
rockfalls detected.

4.4 GSI and RMR as a preliminary slope assessment tool, but for estimating
rock mass characteristics. The GSI and RMR results for
GSI lends itself well to the initial condition assessment of site S042 only differ slightly between the two slope aspects
rock slopes since it was developed for application through with the rock comprising S042-North appearing slightly
visual inspection of a slope, focusing on the surface quality more competent with less obvious signs of weathering than
and structure of a rock mass. The RMR89 was employed S042-South (Table 2). This difference is captured by the
in this study as a tool to verify that the select GSI range current period of change detection results with greater
was appropriate. It should be noted that GSI and RMR volume of rockfalls detachments from the slope detected
cannot capture the impact of the talus slope on the from S042-South. However, these findings do not align with
likelihood, let alone consequence, of rockfalls impacting the large rockfall event from the north slope in 2013. Similar
Highway 742. Additionally, GSI was never intended for use to Q-Slope, GSI and RMR are not directly impacted by
rockfall event volumes or frequencies and receive a limited tools can incorporate, and benefit, from performance
benefit with the integration of change detection results. monitoring through remote sensing.
This limits the practicability of GSI for use, pertaining to a
GAM program, to an initial assessment tool applied though 6 ACKNOWLEDGEMENTS
visual inspection alone. While RMR could benefit from
different forms of performance monitoring, like Q-Slope, The authors wish to thank Klohn Crippen Berger Ltd. and
such as subsurface exploration, this is not possible for the Alberta Transportation and Economic Corridors for their
S042 site due to access limitations. contributions to the work presented in this paper and for
facilitating access to the sites and historic information.
5 CONCLUSION Special thanks to Sonam Choden, and previous students
from the University of Alberta Geotechnical Centre, for
The results of a change detection analysis conducted at the assisting in the collection of field data.
S042 rockfall site near Canmore are presented in this
paper, along with a review of how these results can be 7 REFERENCES
integrated with rock slope and rock mass rating systems to
reduce the uncertainty TEC experiences when allocating Agriculture and Irrigation, Alberta Climate Information
their available resources for the management or Service (ACIS). 2023. Data provided by Agriculture and
remediation of rockfall geohazards along highway corridors Irrigation, Alberta Climate Information Service (ACIS)
in Alberta. TEC’s current condition assessment tool, the Retrieved from https://acis.alberta.ca (March 2023)
GRMP RL ranking, can clearly benefit from change Alberta Transportation and Economic Corridors. 2022.
detection results to provide quantifiable event frequencies Traffic Volumes at points on the highway. Retrieved
when selecting an appropriate PF. There is also a potential from https://open.alberta.ca/opendata/traffic-volumes-
benefit to the CF by reviewing how the measured rockfall at-points-on-the-highway
volumes would impact the respective highway. Other Alberta Transportation and Economic Corridors. Appendix
condition assessment tools such as the RHRS and CRHRS A – GRMP Risk Rating System (Earth slides, Rock fall,
would benefit from performance monitoring, via remote Erosion. 2006
sensing, for select categories which rely on historical data http://www.transportation.alberta.ca/PlanningTools/G
pertaining to rockfalls experienced at the site. Since these MS/Annual%20Landslides%20Assessments/Appendix
methods are designed to quantify both likelihood and %20A%20-
consequence, they have the greatest potential, from the %20GRMP%20Risk%20Rating%20System%20(Earth
methods tested, for use as condition assessment tools as %20slides,%20Rock%20fall,%20Erosion).pdf
part of a formalized GAM program. The other rating AMEC Environment and Infrastructure. 2015. Southern
systems applied to site S042, i.e., Q-Slope, GSI, and RMR, Region Geohazard Assessment, 2014 Annual
do not clearly benefit from the inclusion of remote sensing Inspection Report, Site S42: Highway 742:02, Spray
data as part of rockfall geohazard performance monitoring. Lakes Rock Fall.
This is a reasonable conclusion since these tools were not http://www.transportation.alberta.ca/PlanningTools/G
originally intended to be used as condition assessment MS/Annual%20Landslides%20Assessments/Reg-
tools but were included in this study as industry accepted Southern%20(CMA517,%20CRR)/Inspection%20Sites
tools for assessing rock slopes. /25291_02%20or%20742_02%20(S042)%20-
It is important to note that the change detection analysis %20Spray%20Lakes%20Rockfall/Reports/2014%20S
from the current monitoring period failed to provide a return 42%20Inspection%20Report.pdf
period for large rockfall events, greater than 15 m 3, possibly Barton, N., and Bar, N. 2015. Introducing the Q-slope
misleading a reviewer to think rockfalls are more prevalent method and its intended use within civil and mining
from S042-South. This highlights the importance of engineering projects. In ISRM Regional Symposium-
incorporating the observational rockfall history of a site into EUROCK 2015.Bernhardt, K.L.S., Loehr, J.E., and
the condition assessment and performance monitoring of Huaco, D. 2003. Asset management framework for
rockfall geohazards. Continuation of the remote sensing geotechnical infrastructure. Journal of infrastructure
program at the S042 site is recommended to estimate the systems, 9(3): 107–116. American Society of Civil
return period of large rockfall events similar in volume to Engineers.
the 2013 event which originated from the north slope. Bieniawski, Z.T. 1993. Classification of rock masses for
The technical and economic challenges intrinsic to the engineering: the RMR system and future trends. In
management of rockfall geohazards are an ever-present Rock Testing and Site Characterization. Elsevier. pp.
concern at TEC. There is a clear benefit for TEC from 553–573.
research which reviews the available tools for assessing CloudCompare [GNU GPL software] (2.12) (2022)
and monitoring rockfall geohazard, especially as TEC Retrieved from http://www.cloudcompare.org/
transitions towards a formalized GAM program. The Deane, E., Macciotta, R., Hendry, M.T., Gräpel, C., and
integration of performance monitoring techniques, such as Skirrow, R. 2020. Leveraging historical aerial
remote sensing, with condition assessment tools is an photographs and digital photogrammetry techniques
important part of any comprehensive GAM program. The for landslide investigation—a practical perspective.
results presented in this paper provide a preliminary Landslides, 17(8): 1989–1996.
assessment of how established rock mass characterization
DiFrancesco, P.M., Bonneau, D. and Hutchinson, D.J., Canmore, Alberta, Canada: managing risk calculation
2020. The implications of M3C2 projection diameter on uncertainty in practice. Canadian Geotechnical
3D semi-automated rockfall extraction from sequential Journal, 57(3), pp.337-353.
terrestrial laser scanning point clouds. Remote Pierson, L.A. 1992. Rockfall Hazard Rating System.
Sensing, 12(11), p.1885. Transportation Research Record 1343
Google Earth Pro 7.3. 2022. Central and southern Alberta. Rodriguez, J., Macciotta, R., Hendry, M.T., Roustaei, M.,
Image © 2022 Google LLC. Retrieved in December Gräpel, C., and Skirrow, R. 2020. UAVs for monitoring,
2022.] investigation, and mitigation design of a rock slope with
Hoek, E. 1994. Strength of rock and rock masses. ISRM multiple failure mechanisms—a case study.
News Journal, 2(2): 4–16. Landslides, 17(9): 2027–2040.
Hoek, E., and Brown, E.T. 1997. Practical estimates of rock Russell, C.P., Santi, P., and Higgins, J.D. 2008.
mass strength. International Journal of Rock Modification and Statistical Analysis of the Colorado
Mechanics and Mining Sciences, 34(8): 1165–1186. Rockfall Hazard Rating System. Prepared for Colorado
Klohn Crippen Berger. 2019. Southern Region GRMP Site Department of Transportation.
Inspection Report, S042 - I & II Spray Lakes Rockfall. Tappenden, K.M., and Skirrow, R.K. 2020. Vision for
http://www.transportation.alberta.ca/PlanningTools/G Geotechnical Asset Management at Alberta
MS/Annual%20Landslides%20Assessments/Reg- Transportation. 2020 Canadian Geotechnical
Southern%20(CMA517,%20CRR)/Inspection%20Sites Conference (GeoVirtual).
/25291_02%20or%20742_02%20(S042)%20- Wollenberg-Barron, T.D.G., Macciotta R., Tappenden,
%20Spray%20Lakes%20Rockfall/Reports/2019R%20 K.M., and Skirrow, R.K. 2023. Comparison of Rating
S042%20Inspection%20Report.pdf Systems for Alberta Rock Slopes, and Assessment of
Kromer, R.A., Hutchinson, D.J., Lato, M.J., Gauthier, D. Applicability for Geotechnical Asset Management.
and Edwards, T., 2015. Identifying rock slope failure University of Alberta Geotechnical Centre
precursors using LiDAR for transportation corridor [Unpublished].
hazard management. Engineering Geology, 195,
pp.93-103.
Lague, D., Brodu, N. and Leroux, J., 2013. Accurate 3D
comparison of complex topography with terrestrial laser
scanner: Application to the Rangitikei canyon
(NZ). ISPRS journal of photogrammetry and remote
sensing, 82, pp.10-26.
Lato, M., Diederichs, M.S., Hutchinson, D.J., and Harrap,
R. 2009. Optimization of LiDAR scanning and
processing for automated structural evaluation of
discontinuities in rockmasses. International Journal of
Rock Mechanics and Mining Sciences, 46(1): 194–199.
Lato, M., Kemeny, J., Harrap, R.M., and Bevan, G. 2013.
Rock bench: Establishing a common repository and
standards for assessing rockmass characteristics using
LiDAR and photogrammetry. Computers &
Geosciences, 50: 106–114.
Lato, M.J. 2010. Geotechnical applications of LiDAR
pertaining to geomechanical evaluation and hazard
identification. Library and Archives Canada. Canada,
Ottawa.
Lato, M.J., Diederichs, M.S., Hutchinson, D.J., and Harrap,
R. 2012. Evaluating roadside rockmasses for rockfall
hazards using LiDAR data: optimizing data collection
and processing protocols. Natural Hazards, 60(3): 831–
864.
Macciotta, R., Hendry, M. 2021. Remote sensing
applications for landslide monitoring and investigation
in Western Canada. Remote Sensing 13(3): 366
Macciotta, R. and Martin, C.D., 2019. Preliminary approach
for prioritizing resource allocation for rock fall hazard
investigations based on susceptibility mapping and
efficient three-dimensional trajectory modelling.
Bulletin of Engineering Geology and the Environment,
78(4), pp.2803-2815.
Macciotta, R., Gräpel, C., Keegan, T., Duxbury, J. and
Skirrow, R., 2020. Quantitative risk assessment of rock
slope instabilities that threaten a highway near
Monday, October 2, 2023

HYDROGEOLOGY II
TITLE A TOWED MARINE TIME-DOMAIN ELECTROMAGNETIC
METHOD FOR EVALUATING THE CONTINUITY OF AN AQUITARD

Ashlee Fudge, M.Eng., Jane Dawson, Ph.D. and Drew Fossen B.Sc.
DMT Geosciences Ltd, Calgary, AB, Canada

ABSTRACT

Time Domain Electromagnetic (TEM) methods have long been used for hydrogeological studies. Investigations into
hydrostratigraphic units are common because of the ability to map aquifer properties based on grain size, which is
commonly related to electrical conductivity contrasts in the subsurface. This investigation showcases the use of Geonics
Ltd.’s G-TEM system in a mobile acquisition setup for waterborne TEM. The survey aimed to map the continuity and
composition of a conductive glacial clay till unit underlying a surficial water body. Variations in the clay till may indicate
zones of interest for possible hydrogeological pathways. Preliminary results showed variability in depth to both the top
and base of the clay till. These variations may suggest a connection between structures in the deep underlying units and
the glacially deposited clay till unit. Results from the GTEM compared favourably to previously acquired ground TEM
surveys.

ABSTRAIT

Les méthodes électromagnétiques dans le domaine temporel (TEM) sont utilisées depuis longtemps pour les études
hydrogéologiques. Les études sur les unités hydrostratigraphiques sont courantes car elles permettent de cartographier
les propriétés des aquifères en fonction de la taille des grains, qui est généralement liée aux contrastes de conductivité
électrique dans la subsurface. Cette étude illustre l'utilisation du système G-TEM de Geonics Ltd. dans un dispositif
d'acquisition mobile pour le TEM en milieu aquatique. L'étude visait à cartographier la continuité et la composition d'une
unité de till argileux glaciaire conductrice sous-jacente à une masse d'eau superficielle. Les variations dans le till argileux
peuvent indiquer des zones d'intérêt pour d'éventuelles voies hydrogéologiques. Les résultats préliminaires ont montré
une variabilité de la profondeur à la fois au sommet et à la base du till argileux. Ces variations peuvent suggérer un lien
entre les structures des unités sous-jacentes profondes et l'unité de till argileux déposée par les glaciers. Les résultats
au sol du GTEM ont été comparés favorablement aux études TEM au sol acquises précédemment.

1. INTRODUCTION
2. METHODOLOGY
To understand the interaction between groundwater and
surface water underlying a water body, it is essential to 2.1 TEM
understand the geologic setting. Obtaining such
information under surficial water bodies can be The TEM sounding method is a time domain technique
challenging. Time Domain Electromagnetic (TEM) that resolves the resistivity of the Earth’s subsurface at
methods have long been used for hydrogeological / predetermined time (depth) increments. TEM follows
groundwater studies. Investigations into Faraday’s Law that an induced voltage in a circuit is
hydrostratigraphic units are common because of the proportional to the rate of change of magnetic flux
ability to map aquifer properties based on grain size, through that circuit.
which is commonly related to electrical conductivity
contrasts in the subsurface. Sending electricity through a wire coil induces a primary
magnetic field through that coil. While the current is
The goals of the survey were to map the continuity and constant, the primary magnetic field invariant. Abruptly
composition of a conductive glacial clay till unit reducing the transmitter current to zero will induce a
underlying a surficial water body. voltage due to the sudden change in the magnetic field.
This voltage will induce secondary eddy currents in the
Commonly used as a land method, we have adapted a subsurface, which create secondary magnetic fields.
Geonics G-TEM system for marine acquisition. This These magnetic fields in the subsurface will attenuate
investigation showcases the use of Geonics Ltd.’s G- with time and are measured at discrete time gates (U.S.
TEM system applied in a mobile acquisition setup for Army Corps of Engineers, 1995).
waterborne TEM. The G-TEM system allowed rapid,
high-resolution acquisition across the lake for Instrumentation consists of a transmitter to impart
hydrostratigraphic mapping and evaluation of the clay current into a square single or multi-turn loop of wire laid
till unit. on the ground surface and a receiver coil and module to
measure and record the resulting magnetic field (U.S.
Army Corps of Engineers, 1995).
 continuously and simultaneously. Data was collected at
an average rate of 15 km per day.

Figure 1. Central Loop sounding configuration (US

EPA, 2023).

The secondary magnetic field, as measured at the

surface at incremental time gates, is dependent upon
the electrical properties of the subsurface (US EPA,
2023). In this manner, vertical variations in electrical Figure 3. G-TEM time domain system from Geonics
properties may be resolved. Lateral resistivity variations
are determined by moving the TEM instrument 2.2 Modelling
configuration along the surface and acquiring
measurements at discrete intervals. For better lateral Figure 4 shows an example of a raw TEM sounding. The
resolution, transmitter loops can be overlapped by 50 data points are displayed as square symbols, and each
percent (Nabighain & Macnae, 1991). colour represents a different transmitter frequency:
deep purple represents data collected at 285 Hz, and
For this program, however, the survey array was in blue represents data collected at 30 Hz. Since
constant motion to keep the array in alignment while on increasing time relates to increasing depth, a qualitative
the water. To accommodate for constant motion, the assessment can be made of the data. Generally, a
collection of each TEM sounding was triggered on a negative slope indicates the presence of a material of
time interval. This, coupled with variations in boat speed lower resistivity than the overlying material. Similarly, a
and ever-shifting environmental conditions, meant positive slope indicates the presence of a material with
minor variations in the spacing of the TEM soundings. It higher resistivity than the overlying material.
was determined that the average spacing between each
TEM sounding along a survey line was approximately
18 m.

The general setup for marine acquisition was a boat that

would carry the receiver module while towing a small
boat/dinghy with the receiver coil in it while towing a
smaller multi-turn transmitter loop that was mounted to
a pontoon-style raft further back (Figure 2).

Figure 2. Marine TEM array schematic.

Figure 4: Raw TEM Sounding.
Unlike the central loop configuration in Figure 1, this
survey used an “out of loop” configuration to As shown in Figure 5, the measured results can then be
accommodate the condensed size of the transmitter forward-modelled and inverted to produce a geoelectric
loop. A 5 x 5 metre multiturn transmitter loop was used cross-section of the subsurface with several layers
with a spacing of 20 metres from the receiver coil.

To achieve lateral variation resolution on a water body,

a data collection system that could acquire data
continuously was required. On this project, the G-TEM
system by Geonics was used (Figure 3). The G-TEM is
a time-domain electromagnetic instrument capable of
continuous data acquisition. This system can monitor
data in real-time (Geonics, 2013). Crews could easily
resample survey locations if there were any concerns
with the data. TEM and GPS data were collected
consistent with the inflections found in the raw data transducer and GPS units of the system were mounted
(Interpex.com, 2018). to the transom of the TEM acquisition boat, and
readings for both units were collected at a rate of one
per second and combined using Chesapeake
Technology’s SonarWiz software. The sample rate of
bathymetry data and the speeds the survey crew were
limited to allowed for very dense data along each survey
line with a depth to water bottom occurring
approximately every 0.5 m.

3. RESULTS

Figure 7 shows a typical sounding from the marine TEM

Figure 5. TEM sounding data (left) and resulting depth survey with a preliminary inversion. Data was collected
model in IX1D (right). as a single frequency sweep to optimize integration time
and acquisition efficiency. The final four time gates were
consistently noisy and were stripped from the dataset
With most modelling, there are inherent equivalency
leaving 16-time gates for the inversions.
issues. The modelled data is shown as the solid red line
in Figure 6, while equivalent models are shown as the
dashed green lines. Each model honours the data and
the original model with less than a 5 percent error. To
overcome equivalency in modelled data, secondary
data sets such as drill data should be used to constrain
the models to produce the most accurate results.

Figure 7: Example Sounding.

Initial inversions were carried out over four test lines to

determine system accuracy of the inversion parameters.
One drill hole was available for ground truthing. The
results of a line coincident with the borehole location
shown in Figure 8 showed a good correlation with the
top of the clay till.

Figure 8. TEM inversion results compared to available

borehole data.

A comparison was also made between a TEM line

collected on the ice in the winter and the marine TEM
data (Figure 9). Results are comparable, therefore
Figure 6. Equivalent forward models of the TEM data. increasing confidence in the results of the marine data.

2.3 Bathymetry

The bathymetry data were collected using a BL200

survey grade 200 kHz single beam echo sounder with
an accuracy of 0.01 m +/- 0.1% depth and a beam width
of 9 degrees. Calibration was performed at the boat
launch site by using a surveyor height measuring rod to
determine the correct velocity of the lake water to give
the most accurate measurement of water depth. The
Figure 9. Comparison of TEM data collected on ice (top)
and TEM collected on water (bottom).

3.1 Processing
Data were processed using two different software
packages: IX1D and EM1DTM. The reasoning behind
this approach was to assess which inversion algorithm
would produce the best results and which would be
most efficient given the large number of soundings
collected.

3.1.1 IX1D Figure 10: Bathymetry results.

IX1D is the algorithm DMT uses most commonly for 4. INTERPRETATION & DISCUSSION
ground TEM data. IX1D has the ability to provide both
layered earth and smooth (Occam) inversions Preliminary results in Figure 11 and Figure 12 showed
(Interpex.com, 2018). We have found that both types of variability in depth to both the top and base of the clay
inversions are useful and indeed we commonly carry out till, respectively.
both inversions on the same data set as we did in this
case.

3.1.2 EM1DTM

EM1DTM is the algorithm DMT uses most commonly for

airborne data. The advantage of this inversion is that it
can rapidly process and invert large datasets. With
proprietary add-on software, DMT can do laterally
constrained inversions to improve the model (EM1TDM,
2018).

3.2 Bathymetry

The bathymetry data in Figure 10 was gridded using a

kriging algorithm. Depths ranged from less than 0.5
metres in the southwest and up to 7 metres at its
deepest. The bathymetry shows two distinct trough-like
or trench features in the lake bottom: one trending north-
northeast, the other west-northwest. A third, smaller
trench is trending west from the southern tip of one of
the trenches. The water bottom is the shallowest along
the northern shore of the lake.

Figure 11: Depth to the top of the clay till.

Resistivity values between 8 – 30 ohm-m were chosen
as a conservative estimate to model the clay till unit. A
minimum of 8 ohm-m was chosen as resistivities below
that value would indicate potential for basal waters. A
maximum of 30 ohm-m was chosen as resistivities
above that value indicate more coarse-grained
sediments (Telford, Geldart, & Sherrif, 1990).

These variations suggest a connection between

structures in the deep underlying units and the glacially
deposited clay till unit. The results of this survey will be
used to identify zones of interest based on variations in
the clay till, which will help guide strategic drilling
locations to verify the models and further evaluate
hydrogeological processes into/out of the lake.

Figure 13. Average Resistivity of the clay till. Values

are between 8 and 30 ohm-m.

A section (A) with lower resistivity values from Figure 13

is shown in Figure 14. TEM data showed a well-defined
low resistivity layer with resistivities in the range of 8-30
ohm-m at an approximate depth of 20 metres. The clay
till is continuous and does not vary in thickness.

Figure 12. Depth to the base of the clay till.

The average depth to the top of the clay till varied from Figure 14. Line A from Figure 13. Average resistivity of
1 – 34 metres. A significant drop in depth is present on the clay till layer. Vales are between 8 and 30 ohm-m.
the lake's northeast side. Variations in the clay till may Results show a well-defined continuous clay till layer.
indicate zones of interest for possible hydrogeological
pathways. Due to the conductive nature of the clay till A section (B) with higher resistivity values from Figure
the base was not well resolved. The conductive (low 13 is shown in Figure 15. Within the clay till, areas of
resistivity) properties of the clay till attenuate the signal higher resistivities are visible in the 30-40 ohm-m range.
faster than less conductive sediments, allowing for less These zones of higher resistivity may indicate
depth penetration of the signal. weaknesses within the clay till allowing hydrogeological
pathways. Any faulting present within the clay till will
The results from the TEM survey show variation in also have the potential to affect hydrogeological flow.
average resistivity values of the clay till in Figure 13.
Higher resistivities may indicate a higher risk for
potential hydrogeological pathways as these deposits
may have more sands than clays.
 cases on the edges of the trenches. It is likely therefore
that some common mechanism is affecting both the
bathymetry and the clay till in these areas. One theory
is that faulting in the Precambrian may the root cause.
As discussed earlier integration with marine seismic to
identify any structure such as faulting or collapses in the
Precambrian.

5.0 CONCLUSION
Figure 15. Line B from Figure 13. Results show a
discontinuous clay till layer. Resistivities above 30 ohm- The TEM survey successfully mapped the thickness
m within the clay till may indicate potential for and average resistivity of a clay till layer under a body
hydrogeological pathways. of water. Variability in depth to the top of the clay till may
indicate a connection between the glacially deposited
A number of other interesting features were identified in clay till unit and deep underlying structures. Due to the
the clay till. The top and base of the clay till show an conductive nature of the clay till unit, TEM had a
almost 20 m decrease in elevation along the northern maximum depth of penetration that could not capture
shore of the lake, an example of which is shown in such deep structures. Identifying zones of higher
Figure 16. The underlying mechanism for this dip in the resistivity within the clay till is essential as these regions
clay till is unclear. An integration of this data with land may indicate zones of potential hydrogeological
and air-based TEM may shed light on this issue as well pathways. Limited ground truthing data was available,
as marine seismic to identify possible structures. and like with most geophysical methods, a
complimentary data set would enhance the results.
Additional surveys may be completed with acoustic sub-
bottom profiling to investigate these anomalies further.
The results of this survey will be used to identify zones
of interest based on variations in the clay till, which will
help guide strategic drilling locations to verify the
models and further evaluate hydrogeological processes
into/out of the lake.

6.0 REFERENCES

EM1TDM. (2018). EM1DTM package. Retrieved May 2,

2023, from EM1DTM:
Figure 16: Example of dipping clay till at the northern https://em1dtm.readthedocs.io/en/latest/index
edge of the lake. .html
Geonics. (2013). G-TEM. Retrieved 04 12, 2023, from
In a number of locations anomalous “humps” were Geonics Limited:
identified in the clay till such as that shown in Figure 17. http://www.geonics.com/html/g-tem.html
These features are about 150-300m in diameter. As with Interpex.com. (2018). IX1DV3. Retrieved 05 02, 2023,
the dipping clay till, the underlying mechanism for these from Interpex:
features is unclear. http://www.interpex.com/ix1dv3/ix1dv3.htm
Nabighain, M. N., & Macnae, J. C. (1991). Time Domain
Electromagnetic Prospecting Methods. In M.
N. Nabighain, Electromagnetic methods in
applied geophysics (Vols. Volume 2,
Application, Parts A and B, pp. 427-520).
Society of Exploration Geophysicists.
Telford, W., Geldart, L., & Sherrif, R. (1990). Applied
Geophsics (2nd ed. ed.). Cambridge
University Press.
U.S. Army Corps of Engineers. (1995). Geophysical
Exploration for Engineering and Environmental
Investigations.
US EPA. (2023, May 09). Time-Domain/Transient
Electromagnetics (TDEM/TEM). Retrieved
Figure 17. Example of "hump" in clay till on the left, with May 09, 2023, from US EPA:
a map highlighting the shallowest clay till on the right. https://www.epa.gov/environmental-
geophysics/time-domaintransient-
The bathymetry results were compared with the clay till electromagnetics-tdemtem
surface in Figure 11 to determine if there was any
relationship between the two. There is a clear
manifestation of the same trenches noted in the
bathymetry in the top of clay till. Another interesting
correlation is that the clay till “humps” occur in most
Surface and groundwater interaction in
determining baseflow contribution at Davis
Creek, SK.
Anatoly Melnik, Kei Lo & Nolan Shaheen
Water Security Agency, Moose Jaw, SK, Canada

ABSTRACT
Baseflow is a critical component to sustaining Species at Risk (SARA) and is aggravated by climate variability in semi-arid
regions. The contribution of ecological flow from shallow groundwater is being assessed by evaluating the surface-
groundwater interaction along the Davis Creek, Saskatchewan. The creek is an important habitat area for Mountain
Suckers (Catostomus platyrhynchus) and is listed as a species of special concern under SARA. The primary threat facing
the species includes water availability in the support of ecological flow. Natural groundwater discharge occurs along the
creek and is considered a contributing factor to sustained baseflow of the ephemeral stream.

A desktop study followed by drilling and installation of groundwater monitoring wells was completed along the creek. Water
levels, temperature, and water quality data were collected between 2020-2022 from four monitoring wells and a staff gauge
used to determine the hydraulic responses between the creek and the aquifers. Hydraulic response between bedrock
aquifer and the creek is strong; however, the magnitude and direction of fluxes change depending on season, precipitation,
stream flow, and location along the stream. Isotopic analysis of 18O and 2H of water followed the meteoric water line
suggesting direct recharge from the surface water body. Deeper pools of water occurred along the creek path and are
likely supported by groundwater discharge while other parts of the creek may contribute to groundwater recharge. The
presence of a diversion channel with control gate and a weir upstream from the investigated site added complexity to the
interaction between groundwater and surface water. Overall, the study results suggest that the relationship between the
groundwater resource and surface water is dynamic and seasonally variable along Davis Creek.

1 INTRODUCTION contributing to baseflow are also sensitive to seasonal and

long-term air temperature dynamics (Briggs et al., 2018).
The understanding of watershed hydrology, including both Rivers and streams that support fisheries are likely
groundwater and surface water (GW-SW), is crucial for hydrologically dependent on groundwater to maintain
sustainable water management. The interaction between instream flows. This suggests that variations in the
groundwater and surface water bodies (i.e., rivers, lakes, instream flow needs will be dependent on the local
streams, wetlands) is an important feature of hydrologic groundwater flow regime. The effects of climate change
cycle. However, in the past, components of hydrologic may result in greater water usage and reduce groundwater
cycle have been considered independently (Winter et al., contribution to baseflow in ephemeral streams. It may also
1999) and their interaction overlooked in semi-arid regions alter the timing of the spring freshet and impact the annual
(Fleckenstein et al., 2010). In an interconnected stream patterns and nutrient loading to lakes (Li et al.,
hydrological system, groundwater discharge is integral in 2020). Successful management of the ecological integrity
the contribution to baseflow to rivers and streams in streams will depend on the ability to adaptively manage
(Lewandowski et al., 2020). Groundwater withdrawal by the limited resources. A suitable thermal habitat is also
means of pumping impacts the groundwater and surface influenced by river morphology, riparian vegetation, and
water balance (Alley, 2007; Feinstein & Reeves, 2010; land use (Chu et al., 2008). Since groundwater discharge
Zhou, 2009). Surface water, especially in dry seasons, can into streams has been identified as a critical component of
also be largely affected by groundwater withdrawal and maintaining suitable thermal habitat for fishes, greater
thereby induces considerable fluctuations on the understanding of the complex surface-groundwater
groundwater levels (Sanz et al., 2011). The understanding interaction is required (Miranda et al., 2020).
of the hydrologic interactions in the prairie environment is
more complicated due to uncertainties associated with the The Saskatchewan Milk River population of Mountain
vertical interactions between the glacial and bedrock Sucker (Catostomus platyrhynchus) is listed as species at risk
aquifers and aquifer parameters. under the Canadian Species at Risk Act (SARA). This
species is considered a cool-water species associated with
Climate change is expected to cause an increase in stream streams of moderate flows. Spawning occurs in riffles of
temperature, alter the timing of the spring freshet, and moderate to fast flowing water often adjacent to pools
interrupt the annual stream flow patterns. Availability of (COSEWIC, 2010). Mountain Sucker has a limited
sustainable groundwater discharge for fisheries in streams distribution within southwestern Saskatchewan. Large
is influenced by several factors, including flow, land use, numbers of Mountain Sucker occur in Davis, Caton, and
water quantity and quality. Shallow alluvial aquifers Farewell Creeks. These creeks are ephemeral and are
reduced to a series of isolated pools by mid to late summer. The topographic elevation around Davis Creek ranges
Hundreds of fish congregate in each of these pools and between 950 to 1,250 metres above sea level. The
their survival is dependent on groundwater influxes elevation difference from the headwater to the confluence
maintaining these pools. Low flows and high-water with Frenchman River is approximately 200 m resulting in
temperatures resulting from drought or surface water an average vertical gradient of 0.7%. In some places the
extractions are threats to the species. Natural groundwater channel appears to be braided and the higher gradient
discharge into these ephemeral creeks effectively sustains results in relatively rapid flows forming riffle-pool
the pool habitat during the low flow periods of summer sequences. The surficial sediments of the creek bed and in
months. The buffering of temperature via groundwater the riparian zone consist of materials ranging from pebbles
discharge during both the summer and winter periods to large gravel mixed with sands and silts. The coarse
ensures fish survival through the coldest and warmest sediments of the creek bed are characteristic of a fast-
seasons. Given the importance of groundwater in flowing stream.
supporting this habitat and the economic value from Davis
Creek diversion, determining the significance of the
surface-groundwater interaction in the Davis Creek area
and identification of threats to the groundwater discharge
are important to the survival of the Mountain Suckers as
well as management of water resources.

Vertical and horizontal groundwater gradients combined

with surface water levels infer flow directions between
groundwater and surface water bodies. It is expected that
these gradients will change throughout the year due to
natural fluctuations. To assess the groundwater levels,
long-term monitoring of water levels will provide the most
accurate representation of surface-groundwater system
interaction. Additionally, water chemistry including major
ions, trace elements, and stable and radiogenic isotopes
can be used as a proxy for gaining insight to the surface-
groundwater flow dynamics (Lo and Melnik, 2017).

1.1 Study Area

The study area is located on the southern slopes of the east

block of Cypress Hills between Townships 6-8 and Range
25, West of the 3rd meridian (Figure 1). From its
headwaters in the Cypress Hills, Davis Creek flows south
for over 25 km until it merges with the Frenchman River.
The area is sparsely populated with grazing as the
dominant land use.

The creek has an approximate drainage area of 117 km 2

and all of it is considered effective. Hydrometric data for the
creek was collected during the early part of the 1900s
(1907-1928; hydrometric station 11AC004). Currently, Figure 1. Topography of the Davis Creek area.
there are no active hydrometric stations in the creek. From
the historical observation records, the mean monthly flows
range from 0.040 m3/s to 2.125 m3/s during the open water
season (March to October). In general, peak flows in the
creek occur during the spring runoff months (March to May)
with April being the most frequent month for peak flows.
April flows also account for ~ 55% of the annual runoff
volume. Figure 2 shows the seasonal peak flow conditions.

The average temperatures in Cypress Hills range between

-3.0 and -15 ºC in the winter and 8.6 to 23.2 ºC in the
summer. However, extreme temperatures between winter
and summer could occur in the range of -40.0 to 40.0 ºC.
The total average annual precipitation is 600 mm with 40%
of that being due to snow in the winter.
Figure 2. Typical mean flow in Davis Creek.
2 SITE DESCRIPTION AND DATA COLLECTION
Precipitation data were available from the nearest weather
The Davis Creek site has infrastructure, which includes a station at the Cypress Hills Provincial Park, approximately
weir, and separate diversion channel with a control gate. 20 km west of the site. No stream gauging stations exist
The Davis Creek diversion works includes a 21 m long upstream or downstream of the creak, thus, flow rates and
Coanda weir with 14 panels of fish screen attached onto volumes passing through Davis Creek are unknown.
the included face of the weir to mitigate the deaths of SARA
listed species. A bidirectional rock ramp fishway was Two water sampling events were completed for each
constructed to maintain flow continuity and provided a piezometer in fall 2020 and spring of 2021 to capture
natural fish passage in the creek all year around. The seasonal variation in water quality. Surface water samples
diversion canal has a capacity of 10 m3/s, which connects were collected near the staff gauge in Davis Creek. Each
Davis diversion weir to Belanger Creek. well was purged for a minimum of three well volumes.
Samples were sent to a commercial laboratory for major
To determine the local geological and hydrogeological ions and metals as well as isotopic analyses of oxygen and
environment surrounding Davis Creek, four monitoring hydrogen, including tritium. Duplicate samples were
wells were installed around the site (Figure 3) and one staff collected as a part of quality control.
gauge into Davis Creek. Well details are listed in Table 1.
The location and depth of the piezometers was to ensure
the capture of the glacial and bedrock sediments and to
provide the general lateral and vertical groundwater flow
directions across the site. Shallow wells, Davis 1A and
Davis 1B, are located immediately adjacent to the creek
and the staff gauge. The purpose of these wells is to
measure immediate interactions between groundwater and
Davis Creek and provide vertical gradient directly beneath
the creek. Davis 2 is located further away and downstream
from the creek to investigate the groundwater responses to
creek fluctuations further away from the creek. Davis 3 is
the deepest and furthest well in this study. Its purpose is to
record regional water level fluctuations that are expected
to occur due to a combination of moisture loading and
creek levels.

Table 1. Well completion details.

Monitoring Well Screen Depth (m) Target zone

Davis 1A 13.72 – 16.76 Eastend Fm.
Davis 1B 5.18 – 6.71 Surficial/Eastend
Davis 2 18.29 – 24.38 Eastend Fm
Davis 3 42.67 – 48.77 Bearpaw Fm. Figure 3. Site Layout and location of monitoring wells.
Davis Creek ~1 Davis Creek

The hydraulic properties of the aquifers were estimated

Data collected from four monitoring wells and staff gauge using water level data collected during the sampling event
were used to evaluate the surface-groundwater interaction to realize greater efficiency during the field work. Water
and its influence on Mountain Sucker habitat sustainment. levels were recorded during the purging period at
Currently these piezometers are the sole source of increasing time intervals, similar to a pumping test. The
information on the local geologic conditions and on the flow rate was known, and the effect of casing storage was
temporal water level fluctuations. To determine the eliminated by purging for an extended period, i.e., at least
groundwater flow directions and natural groundwater three well volumes to ensure consistent flow into the well
fluctuations, monitoring wells were equipped with digital from the aquifer. The aquifer parameters were determined
water level recorders. by the Cooper-Jacob straight line method. This method
allows for the determination of hydraulic conductivity
The monitoring wells and staff gauge were instrumented directly around the monitoring well. Average hydraulic
with pressure transducers that recorded the pressure of the conductivities of the completion zones around the wells are
water column in the piezometer. All transducers were in the range of 5×10-6 m/s and 8×10-6 m/s. These values
programmed to measure and record the water and are consistent with very fine sand and silt materials
atmospheric pressures as well as temperature on an hourly observed during well drilling and completion.
frequency. Water levels were compensated for barometric
pressure and plotted on composite hydrographs with
temperature.
3 REGIONAL GEOLOGY AND HYDROGEOLOGY in the area and is composed of black bentonitic shale
overlain by olive-green bentonite, shale, and silt. The
The Davis Creek area of Southwestern Saskatchewan is Frenchman Formation ranges in thickness from 7.6 m to 69
dominated by bedrock formations with thin and limited m (25 to 225 ft) and consists mainly of coarse cross-
glacial sediments (Maathuis and Simpson, 2007). The bedded brown sand with minor amounts of green and grey
bedrock stratigraphy consists of the Cretaceous age shale, silt, and bentonite. The Ravenscrag Formation
Bearpaw, Eastend-Whitemud and Frenchman Formations conformably overlies the Frenchman Formation and is up
and the Tertiary age Ravenscrag and Cypress Hills to 85 m (280 ft) thick and composed of non-marine sands,
Formations (Figure 4). A schematic geological cross silts, clays, and coals.
section along Davis Creek is shown in Figure 5 and is
consistent with lithological observations at the site during
drilling. Figure 6 illustrates the interpreted stratigraphy of
the area through the monitoring wells.

The Cretaceous Bearpaw Formation underlies much of the

study area and is the lowermost unit considered in this
study. It consists primarily of dark marine shale with
occasional sandstone members. Two of these sandstone
members, the Oxarart and Belanger are exposed along the
southern end of Davis Creek and Frenchman River valley.
Stratigraphically, the Belanger Member overlies the
Oxarart sand member by approximately 7.6m (25 ft) with a
thickness of 6 to 7.6 m (20 to 25 ft). The Belanger sand
member consists of brown and grey sandstone mixed with
dark shale containing concretionary layers that are rich in
fossils and commonly outcrops as a prominent ledge. The
Oxarart sand member underlies the Belanger sand
member. The Oxarart member is up to 6 m (20 ft) in
thickness and is composed of massive greenish-grey
sandstone that, at the top, is hard, rusty, and contains
fossilized wood, plant impressions, worm burrowing etc.

Overlying the Bearpaw Formation is the Eastend

Formation. The Eastend Formation is composed of grey to
green very fine sand, silt, and clay with thin coal seams. It
represents the transition from marine to non-marine
conditions of sedimentation and ranges in thickness from
20 m to 30 m (65 to 100 ft). The Whitemud Formation is
from 15 m to 23 m (50 to 75 ft) thick and originated in
conditions of alluvial and subaerial deposition. It is
composed of kaolinized, white sand and clay separated by
a carbonaceous zone. The Battle Formation is Figure 4. Bedrock geology of the Davis Creek area.
discontinuous and overlies the Whitemud Formation,
where present. The deposit is as much as 9 m (30 ft) thick

Figure 5. Schematic geologic cross-section along Davis Creek.

Bedrock water-bearing formations are the main aquifers in respond to precipitation events. However, the water level
the study area. These aquifers are contained within the increased during the 2021/2022 winter season, before the
Cypress Hills Formation that lies unconformably above the spring snowmelt. This is attributed to precipitation and
Ravenscrag and is the uppermost bedrock unit in the study higher temperatures in January and February of 2022.
area. It consists of thick beds of coarse conglomerate and Unstable warmer weather in the Cypress Hills during winter
massive, cross-bedded, hard, grey sandstone. The months is common and likely caused localized melting and
Quaternary sediments within the Davis Creek valley run off generating aquifer recharge. Midwinter melts and
consist of fluvial and colluvium: silt, sand, gravel, and snow cover depletion in the Canadian Prairies have been
bedrock rubble from slopes and terraces. documented by Pavlovskii et al. (2019) and are consistent
with the observations herein.
Belanger/Oxarart and other sands members of the
Bearpaw Formation, Eastend- Ravenscrag Formations Water temperature in parts of the creek during winter and
and the Cypress Hills Formation. The Eastend-Ravenscrag into early spring stayed around zero degrees Celsius and
aquifer consists of sand, silts, clays, and coal. The main began to rise in middle or late April, coincidental with air
aquitards are the Whitemud Formation and shale members temperatures. This rise occurred mostly after the peak
of the Bearpaw Formation. Aquitards likely also exist within water levels and reached as high as 20 degrees Celsius by
the aquifers as thin layers of clay due to heterogeneity. mid-summer. Generally, summer water temperatures
ranged between 12 - 20 degrees.
Groundwater flow is expected to be controlled by
topography and heterogeneities within each aquifer, During the winter months, water temperature in the creek
directed from high elevations in the Cypress Hills towards was measured at near zero degrees Celsius, while shallow
low-lying areas in the south of the study area as well as groundwater appears to be 5 degrees warmer than the
towards erosional valleys, such as Davis Creek. Flowing creek water. This suggests that groundwater seepage is
springs may also exist along the valleys where the aquifers exhibiting strong bidirectional exchange with the surface
are exposed and/or eroded at the outcrops. water where the groundwater is buffering the short-term
temperature swings (Hare et al., 2015). Water temperature
fluctuations in wells Davis 1A and Davis 1B have a smaller
amplitude than the temperature of the water in the creek.
The shallowest well, Davis 1B, shows temperature range
between 4 and 8 degrees Celsius throughout the season.
However, the highest temperatures tend to occur during
middle and late autumn. This means that there is a 3 – 4-
month lag in groundwater temperature response compared
to air and surface water.

Formation of ice in the creek depends on many factors

which are the heat radiating from ground, water turbidity
and salinity, and local freezing index. Water has a high
specific heat capacity (at least 5 times greater than dry
sediments) and combined with the high water table
connected to the creek, the warmer discharging
groundwater could buffer cold temperatures during the
winter season and preventing the complete freeze of the
creek and the Mountain Sucker. Temperature observations
within the creek and shallowest well (Davis 1B) suggest
that sufficient latent heat may be available in the ground to
prevent the formation of ice in the deepest parts of the
pools. In the summer months, cooler groundwater
temperatures yet again buffer the temperatures in the
creek pools and maintain preferred cooler conditions for
the Mountain Sucker.
Figure 6. Schematic of monitoring wells.
Water levels in Davis Creek and Davis 1B (shallowest well)
are nearly identical, except for summer, when the creek
4 RESULTS AND DISCUSSION shows consistently higher water levels between July and
September. In addition, the magnitude of water level
4.1 Water Level and Temperature Response response in the creek to precipitation events is muted
compared to Davis 1A and 1B. It is possible that the control
Groundwater levels and temperatures of Davis Creek, structure and weir located upstream from the monitoring
Davis 1A and Davis 1B are shown in Figure 7. Water levels site create a buffering effect during low flow periods while
at Davis Creek, Davis 1A and Davis 1B locations respond groundwater generally responds to more regional trends.
strongly to precipitation and spring snowmelt. Water levels
between June and November are mostly stable and
 shallow aquifer responds to broader fluctuations from
further upstream of this site.

Monitoring wells Davis 2 and Davis 3 show stable water

levels during winter months followed by an increase in early
spring (Figure 9). The response to the spring melt in these
wells is gradual and delayed compared to Davis 1A and
1B. This result is expected because Davis 2 and Davis 3
are deeper and further away and are unlikely to contribute
directly to the baseflow. Groundwater levels in the summer
and fall generally show decline followed by a stable period.

4.2 Hydraulic Gradient and Discharge Volume

Water level fluctuations exhibited throughout the

monitoring period are in response to precipitation, weir, and
Figure 7. Water levels and temperature data from Davis control structure. Hydraulic gradient generally remained
Creek and Davis 1A/1B wells. directed upward and towards the creek indicating “gaining
stream” conditions throughout much of the year.
Hourly water level data were examined in detail to Occasional high precipitation events resulted in rapid water
document the interaction between the groundwater and level rises in the creek exceeding groundwater levels. With
creek levels (Figure 8). A time lag is noted between the the dissipation of surface run-off, gradient returned to
peaks of water levels on that day. Although the data previous state within several hours.
resolution does not allow for precise determination of time
lag, approximate lag times in response to increased creek Seasonal changes in hydraulic gradients are also apparent
levels are 1-1.5 hours at Davis 1B and 2-3 hours at Davis with higher gradient during the spring months. Using
1A. In addition, the magnitude of response in Davis 1A is Darcy’s Law, average vertical gradient of 0.15 (from
50% lower than in Davis 1B and the creek suggesting a hydrograph), and estimated hydraulic conductivity, the
direct hydraulic relationship between the shallow alluvium discharge volume per square metre of creek bed at the
to the creek. monitoring site was estimated to be 4.5×10-8 m3/s, or 4 L
per day. However, this volume is likely to have significant
spatial variation as the gradient and hydraulic properties of
the aquifer change along the length of the creek.
Installation of seepage meters along the length of the creek
combined with stream gauging could provide a more
detailed representation of groundwater contribution to the
baseflow of Davis Creek.

Figure 8. Detailed water levels showing response time lag.

The vertical separation between the creek bed and screens

of Davis 1A and 1B are 11 m and 2.5 m, respectively. The
additional thickness of sediment separating Davis 1A from
the creek bed creates buffering effect and the time lag with
respect to propagation of pressure into deeper zone due to Figure 9. Water levels and temperature data from
any changes at the surface. Kurylyk et al. (2014) showed monitoring wells Davis 2 and Davis 3.
that the degree of annual signal in the groundwater
seepage zones was related to the depth of the confining till 4.3 Water Quality
layer below the shallow aquifer. The upward vertical
gradient is maintained despite the rise in the creek water Groundwater quality in all wells appears to be similar with
levels on Figure 8 or most other instances of increased flow total dissolved solids (TDS) of 322 – 524 mg/L and Na-
in the creek. This suggests that the lower zone of the HCO3 water type (Figure 10). Surface water quality from
Davis Creek has lower TDS of 249 mg/L and water type of
Ca/Mg-HCO3. The chemical separation of surface water Testing for tritium concentrations allows for a qualitative
from Davis Creek, as shown on the piper plot, is due to estimation of the groundwater age (Clark and Fritz, 1997).
reduced concentrations of sodium and potassium in Tritium or 3H is a radioactive isotope of hydrogen with a
surface water. There is a noticeable difference between half-life of ~12.3 years (4500 days). 3H is a useful isotope
groundwater and surface water at the time of sampling. for age dating groundwater by providing information of
when the water was last in contact with the atmosphere
The isotopic analysis indicates that the surface water is (precipitation). Natural 3H is created in the upper
enriched in 𝛿18O and 𝛿D isotopes relative to groundwater atmosphere from the cosmic bombardment of nitrogen with
and all waters lie on or close to the meteoric water line neutrons (cosmic rays). 3H then combines with oxygen to
(Figure 11). Fractionation or separation of naturally produce tritiated water (H3HO) and enters the hydrologic
occurring oxygen and hydrogen isotopes in water occurs in cycle through precipitation. Prior to 1950’s, natural 3H
the hydrologic cycle during evaporation and condensation. concentrations in precipitation were less than ~15 TU
Isotopically light water molecules evaporate before the (tritium units). However, 3H concentrations increased
heavy ones. As a result, surface water exposed to significantly (orders of magnitude) during 1950s – 1970s,
evaporation tends to be enriched in heavy isotopes relative when thermonuclear devices were detonated in the
to meteoric water and will have a less negative or more atmosphere. Since then, most of the 3H produced by
positive 18O/16O ratio. thermonuclear testing has decayed or been absorbed by
oceans and present concentrations have returned to
“normal” (pre-1950’s).

Tritium in surface water ranged between 5.4 and 8.4 TU

indicative of modern water from recent precipitation.
Concentrations of tritium in groundwater ranged from <0.8
to 1.4 TU. Largest concentration of tritium in groundwater
(1.4 TU) was observed in well Davis 1A, located adjacent
to the creek, and suggests that mixing between recent
surface water and older groundwater is occurring at a local
level.

Figure 10. Piper plot of water quality data from monitoring

locations.

The meteoric water line (Fritz et. al., 1987) represents

natural seasonal variations of stable isotopes in
precipitation for the study area. It is used herein as baseline
to determine sources of groundwater recharge, which
could be from direct precipitation, surface water, or a
combination of both. Data points that plot on the meteoric
water line are indicative of water sourced directly from
Figure 11. Relationship between 𝛿18O and 𝛿D stable
precipitation with little or no evaporation. Any deviation to
isotope ratios for groundwater and surface water samples.
the right of the line will signify water that has been
Smaller points are data from other locations in the province
subjected to evaporation or mixing with water impacted by
for relative comparison.
evaporation. The 𝛿18O values in groundwater ranged from
-20.22 to -21.86 and 𝛿D of -154.8 to -166.7. It shows that
similarity exists between groundwater precipitation. Values
5 CONCLUSION
of stable isotopes from surface water shows that surface
water is enriched in 𝛿18O (-17.52 to -18.63) and 𝛿D (-136.1
The study conducted by WSA improved understanding of
to -143.8). This is to be expected since surface water is groundwater contribution to the ecosystem at Davis Creek
subjected to varying degrees of evaporation. However, in
and demonstrated the value of integrated approach
the case of water from Davis Creek, stable isotopes are
towards groundwater management. There is a better
relatively close to the meteoric line and the enrichment is
understanding of the groundwater contribution to the
limited when compared to isotopic values from other baseflow of the deeper pools in the creek that sustain
surface water bodies across the province. This suggests a
habitat of Mountain Suckers during summer periods of low
strong influence of groundwater or a major precipitation
precipitation and creek flow.
event just prior to sampling.
Water level, temperature and water quality data were Committee on the Status of Endangered Wildlife in
collected from four monitoring wells and the creek. These Canada.
data suggest the occurrence of complex and dynamic Feinstein, D. T., & Reeves, H. W. (2010). Regional
interaction of surface and groundwater. The shallow Groundwater-Flow Model of the Lake Michigan Basin in
groundwater is contributing to baseflow conditions within Support of Great Lakes Basin Water Availability and
Davis Creek. However, baseflow contribution is likely Use Studies Part II. U.S. Geological Survey, (Scientific
variable along the creek due to heterogeneity resulting Investigations Report 2010-5109), 119–248.
from local geological conditions. These results highlight the Fleckenstein, J. H., Krause, S., Hannah, D. M., & Boano,
complexity and importance of the flow dynamics occurring F. (2010). Groundwater-surface water interactions:
in this area in supporting the identified SAR. Further New methods and models to improve understanding of
monitoring and data collection will continue to better define processes and dynamics. Advances in Water
local flow dynamics and the relationship between Resources, 33(11), 1291–1295.
groundwater and Davis Creek. Fritz, P., Drimmie, R. J., Frape, S. K. and O’Shea, K.
1987. The isotopic composition of precipitation and
Given the growing recognition of groundwater contribution groundwater in Canada. In: Isotope Techniques in
to surface waters dynamics, the groundwater flow paths Water Resources Development. International Atomic
need to be better defined for greater reliability to inform Energy Agency, Vienna, 539–550.
process water management of fish habitats to temperature Hare, D., K., Briggs, M.A., Rosenberry, D.O., Boutt, D.F.,
varied climates. Further collection of data and installation and Lane, J.W. 2015. A comparison of thermal infrared
of seepage meters, lidar, and stream gauging instruments to fiber-optic distributed temperature sensing for
have the potential to improve understanding of evaluation of groundwater discharge to surface water.
groundwater contribution that sustains the Mountain Journal of Hydrology. 420-421:142-158.
Sucker habitat. Lidar data would further gain a better Kurylyk, B.L., MacQuarrie, K.T.B., and Voss, C.I. 2014.
understanding of the stream incision into the shallow Climate change impacts on the temperature and
geological environment and in estimating the alluvial magnitude of groundwater discharge from shallow,
aquifer unsaturated thickness. Seepage meters and unconfined aquifers. Water Resources Research.
hydrometric monitoring station will provide the stage and 50:3253-3274.
discharge of the creek throughout the year. Coupled with Lewandowski, J., Meinikmann, K., and Krause, S. (2020).
the generation of a numerical flow model, the data can also Groundwater-surface water interactions: Recent
be used for calibration of this area, which would incorporate advances and interdisciplinary challenges. Water.
both steady-state and transient analysis of the system. 12:296.
Such a model could then be used to predict changes in Li, Ch., Ye, C., Li, Jx., Wei, W., Zheng, Y., Kong, M., and
creek and groundwater levels that may result from Wang, H. 2020. Impact of spring freshet flooding and
changing climate conditions or managed water use. summer rainfall flooding on the water quality of an
alpine barrier lake. Environmental Science Europe 32,
Acknowledgement 57.
Lo, K., and Melnik, A. 2017. Water chemistry and natural
This work is partially supported by the Canada Nature Fund tracer analysis. Water Security Agency. 20 p.
for Aquatic Species at Risk (CNFASAR). Maathuis, H., and Simpson, M. 2007. Groundwater
Resources of the Cypress Lake (72F) area,
Saskatchewan. SRC Publication 11974-1E07. 37 p.
6 REFERENCES Miranda, L.E., Coppola,G., and Boxrucker, J. 2020.
Reservoir fish habitats: A perspective on coping with
Alley, W. M. (2007). Another water budget myth: The climate change. Reviews in Fisheries Science and
significance of recoverable ground water in storage. Aquaculture. 28:4, 478-498.
Ground Water, 45(3), 251. Pavlovskii, I., Hayashi, M., and Itenfisu, D. 2019. Midwinter
Briggs, M.A., Lane, J., Snyder, C.D., White, E.A., Johnson, melts in the Canadian prairies: energy balance and
Z.C., Nelms, D.L., and Hitt, N.P. 2018. Shallow bedrock hydrological effects. Hydrology and Earth System
limits groundwater seepage-based headwater climate Sciences. 23:1867-1883.
refugia. Limnologica. 68:142-156. Sanz, D., Castaño, S., Cassiraga, E., Sahuquillo, A.,
Clark, I.D., and Fritz, P. 1997. Environmental Isotopes in Gómez-Alday, J. J., Peña, S., & Calera, A. (2011).
Hydrogeology, CRC Press/Lewis Publishers, Boca Modeling aquifer-river interactions under the influence
Raton, FL, 328 pp. of groundwater abstraction in the Mancha Oriental
Chu, C., Jones, N.E., Mandrak, E., Piggot, A.R., and System (SE Spain). Hydrogeology Journal, 19(2), 475–
Minns, C.K. 2008. The influence of air temperature, 487.
groundwater discharge, and climate change on the Winter, T. C., Judson W. Harvey, Franke, O. L., & Alley, W.
thermal diversity of stream fishes in southern Ontario M. (1999). Ground water and surface water: A single
watersheds. Canadian Journal of Fisheries and Aquatic resource, USGS Circular 1139, 79 p.
Sciences. 65(2): 297-308. Zhou, Y. (2009). A critical review of groundwater budget
COSEWIC. 2010. Assessment and Status Report on the myth, safe yield and sustainability. Journal of
Mountain Sucker (Catostomus platyrhynchus). Hydrology, 370(1–4), 207–213.
Saskatchewan-Nelson River Populations in Canada.
 Monday, October 2, 2023

REDI (Special Session)

Bridging the Data Gap for a more Equitable,
Diverse and Inclusive Future in
Geotechnical Engineering: Survey Design
and Ethical Considerations
Kaitlyn Diederichs1, Jack Park2,3, and D. Jean Hutchinson2
1 Department of Environmental Studies – University of Victoria,

Victoria, BC, Canada

2 Department of Geological Sciences and Geological Engineering –

Queen’s University, Kingston, ON, Canada

3 BGC Engineering Inc., Calgary, AB, Canada

ABSTRACT
As the benefits of increased Equity, Diversity, and Inclusivity (EDI) in the geotechnical engineering field become more
apparent, organizations need to find ways to collect representative data from their members/employees, most often through
the use of surveys. The combined demographic and experiential data collected within these surveys provides valuable
information to enable organizations to implement changes that promote EDI efforts, attracting candidates, enhancing
performance and improving retention for a diverse range of employees. To maximize the value of the information collected,
surveys must be designed such that the right information is being collected in the right way with the right intentions. This
paper introduces the motivation for EDI initiatives, guidelines for the implementation of a successful EDI cycle, the technical
foundations of survey design and specific considerations that should be made when developing an EDI survey.

RÉSUMÉ
Alors que les avantages des initiatives d’équité, de diversité et d’inclusion (EDI) dans le domaine de l’ingénierie
géotechnique deviennent plus apparents, les organisations doivent trouver moyens de collecter des données
représentatives auprès de leurs membres/employés, le plus souvent par les biais d’enquêtes. Les données
démographiques et expérientielles combinées recueillies dans le cadre de ces enquêtes fournissent des informations
précieuses pour permettre aux organisations de mettre en œuvre des changements qui favorisent les efforts EDI,
améliorant les performances, les expériences et la rétention des employés pour un large éventail d'employés. Pour
maximiser la valeur des informations collectées, les enquêtes doivent être conçues de manière que les bonnes
informations soient collectées de la bonne manière avec les bonnes intentions. Cet article présente la motivation des
initiatives EDI, les lignes directrices pour la mise en œuvre d'un cycle EDI réussi, les fondements techniques de la
conception de l'enquête et les considérations spécifiques à prendre en compte lors de l'élaboration d'une enquête EDI.

1 INTRODUCTION 1.1 Principles of EDI

Increased Equity, Diversity, and Inclusivity (EDI) leads Equity, diversity, and inclusivity (also referred to as
to a broader range of perspectives, increased creativity and inclusion) are values that are closely connected within
collaboration and a larger pool of potential employees for a organizations, including engineering corporations,
company to hire from (National Engineering Month, 2023). government agencies and academic institutions. These
As these benefits become more apparent in the elements of an organization’s culture need to be prioritized
geotechnical engineering field, organizations need to find in order to ensure that employees are empowered and feel
ways to collect representative data from their employees to like part of the team, without discrimination.
establish a baseline of its status of EDI indicators, develop EDI initiatives have existed since the 1960s when they
strategies to improve the employee experience, and then were initiated during the civil rights movement in the United
evaluate progress. This is most often done by distributing States (Equal Justice Initiative, 2021). In recent years, this
confidential and anonymous surveys asking a variety of work has been reignited due to ongoing issues related to
questions about demographics, job satisfaction, workplace EDI and its implementation within the workplace. With
experiences and workplace culture. These provide more organizations taking on various initiatives to engage
valuable information to enable organizations to identify and in EDI-related advancements, including formulating and
implement changes that promote EDI efforts, thereby administering surveys, it is important to have a common
improving employee performance and retention for all understanding and definition of the terminologies that guide
employees. To maximize the value of the information this practice. The following are the definitions defined by
collected, surveys must be thoughtfully designed such that the Government of Canada (n.d.):
the right information is being collected in the right way with
the right intentions. • Equity: “The principle of considering people’s
unique experiences and differing situations, and
ensuring they have access to the resources and
 opportunities that are necessary for them to attain professional activities such as field work,
just outcomes. Equity aims to eliminate disparities conferences and meetings;
and disproportions that are rooted in historical and • Increased awareness of appearance and
contemporary injustices and oppression.” clothing;
• Diversity: “The variety of identities found within • Concern for physical safety;
an organization, group or society. Diversity is • Loss of confidence and/or lowered work
expressed through factors such as culture, productivity.
ethnicity, religion, sex, gender, sexual orientation,
age, language, education, ability, family status or Beniest et al. (2023) provided some early analysis of
socioeconomic status.” data collected from an ongoing survey of European
• Inclusion: “The practice of using proactive geoscientists which indicates that almost half of the
measures to create an environment where people respondents have “sometimes experienced a)
feel welcomed, respected, and valued, and to disrespectful comments or actions, b) people questioning
foster a sense of belonging and engagement. This the respondents’ professional expertise, and c) sexist or
practice involves changing the environment by racist language in their workplace”, with the result that 40%
removing barriers so that each person has equal have considered leaving their employer or changing
access to opportunities and resources and can careers. These findings point to the need to ensure that
achieve their full potential.” organizations take actions and make policy changes to
improve the culture surrounding EDI at all levels.
There are numerous other terminologies that define
and are related to EDI. For a more extensive list of 1.3 EDI Success Cycle
terminology and definitions, refer to the Guide on Equity,
Diversity and Inclusion Terminology under the Resources The success of EDI efforts within a given organization
section of this paper. This is an emerging space and as relies on buy-in at all levels. From the upper levels of
these topics are explored, the terminologies and their decision makers to daily operational staff, this involves
definitions will also change. creating strong principles and committing to making the
organization a place of trust and safety for everyone.
1.2 Motivation for EDI Implementing changes that provide a positive impact
on the organization cannot happen without a sound plan.
There are two value systems that drive the promotion An EDI success cycle (Figure 1) provides a guide for
of EDI in the workplace (Byrd and Sparkman, 2022). The deciding how to plan and implement changes while
moral case (or social justice) for EDI states that equality for effectively communicating with and engaging all
all (and necessarily, diversity and inclusion) has value in stakeholders.
and of itself, and the promotion of these ideals in the In any part of the cycle, the process may be difficult and
workplace is a moral obligation. There is also an increasing even triggering for individuals as it may bring unpleasant or
societal pressure on organizations to take a stance on harmful experiences to the surface. It is important to have
social issues, as current and potential constituents want to in place a support system that can be offered to the
align themselves with organizations that connect to their survey’s participants and to ensure safety for the
values. participants with a disclosure of how the data will be
The business case states that a diverse workforce handled and used.
provides value in the form of creativity, innovation, and Although this paper focuses on this cycle's survey
different perspectives. We would also argue that increasing planning and development, all parts are equally important.
the diversity of the workforce lends resilience and The next subsections explain the importance of each
adaptability to an organization. This rationale for the values component in broad terms.
of EDI is tied to the economic goals of a company and so
may be used to justify the expense of EDI efforts 1.3.1 Training and Education
(educational materials and training opportunities,
surveying expenses and the cost of individual support). Training and education span the entire EDI success
Byrd and Sparkman (2022) state that neither of these cycle. This is a critical component in engaging all
values can successfully exist in an organization without the stakeholders within the organization and providing a
other, despite the business case often being the only common and mutual understanding of the terminologies,
justification used. establishing the purpose of the survey, and setting
When a diverse workforce is not supported by and expectations for implementing changes to the organization.
included in the higher levels of an organization, a number The individual understanding of topics related to EDI
of potential outcomes arise. The categories identified by may vary greatly from one to another. Having a common
Beniest et al. (2023) include: ground where individuals can learn in the way that suits
• Leaving current employment and/or considering a them best and express their thoughts and opinions is
career change; crucial in developing a collective desire to make impactful
• Avoiding colleagues and/or supervisors; changes related to EDI issues in an organization. Training
• Declining professional opportunities for career and education opportunities are also great avenues to
development or advancement and/or skipping disseminate information and communicate with the
stakeholders. Along with the critical need for sponsorship
 Survey Planning &
Development

Ongoing: Survey Design &

EDI Action Plan Development Research Education
Administration
Training Dialogue

Survey Interpretation &

Dissemination of Results

Figure 1: Considerations for EDI Success Cycles.

at the leadership level, this continued learning should be within the process to provide deeper knowledge and
encouraged for everyone, regardless of their position and perspective. Employees for whom EDI is not a job focus,
responsibilities. especially employees of equity-deserving groups, should
Given the importance of engagement and participation not be asked to perform this educational labour as it
of all levels of stakeholders, it is advised that sufficient time unfairly places the burden of resolving workplace inequity
and resources are formally provided for individuals to on those most affected. Whether it is a mandatory training
complete the surveys and training required (i.e., having the or a request for voluntary participation, desired
employees’ time be compensated or setting aside a participation rates should also be communicated to the
designated class time for the students). This reduces the constituents.
extrinsic burden of those who are being asked to
participate (Qualtrics, 2019). It is also advised that experts
who are familiar with topics related to EDI are consulted
1.3.2 Survey Planning & Development 1.3.5 EDI Action Plan Development

The motivation to conduct a survey – in various forms, The EDI action plan development stage is a
such as an anonymous form, confidential interviews, or a (re)evaluation stage that allows the organization to reflect
focus group – and collect information on an organization’s on their current practices, culture, and alignment with how
members/employees may vary from one organization to its members/employees are feeling. Does the demographic
another. However, in general, demographic and profile of the organization match the available talent pool
experience information give an insight into the makeup and and the industry target? Is there a similar feeling of
the current state of the organization’s environment and can belonging and being valued amongst all employees?
uncover issues that may not be apparent. The initial These are the questions that can help develop what comes
development of the survey takes careful consideration, and next.
the developer must be mindful of the sensitivity of what is Action items that arise from the survey findings will vary
being asked of the participants to share. The design of EDI- from one organization to another. A few of these may be:
related surveys should be considered a skilled task, and a • Time-based goals: what is achievable within the
subject matter expert be consulted during the process of short (less than 12 months), medium, and long
survey planning and development. Further considerations (five or more years) term.
and technicalities of survey questions are elaborated in • Benchmarks for the success of the initiatives.
Sections 2 and 3. • A champion that will be responsible for the
delivery of the action.

1.3.3 Survey Administration During the process of carrying out the action items and
implementing changes, it is important to check in with the
When administering the survey, it is important to constituents to make sure the changes are having a
preface and give a clear explanation of the motivation positive impact and to continue engagement with the work
behind the survey, the importance of participation, and how overall. One of the ways this can be accomplished is by re-
the information will be collected and used. Highlighting surveying. Though there are no specific guidelines on the
anonymity, or in the case of interviews and focus groups, frequency, it should be after sufficient time has passed
confidentiality is crucial and will allow the participants’ since the implementation of any changes and when
psychological safety to openly share their personal meaningful engagement can be made with employees (i.e.,
experiences. not conducting the survey so often that the participants
become disinclined to participate).
1.3.4 Survey Interpretation & Dissemination of Results

The interpretation of the survey comes in two folds: 2 CONSIDERATIONS FOR QUESTIONS -
statistics and relationships. Statistical data represent the DEMOGRAPHICS, PROVING THOUGHTFULNESS
demographic questions being asked. This is typically
presented in figures and tables of percentage responding. 2.1 Pre-Survey Testing
It is important that any figures that are presented have
statistical significance in order to maintain the anonymity of When planning to implement a survey, researchers
the participants. should take care to create an effective survey through
Relationship data requires associating the thorough research and by following best practices but
demographic data to one another, or to the experiential should also make time to test the survey amongst
responses. This process requires careful consideration themselves, the EDI Committee members, or a small test
and may take considerable time to do. Both interpretation group of employees before distributing the survey widely.
processes provide the makeup of the organization and give This will help to catch any issues in wording and to have a
insight into the individual demographic groups’ preview of the data collected in case there are missing data
experiences. This information will guide future points that can be added before the official survey. It can
development and implementation of changes within the give insights into the potential for respondents to
organization. experience fatigue during the survey. An accurate time
Dissemination of the survey results and the findings estimate for the completion of the survey determined
should be widely shared within the organization, as it through testing can help alleviate frustration and promote
follows through with the commitment to the respondents on completion while helping manage project costs. An
how the collected information was used. In addition to a accelerated iterative design process such as this can
formal document, the information should also be presented increase the utility of the survey results without increasing
in a collective setting that allows for follow-up questions to the burden on the participants.
be asked. Focus groups to discuss specific aspects of the
results can be a productive way of communicating 2.2 General Considerations: Question
information and receiving feedback that can be analyzed Phrasing and Language Precision
later using qualitative thematic analysis.
When wording the questions for a survey, the
researcher should take care to use language that is easily
understood by the target audience (i.e., the intended
participants of the survey), while also avoiding ambiguous EDI efforts would make major differences to the quality of
wording or regional or colloquial terminology that may not their work experiences. This does not provide productive
be familiar to the entire pool of participants (Tourangeau et information. Option 2 asks a similar question but reinforces
al., 2000). General and broadly distributed surveys should company commitments to the initiative and provides an
not be exclusionary. It may be appropriate to use highly avenue for researchers to discover a need for training or
technical language if the intended participants are a other educational materials.
particular group of subject experts, and that is a decision
the researcher can make with that prior knowledge. 2.3 Demographic Questions
Unsuitable language may cause participants to avoid
certain questions or not complete the survey overall due to A demographic question asks the respondent to detail
frustration. some aspect of their identity. This might include age, race,
In the context of EDI-related surveys, it is also of or gender among other variables (Patten, 2014). These
particular importance to display the care you are taking to questions most directly respond to the “diversity” aspect of
develop the survey through your word choice. Using the the EDI umbrella. The information collected through this
correct, community-accepted terminology to refer to type of survey can also be used to create a profile of the
minority groups is of the utmost importance to ensure both overall organization, the respondents of the survey in
the comfort of the participants and their continued general, or provide context for a cluster of responses.
participation in the survey. For example, when discussing These questions are very important in EDI surveys in
the gender of participants, using words such as woman, particular, as they can illuminate patterns in the identities
man, or non-binary person rather than male and female is of respondents who share similar positive or negative
more inclusive. It is important to acknowledge that trans experiences and identify areas of improvement in the
people are already included within these gender inclusivity of an organization’s culture. However, for equity-
categories. If you would like information specifically about deserving individuals, these questions can target sensitive
the experiences of trans people within your company, it is topics, particularly if a list of identifiers is long, but not fully
best practice to ask that in a second question surrounding inclusive. As a result, demographic questions must be
that topic specifically (e.g., “Do you have lived experience approached with care to avoid discouraging participants
as a trans person?”) When asking questions around from responding to these questions, or to the survey overall
disability and accessibility, it is important to use words (Patten, 2014). Researchers need to take care to word
accepted by the community and do not attempt to use these questions using currently acceptable language in a
euphemisms (e.g. “special needs”). When in doubt, consult way that displays thoughtfulness. It is important with
experts with lived experience in equity-deserving groups. questions that may request responses along a spectrum of
As mentioned in Section 1.3, it is important to provide social desirability or respectability that a researcher (and
training and education to those whom the survey is their wording of questions) does not cast judgment on any
intended for prior to distributing the survey and where particular identity or answer (Tourangeau et al., 2000).
possible, to include definitions and examples within each According to Patten (2014), it is best practice to limit the
question to prime the participants for answering without demographic information to only what is needed to satisfy
any confusion or ambiguity. As the knowledge of EDI the goals of the survey. In EDI research, there are often
advances and subsequent iterations of surveys are assumptions made about what information may be
created, it is important to note that the language and important to collect, but this may result in missing hidden
terminology used may evolve. There may also be or previously unidentified patterns or intersections in
geographical differences in communication and experiences in a given organization. A potential solution to
understanding that need to be considered depending on this is reducing the granularity of the responses requested.
the makeup of your constituents (Engineers Information can still be collected about a participant’s
Canada, 2023). inclusion in an equity-deserving group, but fewer options
A specific consideration for EDI-related surveys is to reduce the intrusiveness of the question. For example,
ensure that the questions you are asking support the instead of asking a participant to give a specific description
mandate of the survey. It is important to not undermine the of their sexual orientation, it may be more appropriate to
impact of the survey and initiatives within the context of the ask whether they identify as belonging to the LGBTQ2SIA+
survey. An example of this is asking questions related to community, as a whole. To further limit the potential for
employee perceptions of EDI initiatives: discomfort to participants in answering questions of this
type, some questions may warrant a “prefer not to say”
• Option1: option, as it allows respondents to skip questions they are
I think EDI surveys are important. not comfortable answering, even with guarantees of
[Yes, No, etc.] anonymity. In addition, if questions are too detailed and
• Option 2: specific, it may be possible to identify an individual from the
I understand why EDI surveys are important. anonymized results, especially in smaller organizations. An
[Yes, No, etc.] example of this may also include requesting information
about a participant’s age. A question that asks for
In the above examples, Option 1 has the potential for a responses within range categories would be much less
majority demographic who have not experienced likely to single out participants than asking for exact
marginalization in the workplace may answer in the numerical values (you may only have one employee born
negative, overwhelming the responses of people for whom in a given year). Another approach is to collect granular
data, but not analyze the groups below a certain threshold. time for respondents to give and for researchers to process
Where appropriate, groups can be re-aggregated for this information. Instead, it might be best to break the
analysis and reporting purposes. In both approaches, it is original question into a series of simplified questions that
recommended that the direct responses (raw data) be only revolve around the original theme (Schaeffer & Presser,
accessible to an external third party to maintain the 2003). This specific example may be reformatted into the
anonymity of the participants. This should also be following statements and responses (among others):
communicated to the participants prior to and during the a) In the past year, I have experienced harassment or
survey, and when the analyzed information is presented. discrimination at work. [Never, Rarely, Sometimes,
Demographic information can also be used to Often, Almost constantly]
determine the relevance of a respondent’s answer. For b) In the past year, I have observed harassment or
example, a person with a disability will likely have more discrimination happening to others at work. [Never,
useful insights into the accessibility of a workplace or the Rarely, Sometimes, Often, Almost constantly]
organization’s attitude towards disability. This type of c) In my time at this organization, instances of
information can be used after the survey has been harassment or discrimination have become less
completed to filter responses, or during the survey to frequent and less severe. [Strongly agree,
display different questions to different participants. Somewhat agree, Neither agree nor disagree,
Somewhat disagree, Strongly disagree]
2.4 Experiential Questions d) I have experienced harassment from the following
categories of colleagues (check all that apply):
Often in EDI surveys, the workplace experiences of the [Managers, Peers, Subordinates, Individuals,
participants, both positive and negative, are of interest to Groups, Other: (type)]
paint a picture of the impacts of organizational culture on
individuals. The information to be gathered from this For questions of this importance and complexity, providing
section of the survey is crucial to gauging the state of equity a place for respondents to indicate that they would like to
and inclusivity within the organization. However, follow up with researchers further (either by allowing for
requesting participants to think about potential negative long answers or opportunities for interviews) may be
experiences may risk furthering the harm of that appropriate.
experience and so should be approached thoughtfully and
with care to provide opportunities to skip these questions if
necessary and to provide resources for participants who 3 TECHNICALITIES OF SURVEY QUESTIONS
would like support in processing these experiences.
When asking about experiences in the organization, it This section provides descriptions of several common
is important for the researcher to consider exactly what survey question formats and examples of appropriate
they would like to learn. It is possible to be too broad and usages and formatting.
miss critical details that could shape future training or
initiatives. A sample of a question that may produce this 3.1 Multiple-Choice Questions
outcome is as follows: “Have you ever experienced or
witnessed harassment at this organization?” Multiple-choice questions allow the researcher to
This question has a few issues. It does not distinguish provide the participant with a finite set of options to select
experiences a respondent may have had as the target of from. Multiple-choice questions can be thought of as a tool
harassment from events they may have witnessed as a for grouping participants based on prior knowledge about
bystander. The results of this question, with demographic the population or the context in which you are asking the
information, might allow researchers to determine (a) if question. When choosing the potential responses for a
there are particular identities or intersections of identities multiple-choice question it is important to balance the goals
who experience discrimination or harassment in the of providing an exhaustive list of options with minimizing
workplace at higher rates and (b) if there are differences in confusion or frustration. It is not ideal to have a huge list of
the recognition of discrimination or harassment between options as it can both confuse participants (too many
demographic groups. In the case of (a), an organization or options may create non-meaningful distinctions if
employer should develop support systems for equity- participants are forced to choose between similar options)
deserving people. Conclusions from both (a) and (b) will and lead to fatigue within the survey (risking non-
help to identify future educational initiatives to educate completion). A good rule of thumb is to aim for
employees on behaviour that is not appropriate in the approximately five options to avoid overtaxing the working
organization and how to respond to it, and to identify it as memory capacity of respondents (Mattarella-Micke and
an ally. Secondly, there are no time boundaries on the Beilock, 2012). This can be achieved by simplifying the
responses to this question. This means that a researcher choices into broader categories or by creating a non-
has no way of determining if an event occurred recently exhaustive list with an option for participants to write in their
(potentially after the implementation of an EDI intervention own “other” responses. In some cases, such as
of interest) or in the relatively distant past, nor how often demographic questions about ethnicity or religion, it can be
the person has encountered such experiences. Simply harmful to create cases where some groups are excluded
asking “Tell me about your experiences relating to or aggregated, and so well-researched care should be
harassment at this organization” as an open-ended taken to make sure the question does not alienate any
question might provide the most details but will take a lot of respondents.
 Researchers can allow for several selections (multiple- numerical. For these questions, it is important to make sure
answer) within a set or restrict the participant to making a that the scales can be easily understood by the survey
single choice (single-answer). These two options for respondents, and that the categories give meaningful and
responses give the researcher flexibility to ensure that the consistent results. The typical variables used in these
question collects the most useful information from the questions (favourability, relevance, importance, frequency,
participants, but also must be used correctly to avoid value) are subjective and may require baselining to achieve
frustrating participants or resulting in missing information. useful or accurate results. Care should be taken when
For this, it is important to have enough background determining the granularity of the scale and when
knowledge of the question being asked to know if the describing the question and the meaning of ranges on the
potential responses are mutually exclusive or not. If there scale.
is no possibility for a respondent to need to select more Scales can be dichotomous or include huge ranges of
than one answer to accurately describe their response, a values. These two ends of the spectrum have advantages
single-answer format would be appropriate and can and disadvantages that should be balanced to optimize the
prevent errors in response. Otherwise, a multiple-answer question. Dichotomous scales include opposite options (in
format is necessary. effect, True or False) and so create very little ambiguity.
An example of a reasonable single-answer multiple- Participants should have a very clear idea of what each
choice question may be as follows, where each choice is response indicates, and responses will be very accurate.
clear and there is no overlap among categories or This extreme of a simplified scale can erase the precision
unaddressed gaps: and nuance of a scale. There is significantly more
How long have you been employed at this company? judgement required on the part of the participant faced with
(a) Less than 1 year a percentage than a smaller range with larger divisions.
(b) 1-5 years The accuracy and reliability of a scale of this large range
(c) 6-10 years may be reduced in the name of assumed precision. To
(d) More than 10 years strike a balance between these two, scales with 5-9 bins
should be used (Qualtrics, 2019). If possible, try to keep
While there are many examples where multiple-answer scales consistent between similar questions to avoid
questions are appropriate, often questions surrounding increasing the participant’s mental load of interpreting the
identity or experience have the potential for more complex scales.
answers than a single choice can represent. It may also be
possible to turn a multiple-answer question into a single- 3.3 Likert-Type Scale Questions
answer by consolidating potential responses into a larger
category if a lesser amount of granularity in response is A Likert-type scale question aims to evaluate a
acceptable. For example, the following question requires a participant’s feelings on a given statement or topic using
multiple-answer format to allow for intersections to be descriptive categories along a spectrum. This type of
represented: question in EDI surveys is commonly used to assess
What is your sexual orientation? participants’ perceptions of the impact of an aspect of
(a) Heterosexual or Straight organizational culture or responses to a given statement.
(b) Homosexual or Gay/Lesbian Generally, the options are defined as follows:
(c) Bisexual • Strongly Disagree
(d) Asexual • Somewhat Disagree
(e) Other (written response) • Neutral
(f) Prefer not to say • Somewhat Agree
• Strongly Agree
By changing the question to be more inclusive but less
detailed, you may capture just as much useful data from This question format can also be used to obtain
the response for your survey purpose but can simplify the estimates of perceived frequency by asking participants to
question to allow for a single answer. The following tell the researcher how often an event occurs, with the
example question may be appropriate, but this type of following scale:
decision requires some thought into the current culture of • Almost Never
your organization and how you will be using the data. • Rarely
Do you identify as a member of LGBTQ2SIA+
• Sometimes
communities?
• Often
(a) Yes
(b) No • Almost Always
(c) Prefer not to say
It is best to use words such as “almost” or “strongly” to
3.2 Scale or Rating Questions temper the extremes of your responses in a Likert-Type
scale as participants are unlikely to select answers that
Questions that allow participants to provide a rating, or imply 100% certainty (Patten, 2014), and the responses
an answer along a scale, can quantify perceptions of a may be artificially clustered as a result.
given variable. These can take the form of Likert-Type A potential variation to the scale is the “forced choice”
questions as described in the following section or can be format, which removes the neutral option from the survey
and requires that a participant responds to a question with
a more binary choice. This is inappropriate in questions as together, they provide critical nuances and context to
where there may be a reasonable neutral answer as it may the experiences shared by the participants. There is no set
create a “false truth” result, but appropriate in situations a rule on the number of participants in focus groups, but Glitz
neutral result does not provide sufficient information for (1997) suggests six to ten individuals, plus a moderator.
critical decision-making. If a forced choice format is Like the survey, focus groups and individual interviews
selected, it is good practice to allow participants to skip the should ensure confidentiality and provide a safe and
question to avoid erroneous selection. supportive space for the participants. As such, these
A note in processing Likert-type scale data is that sessions should be facilitated by an external third party
averages are not necessarily meaningful depending on the (hopefully in coordination with the survey facilitator) without
distribution of the data (Sullivan and Artino, 2013). There is the presence of hierarchical presence or power dynamic,
an assumption within an average calculation that such as the organizational management or human
categories are equidistant in value, which may not be the resources. For focus groups, careful consideration should
case for all questions. Standard parametric tests can be be made to create appropriate community groups that
used for comparison (Norman, 2010); however, the provide a safe space for individuals to come and share their
bimodality of responses may provide valuable information experiences (i.e. women, those dealing with mental health,
about the experiences participants are sharing and or Black, Indigenous, or people of colour groups). The
researchers must be careful not to lose the nuances of questions asked should be open-ended to encourage
unexpected distributions. discussion amongst the participants and in line with the
themes presented within the survey.
3.4 Ranking Questions
3.7 Sampling Considerations for
A ranking question asks a participant to organize Representativeness
objects or options according to a given criteria. This might
include favourability, relevance, or importance. Ranking It is possible to collect the most accurate data points in
questions are most useful when there is real-world the world and not be able to draw any representative
relevance to the comparison between options. Again, care conclusions about the organization itself if the sample of
should be taken to make sure that the number of options participants does not contain: (a) enough people to make
strikes a reasonable balance between having a large valid inferences and (b) enough diversity of individuals to
enough number of options to allow for nuance and reflect the population of the organization. To address (a),
variability and few enough options to ensure reliability and the researcher needs to know the ideal confidence interval
avoid fatigue. The data collected from ranking questions to which they would like to report results to. For smaller
are often more reliable and consistent than rating question organizations (<200 employees/members), it is likely that
responses but may be more difficult to interpret and the goal will need to be to collect as many responses as
analyze, particularly when strict choices between options possible, but in very large organizations, a fraction of the
are unimportant. employees (evenly distributed across job classes and
demographics) may be sufficient. When attempting to draw
3.5 Open-Ended Questions conclusions within different demographic groups (b), it
benefits the researcher to think of each group as a sub-
Open-ended questions are invaluable for collecting population to be sampled. From here, it can be inferred that
nuanced stories. There is a plethora of experiences that the same principles for number or proportions of
cannot be accurately captured into multiple-choice or other respondents in each major demographic category should
closed-ended questions. Oftentimes, these are the be the goal.
experiences that can reveal deeper issues with the
implementation of EDI initiatives and protections for
marginalized individuals, but they do require thoughtful 4 RELEVANT QUESTION CATEGORIES IN
implementation and anonymization as individuals may be GEOTECHNICAL ENGINEERING
more readily identifiable from these responses. They also
require more time to appropriately process the data and Geotechnical Engineering field work often presents
find themes between responses. In these types of data challenges for equity-deserving people. Increasing market
collection, depending on volume of responses, qualitative availability of diverse individuals trained in Geotechnical
data analysis software (such as NVIVO) that allows for Engineering subjects, development of more equitable
thematic coding and compilation can aid in finding hiring practices and evolving and more inclusive company
meaningful conclusions. cultures have all resulted in increasingly diverse employee
pools in the Geotechnical Engineering field in Canada.
3.6 Focus Groups and Individual Training (Hutchinson and Diederichs, 2022) and
Interviews development of company policies are an essential
Focus groups or individual interviews are another component of this work. However, Geotechnical
method of collecting information. This technique provides Engineering field work still frequently presents challenges
a qualitative methodology to explore the opinions, for equity-deserving people at all levels of organizations,
knowledge, perceptions, and concerns of individuals and if EDI work is not continually prioritized, these gains in
(Glitz, 1997). Participation in the survey, focus group(s), or diversity will be lost.
individual interviews do not need to be mutually exclusive,
 4.1 Psychological Safety in the Field of time away from home living and working on a remote job
site, and often presents challenges for regular
Fieldwork is common in geotechnical engineering and communication with supervisors, mentors and allies.
may result in individuals working in environments that Therefore, the list of locations given above may be
introduce additional stress as a result of the potentially expanded to include: remote work camps, remote field
remote location or interactions with people from outside of work, and any other location where your employees work.
the organization. Health and safety (H&S) programs have Best company practice should include the development of
become mandatory in almost all situations and ensure that a specific policies and practices to support employees
the organization’s participants can perform their duties working in the field, to inform employees, clients and sub-
without harm. However, most H&S programs refer to consultants of EDI expectations, responsibilities, reporting,
physical aspects and may not explicitly include the development of response plans to rectify the situation in the
psychological safety of the individual, which can be field, and sanctions for non-compliance. This may include
negatively impacted by EDI-related issues. To capture the ensuring that individual employees are not deployed solo,
experiences unique to fieldwork, it may be worth and that they are supported to leave unsafe situations if
considering developing questions that are specific to this there is no possible and timely resolution to issues that
topic in the survey in collaboration with the health and have the potential to create physical and/or mental harm.
safety team within the organization. Some of the EDI survey questions may address
issues related to whether employees feel safe, both
4.2 External Players in the Field physically and mentally, by evaluating some of the
following issues: inappropriate comments; negative
It is possible that an organization can have well- references to physical or mental ability, race or sexual
developed EDI protections and a positive environment orientation; sexual harassment including constant staring,
internally, but when employees work offsite, this may differ. touching or pressure to date someone; suggestive stories
Fieldwork often requires communication and interactions or offensive jokes; exposure to inappropriate materials; and
with other workers, contractors, client representatives, and any other observed issues in the workplace and
the general public. Although it may be difficult to directly environment. Additional questions should be developed to
control the actions of those external players, it is important understand whether employees requiring accommodation
to capture relevant experiences to inform how to better for disabilities and illnesses – physical and mental – are
prepare employees for such situations, and to ensure a supported in these workplaces.
support system is in place proactively.

4.3 Accessibility in the Field 5 HELPFUL RESOURCES

In geotechnical engineering, in both academia and • Guide on Equity, Diversity and Inclusion
industry, the field is often promoted to be better suited for Terminology (Government of Canada, n.d.)
able-bodied individuals. However, not all fieldwork requires • Questionnaire Research: A Practical Guide
full mobility, and the industry should certainly be more (Mildred Patten, 2014)
inclusive. The same can also be applied to those with • “Question and Questionnaire Design” in the
“invisible illnesses”, such as mental illness, chronic, pain, Handbook of Survey Research (Krosnick, Jon A.
and hearing loss. Those who require accessibility and Stanley Presser, 2010)
assistance may either be excluded altogether or face
difficulties in the field. These experiences should be 6 ACKNOWLEDGEMENTS
considered and captured within the survey. Given the
traditional makeup of geotechnical engineering, those We would like to thank BGC Engineering, in particular
individuals experiencing these challenges may be few and Leanne Johnson and Ashley Perkins, for sharing their
far between and so particular care must be taken to ensure valuable insights surrounding the importance of EDI and its
the anonymity of these individuals when soliciting specific nuances in the geotechnical field. We would also
information and sharing the results. like to thank Sarah Hatherly, a Queen’s University PhD
candidate in Geological Sciences for sharing her
4.4 Potential Questions experiences in running EDI-related surveys. Additionally,
we would like to thank Ashlyn Clarry and Cindy Liu for their
As a result of this ongoing challenge, the questions help in sensitivity reading several sections of this paper.
included in an EDI survey related to workplace culture and
safety should poll employee experience in the field
explicitly, in addition to other locations. In the EGU survey
(Beniest et al, 2023), respondents are asked to identify the
location of their experiences as occurring in one or more
of: the “office, common workspace, laboratory, work social
event, classroom, field setting, work-related travel,
conference or professional society event, meeting, other
(please specify)”. Field work often involves working with
employees of other companies, may involve long periods
7 REFERENCES Mattarella-Micke, A., and Beilock, S.L. 2012. Capacity
Limitations of Memory and Learning. In: Seel, N.M.
Beniest, A., Popp, A., Di Chiara, A., Gürer, D., van (eds) Encyclopedia of the Sciences of Learning.
Rijsingen, E., Li, M., and Pieber, S. 2023. The Springer, Boston, MA. https://doi.org/10.1007/978-1-
European Geosciences Community: insights from a 4419-1428-6_603
survey on workplace diversity and climate, EGU National Engineering Month. 2023. Why is Equity, Diversity
General Assembly 2023. Vienna, Austria, 24–28 Apr, & Inclusion (EDI) Important in Engineering? Accessed
EGU23-16491, https://doi.org/10.5194/egusphere- from: https://nemontario.ca/2023/03/03/why-is-equity-
egu23-16491 diversity-inclusion-edi-important-in-engineering/
Byrd, M. Y., and Sparkman, T. E. 2022. Reconciling the Norman, G. 2010. Likert scales, levels of measurement
Business Case and the Social Justice Case for and the “laws” of statistics. Adv in Health Sci Educ. 15.
Diversity: A Model of Human Relations. Human 625-32. https://doi.org/10.1007/s10459-010-9222-y
Resource Development Review, 21(1), 75–100. Patten, M.L. 2014. Questionnaire research: A practical
https://doi.org/10.1177/15344843211072356 guide, Routledge, Glendale, CA, USA.
Engineers Canada. (2023, March). Managing Transitions. Qualtrics LLC. 2019. The Qualtrics handbook of question
Accessed from: design. https://www.qualtrics.com/ebooks-
https://engineerscanada.ca/reports/managing- guides/qualtrics-handbook-of-question-design/
transitions Schaeffer, N.C. and Presser, S. 2003. The science of
Equal Justice Initiative. 2021. Reconstruction in America: asking questions. Annual Review of Sociology. 29(1).
Racial Violence After the Civil War. Accessed from: 65-88.
https://eji.org/report/reconstruction-in-america/ https://doi.org/10.1146/annurev.soc.29.110702.11011
Hutchinson, D.J. and Diederichs, K.D. 2022. Reflections on 2
Resources: Engaging the Future in Geotechnical Sullivan, G.M. and Artino, A.R. Jr. 2013. Analyzing and
Engineering – Steps towards effective gender interpreting data from Likert-type scales. J Grad Med
inclusivity in our profession. GeoCalgary 2022, Educ. 5(4). 541-2. https://doi.org/10.4300/JGME-5-4-
Calgary, Alberta, Canada. 18
Glitz, B. (1997). The focus group technique in library Tourangeau, R. Rips, L.J., and Rasinski, K.A. 2000. The
research: an introduction. Bulletin of the Medical Psychology of Survey Response, Cambridge University
Library Association, 85(4), 385–390. Press, Cambridge, England.
Government of Canada. n.d. Guide on Equity, Diversity https://doi.org/10.1017/CBO9780511819322
and Inclusion Terminology. Accessed from:
https://www.noslangues-
ourlanguages.gc.ca/en/publications/equite-diversite-
inclusion-equity-diversity-inclusion-eng
Tuesday, October 3, 2023

GEOHAZARDS III
Cost-effective dGNSS system for landslide
monitoring: system description,
deployment, and initial findings at the Chin
Coulee Landslide
Ingrid Berru & Renato Macciotta
Department of Civil and Environmental Engineering – University of
Alberta, Edmonton, AB, Canada
Chris Gräpel & Jorge Rodriguez
Klohn Crippen Berger, Edmonton, AB, Canada
Roger Skirrow & Kristen Tappenden
Ministry of Transportation and Economic Corridors, Edmonton, AB, Canada

ABSTRACT
In recent years, the use of differential global navigation satellite systems (dGNSS) for monitoring landslide displacements
has increased due to their high accuracy and low cost. The Sparkfun dGNSS system, which is commonly used in surveying,
is one such systems. It has a high-frequency measurement of displacements at 60-second intervals, achieves
manufacturer horizontal and vertical accuracy of 14 mm and 10 mm, respectively, and is reasonably priced (capital cost
around $1920 USD per unit). The evaluation of the viability of the Sparkfun dGNSS as a cost-effective landslide monitoring
system is presented in this paper, including its deployment at the Chin Coulee landslide located in southern Alberta, initial
findings, and comparison with commercial off-the-shelf Geocubes, another dGNSS, installed in the area. During a one-
month testing period, the standard deviations achieved by the Sparkfun dGNSS were less than the accuracy threshold set
by the manufacturers and has provided similar displacement trends compared to the Geocubes.

RÉSUMÉ
Ces dernières années, l'utilisation des dGNSS pour surveiller les déplacements de glissement de terrain a augmenté en
raison de leur précision et de leur abordabilité. Le système Sparkfun dGNSS est un système couramment utilisé qui
effectue des mesures de déplacement à haute fréquence avec une précision horizontale et verticale de 14 mm et 10 mm
respectivement, et à un prix raisonnable. La viabilité du système Sparkfun dGNSS en tant que système de surveillance de
glissement de terrain rentable est évaluée dans cet article, y compris son déploiement sur le glissement de terrain de Chin
Coulee, les résultats initiaux et la comparaison avec les Geocubes, un autre dGNSS installé dans la région. Au cours
d'une période d'essai d'un mois, le Sparkfun dGNSS a atteint des écarts types inférieurs au seuil de précision établi par le
fabricant et a fourni des tendances de déplacement similaires par rapport aux Geocubes.

1 INTRODUCTION denser spatial coverage of landslides with relatively high

accuracy (Rodriguez et al. 2021).
In Canada landslides pose a significant risk to people,
infrastructure, and natural resources, with damages 1.1 Differential GNSS System
costing up to $400 million annually (NRC. 2018; Deane et
al. 2019), making landslide monitoring crucial in hazard A cost-effective option for monitoring slope movements is
analysis and risk assessment to detect early warning signs the Global Navigation Satellite System (GNSS). GNSS
of slope collapse and mitigate risks. monitoring of slope movements is being increasingly used
In-situ instruments such as ShapeAccelArrays (SAAs), in engineering practice due to the high frequency of data
inclinometers (SIs), and piezometers are commonly used collection, relatively high degree of accuracy, and gradually
to monitor landslides. They require borehole drilling for decreasing cost over the last few years. Standalone GNSS
installation, which can be costly and limits their use to receivers cannot achieve millimeter accuracy for
projects with sufficient budgets, or conversely, reduces the monitoring landslide displacements. GNSS receivers use
number of monitoring points (Smethurst et al. 2017; the real-time kinematic technique (RTK) to improve the
Rodriguez et al. 2021). Additionally, accessing drill pad relative accuracy of the network, resulting in a system
locations may require construction of access roads and referred to as a differential GNSS system (dGNSS).
associated tree clearing, consideration of wildlife restricted The dGNSS system includes at least two GNSS units,
activity periods, accessing privately owned land, and one fixed in a stable area with known coordinates (the
potential destabilization of sensitive slopes with earthworks base), and another (the rover) whose location will be
or changes in surface water drainage. Such project estimated. The base station compares its fixed location
constraints may require engineers to consider cost- with the location provided by satellites, which have errors
effective, low-impact monitoring instruments that provide caused by atmospheric layers, multipath errors resulting
from signal reflections off nearby objects, internal noise Geocubes provide meaningful results for analyzing
within the device, and satellite orbits (Rodriguez et al., landslide displacements. For example, at the Chin Coulee
2021). The base station determines these errors and landslide, a horizontal daily standard deviation of 0 mm-
corrections and sends them to the rover, which then 5 mm was calculated for a 24-hour moving average, and at
calculates its location with improved accuracy. Currently the 10-Mile landslide in British Columbia, the horizontal
dGNSS can achieve accuracies ranging from 1 mm to daily standard deviation for the 24-hour moving average
20 mm (Takasu and Yasuda, 2009, Deane et al. 2019); one ranged from 1 mm-2 mm, consistent with the specifications
such example are Geocubes, a single-frequency dGNSS provided by Ophelia and Rodriguez J. (2021) estimations.
network that collect data every 60 s and has an (Deane et al. 2019, Rodriguez et al. 2018, Rodriguez J.
approximately capital cost of US$4,500 per unit (Rodriguez 2021). However, other cost-effective alternatives are now
J. 2021). Geocubes are designed to achieve an accuracy available in the market that offer similar capabilities. For
of 1 mm horizontal and 2 mm vertical, as specified by instance, Sparkfun has developed a GNSS receiver that
Ophelia Sensors (Deane et al. 2019; Ophelia-sensors, has a manufacturer specified accuracy of 14 mm horizontal
2023). However, according to Rodriguez J. (2021), the and 10 mm vertical when RTK is used (dGNSS), delivers
accuracy of Geocubes is approximately 5 mm horizontal the same high frequency measurements as the Geocubes
and 10 mm vertical. (60-second interval), and is reasonably priced (a capital
In June 2018, ten Geocube receivers were installed at cost around US$1920 per unit), making it a suitable option
the Chin Coulee landslide site in Southern Alberta (Figure for use as an early warning system.
1), to monitor slope displacements at one-minute intervals. To evaluate the viability of the Sparkfun dGNSS
Nine Geocubes were installed on the slope, with four network for monitoring ground hazards, three Sparkfun
located on the moving mass and the remaining five placed units, consisting of one base and two rovers, were
on assumed stable ground around the landslide perimeter. deployed at the Chin Coulee landslide. This paper presents
The tenth Geocube, serving as the base, was installed an overview of the Sparkfun dGNSS system, its
across a reservoir from the landslide in a stable and fixed deployment, initial findings, and a comparison of its
area (Figure 1). The installation details are provided in monitoring data with that of the Geocubes.
Deane et al. (2020), Deane E. (2020) and Rodriguez J.
(2021) (Figure 2).
2 CHIN COULEE LANDSLIDE

The Chin Coulee landslide, located in southern Alberta

(Figure 3A) between Highway 36 and the Chin Coulee
reservoir's northern slope, is about 50 m high and 350 m
wide (Deane et al. 2020) (Figure 3B). The landslide has a
slope angle that varies from 9o to 13o and is approximately
2 Mm3 in volume (Deane et al. 2020).
The landslide is composed of three main layers: The
top layer is a medium plasticity clay fill that was a result of
highway construction. The layer below the clay fill consists
of stiff to very stiff glacial till, which includes sand and silty
lenses and has low to medium plasticity, ranging in
thickness from 5 m to 40 m. The bottommost layer is
Figure 1. Distribution of Geocubes at Chin Coulee landside bedrock which includes siltstone, sandstone, and medium
(Right photo S. Alberta MDs and Counties, Google Earth to high plasticity silty shale with sub-horizontal coal seams
Image) found within the shale. (Deane et al. 2020; Rodriguez J.
2021; Golder and Associates 1998).

Figure 3. Chin Coulee landside: (A) Location and (B) plan

view (Right photo, Landsat / Copernicus and S. Alberta
MDs and Counties, Google Earth Image)
Figure 2. Typical Geocube field setup
 and enabling the receiver to record monitoring parameters.
Slope inclinometers (SI) and piezometers were An advantage of the Sparkfun units is that there is no need
installed during multiple site investigation campaigns since to upload the settings every time the unit is powered on and
1998. The data collected from piezometers indicated that off, as the system automatically uses the latest settings
the groundwater table depth varies between 3.5 m and stored in the microSD card.
12 m depth in the landslide toe and head scarp,
respectively. SI measurements suggest the landslide is
moving over a sub-horizontal coal seam located
approximately 62 m below the head scarp (Deane et al.
2020; Rodriguez et al. 2021).
The evolution of the landslide has been assessed using
historical aerial photos and photogrammetry, revealing that
the construction of Highway 36 and filling of the reservoir
at its toe had an important role in the development of the
landslide (Deane et al. 2020). Filling of the reservoir and
highway realignment several times between 1950 and
1970, would have contributed to significant displacements
in the landslide area in 1978 and the formation of the
current landslide extension. Minor displacements were
observed between 1980 and 1996. A significant
reactivation occurred in 1997, impacting the highway and
leading to the inclusion of the Chin Coulee landslide in the
Geohazard Risk Management Program (GRMP) under Figure 4. dGNSS Sparkfun field setup
Transportation and Economic Corridors (TEC), and the
initiation of monitoring in 1998 (Deane et al. 2020). Due to
persistent landslide movements and a further relocation of 4 SPARKFUN SYSTEM DEPLOYMENT
the highway in 2016, the monitoring instruments installed
in the moving mass are no longer viable (Deane et al. 2020; In November 2022, three GNSS Sparkfun receivers (one
Rodriguez et al. 2021). In 2018, two SIs were installed base and two rovers) were installed at Chin Coulee
between the relocated highway and the slide head scarp, landslide to evaluate their suitability for monitoring
and the Geocube dGNSS network was also deployed at landslide displacements. The GNSS receivers were
the site. deployed alongside the Geocubes for two primary reasons:
to compare the results of both dGNSS systems and to
power the Sparkfun units using the Geocubes' power
3 dGNSS SPARKFUN SYSTEM supply system which consists of two 12-volt, 100 Ah
batteries, a 10-watt solar panel, and a solar charge
The GNSS Sparkfun receiver is a novel device designed controller.
for surveying purposes that integrates a high-precision
ZED-F9P positioning module, developed by U-blox,
capable of receiving and tracking four global navigation
satellite systems (GPS, GLONASS, Galileo, and BeiDou),
and operating on dual frequency bands.
The Sparkfun unit has a power consumption of 240 mA
and can work as a differential GNSS network (dGNSS)
achieving accuracies of 14 mm horizontal and 10 mm
vertical (according to the manufacturer’s specifications). It
records essential monitoring parameters such as the
number of satellites being accessed, longitude, latitude,
height above ellipsoid (HAE), and accuracy levels. The
monitoring data is stored in a microSD card, and the GNSS
units can support microSD cards up to 32GB in capacity.
The dGNSS requires additional components to be
considered a suitable differential monitoring system, such
as a dual-frequency GNSS antenna (Figure 4A), a point-to-
multipoint radio (XBP9X radio), an interface cable with
SMA Male to TNC Male connectors, a radio cable with JST-
GHR-04V and JST-GHR-06V connectors and a USB A to
C cable to power the units (Figure 4B).
The system can be easily configured through the
configuration page via WiFi, a terminal window, or using
the free software "U-center". The configuration primarily Figure 5. GNSS receivers’ installation locations at Chin
focuses on setting the measurement rate, the logging time, Coulee
 rovers’ units are powered by the solar panels and batteries
using USB plugs connected to the solar controller charger.
Geocubes SM22-6 and SM22-9 were selected as the The rovers and their radios are stored in battery boxes
monitoring points for the Sparkfun rovers based on their (Figure 8B).
continuous displacement trend (Rodriguez et al. 2021).
These locations were also selected for their clear line of
sight to the base installation location. Rover 949E and
rover 9532 were then installed at SM22-6 and SM22-9,
respectively (Figure 5).
The Sparkfun base was installed across the reservoir
from the landslide, at a pump house that has constant
energy supply (Figure 6). The Sparkfun base antenna and
the Geocubes antenna are both mounted on the same
wooden support (Figure 7A). The XBP9X radio was used
for the system, and its antenna was placed on the corner
of the pump house to ensure a clear line-of-sight with the
rovers’ radios. The base and its radio were stored in a
waterproof plastic box (Figure 7B) to protect them from
external factors such as weather.

Figure 8. Typical Sparkfun rover installation arrangement

5 GNSS SPARKFUN PERFORMACE AND

MONITORING RESULTS

The initial findings from a one-month monitoring period

between November 18, 2022, and December 18, 2022,
using the dGNSS Sparkfun network is discussed in this
section.
The GNSS Sparkfun system collected number of
satellites in view, longitude, latitude, height above ellipsoid
(HAE), and accuracy levels every minute. The data was
saved on a micro-SD card and later downloaded in CSV
format using "U-center" software (Ublox, 2023). A Python
Figure 6. Base Sparkfun installation arrangement code was created by the authors to post-process the data,
filtering outliers and removing data that exceeded the
manufacturer's accuracy thresholds (14 mm horizontal and
10 mm vertical) to minimize the dispersion and thus
improving the dGNSS network performance (Sharifi et.al.
2022). The Python code generates various graphs,
including the number of satellites, cumulative
displacements, and velocity.

5.1 Accuracy, cumulative displacement, and velocity

Results

An initial assessment of the Sparkfun dGNSS network's

performance was observed analyzing the accuracies
recorded during the monitoring period. The horizontal
accuracy achieved for both rovers was up to 15 mm, while
Figure 7. Sparkfun base, XBP9X radio and GNSS antenna the vertical accuracy varied between 10 mm to 20 mm
(Figure 9).
The number of satellites in view by the Sparkfun
The rovers' satellite antennas were installed at 0.5 m receivers varied from 19 to 32. The percentage of outliers
above the ground level and mounted on 5/8" threaded steel was relatively low, with 2% for rover 949E and only 1% for
rods. The rods were driven 0.5 m into the ground and rover 9532. However, outliers, peaks, and data above the
connected with brackets to a perforated square steel tube accuracy thresholds accounted for 11% and 13% of the
that was driven 0.8 m into the ground for stability. The raw data for rovers 949E and 9532, respectively. This
XBP9X’s antenna of each rover was mounted at 0.4 m suggests that the overall performance of the system was
above ground level on the steel rods to ensure a clear line- good, and at least 87% of the raw data can be used for
of-sight with the base radio’s antenna (Figure 8A). The analyzing the landslide displacements. The raw data (grey
dots) vs the filtered data (red dots) for the East coordinate Rover 9532, which is located closest to the reservoir,
of rover 949E is shown in Figure 10. exhibited a decelerating trend (Figure 5 and Figure 11B).
The lower movements recorded by rover 9532 during the
winter are consistent with the findings of Rodriguez et al.
(2021), who showed a graph indicating a decrease or
absence of displacement during that season. Rover 949E
experienced a sudden increase in the trend of cumulative
displacement (Figure 11A), mainly due to northward
displacements. However, further monitoring data is needed
to determine the cause or effect of the spike in rover 949E
on the overall displacement trend. Deane et al. (2019)
recommended using larger moving averages for reliable
displacement trend analysis on Chin Coulee landslide as is
classified as extremely slow-moving landslide (Cruden and
VanDine 2013), unfortunately this was not feasible for the
one-month monitoring period. A larger moving average will
be applied when more data is collected as this is a key part
of the ongoing investigation of the system.

Figure 9. Vertical accuracies of Sparkfun rovers: (a) rover

949E and (B) rover 9532

Figure 10. Rover 949E: east raw data vs east filtered data

The cumulative displacements exhibited a wide range

of values, varying between 0 mm and 20 mm, which made
it challenging to identify a discernible displacement trend.
Deane E. (2020) proposed the use of a moving average to
Figure 11. Cumulative horizontal displacement: (a) rover
identify the trends using GNSS monitoring systems. Sharifi
949E and (b) rover 9532
et. al. (2022) evaluated various moving average methods
and found that the Gaussian-weighted moving average
(GWMA) is the most effective method for analyzing data in
Both rovers detected downslope movement towards
the time domain, as it reduces scatter while preserving the
the reservoir, with vertical displacements ranging from
genuine displacement pattern (Sharifi et. al. 2022).
10 mm to 5 mm for the one-month period. Rover 949E
Therefore, the GWMA filter was used in this study to
recorded average horizontal velocities of 0.14 mm/day and
analyze the displacements of the Chin Coulee landslide.
0.13 mm/day for 12-hour and 1-day moving averages,
Based on the monitoring time, 12-hr and 1-day moving
respectively, which is consistent with Rodriguez et al.
averages were selected for filtering the dataset.
(2021) who reported horizontal displacement rates of
After applying the moving average, the displacement
around 0.2 mm/day during November to December. Rover
trend was visible for both rovers. For 1-day moving
9532 registered an average negative velocity of -
average, the horizontal cumulative displacement for rover
0.03 mm/day for the 1-day moving average. It should be
949E and rover 9532 are approximately 7.2 mm (Figure
noted that the average velocity calculated using the 12-
11A) and 2.5 mm (Figure 11B), respectively for the one-
hour moving average was 0.020 mm/day, which may be
month period.
attributed to the short average window applied and the very
slow displacements; therefore, the calculated values could 2 mm for the horizontal and vertical accuracies. The
be within the error of the system at this location. standard deviation for the east and north coordinates for
rover 949E and Geocube SM22-6 were below 5.6 mm and
5.2 Comparison with Geocubes system 2 mm, respectively (Figure 13). These accuracies were
below the threshold set by the manufacturers, indicating
The number of satellites in view differed between the two that the distance (~1 km) between the base and the rovers
Geocubes, with SM-6 receiving signals from 7 to 11 did not affect the performance of the Sparkfun system. This
satellites and SM-9 receiving signals from 12 to 22 corroborates Sparkfun's specifications, which require the
satellites. This is a lower number of satellites compared to rovers and base to be within a radius of 20 km to receive
the Sparkfun units, which may be due to the latter's dual- corrections.
frequency capability that allows it to capture more
satellites.
The cyclical daily pattern of fluctuations in the GNSS
network is a well-known phenomenon caused by
atmospheric conditions and signal reflections from nearby
surfaces (Kintner et al. 2007, Rodriguez et al. 2021). Both
dGNSS networks exhibited a cyclical daily pattern (Figure
12), and the varying number of visible satellites may
account for the differences in trends.

Figure 13. 24-hour daily standard deviation: (a) Sparkfun

rover 949E and (b) Geocube SM22-6

The Geocube SM22-6 had a lower cumulative

horizontal displacement of about 3.2 mm for the 1-day
window size than the rover 949E (7.2 mm). Geocube
SM22-9, on the other hand, exhibited a similar
displacement trend to rover 9532 with a cumulative
displacement of approximately 3.2 mm for the same
Figure 12. Cyclical pattern of north coordinates: (a) window size (Figure 14). Although it is recommended to
Sparkfun rover 949E and (b) Geocube SM22-6 analyze the Chin Coulee data for larger periods of time
(Deane et al. 2020), this initial analysis shows good
agreement between the Geocubes and the Sparkfun
The Hampel filter (Hampel F. 1971) was used to system.
remove outliers in the Geocube data, as recommended by The cost of a single Sparkfun receiver is US$750.
Sharifi et al. (2022). This method involves calculating the However, if we include the cost related to the mounting,
median of data within a window size and identifying outliers radios, power supply, and protection of the unit, the total
as data that falls outside a specified distance from the capital cost for one Sparkfun unit increases to around
median. The outliers were found to represent 6% and 4% US$1920 (Table 1). This cost is more affordable than the
of the raw data for SM22-6 and SM22-9, respectively. A Geocubes, which have a capital cost of around US$4500
higher percentage of data was discarded for the Sparkfun per unit according to Rodriguez J. (2021). However,
receivers, but the difference between the two technologies additional costs apply to both dGNSS networks that are not
was only 6% to 9%. included in the capital cost. For the Sparkfun system, there
To evaluate the accuracy of both GNSS technologies, are labor costs associated with installation data retrieval
a 24-hour daily standard deviation calculation was used. and data processing, which vary widely depending on the
This approach facilitated comparison with the Geocubes, project requirements. For the Geocubes, there are costs
which have a 24-hour error threshold less than or equal to associated with storing and maintaining the monitoring
data in the Ophelia database and monthly plan user costs.
For instance, to store and maintain data from ten effectiveness, three Sparkfun GNSS units, consisting of
Geocubes at the Chin Coulee landslide, an estimated one base and two rovers, were installed at the Chin Coulee
US$250 monthly fee is required. landslide to operate as a dGNSS network. To the author’s
knowledge, this is the first use of this system for ground
hazard monitoring in Canada.
The preliminary results presented in this paper cover
the first monitoring period with a duration of 1 month. The
results indicate that approximately 87% of the raw data is
suitable for analyzing landslide displacements. The
dGNSS Sparkfun has demonstrated horizontal standard
deviations that are significantly lower than the
manufacturer's specified accuracies, indicating excellent
performance in the field. The 24-hr horizontal standard
deviations calculated for both dGNSS networks during the
test are shown in Table 2. The Sparkfun rover 9532
showed a similar overall displacement trend to the
Geocube SM22-9, suggesting small cumulative
displacements in this period.
Overall, the system deployment has shown very
promising preliminary results. Testing of the Sparkfun
system for landslide monitoring continues, to evaluate its
reliability and overall field performance, and to assess its
applicability as part of landslide early warning systems.

Table 2. Manufactured vs testing horizontal accuracies for

a 24-hr period

Item 949E 9532 SM22-6 SM22-9

Figure 14. Comparison of cumulative horizontal Manufacturer
14 1
displacement between Geocube SM22-9 and Sparkfun (mm)
rover 9532 Testing (mm) <5.5 <5.0 <1.8 ≤2.0

Table 1. Breakdown of approximate capital cost for one 7 ACKNOWLEDGEMENTS

Sparkfun unit
This research was made through a collaboration with and
Description Cost (USD) Alberta Transportation & Economic Corridors (TEC) and
Sparkfun unit $750 Klohn Crippen Berger (KCB). The authors wish to offer
Radio System (XBP9X radio and special acknowledgement to Jorge Rodriguez, Aden
$250 Shipton, and Katrina Cereno for helping install the
accessories)
Protection (battery box and accessories) $340 Sparkfun dGNSS system at Chin Coulee landslide.
Mounting Equipment (Steel rods, metal
$160
clamps and solar panel pole)
8 REFERENCES
Power Supply System (solar panel,
$420
batteries, lighter plug and accesories)
Cruden D, VanDine DF. 2013. Classification, description,
Total $1920 causes and indirect effects - Canadian technical
guidelines and best practices related to landslides: a
national initiative for loss reduction. Geological Survey
6 CONCLUSIONS of Canada, Open File 7359, 22 https:// doi. org/ 10.
4095/ 292505
The GNSS Sparkfun units offer dual-frequency bands, Deane E, Macciotta R, Hendry M, Gräpel C, Skirrow R.
high-frequency measurements, a manufacturer accuracy 2019. The use and limitations of modern technologies
of 14 mm horizontal and 10 mm vertical as a dGNSS for slow, vegetated landslide monitoring - Chin Coulee
network and a relatively low cost compared with other landslide. In: 72nd Canadian Geotechnical Conference,
systems. This study has confirmed they are a promising Geo St. John’s 2019. St. John’s, NL, Canada
option for landslide monitoring. Its low cost enables its Deane, E. 2020. The Application of Emerging Monitoring
installation at greater density and multiple locations, Technologies on Very Slow Vegetated Landslides.
creating a strong monitoring arrangement that can
withstand damage to some units without affecting the
interpretation of the displacement. To evaluate their
Deane E, Macciotta R, Hendry M, Gräpel C, Skirrow R
2020. Leveraging historical aerial photographs and
digital photogrammetry techniques for landslide
investigation—a practical perspective. Landslides 17,
1989–1996 (2020). https://doi.org/10.1007/s10346-
020-01437-z
Google Earth. 2022. Alberta, Canada. S. Alberta MD’s
County –July 20, 2015. www.earth.google.com/ [April
27, 2022]
Google Earth. 2022. Landsat / Copernicus
www.earth.google.com/ [April 27, 2022]
Golder and Associates (1998. Geotechnical investigation –
site S5: highway 36:02 – November 1998 Report
Geotechnical Risk Management Plan
Hampel, F. R. 1971. A general qualitative definition of
robustness, Ann. Math. Stat., 42, 1887–1896, 1971.
Kintner PM, Ledvina BM, De Paula ER (2007) GPS and
ionospheric scintillations. Sp. Weather 5:1–23. https://
doi. org/ 10. 1029/ 2006S W0002 60
NRC. 2008. Canada’s Most Damaging Landslides.
Geofacts.
www.publications.gc.ca/collections/collection_2008/nr
can/M4-59-2-2008E.pdf [April 19, 2019].
Ophelia-Sensors. 2023. Product Information Page.
https://ophelia-sensors.com/geocube-solution [May 8,
2023].
Rodriguez, J., Michael T.H., and Renato, M. 2018. Cost-
Effective Landslide Monitoring GPS System:
Characteristics, Implementation and Results,
Geohazards 7, Canmore, Alberta, Canada.
Rodriguez J. 2021. Assessing The Benefits of Cost-
Effective Monitoring Technology on Unstable Slopes
Sensitive to Weather.
Rodriguez J, Deane E, Hendry M , Macciotta R, Evans T,
Gräpel C, Skirrow. 2021. Practical evaluation of single-
frequency dGNSS for monitoring slow-moving
landslides. Landslides 18, 3671–3684 (2021).
https://doi.org/10.1007/s10346-021-01737-y
Sharifi, S., Hendry, M. T., Macciotta, R., and Evans, T.
(2022). Evaluation of filtering methods for use on high-
frequency measurements of landslide displacements,
Nat. Hazards Earth Syst. Sci., 22, 411–430,
https://doi.org/10.5194/nhess-22-411-2022, 2022.
Smethurst J, Smith A, Uhlemann S, Wooff C, Chambers J,
Hughes P, Lenart S et al. 2017. Current and future role
of instrumentation and monitoring in the performance of
transport infrastructure slopes. Q J Eng Geol Hydrogeol
50:271–286. https:// doi. org/ 10. 1144/ qjegh 2016- 080
Ublox, 2023. U-center software information page.
https://www.u-blox.com/en/product/u-center [May 13,
2023].
Design, Installation and Commissioning of
Year-Round GNSS at Downie Slide
Chris Daniel, Glenn Erho, Adrian Colombo & Tom Stewart
British Columbia Hydro and Power Authority, BC, Canada

ABSTRACT
Downie Slide is an ancient, slow-moving landslide roughly nine square kilometres in plan with an estimated volume of 1.5
billion cubic metres. BC Hydro has monitored the slide for more than 50 years and recently installed a new automated
surface displacement monitoring system using GNSS technology. The application is unique because the equipment was
installed on towers up to 5.5 metres in height and because the GNSS receivers are powered by batteries with solar
recharge. This paper describes the design of the system and presents initial performance data.

RÉSUMÉ
Downie Slide est un ancien glissement de terrain à déplacement lent d'environ neuf kilomètres carrés en plan avec un
volume estimé à 1,5 milliard de mètres cubes. BC Hydro surveille le glissement depuis plus de 50 ans et a récemment
installé un nouveau système automatisé de surveillance du déplacement de surface utilisant la technologie GNSS.
L'application est unique car l'équipement a été installé sur des tours jusqu'à 5,5 mètres de hauteur et parce que les
récepteurs GNSS sont alimentés par des batteries à recharge solaire. Cet article décrit la conception du système et
présente les données de performances initiales.

1 INTRODUCTION Ongoing slide movement causes drillhole instruments to

fail in shear. Instrument health and utility is carefully
Downie Slide is an ancient, slow-moving landslide roughly tracked so that key instruments can be replaced prior to
nine square kilometres in plan with an estimated volume of failure, providing crucial cross-over data between new and
1.5 billion cubic metres (Figure 1). Located on Lake failing instruments. Instrumentation upgrade projects are
Revelstoke about midway between Mica Dam and required on a roughly 20-year cycle. These projects are
Revelstoke Dam, the slide is covered in old growth forest logistically and technically challenging. Each hole requires
and typically sees up to 4 metres of annual snow between 10 and 14 days of continuous drilling followed by
accumulation. The site has no cellular service and is 4 to 6 days of instrument installation. Instrument upgrade
accessed by boat or helicopter. Construction of Revelstoke projects and ongoing monitoring represent a significant
Dam caused the toe of the slide to be submerged. BC investment.
Hydro has regulatory obligations to provide drainage to
offset the submergence and to adequately monitor the slide During the most recent instrumentation upgrade project it
to confirm the ongoing effectiveness of the drainage. was decided to design and install a new, automated
surface displacement monitoring system using Global
Two drainage adits were excavated into the toe of the slide Navigation Satellite System (GNSS) technology. The
to more than compensate for submergence of the toe. driver was to reduce reliance on drillhole instruments that,
Monitoring of the slide has been ongoing for more than 50 in addition to high replacement cost, would fail abruptly if
years, including annual visual inspections, annual survey the slide mass were to accelerate suddenly. While this has
of surface monuments, and collection of data from been assessed to be improbable, ensuring continuity of
instruments including nested piezometers, inclinometers, monitoring capability is an important component of
extensometers and weirs. Technological advancements emergency preparedness planning.
have improved monitoring quality, allowing the installation,
logging and remote monitoring of sensors placed in The surface displacement monitoring trial described herein
selected instruments. Historically these have included in- is unique because the GNSS antennae were installed on
place-inclinometer sensors (IPIs), pressure sensors in towers up to 5.5 metres in height to stay above the snow,
piezometers, and water level sensors in weirs. and because the GNSS receivers, which draw about 3 W,
run on local batteries with solar recharge. This paper
IPI sensors placed within the shear zone at the base of the describes the design, installation, and commissioning of
landslide confirm seasonal variation of slide displacement the new GNSS equipment. Data illustrating the
rate that is correlated to groundwater levels. Total annual performance of the power supply systems, tower tilt and
displacements are measured through manual profiling of tower displacement due to slide movement are presented.
inclinometer casings and by comparing successive annual
surveys of permanent monuments. These two
measurements are in good agreement, showing horizontal
movement of 5 to 20 mm per year.
Figure 1. Plan view of Downie Slide showing GNSS tower locations (contours of elevation are in metres)

2 SYSTEM DESIGN 2.1 Site Selection

Continuous GNSS installations for monitoring Measurement locations were preferentially sited on rock
displacement require secure mounting locations, reliable outcrops, in areas of natural or existing clearings, on local
electrical power and either easy access to download data topographic highs and adjacent to existing access roads. A
or a means of remotely transmitting the data. These total of 11 on-slide sites were selected and a twelfth site
requirements are easily met in urban settings but were was selected across the valley to provide a stable
more challenging at Downie Slide. Performance targets reference point and act as a data repeater. Four of the
accounting for these limitations were developed through selected sites are on rock outcrops and the remaining eight
discussions with equipment suppliers, surveyors and the are soil sites.
personnel responsible for monitoring and maintaining the
equipment. The selected targets were: 2.2 Snow Data

 Year-round measurement of daily position; There are two permanent snow stations on Downie Slide
 Daily transmittal of positions to off-site server; that have been monitored monthly through each winter
 Detect movements as small as 20 mm; and, since 1977. One is located near location G-50 (Figure 1)
 Ability to detect tower bending and base rotation. and the other is above the headscarp of the slide. These
data were analyzed to develop design snow heights,
Some key aspects of the design are described in the passive snow load and creep and glide load corresponding
following sections. to a 50-year return period for each tower location (Table 1).
Figure 2. Three sizes of GNSS tower showing key equipment included on each tower (dimensions in millimetres)

Table 1. Site-specific snow design criteria.

50 Year Return Period Values
Site Elevation (m) Tower Height (m) Slope (deg.) Snow Height (m) Passive Load (kN/m) Creep & Glide Load (kN/m)
G-57 662 4.1 6 1.96 1.9 2.8
G-7 686 4.1 0 2.00 2.0 0
G-54 705 4.1 0 2.04 2.0 0
G-56 719 4.1 0 2.06 2.0 0
G-8 741 4.1 0 2.10 2.1 0
G-111 790 4.1 0 2.20 2.2 0
G-107 851 4.7 10 2.32 2.3 6.3
G-52 909 4.7 6.3 2.45 2.4 4.5
G-50 979 4.7 10 2.61 2.6 7.9
G-43 1048 4.7 6.5 2.77 2.7 5.8
G-19 1289 5.5 0 3.45 3.4 0
G-44 1371 5.5 5.3 3.71 3.6 8.2

The snow height governs the required tower height. snow load for each of the three tower heights was used for
Towers of three heights were designed as shown in Figure design. They consist of galvanized HSS 168 x 4.8 and HS
2 and used at the various sites as listed in Table 1. 273 x 7.9 members welded or spliced (Figure 2). The
splices were included to maintain the ability to limit
2.3 Tower Deflection and Rotation component weight for helicopter transport, if required.

The performance cost of installing GNSS antennae on top Base rotation due to lateral loading was mitigated by
of a tower is that the system will record tower bending and selecting rock foundations when practicable. Foundations
base rotation movement in addition to slide movement. at rock sites consisted of four 32 mm diameter vertical
Tower bending was found to be controlled by lateral snow Grade B7 anchor rods grouted into 60 mm diameter holes
load. The potential for base rotation arises due to both to either 1.4 or 2.0 m depth depending on tower size.
lateral snow load and the towers being founded on an
active landslide. For towers of the height required at Soil foundations were pre-cast concrete footings with
Downie Slide, these other causes of movement could integral central columns (Figure 3). The footings were
eclipse the measurement of interest. designed to resist overturning with no passive lateral
For tower bending, a maximum allowable tower top resistance. The footings were either 1.6 m or 2.2 m square
deflection of 10 mm was adopted and the most severe depending on tower size.
 Figure 4. Tiltmeter sign conventions.
Figure 3. Installation of 1.6 m square concrete footing.

Recognizing that some base rotation would be inevitable

due to ongoing movement and deformation of the slide
mass, each tower was equipped with a precision biaxial
tiltmeter just above the base of the tower. A second biaxial
tiltmeter was placed near the top of the tower to detect
tower bending. Uniform tilt measured in the upper and
lower sensors indicates base rotation. Tilt measured by
only the upper sensor indicates tower bending. The
tiltmeters are installed in protective housings on the north
sides of the towers (Figure 2) to reduce solar heating of the
sensors. The planes of measurement and sign
conventions are shown in Figure 4.

2.4 Power Requirements

Solar panels and a battery bank were the only option

available to power the GNSS equipment. BC Hydro
commonly runs monitoring equipment on solar power but
the stated 3 W power draw of typical GNSS receivers is 2-
3 orders of magnitude higher than typical equipment. Due
to the uncertainty about the feasibility of this approach, a
power system trial was considered necessary. A 3 m high
scaffold was built and lashed to an existing helipad
(Figures 5). The prototype power system and prototype
GNSS, datalogging and radio equipment were installed in
2019 and ran over the winter of 2019-2020.

The power system proved to be insufficient for a period of

about 3 weeks in January when daylight hours and
intensity were limited. A useful lesson learned from this
trial was that the GNSS equipment would shut down when
the battery bank voltage dropped below a threshold
amount but would reboot automatically when the voltage
recovered. This knowledge was used to design power-
saving logic. The datalogger programs in the permanent
installations include logic to cut power to the GNSS
receiver when the battery voltage drops below 11.8 V and
leave it off until voltage increases to greater than 12.0 V.
Inclusion of the power saving logic allowed the final power
system design to be of similar capacity to the prototype
system and include four 275 A-Hr 6V deep cycle batteries
and two 180 W C1D2 12V solar panels at each tower.

The power trial also guided the vegetation clearing protocol

that was adopted for the final design, which called for
creation of a 60-degree sky-view in all directions but Figure 5. Power trial arrangement during setup (upper) and
particularly in the southern hemisphere. during a fly-over on February 3, 2020 (lower).
 3.2 Tower Tilt

3.2.1 Daily Trends

The tower tilt data shows very clear daily cycles of tower tilt
driven by differential heating of the tower sides. The side of
the tower facing the sun tends to heat up and expand
longitudinally more than the opposite side, causing the
tower to bend away from the position of the sun. This effect
is more pronounced in the summer.

The effect is illustrated in Figure 9, in which Tower G-52 is

seen to bend to the north and then back to starting position
over the course of each day, reaching a peak northerly
deflection between noon and 14:00. In the east-west
direction, the tower tilts to the west in the morning and the
east in the afternoon. The tower top deflections assuming
uniform tilt of the entire tower are in the order of 5 mm.
Figure 6. Completed 4.1 m tower (G-08).
For hourly GNSS positions, it would be possible to correct
each reading for tower tilt. Over the course of a day,
however, the thermal deflections of the towers are
asymmetric, particularly in the north-south direction, and
thus the thermal deflections tend to introduce a northerly
bias in the GNSS. GNSS measurement error generally
decreases with measurement duration because random
errors average out. In this instance, however, the benefit of
a longer measurement period is negated by the
introduction of seasonally-varying bias error due to thermal
deflection of the towers. The GNSS position accuracy is
thus optimized by reducing the measurement duration to
the period between midnight and 06:00 each day. This will
also decrease power draw and data transmission
requirements by about 75%.

3.2.2 Annual Trends

Figure 7. Completed 5.5 m tower (G-19) with Downie Slide Multiple years of data will be required to assess annual
head scarp in background. trends, but these are comments on preliminary trends.

3 SYSTEM INSTALLATION AND PERFORMANCE In terms of base rotation, the soil foundation towers were
more likely to exhibit some non-recoverable tilt than rock
The twelve GNSS towers were installed and instrumented sites, as expected. Some of the rock foundation sites
during the summer and fall of 2021. Typical completed appear to have undergone almost completely recovered tilt
towers are shown in Figures 6 and 7. Performance data in the east-west direction, roughly corresponding to the
from the first year of operation is presented in the following direction of slide movement.
sections.
Only one tower (G-7) showed a clear trend of tower
3.1 Power Supply bending over the winter. This tower is in a large clearing
with no cross-slope, such that the snow creep load was
Power supply performance varied across the towers and forecast to be zero. It is speculated that the loads that
as expected, was qualitatively correlated to the available cause the bending may have been due to differential
southern exposure. Three of the sites show very good melting of the snowpack around the tower, but it is not clear
performance, operating year-round with no requirement for why this bending would have been at one tower only.
power-saving logic. The remaining sites required between
roughly a week and 6 weeks of power saving logic during There is evidence of snow-driven tower tilt at several other
January and February. A typical example of the latter is sites that may be related. At these sites a trend of tilt in one
shown in Figure 8. Daily tower tilt data discussed in section direction is observed, followed by an abrupt reversal of the
3.2 presents an additional power saving opportunity that trend, presumably due to removal of the driving load (e.g.,
will reduce or eliminate the need for the power saving logic. Figure 10). The abrupt reversal typically occurs in January
or February. It is speculated that these may also represent
differential melting.
Figure 8. Typical battery voltage history showing short-term power-saving logic in January due to low solar input.

Figure 9. Tower G-52 hourly tilt data from upper tiltmeter (August 13 to October 26, 2021).

Figure 10. Northing and Easting displacements for Towers G-7 and G-52 between October 28, 2021, and October 1, 2022.
3.3 Slide Movement September is unknown but does not seem to be caused by
tower tilt.
The GNSS data processing software requires a real-time
data feed and establishing this data flow was delayed while 4 SUMMARY AND CONCLUSIONS
a protective firewall was installed. For this reason, daily
positions from the first year of operation (October 28, 2021 New GNSS installations were designed and installed at
to October 21, 2022) are only available for two towers, G- Downie Slide including eleven on the slide and one
7 and G52 (Figure 10). The daily positions shown were reference station located off-slide. Each GNSS antenna is
calculated using data collected between midnight and attached to the top of a steel monopole tower to allow data
06:00 to eliminate noise related to thermal deflection of the collection year-round despite up to 3.7 m of expected snow
towers. The blue data points are daily positions, and the height. The tower height is such that power supply solar
orange lines are fifteen-day moving-averages. panels and radio antenna are also maintained above snow
level year-round. The controlling factor for the structural
Tower G-52 was installed roughly 100 m from survey design of the towers was lateral snow load and the target
monument M-52. There is no survey monument near of limiting any snow load induced tower lateral deflections
Tower G-7. The annual survey of M-52 was completed 24 to less than 10 mm at the top of the tower. GNSS data
days before the start of data collection at G-52. Despite collection was initiated in October 2021. This paper
this, the annual displacement measured for M-52 provides presented select data from the first year of operation.
a reasonable check of the measured G-52 displacement as
summarized in Table 2. The resultant displacements are Battery voltage data demonstrates that the power system
essentially equal at about 7 mm. The resultant azimuths performance is adequate. The system controls include a
are 31 degrees and 64 degrees for M-52 and G-52 conservation mode to protect battery health during periods
respectively. The cause of this discrepancy is unknown, of low solar recharge and the towers typically only required
but it is noted that the resultant azimuth for M-52 over the this measure for up to 3 weeks in January. Tiltmeters
preceding ten years was 85 degrees, roughly downslope. installed on the towers revealed daily cycles of thermal
deflection that are easily avoided by collecting data
Table 2. Comparison of M-52 and G-52 Results overnight. Tiltmeter data also confirmed that snow load
Displacement M-52 G-52 induced tower tilt and base rotation due to slide movement
Manual Survey Automated System or soil foundation movement only produced tower top
2021-09-21 to 2021-10-15 to deflections less than 10 mm and generally less than 5 mm.
2022-09-13 2022-09-13
Full histories of displacement are available for two towers
Northing (mm) 6.2 3.0
over the first year due to the need to ensure secure data
Easting (mm) 3.7 6.2 transmission. One of these towers was adjacent to a
Resultant (mm) 7.2 6.9 survey monument and the displacement of the monument
Azimuth (degrees) 31 64 was measured to be equal to that of the tower (7 mm) but
with a 33 degree difference in azimuth. Comparison of the
Review of the daily position data provides additional insight GNSS tower displacement to the calculated tower tilt
into timing of displacement that is not available from the displacement histories shows the slide begins to move in
manual survey results collected once a year. Figure 11 April as observed in available IPI histories. Operation of
shows the history of Northing and Easting displacements the GNSS equipment is ongoing and collection of multiple
(blue lines) and the fifteen-day running average (orange annual cycles of movement is expected to establish
lines) for Tower G-52. The two lower plots in Figure 11 baseline surface movement trends that can be used for
compare the fifteen-day running averages in orange to the performance monitoring, augmenting the critical downhole
estimated tilt-induced displacement in green. The latter instrumentation at site at a fraction of the cost of an
assume the tilt measured at the upper tiltmeter is uniform equivalent number of downhole monitoring locations.
over the height of the tower. The GNSS displacement on
G-52 is seen to follow the tilt induced displacement through Acknowledgments
most of the winter and then begin to deviate through the
summer. The work described herein builds on more than fifty years
of previous work by BC Hydro engineers, geologists and
The offset between the GNSS displacement and the tilt- technologists, external consultants and researchers at
induced displacement is interpreted to be the slide Downie Slide. Personnel contributing to the current
movement. The trend of the daily readings gives greater instrumentation upgrade project are too numerous to list in
confidence into the displacement pattern than can be full, but the contributions of Zeljko Cecic, Christine Peters,
achieved from annual readings alone. The onset of slide Mitchell Illerbrun, Chris Wong, John Wardrop, Julia Marsh,
movement in April that is apparent in Figure 11 is Chase Reid and BC Hydro Construction Services are
consistent with IPI data which also show onset of particularly acknowledged. Camilla Loughlin and Mehrdad
movement during the spring melt. The cause of the Soltanzadeh assisted with processing of the tower tilt data.
northing deflection observed in June and then reversed in Special thanks also to Dynamic Avalanche, Monaghan
Engineering and Monashee Survey and Geomatics.
Figure 11. Tower G-52 daily GNSS position data (upper plots) and comparison of GNSS displacement to tilt-induced
displacements estimated from upper tiltmeter data
A Semi-Empirical Oil Sands Mine Pit Wall
Design Update at the Muskeg River Mine
S. Li, R. Hoda and S. Iqbal
Canadian Natural Resources Ltd., Calgary, Alberta, Canada

ABSTRACT
The Canadian Natural Muskeg River oil sands mine near Fort McMurray has conventionally used a 2D limit equilibrium
approach to design pit walls. The mine has been in operation for over two decades, which has provided the opportunity to
learn from the pit wall instrumentation data, bench-scale slides, and panel mining activities. Significant safety, practical,
and economic benefits may be realized with optimized pit wall designs.

Applying experience and learning, the pit wall design approach has evolved from focusing on 2D limit equilibrium methods
to a semi-empirical approach that considers additional potential failure modes (e.g., bench block slides, rich ore flows, and
adverse clay dips). This paper describes the updated pit wall design approach and recent performance results at the
Muskeg River Mine.

RÉSUMÉ
Canadian Natural exploite la mine de sables bitumineux de la rivière Muskeg près de Fort McMurray depuis deux
décennies. L’approche d'équilibre limite 2D a été utilisée de façon conventionnelle pour concevoir les murs de la fosse.
Des apprentissages ont été faits à partir de la performance des murs, de la défaillance des bancs et de l’extraction minière
en mode de panneaux. Une conception optimisée des murs peut permettre des améliorations significatives en matière de
sécurité, d’opérationnalité et au niveau économique.

A partir de l'expérience acquise, l'approche de conception des murs a évolué passant de la méthode d'équilibre limite 2D
à une approche semi-empirique qui prend aussi en compte d’autres modes de défaillance (par exemples: la rupture en
bloc de bancs, des écoulements riches en bitume et des sols argileux défavorables). Cet article décrit l'approche améliorée
de conception des murs et les résultats récents de la performance de la mine de la rivière Muskeg.

1 INTRODUCTION wall design process towards a semi-empirical approach

that considers additional potential failure modes discussed
The Muskeg River Mine (MRM) is an open-pit mine and in Section 4. This paper describes the updated pit wall
bitumen production facility operated by Canadian Natural design approach and recent performance data
Resource Limited (Canadian Natural) as part of its Albian observations at MRM. A general MRM mine pit layout is
Sands joint venture project. The mine is located on presented on Figure 1.
Lease 13, 70 km north of Fort McMurray, Alberta.
Conventionally, a 2D Limit Equilibrium Method (LEM)
approach was used for the final pit wall design at MRM. For
specific projects, deformation analyses were also
conducted using FLAC 2D to ensure the safety of critical
infrastructures (Bidwell et al., 2011).
Pit wall designs using an empirical approach have been
reported for other oil sands mines in the Fort McMurray
region by Cameron et al. (2008), Chen et al.(2016) and Shi
et al. (2021), which demonstrate that design optimization
can gain additional ore while maintaining safety using the
Observational Approach (Peck, 1969). However, the
geological settings at MRM are unique as the mine has a
relatively thin overburden compared with other oil sands
mines, which limits the direct application of learnings from
other mines.
MRM has been in operation for over two decades,
developing significant experience and learnings from pit
wall performance, bench-scale slides, and panel mining Figure 1: MRM Mine Pit
activities. These learnings have helped to evolve the pit
2 GROUND CONDITIONS

2.1 Geology

The stratigraphic profile at MRM is characterized by a thin

layer of Quaternary overburden (1-5 m), underlain by the
Cretaceous McMurray Formation oil sand on top of the
Devonian Waterways Formation limestone and calcareous
shale. The general stratigraphy profile for the MRM area is
shown in Table 1.
The depositional environments for the Holocene and
Pleistocene are continental and glacial, respectively. The
surficial Quaternary overburden is present as Holocene
and Pleistocene units at MRM, and the deposits are mainly
glacial sand/gravel material.

Table 1: General stratigraphy profile at MRM

Age Member

Holocene HOL
Quaternary
Figure 2. LM1w Thickness Map
Pleistocene PL
2.2 Hydrogeology
Clearwater Kc KcW
Understanding the hydrogeological processes is essential
Upper McMurray UM in assessing the effects of groundwater pressure on the pit
Early
wall stability. In general, the groundwater conditions at
Cretaceous
Middle McMurray MM2 MM1 MRM are characterized as follows:
• Before mining, the groundwater generally occurs at
Lower McMurray LM2 LM1 LM1w or near the original ground surface. As mining
progresses, stress relaxation and slope face
Devonian Devonian DW drainage facilitate the natural dissipation of the pit
wall groundwater and pore pressures.
The Upper McMurray (UM) underlies the Quaternary • Deep pumping wells are commonly installed at
deposits and was generally deposited in a marine MRM to dewater the underlying watersands to
environment. Geotechnically, unfavourable facies within support the mining operation. This process also
the UM include Offshore Mud (OM), Marine Lagoonal Mud reduces groundwater flows from confined aquifers
(MLM), and Lower Shoreface (LSF). These clay facies can through the pit wall face, improving the overall
cause localized wedge block failures if adverse clay dips stability of the pit wall and trafficability for equipment
are present. in the pit.
The Middle McMurray (MM) underlies the UM deposit • Generally, the north and east sides of the MRM pit
and consists of inter-bedded estuarine and tidal channel have the greatest thicknesses of watersands. They
sands, which form the major ore horizons. Weak MM form part of a continuous basal aquifer along the
facies, such as Mud Flat (MF), Tidal Flat Mud (TFM), east side of Lease 13. The watersands, however,
Estuarine Channel Mud (EM), Tidal Mud (TM), and Tidal are more sporadic and discontinuous towards the
Channel Abandoned (TCA), can also cause localized western half of the pit.
wedge block slides. • The hydraulic conductivity of the McMurray
The Continental environment Lower McMurray member Formation is very anisotropic, as the layers drain
(LM) is the lowermost part of the McMurray Formation and horizontally, and there may be multiple perched
includes a flushed sandy zone at the bottom called basal water tables in the pit wall.
watersands or LM1w, which is similar to the oil sand with a
low bitumen content and high permeability. This sandy unit 3 DESIGN APPROACH
can enhance the stability of the pit wall due to its dense,
interlocked nature. Figure 2 shows a thickness map of The final pit wall height is about 70 m to 80 m, with multiple
LM1w at MRM. 15 m high benches at MRM; a typical pit wall configuration
The Cretaceous McMurray sits uncomformably on the is shown on Figure 3. A haul road is generally incorporated
Devonian Waterways (DW) Formation limestone. It on the first bench to maintain mine equipment access.
includes clayey facies of geotechnical concern such as the
Calcareous Shale (CALCSH) and Paleosol (PSOL) at the
Devonian contact and frequently governs the pit wall global
stability.
 4 PIT WALL FAILURE MODES

Pit wall failure modes are reasonably well understood

through over 20 years of mining at MRM and the
experiences of other open-pit oil sands mines (e.g.,
Morgenstern et al. 1988, and Chen et al. 2016). The
common pit wall failure modes for open-pit oil sands mines
Figure 3. Typical MRM Pit Wall Configuration are described in the following subsections.

3.1 Semi-Empirical Design Approach 4.1 Crest Retrogression and Slumping

The previous MRM pit wall design primarily relied on the The steep, rapidly excavated slopes in heavily
2D LEM stability assessment of representative cross- overconsolidated soils can stand at near-vertical slopes for
sections to obtain a target Factor of Safety (FOS). Over 20 a short period. With time, pit wall crest retrogression and
years of operational experience provides extensive slumping could happen. Section 2.1 notes that the
performance observations and instrumentation data. Quaternary overburden deposits at MRM are generally thin
Opportunity-based panel mining activities have been and sandy. These overburden units are usually excavated
completed, which involve temporary pit wall excavation together with the underlying McMurray overburden, and
with a steeper than the design slope followed by prompt any shallow sloughing is operationally managed. This type
buttressing with compacted fills. of failure is generally not a concern at MRM.
The semi-empirical approach considers observed
interim and final pit wall slopes with similar geological 4.2 Wedge Failures and Toppling of McMurray Clays
conditions to assess the failure risk of the pit wall section
under design. Section 4 of the paper discusses several Wedge failures and the toppling of jointed McMurray clays
potential failure modes applicable to the MRM pit. A pit wall are shallow block failures that commonly occur in the UM
performance review is presented in Section 5. clay facies (e.g., MLM, LSF) and the MM clay facies (e.g.,
MF, TFM, TFX, TCA, TM). This type of failure is typically
3.2 Pit Wall Stability Criteria limited to the bench scale. Due to the relatively low
hydraulic conductivity of the surrounding oil sands, the pore
The target factor of safety (FOS) criteria for pit walls are pressure in clay beds does not dissipate within the short
provided in Table 2. period between successive pit wall cuts. The resulting low
shear strength in the clay beds and the presence of any
Table 2: Design Target FOS Criteria for Pit Walls unfavourable dip angle of these clay beds can cause block
failures. Figure 5 presents an example of a block failure
Design Case MLC1 RWC2 through TFX clay facies in the Middle McMurray.

Pit Wall Overall Slope 1.2 1.1

Global Failure Through Deformation
1.3 1.1
Tolerant Structure Located at the Crest
Global Failure Through Deformation
1.5 1.3
Intolerant Structure Located at the Crest
1
Most Likely Case (MLC) material parameters are those that, in
the judgment of the design engineer, have the highest probability
of being mobilized in the field, allowing for considerations of strain
compatibility.
2
Reasonable Worst Case (RWC) material parameters are the
worst parameters likely to occur over a scale large enough to
influence the stability of the earth structure.

Each design section is assessed for several potential Figure 5. Middle McMurray Block Slide
failure planes, as shown on Figure 4.
4.3 Global Translational Block Slide Failures

This mechanism refers to the large-scale slides starting

from the pit wall crest (or the crest of an earth structure
near the pit crest) and daylighting at the pit wall face
through weak clay facies within the McMurray
(UM/MM/LM2) or shallow Devonian Waterways. This type
of failure is the target of the stability analyses, and an
Figure 4. Potential Failure Planes adequate FOS is required per Section 3.2. For stability of
the pit wall only (no earth structure at or close to the crest), At earlier stages of mining, SIs were installed near the final
only failures through weak LM2 facies and PSOL/CALCSH pit wall crest but deformations of interest were not always
are examined. A translational block slide through UM or recorded (Bidwell et al., 2011). Figure 8 shows one SI
MM is not expected (high calculated FOS) and can be installed 34 m from the final pit wall crest with an overall
managed operationally. design slope of 2.7H:1V. No discrete shear movement was
A weak layer is typically assigned at the top of a specific observed at this location.
geological member in the stability model. When adjacent
earth structures are on the pit wall crest, this failure
mechanism is checked through all possible clayey facies
within the UM/MM/LM2/DW members, based on the
interpretation of the borehole logs.

4.4 Rich Ore Flow

Rich-ore flow is common with bitumen-rich, poorly

cemented oil sands near the base of the McMurray
members (i.e. estuarine oil sands in the MM and Fluvial oil Figure 8. No discrete shear movement
sands in the LM1) when the ambient temperature is high.
The locations of potential rich-ore flow are depicted on SIs are now installed on the crest of the first bench of pit
Figure 6. The rich ore flow typically stops when the repose walls in some areas at MRM so that the relevant
angle (e.g., 2H:1V) is established. Figure 7 shows an deformations can be better detected. The pit wall
example of the ore flow at a lower bench. excavation in most of these areas triggered discrete shear
movements in the UM, MM, LM2, and DW members.
Figure 9 presents an example of the discrete shear
movements recorded in instruments installed on the first
bench of the pit wall. These shear movements attenuated
upon completion of the excavation and are believed to be
primarily caused by horizontal stress relief, as none
progressed to overall failure.

Figure 6. The potential locations of rich-ore flow

Rich-ore flow Figure 9. Shear movement observed in all members

5.2 Movement Observations from Panel Mining

Panel mining is a construction technique to gain additional

ore by removing a portion of the pit wall base (e.g.
Figure 7. Rich-ore flow at the Fluvial bench temporarily cut into the design slope) and backfilling
immediately with compacted fill in the excavated area to
5 PIT WALL PERFORMANCE support the pit wall stability. Panel mining relies on the
experience that the short-term stability of an open "panel"
Instrumentation data are essential to understanding the pit of limited distance parallel to the pit face is generally
wall behaviour. They are indicators of the influence of greater than the long-term stability of an extended distance
stress relief due to mining, increase or decrease in pore of the pit wall, allowing the slopes to be mined steeper than
pressures from groundwater and stress changes, and other the long term stable slope angle, provided that a buttress
factors influencing pit wall stability. The commonly installed is placed shortly after mining. Mining and buttressing can
geotechnical monitoring instruments at MRM include be done either in a continuous manner or in panels.
Vibrating Wire Piezometers (VWP), Slope Inclinometers Figure 10 shows a typical panel section and backfill
(SI), and Shape Accelerometer Arrays (SAA). This section advance and Figure 11 shows SI data recorded during a
presents some key observations of shear movements and panel mining event. Panel mining is used at MRM on an
pore pressure responses due to mining. opportunity basis to extract additional ore from the
Estuarine ore (above LM2) or the Fluvial ore (above the
5.1 Movement Observations from Mining to the Design LM1w or DW surface) using the Observation Approach
(Peck, 1969).
Before executing a panel mining operation, a detailed • Zone C: The elevation 250 m bench was
geotechnical assessment is performed to identify all the established by the end of November 2018. The
risks per the failure modes discussed in Section 4, with minor shear movement was observed in a MM Mud
adequate instrumentation coverage in the area of interest. Flat layer. The shear movement started by the end
The mine operation personnel are engaged in specific of May 2019 and accelerated near the end of
details and operational requirements. August 2019. The acceleration of the shearing is
due to the excavation of the 265 m bench. The
buttress was established and shear movement
rates slowed, with 12 mm displacement.
• Zones D: Excavations between August and October
2018, triggered approximately 10 mm discrete
shear movement, at the DW contact. Excavation
stopped in October and the toe berm/buttress was
constructed. However, the shear movement
attenuation was slower than expected which is
attributed to the compression of any loose fill within
the buttress.
Excavation over the design limit due to a panel mining
event could result in localized sloughs. These sloughs can
pose challenges during the abutment construction for the
In-Pit Dykes (Hoda et al. 2016, and 2019). Therefore, panel
Figure 10. A typical panel section and backfill advance mining is generally avoided in abutment areas.

As part of a pit wall optimization program at MRM, a

series of instruments were installed across the pit wall to
gain more insights into pore pressure responses, horizontal
stress relief or shearing during the excavations. Figure 12
shows three SIs installed at one cross-section of the final
pit wall on the pit crest, the first bench, and 90 m away from
the pit wall crest.

Figure 12. Shear movement along a cross-section

The following observations are made from the cross-

section with three SIs:
• As expected, the SI installed closer to the crest
recorded higher deformation in all shear zones.
Figure 11 SI responses due to a panel mining event • The discrete shear movements observed in the
McMurray (i.e. MM and LM2) clays are primarily due
Figure 11 shows one SI with four discrete shear zones to horizontal stress relief during mining. The
developed during a panel mining event. Details of shear magnitude of the shear movements in the MM and
movements for each zone are summarized as follows: the LM2 was generally between 2 mm and 14 mm.
• Zone A: No shear movement was observed when • Despite the presence of LM1w, shear movements
the elevation 280 m bench was initially established are still observed at the DW contact for two SIs
in August 2018. The panel excavation between July closer to the pit wall crest. The magnitude of the
5th and 12th, 2019, beyond the long-term stable shear movements at the DW contact is between
slope angle below the 280 m bench, triggered 12 6 mm and 8 mm. A portion of these movements
mm of shear movement within the MM clay, which could be attributed to shearing resulting from a
eased by the end of July 2019. steep slope from the first bench crest to the pit wall
• Zone B: Shear movement occurred in a MM clay toe.
layer and appeared to be associated with the panel • No shear movement in any clay facies was
excavation. The excavation of the 265 m bench was observed from the SI installed at an offset from the
completed around July 05, 2019. crest (~90 m).
The observation of no shear movement in LM2 clays 90 m
from the crest differs from the FLAC modelling results
(Bidwell et al., 2011), which predicted horizontal movement
within the LM2 was 180 mm at 100 m from the pit wall crest.
Bidwell et al. (2011) also noted that the observed horizontal
movement at the time is less than predicted by the
numerical analysis due to a lower in-situ horizontal stress
and more favourable geological conditions than assumed
in the numerical modelling. Therefore, the performance
data provides valuable insights into the pit wall design
process. A summary of the observed overall pit wall slopes
is discussed in Section 6.1.

5.3 Observed Piezometric Conditions

Pore pressures within the pit wall, especially those in the

clay facies, play a crucial role in pit wall stability. Vibrating
Wire Piezometer (VWP) tips are typically installed in the
clay facies during SI installation using the fully grouted
method. A representative piezometric condition is shown
on Figure 13 for different members/clay facies within the oil Figure 14. Piezometric heads dissipating in the MM/LM2
sands formation.
6 PIT WALL DESIGN

The pit wall design at MRM considers the failure modes

discussed in Section 4 and includes the following
components:
• Understand the pit wall performance through a
detailed review of existing instrumentation data
and the maximum observed slopes for various
geological conditions and pit wall height. This
component plays a critical role in the current
design work.
• Perform 2D LEM stability analyses with observed
pore pressures in the critical facies. The 2D LEM
stability results are then used to validate the
design based on observed pit wall slopes.
• Interpret the clay dips in facies of concern along
Figure 13. Typical MRM stratigraphic column with the final pit wall for potential block slides.
piezometric conditions • Incorporate a flatter slope for rich ore benches.

The pore pressures generally dissipate with mining. The 6.1 Observed Pit Wall Slopes
factors that could influence the observed pore pressure
regime within the pit wall are listed below: Although some bench-scale block failures have occurred,
• Two different tips in similar clay units can exhibit no large-scale or full-height pit wall failures have occurred
varying piezometric elevations for similar geometry during two decades of mining at MRM. A summary of
if the drainage in one area is aided by the presence observations during this time includes:
of a nearby coarser material. • Discrete shear movements were observed during
• An increased stress condition due to the presence mining to the design limit and panel mining past the
of an earthen structure near the pit wall crest could design limit in a controlled manner.
introduce stress-related pore pressures. • The observed stable slopes during mining are
• The DW VWPs are typically installed within the top generally steeper than the 2D stability analysis
few meters of the DW clays. Any basal aquifer predicted slopes.
activity (LM1w) can significantly influence these
instruments. For discussion purposes, the MRM mine pit is divided into
three areas (Areas 1 to 3) to allow consideration of earth
Figure 14 presents an example of pore pressure structures (e.g., waste dumps), different hydrogeological
dissipation during the pit wall excavation process. conditions, and pore pressure behaviour in clays. The key
features of each area are discussed briefly in the
corresponding subsection below.
Observations from Area 1: There are no earth structures examined to ensure sufficient offset between the
near the pit wall crest, and the pore pressure within the clay structure and pit wall crest, as shown on Figure 4.
facies dissipates faster than in Areas 2 and 3. The slope
measured from the crest of the haul road to the toe is about 6.3 Clay Dip Interpretation
2.0 H: 1V with a 63 m high wall. The steepest slope
measured from the pit wall crest to the toe is about 2.7 H: The orientation and dip angles of clay beds within the oil
1V with a 72 m high wall (Figure 9). sands play a crucial role in the pit wall stability
(Morgenstern et al. 1988), as they can lead to the
Observations from Area 2: There is an 80 m high waste development of block slide failures. Accurate interpretation
structure near the pit wall crest with a relatively thin LM1w of dipmeter data can identify areas where these features
layer. The pore pressures within clays generally dissipated could pose a high level of operational risk.
slower than in other areas at the MRM. The steepest slope The dipmeter data within the final pit wall were
measured from the pit wall crest to the toe is about 2.8 H: reviewed to assess the potential impact of clays beds on
1V with a 78 m high wall. The slope measured from the the pit wall stability. Based on the experience from
crest of the haul road to the toe is about 2.2 H: 1V with a localized block failures of pit walls at MRM, the following
63 m high wall. A panel mining exercise was completed in clay units are selected for assessment:
late 2018, and a temporary overall pit wall slope of 2.4H:1V • Marine Clay (UM): MLM, USF, MSF, LSF
was established for a 78 m high wall with an immediate • Estuarine Clay (MM): SM, MF, TFM, TFX, TCA, TM
buttress (Figure 11). • Continental Environment (LM2): BS, CO, M, FPOB,
DCM
Observations from Area 3: There are no earth structures When the strike of a geological discontinuity is within ±20
near the pit wall crest, and the dissipation rate of the pore degrees of the strike of the slope face, or if the azimuth of
pressure within clays is higher than that of Area 2. The the line of intersection by two or more geological
steepest overall slopes are 2.4H: 1V with a 75 m high wall discontinuities is within ±45 degrees of the azimuth of the
or 2.0H: 1V with a 69 m high wall (Figure 12). pit wall slope face (Wyllie and Mah, 2004) the strike is
considered unfavourable. Only one geological discontinuity
Deformations in the critical facies (e.g., BS, M, FPOB, and (e.g., bedding plane) is considered during the design stage
PSOL) were acceptable (i.e. 10 mm) and the observed by reviewing the borehole dipmeter data. The following
slopes are generally steeper than the previous pit wall coding criteria were established to quantitatively set the
designs that were based on 2D LEM stability analyses. potential of a block failure at each borehole location based
Therefore, the observed slope configurations could help on the dip angle of unfavourably oriented clay beds in
guide future pit wall designs with similar geological and conjunction with the strike of the dips (±20 degrees).
hydrogeological conditions. • Green arrow for a location where clayey facies have
dip angles less than 5 degrees;
6.2 2D Limit Equilibrium Method Stability Analysis • Yellow arrow for a location where clayey facies
have dip angles between 5 and 10 degrees; and,
2D LEM is widely used for pit wall design in the oil sands • Red arrow for a location where clayey facies have
industry. The adopted material strength parameters, dip angles greater than 10 degrees
modelled pore water pressures, and the pit wall geometry
control the LEM stability results. Figure 15 provides an example of a high potential MM clay
• Material strength parameters utilized in the stability failure due to the adverse clay dips along the final pit wall
analyses are a product of site investigations and if the strike of the pit wall is north-south. The coding criteria
laboratory testing throughout the life of mine and are consistent with shallow block failures observed at
industry experience in general. The shear strength MRM, typically developed through TFM or TFX facies
properties of oil sands are reasonably well within the MM member. At those failure locations, the
understood, and the Oil Sands industry widely uses measured dip angles for TFM and TFX were between 11
the design values (Morgenstern et al. 1988). and 20 degrees.
• Pore pressure definitions play an essential role in
pit wall stability. Due to different hydrogeological The dipmeter interpretation can identify some high-risk
conditions or the interaction between the pit wall areas with potential bench-scale failures. However, it may
and adjacent earth structures, they can vary from not capture all the risks due to borehole spacing and the
area to area. At MRM, the phreatic surface behind dipmeter data limitations (e.g., more than one
the pit wall is generally assumed to drop 50 m discontinuity). Therefore, performance observations and
behind the final pit wall at the crest and 20 m behind routine visual inspections are crucial for successfully
the final pit wall at the toe to mimic the unloading establishing the final pit wall configuration.
condition of the excavated soils.
• The pit wall geometry at MRM is often defined by 6.4 Rich Ore Flow Consideration
the overall pit slope required to satisfy the FOS
criteria for deep-seated failures through the LM2 Rich-ore flow typically occurs in bitumen-rich oil sands
clays and top of Devonian. close to the base of the McMurray formation (i.e., estuarine
• In areas with adjacent earth structures, the deep- oil sands in the MM and fluvial oil sands in the LM1), and
seated failures through UM or MM may also be can potentially compromise pit wall stability. The high
bitumen grade locations are close to the MM/LM2 and • Assessment of Pit Wall Failures: Cracking at the pit
LM1/DW contacts. A rich-ore slope (2.0 H:1V or flatter for crest, localized slumping, excessive ravelling and
the Fluvial benches) is incorporated into the pit wall design block movements should be reported to the
at locations where bitumen-rich facies are present. responsible engineer for pit walls. An assessment
of the failure mechanism and any learnings should
be documented for future reference and operations
improvement.
• Slope Buttress: The stability analyses and the field
observations reflect the best understanding of
geotechnical or geological conditions at the time.
Where instrumentation or visual monitoring
suggests poor performance or an increased risk of
pit wall failure is identified, a toe buttress may be
required.

8 CONCLUSION

A performance-based empirical approach is proposed to

design the pit wall at the MRM, supplemented with 2D LEM
stability analyses. The design approach reflects the current
experience and performance observations. The present pit
wall design slopes vary between 2.4H:1V to 2.8H:1V with
reasonable offsets from the nearby structures, improving
from the previous slope of 2.5H:1V to 3.2H:1V. The design
engineer may further optimize the pit wall slope if more
Figure 15. High-risk clay dip – red arrow borehole evidence is observed during future pit wall excavation.
7 RISK MANAGEMENT ACKNOWLEDGEMENTS
Risk management includes identifying high-risk areas by The authors gratefully acknowledge the permission of
interpreting the dipmeter data from the available boreholes, Canadian Natural Resource Limited to publish the findings
complemented with instrumentation response and field of this study, Scott Martens for the technical review and
geological mapping during mining. MRM final pit walls are Sylvain Lacroix for his assistance in the French translation
monitored using an observational approach. of the abstract of this paper.
Instrumentation results combined with field-level
monitoring will validate design assumptions or provide a REFERENCE
warning of poor performance. The monitoring plan should
be structured to provide sufficient notice to reasonably Bidwell, A., Kostaschuk, R., and Eaton, T. 2011. A case
execute a remediation plan to stabilize an underperforming history of modelling and monitoring of ground movement
area. The risk management for the pit wall generally adjacent to an oil sands mine. 2011 Pan-Am CGS
comprises the following components: Geotechnical Conf., Toronto, Ontario, Canada
Cameron, R., Danku, M. and Purhar, G. 2008. Oil sands
• Excavation of Final Pit Walls: The final excavation mine pitwall design and performance at Syncrude. The
limits should be surveyed and compared with the 61st Canadian Geotechnical Conference and 9th Joint
designed pit shell. Standard practices (e.g., Groundwater Conf., Edmonton, Canada, Sept. 21-24.
establish and dig limit and fall-down slope on the Chen, H. J., Charron, R., Hepp, D., Jeffrey, M., Chen, J.,
upper bench) should ensure the final pit limits do not and Stuart, M. 2016. Pitwall Design and Performance at
exceed design limits. Any deviation from the design Suncor North Steepbank Extension Mine. Vancouver, BC,
will be assessed and documented. Canada
• Surface Water Management System: Management Hoda, R., Iqbal, S., and Martens, S. 2016. In-pit Dykes
of surface water, including seepage from the pit Construction Practices At Shell's Muskeg River and
wall, is a critical component of performance
Jackpine Mines, CAD 2016 Annual Conference, Halifax,
management. Also, ponded water should be
NS, Canada
avoided on the pit wall benches.
• Visual Monitoring During Mining: Monitoring interim Hoda, R., Li, S., Waldick, J., and Martens, S. 2019. In-pit
pit walls during mining is intended to help keep Dykes Design and Performance Observations at the
mining equipment and staff safe. Monitoring should Muskeg River Mine, CAD 2019 Annual Conference,
include regular visual inspections of the pit wall and Calgary, AB, Canada
pit crest by a qualified geotechnical engineer. Any Morgenstern, N. R., Fair, A. E., and McRoberts, E. C. 1988.
evidence of movement should be reported to the Geotechnical engineering beyond soil mechanics - a case
responsible engineer for additional assessment. study. Canadian Geotechnical Journal, 637-661.
Peck, R. B., 1969. Advantages and Limitations of the
Observational Method in Applied Soil Mechanics. The
Ninth Rankine Lecture, Geotechnique, Vol. 19, No. 2.
Wyllie, D. C., & Mah, C. W. (2004). Rock Slope Engineering
Civil and Mining (4th edition).
Shi, R., McLachlan, D.J., Lee, C., Walter, D.J., MacNeil, S.
and MacGowan, T. 2021. Suncor's High Pit Wall Study Part
1: Investigation and Performance Monitoring Results, the
74th Canadian Geotechnical Conference, Ontario, Canada
Insights into “medium term” borrow pit
stability for slopes cut in high plastic clay
Bryce Marcotte, Adam Leik & Calvin Boese
SRK Consulting Canada, Saskatoon, SK, Canada

David Elwood,
University of Saskatchewan, Saskatoon, SK, Canada

Travis Pastachak, Brent McFarlane, Jason Bell

New Gold Inc., Emo, Ontario, Canada

ABSTRACT
Temporary excavations in soft clay soils are prevalent across Canada to enable efficient construction and infrastructure
placement. Vertical or near-vertical cuts can be safely made in low-hydraulic conductivity plastic soils in the short term;
however, these steeper slopes may eventually slump due to stress changes and progressive yielding. This case study
examines two slopes that slumped during winter (not operations) in a borrow pit. A preliminary finite element model and
parametric study is used to estimate the time to failure for various slope geometries, considering the specific site
characteristics. The paper aims to improve material and borrow geometry understanding for future excavations. This
proactive approach enables better planning and management of similar borrow pits, ultimately enhancing excavation
efficiency, long-term planning, and safety.

RÉSUMÉ
Les excavations temporaires dans les sols argileux mous sont courantes au Canada pour permettre une construction et
un placement d'infrastructures efficaces. Des coupes verticales ou presque verticales peuvent être réalisées en toute
sécurité dans les sols plastiques à faible conductivité hydraulique à court terme ; cependant, ces pentes plus abruptes
peuvent finalement s'affaisser en raison de changements de contraintes et de fléchissement progressif. Cette étude de
cas examine deux pentes qui se sont affaissées pendant l'hiver (hors exploitation) dans une fosse d'emprunt. Un modèle
préliminaire par éléments finis et une étude paramétrique sont utilisés pour estimer le temps d'échec pour diverses
géométries de pente, en tenant compte des caractéristiques spécifiques du site. L'objectif de l'article est d'améliorer la
compréhension des matériaux et de la géométrie des emprunts pour les excavations futures. Cette approche proactive
permet une meilleure planification et gestion de fosses d'emprunt similaires, améliorant ainsi l'efficacité des excavations,
la planification à long terme et la sécurité.

1 INTRODUCTION critical to ensure performance is maintained throughout the

design life.
Engineers have recorded unsatisfactory performance
in cut clay slopes for several centuries (Gregory 1844, The challenge with regards to most clay excavations
Skempton, 1948;). For “short-term” excavations, the Factor lies within the definition of “short” and “long” term. The time
of Safety (FoS) can be evaluated using the undrained to failure within a short-term excavation is first, highly
strength parameters in a simple Limit Equilibrium (LEM) or dependent on the nature of the material (mainly the in-situ
numerical analysis (CFEM, 2006; Graham, 2006). The state of stress and hydraulic properties) (Vaughan and
undrained strength, as measured from a conventional Walbancke, 1973). Secondly, the local geology plays a key
compression or shear test, may not represent the stress role as the soil within the slope returns to an equilibrium
path of a slope cut (Leroueil, 2001), however, experience after the removal of stress (Bromhead and Dixon, 2000).
has shown that this methodology is sufficient for most Conventional LEM analysis considers all elements
engineering problems. along the “failure” surface to be at equal strength (drained
The “long-term” stability of clay slopes should be or undrained) as defined by the material properties at the
estimated using effective stress parameters with point of limiting equilibrium failure (FoS~1a). This assumes
equilibrated pore pressures. This too can be approximated the change from stable to unstable is binary (rigid-plastic
with a LEM model, provided that the assumed material behaviour), and the engineer does not require/know
properties for the slope are conservative for the application information regarding the stress-strain behaviour for
– coupled with an appropriate Factor of Safety (FoS). The analysis. Proper pore pressure assumptions and
main challenge with “long-term” stability is choice of shear operational shear strength estimates are critical to a
strength parameters for structured clays. reasonable solution (Cooper et al., 1998).
As identified by Skempton (1948), the steepest natural However, most slope failures in clay slopes are
slope in a London Clay is about 5H:1V (10 degrees), progressive. As outlined by Lo (1971), progressive refers
despite “stable” slope cuts of 3H to 4H:1V standing for over to both space and time. Consider the following idealized
50 years. Proper selection of an “operational” strength is stress-strain curves for a number of different clay soils.
 Given all the complexities with cuts in clay slopes, the
“short” term stability and “long” term stability can generally
be evaluated well with current modelling procedures.
Geotechnical design challenges arise when slopes are cut
at angles steeper than the long-term stable angle but are
required for only a temporary period of time. This time,
coined “medium” term in this paper requires careful
consideration of the soil behaviour and geometry of the
excavation (with consideration for both risk and
consequence of failure). The purpose of the study was to
provide guidance to ensure a borrow pit could remain
functional for multiple seasons.

2 SITE BACKGROUND

The site is operated by New Gold Inc., an intermediate

Canadian based mining company with core assets based
in Canada. The Rainy River mine is located south of
Kenora, Ontario in the Lake of the Woods area. A brief
Figure 1. Typical stress strain behaviour of clay soils (from discussion of the general geology and geotechnical units,
Lo , 1972) as related to the borrow pit, are given below.

Curve I represents behaviour of a strain hardening 1.1 Geological Units

material, such as till soil with low clay content such that the
shear behaviour is dominated by the coarser fraction. After Starting from the lowest unit, undulating bedrock of the
reaching a peak shear resistance (denoted by A) (which Canadian Shield is present on site and visible from surface
could be defined by the maximum stress obliquity) the soil outcrops (390 m), or as low as ~310 m below overburden
continues to increase in resistance with additional strain. If soils within the Tailings Management Area (TMA). Soils
the slope cut geometry permits all elements to achieve the within the TMA were deposited as a result of two main
peak shear strength along the surface, and continue to glacial advances: the Labradorean and Keewatin sectors
harden, then a global failure is unlikely to occur (although of the Laurentide ice sheet (Ontario Geological Survey,
a serviceability problem may arise depending on the 2001; BGC 2021). Sequencing of the geological units is
stiffness of the soil and may require consideration provided in Figure 2 as shown in SRK’s 3D geological
depending on the application). model.
To the contrary, curve VI represents strain-weakening
behaviour. Upon reaching a peak strength, the soil strength
reduces with increasing strain. This can be observed in
overconsolidated clays, or normally consolidated clays with
high clay contents as the clay particle reorientation results
in defined shear zones within the clay (Burland, 1990).
Limited discussion with regards to the volume change
and change in pore pressure are provided in this paper, but
it is important to note that different clays can exhibit
dramatically different responses, for example the Leda
clays of the Eastern provinces (Lefebvre, 1981). Structured
Leda clays of eastern Canada may compress during strain
weakening further generating positive excess pore
pressures. On the contrary, consolidated clays (ex: London
Clay) tend to increase in volume, resulting in a decrease in
pore pressure after reaching the peak strength. These
characteristics play a large role in the rate of movement as Figure 2. Generalised geotechnical geology of the Tailings
the slope progresses to equilibrium conditions. Management Area at the Rainy River mine
Excavation in a clay soil with low hydraulic conductivity
occurs in an undrained state (ie: approximately no volume In a brief summary, the geological history of the TMA
change occurs). Each cut into the clay soil reduces the total area on site is as follows (from bedrock to surface):
stress on the soil elements below it. To compensate, the • WST: Labradorean ice sheet advance towards the
pore water pressures must go negative to the ambient southwest deposits coarse Whiteshell Till (WST) along
condition. The soil, despite having less total stress acting the bedrock. The WST is variable thickness and is
on it, remains (approximately) at the same effective stress, found in bedrock troughs or depressions crosswise
due to a change in pore water pressure (Heureux et al., and on the leeward side of the bedrock highs.
2007; Potts et al., 1997) Thicknesses within the site area typically range from 0
to ~15 m. The WST can be observed to daylight at
 bedrock highs around the site and acts as a confined Table 1. Approximate properties of Whitemouth Lake Till
aquifer – artesian conditions are observed in some present at the Rainy River TMA (BGC 2021, SRK 2023)
locations.
• WYL: The retreat of the Labradorean ice sheet likely Characteristics (%) Mean Range
resulted in a pro-glacial lake over the site area in the
Water content1 37 20-50
period prior to the Keewatin ice advancement. The
Liquid limit 68 24-82
unit, coined the Wylie formation (WYL) is rhythmically
layered sediments, progressing from typically sands Plastic limit 21 12-35
and silts near the bottom (Labradorean sourced) to Clay 64 59-74
increasing more Keewatin ice derived clays near the Peak Phi’ 13-33
surface, reflected by a change in mineralogy Post-Peak Phi’ 12-24
(Keewatin clays containing high volume of expansive
Residual Phi’ 6-10
clay minerals).
• WML: Movement of the Keewatin glacier from west to Void Ratio, e 1.06 0.58-1.46
east resulted in the deposit of Whitemouth Lake Till Coefficient of compression, Cc 0.38
(WML) across the site in a buoyant environment. The Coefficient of recompression, Cr 0.12
clay is a highly plastic montmorillitic soil and is typically Hydraulic conductivity 5.5x10-10 2x10-9 – 3x10-11
overconsolidated (OC) from depths of 4-10 m (due to
desiccation/weathering) and normally to lightly Multiple thin walled Shelby samples from various depths
overconsolidated at greater depths. collected in 2023 have been tested in conventional, CIUC
• BRE: The Brenna layer, deposited within the retreat of triaxial testing. Additionally, direct simple shear, direct
the Kewaatin ice and emergence of the Lockhart shear (with shear reversals), and ring shear testing has
Phase Glacial Lake Agassiz, has variable grain size been completed to fully characterise the behaviour at large
and characteristics due to its dynamic depositional strains. The WML stress-strain properties can be described
environment. It has generally been observed to be as ductile strain-softening.
spatially continuous below elevation 375 m, with
thickness ranging from 0.5 m to 6 m. The depositional
environment varied both temporally and spatially,
resulting in different types of BRE – classified as both
low and high plasticity clay.
• Organics: Thick fibrous peat deposits and topsoil can
be observed around site.

At the Rainy River mine site, organics and BRE are

removed to expose the WML for borrow development. The
WML is used for dam core construction due to its
consistency and low hydraulic conductivity.
Meanwhile, the Whiteshell Till (WST) acts as a confined
aquifer, recharged from areas with bedrock highs and
daylighting WST. Groundwater is present at or near
surface as observed in piezometers, boreholes, and many
local lakes and streams evident in the area. Although the
WST has been observed to be artesian in some areas, the
transmissivity has been observed to be generally low.

2.1 Geotechnical Properties

Extensive work has been performed characterising

both the BRE and WML units as they relate to TMA
performance. New Gold has supported several site
investigations, which have been conducted for both the
TMA, waste rock, and overburden stockpiles. Additionally,
test embankments and numerical modelling has been
undertaken and supported on site (Etezad et al. 2019,
2022)
For the purpose of this study, only the WML will be Figure 3. Results of 2023 Triaxial Compression tests
described in detail. Index properties of the clay are given in (black) along with calibrated Mohr-Coulomb linear elastic,
Table 1. hardening-softening model (red)

The behaviour observed indicates strain-softening in both

over consolidated and normally consolidated samples.
Typically, a peak friction angle (calculated with the
assumption of zero cohesion) between 20-25 degrees is procedure. Instability of the clay is well understood on site,
observed as shown in Figure 3, followed by strain softening with the observational method guiding construction
to a post-peak friction angle around 12-18 degrees. practices.
Two slope failures were noted to occur after the
Slickensided surfaces are present across the site and vary construction season was complete and prior to the borrows
in size of 10 mm to approximately 10,000 mm in size. They being inundated with tailings (it is expected other borrow pit
are assumed to be discontinuous and occur at a number of slopes have also failed, but were inundated with tailings
angles, the majority being subvertical. The shear strength and water, preventing observations).
along the slickenside is assumed to be at residual, Two borrow pits were excavated in 2022, both with
although the bulk strength is assumed to be a combination similar geometries, but to different final depths. The NW
of the peak-post-peak-residual value (ie: “operational”) borrow was sequentially excavated to a final total depth of
(Skempton 1985). ~10 m below original ground as compared to the NE
borrow, down to only a final depth of ~5-6m. The geometry,
as well as the outline of the slumped slopes, are shown in
Figure 5 for the NW borrow. No slope failures or signs of
instability were noted in the NE borrow.

Figure 5. Topography of North West borrow pit with

approximate locations of slump locations highlighted with
dotted lines.

The East slope failure occurred first, approximately

occurring in early November. The failure is located in a
section with an over steepened toe slope. As the borrow pit
was no longer required, the contractor removed a final strip
of clay late in the season and did not regrade the borrow
after excavation, as they did at the South, West and North
slopes.
The North slope failure timeline is uncertain – however,
field observations using microdrones indicate it likely
Figure 4. Images of slickensides observed in WML borrow.
occurred around March-April (2023), nearly 1 year after
excavation had begun. The failure appears to have
At Rainy River, an “operational strength” of 12-15 degrees
occurred in sequences that indicate a series of progressive
is typically used in Limit Equilibrium Modelling (LEM) based
slides.
on the laboratory testing and assessment of onsite
Survey data from Fall and Spring on the East slope
performance (Etezad et al. 2019, 2022).
indicate a final slope angle of 8-10 degrees, as shown in
Figure 6. This demonstrates that the slope progressively
1.2 Borrow Pit Geometry
failed towards a near residual value for the clay soil.
A number of clay borrow pits have been constructed over
the history of the site to construct the clay core tailings
management area (TMA). The past borrow pits were
operated for one season and dug within the TMA footprint
with the expectation that tailings would later infill them.
Local construction experience gained from other
excavations on site, largely the extensive overburden
removal from the open pit, informed the excavation
Figure 6. Profile of E slope before, with approximately
between 15–18-degree slope, and after slump with
approximate estimate slope angle of 8 degrees shown. The main limitation with the model is the degradation of
shear stiffness once plasticity occurs along the shear band.
3 NUMERICAL MODEL The model fails to account for any localised dilation and
negative excess pore pressure, and generally results in an
A numerical model was developed to better understand the overall “failure” soon after yielding occurs. The model also
time to failure within the borrow pits. It is well understood does not account for any rate effects on the shear strain of
that in the “very short term”, vertical 10-15 m high slopes the clay (viscous effects) (Renzo and Alfaro, 2023).
can remain stable. In the “short” term, a slope of 3H:1V may
be appropriate. In the “very long term” a slope of 11H:1V 3.2 Boundary Conditions
may be reasonable (ie: the slope is at ~residual conditions)
Skempton 1969, 1948). Prior to the start of WML excavation, BRE and topsoil were
The numerical model was used to assess the “medium” stripped and removed. In the early spring, the borrow pit,
term stability. Due to the increase in TMA area and was only a few metres deep, was completely filled with
reduction of available clay sources, it is advantageous for water – indicating the assumption of saturated conditions
the mine to keep current (2023) borrow pits open for a to be reasonable. Onsite observations indicate a
second season. Prevention of any slope instabilities groundwater level that can be assumed to be near ground
increases excavation and haul efficiencies in 2024. surface (assumed at 2m below ground surface in model).
The underlying WST aquifer unit has been observed to
3.1 Material Model have variable artesian conditions across site. Observations
during stage 5 borrow excavations indicate that the aquifer
Strain softening soil models are inherently difficult to may have a relatively low transmissivity (due to the thin
model, especially when the formation of discrete shear deposition along the competent bedrock surface). In the
bands is expected (Zabolotnii et al., 2021). The choice of area, it has been observed to have piezometric total heads
model for high-plasticity strain softening clay should be corresponding to the ground surface to 5 m above the
chosen carefully based on the expected timelines and original ground surface. Despite the assumption of low
material behaviour represented. As an expected “failure” transmissivity, the bottom boundary condition was
was expected, the Mohr-Coulomb hardening-softening modelled as a total head equal to the initial water table (ie:
model was used. Model parameters are given below in 2m below ground surface). This boundary condition
Table 2. provides accelerated pore water recharge to the clay
below.
Table 2. Approximate properties of Whitemouth Lake Till The bottom boundary condition is relatively challenging
present at the Rainy River TMA to assign. Although a high total-head boundary condition
results in increased rate of pore water pressure, it also
Characteristics (%) Value results in yielding elements near the base. Observations on
site indicate that when nearing the base of excavations, the
Peak Phi, c’ 21°, 3 kPa
hydraulic pressures are released and the WST aquifer
Post-Peak Phi, c’ 15°, 1 kPa pressures slowly dissipate and can be managed by
Elastic Modulus 7,500 kPa pumping. Therefore, it is anticipated that during excavation,
Ko 0.8 the total head in the WST aquifer in the vicinity of the
Eo 1 borrow pit slowly decrease, and the final aquifer pressure
ksat (field scale) 5x10-9 m/s below the borrow pit is likely a total head 0-2 m above the
base of the pit. However, the assumption of a high aquifer
pressure represents the most conservative case, despite
the improper mechanism of yielding elements along the
base (as the clay likely behaves as a discontinuum C, however, significant displacements do not yet occur as
resulting in fractures that act to release the aquifer the porewater pressure equilibration has not occurred. As
pressure). observed in Figure 7-D, the water table rises slightly
reducing the effective stress. This causes a progressive
3.3 Model softening from the base of the slope that progresses
towards the crest.
The model was created using Geostudio’s Sigma/w. A The FEM model was calibrated (by tweaking the
singular unit of WML clay was modelled using the material hydraulic conductivity primarily) to correctly predict the
properties given in Table 2. failure time observed on site. The FEM model predicts the
The excavation was modelled by removing lifts of soil, general location of the observed slip reasonably well with
approximately equal to the depth and timelines observed field observations. The model is further validated/calibrated
on site. Each sequence was modelled for 20 days to allow by the stable West and South slopes that also remained
for swelling and pore pressure changes to occur prior to the stable in the modelling procedure at the observe timeline
next excavation sequence. (as shown in Figure 9).
A bottom total head equal to the hydrostatic porewater As the primary objective was to better inform borrow pit
pressure (2m below ground surface) provided inflow into excavations in future years, numerous sensitives were
the model. conducted by varying the slope angle and thickness above
“Failure” was defined by observation of yielded the underlying hard-stratum – as both have a significant
elements, significant shear strains, and toe displacement. effect on the time to yield.
Model convergence dropped off significantly once yielding
occurred as large-strains could not be handled by the
modelling software.

4 RESULTS AND DISCUSSION

The FEM model was first calibrated to the known East

slope slump, and for conciseness, will be shown in this
paper. The results of the analysis are shown below in
Figure 7.

Figure 8. Slope angle sensitivity analysis with constant

base thickness of 3m above the underlying aquifer (shear
strains shown as heat map, days given since start of
excavation).

Unsurprisingly, significant improvement can be made by

reducing the slope angle, as shown in Figure 8. However,
the overall depth of the excavation, in relation to the total
depth is significant (as noted by L’Heureux et al. 2009).

4.1 Sensitives

The results of the sensitivities, as well as the observed

stable/unstable slopes is provided in Figure 9. This chart
represents the conditions (depth, material properties,
boundary conditions) at this specific site, and would not
Figure 7. Back analysis of East slope failure geometry apply to other sites.
(shear strains shown as heat map, days given since start
of excavation).

As noted in Figure 7, each excavation sequence results in

increasing shear strains, with shear strains concentrating
at the toe of the slide. Strain softening occurs in Figure 7-
 geological differences, such as the presence of sand
seams which increase the excess porewater pressure
dissipation (or rather, recovery) rate. Sand seams, as a
result of glacial thrusting, are present at Rainy River, and
this is considered a potential risk in the current modelling.
A decrease in underlying aquifer pressures would have
the same effect as a lower clay hydraulic conductivity -
which would reduce equilibration rate and improved
stability in the medium term. A reduced clay thickness (T)
would likely be appropriate if a reduced WST pressure was
observed, but this was not evaluated in the current model.

4.2.2 Uncertainty in Material Properties

There are numerous “knobs” to turn with regards to

material properties, namely, the Ko value, (Postill et al.,
2021), modulus, and the choice of strength parameters.
Sensitivity analysis completed as part of this study
indicated that the models are less sensitive to Ko and
modulus values when compared to the bottom boundary
and hydraulic conductivity (factors controlling the
recharge), as noted by others (Vaughan and Walbancke,
Figure 9. Results of sensitivity analysis on slope angle and 1973). The models were not evaluated for sensitivities to
base thickness. Applicable for a 10 m deep excavation, in mesh size, although this is known to play a significant role
WML clay, with artesian aquifer conditions. in strain-weakening material models (Zabolotnii et al.
2021).
As observed, the angle of the slope plays a key role in A significant uncertainty at Rainy River is the orientation
the time until failure. The model highlights the importance and pervasiveness of the slickensides planes. The initial
of ensuring sufficient depth of clay is present below the static equilibrium will depend on both – as will the
current cut. The additional clay thickness (T) acts to reduce subsequent rate of shear softening. This is a highly difficult
the equilibration of excess porewater pressure upwards parameter to assess and has not been fully captured here.
within the slope and drives the yielding of the clay deeper
along the rigid stratum below, both increasing the time to 4.3 Lessons Learned
failure. As observed in this model, an approximate 5x The adage “all models are wrong, but some are useful”
increase in failure time was noted for each additional metre applies to this paper. Through this modelling exercise, the
of clay present along the base of the excavation. following lessons were learned to guide the future
The known slope failures, shown as East and North in excavations in soft clay at the mine:
Figure 9, fit the prediction well, although the actual depth
• Where possible, the depth of excavation should be
of clay remaining at the base (T) is uncertain and estimated
minimised (ie: excavate wider bench cuts when
based on the geological model developed. Furthermore,
efficient to do so)
the geometry of the North slide, including rockfill placement
• Slopes should be graded to 6H:1V to 7H:1V for longer
on sides, may introduce significant 3D effects not
term excavations.
evaluated in this paper.
The South & West slopes, shown in Figure 9 did not • Slopes should be track-packed and sealed as best
fail. Based on the final slope angle, it would not be possible from infiltration and localised ponding.
anticipated for failure to occur prior to the borrow pit being • Vertical cuts at the toe of the slopes should be
inundated with tailings. regraded or backfill as soon as possible. Out-of-spec
material can be used to buttress the toe of slopes.
4.2 Comment on Model • Test pits should be conducted after each bench to
ensure a minimum thickness of clay is present above
4.2.1 Uncertainty in Hydraulic Conductivity and the underlying aquifer.
Geology • Excavate and load trucks away from toe of slopes.
These guidelines were developed based on the model and
L’Heureux et al. (2009) reported that the rate of slope in conjunction with New Gold’s operations to ensure
failure is dependent on the coefficient of volume change efficient excavation, while being cognisant that the borrow
(cv), which is ultimately a commentary of the hydraulic pits will likely require re-access in 2024.
conductivity of the clay soils. The cv relates the equilibration Although the developed model provides confidence in
of porewater pressures, resulting in the progressive shear the safe excavation of future borrow pits, it must be
softening. Either an incorrect assumption in material emphasized that diligent monitoring and observations is
properties or geology could result in a change in the time fundamental to both improve site understanding and
to yield. As noted by Potts et al. (1997), the likely difference continued calibration of the geotechnical model (Peck,
in failure times in London clay can be attributed to minor 1969). With regards to these slopes, they are not deemed
to be high enough risk/size to warrant detailed monitoring, papers (Etezad 2019, 2022), and the greater geotechnical
such as slope inclinometers or piezometers. It is also committee benefits from their transparency and willingness
anticipated that the slopes will be rather slow moving, like to participate in more academic activities.
the results observed in Cooper et al. (1998). However, the
use of change detection surveys – such as drone images, 7 REFERENCES
can be a valuable tool in evaluating and responding to
changes over time. Bromhead, E.N. and Dixon, N., 2000. Discussion: Delayed
collapse of cut slopes in stiff clay. Géotechnique, 50(2),
5 CLOSURE 203-205.
BGC. 2021. TMA Foundation Characterization and
Stability of cut slopes in soils must be considered carefully Geotechnical Parameters. Report Rev. 1 (Internal)
to ensure a safe and efficient excavation. In the short term Burland, J.B., 1990. On the compressibility and shear
in low-conductivity geomaterials, slope cuts can often be strength of natural clays. Géotechnique, 40(3),329-378.
near vertical and provide satisfactory performance – often Cooper, M.R., Bromhead, E.N., Petley, D.J. and Grants,
measured and approximated using the undrained shear D.I., 1998. The Selborne cutting stability experiment.
strength. In the “short” term, the strength is unlikely to be Géotechnique, 48(1), pp.83-101.
less than a properly assessed undrained strength value. Etezad, M., Riddell, B., Bocking, K., Becker, D., Hamdani,
The long-term stability is assessed using the effective A., Buchana, P., DaSilva, F. 2019. Interpretation and
stress parameters. Appropriate effective stress parameters analysis of a test embankment on soft clay. Geo St.
must be used depending on both the timeline of the John’s 2019.
excavation and allowable strains. Etezad, M., Riddell, B., Bocking, K., Becker, D., Steve, B.
Challenges arise when assessing slopes for “medium” Pastachak, T., Bell, J., Wimmelbacher, E. 2022. Three-
term stability, as a halfway point between the short term dimensional effect on a large stockpile slope stability
and long-term. Medium-term slopes are typically cut using analysis. GeoCalgary 2022.
local experience, rule of thumb, and approximate Graham, J., 2006. The 2003 RM Hardy Lecture: Soil
estimations. They cannot be properly assessed using a parameters for numerical analysis in clay. Canadian
limit equilibrium method (only very approximate results can Geotechnical Journal, 43(2), pp.187-209.
be obtained, if a correct shear resistance is selected). Gregory, C.H. 1844. On railway cuttings and embankments
In this paper, two failures in slopes cut in high-plasticity with an account of some slips in London Clay, Min.
strain weakening clay are presented. A model was Proc. I.C.E. Vol 3, 133.
calibrated to the failures and used to conduct a sensitivity Lefebvre, G., 1981. Fourth Canadian Geotechnical
analysis to guide further excavations. The following Colloquium: Strength and slope stability in Canadian
lessons were learned through the modelling procedure and soft clay deposits. Canadian Geotechnical Journal,
observations on site: 18(3), pp.420-442.
• Shallow slope angles can significantly reduce the rate Leroueil, S., 2001. Natural slopes and cuts: movement and
of strain softening; failure mechanisms. Géotechnique, 51(3), pp.197-243.
• Steeper slopes at the base of excavations should be Lo, K.Y., 1972. An approach to the problem of progressive
avoided, or replaced with additional material; failure. Canadian Geotechnical Journal, 9(4), 407-429.
• Excavations depths should be reduced, as much as L’Heureux, J.S., Leroueil, S. and Laflamme, J.F., 2009.
practical, to increase the thickness above the Evolution of the factor of safety following excavation in
underlying unit; clay. Canadian geotechnical journal, 46(5), 487-493.
• Test pitting during excavation can help employ the Ontario Geological Survey. 2001. Quaternary Geology
observational method to understand current borrow pit Fort-Frances-Rainy River Area. Report 286
geometry the depth to underlying stratum. Postill, H., Helm, P.R., Dixon, N., Glendinning, S.,
Significant uncertainty is present in the model and Smethurst, J.A., Rouainia, M., Briggs, K.M., El-
assumptions, and the time-rate results of this paper should Hamalawi, A. and Blake, A.P, 2021. Forecasting the
be considered applicable only to the site described long-term deterioration of a cut slope in high-plasticity
(although the lessons learned may be broadly applied to clay. Engineering Geology, 280, p.105912.
other similar sites). Potts, D.M., Kovacevic, N. and Vaughan, P.R., 1997.
Despite the best-efforts to calibrate and predict “failure” Delayed collapse of cut slopes in stiff clay.
times of clay slopes, geological uncertainty introduces Géotechnique, 47(5), pp.953-982.
significant error in the predicted timelines. As is often in this Skempton, A.W. 1948. The rate of softening in stiff fissured
profession, the best approach is to continue to model, but clay, with special reference to London Clay, Proc 2nd
not forget the importance of diligent site observations and Int. Conf. Soil. Mech. Found. Engng. 50-53.
re-calibration of our models (Peck, 1969). SRK. 2023. TMA Site Investigation 2023 (Draft, Internal).
Vaughan, P.R. and Walbancke, H.J., 1973. Pore pressure
6 ACKNOWLEGMENTS changes and the delayed failure of cutting slopes in
overconsolidated clay. Géotechnique, 23(4), 531-539.
The authors would like to acknowledge the support of New Zabolotnii, E., Morgenstern, N.R. and Wilson, G.W., 2021.
Gold Rainy River (NGRR) mine for the continued support Mesh sensitivity in numerical models of strain-
in development of this paper. NGRR has been highly weakening systems. Computers and Geotechnics, 136,
supportive in initiatives such as this paper, and previous p.104253.
Updated Understanding of landslides’
kinematics along Thompson River Valley
Amir Soltanieh
Hatch, Edmonton, Alberta, Canada
Renato Macciotta Pulisci
University of Alberta, Edmonton, Alberta, Canada

ABSTRACT
Landslide events are one of most common phenomena which can cause catastrophic consequences and damages for the
area that hosts these events. Monitoring both natural and man-made slopes is one of the most effective and common
methods to mitigate the negative effect of the possible slope failure. Monitoring slopes can help engineers to prevent those
events or to plan for any repair schedule for infrastructures. Early warning of landslide events also could be alert risk
management teams to decrease damages and avoid possible fatalities.
Remote sensing is one of the most effective methods for monitoring, specially in remote areas and can be employed to
monitor slope’s activity in different infrastructures such as open mines, dams, roads, and railways to have accurate
estimation of ground movement. InSAR technology, as a remote sensing method can be used to estimate ground
movement with high accuracy (mm/year) to monitor and have early warning of unordinary movements. In this research,
InSAR data helps to update understanding of landslides’ movement along the Thompson River valley, BC, Canada. This
corridor hosts several relatively active landslides. The real movement of landslides along the valley is estimated using two
different orbits of Sentinel-1 Satellite in this research, and the result is verified using the existed GPS data for Ripley
landslide which is the fastest landslide along the valley. Then the kinematics of landslides is updated using the results of
defined method as well as previous results at the same area.

RÉSUMÉ

La télédétection est l'un des méthodes les plus efficaces pour surveiller l'activité des pentes, en particulier dans les zones
éloignées. Elle peut être utilisée pour surveiller l'activité des pentes dans différentes infrastructures telles que les mines à
ciel ouvert, les barrages, les routes et les chemins de fer afin d'avoir une estimation précise du mouvement du sol. La
technologie InSAR, en tant que méthode de télédétection, peut être utilisée pour estimer le mouvement du sol avec une
grande précision (mm/an) afin de surveiller et d'avoir une alerte précoce des mouvements inhabituels. Les données InSAR
permettent de mettre à jour la compréhension du mouvement des glissements de terrain le long de la vallée de la rivière
Thompson. La vallée de la rivière Thompson, traversée par les chemins de fer CPR et CNR héberge plusieurs glissements
de terrain relativement actifs. Le mouvement réel des glissements de terrain le long de la vallée a estimé à l'aide de deux
orbites différentes du satellite Sentinel-1, et le résultat est vérifié à l'aide des données GPS existantes pour le glissement
de terrain de Ripley, qui est le plus rapide le long de la vallée. Ensuite, la cinématique des glissements de terrain est mise
à jour en utilisant les résultats de la méthode définie ainsi que les résultats précédents dans la même région.

1 INTRODUCTION monitoring techniques, decision-makers can make

informed choices regarding risk management strategies
Since 1771, Canada has experienced numerous landslide and allocate necessary resources to mitigate the potential
events resulting in hundreds of fatalities, significant impact of landslides on infrastructure. (Huntley et al. 2019
infrastructure damage, and costs amounting to millions of and Rodriguez et al. 2020)
dollars. (Choe et al. 2021). Enhanced landslide risk management strategies can be
Climate change can also have a significant impact on effectively planned through the implementation of remote
landslide activity, particularly in northern regions of sensing monitoring techniques. (Deane et al. 2020,
Canada, by accelerating glacier melting and retreat. This Macciotta et al. 2020, Woods et al. 2020, Lan et al. 2021,
observation was confirmed by a study conducted by Mei et Rodriguez et al. 2021 and Macciotta et al. 2021).
al. in 2008. It underscores the importance of ongoing This manuscript presents an update understanding of
monitoring and a comprehensive understanding of the landslide kinematics using satellite InSAR, as one of most
landslide phenomenon to effectively mitigate the adverse effective remote sensing tools, for active landslides along
effects of these events not only in northern areas of Thompson River Valley south of Ashcroft in British
Canada but also worldwide. Columbia, Canada.
The monitoring of landslides can be utilized to establish Researchers have previously conducted studies on the
robust early warning systems, enabling proactive activity and kinematics of landslides in this valley,
adaptation to these events, and facilitating the highlighting its importance. (Huntley et al. 2019, Hendry et
development of suitable maintenance schedules for al. 2015, Macciotta et al. 2016, Journault et al. 2018,
infrastructure. This becomes particularly crucial when Huntley et al. 2021, Huntley et al 2019, Eshraghian et al.
stabilization or avoidance measures are technically or 2007 and Huntley et al. 2021). This paper presents an
economically challenging. By employing effective update on landslides’ kinematics based on the interpreted
stratigraphy of the same area, and InSAR monitoring data 1.2 Thompson River valley landslides
between May 2015 and May 2017. (Huntley et al. 2019,
Huntley et al. 2021, Huntley et al. 2019, Eshraghian et al. Thompson River valley, south of the town of Ashcroft
2007, Huntley et al 2020, Sattler et al. 2021 and Schafer et connects the port of Vancouver to the rest of Canada by
al. 2016). hosting Canadian Pacific Railway (CP) and Canadian
In this presentation, a practical method is proposed for National Railway (CN). (Vallee et al. 2019).
calculating the real displacement of landslides. It involves This corridor is a vital corridor for transportation Canada
combining the geometry of two InSAR orientations and and many investigation programs were planned to monitor
assuming horizontal movement parallel to the average 10 km section of this corridor where up to 14 landslides
slope azimuth in various sectors of each landslide. This have been detected (Huntley et al. 2021, Porter et al.
approach offers a useful tool for accurately assessing 2002).
landslide displacement. The stratigraphy of the ground in this area has been
The proposed method is verified using existing in-place extensively studied by previous researchers, providing
monitoring data from the Ripley landslide, previously valuable insights into the geological layers and formations.
studied by other researchers. This validation process (Journault et al. 2018, Clague et al. 2003, Huntley et al.
ensures the accuracy and reliability of the method. 2014). The landslides along this valley, was formed during
(Macciotta et al. 2015) postglacial period gradually, by Southward Thompson
river’s erosion which down-cut up to 150 m of deposited
1.1 Landslide Remote Sensing and Satellite InSAR. sediments (Journault et al. 2018, Porter et al. 2002).
The presence of a weak silt and clay layer were confirmed
Remote sensing and monitoring techniques are as the responsible factor for the instability of landslides in
increasingly used by geotechnical engineers to study this area in other investigations (Journault et al. 2018,
ground movement and gain insights into landslide activity. Clague et al. 2003, Eshraghian et al. 2007). Previous
(Xu et al. 2016). investigations have utilized back analyses to determine
Remote sensing techniques enable engineers and that a pre-sheared surface in this layer exhibits residual
researchers to study the kinematics of landslides over large strengths characterized by no cohesion and friction angles
areas where in-place instrumentation is not feasible or ranging between 9 to 16 degrees. (Hendry et al. 2015).
practical. (Rodriguez et al. 2020). Figure 1 illustrates the location and extends of known
InSAR utilizes radar sensors carried by satellites to landslides along the Thompson River valley (Hendry et al.
measure surface displacement and monitor landslide 2015). Landslides have different sizes and velocities in this
activities. This method's continuous monitoring and varying valley, between 0.75 to 15 million cubic meters and 10
resolutions make it robust for studying landslides. (Abe et mm/year (extremely slow) to 50 mm/s (very rapid)
al. 2020). Numerous investigations across various respectively (Journault et al. 2018).
research studies have consistently confirmed the efficacy
of the InSAR method for change detection purposes 2 MATERIAL AND METHODS
worldwide, spanning diverse applications. (Motagh et al.
2017, Chen et al. 2022, Liu et al. 2022, Mastro et al. 2022, This paper focuses on presenting the actual movements of
Abdel-Hamid et al. 2021, Dumka et al. 2020, Castellazzi et three highly active landslides that have a significant impact
al. 2021, Solari et al. 2020, Zhai et al. 2018, Lu et al. 2021). on railroads. Initially, the method for calculating these
The InSAR method offers several advantages, including movements is applied to the Ripley landslide and verified
high accuracy (sub-cm/year), good resolution, and its by comparing it with GPS data obtained from that specific
usefulness in areas where installing in-place ground landslide. Once validated, the method is employed to
monitoring instruments is challenging from an economic calculate the movements of other landslides. Additionally,
and technical standpoint. However, it does have limitations. the updated interpretation of landslide kinematics based on
For instance, it is not suitable for detecting rapid changes the results and previous research is investigated in this
due to the time gaps between two consecutive data study.
acquisitions, which currently range from approximately 3
days to 1 week. This limitation prevents the detection of 2.1 Published Displacement Monitoring Data at the
high acceleration in ground movement. Furthermore, Ripley landslide
InSAR is susceptible to atmospheric noise, vegetation
cover, and seasonal effects such as rainfall or snow cover, Three GPS monitoring systems' results, reported in a
which pose challenges for accurate measurements using previous investigation, are used in this paper to verify the
this method. (Choe et al. 2021, Huntley). However, it accuracy and reliability of the method. (Macciotta et al.
became a well-known method for many applications 2015). In that research, ground displacement was
recently (Mancini et al. 2021). monitored from April 2008 to October 2014, with
The abilities of InSAR for monitoring slow movements, intermittent gaps for system repairs and maintenance.
beside the other information such as precipitation or (Macciotta et al. 2015). In this paper, the InSAR method
groundwater fluctuation data, make it a robust method for and GPS data are used to verify the results, even though
understanding landslide mechanisms and its possible they were not collected simultaneously. It is assumed that
triggers (Huntley et al 2021). there were no changes in the direction and acceleration of
ground movement. Figure 2 provides the approximate
locations of the installed GPS systems on the Ripley various locations within the Thompson River Valley from
landslide. November 2014 to Spring 2018. The distance between the
ground and the satellite in the Line of Sight (LOS) direction
is recorded for both ascending and descending orbits of the
Sentinel 1 satellite. Table 1 provides the geometry of the
Sentinel 1 LOS for both ascending and descending orbits.
The table includes the angle (θ) between the LOS and the
vertical plane, as well as the angle (δ) between each orbit
and geographic north.
Figure 3 shows a schematic of the LOS geometry for
Sentinel 1 Satellite. The processed data between May
2015 to May 2017 is used in this research to detect
landslide movements.

2.3 Displacement Calculation at the Surface of

Thompson River Valley’s Landslides

The displacement recorded by the InSAR satellite

represents the displacement along the Line of Sight (LOS)
direction of each orbit, which may differ from the actual or
true displacement. The measured displacement in InSAR
is a projection of the true displacement onto the LOS axis.
To obtain a close approximation of the true displacement,
several assumptions need to be made. For simplicity, the
symbol R is used to represent the true displacement vector
in the calculations.
To calculate the true ground displacement for each area,
the movement is assumed to be parallel to the slope
azimuth horizontally. Using the topography information of
each area, the assumed unit vector of the horizontal
component for the R vector is determined. The direction
and magnitude of the R vector for each landslide can be
calculated by considering the LOS displacement vectors
from both the ascending and descending orbits of the
Sentinel 1 satellite.
This procedure is employed to calculate the R vector and
determine the real ground displacement for each area as
below.
𝑷𝑖 = (𝑹. 𝑺𝑖 )𝑺𝑖 = 𝑀𝑖 𝑺𝑖 𝐴 [1]
While 𝑆𝑖 , 𝑃𝑖 , and 𝑀𝑖 present the LOS unit vector, Projection
of R vector on LOS direction and Magnitude of
displacement along the LOS direction for each orbit
respectively.
Each LOS unit vector (𝑺𝑖 = (𝑒, 𝑛, 𝑢)) could be calculated as
follow:
𝑢 = cos 𝜃 [2]
𝑒 = sin 𝜃 . sin 𝛿 [3]
𝑛 = sin 𝜃 cos 𝛿 [4]
The coordinate system used in this paper corresponds to
(east,north,up). The unit vector for each orbit based on the
geometry is calculated as follow for both orbits:
𝑺𝑎 = (−0.613, −0.123,0.781) [5]
𝑺𝑑 = (0.692, −0.095,0.716) [6]
Figure 1.Landslides along Thompson River valley- South Replacing equation 5 and 6 in equation 1 gives following
of town of Ashcroft equations:
−0.613𝑥 − 0.123𝑦 + 0.781𝑧 = 𝑀𝑎 [7]
0.692𝑥 − 0.095𝑦 + 0.716𝑧 = 𝑀𝑑 [8]
2.2 Satellite InSAR Data The third equation to calculate R vector can be generated
from the horizontal movement’s assumption which was
The satellite InSAR data used in this research was discussed before as below:
𝑥
collected by the Sentinel 1 satellite and processed by the = tan 𝛼 [9]
𝑦
TRE Altamira Group. The data captures ground activity in
The R vector for each area now can be calculated using The LOS displacement for each sector is determined by
equations 7 to 9. averaging the LOS displacements of all scatterers within
that sector. Figure 4 provides information on the data
density of the South extension landslide, along with
selected sectors as an example. The figure demonstrates
sufficient data density from both orbits for each sector,
ensuring reliable calculations of LOS displacement.
Longer time periods (typically on an annual basis) are
selected to calculate the average R value, ensuring that
minor variations over a few days do not significantly impact
the results.

3 RESULTS

This section focuses on presenting the results of the

calculated R vectors using InSAR data collected by
Sentinel 1 for each sector of the Ripley, South Extension,
and Redhill landslides. These landslides are known to
exhibit rapid movement along the valley. The calculated
horizontal components of the R vectors are displayed in a
plan view for each sector, while the vertical-horizontal
movement components are presented in section views.
The cross-sections in these figures are constructed based
on the topography of the area, while the stratigraphy
information is derived from other research conducted in the
same area. (Eshraghian et al. 2007).

Figure 2. Installed GPS systems on Ripley Landslide

Figure 3. Satellite Geometry Schematic

Table 1.Satelite viewing (LOS) angles for the Ascending

and Descending orbits.
Sat Orbit Track Sensor Symb Angle
-llite Mode -ol (°)
Senti Ascending 64 IW θ 38.66
-nel1 δ 11.33
Senti Descending 115 IW θ 44.29 Figure 4. Satellite data density for South extension
-nel1 δ 7.79 landslide
 Figure 8 shows the cumulative LOS displacement recorded
3.1 Ripley landslide by the Sentinel 1 satellite for the South Extension landslide.
It highlights the higher activity at the toe of the landslide in
The Ripley landslide, despite being small in size, is the sections 1 to 6 compared to the areas near the scarp.
fastest in the area. The presence of recorded GPS data
adds confidence to the results obtained using the applied
method.
Due to lower data density on the Ripley landslide, careful
selection of sectors was necessary. The chosen sectors
aimed to cover the entire surface of the landslide. To verify
the method, some sectors closer to GPS locations were
also included. Additionally, areas beyond the landslide
extension were considered to gather more information
about potential retrogression in the region.
Figure 5 shows the cumulative LOS displacement for the
Ripley landslide, along with selected sectors indicating
data density from both ascending and descending orbits.
The toe area of the landslide is highly active, while a less
active region is observed to the south.
In Figure 6, the plan-view displays the horizontal
components of R vectors for each sector of the Ripley
landslide. The recorded horizontal movement of GPS
stations within the landslide is also shown. The figure
indicates ground movement ranging from 1 mm/year at the
backscarp to 81 mm/year in sector 9, closer to the river at
the landslide toe. The velocity gradually decreases away
from the river, and no significant activity is observed at the
backscarp. The results suggest slow to very slow activity
for the Ripley landslide based on the velocity classification
system. (Curden et al. 1996, Hunger et al. 2014).
These results align with previous research findings on the
Ripley landslide, confirming similar magnitudes and
directions of ground movement. The maximum velocity of
82 mm/year, consistent with previous reports, occurs in the
middle of the landslide toe. (Huntley et al. 2021).
The ground movement recorded by in-place GPS
instruments on the Ripley landslide generally agrees with
the results of the current method. However, there is a
discrepancy with GPS 3, which may be due to its
installation location on the retaining wall, as noted in
previous research. (Macciotta et al 2015). The comparison
between GPS data and the approach used in this paper
boosts confidence in applying this approach to other
landslides in the Thompson River valley.
Figure 6 also shows the cumulative movement recorded by
GPS systems installed on the Ripley landslide before.
Figure 7 shows the vertical component of calculated R
vectors for different sectors of the Ripley landslide, along
with the corresponding GPS vertical components. The
comparison confirms the accuracy of the calculated results.
Sectors closer to the river exhibit more horizontal
movement, indicating the presence of a weak clay layer, as
suggested in previous research on landslides in the
Thompson River valley. (Journault et al. 2018, Clague et
al. 2003, Eshraghian et al. 2007).

3.2 South Extension landslide

The South Extension landslide is situated south of the

South landslide, indicating a retrogressive behavior, and
earning its name due to its significant activity in comparison Figure 5. Cumulative LOS displacement for Ripley
to other sections of the South landslide. landslide
 the potential disastrous consequences on the railroads and
Figure 9 presents the horizontal components of the true overall transportation infrastructure.
displacement (R) vectors for different sectors of the South Figure 12 displays the horizontal component of the R
Extension landslide. It reveals a maximum velocity of 64 vectors for each sector of the Redhill landslide. It is evident
mm/year in Sector 4, while sectors near the scarp exhibit that the toe of the landslide, particularly in sections 8 and
lower activity. The direction of horizontal movement differs 9, exhibits higher activity with a velocity of 52 mm/year. In
between the south and north sides of the landslide due to contrast, the remaining sections demonstrate velocities
variations in slope azimuth. Overall, the results indicate low ranging from 5 to 46 mm/year during the period between
to very low activity for this landslide. May 2015 and May 2017.
Figure 10 shows the vertical components of R vectors for In Figure 13, the vertical components of the R vectors for
different sectors of the South Extension landslide. Sectors various sections of the Redhill landslide are depicted. The
closer to the scarp exhibit more vertical movement, figure reveals that there are generally parallel vertical
indicating the presence of a weak horizontal clay layer. movements across different sections of the landslide, with
Retrogression is observed in sections 4, 6, 8, and 10 on the relatively less vertical displacement observed near the toe
southern side of the landslide. Less activity is observed of the landslide.
near the scarp, consistent with Sentinel 1 radar data.

Figure 7. Vertical components for R vectors and GPS

recorded movements.

Figure 6. Horizontal component for R vectors and GPS 4 DISCUSSIONS

recorded movements.
Figure 4 displays sufficient data density for detailed
3.3 Redhill landslide kinematic interpretation of the South Extension landslide,
as well as other landslides in the area.
Although Redhill landslide is situated on the west side of
Figure 5 shows higher activity near the toe of the Ripley
the Thompson River and is not traversed by any railroad
landslide, while the southern extended area exhibits lower
companies, it still presents a significant concern as its
activity.
potential negative impacts on the operation of railroads in
The South Extension landslide, located on the north side of
the vicinity cannot be overlooked.
the South landslide, exhibits greater activity in sectors 2 to
Figure 11 presents the cumulative 2-year Line of Sight
7, while other sectors show relatively less movement
(LOS) displacements at the Redhill landslide, showcasing
according to Sentinel 1 data. The backscarp area appears
the magnitude of ground movement in both ascending and
to be inactive based on Figure 8.
descending orbits. The figure clearly depicts significant
According to Figure 11, the Redhill landslide displays
activity concentrated at the toe of the landslide,
higher activity on sectors 3, 5, 7, and 8 near its toe, while
underscoring the criticality of continuous monitoring to
the activity decreases in other parts as the distance from
establish an early warning system and effectively mitigate
the river increases. The southern side appears to be more
active than the northern side when viewed from an
ascending perspective.
Ripley landslide has the highest maximum horizontal
movement of 81 mm/year on sector 5 near GPS 2, while
the other landslides are classified as very slow-moving
according to Curden and Vernes velocity classification.
(Hunger et al. 2014, Curden et al. 1996).

Figure 9. Horizental components for R vectors on South

extension landslide

Figure 10. Vertical component for R vectors on South

extension landslide

landslide

The horizontal movement for Ripley landslide is calculated

Figure 8. Cumulative LOS displacement for South between 2 to 81 mm/year in Figure 6. In Figure 9, the
extension landslide Figure 6 shows Ripley landslide with horizontal velocities
ranging from 2 to 81 mm/year. South Extension landslide
in Figure 9 has velocities from 3 to 64 mm/year, and Redhill
landslide in Figure 12 ranges from 5 to 52 mm/year.
Figure 7, Figure 10, and Figure 13 display the vertical
components of the calculated R vectors for Ripley, South
Extension, and Redhill landslides, respectively.

Figure 12. Horizontal component of R vectors on Redhill

landslide

In the Ripley and South Extension landslides, the vertical

movement varies with the distance from the river. Sectors
closer to the scarp exhibit more vertical displacement
compared to sectors closer to the river, as observed in the
selected cross-sections. The Redhill landslide also
demonstrates a similar pattern, with more horizontal
movement near the river and less activity closer to the
scarp. These observations suggest the presence of a
weak, semi-horizontal clay layer, which has been identified
as a contributing factor to landslide failure in the area.
(Journault et al. 2018, Eshraghian et al. 2007, Clague et al.
2003).
The failure surfaces interpreted in the figures suggest a
compound retrogressive mechanism closer to the scarp,
accompanied by horizontal transitional movement on the
toe of the Ripley landslide. A similar mechanism can be
proposed for the South Extension landslide, while the
Redhill landslide on the opposite side of the Thompson
River exhibits a likely more retrogressive rotational
mechanism, possibly due to the steeper slope on that side
of the riverbank.
This paper investigates three of the most active landslides
in the Thompson River valley, Ripley, South Extension, and
Redhill landslide, in order of their velocities from fastest to
Figure 11. Cumulative LOS displacement on RedHill slowest. The Thompson River valley is recognized as an
important corridor with a history of reported landslides.
 system proposed by Curden and Vernes (1996) (as
referenced in Clague et al., 2003 and Hunger et al., 2014).
The total range of calculated movement for all the
investigated landslides in this study is reported to be
between 2 to 81 mm/year.
The maximum velocities calculated for Ripley, South
extension, and Redhill landslides are consistent with
previous research, confirming the findings in this study.
The comprehensive analysis of all components of R
vectors for different sectors provides a robust interpretation
of the kinematics of these landslides. Outside the known-
active areas, minimal to no retrogression activity is
observed for the landslides, except for the South Extension
landslide. The analysis extends to include the ground
activities behind the landslide scarps in these areas.
The approach utilized in this study combines satellite
InSAR data recorded on two different orientations with the
assumption of horizontal movement parallel to the average
slope aspect. By averaging the inconsistencies in space
and time between the different line-of-sight (LOS)
orientations, a more comprehensive understanding of the
East-West and vertical components of real displacement is
obtained.
However, it is important to evaluate the assumption of
horizontal movement on a case-by-case basis, as complex
landslide mechanisms can potentially lead to misleading
interpretations of topographic characteristics. Careful
Figure 13. Vertical component of R vectors on Redhill consideration and analysis of the specific landslide
landslide characteristics are necessary to ensure accurate results.

6 ACKNOWLEDMENT
The authors would like to acknowledge Canadian Pacific
5 CONCLUSIONS Railway (CP) and TRE Altamira for providing the InSAR
monitoring results used in this study. This research was
This paper presents an updated understanding of landslide performed originally within the umbrella of the Railway
displacement and kinematics through the analysis of Group Hazard Research Program (RGHRP), which is
InSAR monitoring data collected by the Sentinel 1 satellite. founded by Natural Sciences and Engineering Research
The data spans from May 2015 to May 2017 and includes Council of Canada (NSERC), Canadian Pacific Railway,
measurements from both ascending and descending Canadian National Railway, and Transport Canada, and is
orbits. To obtain a more detailed understanding of the a collaboration that includes Queen’s University, the
landslides' movement, each landslide is divided into Canadian Geological Survey, and the University of Alberta.
multiple sectors, selected based on data density. This paper is also supported by Hatch Company to be
To calculate the 3D movement vector (R) for each sector, presented at Geo Saskatchewan Conference 2023.
the horizontal movement is assumed to be parallel to the The paper after all reviews is improved grammatically using
average slope aspect of the sector. The displacement AI tools.
values recorded by Sentinel 1 between May 2015 and May
2017, in both ascending and descending Line of Sight 7 REFERENCES
(LOS) directions, are used to determine the horizontal
movement. Choe, B.-H.; Blais-Stevens, A.; Samsonov, S.; Dudley, J. Sentinel-1 and
RADARSAT Constellation Mission InSAR Assessment of Slope
The accuracy of the method is verified by comparing the Movements in the Southern Interior of British Columbia, Canada.
calculated results with existing GPS data from the Ripley Remote Sens. 2021, 13, 3999.
landslide. The verification process is discussed in this Mei, S.; Poncos, V.; Froese, C. Mapping millimetre-scale ground
deformation over the underground coal mines in the Frank Slide area,
paper, as well as in a previous manuscript. (Soltanieh et
Alberta, Canada, using spaceborne InSAR technology. Can. J. Remote
al. 2022). Sens. 2008, 34, 113–134.
By utilizing InSAR data and sector-based analysis, this Huntley, D.; Bobrowsky, P.; Hendry, M.; Macciotta, R.; Best, M. Multi-
study aims to provide a more comprehensive technique Geophysical Investigation of a Very Slow-moving Landslide
near Ashcroft, British Columbia, Canada. J. Environ. Eng. Geophys.
understanding of landslide displacement and kinematics, 2019, 24, 87–110.
contributing to improved monitoring and management of Rodriguez, J.; Macciotta, R.; Hendry, M.T.; Roustaei, M.; Gräpel, C.;
these hazardous events. Skirrow, R. UAVs for monitoring, investigation, and mitigation design of
a rock slope with multiple failure mechanisms—A case study.
The landslides investigated in this paper can be classified
Landslides 2020, 17, 2027–2040.
as very slow based on the calculated movement observed
in each sector and according to the velocity classification
Deane, E.; Macciotta, R.; Hendry, M.; Grapel, C.; Skirrow, R. Leveraging Dumka, K.R.; SuriBabu, D.; Malik, K.; Prajapati, S.; Narain, P. PS-InSAR
historical aerial photographs and digital photo-grammetry techniques derived deformation study in the Kachchh, Western India. Appl.
for landslide investigation—A practical perspective. Landslides 2020, Comput. Geosci. 2020, 8, 100041.
17, 1989–1996. Castellazzi, P.; Schmid, W. Interpreting C-band InSAR ground deformation
Woods, A.; Macciotta, R.; Hendry, M.; Stewart, T.; Marsh, J. Updated data for large-scale groundwater management in Australia. J. Hydrol.
understanding of the deformation characteristics of the Checkerboard Reg. Stud. 2021, 34, 100774
Creek rock slope through GB-InSAR monitoring. Eng. Geol. 2020, 281, Solari, L.; Del Soldato, M.; Raspini, F.; Barra, A.; Bianchini, S.; Confuorto,
105974. P.; Casagli, N.; Crosetto, M. Review of Satellite Interferometry for
Lan, H.; Zhao, X.; Macciotta, R.; Peng, J.; Li, L.; Wu, Y.; Zhu, Y.; Liu, X.; Landslide Detection in Italy. Remote Sens. 2020, 12, 1351.
Zhang, N.; Liu, S.; et al. The cyclic expansion and contraction Zhai, C.; Lu, Z. Remote Sensing of Landslides (A review). Remote Sens.
characteristics of a loess slope and implications for slope stability. Sci. 2018, 10, 279.
Rep. 2021, 11, 2250. Lu, Z.; Kim, J. A Framework for Studying Hydrology-Driven Landslide
Rodriguez, J.; Deane, E.; Hendry, M.; Macciotta, R.; Evans, T.; Gräpel, C.; Hazards in Northwestern US Using Satellite InSAR, Precipitation and
Skirrow, R. Practical evaluation of single frequency for monitoring slow- Soil Moisture Observations: Early Results and Future Directions.
moving landslides. Landslides 2021, 18, 3671–3684. GeoHazards 2021, 2, 17–40.
Macciotta, R.; Hendry, M.T. Remote Sensing Applications for Landslide Mancini, F.; Grassi, F.; Cenni, N. A Workflow Based on SNAP–StaMPS
Monitoring, and Investigation in Western Canada. Remote Sens. 2021, Open-Source Tools and GNSS Data for PSI-Based Ground
13, 366. Deformation Using Dual-Orbit Sentinel-1 Data: Accuracy Assessment
Huntley, D.; Bobrowsky, P.; Hendry, M.; Macciotta, R.; Elwood, D.; Sattler, with Error Propagation Analysis. Remote Sens. 2021, 13, 753.
K.; Best, M.; Chambers, J.; Meldrum, P. Application of multi- Vallee, M. Falling in Place: Geoscience, Disaster, and Cultural Heritage at
dimensional electrical resistivity tomography datasets to investigate a the Frank Slide, Canada’s Deadliest Rockslide. Space Cult. 2019, 22,
very slow-moving landslide near Ashcroft, British Columbia, Canada 66–76.
Landslides 2019, 16, 1033–1042. Porter, M.J.; Savigny, K.W.; Keegan, T.R.; Bunce, C.M.; MacKay, C.
Macciotta, R.; Hendry, M.; Martin, C.D. Developing an early warning system Controls on stability of the Thompson River landslides. In Proceedings
for a very slow landslide based on displacement monitoring. Nat. of the 55th Canadian Geotechnical Conference, Niagara Falls, ON,
Hazards 2016, 81, 887–907. Canada, 20–23 October 2002.
Huntley, D.; Rotheram-Clarke, D.; Cocking, R.; Joseph, J.; Bobrowsky, P. Journault, J.; Macciotta, R.; Hendry, M.T.; Charbonneau, F.; Huntley, D.;
Understanding plateau and prairie landslides: Annual report on Bobrowsky, P.T. Measuring displacements of the Thompson River
landslide research in the Thompson River valley, British Columbia, and valley landslides, south of Ashcroft, BC, Canada, using satellite InSAR.
the Assiniboine River valley, Manitoba-Saskatchewan 2020–2021. Landslides 2018, 15, 621–636.
Geol. Surv. Can. 2021, 8838, 3–73. Clague, J.J.; Evans, S.G. Geologic Framework of Large Historic Landslides
Eshraghian, A.; Martin, D.; Cruden, D. Complex Earth Slides in the in Thompson River Valley, British Columbia. Environ. Eng. Geosci.
Thompson River Valley, Ashcroft, British Columbia. Environ. Eng. 2003, 9, 201–212.
2007, 13, 161–181. Huntley, D.H.; Bobrowsky, P.T. Surficial geology and monitoring of the
Huntley, D.; Rotheram-Clarke, D.; Pon, A.; Tomaszewicz, A.; Leighton, J.; Ripley Slide, near Ashcroft, British Columbia, Canada. Geol. Surv. Can.
Cocking, R.; Joseph, J. Benchmarked RADARSAT-2, SENTINEL-1 2014, 7531, 21.
and RADARSAT Constellation Mission Change-Detection Monitoring Hendry, M.T.; Macciotta, R.; Martin, C.D.; Reich, B. Effect of Thompson
at North Slide, Thompson River Valley, British Columbia: Ensuring a River elevation on velocity and instability of Ripley Slide. Can. Geotech.
Landslide-Resilient National Railway Network. Can. J. Remote Sens. J. 2015, 52, 257–267.
2021, 47, 635–656. Eshraghian, A. Hazard Analysis of Reactivated Earth Slides in the
Huntley, D.; Holmes, J.; Bobrowsky, P.; Chambers, J.; Meldrum, P.; Thompson River Valley, Ashcroft, British Columbia. Ph.D. Thesis,
Wilkinson, P.; Donohue, S.; Elwood, D.; Sattler, K.; Hendry, M.; et al. University of Alberta, Edmonton, AB, Canada, 2007.
Hydrogeological and geophysical properties of the very slow-moving Cruden, D.M.; Varnes, D.J. Landslide Types and Processes, Transportation
Ripley Landslide, Thompson River valley, British Columbia. Can. J. Research Board; U.S. National Academy of Sciences, Special Report,
Earth Sci. 2020, 57, 1371–1391 247; National Academy of Sciences: Washington, DC, USA, 1996.
Sattler, K.; Elwood, D.; Hendry, M.T.; Huntley, D.; Holmes, J.; Wilkinson, Hungr, O.; Leroueil, S.; Picarelli, L. The Varnes classification of landslide
P.B.; Chambers, J.; Donohue, S.; Meldrum, P.I.; Macciotta, R.; et al. types, an update. Landslides 2014, 11, 167–194.
Quantifying the contribution of matric suction on changes in stability Soltanieh, A.; Macciotta, R. Updated understanding of the Ripley landslide
and displacement rate of a translational landslide in glaciolacustrine kinematics using satellite InSAR. Geosciences 2022, 12, 298.
clay. Landslides 2021, 18, 1675–1689.
Schafer, M. Kinematics and Controlling Mechanics of the Slow-Moving
Ripley Landslide. Master’s Thesis, University of Alberta, Edmonton,
AB, Canada, 2016.
Xu, D.; Hu, X.-Y.; Shan, C.-L.; Li, R.-H. Landslide monitoring in
southwestern China via time-lapse electrical resistivity tomography.
Appl. Geophys. 2016, 13, 1–12.
Abe, T.; Iwahana, G.; Efremov, P.V.; Desyatkin, A.R.; Kawamura, T.;
Fedorov, A.; Zhegusov, Y.; Yanagiya, K.; Tadono, T. Surface
displacement revealed by L-band InSAR analysis in the Mayya area,
Central Yakutia, underlain by continuous permafrost. Earth Planets
Space 2020, 72, 138.
Motagh, M.; Shamshiri, R.; Haghighi, M.H.; Wetzel, H.-U.; Akbari, B.;
Nahavandchi, H.; Roessner, S.; Arabi, S. Quantifying groundwater
exploitation induced subsidence in the Rafsanjan plain, southeastern
Iran, using InSAR time-series and in situ measurements. Eng. Geol.
2017, 218, 134–151.
Chen, B.; Li, Z.; Zhang, C.; Ding, M.; Zhu, W.; Zhang, S.; Han, B.; Du, J.;
Cao, Y.; Zhang, C.; et al. Wide Area Detection and Distribution
Characteristics of Landslides along Sichuan Expressways. Remote
Sens. 2022, 14, 3431.
Liu, Y.; Yao, X.; Gu, Z.; Zhou, Z.; Liu, X.; Chen, X.; Wei, S. Study of the
Automatic Recognition of Landslides by Using InSAR Images and the
Improved Mask R-CNN Model in the Eastern Tibet Plateau. Remote
Sens. 2022, 14, 3362.
Mastro, P.; Masiello, G.; Serio, C.; Pepe, A. Change Detection Techniques
with Synthetic Aperture Radar Images Experimens with Random
Forests and Sentinel-1 Observations. Remote Sens. 2022, 14, 3323.
Abdel-Hamid, A.; Dubovyk, O.; Greve, K. The potential of sentinel-1 InSAR
coherence for grasslands monitoring in Eastern Cape, South Africa. Int.
J. Appl. Earth Obs. Geoinf. ITC J. 2021, 98, 102306.
Dutchman’s Ridge: Results from the analysis
of geologic, monitoring and LiDAR-derived
data of an extremely slowly moving rockslide
Camila Troncoso Klein & D. Jean Hutchinson
Department of Geological Sciences & Geological Engineering –
Queen’s University, Kingston, Ontario, Canada
Thomas Stewart & Julia Marsh
Dam Safety, Reservoir Slopes – BC Hydro, Revelstoke, British Columbia, Canada

ABSTRACT
Dutchman’s Ridge rockslide is located along the Upper Columbia River, 1.5 km upstream of Mica Dam in British
Columbia. It is underlain by a Basal Fault and was recognized during investigations in the 1960s. Displacements up to
25 mm/year of a 115 million m3 rock mass started after reservoir filling, which were identified in a 1984 review. A
drainage adit was built in 1986-1988, decreasing water pressures and, subsequently, displacement rates. This paper
proposes new boundaries, according to geological and geotechnical evidence, and that the instability is not a single
homogeneous mass, but presents several identifiable zones. The slope is dissected by a feature, that could be the
superficial expression of Lavender Shear, recognized during adit construction. The geometry and shear strength of the
Basal Shear Zone were approximated using Rocscience’s Slide2. Ongoing work using finite element program RS2
explores the effect of increasing water pressures due to a hypothetical loss of drainage capacity.

RÉSUMÉ
L'éboulement Dutchman's Ridge est situé dans le cours du fleuve Columbia Supérieur, 1,5 km loin du barrage Mica. Il
repose sur une faille basale et a été reconnu lors d'enquêtes dans les années 1960. Des déplacements jusqu'à 25
mm/an d'un massif rocheux de 115 millions de m2 ont commencé après le remplissage du réservoir, e ont été identifiés
dans une étude de 1984. Une galerie de drainage a été construite en 1986-1988, diminuant les pressions d'eau et, par
la suite, les taux de déplacement. Cet article propose de nouvelles limites, selon des preuves géologiques et
géotechniques, et que l'instabilité n'est pas une seule masse homogène, mais présente plusieurs zones identifiables.
La pente est disséquée par une caractéristique en forme de crête, qui pourrait être l'expression superficielle du
cisaillement Lavender, reconnu pendant la construction de la galerie. La géométrie et la résistance de la zone de
cisaillement basale ont été approximées à l'aide du Slide2 de Rocscience. Les travaux en cours utilisant le programme
d'éléments finis RS2 explorent l'effet de l'augmentation de la pression de l'eau due à une hypothétique perte de capacité
de drainage.

1 INTRODUCTION Donati et al., 2020) and different domains within the

deforming slope have been recognized (Kalenchuk, 2010;
Rockslides are a type of landslide, common in Westin, 2017). The main concerns with this active
mountainous areas, broadly defined in Cruden & Varnes instability are both a rapid failure that could impact the
(1996) as downslope movements of rock mass, occurring Revelstoke Dam reservoir and the risk of it causing
dominantly on surfaces of rupture or on relatively thin blockage of the river. In order to improve the stability of the
zones of intense shear strain. This definition has been slope by reducing displacement rates, a drainage system
updated, and new categories proposed in Hungr et al. was built between 1974 and 1982, totalling 2430 m of
(2014). It is widely accepted that structural features such tunnel and 13600 m in drain holes (Moore, 1989;
as folds, faults and discontinuities control rock mass Kalenchuk, 2010).
behaviour and the stability of rock slopes (Stead & Wolter, The recognition and monitoring of unstable slopes next
2015), thus, their study is crucial to understand the to dam sites, that could eventually fail and cause
mechanisms involved. displacement waves or reservoir blockages, became a
These phenomena are recurrent in the Upper Columbia global concern after the October 1963 Vajont disaster
region in southeastern British Columbia (Figure 1), where (Paronuzzi & Bolla, 2012; Paronuzzi et al., 2016; Wolter et
several rockslides have been recognized, including al. 2015).
Downie slide, one of the largest rockslides in the world. After the construction of the Vajont Dam, a series of
Downie slide was first recognized during investigations filling-drawdown cycles was carried out at the reservoir
of alternative damsites in the 1950s and has an estimated between 1960 and 1963 (Paronuzzi et al., 2016).
volume of 1.5 billion m3. It is located next to the Revelstoke Movement rates at an unstable slope on the northern side
Dam reservoir, on the western side of the Columbia River of Mt. Toc accelerated during filling stages, later
and has been extensively studied (Brown & Psutka, 1980; decreasing during drawdown cycles. Cumulative horizontal
Kalenchuk, 2010; Kalenchuk et al., 2012; Westin, 2017; displacements ranging from 80 cm to 350 cm were
registered between May 1960 and the time of failure upper area, as are several tension cracks at the crest. The
(Paronuzzi et al., 2016). toe of the slope is covered by both alluvial and glacial
Months after the start of the third filling of the reservoir, deposits. It is underlain by a Basal Shear Zone (BSZ)
movement accelerated, and a new drawdown episode was dipping toward the reservoir.
carried out. However, at some point during this process, Due to its location just 1.5 km upstream of the Mica
the unstable slope continued moving and displacement Dam site, and because a rapid failure or reactivation could
rates accelerated until catastrophically failing, initiating a not be ruled out, the crest of Mica Dam was raised to
landslide wave that overtopped the dam and killed around prevent a wave from overtopping it (CASECO, 1969).
2000 people in the town of Longarone and other nearby Exploratory drilling commenced in 1969 to assess
villages (Wolter et al., 2015). The landslide deposit has an displacement within the rock mass and to better
estimated volume of 270-300 million m3 of rock and debris understand the geological characteristics of the rockslide.
(Paronuzzi & Bolla, 2012). Survey monuments, installed to monitor surface
Prior to the construction of Mica Dam, which started in displacements, did not show rates indicative of incipient
1967, new investigations using air photos and surface instability before reservoir filling started. The same results
mapping were carried out to identify slopes susceptible to were obtained after slope indicator surveys were carried
sliding. Several sites with post-glacial (<10000 years old) out inside three of the boreholes initially drilled at the toe
instabilities were recognized but only the Little Chief and and mid-northeastern slope of Dutchman’s Ridge (DH908,
DH911 and DH912) (CASECO, 1973).
It was not until 1982, after BC Hydro took control of the
dam, that the instrumentation data was re-evaluated,
including inclinometer data that had not been read during
the filling period. Small but consistent displacements of the
rock mass overlying a continuous fault were recognized to
have occurred during reservoir filling (1973-1976) at 5.1 to
18 mm/year, averaging 11.4 mm/year (BC Hydro, 1984).
Between 1973 and 1983, the total movement within the
underlying fault zone was over 20 cm, according to an
inclinometer at the northern toe of the slope. Besides,
during the period 1975-1981, surface displacements at an
average rate of 12 mm/year were measured (BC Hydro,
1984). As a result, in addition to the five previously drilled
holes, two more holes were drilled and instrumented, and
additional geological mapping was carried out in 1982-
1983. The accuracy of the monitoring system was
increased.
It was decided that it was necessary to reduce
displacements, and drainage through the construction of
an adit from which drain holes were to be drilled was
considered. An exploration project initiated in 1985 and
finished in 1986 provided information on landslide
boundaries, nature and extent of the basal fault zone,
groundwater pressures acting on potential slide base and
the effect of seismicity on the potential landslide mass (BC
Hydro, 1987a).
Besides gathering surface information, a total of 26
boreholes (at 23 locations) were drilled on Dutchman’s
ridge during both years of the exploration campaign,
totalling 4116 m. Instrumentation (inclinometers, standpipe
piezometers and multi-port piezometers) was installed
inside most holes.
The construction of the drainage adit lasted two years
from 1986 to 1988. A total of 872 m of adit was excavated
Figure 1. Location of the Upper Columbia region in British and 12700 m of drain holes were drilled (BC Hydro, 1989).
Columbia. Mid-right: Google Earth image showing the The basal fault and several relevant shears were
lower part of Dutchman’s Ridge, 1.5 km away of Mica Dam. intersected, and the geological model was updated.
Lower right: View of Downie slide next to Revelstoke Dam By the end of construction, drainage had been
reservoir. considered successful, by lowering the average drainable
head at the base from 36 m to 10 m (Moore, 1989).
Dutchman’s areas were deemed as potentially hazardous Consequently, the rate of displacement decreased. The
at that moment (CASECO, 1969). drainage system continues to perform as intended, and the
Dutchman’s Ridge (Figure 1) rises to an elevation of slope remains marginally stable, with an average
almost 2000 m a.s.l. and has an average slope angle of
around 30°. Subvertical foliation surfaces are evident in the
displacement of 1 mm/year, recorded by an extensometer 2.1.2 Proposed new boundaries
extending across the Basal Fault within the adit.
The 1988 model considered a rockslide volume of 115 Given the extremely slow displacements at Dutchman’s
million m3 contained within the boundaries shown on Ridge and the relatively small accumulated deformation
Figure 2 and represents the starting point of this work. compared to other extensively-monitored rockslides like
This paper summarizes the results of a review of the Aknes (Ganerød et al., 2008) and Downie slides
existing geological and monitoring data from Dutchman’s (Kalenchuk et al., 2012), typical features such as lateral
ridge rockslide, as well as a LiDAR topography obtained in and head scarps, which are useful to identify the
2014. The work included two field trips, in October 2020 boundaries of a rockslide, were not identified during
and July 2022, which were useful in finding field evidence previous studies as they have either not developed or are
to support what is proposed here. incipiently developing. The LiDAR topography of the slope,
Based on this work, it is proposed that the upper obtained in 2014, has been of immense help to better
boundary of the rockslide be extended upwards, and that define the extent of the rockslide.
the instability is not a single homogeneous mass, but rather A review of the existing data on Dutchman’s Ridge and
presents different zones from a mechanical and work with LiDAR-derived hillshade images, provided new
geomorphological point of view. evidence related to the boundary definition and zonation of
Two-dimensional (2D) models using limit equilibrium the rockslide.
methods were run to approximate the geometry and A ridge-like feature (Figure 2) trending NW, parallel to
estimate the strength parameters of the sliding surface of foliation, became noticeable as it divides the slope in two
the rockslide. Ongoing work is being done to test the areas with different morphologies. Evidence gathered
influence of a hypothetical loss of drainage capacity during this study, shows that this feature fits well with the
through finite element modelling. projection to the surface of Lavender Shear, a fault
encountered at the intersection of the Basal Shear Zone
during adit construction, that not only truncates its base,
2 Dutchman’s Ridge rockslide but also places the hanging wall side (SW) deeper (BC
Hydro, 1989). The intersection of both structures plunges
The instability at Dutchman’s Ridge is an extremely slowly 29° to the SE (138°) (BC Hydro, 1989), parallel to the main
moving (< 16mm/year) rockslide according to the Landslide direction of movement as indicated by inclinometers.
velocity scale in Hungr et al. (2014). The moving mass, As can be observed in Figure 2, the upper boundary of
underlain by a shear zone, was estimated to be 150 million the 1988 outline of the rockslide was extended to include
m3 (BC Hydro, 1993). the upper part of this feature as there seemed to be enough
The stratigraphic sequence is composed of highly evidence that it is part of the instability.
metamorphosed sedimentary rocks, mainly Mica As a result of the change in position of the rockslide
Gneiss/Schist and Hornblende Gneiss with minor boundary, borehole 85-20 (Figure 2), previously
occurrences of pegmatite and other types of Gneiss. considered to be in a stable zone, would now need to be
These rocks are part of the Precambrian Horsethief Creek included within the instability area. The existing geological
Group (BC Hydro, 1989). and geotechnical evidence was further investigated to
The foliation is mostly consistent above and below the support this change.
Basal shear zone, with an average orientation of Even though a shear zone was not recognized in 85-20
120°/50°SW (BC Hydro, 1989). This preferred orientation and 85-20A (drilled few meters away) during core logging
was confirmed during recent field work in July 2022, in 1985, a combination of core loss segments and highly
although steep foliation dipping up to 70°SW was found weathered and disturbed Mica Gneiss, with zones of
near a NW-SE feature in the middle of the slope, especially partially decomposed rock is present in the same segment
in the central and upper parts of the instability. where displacement has been occurring (6-9 m depth from
collar) in the same direction as other areas of the rockslide
2.1.1 Basal Shear Zone (toward the SE). This could be where the BSZ is present at
this location. This is also supported by the fact that the slide
The Basal Shear Zone (BSZ) is a tectonic structure (BC base, according to a projection of the plane using its typical
Hydro, 1993) underlying the rockslide at Dutchman’s Ridge orientation, would reach the surface there.
with an approximate strike of 65°E and a dip between 24° The upper northeastern boundary was extended as
and 35° towards the Mica reservoir (SE). As recognized in well, in order to include the ridge-like feature. A geologic
rock cores, it has a thickness between 1 m and 40 m and structure, named New Potential Fault in Figure 2 was
seems to be a continuous feature composed primarily of proposed to be the upslope limit on this area of the slope,
breccia, gouge and crushed rock (BC Hydro, 1989). A but more evidence is needed to support this interpretation.
minimum displacement of several tens of meters across The downstream (SW) and downslope boundaries
the fault was estimated (BC Hydro, 1993). have remained the same as proposed in 1988 since no
From shear strength testing on rock core samples, the new evidence to modify them was found in this study.
peak friction angle is in the range of 26°-33°, whereas a According to projections of the BSZ and the presence of
peak friction angle of 27° and a residual friction angle of natural boundaries such as Dutchman’s Creek, the base of
24° was estimated from a sample taken at the intersection the rockslide is expected to daylight both near the Creek
in the drainage adit (BC Hydro, 1987b; BC Hydro, 1988; and somewhere at the toe.
BC Hydro, 1989).
 The proposed northeastern boundary does not differ signs of instability such as a crack on the main road,
much from the previously defined boundary, with the presence of talus deposits and gully-forming erosion were
exception that a scarp and some tension cracks just below identified in the field.
inclinometer 85-21 were incorporated (Figure 2). Due to the The new proposed boundaries would add around
absence of instrumentation, there is no proof of deep- 217900 m2 to the area of the rockslide, which would reflect
seated displacement occurring further north. However, it is in only ~1.5 % increase in volume due to the expected
likely that this boundary is located at the northeastern flank lesser thickness near the top of the instability.
of Dutchman’s Ridge near Little Chief Creek, where some

Dutchman’s Creek

Figure 2. LiDAR-derived hillshade image (azimuth 45) showing landslide boundaries. The red outline represents the most
recently defined boundaries from 1988. The blue outline represents the proposed new boundaries of Dutchman’s Ridge.
The burgundy line with question marks indicates the anticipated maximum extents.
2.1.3 Zonation
It is the area located at the west and southwest side of the
In previous work, different areas at Dutchman’s Ridge were slope, where the Basal Shear Zone, as inferred from its
recognized based on surface morphology, however, since slope failures. An incipient scarp would be the upper limit
the Basal Shear Zone seems to be continuous, based on of this domain.
rock core evidence, and the rockslide has shown a low
displacement rate, it was probably difficult to confidently 2.1.3.2 Extremely Slowly Moving Area
define domains.
After a review of geotechnical and geological data, both The central west section of the slope shows significantly
from surface and subsurface, gathered over more than 60 fewer signs of superficial deformation than the surrounding
years; interpretations made from LiDAR data; and areas; the terrain is not hummocky as in the eastern side,
observations from site visits in October 2020 and July and the natural surface drainage system running toward
2022, the subdivision of Dutchman’s ridge rockslide into the reservoir is still preserved.
the six domains shown in Figure 3 is proposed. The only inclinometer within this domain shows
consistent but near-zero displacement since readings of
2.1.3.1 Southwest Flank this instrument started in 1987.

Figure 3. Left: Azimuth 45 LiDAR hillshade showing domains proposed at Dutchman’s Ridge rockslide. Cross-sections
used for numerical analysis in green. Top right: Azimuth 270 LiDAR hillshade showing domains and the location of the
apparent tension crack on the northeastern side of the slope. Bottom left: Photograph taken on July 2022 showing rock
blocks up to 12 m-high in the Slump block area, at the location of the apparent tension crack in the Main Road at
Dutchman’s Ridge.

field work and aerial photographs, although, the

2.1.3.3 Slump Block Area mechanism behind was not fully understood.
On LiDAR-derived hillshade images, the terrain is
This area was early recognized on the northeastern portion hummocky and several internal scarps were identified.
of Dutchman’s Ridge rockslide (BC Hydro, 1993) through During the July 2022 site visit, loose rock blocks were
found throughout this domain but were particularly frequent 3 Numerical analysis
and large (up to 12 m high) near a feature trending NNW
(Figure 3, lower and upper right) that resembles a long and 2-D limit-equilibrium and finite element analyses (ongoing)
deep tension crack. It is proposed here that this structure were performed on Dutchman’s Ridge with the objective of:
is probably related to a fault that was postulated to exist at
depth with a northerly strike (BC Hydro, 1989). During adit a) Testing the influence of toe geometry and Basal
construction, a disruption in stratigraphy was found jnk1bm Shear Fault thickness on the factor of safety (FoS)
nnnat this approximate location, between two drain holes. of the slope.
An interplay between this fault, foliation shears and b) Estimate the shear strength (Cohesion (C) and
other geologic structures would explain rock block Friction angle (ɸ)) of the Basal Shear Zone.
formation and the more extensive damage in this domain. c) Assess the influence of Cohesion (C) and Friction
angle (ɸ) ranges on modelled slope stability.
2.1.3.4 Northern Toe d) Estimate the variation (in %) of the FoS of the
slope if the present drainage system lost capacity
It is located on the upstream side of the slope toe and or an extreme hydrological event caused water
shows the fastest rate of displacement at the BSZ at pressure to increase.
inclinometer 908A (average of 1.3 mm/year). A distinctive
feature of this domain is that all the inclinometers included Two cross-sections (MCS_4 and MCS_6 on Figure 3),
in it (908, 908A, 914 & 916 in Figure 2) show displacement one on each side of the NW Ridge, were built in a direction
zones, moving at between 0.5 and 3.8 mm/year, above the parallel to foliation to represent the variations in
main Shear Zone. topography, Basal Shear Fault geometry and groundwater
Projection of the most relevant foliation shears profile in the slope.
intersected during adit construction would reach the The previous stability analyses were carried out after
surface on the northeastern side of the slope. It is therefore adit construction in 1988, with the updated geological and
likely that these structures are responsible for the above- groundwater models at the time and limited backup data
mentioned active displacements. This could be the case for estimating the parameters (BC Hydro, 1988).
for the Slump block area as well, however, there is a large The analysis presented here is different in that the
area where inclinometers are absent, preventing more upper boundary has been extended upwards, as proposed
detailed analysis. in this work, and the geometry of the rock/alluvial deposit
interface was simplified by removing the alluvial deposits
2.1.3.5 Lower Southwest Area to avoid adding more uncertainties. The surface
topography was obtained directly from cross-sections built
This area is located at the downstream side of the slope on Leapfrog (Seequent, 2020), from a mesh created using
toe and, as opposed to the Northern toe, no shallower the LiDAR point cloud. The topography of the underwater
displacement zones are identified in inclinometers. toe was estimated from contour maps prior to the reservoir.
Another relevant characteristic of this domain is two The piezometric surfaces were developed using
linear features, striking NNW and NE respectively, which assumptions similar to the ones made in the 1988 analysis.
are also partly present on the Extremely slowly moving In this case, both a pre-drainage (post reservoir filling) and
area and that appear as narrow depressions in the field. post-drainage surface were built. The pre-drainage data is
These are likely the result of faults that have been covered from spring-summer of 1986, whereas the post-drainage
by sediments. Similar features have been observed by BC stage was considered to be spring-summer 2020.
Hydro personnel at other landslide sites in the region The groundwater pressures acting at the slide base
(Stewart, 2022). (main gouge within the Basal Shear Zone, as defined in BC
Hydro (1987a), and, when possible, position of peak
2.1.3.6 NW Ridge displacement from inclinometers in this work), are believed
to be the most relevant for the rockslide stability. Therefore,
The previously described NW ridge-like feature was only ports just above, just below and right at the slide base
defined as a singular domain because it seems to were considered for this particular case. The maximum
represent a transition between a smooth and, apparently, piezometric level of the three ports for each piezometer
slightly displaced area to the southwest (Extremely slowly was chosen to build the piezometric surfaces (Figure 5).
moving area) and a more damaged one to the northeast To simplify the analysis, geologic structures such as
(Slump block area). shears or joints were not included, and the Basal Shear
Zone is assumed to be the weakest material where the
2.1.3.7 Northeastern Boundary most deformation occurs.
Numerical analyses are carried out taking into account
It is the area located at the northeastern edge of the that absolute values of FoS may have no meaning on this
rockslide, where clear signs of instability are present, kind of engineering problem and, instead, the rate of
although, due to the absence of instrumented or change of the FoS can be used to judge the impact of
exploratory holes, displacements cannot be confirmed at changes in slope conditions as well as the effectiveness of
the moment. remedial actions (Hoek, 2023).
 In this work, all materials are assumed Mohr-Coulomb, Final models were built for both profiles using the true
meaning that the shear strength is represented in terms of geometry of the BSZ, as recognized in boreholes. Some
cohesion C and friction angle ɸ (Coulomb, 1773). interpretation was done at the toe as well as the top of the
rockslide, due to lack of borehole coverage in these
3.1 Limit equilibrium analysis regions. Some of the results obtained are included in Table
1. Note that the parameters in the table are the ones
Slide2 is a 2D slope stability program that analyzes the expected to be used as a starting point for finite element
stability of slip surfaces using vertical slice or non-vertical analyses and that the considered values are lower for
slice limit equilibrium methods (Rocscience Inc., 2023b). It MCS_6, since it is thought that the rock mass has
was chosen to explore Dutchman’s Ridge configuration undergone higher deformation in that area.
and strength parameters due to its simplicity of use and
how quickly models can be analyzed.
To account for variability of results, three methods to
calculate FoS were chosen: Janbu corrected (Abramson et
al., 1996); Spencer (Spencer, 1967) and
GLE/Morgenstern-Price (Morgernstern & Price, 1965).
Regarding surface search methods, Cuckoo (Yang & Deb,
2009) and Simulated annealing (Kirkpatrick, 1984) were
utilized.
The results from shear strength testing (2.1.1) were
used as the basis to estimate the friction angle of the Basal
Shear Zone. A range from 0 to 400 kPa was considered for
cohesion. Cohesion values from shear testing were not
considered due to the difficulty of measuring this value in
the lab (Wyllie & Mah, 2004; Hoek, 2023).
For the undisturbed rock mass (below BSZ), high
strength parameters were assigned to force the slip surface
to the known location, as was done by Kalenchuk (2010)
for Downie slide: ɸ=46° and C=1x106 kPa. In the case of Figure 4. Factor of safety (GLE/M-P) vs. friction angle for
the landslide mass (above BSZ), peak shear strength three different thicknesses of the Basal Shear Zone (C=0).
properties of ɸ=38° and C=2000 kPa for MCS_4, and ɸ=35°
and C=1500 kPa for MCS_6 were assigned, assuming Table 1. Summary including part of the results from the limit
values higher than at Downie. equilibrium analyses, using the true Basal Shear Zone, for
There is substantial uncertainty regarding the both MCS_4 and MCS_6.
geometry of the toe, in particular the exact point of slide
base daylighting. An analysis of the slide base, using ɸ of Cross- ɸ C Search Janbu GLE/M-
27° and C of 100 kPa, was carried out to compare three section
Stage
(°) [MPa] method corrected
Spencer
P
different geometries of the toe. The intermediate solution
Pre-
was chosen since it provided an intermediate FoS and MCS_4
drain
24 0 S. A.* 0.954 1.101 1.004
seemed more realistic, not being overly flat or steep (Figure MCS_4
Post-
24 0 S. A.* 1.007 1.153 1.053
5). drain
Pre-
The pre-drainage model for MCS_4, with the Slide MCS_4
drain
26 25 S. A.* 1.045 1.176 1.089

Base as a Weak Layer (no thickness) surface, was used to MCS_4

Post-
26 25 S. A.* 1.104 1.237 1.144
drain
calibrate strength parameters. As this represents the Pre-
period when the rockslide was moving at a higher rate, it
MCS_4 31 200 S. A.* 1.518 1.69 1.567
drain
was assumed the FoS should have been ~1. A MCS_4
Post-
drain
31 200 S. A.* 1.59 1.772 1.637
deterministic analysis was performed, with ɸ in the range MCS_6
Pre-
24 0 S. A.* 0.786 0.868 0.792
26°-33° (in 1° increments) and C between 0 and 400 kPa drain
Post-
(in increments of 25 kPa between 0 and 100 kPa, and 100 MCS_6
drain
24 0 S. A.* 0.897 0.967 0.886

kPa above that). As most FoS values were above 1, a ɸ of MCS_6

Pre-
26 25 S. A.* 0.927 0.955 0.915
24° and C of 0 would probably represent shear strength, as
drain
Post-
it provides a FoS the closest to the target value. MCS_6
drain
26 25 S. A.* 0.968 1.011 0.967

For MCS_4, a third analysis was conducted to assess MCS_6

Pre-
drain
27 50 S. A.* 1.115 1.122 1.093
the influence of the thickness of the Slide Base on FS Post-
MCS_6 27 50 S. A.* 1.116 1.178 1.103
values. In addition to the negligible thickness surface, two drain

new models with a 10-m thick and a 20-m thick Shear Zone *S.A. = Simulated annealing search method.
were analyzed. The FoS values resulting from the
GLE/Morgenstern-Price calculation method, using 3.2 Finite elements analysis
simulated annealing search method are shown in Figure 4.
The results follow the expected trend, as thicker shear The final geometry obtained from limit equilibrium analyses
zones yield lower FoS, although the differences are <10% was imported into the finite element software RS2
for a 20-m increase in thickness. (Rocscience Inc., 2023a). The water in the reservoir was
represented as a distributed load (Figure 5).
 Considering all materials to be plastic, the built-in shear derived data. According to this, the upper boundary should
strength reduction method was individually applied to both be extended upwards. There is no new evidence to extend
pre-drainage and post-drainage models on both cross- the southwestern and downslope boundaries. The
sections. Residual values were added for all materials. northeastern boundary is likely located near Little Chief
Preliminary results show that a strength reduction factor Creek, however, due to the lack of instrumentation that
near 1 is only attained with BSZ strength properties greater could not be confirmed.
than those estimated from the Slide2 analyses. Rather than being a single homogeneous mass, the
Ongoing work is being done by building models that rockslide presents six zones showing variable
represent a hypothetical loss of drainage capacity. The morphological characteristics and mechanical behaviour.
results will be given in terms of a percentage of variation in Ongoing numerical analysis using finite element
strength reduction factor. The objective is to assess how modelling with RS2 shows that shear strength values tend
changing water pressures would affect the stability of the to be higher than the ones estimated using limit equilibrium.
slope if the drainage system lost capacity. Work is being carried out modelling a hypothetical loss of
drainage capacity.
MCS_4
Pre-drainage
5 Acknowledgements

This work has been possible due to the support of BC

Hydro, particularly Tom Stewart and Julia Marsh, and the
Geomechanics & Geohazards Group at Queen’s
University. Funding has been provided by NSERC and BC
Hydro.
MCS_4
Post-drainage
6 References

Abramson, L.W., Lee, T.S., Sharma, S., Boyce, G.M. .

1996. Slope Stability and Stabilization Methods (John
Wiley and Sons: N.Y.).
BC Hydro. 1984. "Mica Project Reservoir Slopes
Surveillance - Memorandum on Assessment of
Movement of Ducthman's Ridge."
MCS_6 BC Hydro. 1987a. "Dutchman's Ridge Remedial Works -
Pre-drainage Memorandum on 1985 and 1986 Investigation and
Instrumentation Programs."
BC Hydro. 1987b. "Dutchman's Ridge Remedial Works -
Summary of Information for Panel meeting N°4."
BC Hydro. 1988. "Dutchman's Ridge Remedial Works -
Summary of Information for Panel Meeting No.5 "
BC Hydro. 1989. "Dutchman's Ridge Remedial Works -
Geological and Groundwater Information 1986-1988."
BC Hydro. 1993. "Dutchman's Ridge Remedial Works -
Summary Report "
MCS_6 Brown, Richard L., and John F. Psutka. 1980. 'Structural
Post-drainage
and stratigraphic setting of the Downie slide, Columbia
River valley, British Columbia', Canadian Journal of
Earth Sciences, 17: 698-709.
CASECO. 1969. "Mica Project - Supplementary Report on
Stability of Reservoir Slopes."
CASECO. 1973. "Mica Project - Sixth Supplementary
Report on Stability of Reservoir Slopes."
Coulomb, C.A. 1773. "Essai sur une application des règles
de maximis et minimis a quelques problèmes de
statique relatifs a l'Architecture." Mémoires de
Figure 5. Pre-drainage and post-drainage models created
Matematique et de Physique présentes a l'Academie
for MCS_4 and MCS_6 in RS2.
Royales des Sciences par divers Savants. Paris,
France.
Cruden, D.M., Varnes, D.J. 1996. 'Landslide Types and
4 Conclusions
Processes', Transportation Research Board, U.S.
National Academy of Sciences, Special Report, 247:
New boundaries of Dutchman’s Ridge rockslide are
36-75.
proposed based on geological, inclinometer and LiDAR-
Donati, Davide, Doug Stead, Thomas W. Stewart, and Julia Italy, and a new interpretation of the chronology and
Marsh. 2020. Numerical modelling of slope damage in evolution of the landslide. Landslides, 13: 1067-81.
large, slowly moving rockslides: Insights from the Wyllie, D., Mah, C. 2004. Rock Slope Engineering: civil and
Downie Slide, British Columbia, Canada. Engineering mining (Taylor & Francis Group).
Geology, 273. Yang, X-S. & Deb, S. 2009. "Cuckoo search via Levy
Ganerød, Guri Venvik, Guro Grøneng, Jan Steinar flights." In World Congress on Nature & Biologically
Rønning, Einar Dalsegg, Harald Elvebakk, Jan Fredrik Inspired Computing, 210-14. Coimbatore, India: IEEE
Tønnesen, Vidar Kveldsvik, Trond Eiken, Lars Harald Publications.
Blikra, and Alvar Braathen. 2008. Geological model of
the Åknes rockslide, western Norway. Engineering
Geology, 102: 1-18.
Hoek, E. . 2023. Practical Rock Engineering.
Hungr, O., Leroueil, S., Picarelli, L. 2014. The Varnes
classification of landslide types, an update. Landslides,
11: 167-94.
Kalenchuk, K. S. 2010. Multi-Dimensional Analysis of
Large, Complex Slope Instability. PHD Thesis, Queen's
University.
Kalenchuk, K. S.; Hutchinson, D.J.; Diederichs, M.; Moore,
D. 2012. 'Downie Slide, British Columbia, Canada' in J.;
Stead Clague, D. (ed.), Landslides: Types,
Mechanisms and Modeling (Cambridge University
Press).
Kirkpatrick, S. 1984. Optimization by simulated annealing
--- quantitative studies. Journal of Statistical Physics,
34: 975-86.
Moore, Dennis P. 1989. "Panelist contribution: Stabilization
of Downie Slide and Dutchman's Ridge" In 12th
International Conference on Soil Mechanics and
Foundation Engineering 3063-65. Rio de Janeiro
Morgernstern, N.R. & Price, V.E. . 1965. The analysis of
the stability of general slip surfaces. Geotechnique, 15:
79-93.
Paronuzzi, P., Bolla, A. 2012. The prehistoric Vajont
rockslide: An updated geological model.
Geomorphology, 169-170: 165-91.
Paronuzzi, P., Bolla, A.. Rigo, E. 2016. Brittle and Ductile
Behavior in Deep-seated Landslides: Learning from the
Vajont experience. Rock Mechanics and Rock
Engineering, 49: 2389-411.
Rocscience Inc. 2023a. 'RS2 Version 11.018'.
www.rocscience.com.
Rocscience Inc. 2023b. 'Slide2 Version 9.027'.
www.rocscience.com.
Seequent. 2020. 'Leapfrog Geo 6.0'.
https://www.seequent.com/products-
solutions/leapfrog-geo/.
Spencer, E. 1967. A method of analysis of the stability of
embankments, assuming parallel interslice forces.
Geotechnique 17: 11-26.
Stead, Doug, and Andrea Wolter. 2015. A critical review of
rock slope failure mechanisms: The importance of
structural geology. Journal of Structural Geology, 74: 1-
23.
Stewart, T. (2022). Personal Communication.
Westin, A. M. . 2017. Downie Slide: An integrated remote
sensing approach to characterization of a very slow
moving landslide. Master Thesis, Simon Fraser
University.
Wolter, Andrea, Doug Stead, Brent C. Ward, John J.
Clague, and Monica Ghirotti. 2015. Engineering
geomorphological characterisation of the Vajont Slide,
Tuesday, October 3, 2023

INSITU
INVESTIGATIONS
Converting between power law and
hyperbolic representations of stiffness
decay with strain
Yasmin Byrne & Robert Whittle
Cambridge Insitu, Cambridge, UK

ABSTRACT
Repeatable data for shear modulus is readily obtained from small cycles undertaken as part of a cavity expansion test. In
all soils these data follow a hysteretic path that a high resolution pressuremeter can capture and use to describe the
stiffness degradation with strain characteristics. This response is normally modelled as a power law. The standard cycle is
not expected to see shear strains below 0.01% and hence does not measure the elastic shear modulus, Gmax, found at
shear strains nearer 0.001%.
By contrast, laboratory testing of small strain stiffness is normally represented by a hyperbolic function where Gmax
defines the starting plateau, and in this form is readily incorporated into design software. It is always important that
laboratory and insitu data are compared and if possible aligned. This paper considers potential methods for translating
power law data into hyperbolic relationships.

RÉSUMÉ
Des données répétables pour le module de cisaillement sont facilement obtenues à partir de petits cycles entrepris dans
le cadre d'un test d'expansion de cavité. Dans tous les sols, ces données suivent un chemin hystérétique qu'un
pressiomètre à haute résolution peut capturer et utiliser pour décrire la dégradation de la rigidité avec des caractéristiques
de déformation. Cette réponse est normalement modélisée par une loi de puissance. Le cycle standard ne devrait pas voir
des déformations de cisaillement inférieures à 0,01 % et ne mesure donc pas le module de cisaillement élastique, Gmax,
trouvé à des déformations de cisaillement plus proches de 0,001 %.
En revanche, les essais en laboratoire de rigidité à faible déformation sont normalement représentés par une fonction
hyperbolique où Gmax définit le plateau de départ et, sous cette forme, est facilement intégré dans le logiciel de
conception. Il est toujours important que les données de laboratoire et in situ soient comparées et si possible alignées. Cet
article examine les méthodes potentielles pour traduire les données de loi de puissance en relations hyperboliques.

1 INTRODUCTION reversed. An example of such a test carried out with a pre-

bored pressuremeter is shown in Figure 1.
Cavity expansion testing is acknowledged to be a reliable For a prebored test, the pressuremeter is placed in a
means for determining ground stiffness. Such a test slightly oversized pocket formed by conventional drilling
conducted with a high resolution pressuremeter consists of tools. Although the initial slope of the field curve could be
an expansion phase, where the ground is loaded, a used for stiffness interpretation it is affected by the pocket
contraction phase where the ground is unloaded and forming process and subsequent soil relaxation
several unload/reload cycles where the applied stress is

Figure 1 Field curve – pre-bored test conducted with High Pressure Dilatometer in silty sand
 Consistent and repeatable data for shear stiffness can Either the unload or reload part of cycle can be used,
be obtained from small cycles of unloading and reloading but in practice the second half of the cycle (data obtained
(Hughes, 1982). The response of these cycles is driven by after the turnaround point) is least likely to be affected by
the far field pre-yield properties of the ground. creep effects from the ground or frictional errors introduced
This can be demonstrated by the similarity of by the measurement system. By fitting reloading data with
successive cycles. Any cavity expansion test has the a power curve, 𝛼 and 𝛽 are found. The correlation is
potential to produce high quality stiffness data regardless generally better than 0.99. This method is valid for
of the damage caused to immediately surrounding material measured shear strains in the range between 10-4 (0.01%)
by the insertion process. and the yield strain 𝛾𝑓 (the strain when the full strength of
With high resolution pressuremeters, it is apparent that the material is mobilised). This is typically 10-2 or 1% for
the path followed during these cycles is not linear, as might over-consolidated clays.
be assumed from low resolution measurements, but is The shear modulus at 10-4 shear strain is likely to
hysteretic in all soils. Muir Wood (1990) and Jardine (1992) underestimate the elastic shear modulus (𝐺𝑚𝑎𝑥 ). Although
examine the response of these cycles and propose ways the curve can be projected back beyond 10-4 it is necessary
of obtaining stiffness degradation data from them. Bolton & to know the threshold strain (𝛾𝑒𝑙𝑎𝑠 ) at which stiffness decay
Whittle (1999) give a straightforward method based on a commences. Without this, power curve modelling predicts
power law and it is now customary to apply this approach infinite stiffness at zero strain. The elastic threshold strain
to the standard pressuremeter test. is typically in the order of 10-5.
Small strain stiffness laboratory testing attempts to The pressuremeter lower measurement limit of 10-4 is
determine the elastic shear modulus Gmax and then follow due to a variety of reasons including:
the subsequent degradation as shear strain increases. This
encourages the use of a hyperbolic function to describe the • the limitations of the electro-mechanical
overall response. By contrast, the smallest observable measurement system.
change in the pressuremeter test generally means that the • the possible influence of ground creep
current stiffness is already at an arbitrary point on the • the fine pressure control challenges.
degradation curve (Figure 2) and hence a power law
approach works well. The purpose of this paper is to show In principle the resolution of the displacement
that the two approaches can be reconciled. measuring system is sufficient to see 𝐺𝑚𝑎𝑥 (shear moduli
in excess of 1GPa are routinely measured in rock) but
every change in direction is affected by a small loss due to
friction and dynamic ground effects, loosely termed ‘creep’.
Additionally, stiffness decay will commence after a tiny
proportion of the available strength is mobilised, typically
1%. Hence if a material has a shear strength of 100kPa it
means that all 𝐺𝑚𝑎𝑥 data cease after a change of 1kPa. To
make a credible assessment with the pressuremeter a
minimum of 5 steps of 0.2kPa change would be required.
In a triaxial chamber where all elements are seeing the
same shear stress it is relatively straightforward to arrange
for this fine degree of stress control. However, in the cavity
expansion test the current shear stress seen by an element
of soil depends on its radius relative to the cavity wall. The
field curve is the integration of these infinite stress states.
Consequently there will always be a propensity for the
ground to creep as it attempts to reconcile the different
rates of strain whenever the expansion process is paused
or reversed.
Figure 2 Normalised stiffness/strain response of a
Figure 2 uses data from a self bored pressuremeter test
characteristic SBP test in clay
in Gault Clay, with the part of the curve in red indicating the
information that is provided routinely by the pressuremeter
unload/reload cycle. This is the range of strain applicable
2 THE POWER LAW
to most geotechnical design and hence it is desirable that
this be directly measured insitu.
To describe the non-linear stiffness/strain response, Bolton
& Whittle give the following relationship:
3 STIFFNESS AS A FUNCTION OF STRENGTH
𝐺𝑠 = 𝛼𝛾 𝛽−1 [1]
The stiffness decay process starts at the elastic threshold
Where 𝐺𝑠 is secant shear modulus at plane shear strain 𝛾, strain and terminates at the yield strain required to initiate
𝛼 is the shear stress constant and 𝛽 is the exponent of non- full plasticity. This is largely recoverable. It is convenient to
linearity. 𝛽 takes a value between 0.5 and 1, where 1 is present stiffness decay against the fraction of mobilised
linear elastic. shear stress. This can be written as follows:
 𝑚
𝐺𝑠 𝛾−𝛾𝑒
( ) = 1⁄[1 + ( ) ] [5]
𝐺𝑠𝑛 = 𝛼[𝑛 𝑐𝑢 ⁄𝛼] (𝛽−1)⁄𝛽
[2] 𝐺𝑚𝑎𝑥 𝛾𝑟𝑒𝑓

𝐺
where 0 < 𝑛 ≤ 1 and is the proportion of strength used. 𝑐𝑢 where 𝛾𝑟𝑒𝑓 is a reference shear strain when = 0.5 and
𝐺𝑚𝑎𝑥
is undrained strength and for the drained case will be 𝑚 controls the curvature of decay (Oztoprak & Bolton
𝜏𝑓 (the shear stress required for first yield). denote this 𝛼 but here 𝑚 is used to avoid notation
This is based on Bolton & Whittle, which can be rearranged duplication). Oztoprak & Bolton use Eq.5 to model the
to give shear modulus in terms of shear stress rather than decay response of a wide range of tests conducted in sand:
shear strain using the power law definition. The primary difference between Eq. 5 and earlier
High resolution pressuremeters can make field representations is the curvature exponent. This gives a
measurement adequate for 𝑛 values between 0.05 to 1. more realistic representation of the material response but
This resolution is not enough to see the elastic shear at the expense of an additional unknown.
modulus 𝐺𝑚𝑎𝑥 , since it is likely to degrade after only 1% of The exponent is the only sand specific element of Eq.
the available strength has been mobilized (𝑛 = 0.01). 5. Oztoprak & Bolton found that 𝑚 = 0.88 gave the best
Similarly, it is difficult to recognise the elastic strain average fit to their database of sand tests. The curvature is
threshold, 𝛾𝑒𝑙𝑎𝑠 from measurements made during a cavity potentially related to particle size and will therefore vary
expansion process. However, if either of the parameters with soil type, analogous to the way 𝛽 operates in the
are known, then the power law can be arranged to describe power law method.
the full decay response.
5.1 Application to pressuremeter tests
4 CALCULATING ELASTIC SHEAR MODULUS For Oztoprak & Bolton’s analysis, 𝐺𝑚𝑎𝑥 is known and
stiffness decay is the predicted element. For the
The cavity expansion test is particularly good at identifying pressuremeter test the majority of the stiffness decay
the yield condition. If both the hyperbolic and power law relationship is measured from unload/reload cycles,
descriptions are assumed equivalent, then it can be however, 𝐺𝑚𝑎𝑥 is not. An iterative approach can be applied
inferred from Cao et al (2002) that there is a relationship where successive estimates of 𝐺𝑚𝑎𝑥 can be used in Eq. 5
connecting 𝐺𝑦 , the shear modulus at first yield, to the to find a hyperbolic data set that reproduces the
elastic shear modulus 𝐺𝑚𝑎𝑥 : pressuremeter decay data (Figure 3). Here, the strain axis
is proportion of mobilised shear strain. It starts from zero
𝐺𝑚𝑎𝑥 = 𝐺𝑦 𝐸𝑥𝑝[1/𝛽 ] [3] and ends at the yield strain 𝛾𝑓 .

If measured values of 𝛽 are used then Eq. 3 tends to

give conservative values for 𝐺𝑚𝑎𝑥 . However, in all soils that
can be described as non-linear, at the point of
measurement, 𝛽 represents the particle size and shape at
the microscale. Differential stress will lead to asperity
removal and increasing uniformity. Ultimately, 𝛽 will tend
towards 0.5 as elements approach spherical form. If it is
assumed that 𝛽 is 0.5 when using Eq. 3 then it follows that:

𝐺𝑚𝑎𝑥 = 7.38𝐺𝑦 [4]

This is speculative. However, if justified, it leads to the

conclusion that Eq. 4 gives the maximum that 𝐺𝑚𝑎𝑥 can
be. Once 𝐺𝑚𝑎𝑥 is known, then Eq. 1 or Eq. 2 can be re-
arranged to find the threshold strain 𝛾𝑒𝑙𝑎𝑠 .

5 THE HYPERBOLIC APPROACH

A hyperbolic function is the conventional way of Figure 3 Matching hyperbolic and power curves
representing stiffness decay. 𝐺𝑚𝑎𝑥 is used as the primary
input to define the starting plateau. A simple hyperbolic
arrangement (e.g. Hardin & Drnevich, 1972) is The plane shear strain at failure for the undrained case is
mathematically convenient but is a poor representation of given by Eq. 6. This can be arranged to find the mobilised
the curvier response of real soils. Oztoprak & Bolton (2013) shear stress for any shear strain between zero and 𝛾𝑓 . In
review some variations and suggest the following, based
on Darendeli (2001): Figure 3, this can be used to define the horizontal axis.

1/𝛽
𝛾𝑓 = [𝑐𝛼𝑢] [6]
 𝑚 = 1.5(1 − 𝛽) [7]
For the drained case, 𝑐𝑢 can be substituted with 𝜏𝑓 (the
shear stress at first yield). The shear stress constant 𝛼 There is evidence that where 𝑚 requires additional
must be appropriate for the effective stress level at failure. manipulation, at least one of the other parameters is likely
There are several methods (Bellotti et al, 1989, Whittle to have been wrongly assessed.
& Liu, 2013) for adjusting stiffness and shear stress
constant for stress level. It is common for the first valid
unload reload cycle to be used as an initial approximation. 6 MEASURING 𝑮𝒎𝒂𝒙 WITH THE PRESSUREMETER
The first cycle is often initiated sufficiently close to the insitu
stress state for its power law constant and exponent to be It is clearly desirable to measure 𝐺𝑚𝑎𝑥 directly rather than
used directly. rely on calculation or iteration. This is theoretically
Figure 4 is an example of the potential difficulty. The achievable with an appropriate procedure, despite the
plot consists of 5 cycles taken from a self bored test in difficulties associated with the mechanical limits of the
dense sand. The power law results can be presented as measurement system.
continuous lines. It is also possible to extract stiffness ‘Cycles within cycles’ is one possible procedure. Figure
decay data directly from the experimental measurements, 5 is a sketch of the concept. Having taken the unloading
and these are the points on the same plot. By of plotting part of a cycle, the reloading phase would be only 50% of
both data sets in this way the yield strain can be easily the amplitude of the initial unload. A further unloading down
identified, where measured data separates from calculated to the same lower limit of the cycle would then be made.
trends. If the cycles are large enough then it is common to The process repeated until the change of stress is
slightly exceed the shear stress that applied when the cycle sufficiently small to give a near linear response. The slope
was initiated and the break is therefore the material yield of the linear response would give 𝐺𝑚𝑎𝑥 directly. Once 𝐺𝑚𝑎𝑥
strain. is obtained, the final reloading would return the cycle to the
original loading curve.

Figure 4 Stiffness degradation curves from a drained test Figure 5 Cycles within cycles to identify 𝐺𝑚𝑎𝑥
in dense sand

Any friction at the turn-around point would be the same for

5.2 Summary all cycles within cycles and hence recognisable. Any creep
is a fixed percentage of the process prior to a reversal of
The hyperbolic iterative approach to determining 𝐺𝑚𝑎𝑥 with direction. Reducing the stress amplitude of successive
a pressuremeter has been applied to a number of sites and cycles, makes the impact of creep negligible.
materials of varying permeability grading from clays to When considering soils, creep is a characteristic of
sands. The correlation coefficient between data sets unload/reload cycles, in all soil types. The standard
predicted by the power law parameters and the hyperbolic unload/reload cycle applies shear strain greater than the
parameters is typically 0.99. The iterative procedure stops elastic threshold, and under those conditions all soils creep
when the correlation is at a maximum. due to micro fracturing and particle re-arrangement. Creep
As suggested above there is an inverse relationship is a generic term covering a multiplicity of actions and is
between 𝑚 and 𝛽. Based on tests examined to date, the can be an imprecise definition. Hence clays will in general
following seems to fit most cases: appear to creep more than sands whenever the loading
path is reversed simply because the response tends to be
undrained and any pause initiates loss of excess pore
water pressure. Technically this is not creep, however the
consequence is similar: it introduces a time dependent
deformation for no increase in the applied stress. In
addition, there are rate effects; even a slow pressuremeter
test is rapid compared to the process of laboratory testing.
However, if multiple cycles of ever-reducing size are
carried out at the same point in the test, then creep ceases
to be a significant issue.
The primary challenge is not related to displacement
resolution. Pressuremeter tests in competent rock require
the accurate measurement of sub μm displacements to
give data for stiffness moduli. The problem is applying
pressure increments in small enough steps and reading the
consequent response at a fast enough rate.
One possible solution for this is the use of a load cell
pressuremeter (LCPM) system as described by Hughes
and Whittle (2023). The LCPM is capable of measuring
𝐺𝑚𝑎𝑥 as a fortuitus by-product of making measurements of
the insitu lateral stress. Conceptually, the LCPM is the
reverse of the expansion pressuremeter. The stress in the
ground bears on the surface of a load cell, causing it to
deflect inwards. These sub μm inward movements are
used as feedback for a control system that raises the
internal pressure to match the external stress and hence
restore the load cell to a zero strain condition. This process
takes some time. The readings made whilst this process is
occurring can be interpreted as stiffness information.
However, the LCPM is essentially a research tool and is
seldom used in a wider commercial sense.

6.1 Field tests

Figure 6 A small unload reload cycle examined for 𝐺𝑚𝑎𝑥
As part of the 1992 ground investigations for the
London Crossrail project, cavity expansion tests using a
self-bored pressuremeter (SBP) were carried out with
Figure 7 gives the results from 30m of testing in one
unusually small unload/reload cycles. The purpose was to
borehole from Crossrail. Data are shown for shear modulus
obtain stiffness data at much smaller strain levels than was
at 0.01% shear strain; this is a reliable measurement but
then customary. To achieve this, all cycles were controlled
obtained at too large a strain to be 𝐺𝑚𝑎𝑥 . The results for
by a semi-automatic pressurisation system able to make
𝐺𝑚𝑎𝑥 derived using Eq. 3 and from the measurement
relatively small pressure changes.
technique displayed in Figure 6 are plotted alongside.
If the data from these cycles are examined then it is
These show a stiffer response and seem plausible.
possible to see that following the turnaround point in the
These tests were not designed for the purpose for
cycle, very small strain data are available that may be
which they are being used here. However, the comparison
representative of the elastic shear modulus. Figure 6 is an
between the measured and calculated 𝐺𝑚𝑎𝑥 trends is
example of one such cycle.
encouragingly close.
This cycle mobilises less than 15% of the available
strength. The conventional approach, as in Figure 1,
utilizes at least 50% of the available strength.
Consequently, much of the hysteresis or curvature of the
conventional cycle is not apparent in the very small cycle.
The initial linear portion following cycle turnaround has
been used to quantify 𝐺𝑚𝑎𝑥 . There is good agreement
between this and 𝐺𝑚𝑎𝑥 calculated using Eq. 3. It should be
noted that the 𝐺𝑚𝑎𝑥 estimation from Figure 6 is based on
only three experimental data points. The value obtained
should be considered as proof of concept rather than
reliable measurement.
 • By iteration to find the value that gives the highest
correlation coefficient to the experimental data.
(Eq. 5)
• By using relationships that assume that the elastic
threshold and yield strains are related. (Eq. 3)
• By direct measurement, using modified
unload/reload cycles. (Section 6)

In practice, it is likely that all 3 methods would be

deployed to find a consensus. For direct measurement to
be successful it is necessary that the steps of pressure are
kept very small. This will require an automatic control
system. The reading rate will also need to be faster than
commonly used commercial equipment currently provide;
whilst maintaining the principle of quiet data over quantity
of points.
The cycle procedure itself will need to be modified to
minimise the influence of creep and mechanical
imperfections if the direct measurement of 𝐺𝑚𝑎𝑥 is
intended. Within a test with 3 or more unload/reload cycles,
it should be possible to arrange that one cycle could be
prioritized for extracting small strain stiffness data.
Although most of the examples have come from
undrained tests the suggested methods are applicable to
the drained case. However, unload/reload data from the
same test level will be influenced by variations in the mean
effective stress and an appropriate adjustment for this
would need to be introduced.
Used conventionally the pressuremeter gives
parameters for 𝐺𝐻𝐻 , the lateral stiffness for a lateral
loading. Comparisons with laboratory and other methods
need to take account of this. 𝐺𝐻𝐻 often tends to be the
orientation and loading direction giving the highest stiffness
Figure 7 Deriving 𝐺𝑚𝑎𝑥 from SBP tests in London Clay values.
It must be emphasised that the approaches outlined
here are only applicable to high resolution instruments
7 CONCLUSIONS AND CLOSING REMARKS which make direct measurements of the radial movements
of the cavity wall. It would be unreasonable to expect this
Laboratory based small strain stiffness testing begins by level of discrimination from a volume measuring system.
applying tiny increments of stress that, in the ideal case,
allow both the elastic modulus and elastic threshold strain 8 REFERENCES
to be identified. A cavity expansion test undertaken with a
high resolution pressuremeter uses unload/reload cycles to Bellotti, R., Ghionna, V., Jamiolkowski, M., Robertson, P.
determine stiffness characteristics. These cycles are And Peterson, R. (1989) Interpretation of moduli from
concerned with strain levels greater than the elastic self-boring pressuremeter tests in sand. Géotechnique
threshold but give modulus parameters representative of 39 (2)269 - 292.
those applicable to engineering problems. As the start Bolton, M.D. and Whittle, R.W. 1999. A non-linear
condition is undefined, it is reasonable to use a power law elastic/perfectly plastic analysis for plane strain undrained
to quantify the response. No empiricism is required to expansion tests. Géotechnique 49(1): 133-141.
produce the pressuremeter trends and there are sound Cao, L.F., Teh, C.I. and Chang, M.F. 2002. Analysis of
reasons for considering the power law to be a better undrained cavity expansion in elasto-plastic soils with
representation of the ground response than a hyperbolic non-linear elasticity. International. Journal of Numerical
function. However, power law parameters are not easily and Analytical Methods in Geomechanics 26. 25-52.
incorporated into design packages such as FLAC or Darendeli, B. M. (2001). Development of a new family of
PLAXIS. normalized modulus reduction and material damping
This paper has described methods for translating curves. PhD dissertation, University of Texas at Austin,
between power law parameters and an equivalent TX, USA.
hyperbolic function. To do this it is necessary to identify the Hardin, B. O. and Drnevich, V. P. (1972). Shear modulus
elastic shear modulus 𝐺𝑚𝑎𝑥 , which may be done in at least and damping in soils: design equations and curves.
three ways: Journal of Geotechnical Engineering 98, No. 7: 667–
692.
Hughes, J.M.O. (1982) Interpretation of pressuremeter
tests for the determination of elastic shear modulus.
Proc. Engng Fdn Conf. Updating subsurface sampling
of soils and rocks and their in-situ testing, Santa
Barbara: 279 - 289.
Hughes, J.M.O and Whittle, R.W. (2023) High Resolution
Pressuremeters and Geotechnical Engineering. CRC
Press, ISBN 978-1-032-06094-1, 205-208
Jardine, R.J. (1992) Nonlinear stiffness parameters from
undrained pressuremeter tests. Canadian
Geotechnical Journal, 29, 436-447
Muir Wood, D. (1990) Strain dependent soil moduli and
pressuremeter tests. Géotechnique, 40 (3): 509 - 512.
Oztoprak, S. and Bolton, M. D. 2013. Stiffness of sands
through a laboratory test database. Géotechnique
63(1): 54–70.
Whittle R.W and Liu Lian (2013) A method for describing
the stress and strain dependency of stiffness in sand.
Proc. Symp. ISP6, Paris - September 4, session 3,
paper 7.
Comparison of the nonlinear modulus of the
Cambridge high-resolution pressuremeter in
Clearwater Formation Clayshales
Julian Contreras1, Ana Manderson1, Michael Hendry1, David
Elwood2 & Yasmin Byrne3
1Department of Civil and Environmental Engineering – University

of Alberta, Edmonton, Alberta, Canada

2Department of Civil, Geological and Environmental Engineering –

University of Saskatchewan, Saskatoon, Saskatchewan, Canada

3Cambridge Insitu LTD, Cambridge, England

ABSTRACT
This paper presents the results of thirteen pressuremeter tests conducted in Clearwater Formation Clayshale at Syncrude’s
oil sands operations near Fort McMurray. This paper performs an interpretation of nonlinear stiffness data from undrained
pressuremeter tests. The behavior of the secant shear modulus for several shear strain values is analyzed and a
comparison of the nonlinear stiffness is made using London clay as a reference. In addition, the initial modulus obtained
with shear wave velocity values derived from seismic tests is included. Finally, the stiffness parameters for the Clearwater
clayshale are presented.

RÉSUMÉ
Cet article présente les résultats de treize essais pressiométriques menés dans les schistes argileux de la formation
Clearwater, dans les exploitations de sables bitumineux de Syncrude, près de Fort McMurray. Cet article interprète les
données de rigidité non linéaire provenant d'essais pressiométriques non drainés. Le comportement du module de
cisaillement sécant pour plusieurs valeurs de déformation de cisaillement est analysé et une comparaison de la rigidité
non linéaire est effectuée en utilisant l'argile de London comme référence. En plus, le module initial obtenu avec des
valeurs de vitesse d'onde de cisaillement dérivées d'essais sismiques est inclus. Les paramètres de rigidité de l'argilite de
Clearwater sont enfin présentés.

1 INTRODUCTION underestimate the pre-yield deformations. As such,

Burland (1989), Simpson (1992), Benz et al (2009) and
Standard practice for analyzing the stability of mine pit Clayton (2011) demonstrate that the use of non-linear
walls within the oil sands is to run limit equilibrium analysis elasto-plastic constitutive models, tend to better predict
(LE) using conventional soil parameters along with lateral deformations when compared to conventional linear
application of the observational method. LE outputs do not elastic-perfectly plastic models. The conventional non-
permit the prediction of displacements associated with cut linear models like the Hardening Soil (HS) model (Schanz
slopes that would be an indication of the onset of undesired et al, 2009) however, require inputs from advanced triaxial
performance. In addition, LE models are not capable of testing. These data are easily acquired for most soils,
showing progressive failure within cut slopes or adequately however in HOC (heavily overconsolidated) soils, sampling
deal with stress rotations that occur over time. The criteria and testing is not straight forward.
that relate instrumentation readings to mitigative actions Stiff-fissured clays are typically characterised by their
meant to minimize and control against poor performance brittle behaviour during laboratory testing. However, in
were developed over the 35+ year history of mining in the practice, the reliability of the measured strength and
oil sands and observations of behavior due to unloading. stiffness results can be greatly reduced through sample
With the advancement of computational power, the ability disturbance. Skempton (1964, 1970, 1977, 1985), Bjerrum
to measure in-situ stress-strain characteristics, and mines (1967), Wroth (1984), Fahey (1998), Whittebolle (1982)
advancing into locations markedly different than the pits and many more have shown for fissured HOCs it is difficult
that led to the development of the instrumentation criteria, to obtain a representative sample that is reliable for
application of finite element (FE) numerical models is the laboratory strength measurements. Fissures represent
future. discontinuities within the soil mass, and sample sizes may
The accuracy of calculated displacements for slopes be inadequate to fully capture the overall macrostructure of
and walls from FE models of is typically poor when the soil. If however a series of macro or micro fissures are
compared to measured values. Inaccuracies have been captured in a given test specimen, then preparation of the
attributed to the general simplicity of most linear elastic / sample for testing is often fraught with difficulties and prone
perfectly plastic, constitutive models. Because to pre-mature failure. As a result, it is typical for one to
deformations prior to yielding can be considerable, the select only intact samples for testing. The resulting strength
assumption of linear elasticity is often invalid and tends to values generally represent the upper bound for the soil
mass, while the overall material strength is in actuality geotechnical engineering practice. Figure 2 provides a
closer to the residual value and controlled by the spacing summary of typical stiffness-strain behaviour of soils with
and persistence of the fissures. typical strain ranges for laboratory tests and structures.
The clay shales characteristic of the Canadian Prairies
are generally closer in nature to HOC soils as opposed to
typical shales that are generally indurated and of higher
quality. Certainly, within the upper reaches of the shale 1
Seismic (DST)
Retaining Walls
within Alberta, the clayshale is considered a bedrock by
age only and, if classified using ISRM standards based on Functionalities

Shear Modulus Gi / G0
Uniaxial Compressive Strength (UCS), would be Tunnels
considered an R0 to R1 material. An example of the Very Conventional Soil Testing
distribution of soil undrained shear strength, su versus the Small
Strains Small Strains
UCS strength of various rocks is shown below in Figure 1.

Shear Strain, γ
1e-6 1e-5 1e-4 1e-3 1e-2 1e-1

Dynamics Methods High-Resolution Pressuremeter

Local Gauges

Figure 2. Non-linear stiffness-strain behaviour (adapted

from Atkinson & Sallfors, 1991)

This paper presents the analysis of Cambridge PM data

Figure 1. Distribution of soil shear strength to ISRM UCS within the Clearwater Shale at the Aurora North mine
classification of rocks. located north of Fort McMurray, Alberta. The analysis
focuses on the determination of the non-linear shear
modulus and compares the results with the London Clay.
Since stiff, fissured HOC clays present significant Based on the interpretation of the DST and PM data,
difficulties with sampling and testing, their nature is highly stiffness-strain curves for the Clearwater are developed
conducive to testing with various in-situ test methods. between strain increments of 10-6 and 10-2.
Wroth and Hughes (1973) first presented the development
of the Cambridge style Self-Boring pressuremeter (SBPM). 2 BACKGROUND INFORMATION
Since its inception, Cambridge In-situ has advanced the
design and resolution of the probe to be capable of 2.1 Site Information
accurately measuring the shear modulus of a soil directly.
Bellotti et al. (1989), Muir Wood (1990), Jardine (1991 & The Clearwater Formation has a thickness of between 30
1992), and Bolton and Whittle (1999) all demonstrate the and 40 m based on the borehole information. The
use of the SBPM for evaluation of the non-linear shear Clearwater consists of transgressive/ regressive
modulus of a soil based on unload-reload cycles. sequences of moderately bioturbated clay, silty clay and
Syncrude has recently carried out an extensive field glauconitic silt/sand. Figure 3 shows the distribution and
program within the Clearwater Shale using a high typical thicknesses identified in the study area.
resolution pressuremeters operated by Cambridge Insitu.
Each test was evaluated for both undrained strength 2.2 High-Resolution Pressuremeter Testing
parameters and linear secant and non-linear secant shear
moduli. In addition to the PM testing, downhole seismic The pressuremeter instruments used were designed and
testing (DST) was carried out to determine the very small manufactured by Cambridge Insitu Ltd. The Self-Boring
strain shear modulus at each PM test interval. DST was Pressuremeter Testing (SBPM) has an outside diameter of
conducted following completion of the pressuremeter 88 mm and has a test interval length of approximately 0.5
program and consisted of vertical seismic profiling using 70 m. The Reaming Pressuremeter (RPM) is a rigid cylindrical
mm diameter slope indicator casings installed into the probe a little less than 1 m long and 47 mm in diameter.
pressuremeter test holes. All DST data acquisition and Using these dimensions, the volume of soil sheared during
evaluation of the corresponding shear wave velocities was a given test is approximately 3 orders of magnitude more
conducted by ConeTec Investigations Ltd. than a conventional 38 mm diameter triaxial sample.
The assessment of displacement is associated with soil The pressuremeter test produces an expansion and
stiffness and loading rate for a given soil. Research contraction curve assuming a right cylindrical cavity. The
conducted originally by Hardin and Drnevich (1972) and Cambridge probe measures the radial displacement of the
later supported by Jardine et al. (1986), Jardine (1991) and probe at the borehole wall for a given applied radial stress.
Burland (1989), show that HOC stiffness decreases with Within clay soils, the PM test is performed assuming
increasing strain, showing a non-linear behaviour for shear undrained conditions, therefore the shear strain is derived
strains generally less than 1%. Atkinson & Sallfors (1991) from circumferential strain. In this research it is assumed
illustrate the use of non-linear elasticity in typical that the cavity expands as a right cylinder, therefore, the
average radial (cavity) strain of the 6 probe arms has been 3 STIFFNESS ANALYSIS
used for analysis and stiffness anisotropy has not been
considered. Stiffness is a property that depends mainly on the soil
mineralogy, structure, the degree of cementation, fissuring
A total of 13 pressuremeter tests were performed in 3 or discontinuities. It also depends on the stress path, strain
boreholes as shown in Table 1. path and strain rate as well as and the chosen strain
increment (Sorensen, 2007). The stiffness at small
deformations is typically assumed to be linear and is
directly measured through seismic tests and generally
assumed to be consistent with a strain increment of 10 -6.
Figure 2 shows that at strain increments greater than the
10-6 value, the soil’s stiffness becomes non-linear,
decreasing with increased strain up to a residual (critical
state) value.

3.1 Shear Modulus

Because the boundary conditions of the PM are well known

and the soil is subjected to pure shear during borehole
expansion, the soil shear modulus can be determined with
minimal uncertainty. The simplest interpretation of shear
modulus is the determination of the slope of the bisector
extending through the top and bottom of each unloading-
reloading cycle (Gur). If the bisector is drawn through the
radial stress-cavity strain field curve, the slope must be
divided by 2 to result in a shear modulus. Because there
are an infinite number of moduli that can be calculated from
any given unload-reload cycle, the secant modulus through
the top and bottom of a given cycle is the lowest shear
modulus for that test and the highest strain increment. The
initial shear modulus (G0) is not typically calculated from
the PM as it tends to overestimate the true value.
For this paper, the analysis of the minimum secant and
non-linear shear moduli were calculated from 13 PM tests.
At each of the test intervals, between 3 and 4 unload-reload
Figure 3. North Mine Stratigraphy (from Syncrude, 2022) cycles were performed during either the loading or
unloading phase of each test. For each cycle, the non-
linear secant shear modulus was also calculated using the
Table 1. Characteristics of tests performed. method described by Bolton and Whittle (1999).
# U-R Elevation The tests within the Clearwater indicated strong
Borehole Test Material Instrument similarity to those documented in the London Clay (Muir
Cycles [mASL]
Wood, 1990; Whittle, 1999). The comparison to the London
22-7030 Test 1 3 KCB 263.90 SBPM
Clay could prove valuable for design in the Clearwater
Test 2 3 KCB 262.13 SBPM given the wealth of historical information on the strength
Test 3 3 KCA 258.90 SBPM and stiffness profiles. To assess whether there is a
relationship between the London Clay and the Clearwater,
Test 4 3 KCA 256.94 SBPM
a comparative analysis has been conducted with historical
21-7002 Test 1 3 KCC 264.90 SBPM data obtained in triaxial and pressuremeter tests.
Test 2 4 KCC 260.60 RPM Figure 4 shows the shear modulus calculation for
unload-reload cycle #3 of Test 4 in borehole 22-7030 and
Test 3 3 KCB 256.40 RPM Table 2 shows in summary the minimum secant shear
Test 5 4 KCA 248.50 RPM modulus values determined in each of the loops performed
SBPM
on the Clearwater Formation.
Test 6 3 KCA 247.60
21-7004 Test 1 3 KCC 248.50 RPM
Test 2 4 KCC 279.50 RPM
Test 3 4 KCB 274.30 RPM
Test 4 4 KCA 268.30 RPM
 Where α and β are material parameters determined
from only the reloading portion of the unload-reload cycle
in an PM test. Only the reload portion is selected, given the
difficulty in determining an appropriate strain origin for
analysis. Jardine (1992) discuss the impact of rate effects
on strain origin determination from the unloading phase.
Despite the pressure typically being held prior to unloading,
the actual point where the borehole stops expanding from
either creep or consolidation is not clear. Often, the ‘start’
of unloading is assumed where the reloading portion
crosses the unloading branch. Though not necessarily
incorrect, the accuracy of this assumption negates the
evaluation of the small strain (10-4) moduli. This level is the
smallest value that is reasonably measured using the
technology installed in a Cambridge style probe. It is also
important to note that Jardine (1992) suggests that the cut-
off strain of 10-4 is sufficient for assuming linear elasticity.

Figure 4. Determination of pressuremeter shear

modulus unload/reload (Gur) by linear regression (Tests 2
and 4 Borehole 22-7030).

Table 2. Shear moduli determined from unload-reload

cycles.
Gur [MPa]
Borehole Test
U-R 1 U-R 2 U-R 3 U-R 4
22-7030 Test 1 28.6 25.7 36.5 -
Test 2 104.3 87.4 74.6 -
Test 3 75.0 69.5 86.8 -
Test 4 93.8 73.0 72.0 -
21-7002 Test 1 74.0 68.0 53.0 - (a)
Test 2 28.0 27.0 26.0 39.0
Test 3 23.0 23.0 28.0 -
Test 5 32.0 31.0 35.0 32.0
Test 6 90.0 89.0 97.0 -
21-7004 Test 1 80.0 70.0 67.0 -
Test 2 47.0 52.0 92.0 60.0
Test 3 112.0 76.0 64.0 82.0
Test 4 81.0 77.0 74.0 83.0

3.2 Non-linear stiffness

As discussed above, the stiffness of a soil is not linear. In

most cases however, most constitutive models assume
that the modulus is represented linearly to simplify
calculate deformations. The non-linear behaviour of the (b)
shear stress - shear strain for HOC has been extensively
studied by Jardine et al. (1984, 1986), Burland (1989) and
Hight et al. (2007). The variation of shear modulus (Gi) with Figure 5. Secant shear modulus in Borehole 21-7004.
increasing shear strain can be evaluated discretely (Muir a) each U-R cycle b) Simplified trend for each test.
Wood, 1990) from the small strain version of Palmer (1972)
or fitted by the power law of Bolton and Whittle (1999) using Figure 5 shows the theoretical behaviour of the secant
equation 1. shear modulus for small deformations. These curves were
created from the measurements of the α and β variables of
each reloading loop shown in Figure 5a. To check the
[1]
relative accuracy of the analysis, it would be expected that
if the Clearwater was behaving as an undrained material,
then the modulus degradation curves from a given test The secant shear moduli results obtained from the
interval should plot on top of one another. The plots should Clearwater Formation were compared to data provided by
be similar because if the soil is truly undrained, there is no Jardine (1992) for the London Clay with an OCR of 3.0 and
consolidation that is possible (zero volume change) and to the London Clay analyzed by Whittle & Liu (2013).
therefore the effective stress at the borehole wall following Figure 7 shows the nonlinearity of the shear modulus for
yielding remains constant regardless of the stress state. shear strains between 10-4 and 10-1. From the figure it is
A representative reload cycle of each test interval observed that the Jardine (1992) clay presents a higher
shown in Figure 5a was chosen for further analysis as effective stress therefore the initial modulus is almost twice
shown in Figure 5b. At first glance of the data in Figure 5b, that of the Clearwater shale. In the case of Whittle & Liu
it would appear that the material is quite different in each (2013) clay, the modulus degradation curve is nearly
test as the shear modulus appears to increase for each identical to the Clearwater clay.
test, however, if the calculated shear moduli are
normalized to the vertical effective stress, the degradation
curves are quite similar suggesting that the Clearwater’s
characteristics are consistent with depth and conducive to
normalization as shown in Figure 6. Table 3 shows the
selected non-linear modulus parameters for Clearwater
Formation.

Figure 7. Comparison of secant shear modulus for

Clearwater shale and London Clay.

In order not to bias the behaviour by the initial stress

Figure 6. Secant shear modulus normalized in Borehole state, the secant shear modulus was normalized to the
21-7004 G0.01% modulus as recommended by Jardine (1992). This
strain was defined by Jardine as the cut-off between linear

Table 3. Selected non-linear modulus parameters.

α β Gs0.01% Gs0.1% Gs1% Es0.01% Es0.1% Es1% Gy G50 Gmax

Borehole Test
[MPa] [-] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa]
22-7030 Test 1 5.107 0.677 100 48 23 218 104 49 22 30 97
Test 2 17.350 0.703 267 135 68 591 298 150 53 81 219
Test 3 14.197 0.715 196 102 53 436 226 117 70 89 283
Test 4 13.446 0.675 268 127 60 583 276 131 52 71 230
21-7002 Test 1 11.348 0.684 208 101 49 456 220 106 53 74 231
Test 2 5.111 0.686 92 45 22 202 98 47 21 28 89
Test 3 4.304 0.644 114 50 22 244 108 47 20 30 97
Test 5 5.046 0.627 157 66 28 332 141 60 23 35 114
Test 6 15.218 0.686 274 133 65 600 291 141 76 104 326
21-7004 Test 1 12.688 0.703 196 99 50 432 218 110 52 70 216
Test 2 7.779 0.643 210 92 40 449 197 86 32 46 150
Test 3 19.025 0.724 242 128 68 540 286 152 59 77 234
Test 4 13.566 0.642 367 161 71 783 344 151 55 81 261
elasticity and non-linear elasticity. More importantly, as
described above, this strain represents the smallest strain
that can be reasonably measured with current
pressuremeter equipment. Non-linear shear moduli
normalized to the G0.01% for the Clearwater and London
Clays are shown in Figure 8. Once the stiffness
degradation data is normalized, it is clear that the data from
Jardine (1992) and Whittle & Liu (2013) is slightly lower.

Figure 9. Small-strain Shear modulus from Downhole

Seismic Testing (21-7004).

Table 4. Modulus from Shear Wave Velocity (Vs)

Yield
Elevation Shear G0
Borehole Modulus
[mASL] [MPa] [MPa]
Test 1 264.9 53 124.07
Test 2 260.6 21 140.92
21-7002 Test 3 256.4 20 128.91
Figure 8. Comparison of normalized secant shear moduli Test 5 248.5 23 281.57
for Clearwater shale and London Clay. Test 6 247.6 76 288.19
Test 1 284.5 52 167.85
Test 2 279.5 32 186.96
3.3 Small-strain Shear Modulus 21-7004
Test 3 274.3 59 172.89
Test 4 268.3 55 213.62
The shear modulus for small strains was calculated from
the shear wave velocity (Vs) obtained in the Downhole Test 1 263.9 35 152.36
Seismic Test (DST). The small strain shear modulus may Test 2 262.1 27 112.45
22-7030
then be calculated based on the material density,  and the Test 3 258.9 48 235.77
measured shear wave velocity at the test interval using
Test 4 256.9 89 444.52
equation 2.

[2]
3.4 Shear modulus degradation
A plot of the shear modulus at small deformations along
Having the initial shear modulus (G0) and the secant shear
the borehole shows the difference in stiffness between the
modulus determined from the PM, a graphical coupling of
Clearwater formation and the McMurray formation. In
the moduli was performed using a sigmoidal curve model.
general, the modulus for small deformations of the
The purpose of this was to complete the stiffness
clayshale tends to be homogeneous with mean values
degradation curve and provide a link between the in-situ
between 186 and 236 MPa. The minimum standard
test data and inputs required for non-linear elastic
deviation was measured in Borehole 21-7004 at 20 MPa.
constitutive models like the hardening soil or small strain
Figure 9 shows the shear modulus behaviour with
hardening soil model. The key unknown using the
depth in Borehole 21-7004, including the PM Gur modulus
sigmoidal model, is the performance of the soil in strains
and Gs modulus values for an equivalent strain of 0.0001.
between 10-6 (seismic) and 10-4 (PM). Figure 10 shows the
It is expected that the Gs values calculated from the unload-
field data from borehole 21-7004, fitted using a sigmoidal
reload cycles will be less than those calculated by the DST
model. Where the transition in the curve from linear
given the change in strain increment. The G0 obtained from
elasticity to non-linear behaviour is not clear and has been
the DST is considered truly linear elastic and therefore
assumed. At this time, it is believed that the method
must be higher than values at higher strain increments.
described by Smith et al. (2018) will fill in the missing data
Table 4 presents a summary of the shear moduli
and complete the small strain non-linear modulus profile.
determined from the DST for the 3 boreholes analyzed.
Data has been acquired during this research program and situ and ConeTec, Downhole Seismic testing was carried
is currently being processed. out by ConeTec.

6 REFERENCE

Atkinson, J., Sallfors, G. 1991. Experimental determination

of soil properties. In Proc. 10th ECSMFE, 3, 915–956.

Bellotti, R., Ghionna, V., Jamiolkowski, M., Robertson, P.

K., & Peterson, R. W. (1989). Interpretation of moduli
from self-boring pressuremeter tests in sand.
Geotechnique, 39(2), 269–292.

Benz, T., et al. 2009. Small-Strain Stiffness in Geotechnical

Analyses. Bautechnik, vol. 86, no. SUPPL. 1, pp. 16-
27–27.

Bjerrum, L.1967. Terzaghi Lecture “Progressive failure in

overconsolidated clays and clay shales”. ASCE JSMFE
Vol.93 No. SM5:3-49

Burland, J. B. 1989. Ninth Laurits Bjerrum Memorial

Lecture: “Small is beautiful” – the stiffness of soils at
small strains. Canadian Geotechnical Journal. Revue
Canadienne de Geotechnique, 198911, Vol. 26, Issue
4, pp. 499-516
Figure 10. Shear modulus degradation borehole (21-7004).
Clayton, C. R. I. (2011). Stiffness at small strain; research
and practice. Geotechnique, 61(1), 5–37.
4 SUMMARY AND CONCLUSION
Fahey, M. 1998. Self-Boring Pressuremerer testing in
calcareous soil. Balkema, Rotterdam. ISBN
As Jardine (1992) indicates “the behaviour over the
9061917549.
nonlinear strain range 0.001%<ε<1% is crucially important
in controlling the ground’s response in a wide range of civil
Hardin, B.O., and Drnevich, V.P. 1972.Shear modulus and
engineering problems”.
damping in soils: measurement and parameter effects.
To evaluate the stiffness and non-linear behaviour of
ASCE Journal of the Soil Mechanics and Foundation
the Clearwater Formation Clayshale of the Aurora North
Division, 98 (SM6):603-624.
Mine north of Fort McMurray, 13 PM high-resolution
pressuremeter tests were conducted. The nonlinear shear
Hight, D. W., Gasparre, A., Nishimura, S., Minh, N. A.,
modulus behaviour for small strains was determined by
Jardine, R. J. & Coop, M. R. (2007). Géotechnique 57,
Bolton and Whittle (1999) power law, a comparison was
No. 1, 3–18
made with the degradation of the London Clay, and the
initial shear modulus was determined from the shear wave
Jardine, R.J. 1991. Discussing Strain-dependent moduli
velocities of the DST test.
and pressuremeter tests. Géotechnique 41, No. 4., pp
The shear modulus and elastic modulus ranges of
621-624.
Clearwater Clayshale and its behavior for different strain
levels were defined.
Jardine, R.J. 1992. Nonlinear stiffness parameters from
The graphical representation of modulus degradation
undrained pressuremeter tests. Canadian
generated using a conventional sigmoidal model.
Geotechnical Journal 29(1): 436–447.
It is important to complement this study at a more
detailed level, evaluating the anisotropic behavior of the
Jardine, R.J., Symes, N.J., and Burland, J.B. 1984. The
borehole stiffness and the identification of weak contacts.
measurement of soil stiffness in the triaxial apparatus.
Géotechnique 34(3): 323–340.
5 ACKNOWLEDGMENTS
Jardine, R.J., Potts, D.M., Fourie, A.B., and Burland, J.B.
This research has been carried out as part of an NSERC
1986. Studies of the influence of non-linear stress–
Alliance Grant supported by Syncrude Canada Ltd.,
strain characteristics in soil–structure interaction.
ConeTec Investigations Ltd and Cambridge In-Situ Ltd.
Géotechnique 36(3): 377–396.
The researchers at the University of Saskatchewan and the
University of Alberta would like to thank all participants.
Pressuremeter testing was carried out by Cambridge In-
Liu, L., Chalaturnyk, R., Deisman, N., and Zambrano- Wittebolle, R.J. (1983). The influence of microfabric on the
Narvaez G. 2020. Anisotropic borehole response from engineering properties of glacial till. MSc, Thesis.
pressuremeter testing in deep clay shale formations. Department of Civil and Environmental Engineering,
Canadian Science Publising. 1159-1179. University of Alberta, Edmonton, AB.

Marsland, A., and Randolph, M.F. 1977. Comparisons of Whittle, R.W 1999. Using non-linear elasticity to obtain the
the results from pressuremeter tests and large in situ engineering properties of clay. Ground Engineering,
plate tests in London Clay. Géotechnique,27(2): 217– May, vol. 32, no.5, pp 30-34.
243.
Whittle, R., and Liu, L. A method for describing the stress
Muir Wood, D. 1990. Strain dependent soil moduli and and strain dependency of stiffness in sand. (2013)
pressuremeter tests. Géotechnique, 40, pp 509-512. Proceedings of the 18th International Conference on
Soil Mechanics and Geotechnical Engineering, Paris.
Palmer, A.C. 1972. Undrained plane-strain expansion of a
cylindrical cavity in clay: a simple interpretation of the Wroth, C. P., Hughes, J. M. O., Baguelin, F., Jezequel, J.,
pressuremeter test, Géotechnique 22 No. 3 pp 451- & Palmer, A. C. (1973). Undrained plane-strain
457. expansion of a cylindrical cavity in clay; a simple
interpretation of the pressuremeter test [discussion and
Randolph, M.F., and Fahey, M. 1984. Effect of disturbance reply]. Geotechnique, 23(2), 284–292.
on parameters derived from self-boring pressuremeter
tests in sand. Géotechnique, 34(1): 81–97. Wroth, C. P. (1984). The interpretation of in situ soil tests.
Geotechnique, 34(4), 449–489.
Schanz, T., Vermeer, P.A., Bonnier, P.G. (1999). The
hardening-soil model: Formulation and verification. In
R.B.J. Brinkgreve, Beyond 2000 in Computational
Geotechnics, Balkema, Rotterdam. 281–290.

Schnaid, F. (2005, September). Geocharacterisation and

properties of natural soils by in situ tests. In
Proceedings of the International Conference on Soil
Mechanics and Geotechnical Engineering (Vol. 16, No.
1, p. 3).

Simpson, B. (1992). Retaining structures: displacement

and design. 32nd Rankine Lecture. Geotechnique 42,
No. 4, 541–576.

Skempton, A. W. 1964. Long-term stability of clay slopes.

Geotechnique, 14(2), 77–101.

Skempton, A. W. 1970. First-time slides in over-

consolidated clays. Geotechnique, 20(3), 320–324.

Skempton, A. W. 1977. Slope Stability of Cuttings in Brown

London Clay. Int. Conf. Soil Mechanics. Tokyo, 3, 261-
270.

Skempton, A. W. 1985. Residual strength of clays in

landslides, folded strata and the laboratory.
Géotechnique, 35(1), 3–18.

Smith, L., Elwood, D., Barbour, S.L., & Hendry, M.J., 2018.
Profiling of the in situ compressibility of cretaceous
shale using grouted-in piezometers. Geomechanics for
Energy and the Environment, 14, 29-37

Sorensen, K. K., Baudet, B. A., & Simpson, B. (2007).

Influence of structure on the time-dependent behaviour
of a stiff sedimentary clay. Geotechnique, 57(1), 113–
124.
A CASE STUDY OF GEOTECHNICAL
ASPECTS OF WHARF CONSTRUCTION
AS A PART OF MARINE DOCKING
FACILITIES IN VOISEY’S BAY MINE
Ripon Karmaker and Paul Deering
Stantec Consulting Ltd., St. John’s, NL
Dan McQuinn
Stantec Consulting Ltd., Dartmouth, NS
Arun Valsangkar
University of New Brunswick, Fredericton, NB

ABSTRACT
This paper describes a case study presenting geotechnical aspects of wharf construction with a marine dock as part of
developing nickel-copper-cobalt ore processing facilities at Voisey’s Bay mine site in 2004. The face of the wharf foundation
is constructed using four AS 500 circular cells joined together with six arcs, which also act as a retaining structure for
backfill materials. The very soft clay deposit at the sea bottom within the cellular cells and connecting arcs were not
dredged, necessitating instrumentation and wharf construction monitoring. The paper summarizes the subsurface
conditions in the proposed wharf area based on numerous boreholes drilled. Details of the instrumentation program
consisting of inclinometers, earth pressure cells, piezometers, and strain gages are presented. The data from the
instrumentation obtained during construction and post-construction demonstrates the successful application of the
observational approach for constructing the wharf structure in challenging subsurface conditions coupled with a short
construction season.

RÉSUMÉ
Cet article décrit une étude de cas présentant les aspects géotechniques de la construction d'un quai avec un quai marin
dans le cadre du développement d'installations de traitement de minerai de nickel-cuivre-cobalt dans la mine Voisey's Bay
en 2004. La face de la fondation du quai est construite à l'aide de quatre cellules circulaires AS 500 reliées par six arcs,
qui agissent également comme une structure de soutènement pour les matériaux de remblai. Le dépôt d'argile très molle
au fond de la mer dans les cellules cellulaires et les arcs de connexion n'a pas été dragué, ce qui a nécessité une
instrumentation et une surveillance de la construction du quai. Le document résume les conditions du sous-sol dans la
zone de quai proposée en fonction des nombreux trous de sonde forés. Les détails du programme d'instrumentation
composé d'inclinomètres, de capteurs de pression de terre, de piézomètres et de jauges de contrainte sont présentés. Les
données de l'instrumentation obtenues pendant la construction et après la construction sont présentées pour démontrer
l'application réussie de l'approche d'observation pour la construction de la structure du quai dans des conditions
souterraines difficiles associées à une courte saison de construction.

1 INTRODUCTION an approximate 100-meter berthing face with a minimum

draught of 13.5 meters (ArcelorMittal 2007; CCE 2007).
The Voisey’s Bay Mine site is located approximately 25 km Westmar Consultants Inc. (marine structural design) and
southwest of Nain, NL, on the coast of northern Labrador. Stantec (Geotechnical design) jointly submitted the design,
Access to the mine site is by air, and an airstrip serves the which consisted of a circular steel sheet pile gravity
site on the north side of the site. There are port facilities structure. The face of the wharf foundation is constructed
located to the north on Anaktalak Bay that are used for bulk using four AS 500 circular cells joined together with six
and freight shipping. There is no other developed land or arcs, which also act as a retaining structure for backfill
sea access to the site. materials. A plan view and a typical cross-section of the
wharf are presented in Figures 1 and 2.
The mine is owned and operated by Vale, one of the
world's largest integrated mining companies, with global The construction of the wharf structure was considered
headquarters in Rio de Janeiro, Brazil. Open pit mining for one of the most challenging components of the entire mine
nickel-copper-cobalt ore commenced in 2005, with current development due to the following constraints:
underground developments ongoing that will extend the
mine life to at least 2035. ▪ No Dredging of very soft sediments
▪ Construction within one short ice-free season
The port facility comprises a wharf, a ship loader, and ▪ Robust structure under aggressive arctic operating
a concentrate storage building. Construction of the wharf conditions
began in 2004 and was completed in 2005. The wharf has ▪ Economical form of construction
 site information, providing input regarding recommended
To meet these constraints, an instrumentation plan was additional field investigation, and performing geotechnical
developed to continuously monitor the stability of the analyses.
structure during sheet pile installation and backfilling.
Additionally, a set of curved precast concrete ice impact Based on the information obtained from the marine drill
panels with a reinforced cope beam system was built inside holes, labeled as “Old Borehole” in Fig. 4, the report
to deal with Fednav’s icebreaking bulk carrier. Construction identified that the offshore soils at the wharf generally
of the wharf began in the summer of 2004 (Fig. 3), and the consist of very loose to loose granular soils over a soft to
main structure was completed in December 2004. firm clayey silt layer, underlain by dense granular soils,
then bedrock. The upper layer of granular soils, consisting
of silty sand with gravel, varies in thickness from 1.5 to 3.5
m. Boulders were encountered below Elevation -19 m in
most drill holes, with frequency increasing with depth. The
surface of the bedrock was encountered at a depth of 6.5
m at the shoreline, increasing to approximately a 12.5 m
depth (i.e., Elev. -25 m) below the seabed at the berth face.
Finally, the report recommended additional marine drill
holes for geotechnical design.
Fig 1. Typical plan view of Wharf structure (After The report also recommended considering the impact
ArcelorMittal 2007) of ice loading during the design of the wharf foundation.
However, Westmar Consultants Inc. considered it in their
design by reinforcing the sheet pile structure with concrete
ring wall bracing.

2.2 Geotechnical Investigations and

Recommendations

Based on the recommendations made in the feasibility

study, seven supplementary marine drillings were carried
out by Stantec in 2003, as noted in Fig. 4 as “New
Borehole.” Borehole depth varied from 4.42 m to 21.3 m.
Six Dynamic Cone Penetration (DCP) tests were also
carried out.
Fig 2. Typical cross-section of sheet pile cell (after
ArcelorMittal 2007) The geological soil conditions encountered in this area
were typically glacio-fluvial outwash sequences deposited
The main objective of the present paper is to represent in a marine environment. Consequently, stratigraphic
the geotechnical investigations and the instrumentation layers tend to be discontinuous and often complex to
data to demonstrate a successful observational approach extrapolate from the existing borehole information.
for constructing a wharf foundation in challenging However, the following three distinct layers of subsurface
subsurface conditions within a limited construction period. condition overlying bedrock were recommended by
Stantec for design purposes (JWA 2003a; JWA 2003b):

Layer 1: quite variable in nature, found at the surface,

referred to as clayey sand;
Layer 2: consists of soft to firm, lean clay; and
Layer 3: consists of compact to dense sand with gravel,
cobbles, and boulders.

SPT N-value and DCPT values are presented in Fig. 5.

Based on the in-situ strength interpretation, clayey sand in
Fig 3. Aerial photographs during construction of the Wharf or near the surface is loose to compact relative density, and
in 2004 lean clay is soft to firm in terms of consistency. The
thickness of the lean clay layer varies from 1 m to 2.5 m in
2 PRE-CONSTRUCTION INVESTIGATIONS various boreholes.

2.1 Feasibility studies Gradation analysis results and Atterberg limits for the
disturbed soil samples are presented in Table 1. The lean
Stantec (previously Jacques Whitford and Associates sandy clay consists of an average of 86% silt/clay with a
Limited (JWA)) carried out a feasibility study in 2002 (JWA plasticity index range from 10 to 14. In lean clay samples,
2002). The study included reviewing the available soils and the liquidity index was observed to be greater than unity,
as shown in Fig. 6, which might indicate the presence of
sensitive clay deposits within the foundation soils.

Consolidation and triaxial testing were completed on

the clay samples, and the preconsolidation pressures on
the order of 300 kPa were estimated with undrained shear
strengths ranging from approximately 20 to 50 kPa. Four
Consolidated Undrained (CU) triaxial tests were carried out
with samples collected with the Shelby tube sampler. The
average drained cohesion (c') and internal friction angle
(ɸ') obtained for the Silty Clay (CL) were 10 kPa and 30°,
respectively.

Fig 5. In-situ test results: (a) SPT N-value; (b) DCPT

Table 1. Index properties of the soils

Gradation Atterberg
Analysis Limits

Gravel (%)

Fines (%)
Sand (%)

w (%)

PL (%)
LL (%)

PI (%)
Description

Fig 4. Borehole location around the proposed docking

structure (JWA 2003a)
Silty clayey Sand (SC-SM)

57.1

37.3

18.6
6.5
The shallowest elevation of Bedrock was encountered to Clayey Sand (SC)
at Elev. -22.3 m in borehole BH-N1. The bedrock was
described as light to dark grey gneiss, with fractures Sandy Lean Clay (CL) to
13.2

85.9

30.9
0.9

31

19

12
generally occurring at 70 to 80 degrees to the core axis. Lean Clay (CL)
Based on the Rock Quality Designation (RQD) criteria, the
rock is classified as severely fractured to intact, with RQD 3 CONSTRUCTION OF WHARF STRUCTURE
ranging from 43 to 100%.
The dock has an approximately 100 m berthing face with a
Based on the findings in the geotechnical investigation, minimum draught of 13.5 m, consisting of four cells and six
Stantec issued a geotechnical design brief in 2003 for the connecting arcs, as shown in Fig. 1. The face of the
wharf construction, in which geotechnical instrumentation marginal wharf was built with four AS 500 cells, which
was recommended to establish an observational approach acted as earth retaining structure of backfill materials within
during construction (JWA 2003c). the cells. Each cell was constructed with 150 AS 500
straight-web sheet piles, and four junction piles had a
Sheet piles for the cellular wharf were recommended to diameter of 24.7 m. Each of the six arcs comprised 14
be driven into the dense sand with a gravel layer using a normal AS 500 piles and 12 bent piles (bent by 7°) in
refusal criterion that would prevent tip damage. Although
alternative positions. A minimum interlock strength of 5,500
the available borehole information showed that the
kN/m was ensured between the piles during operation
existence of a significant number of boulders within the
(ArcelorMittal 2007). Concrete ice-impact beams were
sediments was unlikely, it was also recommended to probe
installed to take the horizontal loads.
the sheet-piling location to mitigate the risk of encountering
boulders.

Placing the rockfill behind the cellular structures was

recommended with caution to avoid any possible mounding
of clay around the sheet piles or creating a mud wave.
 Liquidity Index interpretation and pile data of Cell #4 are represented in
-3.0 -2.0 -1.0 0.0 1.0 2.0 3.0
Fig. 8. Below the seabed elevation, clay base, and bedrock
-12 elevation were interpreted from the borehole information.
10 20 30 40 50 An inferred layer is interpreted between the boundary of the
Liquid Limit/Plastic Limit/Water Content (%) clay base and bedrock, referred to as the cobbles layer. As
-14 the proposed wharf foundation was a gravity structure, the
sheet pile system did not necessarily require embedment
into the lower soil layer (ArcelorMittal 2007). Pile tips were
-16 driven near the cobble layer to avoid possible damage to
Elevation (m)

Liquid Limit sheet piles during driving. A total of 154 sheet piles were
Plastic Limit driven in Cell #4 between the clay base and bedrock layers,
-18
Water Content as shown in Fig 8.
Liquidity Index

(a) (b)
-20

-22

-24

Fig 6. Index properties of the Lean clay deposit (c) (d)

3.1 General installation procedures

Cells were installed on the site by using two template

platforms. The platforms were first installed with the help of
supporting pipe piles, as shown in Fig 7(a). the lower
platform was positioned as low as possible inside the
water, while the upper platform was installed as high as
possible from the water surface. Special junction sheet Fig 7. Construction procedures: (a) Installation of templates
piles were positioned vertically and then welded with the and supporting piles; (b) installation of sheet piles; (c)
upper platform. After that, adjacent sheet piles were driving of piles using staggered driving method; (d)
threaded to close the loop of cells between junction piles, backfilling of the cell
as shown in Fig 7(b).
0
After closing the cells, sheet piles were driven using
the staggered driving method, as shown in Fig 7(c). The
-5
upper platform was lowered to an appropriate position in
the next step, and all sheet piles were driven to the final
design depths. -10
Elevation (m)

After placing the sheet piles in position, backfilling using -15

well-graded angular blasted and crushed materials began
and raised/removed the platforms at appropriate times. -20
Finally, the rest of the cell was filled with backfill material
(Fig. 7(d)). -25

3.2 Sheet piles driving and backfilling inspection -30

0 20 40 60 80 100 120 140 160
Sheet Number
Stantec provided full-time inspection during sheet piling Clay Base - BH Interp. Cobbles - BH Interp.
operations and backfilling and recorded daily construction Rock - BH Interp. Pipe Pile Tips
reports from July to December 2004. Final Pile Tips Seabed Elevation

Geotechnical data derived from the investigations were Fig 8. Geotechnical interpretation of soil strata and pile
used to interpolate elevations of the various soil strata and data for Cell #4
develop profiles of subsurface conditions within the dock
area. The developed geotechnical profiles of the Stantec monitored backfilling of the sheet pile
subsurface conditions were used to determine the structures for the dock on-site to provide recommendations
minimum penetration depth requirements for the sheet pile on fill rates, duration of stoppages in filling operation, and
during piling operations. For example, the geotechnical fill placement methods.
 Backfilling of the cellular and arc structures was on 28 October 2004. A gradual reduction of total pressure
completed using a jaw-run (150 mm minus) rockfill of in Cell # 2 was observed as the days elapsed after the
potentially acid-generating (PAG) and non-PAG blast rock. completion of fill placement. The cause of reduction was
PAG rockfill was used below the lowest normal tide, with inferred to be localized arching over the load cell.
the remaining fill to Elev. +3 m being non-PAG rockfill
material. Testing was carried out during the production of 400 18
the fill materials by the on-site soil laboratory. The fill
materials are typical of 30-40% sand-sized particles, with 15
the remainder being gravel to cobble-sized rock. Filling to 300

Total pressure (kPa)

final subgrade elevation was achieved using conventional 12

Fill Height (m)

fill placement and compaction methods.
200 9
3.3 Instrumentation during Construction
Total Pressure @ Cell #1 6
As part of the design for the dock structure, four types of 100 Total Pressure @ Cell #2
instrumentation were installed in the first cell constructed FH @ Cell #1 3
(Cell #4) to monitor the effects of filling on the sheet pile FH @ Cell #2
structure and the underlying clay stratum within the 0 0
confined sheet pile cell. The results and performances of 0 10 20 30 40 50 60 70 80
these instruments were used to determine fill rates to be Elapsed time (days)
applied during the filling of the remaining cell and arc Fig 10. Comparison of total pressure with filling height
structures. during backfilling at Cell #4

The instrumentation included earth pressure cells, 3.3.2 Piezometers

piezometers, strain gauges, and inclinometers. The
location plan for instrumentation is shown in Fig 9. Vibrating wire piezometers were installed within the sheet
pile structures at four locations, as shown in Fig. 9.
Installation was completed by advancing drill casings by
the wash bore method to the required depth. The
placement of the piezometer at the specified location within
the clay stratum was confirmed by obtaining a sample of
the clay before the installation of the piezometer. Piezo 1,
Piezo 2, Piezo 3, and Piezo 4 were installed at Elev. -19.1
m, -18.3 m, -15.4 m, and -15.3 m, respectively.

Readings were obtained from the piezometers at pre-

set times during the backfilling operations to measure the
excess pore water pressure in the clay layer within the
sheet pile structure. The excess pore water pressure within
the clay layer increased as fill placement inside the cell
progressed and was closely monitored as it neared the
Fig 9. Instrumentation location plan– Cell #4 maximum allowable pressure. A threshold value was
established prior to the beginning of fill placement to
3.3.1 Earth pressure cell reduce the risk of pore pressure-induced shear failure of
the in-situ soils within the sheet pile cell structure.
Vibrating wire earth pressure cells were installed within the
cellular structure at two locations. The pressure cells were Figure 11 shows the changes in excess pore water
placed in aggregate bags and covered with concrete sand pressure with the changes in fill height during the backfilling
before placement on the seabed inside the sheet pile cell. period in Piezo 1 and Piezo 4. As backfilling progressed,
The concrete sand provided a bedding material for the excess pore water pressure reached its local peak, and
instruments and protection during rockfill placement. backfilling was stopped to allow the clay deposits to
dissipate the excess pore water pressure with time. After
Pressure Cell#1 and Pressure Cell#2 were placed at reaching the final grade of fill height on 28 October 2004,
Elev. -12.10 m and Elev. -8.70 m, respectively. Figure 10 the excess pore water pressures were observed to be
shows the increase in total pressures during the filling reduced with elapsed days.
operation in Cell#4. Backfilling started on 23 September
2004. The total pressure increased with the filling height In Field, the observations made during the filling of Cell
inside the cell structure till 25 October 2004. The total #4 determined that a fill rate of 750 m3/24 hours period was
pressure is higher at Cell#1, as the pressure cell is placed not to be exceeded and that filling operations be stopped
at a lower elevation. On October 26, 2004, fill height was for a minimum of five days at the time of the cell template
reduced due to the rearrangement of fill materials inside removal (El. +1.5 m) to allow excess pore pressures within
the cell structure. However, the final grade was achieved the clay layer to dissipate. These requirements were
followed when filling all cells and arcs to subgrade In Figs. 12(a) and (b), the maximum compressive
elevation. strains were observed in strain gauges 100-i-2 and 013-i-
2, while the maximum tensile strains were recorded in
100 18 strain gauges 100-o-3 and 013-o-3. However, the changes
in strains became constant with elapsed days, when the
Excess Pore Water Pressure, Umax (kPa)

15 final grade of filling height was attained inside the cell

80
structure.
12
60 Similar trends of strains were also observed in the

Fill Height (m)

9 strain gauges installed in sheet pile#061 and sheet
40 pile#152, as shown in Fig. 9(a). The measured strains were
6 within acceptable limits by structural designers.
Piezo1_Umax
20 Piezo4_Umax 400
3 100-i-1 100-i-2
Piezo1_FH 100-i-3 100-i-4
Piezo4_FH 200 100-o-1 100-o-2
0 0 100-o-3 100-o-4
0 10 20 30 40 50 60

uStrain
Elapsed Time (days) 0
0 20 40 60 80
Fig 11. Comparison of excess pore water pressure with Elapsed Days
-200
filling height during backfilling at Cell #4
-400
3.3.3 Strain Gauge
-600
Spot-welded vibrating wire strain gauges were installed on
(a)
four sheet piles prior to the beginning of driving. Thirty-two -800
gauges were installed, with four being welded to the inner 400
and outer sides of each sheet at predetermined locations,
as shown in Fig 9.
uStrain

0
Elapsed days
0 20 40 60 80
Readings were obtained from the gauges at pre-set
times during and after the backfilling operation to determine -400 013-i-1
013-i-3
013-i-2
013-i-4
the hoop stresses incurred by the sheet pile interlocks as 013-o-1 013-o-2
backfilling progressed. Hoop stresses were measured to -800
013-o-3 013-o-4

ensure the tensile strength of the pile steel was not

exceeded, resulting in declutching or unzipping of the sheet (b)
pile interlock. -1200

Figure 12(a) shows the arrangement of strain gages in Fig. 12 Strain developed on the inside and outside of Cell
sheet pile#100. Strain gauges 100-i-1, 100-i-2, 100-i-3, and #4 during backfilling: (a) Pile #100; (b) Pile #13
100-i-4 were welded at elevations +1.10 m, -4.90 m, -10.90
m, and -13.90 m, respectively, inside the cell structure, 3.3.4 Inclinometers
while 100-o-1, 100-o-2, 100-o-3, and 100-o-4 were welded
on the outside of the sheet pile#100 at the same elevation Inclinometer casings were installed at three locations within
as inside gauges. Figure 10(b) shows a similar the cellular structure. The casings were installed inside a
arrangement of strain gauges for sheet pile#13. Strain protective angle welded to the inside of the individual sheet
gauges 13-i-1, 13-i-2, 13-i-3, and 13-i-4 were welded at pile prior to driving. The casing was lowered into the angles
elevations +0.43 m, -5.57 m, -11.57 m, and -14.57 m inside following the completion of pile driving and backfilled with
of the cell structure, respectively, while 13-o-1, 13-o-2, 13- supplied concrete sand.
o-3, and 13-o-4 were welded outside of sheet pile #13. The
bottom elevations of sheet piles #100 and #13 were -18.57 Readings were obtained from these inclinometer
m and -21.90 m, while the seabed elevations were -5.44 casings during each shift while backfilling operations were
and -13.95 m, respectively. The seabed embedment of ongoing. The results of these readings were interpreted by
sheet piles #100 and #13 was 13.1 m and 7.95 m, Stantec each day to monitor the movements of the pile tips,
respectively. which would indicate lateral movement of the cell structure.
Inclinometer Data were recorded from 0.5 m to 25 m depth
During the backfilling process, no strain readings were with a 0.5-m interval along two axes (i.e., A-axis and B-
observed in strain gauges 100-o-4 and 013-o-4, which axis).
indicates the strain gauges were non-operational. With less
embedment into the seabed, strain gauges installed in An additional inclinometer casing was installed on the
sheet pile#13 show higher compressive strains on both outside of Cell#4, adjacent to sheet pile #140, as shown in
sides of the sheet pile. Fig 9, following the start of backfilling operations. This
casing was installed by driving a drill to the required depth
and installing the inclinometer casing inside. The reading
obtained from the fourth casing was to determine evidence one cell structure (Cell #4), findings from which were
of declutching of the adjacent sheet pile interlocks during applied to three other cells and six arcs construction.
filling operations. Lateral movement of the soil would
indicate declutching of the sheet pile and could be detected ACKNOWLEDGEMENT
from the resultant movement of the inclinometer.
The authors would like to express their sincere thanks to
Figure 13 shows the deflection of the inclinometer with Harold Kullman from Westmar Consultants Inc. and Barry
respect to depths during the filling operation. The positive Powers from Stantec Consulting Limited, who were actively
deflection values indicate the movement of sheet pile #12 involved during the construction of dock facilities.
toward the ocean side, while negative values indicate the
movement toward the shoreline. At the beginning of REFERENCES
backfilling, the bottom part of the sheet pile moved toward
the ocean while the upper part bent inward. However, the ArcelorMittal. 2007. Foundation Solutions for Projects:
upper part of the sheet pile also moved toward the ocean Harbour Construction, Innovative steel sheet pile
side when the cell structure almost backfilled around solutions for modern ports.
October 23, 2004, as shown in Fig. 13. After attainment of Canadian Consulting Engineer (CCE). 2007. Voisey’s Bay
the final grade in the cell structure on October 28, 2004, Wharf, Article, January 1, 2007.
very little deflections were observed in the inclinometer JWA. 2002. Feasibility Study: Geotechnical Engineering
between the period of October 31 and November 15, 2004. Review, Port Site, Voisey’s Bay, Labrador, Project No.
Deflection (mm) BCV40769.
JWA. 2003a. Factual Report: Supplementary Geotechnical
-80.00 -60.00 -40.00 -20.00 0.00 20.00 40.00 60.00
Investigation for Proposed Port Facilities, Port Site,
0
Voisey’s Bay NL, Canada, Project No. NFS08631-
Ocean side
0003.
Shore side
JWA. 2003b. Final Report: Geotechnical Investigation for
5 Proposed Marine Docking Facilities, Voisey’s Bay, NL,
Canada, Project No. NFS09166
JWA. 2003c. Final Report: Geotechnical Design
Parameters and Recommendations, Proposed Marine
10 Docking Facility, Port Site, Voisey’s Bay, Labrador,
Depth (m)

Day of filling Project No. NFS09133-201.

9/24/2004 JWA. 2005. Summary report of Geotechnical/Inspection
9/25/2004 services, Dock Structure, Voisey’s Bay Project,
9/26/2004 15
Anaktalak Bay, NL, NGL Project No. 10273-0003.
9/29/2004
10/1/2004
10/3/2004
10/10/2004 20
10/23/2004
10/31/2004
11/15/2004
25
Fig. 13 Deflections of inclinometer observed in sheet
pile#12 during backfilling operation

4 CONCLUSIONS

A feasibility study was carried out by Stantec prior to the

design of the wharf foundation in 2002. Based on the
recommendations made in the report, a detailed
geotechnical investigation was carried out in 2003. Three
distinct soil layers were encountered in the subsurface
conditions. A soft to firm layer of sensitive clay deposits
was identified, which could not be dredged out for
economic reasons and a short construction period.
Therefore, an instrumentation system comprising earth
pressure gauges, piezometers, strain gauges, and
inclinometers was developed instead to ensure a
successful construction of the wharf foundation.

Based on the instrumentation responses, the rate of

filling, duration of filling stoppage, and methods of filling
were determined. The instrumentation was developed in
Geotechnical Characterization of Iron
Tailings from a Mine in South America using
the Seismic Cone Penetration (S-CPTu) and
Laboratory Tests
F Espinoza & V Lanza
Klohn Crippen Berger, Belo Horizonte, Minas Gerais, Brazil
L Chahua & J Menacho
Klohn Crippen Berger, Miraflores, Lima, Perú
F Pereira & C Andrade
Vale S.A., Nova Lima, Minas Gerais, Brazil

ABSTRACT
This paper summarizes and discusses the results from the Seismic Cone Penetration Test (S-CPTu) and Laboratory Test
of iron tailings disposed of at different times, classified as either old or new tailings. This temporal variation generated
differences in undrained strength, stiffness, and microstructure. These tailings are discretized based on the Soil Behaviour
Type Index (Ic) to review contractive and dilative behaviour and identify possible zones of brittleness using different
interpretation approaches. This characterization will be complemented with the mineralogical composition information
obtained from X-ray diffraction tests to evaluate the presence of microstructures and their influence on the contractive-
dilative response of the tailings. A discussion of some limitations of conventional S-CPTu classification and empirical
correlation methods, when applied to this type of material, is presented in this paper, along with a framework for interpretation
and classification.

RÉSUMÉ
Cet article résume et discute les résultats du test de pénétration du cône sismique (S-CPTu) et du test de laboratoire des
résidus de fer qui ont été éliminés à différents moments, qui sont classés en anciens résidus et nouveaux résidus, cette
variation temporelle a généré des différences en termes de résistance, rigidité et microstructure non drainées. Ces résidus
sont discrétisés sur la base de l'indice de type de comportement du sol (Ic) pour l'examen du comportement contractif et
dilatatif et l'identification d'éventuelles zones de fragilité en utilisant différentes approches d'interprétation. Cette
caractérisation sera complétée par les informations de composition minéralogique obtenues à partir d'essais de diffraction
des rayons X pour évaluer la présence de microstructures et leur influence sur la réponse contractive-dilatative des résidus.
Une discussion de certaines limites de la classification S-CPTu conventionnelle et des méthodes de corrélation empirique,
lorsqu'elles sont appliquées à ce type de matériau, est présentée dans cet article, ainsi qu'un cadre d'interprétation et de
classification.

1 INTRODUCTION interpretation concerning laboratory testing data.

Discussion on the influence of microstructure on the
Tailings characterization for the geotechnical design of interpretation of results is given using the framework
tailings storage facilities (TSF) has become a topic of great proposed by Robertson (2016). Additionally, a statistical
interest in the mining industry and a challenging process analysis is performed to estimate the undrained shear
due to the variability of mineral composition, form of strength (peak) of the tailings using the correlations from
disposal, and storage time, which influences the variability the following authors with S-CPTU:
of the physical properties of the material and the resistance • Olson & Stark (2003), and
properties due to the rearrangement of the particles during • Sadrekarimi (2014).
storage.
In this context, it is common to develop an investigation 2 SITE DESCRIPTION
program focused on collecting data to characterize the in-
situ properties of the tailings, such as field tests and This field investigation campaign was executed in a local
laboratory tests. This paper presents the results of the mine within the “Iron Quadrangle” in the jungle of South
geotechnical investigation of an iron tailings storage facility America in Brazil, shown in Figure 1, where high rainfall is
(TSF), which consists of 60 Seismic Piezocone Penetration expected from December to March. The site is an iron ore
tests (S-CPTu) (with a total length of more than 1200 m), tailings deposit, the tailings were deposited through tubes
21 Standard Penetration Tests (SPT), and laboratory tests in paste, and the dam was constructed with four upstream
such as triaxial test, X-ray diffraction tests, etc. raisings and a buttress during the 70s.
The results from field test data were reviewed to The site presents average annual precipitation more
establish site-specific correlations, soil-type significant than 1600 mm distributed within the rainy and
characterization, behavior-type characterization, and
dry seasons at irregular intervals throughout the year. The characteristics expected to influence the stability of the
main characteristic of the site is described in Table 1. tailings stack.
The analysis approach is based on the idea that the soil
properties are based on its particle characteristics (e.g.,
particle size, particle shape, composition, etc.) and are,
therefore, independent of the in-situ state of the soil
(density, stress level, etc.). Soil behaviour (strength,
compressibility, etc.) is a function of the properties and
state. This approach is helpful because it avoids treating
each soil density as having its properties. For this reason,
the tailings were classified according to their particle size:
fine, intermediate, and coarse.
The characteristics of the tailings disposed to the TSF
vary depending on the ore body encountered, changes in
the mill operating procedures, and mill throughput rate
(Erickson et al., 2017). Tailings placement began at the
TSF in early 1971, and the stack has been progressively
developed. The original stack is defined as the placement
area deposited from 1970 to 2019 (reservoir of the Dam)
Figure 1. General site location (De Vicq et al., 2015) and two pits where the disposal tailings from 2012 to 2020.
Currently, the perimeter of the reservoir and the pits are
Tailings deposition occurred in three areas in different part of a single structure. The tailings placed in the original
years. Tailings were placed in the first area from 1971 to stack between 1971 and 2012 are referred to as “old
2019, while the second and third areas were from 2012 to tailings”, while the tailings placed since 2012 are referred
2020. This difference in when the tailings were placed has to as “new tailings”. There are differences in properties,
caused variations in the properties of the tailings, as shown density state, and behaviour of the old and new tailings, as
in Figure 2. described in this document.
As previously noted, the geotechnical characterization
of iron tailings is challenging. It requires a range of tests in
the field and laboratory to evaluate the behaviour and
performance of tailings and to obtain best-judgment design
parameters. The following framework was developed in
Figure 2 Cross section of disposed tailings this paper:
• Assess soil properties, such as gradation,
The field investigation campaign encountered moisture content, plasticity, specific gravity of soils, and
numerous challenges, such as vibration limits generated by variability in fine, intermediate, and coarse tailings.
the equipment, restrictions on the use of water during • Assess the behaviour/response in terms of
drilling, requirements for the use of unmanned equipment dilative or contractive response based on SCPTUs and
for the execution of the SCPTus, limited accessibility to the observe if there is evidence of sensitive or non-sensitive
areas of interest, size, and capacity of investigation soils to compare these results with the triaxial test.
equipment, and climatic effects. These challenges limited • Develop a statistical analysis according to
the number and type of investigations and the size of the empirical correlations available to estimate the peak
samples collected on the field. undrained resistance and compare them with the values
obtained from the undrained triaxial tests.
Table 1. Characteristics of the Site • Assess the response of strain-softening or strain-
hardening observed during undrained shear.
Characteristics Site • Assess the microstructure (chemical composition,
Average Elevation 1140 masl particle scale, aging, bonding (cementation).
Average Precipitation 1650 mm
4 FIELD INVESTIGATIONS
Tailings Height 44 m
Main Rock Type Phyllite – Itabirite The investigation methods included shallow and deep
Tailings Type Iron investigations in 22 hectares (perimeter of the reservoir
and the pits). The shallow investigations consisted of 20
3 EVALUATION OF FRAMEWORK test pits excavated with an excavator, collecting tailings
material from 2.0 m to 3.5 m deep. These disturbed
The geotechnical characterization of the iron tailings was samples were obtained for laboratory testing, such as
performed by compiling data from historical laboratory and index properties (grain size granulometry, moisture
field data on the tailings and a recent SCPTu campaign and content, plasticity, the specific gravity of soils, etc.) and
laboratory program designed to support this advanced tests, such as Consolidated Isotropic Undrained
characterization work. The combined database was then Triaxial Compression (TX-CIU) and Consolidated Isotropic
summarized and interpreted to assess the critical Drained Triaxial Compression (TX-CID).
 Deep investigations consisted of 21 drill holes (rotary reservoir (old tailings). In contrast, the superficial tailings in
drilling) with a diameter core of 66.5 mm, where only the the reservoir had a sander matrix (new tailings). The old
Standard Penetration Tests (SPT) were performed at tailings have slightly higher sand content than the new
intervals up to 1 m to a maximum depth of 36 m, registering tailings, on average, but the gradation is similar overall.
a total of 307 SPT tests. During the execution of the SPTs, Based on Atterberg limits, they are similar in having low
only the material that was in the SPT sampler scoop was plasticity and activity. According to the moisture content
collected; conventional samplers such as the Shelby or shown in Table 3 in both types of tailings, a clear trend was
Denison tube were not used. These rotary drill evidenced for the in-situ condition and type of samples
investigations were carried out only within the dam collected.
reservoir.
Additionally, 60 SCPTu were executed, drilling 1200 m CLAY SILT
into tailings, including 25 pore pressure dissipation tests
(performed every 3m) and 31 shear wave velocity tests by
hitting a steel beam with a hammer from the ground surface
(performed every 1m). Most of these tests were carried out
on tailings; however, upstream raising material (SCPTus
were pushed along the dam’s crest) and foundation
material were also recorded. A tip with an effective area of Legend:
Old Tailings
10 cm2 was used, which was pushed with a force of 16 tons
Old Tailings (Structural)
generated with a 1-tonne lightweight rig anchored to the
New Tailings
ground. These SCPTus were located inside the reservoir
and the pits. SAND GRAVEL
In addition, six drill holes were executed with Hollow
Stem Auger next to each SCPTu test, reaching depths of
30 m in the tailings. In these boreholes, a direct-push Figure 3. Particle size distribution of tailings – ABNT
sampler was used; this sampler was 1.15 m long and (Associação Brasileira de Normas Técnicas)
0.038 m in diameter. These drillings were carried out within
the dam reservoir, which collected deep tailings material. Table 3. Water content results (w%)
These samples helped to characterize the tailings that
were deposited since 1971 and investigate the difference 25th
between the properties of the old and new tailings, which Min. Max. Average Median
Tailings percentile
will be discussed in this paper. Table 2 summarizes all the % % % %
%
field investigations carried out on the site.
Old Tailings 11.9 26.6 16.3 14.9 13.2
Table 2. A field investigation was conducted at the site. New Tailings 8.9 25.0 18.9 19.5 15.8

Old New 5.2 Mineralogy

Investigation Type
Tailings Tailings
Shallow The mineral composition of the tailings has been assessed
Test pits - 20 using X-ray diffractometry on three samples from one test
investigation
pit (one shallow sample) and one borehole (two deep
Drill holes
21 - samples), as summarized in Table 4. The relative lack of
(SPT)
Deep clay minerals composed chemically of hydrated aluminum
investigation
S-CPTu holes 31 29 silicates may explain the zero plasticity and activity of
Hollow stem auger tailings due to kaolinite being presented in a low
6 - percentage. Sample 1 collected from the pit, considered
holes
new tailings, has a lower percentage of hematite and
goethite compared to the samples obtained from the
5 MATERIAL CHARACTERIZATION
reservoir at a greater depth, considered old tailings. These
minerals are found in almost double the percentages,
5.1 Index Properties
which directly influences the weight of these.
Index properties of the tailings are plotted in Figure 3 and
summarized in Table 3. The tailings primarily comprise silt-
size particles with some fine sand and trace clay. The
tailings collected in the pits present a significant variability
in particle size. They have a higher fines content than the
material collected from the reservoir. During the opening of
the samples collected with direct push, it was possible to
identify three types of tailings. The particle size analysis is
also shown in Figure 2, which shows fine, intermediate,
and coarse tailings. It was possible to identify that tailings
with a finer matrix were deposited at a greater depth in the
Table 4. Tailings Mineralogy This trend, shown with the SPT, was also observed with
the SCPTus executed in the old tailings. Figure 4 shows a
Percent by Weight (Average) slightly increasing trend of clean sand equivalent (Qtn-cs)
Mineral Chemical at depth.
Type Formula Sample 1 Sample 2 Sample 3
(1.25 m) (15.6 m) (21.6 m)
Hematite Fe2O3 34.5 80.0 61.1
Goethite FeO(OH) 13.7 7.7 12.7 Qtn,cs ≥ 0.70
Quartz SiO2 42.5 7.0 14.3
Kaolinite Al2Si2O5(OH)4 4.7 0.0 3.5
Gibbsite Al(OH)3 4.7 5.4 8.4

5.3 Specific Gravity

The specific gravity of the tailings is summarized in Table

5. The tailings from the reservoir show an average specific
gravity of 4.5, while the tailings disposed of in the pits show
values of 3.7. These values are higher than most natural
soils because they contain a significant percentage of
hematite with a minimum specific gravity of 5. The
secondary component is goethite, which has an average
specific gravity of 4.
This possible difference between the old and new
tailings may be because the tailings from the reservoir are
older than the tailings from the pits, and possibly during the
production of tailings, they were obtained with a higher
content of iron ore, which would cause their mass to be
more significant. This may be because the industrial
processes for mineral extraction historically had lower
efficiency than current ones. Figure 4. Clean and sand equivalent (Qtn-cs) - Old Tailings

Table 5. Summary of specific gravity for old and new 6.2 Seismic Piezocone Penetration Test (S-CPTu)
tailings
The data of all 60 SCPTu executed in the reservoir (old
tailings) and pits (new tailings) were plotted on the Qtn-Fr
Type of Tailings Old Tailings New Tailings soil behaviour type (SBT) chart from Robertson (2016).
Mean 4.53 3.72 Figure 5 and Figure 6 show the results for two
Minimum 4.14 3.54 representative SCPTu executed on new and old tailings,
respectively.
25 percentile
th
4.22 3.65
New Tailings
Maximum 4.77 3.97 1000

6 IN-SITU TEST RESULTS

6.1 Standard Penetration Test (SPT)

100

NSPT values were only measured in the old tailings at 1 m

intervals. These values were corrected to obtain the N160,
as shown in Table 6. It is no observed increase in the N160
values with depth. N160 shows values between the ranges
of 1 to 42 blows, with most values between 2 and 4, 10

indicating the loose state of the old tailings.

Table 6. Tailings drill hole analysis

30th 1
Methodology Min. Max. Avg. Med. 0.1 1 10
Perc.
Figure 5. CPT Soil Behavior Charts (SBT) for the New
NSPT 1.0 50.0 7.0 3.8 5.0 Tailings
N160 0 42 3.9 1.9 2.8
 Old Tailings During the review of the samples collected in the
1000 reservoir (old tailings), it was evidenced (visual
classification) the presence of three types of tailings,
classifying them as fine tailings, intermediate tailings, and
coarse tailings. Following this, the values estimated for
each unit were evaluated separately, as shown in Table 8
100 and Table 9. The old tailings with a sandier matrix have a
peak undrained strength value higher than the intermediate
and fine tailings. The fine tailings have lower peak
undrained strength values than the other types.

10 Table 8. Summary of undrained peak strength by type of

Old Tailings

Olson & Sadrekarimi Sadrekarimi

Old Tailings Stark Compress. Simple Sh.
1
0.1 1 10 (2003) (2014) (2014)
Figure 6. CPT Soil Behavior Charts (SBT) for the Old 10th percentile 0.22 0.23 0.20
Tailings 30th Percentile 0.24 0.24 0.21
Mean 0.25 0.25 0.22
The tailings chart plot on both sides of the
90th percentile 0.29 0.27 0.24
contractive/dilative boundary. However, 90% of all the
new tailings data are contractive, while only 78% of old
tailings are on the contractive side. This observation will Table 9. Summary of undrained peak strength by type of
be validated with the state parameter values in the Old Tailings
following items (as shown in
Figure 7 and Olson & Sadrekarimi Sadrekarimi
Figure 8). New Tailings Stark Compress. Simple Sh.
Table 6 shows that the predominant behaviour of both (2003) (2014) (2014)
tailings is clay-like contractive but in a higher percentage 10th percentile 0.21 0.22 0.19
for new tailings. Most of the dilative points are sand-like for
30th Percentile 0.21 0.22 0.19
old tailings, whereas, for new tailings, it is also sand-like. It
is essential to indicate that, on average, 9% of both tailing Mean 0.22 0.23 0.20
types are within the “clay-like-contractive-sensitive” zone, 90th percentile 0.23 0.24 0.21
indicating that the tailings are likely to be very sensitive.
6.4 State Parameter
Table 7. SBT data is presented in percentages.
The methodology used to obtain the state parameter (𝜓)
CC CCS CD SC SD TC TD from de S-CPTu is taking as a reference the frameworks
Tailings
% % % % % % % elaborated by Plewes (1992) and Robertson (2010). The
Old Tailings 36 10 3 16 16 16 3
results are plotted versus depth in
Figure 7 and
New Tailings 45 8 1 20 8 17 1
Figure 8, and they show that they are both generally
contractive and with some dilative layers near the surface
6.3 Undrained Strength of the tailings. The surficial dilative layer is generally less
thick in the reservoir (old tailings).
Undrained strengths (Su) were assessed using SCPTU.
The undrained strength was evaluated in the zones where
the lowest normalized tip resistance (Qt) values were
measured, and the behaviour was contractive. Different
approaches were evaluated to assess the peak undrained
strength ratio: Olson and Stark (2003) and Sadrekarimi
(2014). These values were conservative with each other,
obtaining values between the ranges 0.19 to 0.22. Other
approaches used were those of Lunne (1997), Robertson
(2012), and Thakur (2016), and much higher values were
obtained, ranging from 0.23 to 0.30. As is evident, there is
no fixed trend to estimate the peak undrained strength
ratio. However, as described in Section 6.5, the
microstructure influenced mainly by cementation can
generate very high undrained resistance and stiffness
values.
 most significant peak at 𝜓 = 0.0 and a second smaller peak
Robertson (2010) at 𝜓 =+0.2. These results are consistent with the
1140
interpretation of tip resistance because the less contractive
zones likely correspond to the surface of dry/compacted
1135
lifts, and the more contractive zones in-between are loose.
These plots considered more than 10 000 data collected in
1130 the reservoir (old tailings) and more than 8 000 collected in
the pits (new tailings).
Elevation (m)

1125
Table 10. Tailings State Parameter
1120 Dilatant Contractive
Old Tailings New Tailings
1115 Tailings Type Plewes Robertson Plewes Robertson
(1992) (2010) (1992) (2010)
1110
10th percentile -0.06 -0.05 -0.05 -0.03
Median -0.01 -0.02 0.03 -0.01
1105
80th percentile 0.09 0.02 0.15 0.03
1100
90th percentile 0.13 0.04 0.18 0.06
-0.5 -0.3 -0.1 0.1 0.3 0.5
State Parameter 6.5 Microstructure and ageing

Figure 7. State Parameter for the Old Tailings Robertson (2016) described a microstructure as the
features of a soil deposit at the particle scale that can lead
to its in-situ behaviour being different from what is
Robertson (2010) observed for reconstituted “ideal (unstructured) soils” in the
1140 laboratory. Microstructure tends to influence the small-
strain behaviour of soil more than the large-strain
1135 behaviour (Robertson et al., 2017). Based on this, the
shear wave velocities (Vs) from the SCPTu were used to
1130 evaluate the microstructure. Robertson (2016) defines a
normalized small-strain rigidity index (𝐾∗G) between 100
and 330 for young uncemented soils and greater than 330
Elevation (m)

1125
for soils with significant microstructure. The old and new
tailings 𝐾∗G data were plotted in
1120 Dilatant Contractive
Figure 9 and
Figure 10 to evaluate the microstructure. The results
1115
indicate that the old tailings have more microstructure than
the new tailings. Depth profiles of 𝐾∗G, as shown in the
1110 figures, indicate that the zones with the most
microstructure occur in the upper 10 m.
1105 The microstructure can be caused by secondary
compression, thixotropy, cementation, cold welding,
1100 suction hardening, and ageing (Robertson, 2016;
-0.5 -0.3 -0.1 0.1 0.3 0.5 Robertson et al., 2017). There is a possibility that the
State Parameter tailings generated with older industrial processes and
ageing (around 50 years ago) have an impact on the
microstructure. However, as shown in the figures below,
Figure 8. State Parameter for the New Tailings
since the age of the tailings increases with depth, but the
microstructure decreases, this suggests that ageing is
Additionally, these plots show that the most strongly
probably not the primary contributing factor to the observed
dilative zones (𝜓 < -0.15) occur in the upper 5 m of the
microstructure. As observed with the instrumentation
reservoir (old tailings) due to the material of the upstream
installed, the degree of saturation also likely increases with
raising, and some individual S-CPTs show slight trends of
depth. Since the water level is, on average, 10 m below the
increasing state parameter with depth. However, no overall
ground surface, the data suggests that the degree of
trends in the data are evident.
saturation is generally high (under 10 m). Suction
A summary of the results is shown in Table 10. The
pressures are expected to be less, so the influence of
distribution of old tailings is unimodal because only one
suction hardening may not be significant to generate
peak is observed for each approach used. For example, for
microstructure under 10 m. However, it maybe is a factor
the old tailings, the most prominent peak at 𝜓 = 0.0 for both
above this depth.
approaches. However, the distribution of new tailings is Usually, the microstructure is more likely to be caused
bimodal, using the approach developed by Plewes, with the
by cementation. This forms chemical precipitates on the
surface of the tailings during dry periods, which are washed 50 kPa to 400 kPa in the new tailings. The results indicated
away in wet weather. Remember that the mine is in a a peak drained strength of 33.5º, as presented in the stress
tropical zone, with heavy but short rains and days with path of
powerful sun, leading to the conditions described. Figure 11. Regarding the results of undrained peak
strength, the mean and 30th percentile results were
1000 analyzed, obtaining values of 0.27 and 0.24, respectively,
Old Tailings (0-5m)
as presented in
Old Tailings (5-10m) Figure 12. These values are consistent with the results
Old Tailings (10-15m) obtained in the statistics of the SCPTu for the new tailings.
Old Tailings (15-20m)
100
Old Tailings (20-25m)

Old Tailings (25-30m)

Qtn

Old Tailings (30-35m)

10

1
1 10 100 1000
I G = G 0 / qn

Figure 9. S-CPTu Microstructure Profiles for the Old

Tailings

1000
New Tailings (0-5m)
New Tailings (5-10m)
New Tailings (10-15m)
New Tailings (15-20m) Figure 11. Drained Strength for the new tailings
New Tailings (20-25m)
100 New Tailings (25-30m)
New Tailings (30-35m)
New Tailings (35-40m)
New Tailings (40-45m)
Qtn

New Tailings (45-50m)

10

1
1 10 100 1000
I G = G 0 / qn

Figure 10. S-CPTu Microstructure Profiles for the New

Tailings
Figure 12. Undrained Strength for the new tailings
Soils with microstructure tend to have higher yield
stress, peak strength, and small-strain stiffness than the The undrained resistances for the new tailings were
same ideal soil at the same density and confining pressure estimated from the confining stress and the peak p' and q
(Robertson, 2016). This may partly explain the generally values. Equation [1] shows the calculation used.
high undrained strengths inferred from the SCPTu data
(Section 6.3.2). Su/v = qpeak / c [1]

7 TRIAXIAL TESTS Triaxial tests were not executed on samples collected

in the region where the old tailings were identified. This is
The samples collected from the new tailings were obtained because access to the equipment necessary to collect the
through 20 surface test pits. The material was material is restricted by law, making it impossible to
homogenized and remoulded in the laboratory to obtain perform tests for comparison with the statistical analysis
more excellent compatibility with field conditions and thus presented.
be able to perform triaxial tests.
Thus, it was possible to perform conventional TX-CIU and
TX-CID triaxial tests with confining stresses ranging from
8 DISCUSSION Robertson, P.K. (2016). Cone penetration test (CPT)-
based soil behaviour type (SBT) classification system – an
A geotechnical characterization of the iron tailings was update. Canadian Geotechnical Journal. 53: 1910-1927.
performed by compiling and interpreting the historical data. Robertson, P.K. (2022). "Evaluation of Flow
The tailings have a high fines content of about 60% to 90%, Liquefaction and Liquefied Strength Using the Cone
a low clay content of around 10% or less, and are non- Penetration Test: An Update." Canadian Geotechnical
plastic. The in-situ condition of the tailings is contractive Journal. 59(4): 620–624.
behaviour in the majority, and the contractive zones are Robertson, P.K. & Campanella, R.G. (1983).
generally thicker in the old tailings. Interpretation of Cone Penetration Tests: Sands and Clays.
Typical state parameters of +0.02 in old tailings and Robertson, P.K., Viana da Fonseca, A., Ulrich, B., and
+0.03 in new tailings were interpreted from the SCPTu Coffin, J. 2017. “Characterization of unsaturated mine
data. A comparison between new and old tailings shows waste: a case history.” Canadian Geotechnical Journal. 54:
that the tailings have become more contractive over time 1752-1761.
due to increased confining stress as the tailings stack was Sadrekarimi, A. (2014). Effect of the Mode of Shear on
raised, and chemical weathering may also be involved. Static Liquefaction Analysis.
The tailings show significant microstructure in the field Thakur, Vikas & Fauskerud, Odd & Gjesvik, Vidar &
that decreases with depth and is believed to be primarily Christensen, Stein & Oset, Frode & Nordal, Steinar &
caused by cementation. Microstructure is typically Viklund, Margareta & Strand, Stein-Are. (2016). A
expected to increase the stiffness and peak strength of the procedure for assessing the undrained shear strength
tailings compared to reconstituted or disturbed samples profile of soft clays. 17th Nordic Geotechnical Meeting
tested in the laboratory. However, for the new tailings, the (NGM-2016).
lower bound (30th percentile) peak undrained strength ratio
from laboratory tests showed good agreement with the
SCPTu data and is unlikely to be significantly affected by
microstructure.
The influence of the sub-horizontal layering of finer and
coarser tailings was not identified but could be investigated
through laboratory testing, such as DSS testing.

9 REFERENCES

De Vicq Ferreira Da Costa, R., Leite, M. G. P.,

Mendonça, F. P. C., & Nalini, H. A. (2015). Geochemical
mapping of arsenic in surface waters and stream
sediments of the Quadrilátero Ferrífero, Brazil. REM.
Revista Escola De Minas; SciELO.
Erickson, B., Butikofer, D., Marsh, A., Friedel, R.,
Murray, L., & Piggott, M. J. (2017). Filtered tailings disposal
case history: Operation and design considerations, part I.
In Proceedings of Tailings and Mine Waste.
Lunne, T., Robertson, P.K. and Powell, J.J.M. (1997)
Cone Penetration Testing in Geotechnical Practice.
Blackie Academic & Professional, London, 312 p.
Olson, S. M., & Stark, T. D. (2003). “Yield strength ratio
and liquefaction analysis of slopes and embankments.” J.
Geotech. Geoenviron. Eng., 129(8), 727–737.
Plewes, H.D., Davies, M.P., & Jefferies, M.G. (1992).
CPT-based screening procedure for evaluating liquefaction
susceptibility. In Proceedings of the 45th Canadian
Geotechnical Conference, Toronto, Ont. 1992. BiTech
Publishers Ltd., Richmond, B.C. pp. 41–49.
Robertson, P.K. (2010). “Estimating In-Situ State
Parameter and Friction Angle in Sandy Soils from CPT,” in
Proceedings of the 2nd International Symposium on Cone
Penetration Testing (CPT’10), May 2010, Huntington
Beach, California.
Robertson, P.K. (2012). "The James K. Mitchell
Lecture: Interpretation of In-Situ Tests–Some Insights."
Geotechnical and Geophysical Site Characterization. 4:
pp. 3–24.
Laboratory investigations on geotechnical
characteristics of sensitive clay foundation:
a case study of a mine tailings stack
Ali Reza Zafarani, Mireille Sandrine Ewane, Amir Hossein Zamani,
Philip Gomes
Sustainable Mining Development Department, SNC-Lavalin,
Montréal, Québec, Canada
William Fresser, Jean-François Painchaud
Newmont Corporation, Éléonore Mine, James Bay, Québec, Canada

ABSTRACT
Eastern Canada has substantial deposits of sensitive clay in southern Québec and southeastern Ontario. Sensitive clays
have the potential to rapidly lose their shear strength if remolded, which makes them prone to potential landslides. These
landslides have the potential to retrogress under certain circumstances, with significant soil strength loss which results in
the soil behaving like a viscous liquid. To ensure the best possible performance of infrastructures built on this type of soil,
a precise assessment of the strength and compressibility response of the sensitive clay is necessary. Inaccurate estimates
may have a significant impact on the stability and settlement estimates of the site infrastructure prior to construction and
may result in corrective remedial work during construction, which adds to the project duration and cost. This paper aims to
present a detailed geotechnical characterization of a sensitive clay foundation which exists at the Éléonore Mine filtered
tailings storage facility (TSF), located in James Bay, Québec, Canada. Accordingly, laboratory test results from intact clay
samples were interpreted to quantify the geotechnical properties of the foundation clay. Obtained results from oedometer
tests, monotonic triaxial and simple shear tests, and cyclic direct simple shear tests were analyzed and compared to the
clay data already available in the Québec region. The findings in this study should be helpful to the engineering practice
to support and improve the assessment of sensitive clay behaviour.

RÉSUMÉ
L'est du Canada possède d'importants dépôts d'argile sensible dans le sud du Québec et le sud-est de l'Ontario. Les
argiles sensibles peuvent rapidement perdre leur résistance au cisaillement si elles sont remaniées, ce qui les rend sujettes
à des glissements de terrain potentiels. Ces glissements de terrain peuvent, dans certaines circonstances, subir une
diminution importante de la résistance du sol et se comporter comme un liquide visqueux. Pour garantir une meilleure
performance possible des infrastructures construites sur ce type de sol, une évaluation précise de la réponse en termes
de résistance et de compressibilité de l'argile sensible est nécessaire. Des estimations imprécises peuvent avoir un impact
significatif sur l’évaluation de la stabilité et du tassement de l'infrastructure avant construction et peuvent entraîner des
travaux correctifs pendant la construction, ce qui augmente la durée et le coût du projet. Cet article vise à présenter une
caractérisation géotechnique détaillée de la fondation d'argile sensible qui existe au parc à résidus filtrés de la mine
Éléonore, situé à la Baie-James, Québec, Canada. Par conséquent, les résultats d'essais au laboratoire réalisés sur des
échantillons d'argile intacte sont interprétés afin de mieux quantifier les propriétés géotechniques de l'argile de la fondation.
Les résultats obtenus à partir d'essais oedométriques, d'essais triaxiaux monotones et d'essais de cisaillement simple,
ainsi que d'essais de cisaillement direct cyclique ont été analysés et comparés aux données déjà disponibles pour l'argile
dans la région du Québec. Les résultats de cette étude devraient être utiles à la pratique de l'ingénierie pour soutenir et
améliorer l'évaluation du comportement des argiles sensibles.

1 INTRODUCTION source of this peculiar behaviour (Leroueil et al., 1983;

Konard, 1987; Mayne et al., 1998).
Large portions of Eastern Canada are covered by a mantle A significant formation of soft sensitive clay known as the
of sensitive clays, which have caused dramatic landslides Grande-Baleine clay lies in northwest Québec. These
resulting in loss of key infrastructure and loss of human life. complex marine sediments were formed approximately
The sensitivity of a clay (St) is defined as the ratio of the 7,000 years ago in the Tyrrell Sea, a cold brackish sea with
undrained shear strength of the intact clay (Su) to the a salinity of at least 8 g/l (Locat, 1982). According to
undrained shear strength of the same remolded clay (Sur), mineralogical studies, this soil contains a significant
both at the same water content (Skempton and Northey percentage of felsic minerals (55 to 77%) and 10% to 20%
1952; Rosenqvist 1953; Terzaghi et al. 1996; L'Heureux et of clay minerals. Geophysical and seismological data
al. 2014). indicate that these sensitive clays are susceptible to
The mechanical properties of this material have been seismic wave amplification due to the extremely high
investigated for a very long time to better characterize the impedance contrast between the bedrock and the porous
soils (Crow et al., 2011). However, there is little information
 Figure 1. The Éléonore TSF layout, 2022

available on the cyclic softening potential of these clays stability of the tailings stack is crucial, which is the primary
(Lee, 1979; Lefebvre and Pfendler, 1996). objective of this study.
This paper aims to present a detailed geotechnical
characterization of the Grande-Baleine sensitive clay
which exists at the foundation of the Éléonore Mine filtered
tailings storage facility (TSF). This was accomplished by
evaluating and analysis of a comprehensive geotechnical
investigation that included high-quality intact sampling and
laboratory testing. Accordingly, results of the oedometer
tests, monotonic triaxial and simple shear tests, and cyclic
direct simple shear tests (CDSS) were analyzed and
compared to the data already available for the clays in the
Québec region.

2 STUDY SITE INFORMATION

The Éléonore Mine, which is owned and operated by

Newmont since 2019, is in the James Bay region,
approximately 190 km east of the Cree community of
Wemindji and 350 km north of the town of Matagami. Since
Figure 2. Plan view of the Phases 3B and 4A of the
the beginning of mine operations in 2014, the tailings are
Éléonore Mine TSF with clay layer thickness
hauled, placed, and compacted in the tailings stack whose
gradual expansion is planned in different phases as shown
3 LABORATORY TESTING PROGRAM
in Figure 1. Phases 1, 2, and 3A are already completed,
and they are configured with a starter dam around the
The characteristics of the foundation clay have been
tailings stack perimeter. The TSF also has a waste rock pile
determined based on various geotechnical investigations
and a water collection pond. The same tailings pile and
conducted in 2008, 2010, 2011, 2019, and 2022. However,
starter dam layout is considered for Phase 3B and Phase
for the sake of brevity, only the results from the laboratory
4A.
tests conducted in 2022 will be discussed in this paper.
The foundation soil layer under the TSF consists of a
The sampling during the geotechnical investigation was
clay crust (silty clay), silty clay, till, and bedrock. Figure 2
done with a thin-walled pipe known as a "Shelby" tube
displays the thickness of the clay deposit as determined by
(ASTM D1587). The laboratory testing program consists of
the field investigation data. Borehole data shows the
the following tests: index testing, oedometer test,
presence of a clay layer greater than 5 m thick in Phases
monotonic triaxial testing and simple shear testing, and
3B and 4A, with a sensitivity ranging from sensitive (St=4-
CDSS tests. The results of these tests are discussed in the
8) to ultra-sensitive (St>64). Thus, the evaluation of the
following sections.
characteristics of the foundation clay layer in the Phases
3B and 4A and its effect on the overall static and seismic
3.1 Index tests

Figure 3 shows a representative stratigraphy with main

index properties of the tailings stack foundation, including
natural water content (Wn), liquid limit (WL), and plastic
limit (Wp). Overall, the mean index values in the clay layers
are as follows: Wn = 45%, WL = 37%, and plasticity index,
PI = 14%. Based on the analysis of particle size, the
percentage of clay in the samples ranged from 12 to 72%.
Using the Unified Soils Classification System (USCS), as
defined in ASTM D 2487, which is shown in Figure 4, the
foundation clay at the Éléonore Mine site is classified as a
low plasticity clay (CL) and/or silty clay (ML) or a sensitive
to ultra-sensitive clayey silt (CL-ML).
The undrained shear strength obtained using the fall
cone ranges from 25 to 43 kPa for the foundation clay. The
Figure 4. Classification of the Éléonore mine foundation
sensitivity measured by fall cone test is higher than 500 for
clay in the Casagrande plasticity diagram
some samples which is considered ultra-sensitive clay
(Leroueil et al. 1983; Terzaghi et al. 1996). Based on the
Figure 5 shows this relationship using representative clay
liquidity indices and remolded shear strength (< 1 kPa), this
samples from eastern Canada for which the strength has
type of clay can be described as a "Quick Clay."
been determined using the Nilcon vane test. According to
the proposed relationship, the shear strength (Su) of the
tested clay samples (presented in Table 1) is about 50 kPa
based on the determined pre-consolidation stress values
and plasticity index (PI).
This range of shear strength is relatively higher than the
shear strengths obtained by the fall cone test which could
be due to the sample quality.

Table 1. Oedometer test results on the clay foundation

Test Test 𝜎𝑝′ ′

𝜎𝑣0 OCR Cr Cc
ID. kPa kPa ′
𝜎𝑣0 <360 kPa

Oed-01 Vs 230 80 2.9 0.021 0.99

Oed-02 Vs 220 80 2.8 0.028 0.85

Oed-03 Cv 225 100 2.25 0.020 1.95

3.2.1 Shear wave velocity measurement

The variation of shear wave velocity (Vs) was determined

Figure 3. Conceptual subsurface profile of the Phases 3B for two of the samples using the P-RAT technique
and 4A of the Éléonore Mine TSF with plasticity indices developed in the geotechnical laboratory at Sherbrooke
University (Karray et al., 2015). The approach can be
3.2 Oedometer tests integrated into traditional geotechnical apparatus such as
triaxial and oedometer cells. This technique has been
Three Oedometer tests were performed as part of the 2022 included into an oedometer equipment to measure shear
testing program. Two of these tests were carried out with wave velocity during a consolidation test. The Vs values
the use of a device that measures the shear wave velocity can be used to develop a relationship between the
during the test (see section 3.2.1). The third test was normalized shear wave velocity (VS1) and the void ratio of
conducted using a standard method that measured the tested samples. The following equation can be used to
consolidation as a function of time for each loading level. estimate the value of the VS1 (Youd et al. 2001):
Table 1 presents the results of the Oedometer tests.
The yield stress (𝜎′𝑝 ) values obtained between 220 and 𝑉𝑠1⁄ 𝑃𝑎
𝑂𝐶𝑅 𝛼 = 𝑉𝑠 ( ⁄𝜎 ′ 𝑣 )
0.25
[1]
230 kPa confirm that the clay layer is an over-consolidated
clay with an over-consolidation ratio (OCR) in the range of
Where 𝑉𝑠1 is the normalized shear wave velocity at
2.25 to 2.9. Leroueil et al. (1983) proposed a correlation
effective stress 100 kPa, 𝑃𝑎 is the atmospheric pressure
between the normalized measured shear strength ratio
𝑆 (=101.3 kPa), 𝜎 ′ 𝑣 is the effective vertical stress (kPa), OCR
( 𝑢⁄𝜎 ′ ) and plasticity index for the Québec clays. is the over-consolidation ratio, and power 𝛼 generally
𝑝
varies between 0 and 0.25 with the variation of the plasticity stress-strain and porewater pressure versus strain
index (Hardin and Black, 1968; Hardin and Drnevich, measurements on a clay specimen at 100 kPa confining
1972). pressure is shown in Figure 7. The results show that the
peak shear stress develops early at around 2% strain,
while excess pore pressure keeps increasing, reaching a
maximum value of about 83 kPa later at about 20% strain.
Similar stress-strain patterns for sensitive clays have been
reported in the literature (Demers et al., 2014).
The effective stress path followed during shearing is
shown in Figure 8. It should be noted that all the samples
were consolidated under isotropic stress conditions with no
shear load and zero shear strain. The results indicate that
peak shear stress levels increase as confining pressures
increase. Indeed, higher levels of stress may cause
significant remolding in the soil samples. Based on the
stress patterns shown in Figure 8, the shear strength ratio
𝑆
( 𝑢⁄ ) ranges from 0.24 to 0.58 with an average value of
𝜎′𝑣
S
Figure 5. Variation of u⁄σ′ with plasticity index for eastern 0.29, which is typically lower than the reported values for
p the sensitive clays (i.e., 0.30 to 0.33) (Leroueil et al., 1983;
Canadian clays (Leroueil al., 1983) Konard, 1987; Lefebvre and Pfendler, 1996). A number of
extension triaxial tests have also been performed on the
Figure 6 shows the relationships between normalized samples in order to assess the stress-strain behaviour of
shear wave velocity (Eq. 1) as a function of void ratio for the clay material. However, the results are not presented
the two samples presented in Table 1. The results show here for the sake of brevity. According to the compression
𝑉
that the 𝑠1⁄𝑂𝐶𝑅𝛼 varies between 95 to 130 m/s for the and extension triaxial results and considering the effect of
𝑆
range of considered void ratios (i.e., 0.7 to 1.7). The slight anisotropy, the mean value of 𝑢⁄ ratio is around 0.2.
𝜎′𝑣
difference in shear wave velocity curves between two Based on triaxial results, the cohesion of the sensitive clay
samples could be due to sample quality, particle size material of the foundation varies between 9 to 14 kPa and
distribution, and fines percentage (Karray et al. 2011; Choo the effective friction angle is between 28˚ to 32˚.
and Burns 2015). 200
Shear Stress
175 Excess Pore Pressure
Peak Shear strength, qmax=119.5 kPa
Shear Stress and Pressure (kPa)

150 at Ԑ=2%

125
max= 83.43 kPa

100
= 55.5 kPa
at qmax
75

50
𝜎 ′=103.6 kPa
25 Z = 8.5 - 9.1 m

0
0 5 10 15 20 25
Axial strain, Ԑa (% )

Figure 7. A representative monotonic triaxial test results in

terms of stress-strain and porewater pressure versus strain
measurements on a clay specimen at 100 kPa confining
Figure 6. Variation of the normalized shear wave velocity pressure.
as a function of void ratio
3.4 Monotonic simple shear tests
3.3 Monotonic triaxial tests
Simple shear tests are generally used to estimate shear
A series of triaxial compression tests were performed on strengths in a horizontal plane. The shear strengths
undisturbed samples at confining pressures ranging from obtained with this type of test are generally comparable to
𝜎′𝑣 = 100 kPa to 𝜎′𝑣 = 600 kPa. A representative set of those obtained with a vane shear test. The simple shear
 250
σ'v=600 kPa
σ'v=310 kPa
σ'v=105 kPa
Shear Stress, q/2 = 0.5 (σ'1-σ'3) (kPa) 200 σ'3=0

=146.10 kPa
150

100
=79.95 kPa

=59.75 kPa
50

0
0 50 100 150 200 250 300 350 400 450 500 550 600 650
Effective Stress, p' = 0.5 (σ'1+σ'3) (kPa)

Figure 8. Stress path of compression triaxial tests on clay specimens of normally-consolidated

tests were performed using the constant volume method in

this study. The test consists in first consolidating the
sample to the desired vertical stress under laterally
restrained conditions. Following consolidation, a
progressive deformation with a constant loading rate was
applied to the desired shear deformation.
Results of simple shear testing of representative
samples are summarized in Table 2. These results indicate
that shear strengths are in the range of 40 to 47 kPa for
over-consolidated samples and 60 to 150 kPa for tests
performed in the normally consolidated range.

Table 2. Results of the monotonic simple shear tests

Test e σ′v Su -Peak u εa

ID. consolidated kPa kPa kPa (%) Figure 9. Variation of the normalized shear strength with
MSS-01 1.375 150 40.5 47.2 2.0
respect to the pre-consolidation stress (𝜎′𝑝 ) or the applied
vertical stress (𝜎′𝑣 ) versus shear strain
MSS-02 1.547 200 46.8 64.5 3.3
MSS-03 1.460 294 73.0 105.1 4.8 3.5 Cyclic direct simple shear tests (CDSS)
MSS-04 1.250 585 150.0 178.1 3.5
MSS-05 1.237 300 60.4 103.0 3.3 A series of CDSS tests were performed on a couple of
tested samples. The testing was conducted using an
automated CDSS device with a specimen size of 70 mm in
Figure 9 shows the variation of the normalized shear diameter and 20–25 mm in height. Following sample
strength with respect to the pre-consolidation stress (𝜎′𝑝 ) preparation, the specimens were consolidated at effective
or the applied vertical stress (𝜎′𝑣 ) versus shear strain for a vertical stresses ranging from 50 to 150 kPa before cyclic
couple of tested samples. According to the results, the testing.
𝑆 𝑆 The samples were subjected to constant-volume
average peak value of the 𝑢⁄ or 𝑢⁄ ratio is around
𝜎′𝑣 𝜎′𝑝
loading following the consolidation phase. Throughout the
0.20 which is in accordance with the triaxial compression testing process, the total vertical stress, σ’v, and shear
and extension results. strain, 𝛾, were continually monitored by instruments
connected to the device. The samples were subjected to
constant cyclic stress ratio (CSR) values of 0.20, 0.27, and
0.32 with the application of unidirectional, symmetrical,
sinusoidal shear stress at a frequency of 1 Hz. The CSR is
the ratio of the applied shear stress, 𝜏, to the initial vertical
effective overburden stress, σ’v0. The cyclic loading phase
of all samples was continued until shear strain reached
±3.5%, at which point testing was stopped. According to
Castro (1977), National Research Council (NRC) (1985),
Kramer (1996), and Youd and Idriss (2001), a shear strain
value of this magnitude (i.e., 3.5%) was taken to be the
point at which liquefaction/cyclic softening is initiated.
Figure 10 shows typical CDSS testing results for a
representative sample, which was consolidated to 𝜎′𝑣 =
151 kPa and subjected to a cyclic shear load with an
amplitude of 30.2 kPa (CSR = 0.20). Figures 10a-10e are
detailed below:
• Figure 10a: This plot illustrates the CSR applied to the
sample during cyclic testing. The amplitudes of the
loading are relatively uniform during the CDSS
testing.
• Figure 10b: The amplitude of the shear strain, 𝛾,
developed gradually until the 67th cycle of loading,
when it reached 3.5% shear strain (i.e., initiation of
liquefaction/cyclic softening)
• Figure 10c: The excess pore-water pressure, u, as
indicated by the excess pore-water pressure ratio, ru
( 𝑢⁄ ), built up with each cycle of loading,
𝜎′𝑣0
increasing cumulatively until it reached its maximum
value of 0.7.
• Figure 10d: The graph of 𝜏 versus 𝜎′𝑣 displays the
typical stress path for clays under cyclic loading. A
decrease in effective stress is seen as a result of
excessive pore water pressure development. The
cyclic stress path eventually shifted leftward as it
approached the origin at the liquefaction triggering.
• Figure 10e: The total strain softening with each cycle
of loading is seen in the 𝜏 vs 𝛾 plot.
The responses of the other specimens were comparable to
those presented previously, with the exception that the
number of cycles required to reach a strain of 3.5% varied
inversely with the magnitude of the applied CSR.
Table 3 presents a summary of the CDSS test results
using a single amplitude shear strain of 3.5% as the
starting point for cyclic softening. According to the pore-
water pressure response, 𝛾 = 3.5% developed in each
sample before ru = 1.0.
Figure 11 shows the variation of the normalized cyclic Figure 10. Representative CDSS test results (𝜎′𝑣 = 151
shear resistance (𝜏 𝑦 ) with respect to the static shear kPa, CSR=0.20). (a) CSR versus N; (b) 𝛾 versus N; (c) ru
resistance (Su) as a function of the number of cycles (N). versus N; (d) 𝜏 versus 𝜎′𝑣; (e) 𝜏 versus 𝛾
Comparing the obtained shear resistance ratio to the limits
given by Boulanger and Idriss (2007) for non-sensitive Table 3. Summary of the cyclic simple shear (CDSS)
clays indicates that the cyclic resistance of the tested clay testing
samples is close to the upper border of the limit defined by
Boulanger and Idriss (2007). Test e e σ′v CSR Nliq
ID. Initial Consolidated kPa
CDSS-01 1.550 1.410 145 0.32 5
CDSS-02 1.498 1.390 146 0.27 34
CDSS-03 1.495 1.350 151 0.20 67
 capture the behaviour of sensitive eastern Canadian clays
(Abdellaziz et al., 2020; 2022).
1.00
(a)
0.90

0.80

Shear modulus ratio, G/Gmax

0.70

0.60

0.50

0.40
PI=0%
0.30
Tested sample PI=50%
0.20
Darendeli (2001), PI=0-50 %
0.10 Vucetic and Dobry (1991), PI=0-50 %
Sigmoidal model (Sig4)
0.00

Figure 11. Variation of the normalized cyclic shear 30

(b)
resistance ( cyc ) with respect to the static shear (Su) 25
resistance as a function of the number of cycles (N)
20

Three (3) samples were subjected to post-cyclic testing,

Damping, ξ (% )
PI=0%

consisting of monotonic, undrained shear tests with a 15

maximum shear strain of 28% and a shear strain rate 10

ranging from 0 to 10% per hour. The test results are
summarized in Table 4. The average liquefied strength 5
PI=50%
ratio, Su-PC/𝜎′𝑣0 , was 0.16, and the post-cyclic undrained
shear strengths, Su-PC, ranged from 13.9 to 30.4 kPa. 0
0.0001 0.001 0.01 0.1 1 10
Shear Strain, (% )
Figure 12. Comparison of (a) shear modulus degradation
curve and (b) damping curve of a sample consolidated at
Table 4. Summary of the post cyclic testing results effective stress of 300 kPa with the literature models

Test σ′v Su-PC γpc Su-PC/σ′v0 4 CONCLUSION

ID. kPa (kPa) (%)
PCDSS-01 151 30.4 12.9 0.20 In this study, a detailed geotechnical characterization of the
Grande-Baleine sensitive clay which exists at the
PCDSS-02 146 30.4 11.6 0.20 foundation of the Éléonore Mine filtered tailings stack is
PCDSS-03 145 13.9 15.2 0.09 presented. Laboratory test results from intact clay samples
were interpreted to quantify the geotechnical properties of
3.6 Shear-modulus degradation and damping the foundation clay. Obtained results from the oedometer
tests, monotonic triaxial and simple shear tests, and cyclic
In order to determine the degradation (G/Gmax) and direct simple shear tests were analyzed and compared to
the data already available for the clay in the Québec region.
damping curves, the Hardin type resonance column (RC)
with a semi-fixed configuration modified at the These geotechnical characterization studies show that the
Geotechnical Laboratory of the University of Sherbrooke clay foundation underlying the tailings stack is an over-
(Karray et al., 2022) was used in this study for consolidated clay with OCRs typically more than 2 and has
a high degree of sensitivity in a significant portion of the
deformations of less than 0.1%. The shear moduli and
deposit.
damping at large deformations (𝛾>0.1%) were established
According to the results of the monotonic triaxial and
using CDSS tests (described in section 3.5). Figure 12
simple shear testing, the clay foundation has a low to
presents the modulus reduction and damping curves of a
medium undrained shear strength that is typically above
sample which was consolidated at a confining stress of
30 kPa. An average value of the undrained shear strength
300 kPa. In comparison to available literature models such
𝑆
as Darendeli (2001), Vucetic and Dobry (1991), and the ratio ( 𝑢⁄ ) of 0.20 was determined based on the triaxial
𝜎′𝑣
sigmoidal-4 (Sig4) model (Itasca 2017), investigated compression, extension, and simple shear test results.
sensitive clay samples seemed to have different G/Gmax The dynamic response of the clays was found to be
and damping curves. The comparison of the damping similar to that of sensitive clays with respect to the shear
curve shows that (Figure 12b), for small shear strains, the modulus reduction and cyclic softening resistance. The
tested sensitive clay samples are almost outside the range post-cyclic response of the clay samples showed that the
defined by Darendeli (2001) and Vucetic and Dobry (1991); liquefied strength ratio, Su-PC/𝜎′𝑣0 , was 0.16, and the post-
however, it is similar to the range defined by the sigmoidal- cyclic undrained shear strengths, Su-PC, ranged from 13.9
4 model (Itasca, 2017). For large deformations, it appears to 30.4 kPa.
that the Vucetic and Dobry (1991) model is comparable to The results of the presented laboratory program will be
the experimental data. These findings support earlier used for the development of the tailings deposition plan,
observations that the models for non-sensitive clays that calibration of the instrumentation in the foundation, and
are currently available in the literature do not accurately
ongoing static and dynamic stability analysis (i.e., design) shear wave velocity and cone tip resistance. Canadian
of the Éléonore Mine TSF. The findings of this study should Geotechnical Journal, 48(4), 599-615.
be helpful for the engineering practice to support and Karray, M., Ben Romdhan, M., Hussien, M.N., and Éthier,
improve the assessment of sensitive clay behaviour, Y. 2015. Measuring shear wave velocity of granular
particularly in the northwestern Québec region. material using the piezoelectric ring-actuator technique
(P-RAT). Can. Geotech. J. 52(9): 1302–1317.
5 REFERENCES doi:10.1139/cgj-2014 030
Karray, M., Lashin, I., & Abdellaziz, M. 2022. Discussion of
Abdellaziz, M., Karray, M., Chekired, M., Delisle, M.-C., Small strain shear modulus and damping ratio of two
Locat, P., Ledoux, C., and Mompin, R. 2020. Shear unsaturated lateritic sandy clays. Canadian
modulus and damping ratio of sensitive eastern Geotechnical Journal, 59(12), 2202-2204.
Canada clays. Can. Geotech. J. 58(8): 1118–1134. Konrad, J. M., & Roy, M. 1987. Bearing capacity of friction
doi:10.1139/cgj-2020-0254. piles in marine clay. Geotechnique, 37(2), 163-175.
Abdellaziz, M., Karray, M., Chekired, M., Delisle, M. C., Kramer, S.L. 1996. Geotechnical earthquake engineering.
Locat, P., Ledoux, C., & Mompin, R. 2022. New Model Prentice Hall Inc., Upper Saddle River, N.J
of Shear Modulus Degradation and Damping Ratio L’Heureux, J.S., Locat, A., Leroueil, S., Demers, D., and
curves for Sensitive Canadian Clays. Canadian Locat, J. 2014. Landslides in sensitive clays-from
Geotechnical Journal. geosciences to risk management. In
American Society for Testing and Materials (Philadelphia, Landslides in sensitive clays. Springer, Dordrecht. pp.
Pennsylvania). 2017. ASTM D2487-17e1: Standard 1–12.
Practice for Classification of Soils for Engineering Lee, K. L. 1979. Cyclic strength of a sensitive clay of
Purposes (unified Soil Classification System). ASTM. eastern Canada. Canadian Geotechnical Journal,
ASTM D 1587. Standard Practice for Thin-walled Tube 16(1), 163 - 176.
Sampling of Fine-grained Soils for Geotechnical Lefebvre, G., LeBoeuf, D. and Demers, B. 1989. Stability
Purposes. threshold for cyclic loading of saturated clay. Canadian
Boulanger, R.W., Idriss, I.M. 2007. Evaluation of Cyclic Geotechnical Journal, 26(1), 122-131.
Softening in Silts and Clays, Journal of Geotechnical Leroueil, S., and Vaughan, P.R. 1990. The general and
and Geoenvironmental Engineering, Volume 133 Issue congruent effects of structure in natural soils and weak
6 rocks. Géotechnique,40(3):467–488.
Castro, G., & Poulos, S. J. 1977. Factors affecting doi:10.1680/geot.1990.40.3.467.
liquefaction and cyclic mobility. Journal of the Leroueil, Serge, Tavenas, F., and Bihan, J.-P.L. 1983.
Geotechnical Engineering Division, 103(6), 501-516. Propriétés caractéristiques des argiles de l’est du
Choo, H., & Burns, S. E. 2015. Shear wave velocity of Canada. Canadian Geotechnical Journal 20(4): 681
granular mixtures of silica particles as a function of finer 705. doi:10.1139/t83-076.
fraction, size ratios and void ratios. Granular Matter, 17, Locat, J. 1982. Contribution à l’étude de l'origine de la
567-578. structuration des argiles sensibles de l'Est du Canada.
Crow, H.L., J.A. Hunter, D. Motazedian, A.J.-M. Pugin. Ph.D. thesis, University of Sherbrooke, Québec,
2011. In situ low-strain damping measurements using Canada.
a mono-frequency spectral ratio approach. Soil Mayne PW, Robertson PK, Lunne T. 1998. Clay stress
Dynamics & Earthquake Eng. J., 31(12), 1669–1677. history evaluated from seismic piezocone tests.
Darendeli, M. B. 2001. Development of a new family of Geotech Site Charact (Proc. ISC-1, Atlanta), Balkema,
normalized modulus reduction and material damping Rotterdam, 2:1113–1118.
curves. PhD dissertation, University of Texas at Austin, National Research Council (NRC). 1985. Liquefaction of
Austin, Tex. soils during earthquakes. National Academy Press,
Demers, D., Robitaille, D., Locat, P., & Potvin, J. 2014. Washington, D.C. Report.
Inventory of large landslides in sensitive clay in the Rosenqvist, I.T. 1953. Considerations on the sensitivity of
province of Québec, Canada: preliminary Norwegian quick clays. Géotechnique, 3(5): 195–200.
analysis. Landslides in sensitive clays: from doi:10.1680/geot.1953.3.5.195
geosciences to risk management, 77-89. Skempton, A.W., and Northey, R.D. 1952. The sensitivity
Dobry, R., and Vucetic, M. 1987. Dynamic properties and of clays. Géotechnique, 3(1): 30–53. doi:10.1680/geot.
seismic response of soft clay deposits. Department of 1952.3.1.30.
Civil Engineering, Rensselaer Polytechnic Institute. Terzaghi, K., Peck, R.B., and Mesri, G. 1996. Soil
577. mechanics in engineering practice. John Wiley & Sons.
Hardin, B. O., & Black, W. L. 1968. Vibration modulus of Youd, T. L., & Idriss, I. M. 2001. Liquefaction resistance of
normally consolidated clay. Journal of the Soil soils: summary report from the 1996 NCEER and 1998
Mechanics and Foundations Division, 94(2), 353-369. NCEER/NSF workshops on evaluation of liquefaction
Hardin, B. O., & Drnevich, V. P. 1972. Shear modulus and resistance of soils. Journal of Geotechnical and
damping in soils: design equations and curves. Journal Geoenvironmental Engineering, 127(4), 297-313.
of the Soil mechanics and Foundations Division, 98(7),
667-692.
Karray, M., Lefebvre, G., Ethier, Y., & Bigras, A. 2011.
Influence of particle size on the correlation between
Tuesday, October 3, 2023

PIPELINE
GEOTECHNICS
Investigation of small-diameter steel pipes
subjected to axial ground movement
Author Darren Andersen and Ashutosh Sutra Dhar
Department of Civil Engineering – Memorial University of
Newfoundland, St.John’s, NL, Canada

ABSTRACT
Small-diameter steel pipes are often used for distribution of gas to
residential homes. These pipelines sometimes run through landslide-prone areas, highlighting a need to better understand
their response to ground movements. Many studies were conducted in the past by pulling a pipe through a static soil mass
to investigate the behaviour of pipe subjected to axial ground movements. However, in real landslides, a soil mass moves
when the pipe can be restrained in the stable ground. It is believed that the effect would be similar to that of pulling a pipe
through static soil. A test facility was designed at Memorial University, where a soil mass can move against a pipe
restrained on one end, simulating the condition expected during real ground movement. This paper presents the results of
a test program completed using the test facility to investigate steel pipes subjected to axial ground movements. Steel pipes
of different diameters (26.7 and 60.3 mm) were buried in sandy soil compacted to 96% relative compaction inside a
moveable steel tank. Pipes were anchored at one end and free at the other. The tank was pulled at 0.5 mm/min until
reaching 35 mm displacement. Varying relaxation times were applied to investigate stress relaxation during ground
movements. Pipe axial force, free-end displacement, strain and tank displacement were recorded. Results showed that
axial force can be higher during soil pulling than pipe pulling. No reduction of axial force was observed during relaxation.
The axial force was found to reduce with repeated cycles of ground movements.

RÉSUMÉ
Les tuyaux en acier de petit diamètre sont souvent utilisés pour la distribution de gaz aux résidences. Ces pipelines
traversent parfois des zones sujettes aux glissements de terrain, soulignant la nécessité de mieux comprendre leur
réponse aux mouvements du sol. De nombreuses études ont été menées dans le passé en tirant un tuyau à travers une
masse de sol statique pour étudier le comportement du tuyau soumis à des mouvements axiaux du sol. Cependant, dans
les glissements de terrain réels, une masse de sol se déplace lorsque le tuyau peut être retenu dans le sol stable. On
pense que l'effet serait similaire à celui de tirer un tuyau à travers un sol statique. Une installation d'essai a été conçue à
l'Université Memorial, où une masse de sol peut se déplacer contre un tuyau retenu à une extrémité, simulant la condition
attendue lors d'un mouvement réel du sol. Cet article présente les résultats d'un programme d'essai réalisé à l'aide de
l'installation d'essai pour étudier les tuyaux en acier soumis à des mouvements axiaux du sol. Des tuyaux en acier de
différents diamètres (26,7 et 60,3 mm) ont été enfouis dans un sol sableux compacté à 96 % de compactage relatif à
l'intérieur d'un réservoir en acier mobile. Les tuyaux étaient ancrés à une extrémité et libres à l'autre. Le réservoir a été
tiré à 0,5 mm/min jusqu'à atteindre un déplacement de 35 mm. Des temps de relaxation variables ont été appliqués pour
étudier la relaxation des contraintes lors des mouvements du sol. La force axiale du tuyau, le déplacement de l'extrémité
libre, la déformation et le déplacement du réservoir ont été enregistrés. Les résultats ont montré que la force axiale peut
être plus élevée lors de l'extraction du sol que lors de l'extraction du tuyau. Aucune réduction de la force axiale n'a été
observée pendant la relaxation. On a constaté que la force axiale diminuait avec des cycles répétés de mouvements du
sol.

1 INTRODUCTION available on axially loaded buried steel pipes.

Murugathasan et al. (2021), Al-Khazaali and Vanapalli
Buried pipelines are essential to urban infrastructure, (2019), and Wijewickreme et al. (2009) have investigated
transporting oil, gas and water to residential and axial pipe loading by pulling a pipe through static soil
commercial properties. Typically the distribution lines masses and collecting the responses. However, during a
connecting to the final destination, sometimes termed the ground-movement event, unstable soil will move around a
“last mile,” are the smallest in the pipe network. As many pipe anchored by stable ground, causing stresses and
properties are developed within landslide and earthquake- strains in the pipe. It is believed that pulling a pipe through
prone areas, these pipelines must be placed in these zones static soil is comparable to the ground movement event.
and designed to withstand such conditions. For these Some studies investigate the effects of moving soil on a
reasons, it is important to understand the pipe-soil static pipe; Oskouei et al. (2019) and Jalali et al. (2016) do
behaviour during ground movement events. so from a strike-slip fault angle focusing on soil bearing and
During a ground-movement event, the pipeline may be pipe bending; both conducted full-scale testing on varying
positioned parallel to (axial movement), perpendicular to diameters of steel pipe in sandy soil. Oskuouei et al. (2019)
(lateral movement) or at some angle to (oblique movement) simulated strike-slip faulting by burying pipes of 19, 63.4,
the ground movement, which will affect the pipe and soil 88.9 and 168 mm diameters in two large steel boxes, fixing
behaviour and response. Many studies are currently one and moving the other laterally left and right. Jalali et al.
(2016) tested 114.3 and 168.3 mm diameter pipes buried frictionless system movement. A 4 x 2 x 1.5 m steel soil
in an 8.5 m testing box. Hydraulic actuators moved half of tank is then welded to the top of the carriage. Support arms
the soil box vertically at a 45-degree while maintaining the are welded to the tank, see Figure 2a, which connects it to
other half fixed. Pipe strain and movement were measured the shaft of a hydraulic ram capable of moving the tank up
in both cases. Others, such as Robert et al. (2016), have to 150 mm.
applied soil displacement to lateral-pipe orientation. A The carriage has a travel limit of 300 mm. A 177 mm
114.3 mm diameter steel pipe was buried using different hole is cut into both tank end walls to enable pipe entry.
sandy soils in a testing cell capable of movement. The pipe Two gasket plates, one steel and one rubber, are bolted
is connected to a load cell, fixed outside the tank, via steel over each hole. Each set of gaskets has a centre hole
rope. When the tank is moved, soil bears on the pipe slightly larger than the pipe diameter being tested to allow
causing lateral bending. Axial pullout responses were not each pipe diameter to be placed, avoiding friction with the
investigated in either study. test cell wall during movement. The pipe is then pushed
This paper presents the results from a novel testing into the tank from the ram end and rests on a bed prepared
facility created at Memorial University of Newfoundland, to the desired condition. To fully restrain one end of the
where a soil mass can be moved at a fixed rate along a pipe, a large steel channel is welded vertically to the
pipe fully restrained at one end and unrestrained at the frame’s backstop with holes drilled in a rosette pattern
other. The axial force transferred from the soil to the pipe securing a 22 kN (5,000 lb) or 88 kN (20,000 lb) capacity
via the pipe-soil interaction was investigated using pipe pancake load cell as needed. A custom-built connector is
elongation, soil-tank displacement, and the strain along the bolted to the load cell, with the pipe pinned to the opposing
pipe’s length. Multiple pipe diameters are considered in end, see Figure 2b. To further facilitate the connection,
these tests, including 26.7, 60.3, and 114.3 mm outside each pipe had a steel plate welded to one end with a tab
diameter; however, only results for 26.7 and 60.3 mm pipes attached. The tab was fit into the connector, and both were
are presented herein. The goal is to confirm whether secured using a bolt as a shear pin. The other end of the
displacing the soil is comparable to pulling the pipe, as is pipe is kept unrestrained while a Linear Velocity
widely performed. Displacement Transistor (LVDT) measures any possible
elongation. See also Figure 2.

2 TEST SETUP

2.1 Testing Facility

LVDT
The current test facility was created at Memorial University
of Newfoundland by retrofitting an existing facility. The new
Ram connection
setup is designed to anchor the buried pipe while moving
the soil tank. The setup, see Figure 1, consists of a 4.8 m
base frame bolted to the floor and braced against the
(a) (b)
hydraulic ram’s reaction frame, on which a 4 x 2 m carriage
frame is placed. Between the carriage and base frame are
Figure 1. a) Tank ram-end view with LVDT and tank-ram
six linear bearings (three along the length of each side).
connection visible; b) Close-up of the connection between
The linear bearing’s track is bolted to the frame, and the
the load cell and pipe using the connector.
carriage bolted to the sliding blocks to facilitate virtually

(a) (b)
Figure 2. Testing facility frame configuration; a) without soil tank, and b) with soil tank
 To fill the tank with soil, an overhead crane of 10 T 3 PIPE INSTRUMENTATION
capacity was used to lift large sandbags, which were
emptied into the tank. Soil is placed in roughly 100 to 150 A fibre-optic strain sensor was used during test ST-02 to
mm lifts and then compacted using a battery-powered capture detailed, continuous strain information along the
vibratory plate tamper to obtain the desired density. To length of the pipe. Figure 3 shows the fibre-optic sensor
ensure the appropriate density conditions are met, sand- equipment.
cone replacement tests are performed at the surface level
prior to testing as per ASTM D1556 (2016).

2.2 Testing Programme

The results presented in this paper cover two separate

pipe tests. Two differing diameters were tested with varying
parameters. Both pipes tested were SCH40 steel pipes
with a yield strength of 310 MPa or greater. Table 1 shows
the test name, along with the corresponding parameters
All tests were performed using a 650 mm burial depth,
from pipe springline to surface, with a 0.5 mm/min soil-pull
rate. Moisture contents were limited to <3% and were not
expected to contribute significantly to the axial pull-out
force Saha (2021) reported that suction force due to
moisture content has a negligible effect on the pipe-soil
interaction for moisture content below 3%. Figure 3. Fibre-optic strain sensing equipment

Table 1. Testing Variables The strain setup includes a multi-channel sensing platform
by Luna Innovations Inc. called an optical distributed
sensor interrogator (ODiSI), a laptop controller, a 10 m
Parameter ST-01 ST-02 ST-03
standoff cable, a remote module, and the fibre-optic strain
Outside Diameter (mm) 26.7 26.7 60.3 sensor of 5 m length. As per Luna Innovations Inc. (2020)
User Guide, the ODiSI system operates using “an optical
Wall Thickness (mm) 2.9 2.9 3.9 frequency domain reflectometry (OFDR) technique to
interpret the signal from the [fibre] optic sensors”; A laser
Burial Depth (mm) 650 650 650 is generated and shone through fibre-optic cabling to the
H/D ratio ~24 ~24 ~11 end of the sensor, where there is a termination, which
reflects the light to the sensing platform. Based upon the
Relative Compaction (%) 96.5 1
96 96.3 amount of light that has been reflected, differing due to
backscatter, the system computes the amount of strain in
Moisture Content (%) 2.6 1
2.0 1.3 the sensor and, thus, in the item it is bonded to.
Soil Pulling rate (mm/min) 0.5 0.5 0.5
Luna Innovations Inc. (2017) ODiSI Fiber Optic Sensor
Installation Guide describes the recommended procedure
Instrumented N Y N for adhering a fibre-optic sensor to a material; First, the
surface should be cleaned using an alcohol wipe and
Assumed based on previous testing values
1
sanded with coarse, then finer, sandpaper. After sanding,
the material should be cleaned again before the strain
sensor’s path is laid out. The sensor should be temporarily
Locally manufactured sand is used during the test
held down using a Kapton dot every 300 mm, and a
programme, having properties as described by Saha et al.
polyurethane epoxy applied to each end, anchoring the
(2019); well-graded with an average particle size (D50) of
sensor in place. Once the epoxy has cured, a catalyst is
0.742 mm, a uniformity coefficient (Cu) of 5.81, fines of
applied between Kapton dots and a cyanoacrylate
1.3% and gravel of 0.87%. They also conclude that the
adhesive was carefully spread to bond the sensor to the
maximum dry unit weight of the sand is 19.3 kN/m3 as per
surface (Figure 4). After bonding, a thin layer of silicone
standard Proctor hammer testing.
caulking is added, protecting it from separation during
To perform tests ST-01 and ST-03, the soil tank was
testing and from rupture during soil compaction.
moved by 35 mm at 0.5 mm/min. Once the tank reached
During testing, the ODiSI’s gage pitch was set to 2.6
the desired displacement, it was stopped and held in that
mm. Thus, the system was set to collect a strain reading
location for some time, allowing the pipe to relax. This cycle
every 2.6 mm. The fibre optic strain sensor has a strain limit
of soil movement and pipe relaxation is repeated three
of 3%; however, the system can only calculate capture
times, after which the soil tank is pushed back to its original
strain up to 1.5%.
location. To perform ST-02, the soil tank is immediately
moved by 7 mm to break the pipe-soil bond, then moved at
the aforementioned rate of 0.5 mm/min until 35 mm of
displacement has been achieved.
 Figure 6a shows the force-displacement response for
each test, with reference to peak forces from some notable
literature sources. Note that the diameters of the pipes are
different for those tests (See Table 2) and therefore,
different pullout forces are expected. In general, the axial
pullout force is higher for pipes with larger diameters, as
expected. The peak forces, Tu (kN/m), calculated using
ALA (2005) guidelines, are also included based on
(a) (b) Equation 1.
𝐾𝑜 +1
𝑇𝑢 = 𝜋𝐷𝛼𝑐 + 𝜋𝐷𝐻𝛾̅ 𝑡𝑎𝑛𝛿 (1)
2

Where; D is the outer diameter of the pipe (m), α is an

adhesion factor, c is the soil cohesion (0 kPa for sand,
cancelling the first term), H is the depth (m) from surface to
springline, 𝛾̅ is the effective unit weight of soil (kN/m3), and
(c) δ (°) is the pipe’s coating dependent factor (0.8 for rough
steel) times the friction angle. The dry unit weight of soil
Figure 4. a) Strain sensor path mapped and temporarily was used, given that the soil had low moisture content.
adhered using Kapton dot; b) Application of cyanoacrylate It can be seen in Figure 6a that pipe pulling produced
adhesive between Kapton dots; c) Silicone caulking relatively higher peak forces than the value available in the
applied to pipe after adhesive curing current literature, with the peak for ST-01 and ST-03 being
~13 and ~16 kN, respectively, while Murugathasan et al.
(2021) obtained almost 20 kN at peak for 178 mm diameter
4 RESULTS AND DISCUSSION pipes and Bilgin and Stewart (2009) recorded ~18 kN for
175 mm diameter pipes. However, the results of Sheil et al.
Figure 5a shows the axial force distribution and soil tank (2018) for 350 mm diameter pipes are an exception, as the
displacement with time for test ST-01. A peak force of peak forces developed are lower than the results from ST-
13.25 kN was achieved after roughly 28 minutes, 03, at ~14 kN. ALA guidelines vastly under-predicted the
corresponding to a soil displacement of about 14 mm, peak force, calculating ~2 kN given ST-01 parameters and
given the movement rate of 0.5 mm/min. After this, the load ~4 kN given ST-03 parameters.
decreased steadily until test end as soil softening occurred, The pulling force depends on pipe diameter and buried
reaching roughly 10.3 kN. Figure 5b plots the axial force depth and is proportional to H/D ratio; the highest ratio
and pipe free-end displacement, assumed to be pipe corresponds with the highest peak force, and the lowest
elongation with time. Pipe elongation started after 16 corresponds with the lowest peak force.
minutes, corresponding to about 8 mm of soil displacement For comparison, the normalized pulling forces are
with a registered axial force of 10.12 kN in the pipe. plotted in Figure 6b. Figure 6b plots the normalized axial
Elongation continued steadily until the peak force was force, Na, against the soil tank displacement. Na is
reached, whereby the pipe shrank back to its original calculated based on Equation 2 below:
length. The total elongation captured was roughly 1.5 mm.
During subsequent testing phases, the pipe did not 𝑁𝑎 =
𝑃
(2)
elongate further. Based on these results, the steel pipe is 𝛾ꞌ𝜋𝐷𝐻𝐿
assumed to behave as a rigid body with negligible
elongation

(a) (b)
Figure 5. Force and displacement curves for Pull 1 of ST-01 plotted against time in minutes: a) Pipe axial force and soil
tank displacement; b) Pipe axial force and pipe elongation
 (a) (b)
Figure 6. Force-displacement curves for Pull 1 of ST-01, ST-03 plotted against peak forces obtained from literature: a)
Pipe axial force in kN; b) Pipe axial force normalized.

Table 2. Testing parameter comparisons

Murugathasan et al. Bilgin and Sheil et al. Wijewickreme et al.

Parameter ST-01 ST-03 (2021) Stewart (2018) (2009)
[T3] (2009) [K3] [AB-4]
Pull Rate (mm/min) 0.5 0.5 0.5 N/A N/A 2 to 50
D (m) 0.0267 0.0603 0.178 0.175 0.350 0.457
H (m) 0.650 0.650 0.831 0.760 1.18 1.14
H/D 24.34 10.78 4.67 4.34 3.37 2.50
L (m) 4.0 4.0 2.70 3.05 1.31 3.80
γ (kN/m3) 18.10 18.10 17.17 14.76 16.40 15.70
Relative Compaction (%) 96.5 96.0 88.4 95.0 75.0
1
75.0
Sheil test K3 is said to have full compaction above the pipe using mechanical means; a minimum of 75% relative compaction assumed
1

Where; P is the pipe’s axial load (N), and γꞌ is the soil

effective unit weight (kN/m3). Again, the dry unit weight of
soil was used instead of the effective unit weight for the
calculation.
As shown by Figure 6b, when testing results are
normalized based on individual test pipe diameter, burial
depth and soil density, the axial forces developed during
pipe pull are considerably lower than those developed by
soil pull. ST-01 produces normalized results of almost 3.5,
whereas other studies produced peak results of 1.0 or
below. When normalized, peak forces calculated from ALA
(2005) guidelines just 0.46 using parameters
corresponding to each test. Thus, higher axial forces were
obtained from soil pulling than from pipe pulling.
Test ST-01’s axial force variation is plotted against time
in Figure 7. The peak during each pulling phase is easily
identified, followed by a quick reduction in force as the pipe Figure 7. Pipe axial force variation during each phase for
enters the subsequent relaxation phase. The sharp test ST-01.
decrease in load at these points can be attributed to the
hydraulic system being turned off due to the time required; The first relaxation phase was performed over roughly
consequently, no pressure held the ram or soil tank steady, three days, while the following two were allowed one day
causing unloading of the pipe. each, evident in the horizontal lines plotted after each
pulling phase; dotted lines between phases indicate the bond has weakened and the soil has reached its shearing
reduction of axial force due to turning off of the hydraulic limit. The soil’s ability to transfer load to the pipe will
system. Figure 7 shows that regardless of relaxation continue to decrease, mirroring critical state behaviour.
length, the force decreases rapidly to roughly the same After the third pulling and relaxation phases, the soil is
amount, 4 kN, and is steady during this phase. These pushed back to its starting position of 105 mm, whereby the
findings demonstrate that no time-dependent behaviour axial force drops to a negative value indicating
exists for pipe or soil. If time-dependent behaviour was compression. The high unloading rate is responsible for the
found, it would be ascribed to the soil, as steel is not a time- large gap between the final pulling and pushing phases.
dependent material. Interestingly, this phase produces no significant
It can be seen from Figure 7 that there is a reduction in compression force in the pipe, again indicating the broken
peak force during each subsequent pulling phase. pipe-soil bond.
However, it is more easily shown in Figure 8a, this plots the Figure 8b shows consistently decreasing peak forces
force based on soil tank movement for each phase of test obtained during the pulling phases, represented by discrete
ST-01. Relaxation phases are omitted for clarity. data points and the peak force calculated using ALA
(2005). The difference in peak force between the first and
second pulling phases is 10%, while the difference
between the second and third is slightly higher at 13%. A
trend can be inferred whereby the force decrease equals
5.4% of the soil’s displacement, the slope of the trend line.
Sheil et al. (2018) also reported a drop in peak loads with
repeated loading cycles for compacted sand. However,
their tests investigated cyclic axial push-pull displacements
of the pipe rather than pull-relax cycles for soil movement.
Thus, they present average peak values from a push-pull
cycle. Calculating the peak axial forces from normalized
results presented by Sheil et al. (2018) for test K3, the first
set of cycles produces an average peak force of 7.8 kN, the
second 4.2 kN, the third is 3.3 kN, and from the fourth
onward settles at 3 kN. Based on these findings, it can be
inferred that any repeated movement between soil and
pipe will cause a continual reduction in the peak forces
developed by those movements.
(a)

5 CONCLUSIONS

This paper presented results from a novel testing setup

created at the Memorial University of Newfoundland. A
pipe, anchored at one end and free on the other, was
placed through a moveable steel soil tank. The soil tank
was filled with soil and compacted to a relative density of
96%, then pulled at a rate of 0.5 mm/min until reaching a
total displacement of 35 mm. Once stopped, the system
relaxes for varying timeframes, and the pull-relax cycle
starts anew. The final testing phase pushed the tank back
to its starting position. During testing, pipe axial force, pipe
free-end displacement, soil-tank displacement, and strain
(b) along the pipe were recorded. Based on the data
presented, the following conclusions were drawn;
Figure 8. Force-displacement results from test ST-01
during; a) pulling and pushing phases; b) pulling phases a) Pulling a pipe axially through a static soil mass
only. The graph shows discrete peak force values, produces a different response than pulling the soil
including those calculated by ALA (2005). mass around a pipe fixed at one end.
b) Soil pulling produces much higher pulling force
than pipe pulling (upwards of 3 compared with 1)
The peak force developed during pulling phase 1 is when force is normalised based on soil unit
roughly 13 kN that was reduced to 12 kN in pulling phase weight, pipe length, diameter and burial depth,
2, and to 10 kN in pulling phase 3. The steep reduction in c) Steel pipe acts as a rigid body, with negligible
load between each phase is also evident, given that the elongation when the soil mass is pulled.
ending load of pulling phase 1 is above 10 kN, whereas d) ALA guidelines drastically under-predicted the
pulling phase 2 starts at ~4 kN. The peak force reduction generated axial forces for the soil-pulling case in
in consecutive pulling phases indicates that the pipe-soil dense sand, as with pipe-pulling in dense sand.
 e) Axial load captured during the relaxation phases Robert, D. J., Soga, K., O’Rourke, T. D., and Sakanoue, T.
was steady and consistent, implying no time- 2016. Lateral load-displacement behavior of pipelines
dependent behaviour by either the pipe or the soil. in unsaturated sands. Journal of Geotechnical and
f) Forces developed due to consecutive pull-relax Geoenvironmental Engineering, 142(11): 04016060.
cycles will decrease consistently by around 10%.
Saha, R. C. 2021. Shear strength assessment of a
manufactured well-graded sand (Masters Thesis,
6 ACKNOWLEDGEMENTS Memorial University of Newfoundland).

The authors would like to gratefully acknowledge the Saha, R.C., Dhar, A.S., and Hawlader, B.C. 2019. Shear
financial support and/or in-kind support provided by the strength of a well-graded clean sand, Geo St. John’s
Alliance Grants program of the National Science and 2019, Canadian Geotechnical Society, St. John’s, NL,
Engineering Research Council of Canada, FortisBC CA.
Energy Inc., SaskEnergy Canada Inc., and WSP Canada
Inc. for conducting this research. Sheil, B. B., Martin, C. M., Byrne, B. W., Plant, M., Williams,
K., and Coyne, D. 2018. Full-scale laboratory testing of
a buried pipeline in sand subjected to cyclic axial
7 REFERENCES displacements. Géotechnique, 68(8): 684-698.

ALA (American Lifelines Alliance). 2005. Guidelines for the Wijewickreme, D., Karimian, H., and Honegger, D. 2009.
design of buried steel pipe. Reston, VA, USA: ASCE. Response of buried steel pipelines subjected to relative
axial soil movement. Canadian Geotechnical Journal.
Ali-Khazaali, M., and Vanapalli, S. K. 2019. Axial force- 46(7): 735-752.
displacement behaviour of a buried pipeline in
saturated and unsaturated sand. Géotechnique,
69(11): 986-1003.

ASTM (American Society for Testing and Materials). 2016.

Standard Test Method for Density and Unit Weight of
Soil in Place by Sand-Cone Method. West
Conshohocken, PA, USA.

Bilgin, Ö., and Stewart, H. E. 2009. Pullout resistance

characteristics of cast iron pipe. Journal of
transportation engineering, 135(10): 730-735.

Jalali, H. H., Rofooei, F. R., Attari, N. K. A., and Samadian,

M. 2016. Experimental and finite element study of the
reverse faulting effects on buried continuous steel gas
pipelines. Soil Dynamics and Earthquake Engineering,
86: 1-14.

Luna Innovations Inc. 2020. User’s Guide ODiSI 6, Luna

Innovations Inc., Blacksburg, VA, USA.

Luna Innovations Inc. 2017. ODiSI Fiber Optic Sensor

Installation Guide, Luna Innovations Inc., Blacksburg,
VA, USA.

Murugathasan, P., Dhar, A. S., and Hawlader, B. C. 2021.

An experimental and numerical investigation of pullout
behavior of ductile iron water pipes buried in sand.
Canadian Journal of Civil Engineering, 48(2): 134-143.

Oskouei, A. V., Tamjidi, A., and Pourshabani, P. 2019.

Effects of burial depth in the behavior of buried steel
pipelines subjected to strike-slip fault. Soil Dynamics
and Earthquake Engineering, 123: 252-264.
Geotechnical Risk Management for the
Saddle River Horizontal Directional Drill
Panagiotis Berdousis & Jonathan Woods
Worley Group, Calgary, Alberta, Canada
Alistair McClymont
Formerly Worley Group, Currently BGC Engineering, Calgary,
Alberta, Canada
Jan Bracic & Mick Helsen
Pembina Pipeline Corporation, Calgary, Alberta, Canada

ABSTRACT
This paper focuses on the approach that Pembina Pipeline Corporation and their geotechnical consultant Advisian, Worley
Group adopted to manage the geotechnical risk related to a horizontal directional drilling (HDD) crossing under the Saddle
River in Alberta, Canada. A focused desktop study was completed, with main reference to the well-documented Rycroft
Landslide, which occurred in 1990 and is situated approximately 3 km away from the site. In addition, a geotechnical and
geophysical site investigation comprising four boreholes, laboratory testing and Electrical Resistivity Tomography (ERT)
was completed. The geological model and the interpretation of the extent and mechanism of the Rycroft Landslide were
considered to determine the “no-drill zone” of the HDD crossing, which was successfully completed in 2019, exceeding a
length of 2.2 km and reaching depths of 170 meters below ground surface.

RÉSUMÉ
Cet article porte sur l’approche adoptée par Pembina Pipeline Corporation et son consultant en géotechnique Advisian,
Worley Group pour gérer le risque géotechnique lié à un franchissement de forage directionnel horizontal (HDD) sous la
rivière Saddle en Alberta, au Canada. Une étude documentaire ciblée a été réalisée, avec une référence principale au
glissement de terrain bien documenté de Rycroft, qui s’est produit en 1990 et qui est situé à environ 3 km du site. De plus,
une étude géotechnique et géophysique sur le terrain comprenant quatre forages, des essais en laboratoire et une
tomographie par résistivité électrique (TEI) a été réalisée. Le modèle géologique et l’interprétation de l’étendue et du
mécanisme du glissement de terrain de Rycroft ont été pris en compte pour déterminer la « zone sans forage » de la
traversée HDD, qui a été achevée avec succès en 2019, dépassant une longueur de 2,2 km et atteignant des profondeurs
de 170 mètres sous la surface du sol.

1 INTRODUCTION 2 DESKTOP STUDY

Trenchless crossings in complex landslide terrain pose The desktop study comprised review of available
significant risks to a pipeline project, both during geological maps and considered historic landslides that
construction and over the designed operational lifetime of had developed along the Saddle River area and especially
the pipeline. As such, comprehensive pre-construction the Rycroft Landslide, which developed in 1990 and is
geotechnical assessments are crucial to limit unexpected situated approximately 3 km downriver from the Site.
conditions that may lead to construction cost overruns, or
worse, future slope failures that would compromise the 2.1 Geological Setting
safe operation of the pipeline.
In 2018 Pembina Pipeline Corporation and Advisian, Based on the Alberta Geologic Survey (AGS) surficial
Worley Group (Worley) performed a desktop study and site geology map titled “Map 601 - Surficial Geology of Alberta”
investigation in support of a horizontal directional drilling surficial deposits in the region adjacent to Saddle River
(HDD) design for a replacement crossing under the Saddle likely consist of alluvium, comprising poorly to well sorted,
(Burnt) River. stratified to massive sand, gravel, silt, clay, and some
The crossing was along a NPS10 pipeline from organic sediments. The map also identifies glaciolacustrine
Gordondale to Belloy, 16 km off Belloy, Alberta (the Site). deposits comprising clays, silts, and fine sands, as well as
Figure 1 shows a view of the Site. moraine or till with a mixture of clay, silt, sand and minor
To define and manage geotechnical risks the pebbles, cobbles and boulders within the region (Fenton et
geotechnical teams of Worley and Pembina performed an al., 2013).
extensive desktop study followed by a detailed site A review of the AGS Map 600, “Bedrock Geology of
investigation, as described below. Alberta” identifies that the Site is located within an area of
bedrock referred to as the Smoky Lake Group Kaskapau
Formation, consisting of dark grey, locally calcareous shale
and silty shale with locally abundant, concretionary,
sideritic nodules and is of marine origin (Prior et al., 2013).
 River using a tensioned line. The profile started in the
2.2 Rycroft Landslide canola field west of the river valley and finished in another
canola field east of the river valley. The ERT alignment
The geology in the area of the Rycroft Landslide and an followed an earlier proposed HDD bore path. The location
assessment of the causes that led to soil failure are of the ERT section is shown in Figure 6.
described in Cruden et al. (1993). ERT is a technique for mapping the distribution of
The Rycroft Landslide was a massive landslide that subsurface electrical resistivity (or its inverse; conductivity)
developed progressively and impacted an area in a cross-sectional format. Resistivity data are collected
approximately 800 m long, 1000 m wide and 100 m deep, through a linear array of electrodes coupled to a direct
displacing approximately 40 Mm3 of soil. Figure 2 shows a current (DC) resistivity transmitter and receiver, and an
photo of Rycroft Landslide. A cross-section of the landslide electronic switching box. The final product is a 2-D cross-
is shown in Figures 3 and 4. The geology in the area where section plotting resistivity (in ohm-m) versus depth. The
the landslide developed is shown in Figure 3 and in more resistivity of earth materials is dependent on a number of
stratigraphic detail in Figure 5. factors including mineralogy, porosity, and pore-water
The geology comprises glacial lake deposits (clay and conductivity; however, in general, fine-grained material
silt), underlain by firm to very stiff clay till, underlain by (clays and silt) will have lower resistivities than coarser
preglacial lake high plastic clay, underlain by interbedded grained materials (e.g., sands and gravels).
clay and silt layers, underlain by sand and gravel, underlain
by bedrock consisting of siltstone and shale of the Smoky 3.2 Geotechnical Investigation
Group. The preglacial lake clay deposits were encountered
between approximate elevations 475 and 460 m, whereas The geotechnical investigation was executed from August
bedrock was encountered at approximate elevation 390 m. 8 to September 6, 2018 and involved the drilling of four (4)
In accordance with Cruden et al. (1993), the main boreholes (BHSR18-A, BHSR18-B, BHSR18-C, and
cause that triggered the landslide was lateral and vertical BHSR18-D) to depths of 60.5 m, 79.6 m, 84.9 m and 62 m
erosion of the river slopes following heavy rainfall, which respectively, and the installation of four (4) slope
exposed the high plastic low strength preglacial clay at the inclinometers and 12 vibrating wire (vw) piezometers. The
toe of the slope and led to the progressive development of borehole locations are shown in Figures 6 and 7.
slip surfaces that daylighted through the preglacial clay, Drilling operations were conducted using a track
leading in the end to the generation of the landslide. mounted Fraste FS-250 drilling rig. All boreholes were
A schematic of the ground conditions at the Rycroft drilled using wet rotary drilling methods, and where
Landslide site prior to the event is shown in Figure 4. bedrock was encountered, HQ3 coring was used.
Pembina’s geotechnical team, during a helicopter Additionally, HQ3 coring was advanced in soils
patrol of the Saddle River area, identified an additional, encountered at BHSR18-A and BHSR18-B from 30.5-
older, larger valley-scale landslide a few kilometers 60.5 m and 15.2-48.2 m, respectively. Water was
downriver from Rycroft Landslide. It was particularly noted circulated for both wet rotary drilling and HQ3 coring.
that, like the Rycroft Landslide, this older slide was also on Standard Penetration Tests (SPTs) in accordance with
the northwest valley wall of Saddle River. This suggests ASTM D1586-11 were performed at 1.5-3.0 m intervals.
that valley-scale landslides are occurring as the river head-
cuts upriver, implying susceptibility for a future similar
landslide at the Site, which is upriver of the Rycroft
Landslide.

3 SITE INVESTIGATION

To identify and characterize the soil conditions at the

proposed HDD location, a subsurface investigation was
undertaken by Worley between August 8 and September
6, 2018. The investigation comprised a geophysical survey,
followed by a geotechnical investigation, consisting of four
(4) boreholes, installation of piezometers and inclinometers
and laboratory testing.
The key elements of the subsurface investigation are
described in the below sections.

3.1 Geophysical Survey

The geophysical investigation was executed from August

1, 2018, to August 3, 2018, consisting of a 2,600 m long
Electrical Resistivity Tomography (ERT) profile.
To ensure continuous coverage across both banks and
under the river, cables were suspended across Saddle
Figure 1. The Saddle River site (the Site), with pre-HDD pipeline alignment shown in yellow (photo date: May 17, 2018)

Figure 2. The Rycroft Landslide, 3 km down-river of the Site. Slide extents shown in red (Photo date: May 17, 2018)
Figure 3. Section through the Rycroft Landslide (after Cruden et al., 1993)

Figure 4. Reconstruction of displaced material to show configuration prior to landslide (after Cruden et al., 1993)

Figure 5. Geological section representative of the Rycroft Landslide area (after Cruden et al., 1993)
Figure 6. Site layout including ERT section (in yellow) and borehole locations (in red)

Distance (m)

Figure 7: ERT cross-section showing resistivity values. Interpreted bedrock surface shown with the dashed line
West Approx. 430 m Approx. 210 m
East
Ground Profile

Interpreted
Bedrock

Figure 8: Profile cross-section along proposed HDD alignment (ADIT, 2018), including “No-Drill” Zone shown in red hatch

Figure 9: Plan view showing borehole locations and proposed HDD path (ADIT, 2018), outlined in dark red
 The SPTs were advanced using a 35 mm inside interbedded layers of silt and sand, underlain by clay shale
diameter split-spoon sampler, driven 0.45 m into native bedrock.
ground. Alluvium deposits comprising loose to dense sands and
In addition, two relatively undisturbed samples from gravels interbedded with clay/firm sandy silty clays were
BHSR18-B and BHSR18-D were collected, using a thin- encountered above the silty clay till to 6 m depth at
walled, open Shelby tube sampling device. boreholes BHSR18-B and BHSR18-C, adjacent to Saddle
Soil logging was completed on recovered split spoon River.
samples from SPTs, Shelby tube samples and grab An extract of the geotechnical cross-section, based on
samples recovered from coring. Where relevant, grab the interpretation of the soil conditions in the completed
samples were also collected and placed into sealed plastic boreholes and ERT results, is presented in Figure 10.
bags. Rock logging was completed on recovered rock core,
which was subsequently placed into wooden HQ core 4.3 Bedrock (Clay Shale)
boxes and photographed.
The collected soil samples were submitted to an Clay shale bedrock was encountered beneath the
accredited geotechnical laboratory in Calgary, AB for laminated silty clay deposits at boreholes BHSR18-B and
testing on select samples per ASTM standards as BHSR18-C, at depths of 67.1 mbgs and 30.5 mbgs,
applicable. respectively (generally between elevations of 450 and
Testing included moisture content; Atterberg limits; 487 m). The clay shale was generally fine-grained and dark
grain size analysis, including hydrometer as required; bulk grey with alternating light grey-brown banding throughout.
density; Unconfined Compressive Strength in soil samples; Carbonate cemented layers and marine fossils were
and Uniaxial Compressive Strength (UCS) in rock samples. common throughout the shale.
On the west side of Saddle River, the clay shale
encountered at borehole BHSR18-B was highly to
4 INTERPRETED SOIL CONDITIONS moderately weathered and highly fractured. Strength
ranged from extremely weak (R0) towards the top of the
The subsurface soil conditions were interpreted based on shale and increased to very weak (R1) with depth.
the findings of the boreholes and the ERT. The boreholes Extremely weathered zones characterized by orange
were the basis of the characterization and helped the discolouration and sand-sized particles were noted within
calibration of the geophysical testing results. The ERT the top 2 m. Trace of gypsum crystals was noted
results, once calibrated, helped create a continuous soil throughout the clay shale at borehole BHSR18-B.
model. The clay shale encountered on the east side of Saddle
River at borehole BHSR18-C was divided into two (2)
4.1 Geophysical Survey Interpretation zones:
• the upper zone, which extended between
The ERT cross-sectional results are depicted in Figure 7. elevations of 485 and 470 m, was uncemented to
Resistivity values above bedrock, for the cross-section, are weakly cemented with prominent weak clay-rich zones
interpreted as generally being clay dominant soils for and low strength (UCS=196 kPa, based on one UCS
resistivity values below 15 ohm-m, silt from 15 ohm-m to test). Additionally, this clay shale had a laminated to
25 ohm-m, and sand-gravel-cobble dominant soils massive texture; and
exceeding 25 ohm-m. The geophysical interpretation • the lower zone, which was encountered below
considered the findings of the boreholes (i.e., the 45.7 m depth, had increased cementation ranging
stratigraphy was calibrated based on the soil conditions from weakly cemented to moderately cemented, and
encountered in the boreholes). more of a fissile/laminated texture at 0-15°. UCS tests
Bedrock was interpreted as the base of a clay dominant conducted on this lower shale varied from 2.5 MPa to
soil between depths of approximately 30 to 100 metres 4.4 MPa, with an average of 3.45MPa.
below ground surface (mbgs), which is corroborated by the Rock Quality Designation (RQD) values were logged
encountered shale in boreholes BHSR18-B and BHSR18- when the core was initially retrieved, and the
C. Resistivity values suggest that the bedrock along the encountered bedrock ranged from very poor to fair in
ERT alignment may be predominantly shale or siltstone. borehole BHSR18-B and very poor to excellent in
Sand-gravel-cobble dominant soils may be of interest borehole BHSR18-C.
to drilling activities; therefore, resistive areas of interpreted
sand-gravel-cobble dominant soil have been marked on
the cross-section with the pink-colored regions with hatch 5 HDD FEASIBILITY
shading. These interpretations are corroborated by thick
sand deposits visually seen near line station 1190 m. To determine the HDD feasibility the geotechnical team
considered the results of the desktop study, with special
4.2 Interpreted Subsurface Profile focus on the Rycroft Landslide mechanism, as well as the
interpreted subsurface conditions based on the
The subsurface profile at the Site generally consists of geotechnical boreholes and ERT results.
glaciolacustrine deposits comprising firm to very stiff silty
clays, underlain by stiff to hard silty clay till, underlain by
very stiff to hard medium plasticity laminated silty clays with
 HDD ENTRY/EXIT POINT

OF SADDLE RIVER
APPROXIMATE C/L
BHSR18-B
BHSR18-A
WEST

WEST

GROUND PROFILE

NO DRILL ZONE
El. +480 m
20m MIN CLEARANCE

Figure 10: Extract from geotechnical cross-section along proposed HDD alignment, including “No-Drill” Zone (west side)

Figure 11: Plan view showing proposed HDD path (red colour), borehole locations and ERT section (west side)
5.1 Recommended “No Drill Zone”

The geology at the Site resembles conditions at the Rycroft 6 CONSTRUCTION

Landslide (silty clay, clay till, laminated clay interbedded
with silt and sand, clay shale are common in both areas), The HDD was successfully constructed as per the design
although on Site preglacial high plastic clay was not between January to May 2019. It was an “intersect” HDD
encountered and the bedrock was encountered higher more than 2,290 m long and was installed at depths that
(between elevation of 450 to 487 m, as opposed to 390 m reached 170 m. Casing was used on both HDD entry
at the Rycroft Landslide area). Nevertheless, the top 15 m locations for 29 m each, at an entry angle of 18 degrees on
of the clay shale encountered on the east of the Saddle each side. The HDD drilling and pipeline pullback were
River (Borehole BHSL18-C), between approximate completed between January and March 2019, with final tie-
elevations of 485 and 470 m, was generally uncemented to in completed May 2019.
very weakly cemented, with clay-rich zones of low strength.
This weak clay material, if exposed following erosion of
overburden soils, is prone to softening and development of 7 CONCLUSIONS
tension cracks, especially after repeated wet-dry cycles
over time, which can lead to slope instabilities, i.e., it may To manage the geotechnical risk associated with the
behave similar to the preglacial clays that enabled the Saddle River HDD, the geotechnical team conducted an
Rycroft Landslide. In addition, visual inspection of the extensive desktop study and a site investigation comprising
banks of the Saddle River by Pembina has identified other four deep boreholes, laboratory testing and geophysical
areas where similar landslides to the Rycroft Landslide testing (ERT). It was identified that the development of a
have historically developed. landslide through clay-rich zones of low strength
Taking into account the importance of the NPS10 encountered within the clay shale, similar in size to the
pipeline, the presence of the low-strength clay-rich zones Rycroft Landslide, is a possible geohazard at the Site. For
between elevations of 485 and 470 m on the east side of this reason, the HDD entry/exit points and profile depth
Saddle River and the historical evidence of development of were designed, and constructed, such that the new pipeline
landslides along the Saddle River, it was decided that the avoids a similar size and type landslide to the Rycroft
“No Drill” zone resembles the Rycroft Landslide (refer to Landslide.
Figures 3, 4). For this reason, the “No Drill” zone extended HDD pipeline crossings are situated generally in areas
approximately 430 m behind the crest of the west slope and of difficult access. The strategy of performing a desktop
approximately 210 m behind the crest of the east slope, study followed by a site investigation comprising of
with near vertical phase, and to an elevation of 480 m. It boreholes and geophysical testing (ERT and seismic
was also decided that the HDD drill path have a minimum refraction if necessary) is a recommended practice to
radial clearance from the “No Drill” zone of 20 m. The identify subsurface geotechnical conditions and manage
extent of the “No Drill” zone to the west is longer than this geohazards to the pipeline.
to the east to account for the 3-dimensional (3D) The geophysical testing was calibrated with the
topography of the slope. The proposed “No Drill” zone is borehole data, and in that way a continuous geotechnical
shown in profile cross-section presented in Figure 8. The model could be derived.
detail of the west side of the geotechnical cross-section
with the “No Drill” zone is shown in Figure 10.
3 REFERENCES
5.2 Entry and Exit Considerations
Cruden, D.M., Keegan, T.R. and Thomson, S. 1993. The
Due to the length of the proposed HDD an intercept is landslide dam on the Saddle River near Rycroft,
necessary and entry/exit points are needed at each end of Alberta. Can Geotech. J. 30, 1003-1015. 1993.
the drill path.
Considering the extent of the Rycroft Landslide, it was Fenton, M.M., Waters, E.J., Pawley, S.M., Atkinson, N.,
decided that the entry/exit points be located a minimum of Utting, D.J. and Mckay, K. 2013. Surficial Geology of
approximately 500 m away from the crest of the slopes Alberta (Map 601), Scale 1:1000000, Alberta
(300 m setback and 200 m buffer). The selected location of Geological Survey.
the west entry/exit points and relevant clearance is shown
in Figure 11. Prior, G.J., Hathway, B., Glombick, P.M., Pana, D.I.,
Based on our understanding of the area, the entry and Banks, C.J., Hay, D.C., Schneider, C.L., Grobe, M.,
exit conditions at the crossing location were expected to Elgr, R. and Weiss, J.A., 2013. Bedrock Geology of
encounter alluvium materials, fine silty clay glaciolacustrine Alberta (Map 600), Scale 1:1000000, Alberta Geologic
material underlain by glacial clay till. Small size localized Survey.
sand and gravel zones (as indicated in the ERT cross-
section) were also likely to be encountered. ADIT Engineering Inc., 2018. Saddle (Burnt) River 2-13-78-
Casing and possibly bentonite drilling fluid to stabilize 4 W6 Horizontal Directional Drill. Issued for
these weak surficial materials and assist in drilling fluid Construction Drawing.
control, was likely to be necessary, subject to confirmation
by the HDD contractor.
Loading-unloading-reloading responses of
MDPE pipes in sand subjected to axial
ground movement
Sudipta Chakraborty1, Saifa Anjum1,
Ashutosh Sutra Dhar1, and Mujib Rahman2
1Department of Civil Engineering – Memorial University of

Newfoundland, St. John's, NL, Canada

2FortisBC Energy Inc., Surrey, BC, Canada

ABSTRACT
Medium-density polyethylene (MDPE) pipes are one of the common means of gas distribution in Canada and worldwide.
These pipes are sometimes exposed to ground displacements caused by landslides and other geohazards. Pipelines can
be subjected to load-unload-reload cycles in areas where landslides occur in succession. Repeated loading-unloading-
reloading on buried MDPE pipes can affect behavior at the pipe–soil interface and the corresponding pipeline responses.
However, very limited study is currently available on the behavior of pipes subjected to loading-unloading-reloading due to
landslide. This paper presents a study on the loading-unloading and reloading behavior of MDPE pipes under axial ground
movement. A full-scale laboratory facility developed at the Memorial University of Newfoundland was used in the tests.
The facility was designed to simulate the axial ground movement where a soil box is moved over two rails. During testing,
the pipe was instrumented with fiber optic strain sensor along the entire length to monitor axial strain. A significant drop in
axial soil resistance was observed after the first cycle, indicating the damage of soil–pipe interface bonding. Pipe wall
strains gradually decrease from the fixed end of the pipe to the free end along the length.

RÉSUMÉ
Les tuyaux en polyéthylène à densité moyenne (MDPE) sont l'un des moyens courants de distribution de gaz au Canada
et dans le monde. Ces conduites sont parfois exposées à des déplacements de terrain causés par des glissements de
terrain et d'autres géorisques. Les pipelines peuvent être soumis à des cycles de chargement-déchargement-
rechargement dans les zones où les glissements de terrain se succèdent. Le chargement-déchargement-rechargement
répété sur des conduites enterrées en MDPE peut affecter le comportement à l'interface conduite-sol et les réponses
correspondantes de la conduite. Cependant, des études très limitées sont actuellement disponibles sur le comportement
des conduites soumises à des chargements-déchargements-rechargements dus à des glissements de terrain. Cet article
présente une étude sur le comportement de chargement-déchargement et de rechargement des tuyaux en MDPE sous
mouvement axial du sol. Une installation de laboratoire à grande échelle développée à l'Université Memorial de Terre-
Neuve a été utilisée dans les tests. L'installation a été conçue pour simuler le mouvement axial du sol lorsqu'une caisse
de sol est déplacée sur deux rails. Pendant les essais, le tuyau a été équipé d'un capteur de contrainte à fibre optique sur
toute la longueur pour surveiller la contrainte axiale. Une baisse significative de la résistance axiale du sol a été observée
après le premier cycle, indiquant l'endommagement de la liaison de l'interface sol-tuyau. Les contraintes sur la paroi du
tuyau diminuent progressivement de l'extrémité fixe du tuyau à l'extrémité libre sur toute la longueur.

1 INTRODUCTION contact stresses and strain on the pipe wall) can be

affected greatly due to repeated loading-unloading-
Pipelines are the most common means of transportation reloading. Moreover, cyclic loading also affects soil
and distribution of gas and fluids. Among them, medium- responses, resulting in fatigue damage to the pipeline
density polyethylene (MDPE) pipes are greatly used in (Cullin et al. 2015). Stress–strain behavior and distribution
Canada and worldwide, especially for gas distribution. The in soils can vary considerably when subjected to repeated
pipes are sometimes exposed to ground displacements loading and unloading (Wijewickreme and Soysa 2016;
induced by landslides, ground subsidence, liquefaction, Nova and Hueckel 1981). For MDPE pipes, the flexibility
and other geohazards. As gas distribution pipelines are and time-dependent stress–strain behavior of material
mostly in residential areas, these ground displacements (Das and Dhar 2021ab) can also contribute to the pipeline
can cause severe damage to the pipe networks and responses (Reza et al. 2023; Reza and Dhar 2021;
surrounding communities. Pipes can be subjected to load- Wijewickreme and Weerasekara 2015; Weerasekara and
unload-reload cycles in areas where landslides or other Wijewickreme 2008). However, studies currently available
ground movements occur in succession. Even the pipes on the behavior of MDPE pipes exposed to load-unload-
can experience loading-unloading-reloading due to the reload cycles due to landslides are very limited.
temperature variation of the content and during start-up This paper presents a full-scale laboratory
and shut-down. The pipe–soil interaction for the buried investigation of 42.2-mm diameter MDPE pipes buried in
pipelines and the corresponding pipeline responses (i.e., dense sand subjected to load-unload-reload cycles. The
testing facility at the Memorial University of Newfoundland includes a 10 m standoff cable and remote module for each
is designed to simulate axial ground movement where a channel and a ruggedized shipping case. The sensors
soil box is moved over two rails. The pipe buried in the soil used for this study provided a unique Rayleigh scatter
box is fixed at one end and free on the other end. Ground signature that the ODiSI-6 transformed into distributed
loads are applied to the pipe by subsequent pulling and strain (LUNA 2021). The sensors had a measurement
pushing of the tank. After each loading or unloading, the range of around 1.2% with instrument accuracy of about
test facility was left in a static condition to examine the ±1 µε. Each sensor has a designated key to initialize the
stress relaxation. The pipe was instrumented with fiber properties and reference of measurement of the sensor to
optic strain sensor along the entire length to monitor axial the controller. The remote module and standoff cable are
strain. This study aims to examine the pipe responses used to modify the length to secure a safe connection
subject to axial landslide while the soil is disturbed. between the sensor and controller. The FO sensors were
attached to the pipe wall using cyanoacrylate adhesive
from Micro-Measurements, which provides strong
2 TEST FACILITY adhesion at ambient temperature and is suitable for short-
term applications.
Figure 1 shows the test facility at Memorial University of
Newfoundland. The facility had recently undergone
redevelopment, which was already equipped with a range
of state-of-the-art equipment, including a 4.0m (L) x 2.0m (a) 4
(W) x 1.5m (H) steel tank, a hydraulic ram with an actuator
system, a data acquisition system (DAQ), and associated 5
instrumentation. The details of the test tank are available in
Murugathasan et al. (2020) and Reza and Dhar (2021). 3
One of the key features of the redeveloped facility was the
introduction of a sliding tank system. This system consisted
of a fixed base frame (1) with segments of rails (2) along
the long side and a steel carriage frame (3) with wheels
running over the rails. The steel tank is fixed on the carriage
frame that can move with the frame over the rails. A back
vertical frame (4) holds the load cells where the test pipe 1
segment is restrained. A connection frame (6) is attached
to the steel tank (7). The connector frame is attached to a 2
hydraulic ram that can pull the tank over the rails.

2.1 Distributed Fiber Optic Sensors (b)

7
During the tests, pipe strains were measured using
distributed fiber optic (FO) strain monitoring system instead
of the conventional discrete foil strain gauges. The
conventional strain gauges affect the test results due to the
involvement of wires and protective wraps (Reza et al.
2023). As a distributed sensor (or sensing cable) is
sensitive at every point along its length, it can replace
thousands of discrete sensors. Moreover, distributed 6
sensors are less difficult and more economical to install
and operate as they require a single connection instead of
a large number of connecting cables like wired discrete
sensors. The primary mechanism by which the fiber optic
sensors work is by analyzing the backscattered light in the
silica core of the fiber to provide information relating to the
condition of the fiber (Agrawal 2001). No. Description No. Description
Among the three categories of backscatter, namely
Raman, Brillouin, and Rayleigh, the third one is caused by 1 Base frame 5 Load cell holder
reflections of imperfections along the fiber optic cable.
Therefore, Rayleigh-based systems would provide a Connection frame
2 Steel rails 6
complete understanding of the strain behavior along the to ram
pipe length with micro strain-level accuracy (Kreger et al. 3 Carriage frame 7 Steel Tank
2006).
The current study employed ODiSI-6100 series system 4 Back frame
by Luna Technologies for measuring strains. The system
comprises an instrument controller laptop, an interrogator
with eight enabled channels, HD standard 5 m length mode Figure 1. Test facility: (a) Base frames and (b) tank
sensors, and ODiSI application software (Figure 2). It also on top of sliding system
 shallow depths from 0.3 m to 1.5 m (Groves and
Wijewickreme 2013).
Instrument
controller Free
end
Standoff

Load cell
Sensor
cable Interrogator
1. Tank Pull

Fiber optics

Load cell
2. Relaxation

Remote Sensor key Connector

module Cleaner

Load cell
Figure 2. Components of ODiSI 6100 system
3. Tank push

2.2 Pipe Instrumentation Procedure

Figure 3. Schematics of the test procedure for one cycle
The process of mounting the FO sensor on the pipe follows
four major steps. Firstly, the route of the sensor was
planned. Then the routed surface of the pipe was cleaned The main difference between Test-1 and Test-2 is that
and sanded to a suitable roughness for epoxy adhesion. the first one is instrumented with fiber optics throughout its
The sanding was done twice in a hatch abrasion pattern length, but the second one has only one discrete strain
using 220-grit sandpaper followed by 400-grit sandpaper. gauge at the fixed end of the pipe. Moreover, two cycles of
Surface preparation was completed by wiping the area with loading-unloading-reloading tests were done for Test-1,
alcohol wipes. The second step is to tape down the fiber but for Test-2, it was one cycle.
sensor in the desired layout using Kapton dots. The tape For the first cycle of Test-1, the tank was pulled for 120
dots were applied at a maximum 30 cm interval to prevent mm with a subsequent relaxation of 20 hours, and then the
slack buildup. tank was pushed back in the opposite direction of pulling
The third step was to apply and cure the adhesive. As for 100 mm. Figure 3 shows the loading steps in one cycle.
the bond line quality determines the strain transfer During this push of the tank, the pipe buckled toward the
effectiveness, the adhesive amount must be just enough fixed end, so the tank was slightly pulled to straighten the
for adhesion. Otherwise, excessive epoxy can accumulate pipe. Then the second cycle of loading was performed. The
between the fiber and the pipe surface. The cyanoacrylate details of the test program are summarized in Table 1.
epoxy was squeezed onto the foam swab, and the swab
was dragged on the exposed portions of the fiber. Then the
Kapton dots were removed, and the epoxy coating was Table 1. Test program details
applied in these locations. When the adhesive was cured, Parameters Test-1 Test-2
urethane epoxy was applied at the two ends to anchor the
fiber and hold it in place. Pipe diameter, mm 42.2 42.2
Finally, after the installation, the sensors were covered Soil cover, mm 600 600
by an additional layer of GE-7000 sealant, which provided Test Cycle 2 1
protection from abrasion by soil particles. This additional Fiber Strain
Instrumentation
protective layer was added to prevent accidental damage optics gauge
to the fiber optics wire during pipe installation and preserve Pull-1 120 95
the integrity of the bonded fiber sensor. Then the pipe was Tank
left to cure for 24 hours. Pull-2 60 N/A
displacement,
mm Push-1 100 25
2.3 Test Program Push-2 70 N/A
The test program is planned to investigate the behavior of Duration, Relax-1 20 20
MDPE pipes in the sand under the loading-unloading- hours Relax-2 20 20
reloading cycle. Two tests (Test 1 and Test 2) were
conducted on 42.2-mm diameter MDPE pipes, which were
4.6 m long (4 m inside the cell) with Standard Dimension All instruments were connected to a data acquisition
Ratio (SDR) 10. In each test, the pipes were buried at a system equipped with a personal computer to record data
depth of 600 mm (H/D = 14), which is within the range of simultaneously. The backfill materials and pipe installation
standard practice of installation of gas distribution pipes at procedures used are discussed in the following section.
2.4 Backfill Material Preparation and Pipe Installation 3 RESULTS AND DISCUSSION

After instrumentation, the pipe was laid inside the tank on 3.1 Load–displacement Responses
a prepared (compacted) sand bed, and the load cell was
connected with it (Figure 4b). Then, sand was dumped into Figure 5 shows the loading sequence and the responses
the tank using an overhead crane, maintaining a height of of the pipe during the tests. The load was applied by
approximately 1 m, and spread using hoes and rakes in moving the soil tank against the pipe segment fixed at one
150-mm-thick lifts. Each layer was leveled using a shovel end. The load applied to the pipe by the moving soil was
and then compacted by using a hand compactor having a measured by the load cell at the fixed end (i.e., reaction
weight of 4.5 kg. Each lift was compacted using five passes force). The soil tank was moved away from the pipe (i.e.,
of the compactor. The moisture content of sand was loading) for a duration of about 4 hrs at the rate of 0.5
maintained at around 2.5% for the tests. After the pipe mm/min to a maximum displacement of 124 mm and kept
installation, the fiber optics were carefully connected to its in a static position for about 20 hrs and then the tank was
controller system. To obtain strain data at regular intervals moved toward the pipe at the same rate (i.e., unloading).
of 2.6 mm along the pipe length, a gauge pitch of 2.6 mm Then, the procedure was repeated (reloading was applied).
was selected for the sensors. Strain data for the points of
interest was collected by setting gauge points either by
mentioning the distance from the terminator or using touch-
to-locate method. The complete test facility during a test is
presented in Figure 4c.
A locally available well-graded sand (SW, according to
USCS) is used as the backfill material for the pipe. Saha et
al. (2019) studied the mechanical properties of this sand.
The maximum dry density was estimated as 19.3 kN/m3
using standard Proctor compaction tests (ASTM D698-12;
Saha et al. 2019).

(a) (b) I: Pull-1 IV: Pipe straightening

II: Relaxation-1 V: Pull-2
III: Push-1 VI: Relaxation-2
VII: Push-2

Figure 5. Test sequence and pipe responses (Test 1)

As seen in Figure 5, the reaction force increases with

loading and reaches a constant peak value of 6.3 kN at the
soil tank displacement of beyond 100 mm. When the
(c) hydraulic pump was stopped at the tank movement of
around 125 mm, the reaction force suddenly dropped to 5.6
kN, probably due to the sudden release of energy. After
that, the reaction force gradually reduced due to relaxation.
The reduction of the force was nonlinear in the first couple
ODiSI of hours and then linear at the reduction rate of 0.31% per
LVDT hour. The relaxation behavior might be due to the time-
DAQ dependent behavior of pipe materials and the soil–pipe
interface.
RAM
Axial force due to soil movement caused axial
deformation, resulting in the movement of the free end of
the pipe with the movement of the soil mass. The free-end
movement is equal to the elongation of the pipe due to the
axial force. Figure 5 shows that the free-end movement of
Figure 4. A Typical test setup: (a) pipe preparation, the pipe (i.e., elongation) follows the soil movement and the
(b) Pipe installation, (c) test facility prepared for a test axial force development, reaching 59 mm at the peak soil
movement of 125 mm. The elongation was slightly reduced
to 57.2 mm due to the stopping of the hydraulic pump and
After completing each test, sand cone tests were then relaxation, similar to the reaction force. Thus, the pipe
conducted at three different locations inside the tank, which deformation is associated with the redistribution of stress
yielded an average relative compaction of the backfill resulting in the reduction of axial force.
material 96% of the standard Proctor maximum dry density. Note that soil mass (i.e., test cell) movement is greater
than the pipe elongation (free end movement) from the very
beginning, indicating that the relative ground movement is the buried portion of the pipe in Test-2. The peak axial
mobilized over the entire pipe length from the very forces of 6.3 and 3.7 kN were observed during soil tank
beginning. However, when a pipe is pulled against a static movement for Test-1 and Test-2, respectively. The axial
soil mass, the relative ground movement is gradually force is significantly higher in Test-1 due to the resistance
mobilized from the pulling end to the trailing end (Reza and at the pipe surface caused by the fiber optic strain sensor
Dhar 2021, 2023; Wijewickreme and Weerasekara 2015). and caulking. Caulking was essential to protect the strain
The trailing end did not move immediately after the sensor from the surrounding soil. The load–displacement
application of load in their tests. Thus, load transfer response was also more ductile for the pipe with the FO
mechanisms on the pipe from the surrounding soil are strain sensor. For the pipe without the strain sensor, the
expected to be different when a soil mass moves against a peak pulling force was reached at the soil movement of
static pipe than those when a pipe is pulled against a static around 40 mm, while for the pipe with the strain sensor, the
soil mass. As a soil mass often moves against a pipe peak force was reached at the soil displacement of beyond
anchored in the stable ground, tests with a movement of 100 mm (Figure 7a). Thus, surface roughness induced by
ground against a static pipe would provide a more realistic the strain sensor may significantly affect the test results.
loading scenario expected in the field.
During reverse movement of the soil box, the axial
reaction force and the pipe elongation (free-end
displacement) reduce (Figure 5). However, unloading did
not follow the loading path as the load transfer mechanisms
were different. With about 31 mm of soil box movement
(94.1 mm tank displacement from the original position), the
reaction force was zero, and the pipe elongation was
reduced by about 22.4 mm to 34.8 mm. A larger movement
of the soil box than the reduction of pipe elongation
indicates again that relative movement of the soil was
mobilized over the entire pipe length. Further movement of
the soil box caused compressive reaction force. As a result,
the exposed end of the pipe kinked (Figure 6). The soil tank
movement was then stopped and pulled in the opposite
direction to straighten up the pipe segment. After that,
reloading was applied.
The pipe responses during reloading were similar to
those during loading, except that the peak reaction force
and overall pipe elongation were less. The maximum
reaction during reloading was 5.3 kN, and the maximum
pipe elongation was 52.3 mm. Thus, soil resistance was
reduced by about 10% during the second load cycle. A
reduction of soil resistance during repeated loading was
reported earlier in Sheil et al. (2018).

straightening
on process

Figure 7. Load–displacement response: (a) with tank

displacement (b) with pipe free-end displacement
Figure 6. Kink observed and resolved during moving of
soil box
During a relaxation period of about 20 hrs, the axial
force reduced from 6.7 kN to 5 kN (by 21%) in Test-1 and
The axial force developing on the pipe depends on the
from 3.7 kN to 3.2 kN (13.5%) in Test-2. Pipe elongation
soil displacement, while pipe elongation is related to the
also followed the axial force and was reduced during the
mobilized axial force. Figure 7 plots the axial force against
relaxation. During reverse movement (push) of the soil box,
the tank/soil displacement (Figure 7a) and the pipe
the axial force and pipe elongation was further reduced.
elongation (free end displacement, Figure 7b) for the tests
The pushing of the soil box induced compressive force,
(Test-1 and Test-2). Pipe responses at various stages of
reducing the tensile axial force and pipe elongation. The
loading (i.e., pull, push, relaxation, etc.) are identified in the
load transfer mechanisms to the pipe during compression
figure. The pipe in Test-1 was instrumented with FO sensor
were different from those during tension, as the interface
over the full length of the pipe, while no sensor was within
friction is mobilized from the free end toward the fixed end 3.2 Pipe Wall Strain
of the pipe. The pipe section kinked at the axial
compressive force of 5.6 kN in Test-1. The peak frictional Axial strain distributions along the pipe length measured by
resistances of interface soil under compression loading the fiber optic strain sensor in Test-1 is depicted in Figure
were not reached during the tests. 9 at various level of soil displacements. Figure 9a plots the
The maximum frictional force calculated according to strain during loading, and Figure 9b shows the strains
ALA (2005) and PRCI (2017) design guidelines is during relaxation. Pipe axial strains measured outside the
compared with the axial force measured during the test in test cell using an electrical strain gauge in Test-2 (without
Figure 8. Results for Test-2 were used for this comparison the fiber optic sensor) are also included in Figure 9a. The
as no strain sensor was used for this test, and thus, the strains measured by the electrical strain gauge in Test-2
interface roughness is more representative of the field are denoted by "SG" in the figure. Strain measured outside
condition. Researchers reported earlier the limitations of the box in Test-2 is consistent with the strain measured
the current design guidelines in calculating the maximum using the FO sensor at a similar location in Test-1 at a low
axial pulling force (Weerasekara and Wijewickreme 2008; soil displacement (i.e., 5 mm), validating the strains
Reza and Dhar 2021, 2023). Reza and Dhar (2023) measured by the FO sensor. At low soil displacement, the
modified the ALA (2005) equation, including the effects of frictional resistance was not mobilized on the pipe surface,
compaction-induced lateral earth pressure and the contact and therefore, the contribution of surface roughness by the
pressure changes due to shear-induced dilation in dense
sand and/or pipe diameter changes, as shown in Eq. 1: (a)

H + 𝐾0 H + com
p = D( )tan () [1]
2

where p denotes the maximum soil resistance per unit

length of the pipe, γ is the unit weight of soil, H is the depth
of soil cover measured from the ground surface to pipe
springline, D denotes the pipe outer diameter, K0 is the
lateral earth pressure coefficient of soil at rest, and d is the
interface friction angle between pipe and the surrounding
soil. com (compaction-induced lateral earth pressure) can
be calculated using the method described in Duncan and
Seed (1986) and  (normal stress adjustment factor)
depends on interface stress level, internal pressure,
stiffness of the soil and pipe materials, and interface friction
angle. Figure 8 shows that the maximum axial force was
underestimated by the method recommended in ALA
(2005) and overestimated by the PRCI (2017) method. The (b)
method proposed in Reza and Dhar (2023) reasonably
estimated the peak axial force.

Figure 9. Axial strain distribution along the pipe length

during (a) pulling and (b) relaxation for Test-1

FO sensor on the pipe response was negligible. However,

Figure 8. Comparison of axial pullout load with prediction
as the soil displacement increased, the surface roughness
contributed to the pipe responses significantly, and dropped by 50%, while the axial force was reduced by 20%
therefore, the mobilized axial force in Test-1 was greater due to the shutting down of the hydraulic ram for the stress
than those in Test-2. As shown earlier in Figure 7a, the relaxation phase.
axial forces measured in Test-1 and Test-2 were almost the
same up to the soil displacement of ~ 5 mm, beyond which
the axial forces in Test-1 were greater than those in Test- 4 CONCLUSIONS
2. The difference between the axial forces (Figure 7a)
increased with the increase of soil displacement, which is This study describes a full-scale laboratory facility
consistent with the difference between the strains designed to simulate axial ground movement by moving a
measured in Test-1 and Test-2 (Figure 9a). soil box over two rails. A 42.2-mm diameter MDPE pipe
In Figure 9a, strains measured by the FO sensor are buried in dense sand backfill in the soil box was fixed at
zero beyond the distance of 1 m at the soil displacement of one end and free on the other. Subsequent pulling and
5 mm. The non-zero strain extends gradually beyond 1 m pushing of the tank were employed to apply ground loads
with the soil displacement. Thus, the axial strain is to the pipe. The objective of the study was to explore the
gradually mobilized from the fixed end to the free end of the pipe–soil interaction responses subject to load-unload-
pipe as the soil box moves away from the fixed end. A reload cycles. The major findings from this research are as
similar mechanism was reported from axial pullout tests of follows.
pipe against a static soil mass in Reza and Dhar (2021,
2023). • Tests performed with a movement of a soil mass
As the axial strain depends on the axial stress/force, against a static pipe provide a realistic loading
which in turn depends on the pipe-soil interface shear scenario expected in the field, as a soil mass often
stress, it can be concluded from the measurements of moves against a pipe anchored in the stable ground.
strains that the interface shear stress also mobilizes Load transfer mechanisms during the moving of soil
gradually from the fixed end to the free end of the pipe. against a static pipe were different from those of
Figure 9a shows that the strain is linear up to a certain pulling a pipe against a static soil mass.
distance from the fixed end and then nonlinear. The strain • Test cell movement was greater than the overall
distribution would be linear if the interface shear stress is elongation of the pipe during the tests, indicating
constant, and hence, the axial force/stress varies linearly relative movement of the entire soil mass. However,
with distance. Similarly, the strain would be nonlinear if the elongation is considered equivalent to the ground
interface shear stress is linear or nonlinear (the axial force movement until the trailing end moves for pipe pull
or stress changes nonlinearly). Thus, based on the strain tests.
distribution in Figure 9a, shear stress along the pipe length • As in pipe-pulling tests, the axial stresses/strains
can be assumed as in Figure 10. Wijewickreme and were gradually mobilized from the fixed end to the
Weerasekara (2015) assumed a similar distribution of free end.
mobilized axial force for the calculation of axial strains in • Pipe axial force was reduced during the relaxation
MDPE pipes under axial pullout loading. period when the test cell was static. The reduction of
the force was nonlinear in the first couple of hours and
then linear at the reduction rate of 0.31% per hour.
Soil boundary Point of strain/ shear The relaxation caused redistribution of the stress field
stress mobilization that affected the overall deformation of the pipes.
• Soil resistance is reduced during cycles of loading. A
reduction was about 10% in the second load cycle.
Figure 10. Distribution of interface shear stress • Peak axial resistance was underestimated by ALA
(2005) and overpredicted by PRCI (2017) design
methods. The method proposed in Reza and Dhar
The interface shear stress can be constant when the (2023) was successful in predicting the peak axial
maximum shearing resistance of the soil is fully mobilized. force.
Then, the relative movement of a point on the pipe surface • Change in pipe surface roughness by the FO sensors
(with respect to the surrounding soil) would be beyond the can significantly affect the test results. However, the
elastic limit. As the displacement of the soil box increases, fiber optic sensor can provide valuable information
the maximum shear strength is mobilized over al longer about the load transfer mechanism along the pipe
distance from the fixed, as observed in Figure 9a. At the length. Strains greater than 1.2% could not be
test cell displacement of 120 mm, the maximum shear measured with the fiber optic strain sensor. This
strength is mobilized over the entire pipe length within the strain limit corresponds with the strain measurement
soil, and thus, the strain distribution is linear. Note that range of the fiber optic strain measuring controller
strain higher than 12000 µε (i.e., 1.2%) could not be system.
measured using the fiber optic strain sensor as the limiting
• Strain along the pipe length is linear when the
strain value with the controller system was 1.2%. The FO
maximum shearing resistance of the soil is mobilized
sensor can be damaged at a strain level beyond 3.0%.
and the interface shear stress is constant. Otherwise,
Some redistribution of strains occurred during the
the strain distribution is nonlinear due to non-constant
relaxation phase, but the strain distribution along the length
interface shear stresses.
remained linear (Figure 9b). The strain immediately
 • Based on the strain distribution, the distribution of PRCI. 2017. Pipeline Seismic Design and Assessment
shear stress along the pipe length can be assumed. Guideline, Pipeline Research Council International,
Chantilly, VA, USA.
Reza, A. and Dhar, A.S. 2021. Axial Pullout Behavior of
5 ACKNOWLEDGMENT Buried Medium Density Polyethylene Gas Distribution
Pipes, International Journal of Geomechanics, ASCE,
The authors would like to gratefully acknowledge the 21(7): 04021120.
financial support provided by the Alliance Grants program Reza, A. and Dhar, A.S. 2023. Finite–Element Modelling of
of the National Science and Engineering Research Council Axial Movements of Polyethylene Pipes in Dense Sand
of Canada, FortisBC Energy Inc., and SaskEnergy Canada (under review).
Inc. for conducting this research. Thanks to Billy Bidgood Reza, A., Dhar, A.S. and Rahman, M. 2023. Strain
and Jamal Tinkov for their technical assistance in Assessment of Polyethylene Pipes in Dense Sand
developing the laboratory test facility. Subjected to Axial Displacement, Geosynthetics
International, In Press.
Saha, R. C., Dhar, A.S. and Hawlader, B.C. 2019. Shear
6 REFERENCES Strength Assessment of a Well-graded Clean Sand, In
Proceedings of 72nd Canadian Geotechnical
Agrawal, G. 2001. Nonlinear Fiber Optics, 3rd Ed., Conference, Geo St. John's 2019, St. John's, NL,
Academic Press, San Diego, CA, USA. Canada, September 29-October 2.
ALA. 2005. Guidelines for the Design of Buried Steel Pipe, Sheil, B.B., Martin, C.M., Byrne, B.W., Plant, M., Williams,
American Lifelines Alliance, Reston, VA, USA. K. and Coyne, D. 2018. Full-Scale Laboratory Testing
ASTM D698-12. 2021. Standard Test Method for of a Buried Pipeline in Sand Subjected to Cyclic Axial
Laboratory Compaction Characteristics of Soil Using Displacements, Géotechnique, 68(8): 684–698.
Standard Effort, ASTM International, West Weerasekara, L. and Wijewickreme, D. 2008. Mobilization
Conshohocken, PA, USA. of Soil Loads on Buried, Polyethylene Natural Gas
Cullin. M.J., Petersen, T.H. and Paris, A. 2015. Corrosion Pipelines Subject to Relative Axial Displacements,
Fatigue Failure of a Gray Cast Iron Water Main, Journal Canadian Geotechnical Journal, 45: 1237–1249.
of Pipeline Systems Engineering and Practice, ASCE, Wijewickreme, D. and Soysa, A. 2016. Stress–Strain
6(2): 05014003. Pattern-based Criterion to Assess Cyclic Shear
Das, S. and Dhar, A.S. 2021a. Nonlinear Time-Dependent Resistance of Soil from Laboratory Element Tests,
Mechanical Behavior of Medium-Density Polyethylene Canadian Geotechnical Journal, 53: 1460–1473.
Pipe Material, Journal of Materials in Civil Engineering, Wijewickreme, D. and Weerasekara, L. 2015. Analytical
ASCE, 33(5): 04021068. Modeling of Field Axial Pullout Tests Performed on
Das, S. and Dhar, A.S. 2021b. Modeling Time-Dependent Buried Extensible Pipes, International Journal of
Behavior of Medium-Density Polyethylene Pipes, Geomechanics, ASCE, 15(2): 04014044.
Journal of Pipeline Systems Engineering and Practice,
ASCE, 12(4): 04021049.
Duncan, J.M. and Seed, R.B. 1986. Compaction-induced
Earth Pressures under K0-Conditions, Journal of
Geotechnical Engineering, ASCE, 112(1): 1-22.
Groves, A. and Wijewickreme, D. 2013. Field Monitoring of
Buried Polyethylene Natural Gas Pipelines Subjected
to Ground Movement, In Proceedings of 66th Canadian
Geotechnical Conference, Geo Montreal 2013,
Montreal, Quebec, Canada, September 29-October 3.
Kreger, S.T., Gifford, D.K., Froggatt, M.E., Soller, B.J. and
Wolfe, M.S. 2006. High Resolution Distributed Strain or
Temperature Measurements in Single-and Multi-mode
Fiber Using Swept-Wavelength Interferometry, Optical
Fiber Sensors, Optical Society of America, Paper
ThE42.
Luna. 2021. Optical Backscatter Reflectometer Model
6100 User Guide, Luna Technologies, Blacksburg, VA,
USA.
Murugathasan, P., Dhar, A.S. and Hawlader, B. 2020. An
Experimental and Numerical Investigation of Pullout
Behavior of Buried Ductile Iron Water Pipes in Sand,
Canadian Journal of Civil Engineering, 48(2): 134–143.
Nova, R. and Hueckel, T. 1981. An Engineering Theory of
Soil Behaviour in Unloading and Reloading,
Meccanica, Springer, 16(3): 136–148.
Performance review of GNSS survey
monitoring spikes in a pipeline geohazard
management program
Doug Dewar and Heidi Manicke
Pembina Pipeline, Calgary, Alberta, Canada
Lisa Monk and Ryan McMahon
Measurement Sciences Inc., Calgary, Alberta, Canada

ABSTRACT

A performance review of a Real Time Kinetic Global Navigation Satellite System survey monitoring program was conducted
for a 15 000 km pipeline system in Western Canada. The data set consisted of 6165 monitoring spikes within 210 slope
polygon monitoring arrays at 85 sites. Monitoring epochs ranged from 2 to 24 per site, with some surveys dating back to
2011. Individual monitoring spikes were assessed for displacement types and slope polygons were classified based on
movement scenarios plus pipeline interactions. The discussion focuses on the advantages and disadvantages of survey
monitoring spikes and the complementary role of other ground and pipeline monitoring techniques. Guidance is provided
on the use of monitoring spikes within a pipeline geohazard management program.

RÉSUMÉ
Un examen du rendement a été effectué sur un programme de surveillance par sondage du système mondial de navigation
par satellite cinétique en temps réel pour un réseau de pipelines de 15 000 km dans l'Ouest Canadien. L'ensemble de
données consistait à 6165 pointes de surveillance dans 210 réseaux de surveillance de polygones de pente sur 85 sites.
Les ères de surveillance variaient de 2 à 20 par site avec des relevés datant depuis 2011. Les pointes de surveillance
individuels ont été évalués pour les types de déplacement, les polygones de pente ont été classés en fonction des
scénarios de mouvement et les interactions avec les pipelines. La discussion porte sur les avantages et les inconvénients
des pointes de surveillance et le rôle complémentaire d'autres techniques de surveillance du sol et des pipelines. Des
conseils sont fournis sur l'utilisation de points de surveillance dans le cadre d'un grand programme de gestion des
géorisques des pipelines.

1 INTRODUCTION NRT- Near Real Time

SID- Single pipeline segment within a polygon
Wang et. al. (2017) concluded that conventional and GNSS SMS- Survey Monitoring Spikes
survey technology have become reliable sources of slope SMP- Survey Monitoring Program
monitoring data for pipeline applications. A review was RoW- Pipeline rights of way, typically 10 to 20 m wide
conducted to determine the effectiveness of survey RTK- Real Time Kinematics
monitoring spikes as part of geohazard management Unk- Unknown
program for a 15 000 km pipeline system. Additionally, data
was reviewed to develop program guidelines to manage
surveys going forward. Previously, monitoring spikes were 3 METHODOLOGY
installed and monitored on a site-by-site basis.
The paper is divided into two parts 1) data results and 3.1 Dataset
review and 2) general and pipeline specific commentary
based on the authors’ experiences and opinions. Note that The dataset is limited to sites within the Western Canadian
45 data attributes were assessed for each survey site; Sedimentary Basin, Rocky Mountains and Interior Plateau
therefore, this paper is a summary of key findings. of Alberta and British Columbia, Canada. Data was from:

• 6165 monitoring spikes, within

2 NOMENCLATURE/TERMINOLOGY • 210 slope polygons, at
• 85 sites.
Landslide terminology follows Cruden and Varnes (1997)
and soil to pipeline interactions are described using Dewar Figure 1 shows the epochs per site. Based on this, there is
(2019). Acronyms/Definitions are: a total of 352 individual site surveys. Mature sites have 4
or greater epochs. Two sites at critical pipeline crossings
SI- Slope inclinometer, including ShapeArrays™ had 22 and 24 epochs due to monthly monitoring to verify
Epoch- A round of survey pipeline FFS prior to, during and after mitigation. Sites are
DoC- Depth of Cover, measured to the top of a pipeline generally limited to active, dormant or relict deep-seated
EA- Engineering Assessment landslides and/or shallow slides/flows within pipeline
FFS- Fitness for Service backfill or disturbed soil. Rates of movement range from
GMP- Geohazard Management Program extremely slow (<16 mm/yr) to slow (<5 m/yr).
GNSS- Global Navigation Satellite System
 1. Data was compiled for all GMP SMS surveys
with greater than a single epoch.
2. Individual site surveys were reviewed for:
a. Displacement types (Section 4.1) by
plotting criteria based on local slope fall
line and reviewing total displacement
vectors visually and in spreadsheets,
b. SMS performance including required
replacements and additions to the
Figure 1: Epochs per survey site. survey arrays, and
c. Displacement rates in the horizontal
and vertical direction. Includes
3.2 Survey maximums, minimums, and average
rates. Only maximum horizontal rates
3.2.1 Survey Technique were reviewed as part of this paper.
3. Individual polygons within sites were interpreted
RTK surveying using GNSS has been the primary method for the movement scenarios (Section 4.4.1) by
utilized for the program. RTK was chosen as these RoWs the corresponding author then reviewed by a
are mostly clear of vegetation, providing an unobstructed contributing author.
sky-view. In addition, RTK has proven to provide consistent 4. Individual polygons were reviewed independent
results, and the time/cost value is significant when of the survey data for indications of confirmed or
compared to other methods of survey, such as short-term likely soil to pipeline interactions (Section 4.4.2)
static or conventional surveys. primarily using ILI IMU and axial strain tool data.
This method is dependent on a strong network of 5. SIDs within polygons were reviewed for
survey control monuments located in stable areas outside indications of overgrown or recent landslide
landslide extents. Monitoring spikes are measured using scarps during field inspections. The data was
geodetic grade GNSS equipment and a rigorous survey compared to monitoring scenarios and
technique developed to mitigate random and systematic interaction types.
errors associated with equipment calibration, atmospheric
and ephemeris errors, and ambiguity resolution.
4 DATA RESULTS AND REVIEW
3.2.2 Monitoring Spikes
4.1 Displacement Types
Survey spikes are 203 mm long, 6 mm diameter, steel
spikes with a 12 mm machined head installed with a 38 mm The following displacement types are used for the study:
diameter washer, as shown in Figure 2.
Table 1. Displacement Types and Counts

Type Criteria Count (%)

No No movements / vectors 4729 (77%)
Movements shown on survey
Drift Clear random errors around 283 (5%)
>25 mm a point with no trends
Nebulous Movement with trends not 127 (2%)
>25 mm attributed to downslope
displacements
Downslope Vector is within +/- 22.5 1026 (17%)
Figure 2: Monitoring Spike >25 mm degrees with fall line of
local slope
Conventional survey geotechnical practice would not
specify the relatively short SMS in favor of much deeper Figure 3 provides examples of clear downslope
installation. An example would be a 1.0 to 1.5 m long, movements at an active spike and no movements with drift
18 mm smooth steel bar or a cemented survey monument. around a static spike.
The SMS is specified due to pipeline ground disturbance
standards that require extensive efforts to verify pipeline
positions for any work below 300 mm depth.

3.3 Analysis Method

The following methodology was followed for data analysis:

 Table 3. Horizontal Displacement Rates

Rate Count (%)

Slow (>1.6 m/yr) 4 (4%)
Very slow 2b1 (>160 mm/yr) 7 (7%)
Very slow 2a1 (>16 mm/yr) 35 (37%)
Extremely slow (<16 mm/yr) 18 (19%)
None (0 mm/yr) 31 (33%)
1
Porter et al. (2022) modified from Cruden and Varnes (1996).

Almost 90% of observed maximum horizontal movements

are below an average 160 mm/yr. Movements above these
average velocities were shallow non-interacting slides as
previously noted.

Figure 3: Example of static versus active SMS. 4.3 Reliability

Of the 6165 monitoring spikes, a total of 1228 (20%) was Sources of error and causes of spikes not functioning
added after the initial epoch. The reasons were: within a +/-25 mm drift as per Section 4.1 include:

• 88 spikes (1%) replaced after unplanned ground • Survey drift (5 to 7%)

disturbance or third party activities, • Bioturbation or Frost heave (<2%)
• 202 spikes (3%) replaced after planned ground • Erosion (rare, <0.01%)
disturbance, and • Third party disturbance due to
• 938 spikes (15%) were added to the existing array unplanned/unauthorized ground disturbance
to better characterize slope movements. driving or vegetation management activities (1%)
• Planned construction/ground disturbance (3%)
Table 2 presents the data for 2390 spikes from mature
surveys with 4 or more epochs: Considering all factors other than ground disturbance, SMS
provide meaningful data to interpret 93% of the time in the
total dataset and 91% in the mature dataset. Note that the
Table 2. Displacement Types and Counts ground disturbance factors are not included as SMS were
mainly disturbed during planned landslide mitigations. Drift,
Type Count (%) when exceeding 25 mm, is often not an issue when you
No Movements 1396 (59%) have mature surveys that have clear indications of
Drift >25 mm 175 (7%) consistent movement at adjacent points (refer to Figure 3).
Nebulous >25 mm 77 (3%)
4.4 Slope Polygon Classification
Downslope >25 mm 742 (31%)
4.4.1 Movement Scenarios
Note that the variations in verified downslope movement
>25 mm (17 to 31%) reflect that the mature surveys Slope polygon movement scenarios were classified using
targeted areas of known movements, whereas many of the the system shown in Figure 4. Results are shown in
newer surveys are being used to verify landslide activity at Table 4.
sites which may be dormant or relict. More recent surveys
reflect a program trend towards SMS rather relying of SI Table 4. Movement Scenarios
technology.
Status Polygon SID Polygon
4.2 Displacement Rates Count Count Percent
No Movements 21 41 13%
Table 3 presents the maximum observed horizontal Sporadic Movements 35 70 22%
displacement rates per survey site. Generally, maximum Localized Movements 31 64 19%
average annual rates of movement of “slow” corresponded Block Movements 19 31 12%
Multiple Movements 27 45 17%
to shallow root reinforced slides not interacting with the Complex Movements 29 39 16%
pipelines. Maximum measured displacements in the Total Sample 159 290 100%
sample are at the same polygon/SID corresponding with a
deep backfill failure over a pipeline:

• Horizontal downslope: 14 031 mm (over 3 years)

• Vertical drop: -5 227 mm
• Vertical rise: +428 mm
 4.4.2 Soil to Pipeline Interactions

Slope interactions (or lack thereof) were classified using

Dewar (2019) based on a review of existing ground and
pipeline monitoring data with a focus on detailed
assessments of pipeline shape and caliper data derived
from inline inspections (Dewar 2017, 2018 and 2020).

Figure 5: Interacting Parallel, Oblique and Perpendicular

landslides (from Dewar 2017).

Table 5. Soil to Pipeline Interactions

Interaction Subtype Polygon SID Percent

Type1 Count Count SIDs
None 56 117 40%
Non- parallel 38 55 19%
Interacting
perpendicular 4 11 4%
oblique 4 5 2%
Exposing perpendicular 1 6 2%
Interacting parallel 48 84 29%
perpendicular 0 0 0%
oblique 8 12 4%
Total 159 290 100%
Sample
1
Subtypes not identified are not reported except interacting
perpendicular.

Note that the 56 “None” corresponds with the 56 “No

Movements” or “Sporadic Movements” in Table 4. The data
shows that most interactions are parallel with some oblique
movements. This is due to the Western Canadian
topography, with pipelines having perpendicular valley
slope crossings. Many of the oblique movements are due
to local topography within major river valleys. It should be
cautioned that this dataset would not be representative of
other geographic regions such as the US Appalachian
Mountains, where most interactions are due to
perpendicular or oblique movements.

4.5 Survey versus Visual Inspection

Table 6 provides a comparison of SMS movement

scenarios versus visual observations of landslide features
during field inspections at each survey site conducted by a
Figure 4: Movement scenarios. qualified geotechnical engineer(s).
Table 6. Movement Scenarios versus Visual Inspection 5 DISCUSSION

Scenario Overgrown Scarps Fresh Scarps 5.1 Accuracy and Reliability

Yes No Yes No
None or Sporadic 26 2 11 17 The horizontal errors around a point of +/- 25 mm was
Localized 17 2 4 15 used. Absolute horizontal accuracy of a RTK solution is +/-
10mm. The additional +/ 15mm takes into consideration
Multiple 16 1 7 10
the relative accuracy between the control network and
Block 13 0 5 8 monitoring spikes and operator experience. Measurement
Complex 18 0 9 9 repeatability among the stable control monuments network
at each site is typically less than +/-10mm.
Data generally shows that most sites have some sort of 5.2 Advantages and Disadvantages of Technique
scarp feature that is identified as overgrown by field
inspectors. This is not surprising as this field observation The main disadvantages of the methods:
along with a review of LiDAR data would trigger the
installation of an SMS array to confirm landslide activity. 1. Delayed Detection: Confirmation of movements
Where there are confirmed ground movements that are typically required >25 mm of movement, whereas
likely below pipeline depth (multiple, block or complex), SI technology may identify movement with as low
44% show indications of recent landslide activity via fresh as 2 to 5 mm of casing deflection.
scarps. This confirms the observations previously reported 2. Shallow versus deep: Often it is difficult to
by Barlow (2000) where up to 75 mm of measured annual distinguish multiple shallow movements versus
ground movement in northern Alberta had no visual complex movements where deep-seated
indications of active ground movement. interacting movements are overlain by surficial
Table 7 compares the visual observations versus slides. Typically, an additional source of data such
known or probable soil to pipeline interactions. as SI’s or ILI IMU is required to confirm deeper
movements.
Table 7. Interaction types versus visual field inspection1 3. Contractor specialization: General legal
surveyors focused on property surveys often have
Interaction Type2 Overgrown Scarps Fresh Scarps difficulty conducting geotechnical work.
Yes No Yes No Specialized survey contractors are required.
Non-interacting 53 3 20 36 4. Horizons and coverage: Encroaching vegetation
Exposing/Interacting 36 3 16 23
and poor satellite coverage may limit accuracy.
1
Note that interaction types are not determined solely by the Often can be supplemented with conventional
survey.
2
Dewar (2019)
surveys (Wang et al. 2017).

Only 41% of confirmed or probable soil to pipeline The main advantages are:
interaction polygons show evidence of ground movement
via fresh scarps. Every confirmed or probable soil to 1. Efforts: significantly less installation efforts than
pipeline interaction polygon had an overgrown scarp other technologies.
identified. 2. Modifications/flexibility: SMS array networks can
be refined with additional spikes, post initial
4.6 SI Comparisons installation.
3. Reliability: SMS do not become blocked or
It is difficult to compare these measurements due to the sheared like SI technologies.
different sources of errors (Mikkelsen 2003) where the SI 4. Range: SMS have an unlimited range when
accuracy is mainly dependent on installation quality/shape and monitoring block/slide movements where the
reading repeatability, whereas the SMS is dependent on the spikes are not destroyed by flows/scarp extension
survey methodology and quality of satellite data. The issue is or other landslide mechanisms. These are rare in
that there is no third technology that can be relied on to the case study as pipelines are generally not
determine which method is correct. Previous unpublished work routed across landslides with rates of movement
by the authors indicates that SMS report between 50% to greater that slow.
200% the horizontal movements of an SI. One of the main 5. Replaceability: SMS can be easily replaced if
complicating factors is whether the upper casing deflection damaged or disturbed.
(Manicke and Dewar 2022) is included in the comparison. 6. Vectors: SMS report displacements in x, y and z
Also, SMS may help identify when an SI does not terminate co-ordinates whereas SIs report displacements
below the lowest slip surface in a landslide. on a horizontal plane. Technologies such as
LiDAR and InSAR report a change in height
resolved based on angle of incidence and change
in length of a target rather than true x, y and z co-
ordinates.
 7. Ground disturbance: pipeline operator standards Table 8 details go forward recommendations for specifying
typically require the exposure of active pipelines monitoring arrays.
prior to work and do not permit drilling within 5 m
of the outside edge of a pipeline for any Table 8. Monitoring array specifications
excavation over 300 mm. For pipelines with deep
cover this may be a project showstopper. There State Array Layout Specification
are many well-known logistical challenges for BI Immediate Threat Geotechnical Specialist
installation and monitoring associated with
budgetary concerns, site access, steep slope
Threat: interactions evident Geotechnical Specialist
work and drilling logistics/break downs. Additional in very slow-moving
pipeline related concerns are documented in landslides. Pipeline
Manicke and Dewar (2022). approaching FFS limit.
8. Pipeline position: Most pipelines interacting or
potentially interacting with landslides are installed
Potential Threat: interactions Surveyor specifies with
conventionally in trenches with DoCs ranging evident in extremely slow- geotechnical review prior to
from 0.6 to 1.5 m. As the depth of the slip surface moving landslides. Not installation
becomes deeper, surface measurements either approaching FFS limit.
from SMS and/or LiDAR/InSAR, generally
become more relevant in defining potential Hazard: No interactions Surveyor with geotechnical
interactions, as demonstrated in Figure 6 below. evident but overgrown input if local, multiple, block or
landslide feature present. complex movements detected
Activity unknown.
Condition: Slope present Not Applicable
with no indications of
interactions or slope
movements

5.5 SI Replacement

An adjacent monitoring spike and a point in the top of the

SI casing should be installed to continue monitoring once
the SI has been blocked by movements. Install an SA to
further increase the SI install life and SMS comparison time
(Manicke and Dewar 2022). As the depth of the slip
surface(s) has been established, an SI replacement should
not be required.
Figure 6: Depth of cover versus slip surface depth in
conventionally trenched pipelines from Manicke and Dewar 5.6 Monitoring Methods and Frequencies
(2022).
Recommended SMS monitoring frequencies are shown in
5.3 RTK versus Short Static GNSS Table 9.
As illustrated in Wang et. al. (2017), typically, a RTK Table 9. Monitoring frequencies based on threat
solution is half as accurate as a short-term static solution.
Horizontal accuracy of a RTK solution is +/- 10mm, the State (as per Method Frequency
accuracy of a short-term static solution is +/- 5mm. For this Table 7)
SMP[DD1], the quality of the observations obtained using RTK Immediate NRT NRT reviewed on a frequency as
do not undermine the findings of the survey. Repetitive Threat GNSS1 per EA
observations over multiple epochs have provided results in
exhibiting movement trends on active slopes. Given the Threat SMS Monthly to Semi-Annually as
strain demand limits of most pipelines, gaining 5 to 10 mm confirmed by EA
of accuracy is not required. Potential Threat SMS Annually as confirmed by EA

5.4 SMP Placement/Density Hazard SMS Annually for 3 years. If no

movements – resurvey “on
trigger”2 afterwards.
Monitoring spikes are typically placed in rows
perpendicular to the pipeline in rows of 2 to 3 spikes per Condition None Not Applicable
RoW. Rows are generally no more than 30 m apart, in 1
As per Bracic and McMahon (2020)
addition, the areas above and below scarps and at slope 2
Trigger could be an event such as a 1:50 year rainfall/flood event
breaks are monitored. The initial layout is set by the or a threshold exceedance.
surveyor with refinements to the survey array being
provided with geotechnical input at most survey sites.
5.7 Data Presentation 9 REFERENCES

For interpretation, the authors recommend the following be Babcock, J., Dewar, D, Webster, J, And Lich, T 2020. Deer
incorporated into any plots: Mountain Case Study: Integration of Pipe and Ground
Monitoring Data with Historical Information to Develop
1. Vectors at the actual reported horizontal a Landslide Management Plan. Proceedings 13th
displacement on plots. Excessive displacements International Pipeline Conference, Calgary.
should be noted with a red color and a reduced Barlow, J.P. 2000. Slope Movement Patterns in Young
scale. Valley Slopes in Northern Alberta, Canada. Landslides
2. Vertical displacements annotated beside the point in Research, Theory and Practice, Thomas Telford,
with a sign convention for drop (-) or rise (+). London.
3. Excel tables are provided with all data for Bracic, J. and McMahon, R. 2020. The Use of Remote
interpretation. Real-Time GNSS to Monitor a Pipeline in an Active
Landslide. Proceedings 13th International Pipeline
5.8 Near Real Time GNSS Conference, Calgary.
Cruden, D.M. and Varnes, D.J. 1996. Landslide Types and
Surface displacements can be tracked near real time as Processes. Chapter 3, Landslides Investigation and
reliably as SI slip surface displacements. Bracic and Mitigation. Special Report 247. Transportation
McMahon (2020) have demonstrated the viability of near Research Board. National Academy Press
real time GNSS survey monitoring in pipeline applications. Washington, D.C. USA: 36-75.
Dewar, D., Tong, A., and McClarty, E. 2017. Assessing and
Monitoring the Impacts of Very Slow-Moving Deep-
6 CONCLUSIONS Seated Landslides on Pipelines. Proceedings 75th
Canadian Geotechnical Conference, Ottawa, Ontario.
The following conclusions are given: Dewar, D., ElSeify, M., Van Boven, G., Bjorn, P, and
Bruce, N. 2018. Operational Experiences with Axial
1. RTK survey monitoring spikes are providing Strain Inline Inspection Tools. Proceedings 12th
meaningful results greater than 91 percent of the International Pipeline Conference, Calgary.
time within a conservative error +/- 25 mm of Dewar, D. 2019. A Suggested Soil and/or Rock to Pipeline
horizontal displacement. Landslide Interaction Classification System.
2. Survey monitoring spikes do not appear to be Proceedings 77th Canadian Geotechnical Conference,
significantly affected by frost heave. At most, 2% St. John’s.
of the sample could have been impacted by frost Dewar, D, 2020. Incorporating Inline Inspection Internal
heave. Measurement Unit Data Analysis into Integrity
3. Visual field inspections should not be considered Management Programs. Proceedings 13th
as a reliable method of determining landslide International Pipeline Conference, Calgary.
activity, nor predicting if there is soil to pipeline Manicke, H. and Dewar, D. 2022. Performance Review of
interactions. Slope Inclinometers in a Pipeline Geohazard
Management Program. Proceedings 80th Canadian
Geotechnical Conference, Calgary.
7 FURTHER RESEARCH Mikkelsen, E., 2003. Advances in Inclinometer Data
Analysis. Symposium on Field Measurements in
The data review presented could be refined once more Geomechanics, FMGM 2003. Oslo, Norway.
surveys reach a mature state with 4 or more epochs. Porter, M., VanHove, J. and Barlow, P. 2022. Analysis of
Additionally, data may be reviewed in more detail for Dynamic System Risks where Pipelines Cross Slow-
individual SID’s rather than by site or polygon. Moving Landslides. Proceedings 14th International
Pipeline Conference, Calgary.
Wang, YY., West, D., Dewar, D., Hart, J., Mckenzie-
8 ACKNOWLEDGEMENTS Johnson, A. and Gray, D. 2016. Management of
Ground Movement Hazards for Pipelines. J.I.P. report
Anne-Marie Gagnon for abstract translation. Beverly Grey prepared by Center for Reliable Energy Systems,
for proofreading services. Count Von Heinrich (Heinz) and Dublin, Ohio.
Baroness Genoa-Krimpli-Chorizo experienced shortened
walks during the preparation of this paper. Draven Moss
sorted through site specific visual inspections and reviewed
interactions and movement scenarios.
Investigation of buried pipes under axial tension
and compression loadings
RezaA1, KuwataY2, DharA3
1Memorial University of Newfoundland, St. John’s, NL, Canada
2Kobe University, Kobe, Japan
3Memorial University of Newfoundland, St. John’s, NL, Canada

ABSTRACT
Buried pipes exposed to ground movements can experience soil loads in various scenarios. When the ground movement
is parallel to the pipe axis, it experiences axial/longitudinal forces. The longitudinal force can be tension or compression
depending upon the locations of the pipe to the moving ground. A downstream pipe segment experiences compression,
while an upstream pipe experiences tension. Previous studies focused on investigating the pipe responses to tensile force
using axial pullout tests. The compression loading, as well as loading-reloading effects, were not investigated well. The
loading-reloading is expected when the pipeline is subjected to intermittent ground movements. This study investigates
high-density polyethylene (HDPE) and ductile iron pipes subjected to tension and compression loadings and loading-
reloading cycles. For tension loading, a pipe buried in a soil box was pulled axially. For compression loading, the soil box
with the buried pipe was placed on a shaking table. Then, the shaking table was moved to push the pipe against a reaction
frame. The test results revealed that the maximum axial force needed to mobilize the interface frictional resistance over
the entire pipe sample length is the same for tension and compression loadings for the ductile iron pipe. However, for the
HDPE pipe, the corresponding axial force was significantly higher for compression than for tension. The higher axial force
for the flexible HDPE pipe during compression might be due to the increase in diameter. For the ductile iron and HDPE
pipes, the axial force for mobilization of interface frictional resistance reduced with the loading cycle, likely due to the
arching effect after releasing the interface bonding in the first cycle. The method recommended in the current design
guidelines significantly underestimated the maximum pulling forces observed during the tests.

RÉSUMÉ
Les tuyaux enterrés exposés aux mouvements du sol peuvent subir des charges de sol dans divers scénarios. Lorsque le
mouvement du sol est parallèle à l'axe du tuyau, il subit des forces axiales/longitudinales. La force longitudinale peut être
une tension ou une compression selon les emplacements du tuyau par rapport au sol en mouvement. Un segment de
tuyau en aval subit une compression, tandis qu'un tuyau en amont subit une tension. Des études antérieures se sont
concentrées sur l'étude des réponses des tuyaux à la force de traction à l'aide d'essais d'arrachement axial. Le chargement
de compression, ainsi que les effets de chargement-rechargement, n'ont pas été bien étudiés. Le chargement-
rechargement est prévu lorsque la canalisation est soumise à des mouvements de terrain intermittents. Cette étude porte
sur des tuyaux en polyéthylène haute densité (HDPE) et en fonte ductile soumis à des charges de traction et de
compression et à des cycles de chargement-rechargement. Pour le chargement en traction, un tuyau enfoui dans une
caisse de sol a été tiré axialement. Pour le chargement en compression, la caisse de sol avec le tuyau enterré a été placée
sur une table vibrante. Ensuite, la table vibrante a été déplacée pour pousser le tuyau contre un cadre de réaction. Les
résultats des tests ont révélé que la force axiale maximale nécessaire pour mobiliser la résistance de frottement de
l'interface sur toute la longueur de l'échantillon de tuyau est la même pour les charges de traction et de compression pour
le tuyau en fonte ductile. Cependant, pour le tuyau en HDPE, la force axiale correspondante était significativement plus
élevée pour la compression que pour la traction. La force axiale plus élevée pour le tuyau flexible en HDPE pendant la
compression pourrait être due à l'augmentation du diamètre. Pour les tuyaux en fonte ductile et en HDPE, la force axiale
pour la mobilisation de la résistance de frottement de l'interface a diminué avec le cycle de chargement, probablement en
raison de l'effet de voûte après avoir relâché la liaison de l'interface au premier cycle. La méthode recommandée dans les
directives de conception actuelles sous-estimait considérablement les forces de traction maximales observées lors des
essais.

1 INTRODUCTION displacements due to permanent ground movements

caused by earthquakes or slope instability (O’Rourke et al.
Buried pipelines are important structures that carry 1995; Rajani et al. 1995; Chan and Wong 2004). For
essential commodities, such as water, wastewater, oil, and example, when pipelines cross river valleys, they are often
gas. The overall responses of the buried pipe to various laid parallel to the slope of the valley wall. If the valley wall
loading are highly dependent on its interactions with the becomes unstable, the pipeline will be subjected to
surrounding soil (Psyrras and Sextos 2018; Tsinidis et al. longitudinal forces along its length. This force can cause
2019). It is common for pipelines to experience axial strains in the pipe, potentially leading to its failure. A pipe
segment located toward the downstream boundary of the identify the influence of tensile and compressive loadings
moving ground experiences axial compression, while the on the pipes and (2) to examine the effect of loading-
pipe segment upstream of that boundary experiences reloading on axial soil resistance. Predictions made using
tension. The pipeline can experience significant tension the current design guidelines are compared with the
strains at the crest and compression strains at the base of measured data.
a slope. Pipes beyond the boundaries of the moving
ground are usually fixed in the stable ground but can
experience wall strains, depending on the magnitude of the 2 TEST METHODOLOGY
axial force.
Most of the previous laboratory studies examining the 2.1 Test Facility
axial pipe–soil interactions were conducted under
monotonic tensile loading of pipes, referred to as an axial All the tests were performed in a specially designed tank of
pullout test (Paulin et al. 1998; Scarpelli et al. 2003; 1.6 m long, 0.84 m wide, and 1.02 m deep at Kobe
Anderson et al. 2004; Weerasekara and Wijewickrame University in Japan. The tank was fabricated of thick steel
2008; Wijewickreme et al. 2009; Murugathasan et al. 2021; plates and braced using steel angles, which provided
Reza et al. 2023). The pipelines subjected to monotonic additional rigidity to the boundary walls. The longer side of
compression loading have received much less attention the box had two circular holes on opposite walls to allow
and have not been thoroughly investigated. Weidlich and the pipes to pass in and out of the box. The relative
Achmus (2008), Bilgin and Stewart (2009), and Sheil et al. displacement between the pipe and the soil tank was
(2018) performed repeated cycles of axial pull and push measured using two displacement transducers installed,
tests, with the compression loading resulting from the one at each side between the pipe and test cell boundary
backward movement of the pipes (push test) after the wall. A load cell with a capacity of 10 kN was fixed at one
forward movement (pull test). During the first forward axial end of the pipe to measure the reaction load, which is the
displacement phase of the cyclic tests, it was assumed that frictional force corresponding to the relative displacement
a grain orientation was established around the pipe. When between the pipe and the soil. A total of four tests were
the direction of the movement was changed, the orientation conducted, which included two loading-reloading
may have been disrupted, leading to changes in the stress compression tests and two tensile axial pullout tests. These
and density states at the interface between the soil and the tests were performed on ductile iron (DI) and high-density
pipe (Weidlich and Achmus 2008). These changes in the polyethylene (HDPE) pipes with outer diameters of 93 and
stress state of interface soil can result in reduced friction 89.5 mm, respectively. The Standard Dimension Ratio
forces during the axial push tests. Despite this, the results (SDR), which is the ratio of the pipe outside diameter to
of a tensile loading (i.e., pullout) test are sometimes used wall thickness, was 15.5 and 10 for the DI and HDPE pipes,
in the analysis of pipes under compression loading. respectively. The full length of both pipes was above 2.2 m,
Apart from that, loading and following reloadings in with 1.6 m long section of pipe being in contact with the
underground pipelines are expected in real-field situations sand inside the box. Table 1 provides a summary of the
when the pipeline is subjected to intermittent ground test program undertaken.
movements. Intermittent ground movement refers to The sand used in the tests was uniformly graded river
changes in soil stress or strain that occur repeatedly over sand. The key physical properties of which are given in
time rather than continuously. This type of ground Table 2.
movement can be caused by a variety of factors, including
changes in the water table, temperature fluctuations, and 2.2 Pipe Installation
human activities such as construction or excavation. This
repetitive nature of the loading affects the soil response The backfill soil and the pipe were placed in the test box in
and eventually causes fatigue damage to the pipeline, as a few steps. Before inserting the pipe into the testing tank,
reported by Cullin et al. (2015). Weidlich and Achmus the sand was placed in the box in layers, levelled, and
(2008) reported that the influence of repetitive axial compacted to achieve a dense state through kneading.
displacement of buried pipes in sand was dependent on the Once the level of sand reached the invert elevation of the
relative density of the backfill. A progressive increase in pipe, the pipe was positioned parallel to the longer direction
axial soil resistance was observed in loose sand, whereas of the test cell on a compacted soil surface, passing
a drop in resistance was recorded in compacted dense through the holes in the box. The backfilling and
sand. These findings were also supported by Sheil et al. compaction process then continued in the same manner
(2018). However, the widely used design guidelines for until the desired depth of cover was reached. For all tests
calculating the axial resistance of buried pipelines (ALA conducted in this study, the pipes were buried at a depth of
2005; PRCI 2009; JGA 2013) do not consider the influence 635 mm, resulting in a ratio of soil cover (measured from
of repetitive axial displacement cycles, probably due to a springline) to pipe diameter (H/D) of 6.8 and 7 for DI and
lack of extensive study on the behaviour. HDPE pipe tests, respectively. The density measurements
This paper presents an investigation of the behaviour of the compacted soil were taken at four different locations
of buried pipes subjected to axial tension and compression at the pipe springline level using 98.125 cm 3 steel cylinder
loading. Two types of pipes were used for the tests: ductile collars. The average dry unit weight of the compacted sand
iron (DI) and high-density polyethylene (HDPE), commonly ranged from 13.7 to 14.2 kN/m3. The measurements
used pipes for water distribution systems in Canada and indicated that consistent compaction was achieved during
Japan. The main objectives of the present study are: (1) to backfilling for each test, and the compacted backfill had an
average relative compaction of roughly 93% of the shaking table used in this experiment consists of a 2 m × 3
standard Proctor density. The moisture content ranged m horizontal table having a single displacement
from 6.0 to 6.8%, except in Test 4, where the average (horizontal) degree of the freedom movement. A schematic
moisture content was 9.0%. of the test facility is shown in Figure 1. The test pipe is
The sand and pipe were removed from the box after passed through holes in the walls of the soil box and
each test, and the pipe installation procedure was repeated extends beyond the test cell boundaries. As the shaking
for the next test. table moves horizontally toward a reaction wall, the soil
tank also moves in the same direction. Once one end of the
pipe hits the reaction wall, further displacement of the
Accelerometer 1600 (a) shaking table causes compression on the pipe and relative
displacement between the pipe and the surrounding soil.
Displacement Displacement After certain relative displacements, the shaking table
transducer transducer moves in the opposite direction when the pipe is released
Tank
1020

from any load. The table then reverses its direction again,
and pipe hits the reaction wall, resulting in the pipes being
Pipeline Reaction wall reloaded. A total of three cycles of loading were selected
for each test. The shaking table movement rate was
Load cell
constant (3000 mm/min) for all the cycles.
In the second type of experiment, pipe response to a
Shaking table tensile force is simulated by pulling a pipe through a fixed
soil mass in the test box, as shown in Figure 2. During the
pullout of the pipe, the surrounding soil offers resistance to
(b)
the pipe movements resulting in tension loading on the
1600 pipe. In tension loading tests, the pulling rates of pipe were
approximately 2.5‒5.5 mm/min that was manually
controlled.
Tank
840

Reaction wall Table 1. Summary of the test program

D

Load cell Type 1 Type 2

Test type (loading-reloading (monotonic axial
Shaking table compression test) pullout test)

Figure 1. Test facility for loading-reloading compression Test no. Test 1 Test 2 Test 3 Test 4
test: (a) side view; (b) plan view (unit: mm)
Ductile Ductile
Pipe material HDPE HDPE
iron iron
Displacement Total pipe length
1600 transducer 2.26 2.23 2.26 2.23
(m)
Buried pipe
length, 1.6
L (m)
840

Pipe diameter,
93 89.5 93 89.5
D (mm)
Pipe wall thickness,
Load cell 6 9 6 9
Displacement
Displacement t (mm)
transducer
transducer
Pipe self-weight,
Figure 2. Test facility for tensile axial pullout test: plan 40.1 8.8 40.1 8.8
w (kg)
view (unit: mm)
Burial depth,
0.635
H (m)
2.3 Test Procedure
H/D 6.8 7.0 6.8 7.0
The test program was planned to investigate the axial
behaviour of buried pipes in sand under moist conditions. Test speed 2.6 5.35
3000 3000
Two types of tests were performed to measure the axial (mm/min) (avg.) (avg.)
soil resistance for the pipes. The first type of experiment Type 1 Type 2
involved compressive loading-reloading tests in which the Test type (loading-reloading (monotonic axial
test tank was placed and fixed on a shaking table. The compression test) pullout test)
 However, the post-peak degradation was not observed in
Test no. Test 1 Test 2 Test 3 Test 4
Test 1 under compression loading of DI pipe, potentially
Average dry unit due to the increase of soil density by the particle
weight of soil, γ̅d 14.24 14.08 13.72 13.67 rearrangement.
(kN/m3)
6
Water content, w 6.8 5.98 6.45 9.0 (a) DI pipe

Relative 5
compaction, R 95 94 91 91
(%)

Frictional resistance (kN)

4

Table 2. Backfill sand properties 3

Property Values
2
Standard Proctor max. dry density (kN/m3) 15.04
Cycle 1 Compression
1 Cycle 2 (Test 1)
Optimum water content (%) 11 Cycle 3
Axial pullout (Test 3)
0
Fines content (%) 2.0
0 10 20 30 40 50
Sand content (%) 98.0 Relative displacement (mm)

Gravel content (%) 0 6

(b) PE pipe
Mean particle size, D50 (mm) 0.33
5
Coefficient of uniformity, Cu 1.57
Frictional resistance (kN)

4
Coefficient of curvature, Cc 1.09

3 TEST RESULTS
2
3.1 Force‒Displacement Responses
Cycle 1
1 Cycle 2 Compression
The force–displacement responses for both DI and PE pipe Cycle 3
(Test 2)
segments, measured from compression and tension Axial pullout (Test 4)
loading, are plotted in Figure 3. The force is termed herein 0
as the frictional resistance, which is equivalent to the 0 10 20 30 40 50
applied force. As seen in Figure 3, during both tension and Relative displacement (mm)
compression loading (Cycle 1 for compression loading),
the mobilized frictional resistance increases with the Figure 3. Load–displacement response in loading-
movement (relative displacement) until reaching a peak reloading and axial pullout tests: (a) DI pipe; and (b) PE
value, beyond which the frictional resistance is stabilized. pipe
In general, the frictional resistance is mobilized along the
pipe length with relative displacement, increasing the
pulling force (hence the frictional resistance). Once the For the ductile iron pipe (Figure 3a), the peak value of
maximum frictional resistance is mobilized over the entire the frictional resistance is reached at the value of ~5 kN at
pipe length, the overall frictional resistance is stabilized, the displacement of 3.5 mm, and then the frictional
and the pipe segment moves as a rigid body translation (a resistance was slightly increased at a much lower rate. The
constant translation without further elongation). The rigid increase of frictional resistance in the first cycle might be
body translation of the trailing end can be monitored using due to the densification of the soil during pipe movement.
the displacement transducer at that end. Beyond the point The initial peak value from first compression cycle match
of full mobilization, the mobilized shear stresses can be with the peak value from the tensile loading test at the
reduced due to the wear and tear of sand grain asperities, same level of displacement (i.e., 3.5 mm). Thus, the
which leads to particle rearrangement under these responses of the ductile iron pipe under compression and
displacements (Johnson et al. 1987; Foray et al. 1998). tension loading are expected to be the same. However, for
the HDPE pipe (Figure 3b), the frictional resistance during
compression is significantly higher than the resistance a burial depth of 0.58 m. The guidelines recommend a soil
during tension. The higher resistance during compression spring constant of 6000 kN/m3 regardless of the soil
and the lower resistance during tension is likely due to stiffness, and this value was used to calculate the relative
changes in pipe diameter during axial loading of the flexible displacement at τcr, which was found to be 0.82 mm. In the
HDPE pipe. Reza and Dhar (2021) and Reza et al. (2023) ALA and PRCI guidelines, a constant value of 3 mm is
revealed that the diameter of flexible polyethylene pipe is recommended for the pipes in dense sand.
reduced during axial pulling that can reduce the contact
normal stress and hence the interface frictional resistance.
While the diameter change during axial pushing (a) DI pipe
(compression test) was not investigated earlier, it can
cause an increase in pipe diameter, increasing the contact
normal stress and the frictional resistance.
Figure 3 also reveals that the peak frictional resistance
is reduced during repeated cycles of compression loading.
During the first loading cycle, the frictional resistance
between the pipe and the surrounding soil is the highest. PRCI (2009)
As the pipeline experiences subsequent loading cycles, the
axial frictional resistance drops. This phenomenon can be
attributed to the release of the peak shearing resistance in
the first cycle. In the subsequent cycles, the residual JGA (2013)
shearing resistance was reduced. Similar mechanisms
ALA (2005)
were observed by Weidlich and Achmus (2008) and Sheil
et al. (2018) for dense sand backfill. The breakage of the
structure of interface soil might has contributed to the
reduction of shearing resistance. Besides, cyclic loading
can cause particle breakage in the surrounding soil, which
also affects the circumferential distribution of normal
contact stress on the pipe (Saberi et al. 2022).

3.2 Comparison with Current Guidelines (b) PE pipe

The current industry-practiced design guidelines in North

America (e.g., ALA 2005; PRCI 2009) recommend using
Eq. (1) to estimate the maximum axial force acting per unit
length of pipes buried in cohesionless soils. Eq. (1) is
based on the normal stresses acting on the pipe and the
frictional characteristics of the soil–pipe interface, where
the normal stress is taken as the arithmetic mean of the
vertical overburden pressure (γH), and the at-rest lateral
earth pressure (K0γH) at the pipe springline level. PRCI (2009)

1+K0
Pu = πDγH ( ) tanδ [1] JGA (2013)
2
ALA (2005)
where Pu = the maximum axial force on the pipe (i.e., soil
resistance); γ = unit weight of soil; H = burial depth
measured from the ground surface to pipe springline; L =
pipe length; D = pipe outer diameter; K0 = at-rest lateral
earth pressure coefficient; and δ = interface friction angle
between the pipe and the surrounding soil. Eq. (1) of PRCI
guidelines recommends a range of effective coefficients of Figure 4. Comparison of frictional resistances with guidelines :
lateral earth pressure K which may vary from the value for (a) DI pipe; and (b) PE pipe
at-rest conditions for loose soil (K0) to values as high as 2
for dense dilative soils. According to the guidelines, the Figure 4 shows the mobilized axial resistances per unit
value of δ can be estimated as δ = fϕ′, where ϕ′ is the friction length of the pipe measured during the tests and calculated
angle of the backfill material and f is an interface friction based on the guidelines followed in North America and
reduction factor. On the other hand, the Japan Gas Japan. Test results from compression loading (Tests 1 and
Association (JGA 2013) defined the critical shear stress 2) are presented against relative displacement from the
(τcr) at full mobilized shear strength (i.e., at the time of slip initial values. It reveals that the pipes have permanent
occurrence) as 15 kN/m2 for a burial depth of 1.8 m magnitudes of relative displacement after each cycle,
(measured from the ground surface to pipe crown). In the confirming that structures of interface soil are different in
current study, a value of τcr = 4.90 kN/m2 was calculated for each cycle, resulting in different peak pulling resistance.
For the rough DI and smooth PE sections of the test pipe, induced stress on the pipe could be insignificant. By
the values of f = 0.8 and 0.6 have been adopted, contrast, the normalized resistance obtained by Weidlich
respectively, and a peak friction angle of 40° has been and Achmus (2008) is higher than the current test results,
selected for the test sand for the pulling resistance which could be due to the lower burial depths (0.42 m) and
calculation. A value of 0.5 has been calculated for K0 using higher soil density, resulting in greater soil dilation and
the relationship with the soil’s Poisson ratio (ν): K0 = ν/(1−ν) compaction-induced effects on the pipes.
(where ν is assumed to be 0.33, which falls within the
typical values for dense sand). Figure 4 shows that the 2.5
peak resistance values predicted by ALA and JGA are Test 1 (w/o self-weight_DI)
Test 2 (PE)
significantly lower than those measured during the tests. Test 3 (w/o self-weight_DI)
The underestimation is less for the PRCI guideline, which Test 4 (PE)
takes into account the normal stress increase on the pipe 2 Steel pipe (Wijewickreme et al. 2009)
surface due to the volumetric expansion (i.e., dilation) of DI pipe (Murugathasan et al. 2021)
the surrounding soil during shearing (as adopted from Steel pipe (Sheil et al. 2018)
Wijewickreme et al. 2009). Note that the difference HDPE-coated pipe (Weidlich and Achmus 2008)
between the prediction based on the PRCI guidelines and 1.5

Pu/πDLγH
the result of the test is more significant for the PE pipe
when subjected to compression loading. This is because,
under compression loading, increasing the diameter of the
PE pipe results in a higher level of frictional resistance. In 1
addition, soil compaction during backfilling can further
increase the interface contact stress, resulting in higher
frictional resistance for the pipe (Reza and Dhar 2023).
However, none of the guidelines accounts for the soil
0.5
compaction effect.

3.3 Effects of Loading Cycles

0
The influence of the number of loading cycles on peak axial 0 1 2 3 4
resistance is presented in Figure 5. The contribution of self-
Cycle number
weight to frictional resistance for DI pipe is calculated and
subtracted from the peak response in Figure 5 (denoted as Figure 5. Effects of loading cycles on normalized peak
w/o in the figure). The weight of the pipe is considered resistance
negligible for the lightweight PE pipe. To assist comparison
of the results for pipes with different diameters, lengths,
and burial depths, the peak resistance is normalized into a 3.4 Influence of Loading Speeds
dimensionless term by πDLγH. The normalized peak axial
resistances obtained from previous studies (Weidlich and In the present tests, no significant effects of the loading
Achmus 2008; Wijewickreme et al. 2009; Sheil et al. 2018; speed on the soil resistance were observed. The shaking
Murugathasan et al. 2021) are also included in Figure 5. table, used in Tests 1‒2, was moved at a constant rate of
The results indicate that a large frictional force degradation 3000 mm/min, while the pipe pulling speed in Tests 3‒4
occurs during the initial cycles, and the rate of degradation was varied between 2.5‒5.5 mm/min. These speeds were
reduces during subsequent loading cycles. categorized as low-speed cases based on the previous
Weidlich and Achmus (2008) observed that the axial study conducted by the second author (Kuwata et al.
friction force stabilizes after 5‒8 cycles and introduced a 2018). The study by Kuwata et al. (2018) reported a
degradation factor (CF) as the ratio of the maximum friction detailed investigation on the velocity dependency of soil
force in the first cycle to the residual friction force after five friction in the pipe’s axial direction. They categorized the
cycles. The CF-value increases with soil density, but for the shaking table speed as low (1200 mm/min), medium
loosest state of sand, the friction force increases with (6000‒12000 mm/min), and high (24000‒36000 mm/min)
loading cycles instead of decreasing. This indicates that if speed cases. For the sandy soil with low water content,
the backfill soil density is too high, it may have the opposite which was similar to the sand used in the current study, the
(beneficial) effect of what was intended, and loose sand frictional resistance was higher for the low-speed tests than
backfill may have negative effects during loading-reloading for the medium- and high-speed tests. They argued that in
cycles. It would be interesting to conduct additional the high-speed tests, the shear deformation occurred so
experiments exploring this behaviour with varying soil rapidly that the soil particles could not be rearranged in the
densities. short time, and the frictional force was not transferred to
Moreover, Figure 5 demonstrates a noticeable the pipe. These findings were consistent with the
difference in normalized peak resistance between the observations made by Takada (1976), where the dynamic
current test results and those reported by others (i.e., frictional force was found to be less than the static frictional
Weidlich and Achmus 2008 and Sheil et al. 2018). Sheil et force in sandy soil. Therefore, the present tests,
al. (2018) conducted their tests at deep burial depths of considering them low-speed, provided an upper-bound
1.355 m, where the effects of soil dilation and compaction- measurement of the axial resistance.
4 CONCLUSIONS 6 REFERENCES

This paper presents four experimental tests performed on ALA (American Lifelines Alliance). 2005. Guidelines for the
ductile iron (DI) and high-density polyethylene (HDPE) design of buried steel pipe. Washington, DC and Reston,
pipes buried in moist sand under axial loading conditions. VA, USA: Federal Emergency Management Agency
Two types of tests were performed: loading-reloading (FEMA) and American Society of Civil Engineers
compression tests and monotonic tension loading tests. (ASCE).
The soil frictional resistances and relative displacements Anderson, C., Wijewickreme, D., Ventura, C. and Mitchell,
between the pipe and the soil tank were measured during A. 2004. Full-scale laboratory testing of buried
the tests. The test data are presented and compared with polyethylene gas distribution pipelines subject to lateral
the current design guidelines followed in North America ground displacements. In Proc., 13th World Conference
and in Japan. The emphasis is placed on investigating the on Earthquake Engineering, Aug. 1‒6, Vancouver, B.C.
effect of the loading-reloading cycles on soil resistance. Canadian Association of Earthquake Engineering. Paper
The major findings from the study are as follows. 1543.
Bilgin, Ö. and Stewart, H.E. 2009. Design guidelines for
• The maximum axial force required to mobilize the polyethylene pipe interface shear resistance. Journal of
interface frictional resistance is the same for tension Geotechnical and Geoenvironmental Engineering,
and compression loadings for DI pipes. However, 135(6): 809–818.
for HDPE pipes, the axial force required is Chan, P.D. and Wong, R.C. 2004. Performance evaluation
significantly higher for compression than for of a buried steel pipe in a moving slope: a case study.
tension. Canadian Geotechnical Journal, 41(5): 894–907.
Cullin, M.J., Petersen, T.H. and Paris, A. 2015. Corrosion
• The industry-standard ALA and JGA guidelines fatigue failure of a gray cast iron water main. Journal of
significantly underestimate the peak axial Pipeline Systems Engineering and Practice, 6(2):
resistance of pipelines in soil compared to the PRCI 05014003.
guideline. However, the PRCI guideline also Foray, P., Balachowski, L. and Colliat, J.L. 1998. Bearing
underestimates the frictional resistance of PE pipes. capacity of model piles driven into overconsolidated
sands. Canadian Geotechnical Journal, 35(2): 374–385.
• Axial resistance of pipelines in compacted soil
JGA (Japan Gas Association). 2013. Seismic design
decreases significantly after the first loading cycle,
guideline of medium/low pressure gas pipeline, (in
which is attributed to the release of interface
Japanese).
bonding and rearrangement of particles.
Johnston, I.W., Lam, T.S.K. and Williams, A.F. 1987.
• The frictional resistance degradation continues Constant normal stiffness direct shear testing for
during the three cycles applied in the tests, and the socketed pile design in weak rock. Geotechnique, 37(1):
rate of degradation reduces in the last cycle. 83–89.
Compared with the previous studies, it can be Kuwata, Y., Inase, T. and Sawada, S. 2018. Experimental
expected that the frictional resistance would study on the velocity dependency of axial soil spring for
become constant after some additional loading buried pipe by shaking table (part 2). Journal of Japan
cycles. Society of Civil Engineers, Ser. A1 (Structural
Engineering & Earthquake Engineering (SE/EE)), 74(4):
• The results of the loading-reloading tests have 568–577.
revealed that laying pipes in dense sand backfills Murugathasan, P., Dhar, A.S. and Hawlader, B.C. 2021. An
may have a beneficial effect on the reaction experimental and numerical investigation of pullout
developing on the pipe. behaviour of ductile iron water pipes buried in sand.
• No effects of loading speed on the measured Canadian Journal of Civil Engineering, 48(2): 134–143.
resistances were observed during the tests for the O’Rourke, M.J., Liu, X. and Flores-Berrones, R. 1995.
selected speed range. Steel pipe wrinkling due to longitudinal permanent
ground deformation. Journal of Transportation
Engineering, 121(5): 443–451.
5 ACKNOWLEDGEMENTS Paulin, M.J., Phillips, R., Clark, J.I., Trigg, A. and Konuk, I.
1998. A full-scale investigation into pipeline–soil
The laboratory tests were conducted using the facilities at interaction. In Proc., International Pipeline Conference,
Kobe University. The authors acknowledge the support of IPC1998, ASME, Calgary, Alberta, June 7–11, pp. 779–
undergraduate and graduate students of Kuwata Lab at 788.
Kobe University during the laboratory tests. Prof. Sawada PRCI (Pipeline Research Council International). 2009.
at Disaster Prevention Research Institute of Kyoto Guidelines for constructing natural gas and liquid
University was the host collaborator who provided office hydrocarbon pipelines through areas prone to landslides
space and administrative support during the visit of and subsidence hazards. Report prepared for the
Canadian researchers. The financial support of this Design, Material, and Construction committee, Chantilly,
research was provided by the Natural Science and VA, USA: Pipeline Research Council International, Inc.
Engineering Research Council of Canada through its Psyrras, N.K. and Sextos, A.G. 2018. Safety of buried steel
Alliance International Catalyst grants to the third author. natural gas pipelines under earthquake-induced ground
 shaking: A review. Soil Dynamics and Earthquake
Engineering, 106: 254–277.
Rajani, B.B., Robertson, P.K. and Morgenstern, N.R. 1995.
Simplified design methods for pipelines subject to
transverse and longitudinal soil movements. Canadian
Geotechnical Journal, 32(2): 309–323.
Reza, A. and Dhar, A.S. 2021. Axial Pullout behavior of
Buried Medium Density Polyethylene Gas Distribution
Pipes. International Journal of Geomechanics, 21(7):
04021120.
Reza, A., Dhar, A.S. and Rahman, M. 2023. Strain
assessment of polyethylene pipes in dense sand
subjected to axial displacements. Geosynthetics
International, (Accepted, ahead of print).
Reza, A. and Dhar, A.S. 2023. Finite-element modelling of
axial movements of polyethylene pipes in dense sand,
(under review).
Saberi, M., Annan, C.D. and Sheil, B.B. 2022. An efficient
numerical approach for simulating soil-pipe interaction
behaviour under cyclic loading. Computers and
Geotechnics, 146: 104666.
Scarpelli, G., Sakellariadi, E. and Furlani, G. 2003.
Evaluation of soil–pipeline longitudinal interaction forces.
Rivista Italiana di Geotecnica, 4(3): 24–41.
Sheil, B.B., Martin, C.M., Byrne, B.W., Plant, M., Williams,
K. and Coyne, D. 2018. Full-scale laboratory testing of a
buried pipeline in sand subjected to cyclic axial
displacements. Géotechnique, 68(8): 684–698.
Takada, S. 1976. Restoring force characteristics of buried
pipeline and ground system and earthquake response
analysis. Journal of Japan Society of Civil Engineers,
249: 1–11, (in Japanese).
Tsinidis, G., Di Sarno, L., Sextos, A. and Furtner, P. 2019.
A critical review on the vulnerability assessment of
natural gas pipelines subjected to seismic wave
propagation. Part 2: Pipe analysis aspects. Tunnelling
and Underground Space Technology, 92: 103056.
Weerasekara, L. and Wijewickreme, D. 2008. Mobilisation
of soil loads on buried polyethylene natural gas pipelines
subject to relative axial displacements. Canadian
Geotechnical Journal, 45(9): 1237–1249.
Weidlich, I. and Achmus, M. 2008. Measurement of normal
pressures and friction forces acting on buried pipes
subjected to cyclic axial displacements in laboratory
experiments. Geotechnical Testing Journal, 31(4): 334–
343.
Wijewickreme, D., Karimian, H. and Honegger, D. 2009.
Response of buried steel pipelines subjected to relative
axial soil movement. Canadian Geotechnical Journal,
46(7): 735–752.
Performance review of strain gauges in a
pipeline geohazard management program
Heidi Manicke, Doug Dewar and Joel Babcock
Pembina Pipeline, Calgary, Alberta, Canada

ABSTRACT
Vibrating Wire Strain Gauges (VWSG) have a long precedence in
geotechnical pipeline instrumentation due to their ruggedness, ease of
installation and lack of ongoing calibration requirements. This performance review of VWSG in a pipeline Geohazard
Management Program in Western Canada examines data collected from over 500 individual gauges at 21 different sites.
Individual gauges were assessed for range, installation quality and failure mode. Installation configurations were examined
for trends and efficacy of monitoring pipelines during strain relief and long-term monitoring. The discussion will focus on
the advantages and limitations of VWSG in a complementary role to other monitoring technologies. Advice will be provided
for installation, monitoring frequencies and data interpretation. Additional topics will include VWSG limitations and sources
of error.

RÉSUMÉ
Les Jauge de contrainte a corde vibrant ont une longue préséance en instrumentation géotechnique de pipeline en raison
de leur robustesse, la facilité de l’installation et aux faibles exigences d’étalonnage. L’examen de performance de
l’utilisation de la jauge de contrainte vibrante par le programme de gestion des géorisques des pipelines de l’ouest
Canadien examine les donnes saisies de 500 jauges de contraintes individuelles provenant de 21 différents sites. Les
Jauges individuelles ont été évaluées selon leur portée, la qualité de l’installation et le mode de défaillance. Les
configurations d’installation ont été évaluées selon les tendances et l’efficacité de la surveillance des pipelines lors de la
décharge de traction et la surveillance a long terme des performances des pipelines. La discussion portera sur les
avantages et les limites des jauges de contrainte dans un rôle complémentaire aux différentes technologies de
surveillance. Des conseils seront fournis pour l’installation, la fréquence de surveillance et l’interprétation des données.
Des thèmes supplémentaires incluront les limitations de la technologie de Jauge de contrainte a fil vibrant et les sources
d’erreur.

1 INTRODUCTION FFS – Fitness for service, refers to a pipeline operating

within design and operating limits set out by applicable
Strain gauges have been employed in the pipeline industry regulatory codes, industry guidelines and operator
since the 1980s (Bukovansky and Major, 2002). This paper standards
updates Dewar et al. (2016), which is based on 200 gauges Flange gauge – flanges along the entire length of the VW
making 70 installs at 2 sites, with a much larger dataset of spot welded to pipeline (refer to Figure 1)
539 gauges making 87 installs (56 of 87 having a Gauge position – Measured clockwise looking in
redundant set of gauges) at 21 sites. Work presented in downstream flow direction
this paper is part of a program review to develop a strain GEO – Caliper module mounted on inline inspection tool to
gauge standard. Previously, strain gauges were installed measure internal circumferential shape of pipeline
ad-hoc on a site-by-site basis and not according to any GMP – Geohazard Management Program
standard or specifications. ILI – Inline inspection tool - pipeline smart pigs
The paper is divided into three parts 1) data results ILI IMU GEO – combo tool with ILI and GEO tools
and review, 2) case studies, and 3) general and pipeline- IMU – Inertial measurement unit
specific commentary based on the authors’ experiences Install – primary and redundant sets (if present) of gauges
and opinions. at one location along pipeline
SI – Slope inclinometer, including ShapeArrays™
SG – Strain Gauge
2 NOMENCLATURE & TERMINOLOGY SG Response – defined with tension positive and
compression negative
Landslide terminology will be as per Cruden and Varnes, RoW – Pipeline(s) rights of way typically 10 to 20 m wide
(1996). Soil to pipeline interaction terminology follows με – microstrains, expressed in parts per million
Dewar (2019). Acronyms /Definitions are: Unk – Unknown
VW– Vibrating wire
Block gauge – VW suspended between two independent VWSG – Vibrating wire SG
blocks spot welded to a pipeline (refer to Figure 1)
CSCC – Circumferential stress corrosion cracking
DoC – Depth of cover, from top of pipeline
 Not all records could be considered complete because raw
data sets were not always available and often there were
gaps within the actual data recorded, particularly for
installation data. Given that 539 records/datasets with 34
individual data fields per record were reviewed, only the
key findings are presented herein.

4 RESULTS AND DISCUSSION

4.1 Install Configuration and Observations

The SGs in the dataset were installed in two main

Figure 1. Block-type VWSG (left) and flange-type VWSG configurations 3T, with gauges at the 12, 3 and 9 o’clock
(right) positions around the pipe, or 4C with gauges at 12, 3, 6,
and 9 o’clock positions around the pipe (Figure 2). These
are referred to as “installs” and are either installed with a
3 METHODOLOGY single “set” of gauges at each position or with a primary
and redundant set of SGs at each clock position.
3.1 Dataset Overview

The dataset is limited to SG installations at 21 geohazard

sites within the Western Canadian Sedimentary Basin,
Rocky Mountains and Interior Plateau of Alberta and British
Columbia, Canada. The dataset is 539 individual SGs,
composed of the following:
• 114 flange-type spot-welded VWSG;
• 323 block-type spot-welded VWSG;
• 9 Biaxial Resistance Rosettes bonded to
pipeline;
• 64 Axial resistance gauges mounted on spot
welded foil flanges; and,
• 29 Unk, no installation records.

The SGs were installed for:

• 61% Long-term landslide monitoring. SGs were
often installed during planned pipeline integrity Figure 2. Typical pipe configurations at clock positions in
digs that intersect active or potentially active the direction of pipeline product flow. Redundant gauge
landslides. The digs were scheduled outside of sets not shown in this figure for 3T2 or 3C2 installs.
GMP activities to verify and repair pipeline
anomalies detected during ILI activities. The breakdown of install configuration in the dataset was
• 20% Short-term monitoring during strain relief. as follows:
• 18% Short-term strain relief then transitioned to • 4C – 17%
long-term landslide monitoring. • 4C2 – 72%
• 1% Settlement monitoring post highway • 3T – 10%
embankment construction and welded pressure • 3T2 – 1%
containment sleeve installation.
The 4C configuration, and to a lesser extent the 3T
Most of the SGs (75%) were installed on pipelines that configuration, allows for easy interpretation of the pipeline
were backfilled following installation, while 19% were behavior from the SG data. Axial tension, compression and
installed on surface pipeline segments and the remaining bending can be assessed without complex calculations.
6% were installed on pipeline segments that were left Note that 3C configuration is recommended by some
daylighted in a ditch for the duration of the monitoring SG suppliers; however, as the pipeline must often be
period. undermined for installation or inspection, there is minimal
additional effort required to install a 4C configuration.
3.2 Analysis Parameters VWSG cannot be oriented around the axis of the pipeline
due to pipeline curvature. Resistance type gauges may be
Parameters for data analysis included: multidirectional.
• Installation – Configuration, type, range Redundant sets are installed for almost all long-term
• Performance – Max./min. recorded strain per SG monitoring applications, as the SG install and materials are
• Qualitative review of data – Install strain signature only a small portion of any excavation costs. In a 4C
configuration, should one gauge become defunct, it is still
possible to understand the trends in pipeline movement.
However, with less than 4 gauges, more inference must be
made, and with less than 3, it becomes significantly more
difficult to interpret trends. Installing a redundant set of SGs Axial Compression
allows for data verification in the case of erroneous
readings or SG failure.

4.2 Recorded Strain Vs. Manufacturers Spec

Axial Tension
Manufacturers specified total gauge range in this study was
from 2400 to 5000 με. Most of the dataset consisted of SGs
that came centralized from the manufacturer or were
centralized prior to installation to accommodate equal
amounts of tension and compression;
• 82% - Centralized
• 1.5% Compression Vertical Bending
• 1.5% Tension
• 15% - Not Documented

Maximum delta tension recorded was 2460 με on a

centralized flange-type centralized SG with a total range of
+/- 1200 με in either tension or compression (i.e. 2400 με Horizontal Bending
total). The range of the SG was exceeded by over 200%
Figure 3. Typical Trend Signatures
but the SG continued to function. This pipeline was
subsequently strain relieved and the SG recorded a
recovery of 1210 με. While the SG remained functional, the
set was removed and replaced with new SGs. The three Table 1. Breakdown of Trends
highest recorded strains came from the same site, which is
discussed in case study Location 3. Trend Percentage (%) Number of
Installs
4.3 Observed Trends Axial Tension 21% 18
Axial Compression 6% 5
The following SG responses are documented in Dewar et Vertical Bending 2% 2
al. (2016) for background pipeline effects:
Composite Trends 6% 5
• Strain inversely tracks temperature changes;
No Significant Changes1 55% 48
therefore, a temperature increase results in a
reduction in tension. Insufficient Data 10% 9
• Strain tracks pressure changes; therefore, a
1
Less than 100 με change, likely related to temperature
pressure increase results in an increase in fluctuations and changes in pressure.
tension. This is caused by the pipeline being
restrained by the surrounding soil. There were no incidents of strictly horizontal bending
• Bending strains; two opposing gauges have observed, as there was always some component of axial
opposite strains around a neutral axis. As an movement, so these are categorized under “composite”
example, dig related settlement is characterized where there were two or more signatures detected.
by compression of the top, tension at the bottom The maximum strain recorded at an individual SG was
and no significant change at the sides of the plotted based its position on the pipeline in a violin plot
pipeline. (Figure 4) to show the distribution of strain for a particular
• Axial strains would have the same reaction of all gauge position.
the gauges either in tension or compression. In general, most of the data plotted above the neutral
axis, indicating a trend towards observing axial tension.
Trends were evaluated for 83 SG Installs. Disregarding However, the dataset is biased, as in general the SGs were
obvious backfill signatures, most trends recorded by the installed to monitor this type of movement. Axial tension
SG sets could be placed in one of the categories seen in tends to be easier to detect than compression, which
Figure 3. The breakdown of movement is described in occurs in more discreet areas and is often transient, as
Table 1. seen in Location 4 in the case studies.
 4.4 Reliability

Most pipelines compressional or tensional strain demand

limits are in the range of 2500 to 3000 με based on the
pipelines specified minimum yield strength. Table 2
provides a summary of the operational status of the VWSG
in this study.

Table 2. Summary of VWSG functionality

Description Flange-Type Block-Type

Total Number 114 328
Operating 27 (24%) 208 (63%)
Disconnected/Unk. 69 (60%) 55 (17%)
Replaced 9 (8%) 14 (4%)
Defunct 9 (8%) 51 (16%)
Figure 4. Frequency of recorded maximum reported strain
per SG position
While there is a higher percentage of operating block-type
Compressive strains were typically found at the top of gauges, this is because a number of flange-type gauges
pipeline (12 o’clock) and tensile strains typically at the have been disconnected or monitoring has been
bottom of pipeline (6 o’clock) as a relic of pipeline backfill, discontinued and is not a reflection of gauge reliability.
referred to here as a “standard backfill” signature, with 66% Furthermore, block-type gauges appear to become defunct
of the buried SGs showing the standard backfill signature. twice as frequently as flange-type gauges. Of the 51
The largest initial strain would be seen immediately defunct block SGs, 4 were damaged during re-excavation,
following backfill; however, as most pipeline construction 10 during backfill and for 37 the cause is unknown. Of the
occurs during the winter, there were ongoing increases in 9 defunct flange SGs, 1 was over-ranged, 3 had unknown
strain settlements throughout the spring. reasons for failure and 5 were damaged from backfill – all
Hoop stress or internal pressure was generally not the defunct flange SGs were at Location 3 in the case
considered in this analysis; however, in numerous studies.
occasions pipeline depressurization was observed due to
pipelines being taken permanently or temporarily out of
service. 5 CASE STUDIES
In two cases where known external forces had been put
on the cables of the SGs, increases in compression were 5.1 Location 1
observed that correlate to the timing of when the cables
were pulled. These are discussed further in Section 5, A pipeline crosses perpendicular to a 780 m wide, relict,
Case Studies for Location 1, and Location 3. deep-seated landslide. A 180 m cut-out and replacement
This was repeated in a controlled environment where were completed due to the presence of CSCC. Flange
two flange-type VWSGs were affixed to a plate, one as a gauges (Install SG-00) were installed in a bell hole prior to
test gauge and the other as a control. The temperature was daylighting and pipeline cut-out/tie-in. Initial backfill of the
kept consistent within 0.5°C for both gauges throughout the bell hole introduced the standard backfill signature of ~500
test. Weight was applied to the cable of the test gauge and µε vertical bending, and -250 µε axial was recorded
the changes in frequency of the VWSG of both were following the cut-out. Once the section of pipeline was
recorded and are plotted in Figure 5 below. As force was inserted, four SG installs were installed along the new
applied to the cable of the test gauge, increases in section of pressurized pipeline. Standard backfill
compressive strain were observed. signatures were observed at all installs. Since backfill,
small ongoing increases in axial tension have been
observed at all sets with some sets showing slight
horizontal bending. The axial tension is attributed to
oblique landslide movement.
Following the completion of construction activities on
site, an automated monitoring station was installed and
temporarily secured using T-bars driven into the ground as
the -30°C temperatures would not allow for concrete to
cure as a base for the monitoring station. With spring thaw,
the monitoring station began to tip over. The VWSGs at
SG-00 were procured on a short timeline and the cables
were too short to allow for cable slack. As the monitoring
Figure 5. Cable-pulling plate test results station tipped over, a ~300 µε increase in compression
strain was recorded. Most of this strain was immediately
relieved when the monitoring station was righted and the
tension on the cables decreased, indicating that 5.3 Location 3
compression can be induced on the SGs by pulling on the
cables. The pipeline traverses a 70 m long, active landslide that
moves obliquely to the pipeline. Two generations of SGs
5.2 Location 2 and SIs have been installed at this site; the original set in
2016 and replacements in 2021.
Babcock et al. (2020) describes a 4 km pipeline segment The SG install locations were chosen following a
which traverses a 7.3 km long, complex landslide, stress analysis based on ILI IMU survey and ground survey
characterized as active translational/spread-type results. The original flange VWSGs were installed during
movement, with smaller translational, rotational, and flow- the strain relief and were intended for long-term landslide
type movements within the larger slide. Multiple SG installs monitoring. After backfilling, the 12 o’clock and 6 o’clock
have been completed throughout this segment as gauges at one install showed over 2000 µε in compression
described in Table 3 below. and tension respectively, which exceeded the
manufacturer’s specified range of +/- 1200 µε. The original
Table 3. Location 2 SG Installs SIs installed in 2016 sheared off after baseline and before
follow-up readings could be taken, so accurate
Install Date Purpose Basis for location selection movement/displacement data was not available. The
(# of Installs) pinched point of the Sis was within the backfill and poor
backfill likely contributed to the premature failure of the SIs
2014/2015 Long-term IMU strains, adjacent
landslide pipeline failures, cut-outs
and excessive backfill strains on the pipeline measured by
(4 VWSG)
monitoring the SGs.
The spring following installation, the 6 o’clock gauge
2019/2020 Short-term strain IMU strains, LSM
(11 VWSG) relief transitioned anomalies showed a sudden movement to compression, which
to long-term continued until the pipeline was daylighted in 2021. Initially
landslide it was thought this gauge had become defunct or detached
monitoring from the pipeline but following inspection of the cables after
2021 Long-term Surface pipeline daylighting in 2021, it was determined that ground
(8 Resistance landslide horizontal bends (stress movement had pushed the fiberglass protective coating
SG) monitoring analysis) to verify 1.5 m downslope from the SGs (Figure 7) along the pipe
structure is operating and induced compression on the SGs by pulling on the
within design parameters cables, some of which had torn apart from the gauge.

No appreciable strain change has been observed in three

of the four 2014/2015 installs, including in locations where
adjacent SIs have since recorded ground movements of up
to 10 mm. One installation has recorded 150 to 250 µε
increase in compression 140 m downslope of an SI which
recorded ground movements up to 20 mm over the same
period (Figure 6).

Figure 7. Location 3 daylighting of installed SGs

5.4 Location 4

The pipeline was installed in a conventional trench

Figure 6. Schematic of one SG install within Location 2 which parallels the slope fall-line. Stress relief excavation
was performed between July and September 2020, after
The 2021 resistance SGs were installed with an ILI survey identified anomalous landslide induced
thermocouples along a new surface pipeline segment to bending strains at the toe of the slope. The landslide
thermally compensate for pipeline temperature effects on movement consists of the reactivation of an older landslide
the SGs. Following pipeline tie-in and upon startup of the feature, with tension cracks forming at the main scarp and
cathodic protection system, the thermocouples failed and a toe bulge at the toe of the slope. The extent of the
had to be re-installed by electrically isolating them from the landslide was approximately 90 m along the length of the
pipeline. Accumulations up to 500 µε have been recorded pipeline.
since installation and displacement sensors installed at the Following stress relief, four SG installs were placed
same bends have recorded horizontal displacements of for long-term landslide monitoring near the main scarp,
30 to 90 mm and vertical settlements up to 150 mm. midslope, and at a non-formed (landslide-induced)
overbend and formed (constructed) sagbend near the toe.
Over a two-year monitoring period, SGs near the scarp
recorded increases in tension up to 390 µε and SGs at the tools being run on as low as weekly frequency for critical
toe recorded increased compression up to 420 µε, FFS issues. For the economic operation of a pipeline in the
coinciding with up to 85 mm observed ShapeArray long-term, it is usually only done during the implementation
displacement. Discrete element modelling completed to of immediate risk control measures. SGs can be used to
assess pipe-soil interaction and set instrument monitoring confirm changes in strain to complement ILI IMU. Also,
thresholds indicated that the location of peak pipeline SGs work well for immediate indications of movement and
bending strains near the slope toe SGs was expected to interaction to complement near real-time SIs and GPS
change over time as the bends were roped through the survey monitoring hubs. For long-term monitoring, ILI IMU,
landslide. SG, SI, and survey can be coordinated and combined to
verify pipeline FFS.
Most landslide-related FFS incidents and failures in
6 DISCUSSION Canada are compressive failures at the toe of a landslide
because of wrinkles/buckles at/or adjacent to constructed
6.1 General sag and overbends. These failures tend to be exacerbated
by unfavorable pipeline geometry within zones of
As discussed in Dewar (2016), the primary limitations of accumulation/compression in landslides. This is because
SGs are: most of the soil to pipeline interactions in Canada are the
1. Limited coverage in relation to the total length of Interacting Parallel type (Dewar 2019, refer to Figure 8).
pipeline being monitored as VWSGs only measure a This is based on the authors’ experiences, as there is no
very small area on a pipeline. Wang et al. (2016) notes publicly available definitive work characterizing soil to
that structural fiber optic cables may be a viable pipeline interactions by interaction type for pipeline
technology in the future. At this time, the authors are incidents. Murray and Guthrie (2016) provide an example
not aware of any successful implementation of the of a localized compressive failure located at the toe of an
technology on an operating pipeline. The most recent active landslide at an Horizontal Directional Drill entry
example of an attempted experiment is documented in point.
Murray et al. (2022). While SGs can identify bending and compression, it is
2. Strains measured are only changes in strain and do unlikely that the SGs will be placed at the location of peak
not include previous induced strains from pipeline strain as presented at Location 4 in the case studies.
manufacture, construction, operation or outside Additionally, if the SGs were placed on the actual section
forces. Operating and seasonal variations in of pipeline that wrinkles, it would fail at some point during
temperature and pressure must also be accounted for deformations. The operator would not be able to verify the
when interpreting data. Spot welded VWSGs have the actual mode of failure without significant investigation and
advantage of having the same thermal modulus as the potentially dangerous ground disturbance.
pipeline whereas resistance type SGs often require
additional calculations to compensate for different
moduli.

Therefore, SGs cannot be solely relied on to determine the

FFS of a pipeline. Ground monitoring including LiDAR,
InSAR, survey or SI do not provide actual information on
the FFS of a pipeline (Dewar 2017). ILI IMU technology
allows for the interpretation of pipeline shapes and
changes in shape along with pipeline anomalies
(wrinkles/buckles/cracking) to assess potential interactions
with landslides. Hart et.al. (2019) provides a detailed
explanation of the technology and Dewar (2020) provides
a framework for using ILI IMU as part of a GMP. Figure 8: Interacting Parallel, Oblique, and Perpendicular
The main blind spot for ILI IMU is that it cannot detect landslide (from Dewar 2019)
pure tensional extension, nor can it detect gentle roping
often caused by interacting movements perpendicular to a Katebi et al. 2019 concluded that SGs showed no response
pipeline for single or multiple runs as the strain or changes in active parallel to pipeline landslides because “the loading
in strain may be below any detection criteria. Axial Strain due to slow landslides increases over time as the
ILI technology (Wang et al. 2017 and Dewar 2018) allows displacements accumulate until a maximum frictional
for the measurement of the elastic stresses but currently interface capacities reached at this stage, the soil slides
does not have widespread adoption. While axial strain over the pipe as the interface becomes plastic and as a
technology is proven, the implementation of axial strain result, the soil loading remains constant” (p.1788). An
data into GMPs is still developing. alternate explanation is that the SGs are located within a
Regardless, of the ILI technology used, its main slide block and have enough lateral restraint not to
limitation is that it is run on a prescribed frequency of 5 to experience significant strains as demonstrated in Location
10 years for normal pipeline integrity related concerns. 2 in Section 5: Case Studies. Dewar (2017) explains that
Rarely is an ILI schedule frequency reduced for most pipeline deformations are concentrated at areas of
geotechnical concerns, but the authors have seen ILI IMU differential movement associated with deep-seated
movements, as opposed to areas of cumulative movement 3. Monitoring soil to pipeline interactions along with
typical of pipelines interacting with extremely slow-moving other pipe monitoring technologies where strain
earth flows. demand limits can be established.
Where interactions are oblique or perpendicular, SGs 4. Monitoring of pipeline adjacent to welded
may be more useful in helping monitor and determine the pressure containment sleeves where there is a
FFS of a pipeline. This is because failure in these scenarios potential for settlement to create stress risers.
is often tensile as noted by incidents referenced in USDOT
(2022) and detailed in USDOT (2022a). Combined SG and Gauges should be specified and installed:
ILI IMU bending strain data may be used along with 1. Gauge type and range:
methods presented by Liu et al. (2022) or using finite a. VW flange type gauges,
element/discrete element models. These failures often b. +/- 2500 με range,
exploit flawed girth welds in vintage pipelines or at 2. Install type:
undermatched heat affected zone in modern (2010+) high a. Short-term strain relief only: 3T
strength pipelines as detailed in Wang et al. (2016). (perpendicular or oblique interactions) or 4C
(parallel interactions).
6.2 Strain relief transitioning to long-term monitoring b. Long-term monitoring: 4C2 or 3T2 to avoid
Recommendations backfill induced strains if 360° inspection of
the pipeline is not required.
For strain relief, SGs should be installed in bell holes at key 3. Install Specifications:
locations then reburied prior to the actual strain relief. a. SGs should be oriented with the cable end
Wires should be run vertically up from the top of the pointing downslope to avoid being bent
pipeline to allow for re-exposure during the actual strain backwards by ground movement.
relief. This allows for complete assessment of the b. The cables should be housed in protective
effectiveness of the strain relief in the SGs in over short conduit and either routed directly up or along
pipeline segments. Following strain relief, the SGs can be the pipe with consideration given to cable
reburied to switch to long-term monitoring. location during future re-excavation, as well
as impacts from ground movement.
6.3 Program Recommendations c. Approximately 1 to 3 m of cable slack should
be placed under rock guard at the pipeline to
55% of SG installs report no significant changes. A portion accommodate ground movement without the
of these are installed at sites where no changes in strain cables and gauges being put under tension.
are detected, but are within a known larger interacting d. Careful consideration should be given to the
landslide. Some of the SGs are located at sites with re-coating to minimize displacement or
confirmed active, but not interacting, landslides. Others are interaction with ground movement (i.e. wraps
confirming the performance of a previous mitigation. Some “shingled” to resist movement).
SGs have been placed at locations with a low probability of 4. Install records:
actual soil to pipeline interactions where alternative, more a. Installation method, photos and other
cost-effective methods would have been sufficient to verify pertinent details should be documented for
the lack of landslide interactions. This is mainly a product future reference.
of the ad-hoc, site-specific SG scoping that has previously b. Readings taken during installation and
occurred. Additionally, there is the cost of reading and backfill must be recorded and saved. Often
maintaining the SGs. Once a monitoring device is installed, readings during installation are manual.
there is always the tendency to continue reading it, 5. Data collection method:
regardless of the quality of data provided. a. Manual: not recommended.
The following recommendations are given for b. Manual download datalogger: landslides with
specifications operators/consultants using SGs as part of extremely slow rates of movement.
their GMP to optimize reliability and effectiveness based on c. Near real-time monitoring: very slow to slow
the results of this study. Gauges should be specified when: rates of movement, or any movement that
1. Conducting strain relief (as per Section 6.2). can be characterized by acceleration in rates
2. Performing an integrity dig activity where: of movement.
a. There are clear indications of ground 6. Sampling frequency: 12 hours, noon and midnight
movement below the invert of any pipeline, or hottest/coldest time of day,
b. There are clear indications of interactions in 7. Data download frequency for manual dataloggers
ILI IMU or axial strain data, based on maximum accumulation of strain other
c. There is a ground movement with an than changes related to temperature or pressure,
unknown activity where there are pipeline a. < +/- 50 με – annually,
features that could be caused by soil to pipe b. +/- 50 to 100 με – twice a year,
interactions such as double dents, buckles, c. > 100 με – frequency determined by detailed
wrinkles and CSCC, and assessment but recommend near-real time
d. There is a potential landslide zone of tension monitoring; and
including potential interacting perpendicular 8. Trench backfill procedures should:
or oblique landslides. a. Minimize the potential for install damage, and
 b. Minimize settlements to avoid inducing Cruden, D.M. and Varnes, D.J. 1996. Landslide Types and
excessive backfill strains on the pipeline. Processes. Chapter 3, Landslides Investigation and
Mitigation. Special Report 247. Transportation
Research Board. National Academy Press
7 CONCLUSIONS Washington, D.C. USA: 36-75.
Dewar, D., Tong, A., McClarty, E. and Van Boven, G. 2016.
The following conclusions are given: Technical and Operational Guidelines When Using
• SGs are a complementary monitoring tool amongst Strain Gauges to Monitor Pipelines in Slow Moving
other monitoring techniques that can provide regular Landslides. Proceedings 11th International Pipeline
information about impacts on a pipeline; however, their Conference, Calgary.
limitations should be understood. Dewar, D., Tong, A., and McClarty, E. 2017. Assessing and
• SGs that are monitored in near real-time can Monitoring the Impacts of Very Slow Moving Deep-
complement other near real-time ground monitoring to Seated Landslides on Pipelines. Proceedings 75th
provide alerts for landslide reactivations and/or Canadian Geotechnical Conference, Ottawa, Ontario.
accelerations. Dewar, D. 2019. A Suggested Soil and/or Rock to Pipeline
• SGs should not be relied on as the sole source of Landslide Interaction Classification System.
information of monitoring soil to pipeline interactions Proceedings 77th Canadian Geotechnical
based on their limited coverage. Conference, St. John’s.
• SGs can be used to complement ILI IMU GEO and Dewar, D, 2020. Incorporating Inline Inspection Internal
Axial Strain technologies for pipeline monitoring. Measurement Unit Data Analysis into Integrity
• For strain relief, SGs should be used to confirm Management Programs. Proceedings 13th
expected strain responses during work and to assess International Pipeline Conference, Calgary.
the relative effectiveness of the strain relief. Dewar, D., ElSeify, M., Van Boven, G., Bjorn, P, and Bruce,
• SGs are more useful in zones of tension than zones of N. 2018. Operational Experiences with Axial Strain
compression due to failures in compressions being Inline Inspection Tools. Proceedings of the 11th
localized buckles and wrinkles that are difficult to International Pipeline Conference, Calgary.
target and that SGs cannot effectively assess. Katebi, M., Maghoul, P. and Blatz, J. 2019. Numerical
• SGs do not provide a reliable indication of FFS. The Analysis if Pipeline Response to Slow Landslides:
exact residual, construction stress, movement and Case Study” Canadian Geotechnical Journal,
thermal related stress prior to install are unknown and 56:1779-1788.
can only be estimated. Hart, J.D., Czyz, J.A., and Zulfiqar, N. 2019. Review of
Pipeline Inertial Surveying for Ground Movement-
Induced Deformations. Proceedings of the
8 FURTHER RESEARCH AND WORK Conference on Asset Integrity Management – Pipeline
Integrity Management under Geohazard Conditions,
Additional work to characterize SG signatures, failure AIM-PIMG2019-1009, Houston.
mechanisms and correlating to ground movement or other Liu, B., Wang, YY, and Chex, X. 2022. Application of
causes. Strain-Based Assessment in Support of Operational
and Mitigation Decisions. Proceedings 13th
International Pipeline Conference, Calgary.
9 ACKNOWLEDGEMENTS Murray, C.M. and Guthrie, R 2016. 16TAN North
Saskatchewan River Crossing – Geotechnical
Anne-Marie Gagnon for abstract translation. Beverly Grey Investigation Report (Final). Report to Husky Energy.
for proofreading and formatting services. Count Von Murray, C.M., Navjeeb, A., Jailitian, E., Onwude, L. and
Heinrich (Heinz) and Baroness Genoa-Krimpli-Chorizo Hossain, T. 2022. High Fidelity Distributed Fiber Optic
missed numerous walks during the preparation of this Sensing for Landslide Detection Proceedings 13th
paper. However, Mittens the cat enjoyed many desk naps International Pipeline Conference, Calgary.
in the company of her human during this time. United States Department of Transportation, Pipeline and
Hazardous Materials Safety Administration (USDOT)
2022. Pipeline Safety: Potential for Damage to
10 REFERENCES Pipeline Facilities Caused by Earth Movement and
Other Geological Hazards. Pipeline and Hazardous
Babcock, J., Dewar, D, Webster, J, And Lich, T 2020. Deer Materials Safety Administration updated Bulletin.
Mountain Case Study: Integration of Pipe and Ground United States Department of Transportation, Safety Board
Monitoring Data with Historical Information to Develop 2022a. Enbridge Inc. Natural Gas Pipeline Rupture.
a Landslide Management Plan. Proceedings 13th Pipeline Investigation Report: PIR-22/01.
International Pipeline Conference, Calgary. Wang, YY., West, D., Dewar, D., Hart, J., McKenzie-
Bukovansky, M.B. and Major, G. 2002. Twenty Years of Johnson, A. and Gray, D. 2016. Management of
Monitoring Pipelines in Landslides. Proceedings of the Ground Movement Hazards for Pipelines. J.I.P. report
First European Conference on Landslides, Rybar, J., prepared by Center for Reliable Energy Systems,
Stemberk. J. and Wagners, P. (Eds). A.A Balkema, Dublin, Ohio.
Lisse. p.507 to 516.
 Tuesday, October 3, 2023

NUMERICAL MODELS I
Performance of DSM walls on shoring
excavation subjected to influence of nearby
high building loads and rapid drawdown in
the lower mainland, Vancouver, BC
Sounik Banerjee Ph.D. & Yasser Abdelghany Ph.D., P.Eng., PMP
EXP Services Inc., Burnaby, British Columbia, Canada

ABSTRACT
The use of deep soil mixing (DSM) as a method of shoring for excavations is a common practice in engineering. This study
presents a finite element 2D model in Plaxis to analyze the support of a 4.3 m excavation for a new high-rise development
using a DSM wall reinforced up to a depth of 10m, while the rest remains without reinforcement up to another 22 m below
ground level. The design of the DSM wall aims to protect the excavation against ground collapse from the self-weight and
the presence of a 16-storey existing high-rise building. To consolidate the ground, stone columns were constructed up to
the less permeable soil layers for soil densification, and a preload height of 3.5m was applied. The DSM wall was
numerically designed with respect to the superstructure, consolidation settlement for the soft clay layer, and the rapid
drawdown during the excavation process. Additionally, the horizontal deflection and vertical settlement of both the
excavation wall and the high-rise building were analyzed based on different water content for DSM cement mixing with two
target layers viz., the loose sand and clayey silt. The study provides valuable insights into the design of DSM walls for
excavations, taking into consideration various factors that can affect the stability of the structure.

RÉSUMÉ
L'utilisation du mélange de sol profond (MSP) comme méthode d'étayage pour les excavations est une pratique courante
dans l'ingénierie. Cette étude présente un modèle 2D à éléments finis dans Plaxis pour analyser le soutien d'une
excavation de 4,3 m pour un nouveau développement en hauteur à l'aide d'un mur DSM renforcé jusqu'à une profondeur
de 10 m, tandis que le reste reste demeure sans renforcement jusqu'à 22 m sous le niveau du sol. La conception du mur
DSM vise à protéger l'excavation contre l'effondrement du sol dû au poids propre et à la présence d'un immeuble de
grande hauteur de 16 étages présent depuis environ 50 ans. Pour consolider le sol, des colonnes de pierre ont été
construites jusqu'aux couches de sol les moins perméables pour densifier le sol, et une hauteur de précharge de 3,5 m a
été appliquée. Le mur DSM a été conçu numériquement en ce qui concerne la superstructure, le tassement de
consolidation pour la couche d'argile molle, et l'abaissement rapide pendant le processus d'excavation. En outre, la
déflexion horizontale et le tassement vertical du mur d'excavation et de l'immeuble de grande hauteur ont été analysés.
L'étude fournit des indications précieuses sur la conception des parois DSM pour les excavations, en tenant compte de
divers facteurs susceptibles d'affecter la stabilité de la structure.

Keywords: Soil Mechanics, Numerical Modeling, Deep Soil Mixing, Geohazards

1 INTRODUCTION and silts. Yapage et al. 2014 studied the effect of soft
ground of the performance of DSM walls.
Deep soil mixing (DSM) is a popular ground improvement Other studies have focused on the use of quality control
technique used to enhance soil properties and increase soil measures during DSM wall construction. A study by
strength. DSM walls have been used in numerous civil Ishibashi et al. (2017) investigated the use of a continuous
engineering projects for excavation support, foundation mixing method for DSM walls to improve quality control and
construction, and environmental remediation. The use of ensure the uniform distribution of the binder. The study
DSM walls can help to reduce permeability, increase found that the continuous mixing method was effective in
stability, and prevent seepage in soil conditions where achieving consistent binder distribution and that the
conventional construction methods are not feasible. resulting DSM walls had higher strength and durability
However, DSM walls can also face challenges in certain compared to those constructed using conventional
soil conditions, such as those with seepage, ground methods.
heaving, and the presence of loose sands and silts. Despite these findings, there is still a need for further
Previous studies have investigated the performance of research to address the specific challenges posed by the
DSM walls in challenging soil conditions. A study by Han effect of water content in the DSM mix for cohesive and
and Ye (2002) evaluated the effectiveness of DSM walls in non-cohesive materials.
preventing seepage in a riverbank. The study found that Overall, the literature suggests that DSM walls can be an
the DSM walls were effective in reducing seepage, but that effective solution for challenging soil conditions, but that
the design of the wall and the binder type and dosage were careful consideration must be given to factors such as soil
critical factors in achieving optimal performance. Similarly, type, binder type and dosage, mixing depth, and quality
a study by Rutherford et al. 2007 investigated the use of control measures during construction. The findings of this
DSM walls for excavation support in a site with loose sands paper contribute to this area of research by presenting a
 comprehensive evaluation of DSM wall performance in an 2013). This value ensures an accurate representation of
excavation project with seepage, ground heaving, and the stiffness and behavior of the DSM walls in the analysis.
loose sands and silts, providing valuable insights into the
factors that contribute to successful DSM wall construction
in these conditions. 2.1 Stages of construction

2 METHODOLOGY The construction process for the study can be divided into
eight stages. The detailed stages are described in Table 2
In this study, the soil is modeled using 15 noded plain strain and also the important stages are shown from the Plaxis
elements in Plaxis 2D. The soil layers used in the model in Figure 2. Stage 1 involves establishing virgin soil
simulation are representative of those typically found in the conditions with zero displacement. In Stage 2, the high-rise
Richmond area near the Brighouse Station. The soil layers building is constructed. Once construction is complete,
include a top layer of fill, followed by layers of loose sand, Stage 3 begins, which marks the end of consolidation at
clayey silt, and a compact layer of sand. Since the loading the position under building load. Stage 4 is reached at the
is static in nature and the problem is not expected to end of consolidation after preload. In Stage 5, preload
undergo large deformation, a simple Mohr-Coulomb removal is complete, and consolidation continues. At this
material model is used for all the soil layers. The analysis point, the lateral deflection and vertical settlement values
also considers possible consolidation of all the soil layers. are determined. Stage 6 marks the end of the first stage of
The DSM column is designed such that each cylindrical excavation, while Stage 7 marks the end of the second
column has a diameter of 0.6 m and an area of mutual stage of excavation. Stage 8 includes the end of the third
overlap. The center-to-center distance between the DSM stage of excavation and final consolidation. The lateral
columns is considered as 1.3 m. In each alternative deflection and vertical settlement values discussed in the
column, an I section reinforcement is considered to better results and discussions section were measured at the end
resist the flexural load from the soil layer and the nearby of consolidation, which is at the end of Stage 8. These
structures. Table 1 provides more details about the stages provide a clear timeline for the construction process
reinforcement design. and help to contextualize the findings of the study.
The DSM section is checked against moment and shear
resistances based on Figure 1, but these details are not Table 2. Stages of construction followed in the tests
explained in detail in this paper for the sake of brevity.

Stages Description
S1 Establishment of virgin soil conditions (zero
displacement)
S2 Construction of the high-rise building
S3 End of consolidation the position under
building load
S4 End of consolidation after preload
S5 End of Preload removal+ End of
consolidation after preload removal
S6 End of first stage of excavation
S7 End of second stage of excavation
S8 End of third stage of excavation + final
consolidation
Figure 1. Details of the DSM wall configuration showing the
cross-section of the DSM mix and the I section
reinforcement in alternate columns.
Building UDL
Table 1. Details of the I section used for the reinforcement Preload
of DSM walls.

Metric Depth Width Thickness Web Section

(mm) (mm) (mm) thickness Modulus
(mm) (mm3)
460× 260 509 289 40.4 22.6 5650

The DSM walls in this study are modeled as elastic plate

elements, incorporating appropriate interface elements. (a)
The elastic modulus (E value) for the DSM walls is
determined based on the recommended value of 380 times
(as per the FHWA for Embankments and Foundations,
 Table 4. Soil stratification considered in the current study

Soil layers Top1 (m)

Case 1 Case 2 Case 3 Case 4 Case 5
Excavation
Sand Fill 0 0 0 0 0
Clayey Silt 0.8 0.8 0.8 0.8 0.8
Loose Sand 3.9 3.9 3.9 4.9 5.9
Compact 4.5 5.5 6.5 5.5 6.5
Sand
(b) 1
Difference between top of each layer gives layer thickness. Case
1 to 3 assumes variations in the Loose Sand layer across the site,
Figure 2. Model diagrams in Plaxis showing (a) preloading whereas Case 1, 4 and 5 represents changes in the Clayey Silt
after construction of the DSM wall, (b) excavated section Layer.
and steady state removal of water.
(a)
2.2 Material Properties

The material properties of the soil layers are presented in

Table 3. The soil layers are composed of a medium dense
sand fill, clayey silt which has relatively low cohesion and
low friction angle, followed by loose sand and followed by
a till-like layer which here is presented as a compact sand.

Table 3. Strength properties of the soil layers

Characteristics Sand Fill Clayey Silt Loose Compact

Sand Sand
Young’s 71.3e3 21.45e3 25e3 85e3
Modulus, E in
kN/m2 (b)
Poisson’s Ratio 0.30 0.39 0.31 0.29
Liquid limit 3 35 - 3
Plastic limit - 22 - -
Friction angle 32 22 30 36
∅ in °
Dilation angle 1 0 0 3
Ѱ in °
Cohesion 0 20 0 0.5
c in kPa

2.3 Variation in stratigraphy

Assuming that the typical DSM wall extends approximately

50 m across the plane, it is reasonable to assume that the
two critical layers, Loose Sand and Clayey Silt, will vary Figure 3. Variation of compressive strength of soil-cement
across the site. In order to model the soil profile in 2D, mix at the end of setting for (a) sand, (b) clays (modified
Table 4 describes the possible different sections that need from Topolnickli 2016)
to be considered. Cases 1, 2, and 3 consider the depth of
the Loose Sand layer to be 0.6 m, 1.6 m, and 2.6 m, 2.4 Water content
respectively. Cases 1, 4, and 5 capture the variation in the
Clayey Silt layer at 3.1 m, 4.1 m, and 5.1 m, respectively. Apart from the variation in the thickness of the soil layers,
It should be understood that the variation in these important it is also important to focus on the water content of the
layers might significantly affect the serviceability of the design mix which uses soil from these layers to prepare the
DSM wall and, therefore, must be considered in the DSM wall. It was pointed out by Topolnickli (2016), that the
analysis. Specifically, the variation in these layers might effect of water content on the 28 day Uniaxial Compressive
affect the lateral movement of the DSM wall, its settlement, Strength (UCS) of sands are different from that observed
and the settlement of the 16-storey building with a total in clays when they are mixed with cement. Figure 3 shows
uniform load of 210 kN/m2, which is assumed to be at a that increase in the total water content TWC can
fixed distance of 7 m from the excavation. significantly and monotonically decrease the UCS values
of concrete when sand is used in mixed design for DSM Figure 3. However, as the thickness of the loose-soil layer
walls. This trend is different in clays, where increase in WC increased, as represented by Case 2 and Case 3, the
might increase the UCS value up to an optimum maximum deflection exhibited a gradual increase but appeared to
value in the mix strength is obtained beyond which there is saturate after a certain depth. In the case of clay layers,
steady decline in its strength. In the current study, we there seemed to be a specific design TWC that resulted in
assume that the design of the DSM walls will involve the least effect on the u value. Moreover, the impact of
different combinations of the WC in the loose sand layer TWC on the reduction of the u value was relatively lower
and the clayey silt layer. The WC in the fill and the compact when compared to the loose sand mix. Furthermore, the
sand layer are considered as constant. It is interesting to influence of increased clay layer thickness, as depicted in
note how the DSM wall bends inwards from the combined Case 4 and Case 5, exhibited diminishing effects on the u
effect of excavation and the lateral force exerted by the 16 value. The changes in u became progressively lesser as
storey building. The values of UCS can be usually the clay layer thickness increased. Understanding these
connected with the Young’s Modulus E value of a DSM settlement patterns is crucial for the design and
deep-mixed material empirically as 380 times the UCS construction of structures in similar geotechnical
value of the specimen (Topolnickli (2016)), conditions. It allows engineers to anticipate and mitigate
potential settlement issues, enabling the development of
3 RESULTS AND DISCUSSIONS more robust and resilient foundation systems for high-rise
buildings.
The investigation focused on the construction of the
Deep Soil Mixing (DSM) wall and subsequent excavation.
The variation in the soil profile across a 50 m out-of-plane
length excavation was captured using five different 2D
plane strain sections, namely Case 1 (Sand & Clay), Case
2, Case 3, Case 4, and Case 5. These variations had
significant effects on the lateral deflection (u) of the DSM
wall nearest to the high-rise building, as well as the
settlement of the high-rise building (v).

v
u

Figure 5. Lateral deflection of the top right corner of the

DSM wall in response to varying water content in the DSM
cement-soil mix design for loose sand and clayey-silt
layers.

Figure 6 illustrates the impact of changes in the unconfined

compressive strength (UCS) of the soil mix and the
Figure 4. The deformed mesh showing the lateral increased consolidation resulting from the heightened clay
deflection of the DSM wall and the vertical settlement of the layers on the settlement of the left corner of the high-rise
close building. Note : Settlements are exaggerated 50 building. These settlement observations are a
times for better visualization consequence of the excavation and redistribution of in-situ
stresses.
Figure 5 illustrates the lateral deflection of the DSM wall at The influence of UCS reduction on the vertical settlement
the top right corner, providing insights into the behavior of (v) of the building is prominently observed in the sand layer,
the wall under different scenarios. An envelope of curves where consolidation has no effect. The settlement in this
was plotted, considering variations in water content layer is primarily governed by the reduction in UCS.
exclusively for the sand layer. It was observed that there However, the behavior of the clay layer differs, as it
exists a monotonic, almost power-law relationship between deviates from the initial loss of strength curve, which
the increase in total water content (TWC) and the lateral exhibited an inverted "V" shape. As the thickness of the
deflection. For instance, at TWC = 20%, the u value was clay layer increases, the settlement of the building due to
found to be 53 mm, while for TWC increased to 32%, the u layer consolidation becomes more significant than the loss
value increased significantly to 94 mm. This trend closely of lateral stress caused by the excavation.
aligns with the decrease in compressive strength observed
in cement-sand mixtures as TWC increases, as shown in
 consolidation of the silty clay layer played a relatively
smaller role.

Furthermore, the settlement of the high-rise building was

influenced by both the excavation process and
consolidation of the soil layers affected by the excavation.
In sections where the thickness of the silty clay layers
exceeded 2 m, vertical settlement was predominantly
governed by consolidation rather than the effect of the
DSM wall. These findings highlight the importance of
considering water content and variations in soil layers
when designing and constructing DSM walls. The results
also emphasize the need for careful monitoring and
analysis during the construction process to mitigate
potential settlement issues and ensure the stability and
performance of both the DSM wall and adjacent structures.

6 REFERENCES
Figure 6. Vertical settlement of the left corner of the
highrise building in response to varying water content in the Algulin, J.O.E.L. and Pedersen, B.J.Ö.R.N., 2014.
DSM cement-soil mix design for loose sand and clayey-silt Modelling of a piled raft foundation as a plane strain model
layers. in PLAXIS 2D. Master of Science, Department of Civil and
Environment Eng, Division of Geo-Engineering, Chalmers
These findings highlight the contrasting behaviors of the University, Sweden.
sand and clay layers. The sand layer primarily experiences
settlement driven by the reduction in UCS, whereas the Bruce, M.E.C., Berg, R.R., Filz, G.M., Terashi, M., Yang,
clay layer exhibits settlement influenced by both D.S., Collin, J.G. and Geotechnica, S., 2013. Federal
consolidation and lateral stress reduction. The deviation highway administration design manual: Deep mixing for
from the original loss of strength curve suggests the embankment and foundation support (No. FHWA-HRT-13-
complex interaction between the clay layer and the 046). United States. Federal Highway Administration.
surrounding soil during the excavation process. Offices of Research & Development.

Overall, the results demonstrate the significant influence of Han, J., Zhou, H.T. and Ye, F., 2002. State-of-practice
soil variations on the behavior of the DSM wall. The review of deep soil mixing techniques in China.
relationship between water content, soil layer thickness, Transportation research record, 1808(1), pp.49-57.
and lateral deflection provides valuable insights for the
design and construction of DSM walls, highlighting the Topolnicki, M., 2016, February. General overview and
importance of considering these factors to ensure the advances in Deep Soil Mixing. In XXIV geotechnical
stability and performance of such geotechnical structures. conference of torino design, construction and controls of
soil improvement systems (pp. 25-26).

5. CONCLUSIONS Rutherford, C.J., Biscontin, G., Koutsoftas, D. and Briaud,

J.L., 2007. Design process of deep soil mixed walls for
In this study, a comprehensive analysis of the Deep Soil excavation support. ISSMGE International Journal of
Mixing (DSM) wall construction and its interaction with the Geoengineering Case Histories, 1(2), pp.56-72.
surrounding soil and nearby high-rise building was
conducted using Plaxis 2D fully coupled simulation. The Taki., O..and Yang, D. 1991, Soil Cement Mixed Wall
construction process was divided into eight phases, Technique, Geotechnical Engineering Congress, ASCE,
including wall preparation, preloading, and excavation of New York, Specia l Publication, 27: 298-203
the inner soil in three stages. The variability in the soil
layers around the site was carefully considered, focusing Yapage, N.N.S., Liyanapathirana, D.S., Kelly, R.B.,
on two critical layers: loose sand and clayey silt. Five Poulos, H.G. and Leo, C.J., 2014. Numerical modeling of
different sections were analyzed, accounting for the an embankment over soft ground improved with deep
variations in these target layers. cement mixed columns: case history. Journal of
Geotechnical and Geoenvironmental Engineering,
The results of the study revealed that the water content and 140(11), p.04014062.
differences in thickness between the sand and silt layers
significantly influenced the performance of the DSM wall.
Specifically, the lateral deflection of the wall and settlement
of the nearby high-rise building were found to be greatly
affected by these factors. The lateral settlement primarily
resulted from unloading due to excavation, while
Preliminary impact force estimation of the
Deschaillons Landslide, in Québec, using
the Material Point Method
John Forero, Félix St-Pierre and Ariane Locat
Département de génie civil et de génie des eaux – Université
Laval, Québec, Canada
Floriane Provost
Ecole et Observatoire des Sciences de la Terre [EOST] –
Université de Strasbourg, Strasbourg, France
Jacques Locat
Département de géologie et de génie géologique – Université Laval, Québec, Canada
Scott McDougall
Department of Earth, Ocean and Atmospheric Sciences – University of British Columbia, British Columbia,
Canada
Pascal Locat and Rémi Mompin
Ministère des transports et de la Mobilité durable du Québec, Québec, Canada

ABSTRACT
In eastern Canada, superficial landslides are a frequent type of landslides in clay slopes. On April 27, 2019, such a landslide
occurred in Deschaillons-sur-le-Saint-Laurent along the Saint-Laurent River in Québec. To investigate the post-failure
impact force of this case, a model based on the Material Point Method (MPM), MPM-PUCRio, and another based on
Smoothed Particle Hydrodynamics (SPH), DAN3D, are used in this study. The pre- and post-failure surfaces used in the
models are generated from available Digital Elevation Models (DEM). The clay material that constitutes the sliding mass
is considered viscoplastic and is simulated using the Bingham model. The results from the MPM-PUCRio software are
compared with the results from the DAN3D software in order to define and validate the impact force estimation for the
Deschaillons superficial landslide.

RÉSUMÉ
Dans l'est du Canada, les glissements de terrain superficiels sont le type de glissement de terrain fréquents dans les
pentes argileuses. Le 27 avril 2019, un tel glissement de terrain s'est produit à Deschaillons-sur-le-Saint-Laurent, le long
du fleuve Saint-Laurent au Québec. Pour étudier la force d'impact lors de la post-rupture de ce cas, un modèle basé sur
la méthode Material Point Method (MPM), MPM-PUCRio, et un autre basé sur le Smoothed Particle Hydrodynamics (SPH),
DAN3D, est utilisé dans cette étude. Le pré- et post-rupture utilisées dans les modèles sont générées à partir des modèles
numériques d'élévation (DEM) disponibles. Le matériau argileux qui constitue la masse de débris est considéré comme
viscoplastique et est simulé à l'aide du modèle de Bingham. Les résultats du logiciel MPM-PUCRio du logiciel DAN3D sont
comparés afin de définir et de valider l'estimation de la force d'impact pour le glissement superficiel de Deschaillons.

1 INTRODUCTION 𝐹 = 𝛼𝜌𝜈2 ℎ𝑤 [1]

Superficial landslides are fast and can reach large Where 𝜌 is the bulk density of the material (𝑘𝑔⁄𝑚3 ), 𝛼
distances in relatively short periods of time. These is the dynamic coefficient, 𝑣 is the velocity (𝑚⁄𝑠), and ℎ
landslides occur mainly in areas of steep slopes (more than (𝑚) and 𝑤 (𝑚) denote the debris flow depth and the
20°) and can be triggered by heavy rainfall. The sliding channel width, respectively. A detailed literature review for
mass can contain a mixture of water and sediments of the estimation of impact force was conducted by (Provost
various sizes (Causes 2001). In Eastern Canada, Alaska, et al. 2022).
and Norway, the source of landslides can often contain In recent years, advanced numerical methods have
sensitive clay (Geertsema et al. 2018). According to been developed to allow the analysis of engineering
Cruden et al. (1996), the volume of the displaced mass, the problems involving large deformations (e.g. Troncone et al.
type of movement, and the height and inclination of the 2020). Numerical particle-based methods are frequently
slope are factors that determine the velocity of the debris. used to solve large-deformation problems. Currently, the
The velocity of the displaced mass also strongly depends most relevant are the Smoothed Particle Hydrodynamics
on the thickness of the mass and the width of the flow method (SPH) developed by Lucy (1977) and Gingold et al.
channel. These factors are frequently used in (1977), the discrete element method (DEM) developed by
hydrodynamic methods to determine the impact force of Cundall et al. (1979) and the Material Point Method (MPM)
debris. As reported by Tan et al. (2019), the hydrodynamic developed by Sulsky et al. (1994).
impact force estimation method first proposed by Hungr et SPH is a mesh-free technique based on a pure
al. (1984) is calculated as: Lagrangian description. It has been used to solve problems
such as fluid-structure interaction (FSI) (e.g. Dai et al.
2017) and large-scale landslide motion (e.g. Dai et al. Additionally, the observed damage to the structures
2014; Peng et al. 2022). Several software programs have provides valuable insights. The approximate thickness of
incorporated viscoplastic models, such as the Frictional the debris was determined using a scale. This measured
and Bingham models, based on the SPH method. For thickness is later used for comparison with the thickness
instance, Cuomo et al. (2017) utilized the Frictional model obtained through numerical modelling using DAN3D and
to investigate the propagation of debris avalanches, while MPM-PUCRio software.
Dai et al. (2017) employed the Bingham model in clays to
investigate fluid-structure interaction and debris flow
impact estimation.
MPM, which was developed for simulating large
deformations with a free mesh, has been applied in several
studies for post-rupture slip simulations, including Llano-
Serna (2016), Soga et al. (2016), and Conte et al. (2019).
Additionally, MPM has been utilized to calculate the impact
forces of landslides on barriers, as demonstrated in studies
by Cuomo and Martinelli (2022), Cuomo et al. (2021), and
Ceccato et al. (2017). Several authors have employed the
Mohr-Coulomb constitutive model in MPM for landslide
modelling. For instance, Bandara (2016) used it to
determine the hydro-mechanical behaviour, while Nguyen
et al. (2021) performed a parametric study of shallow
landslides using this model. Andersen et al. (2008) also
used it to model landslides in fine-grained soils induced by
rainfall, based on the generalized interpolation material
point (GIMP) method. Conversely, Llano-Serna et al.
(2016) and Xu et al. (2019) modelled landslide run-out
processes using MPM with the von Mises model and
Drucker-Prager model with viscosity.
To the best of our knowledge, no studies have yet
simulated landslides in sensitive clays using the Bingham
viscoplastic model with MPM. Therefore, this work aims to
contribute to the field by applying the Material Point Method
to model superficial landslide runout in sensitive clays Figure 2. Topography of the site before the Deschaillons
using the Bingham model. landslides, showing the location of cross section A-A’ and
Additionally, this work aims to define and validate the of two observation points located near buildings affected by
calculation of the impact force, by comparing results from the landslide: point 1 near the toe of the slope and point 2
DAN3D and MPM-PUCRio. To achieve this, the on the side of a building impacted by the flow.
Deschaillons landslide that occurred on April 27th 2019, is
used in this study (Figures 1 and 2). The debris, made of
clay and vegetation, impacted buildings of a marina located 2 THE DAN3D AND MPM-PUCRIO SOFTWARE
near the toe of the slope, fortunately causing no deaths.
Information about the conditions before and after the The software MPM-PUCRio, was developed by Fernández
landslide was also acquired by Québec Ministry of (2020) based on the Material Point Method. The domain of
Transportation engineers and made available to Université MPM is represented by Lagrangian points called MPs and
Laval for this research project. an Eulerian mesh. The method was developed to solve
large deformation problems. The MPs are moving through
an Eulerian computational mesh. The MPs carry all
physical properties of the continuum, such as velocities,
stresses, strains, density, momentum, material
parameters, and other state parameters, whereas the
computational mesh is used to solve the balance equations
without storing any permanent information.
Figure 3 illustrates a schematic of the MPM method. On
Figure 3a, the variables are transferred from the particles
to the nodes by mapping functions. The balance equations
are later solved at the nodes (Figure 3b). On Figure 3c,
subsequently, the solution is interpolated back to the
particles. The position of the particles and the variables are
Figure 1. Deschaillons landslide located on the south shore updated (Figure 3d). This process is repeated at each time
of the St. Lawrence River east of Québec City. (Québec step until the end of the simulation.
Ministry of Transportation -MTQ)
Figure 1 shows a photo of the debris at the base of the
slope resulting from the Deschaillons landslide.
 were imposed, which allowed particles to only move in the
direction of the slide (Y axis). The limits of the Eulerian
mesh were set to (0,0,0) for the initial coordinates
(𝑋0 , 𝑌0 , 𝑍0 ) and (50,110,50) for the final coordinates
(𝑋𝑛 , 𝑌𝑛 , 𝑍𝑛 ). During the simulation, a contact method with a
friction coefficient of 0.55 was employed. Using lower
coefficients resulted in an increase in particle velocity. The
buildings were not considered in this case.
In DAN3D, the 3D model was constructed based on the
same topography and source as in MPM-PUCRio.
However, the construction of the 3D model in DAN3D
differs, as it does not require a particle distribution or
generation of a base volume. The software interface allows
for the use of previously generated (.grd) files from LiDAR
data as input. This method allows for a computationally
faster and more efficient model. Additionally, the interface
enables the use of a Digital Elevation Model (DEM) as a
background image, aiding in locating debris at each time
step. DAN3D also provides a friendly and easy-to-use
interface. The simulation employed 1000 particles. The
simulation duration was set to 60 seconds. DAN3D
software enables data export at specific time steps, offering
Figure 3. Schematic of the MPM method showing: (a)
information on node and particle thickness and velocity.
transfer of information from moving particles to fixed nodes,
Similar to the MPM-PUCRio model, the presence of
(b) solution of balance equations at nodes (c) transfer of
structures at points 1 and 2 was not considered.
information back from nodes to particles and (d) update to
particle positions.

The software DAN3D, developed by McDougall and

Hungr (2004), is based on the SPH method. SPH uses a
set of particles that interact with each other through a
smoothing function, which allows the calculation of fluid
properties at any point in the simulation domain. Each
particle has a volume, position, and velocity, and these
properties are updated based on the forces acting on them.

2.1 Numerical 3D model

The numerical model representing the Deschaillons

superficial landslide was constructed using digital elevation
information based on LiDAR data. Two sets of data were
required, one from 2017 and the other from 2019, taken
before and after the landslide occurred, respectively. The
difference between these elevation points was used to
determine the source of the landslide. Additionally, both
data sets were combined to obtain the topography along
which the landslide flows in the numerical analysis. Figure
4 shows the 3D model of the Deschaillons superficial
landslide obtained from these data sets and used in the Figure 4. A 3D model of the Deschaillons superficial
MPM-PUCRio software. landslide was created using the MPM-PUCRio software.
The surface for MPM-PUCRio was constructed using The landslide is represented in red and the surface material
elevation points spaced at 0.25 m intervals. A volume was is represented in blue.
generated from the surface, where each cell had
dimensions of 1 m x 1 m x 1 m and contained 16 material 2.2 Bingham rheological parameters
points. Decreasing the dimensions of this configuration
caused convergence issues during the simulation. The The Bingham model is a rheological model used to
same particle distribution was used to construct the source. describe the behaviour of viscoplastic fluids, which are
Two materials were considered for the numerical those that exhibit a threshold or yield stress before they
simulation: the base volume generated from the begin to flow. According to Dai et al. (2017), in this model,
topography and the volume of the source (171 𝑚3 ). To the relationship between shear stress and shear strain rate
perform this work, both ArcGIS software and the Python of a viscoplastic fluid is described using the following
programming language were utilized. The simulation equation:
involved a total of 852,075 particles. Contour conditions
 𝜏 = (𝜇𝐵 + (𝐷
𝜏𝑜
)𝐷 [2] the actual landslide limit and the debris from the MPM-
2𝑑 )
1/2
PUCRio analysis stays inside it.
The simulated velocities, thicknesses and impact
Where 𝜏 is the shear stress, 𝜏𝑜 is the yield stress, 𝜇𝐵 is the forces along cross-section (A-A') and points 1 and 2 were
viscosity, 𝐷 is the tensor of strain rates, and 𝐷2𝑑 is the also analyzed. These points were located near the
second invariant of the tensor of strain rates. buildings to estimate the impact forces on structures (see
The Bingham viscoplastic model was utilized in this Figure 2 for locations). Figure 6a illustrates the maximum
study to represent the behaviour of sensitive clays. The velocity at a given distance obtain from the analysis along
rheological parameters were obtained from Locat et al. cross-section A-A' using DAN3D and MPM-PUCRio. The
(1988), who presented a relationship between viscosity coordinate (0,0) is the intersection between the line A-A'
and yield stress. Linear interpolation was employed to and the light blue line, at the top. In the DAN3D simulation,
determine the values listed in Table 1. Laboratory tests the maximum velocity is 12.7 m/s, occurring at 27.5 m and
should be carried out in future studies to validate these 3 seconds (dashed red line). Similarly, the MPM-PUCRio
parameters. software shows a maximum velocity of 13.3 m/s at 26 m,
reached within 5 seconds (dashed blue line). However, the
Table 1. Parameters of the Bingham model velocities do not reach zero by the end of the simulation,
primarily due to the stopping criteria of the software.
Software 𝑬 𝒗 𝝆 𝝉𝒐 𝝁𝑩 Generally, the maximum time value is considered as the
(𝒌𝑷𝒂) (𝒌𝒈 ⁄𝒎𝟑 ) (𝒌𝑷𝒂) (𝒌𝑷𝒂 ∙ 𝒔) stopping criterion in numerical modelling, rather than a
DAN3D 1938 5 0.005 specific variable value. In this case, a time greater than 60
seconds was not considered.
MPM- 5.e3 0.3 1938 5 0.005 Figure 6b depicts the maximum thickness at a given
PUCRio
distance along cross-section A-A' using DAN3D and MPM-
PUCRio. The maximum thickness observed in both cases
3 RESULTS are consistent. A maximum thickness of 0.75 m is observed
approximately at 30 m, coinciding with the maximum
Figure 5 illustrates the area impacted by the velocity observed in Figure 6(a). In addition, it is observed
Deschaillons landslide (light blue line) compared with the that the maximum peak is close to the red and blue dashed
simulations using DAN3D and MPM-PUCRio. In both lines that indicate the moment at which the maximum
cases, the particle cluster representing the debris covers velocity peak is reached. Between 50 m and 80 m, an
the actual landslide area shown by the light blue dashed average maximum thickness of 0.30 m is observed.
line. The debris from the DAN3D analysis goes slightly past

Figure 5. Area of impact of the final debris, comparing the DAN3D and the MPM-PUCRio software and the area of actual
impact indicated by a dashed light blue line.
The approximate thickness (observed data) depicted in as the hydrodynamic factor decreases, the results from the
Figure 1 was added to Figure 6b (red dots) to facilitate two software converge.
comparison with the numerical models.
Figure 6c displays the maximum impact force at a given
distance along cross-section A-A' using maximum
velocities and thicknesses from DAN3D and MPM-
PUCRio, with a hydrodynamic coefficient of α = 0.9 in both
cases. The α value is based on Wendeler (2016), who used
a value of 0.7 for low-density clay debris flows. The density
of sensitivity clay from the Deschaillons landslide is higher,
as shown in Table 1, therefore the selected value of α is
higher. However, these values are preliminary and must be
corroborated in the laboratory or in the field. Notably, the
maximum impact force occurs at approximately 30 m, close
to the peak of the maximum velocity observed in Figure 6a.
These results indicate a significant risk factor for any
structure located at this distance along cross-section A-A'.
It is important to reiterate that the impact force is not
directly calculated by the DAN3D and MPM-PUCRio
programs themselves. Instead, it is derived from the
analysis of the pre-existing thickness and velocity data
provided by these programs. These data are utilized in the
study to estimate the impact force using the traditional
hydrodynamic equation.
Figure 7a depicts the velocity at points 1 and 2 (see
Figure 1 for locations). These velocities are specific to
these points and do not correspond to the maximum
velocities recorded along cross-section A-A’ in Figure 6a.
In Figure 7a, a peak velocity of 1.0 m/s for DAN3D and 2.2
m/s for MPM-PUCRio is observed at point 1 within 15
seconds. The peak velocities occur after about 30 s at point
2. In both cases, the velocity decreases over time.
Figure 7b presents the thickness at points 1 and 2. It is
observed that the results obtained using MPM-PUCRio are
adequately adjusted to those obtained using DAN3D. At
point 1, the average thickness measures 0.25 m, while at
point 2, it measures 0.120 m. In both cases the average
thickness is less than the observed thickness of 0.70 m.
However, the observed thickness includes the
accumulation of debris in the structures, whereas the
thickness calculated by DAN3D and MPM-PUCRio does
not take into account the presence of structures.
In Figure 7c, the impact force at points 1 and 2 is
displayed. Point 2 has a lower impact force than point 1. A
value of the hydrodynamic coefficient of α = 0.9 was used
for DAN3D and MPM-PUCRio. At point 1, the maximum
peak of the impact force, determined by the velocity and
thickness calculated using MPM-PUCRio, is nearly three
times higher compared to DAN3D. This difference can be
primarily attributed to the velocity estimation depicted in
Figure 7(a). At point 2, the impact force values are similar.
This is because both DAN3D and MPM-PUC Rio produce
consistent results in terms of velocity and thickness
calculations. However, the impact force value is sensitive
to the hydrodynamic factor. For sensitive clays, this factor
is not yet well-defined in the literature. Figure 6. Comparation between DAN3D and MPM-PUCRio
The hydrodynamic factor considers the angle between along cross-section A-A': (a) maximum velocity, (b)
the velocity vector and the infrastructure (barrier), which maximum thickness, (c) maximum impact force. Solid black
must be measured in the field or in the laboratory. An lines are used for the DAN3D results and dashed black
analysis of the influence of the hydrodynamic factor on the lines are used for the MPM-PUCRio results.
impact force is presented in Figure 8. The results show that
 Figure 8. Comparison between DAN3D and MPM-PUCRio
with various hydrodynamic factors (α) used to estimate
impact forces at points 1 and 2.

4 DISCUSSION

The results presented in Figure 6 only represent the

maximum velocities, maximum thicknesses and maximum
impact forces simulated along cross-section A-A'. They do
not provide information about the overall extent of the
landslide, leaving uncertainty about the specific areas
where the velocities, thicknesses, and impact forces are
highest. However, it is evident that between 20 and 40
metres, the impact force reaches potentially dangerous
maximum values, ranging from 200 KN/m to 250 KN/m.
On Figure 6b, the observed data (red dots) are not
exact representations of the actual thicknesses, but rather
approximations. Ideally, precise measurements would
have been obtained from digital information systems;
however, this was not feasible due to the removal of a
portion of the material deposited at the slope base prior to
the digital data collection using LiDAR. Furthermore, this
observed data may exhibit variations depending on the
image location and projection. Nevertheless, this approach
was employed effectively for the purpose of comparing the
DAN3D and MPM-PUCRio software.
The simulation utilized the contact method
implemented in MPM-PUCRio, which effectively prevents
particle interpenetration between different materials by
controlling particle velocities through a designated
function. For the specific case of sensitive clays, a
coefficient of friction of 0.55 was employed. Additional
simulations, which are not presented in this study, revealed
that a decrease of the coefficient of friction leads to higher
particle velocities.
In Figure 6c, the calculated impact forces are similar
throughout the entire process in both models, indicating
consistent results between the two software programs
Figure 7. Comparison between DAN3D and MPM-PUCRio
despite their numerical differences.
at points 1 (black lines) and 2 (blue lines): (a) velocity, (b)
In Figure 7, as anticipated, the impact force is lower at
thickness, (c) impact force. Solid lines are used for the
point 2 due to its distance from the starting point of the
DAN3D results and dashed lines are used for the MPM-
landslide. The calculation of the impact force in Figures 6
PUCRio results.
and 7 utilized the hydrodynamic force, while the hydrostatic
force was disregarded, as it is typically lower, as noted by
Tan et al. (2019).
 The objective was to estimate the impact forces during 6 ACKNOWLEDGEMENTS
the post-failure phase, and the influence of material
accumulation at points 1 and 2 was not taken into account. Special thanks are extended to the Ministry of
It would be worth incorporating structures into the Transportation (MTQ) for providing the databases and
simulations and assessing the impact forces using various Fabricio Fernández author of MPM-PUCRio for provided
calculation methods depending on the type of structure and excellent feedback and the authors for special technical
whether it is rigid or flexible, as suggested by Tan et al. advice. This research was carried out with the financial
(2019). However, this is not easy to implement in a depth help of Cadre pour la prevention des sinistres du
averaged model like DAN3D. Gouvernement du Québec (CPS 21-22-21).
The estimation of impact force was conducted using
both DAN3D and MPM-PUCRio software. The magnitudes 7 REFERENCES
of the impact forces obtained from both software programs
are consistent. However, it should be noted that the results Andersen, S. M., and Andersen, L. (2008). Modeling of
for velocity and thickness are still preliminary. In order to landslides with the material point method. In
obtain more accurate results using MPM-PUCRio, Proceedings of the Sixth International Conference on
simulations with models employing different particle Engineering Computational Technology. Civil-Comp
discretization techniques need to be performed. Press.
Additionally, the Bingham model utilized in MPM-PUCRio Bandara, S., Ferrari, A., & Laloui, L. (2016). Modelling
should be validated using experimental data. landslides in unsaturated slopes subjected to rainfall
The findings of this study contribute to the protection of infiltration using material point method. International
people and the safe design of infrastructure. The current Journal for Numerical and Analytical Methods in
project did not include a parametric analysis of the Geomechanics, 40(9), 1358-1380.
independent input variables, but it is crucial for future Causes, L. (2001). Landslide types and processes. US
studies to consider estimating velocity based on various Geological Survey: Reston, VA, USA.
scenarios, including factors such as mass, height, slope, Ceccato, F., Simonini, P., di Prisco, C., and Redaelli, I.
and thickness, as recommended by Cruden et al. (1996). (2017). The effect of the front inclination on the impact
Additionally, it is important to estimate the rheological forces transmitted by granular flows to rigid structures.
parameters through laboratory tests instead of relying on In Advancing Culture of Living with Landslides: Volume
linear projections, as done in this study. 2 Advances in Landslide Science (pp. 593-599).
By incorporating these recommendations into future Springer International Publishing.
research, a more comprehensive understanding of the Conte, E., Pugliese, L., and Troncone, A. (2019). Post-
process can be achieved, leading to improved prevention failure stage simulation of a landslide using the material
strategies and more accurate design of protective point method. Engineering Geology, 253, 149-159.
measures. Cuomo, S, and Martinelli, M. (2022). Empirical formulation
for debris flow impact and energy release.
5 CONCLUSION Geoenvironmental Disasters, 9.1.
Cuomo, S., Di Perna, A., and Martinelli, M. (2021). MPM
The DAN3D and MPM-PUCRio software utilize modelling of buildings impacted by landslides.
different formulations, based on the SPH and MPM Understanding and Reducing Landslide Disaster Risk:
methods, respectively. However, both 3D models used the Volume 5 Catastrophic Landslides and Frontiers of
same input topography, source locations and rheological Landslide Science 5th, 245-266.
properties. The preliminary results of velocity, thickness, Cuomo, S., Cascini, L., Pastor, M., and Petrosino, S.
and impact force simulated by the two programs are (2017). Modelling the propagation of debris avalanches
reasonably consistent. This demonstrates the potential of in presence of obstacles. In Advancing Culture of Living
the Material Point Method for calculating the impact force with Landslides: Volume 3 Advances in Landslide
of flow slides on infrastructure. However, DAN3D offers Technology (pp. 469-475). Springer International
greater ease in obtaining the output variables, such as Publishing.
thickness, compared to MPM-PUCRio. In contrast, utilizing Cundall, P. A., and Strack, O. D. (1979). A discrete
the MPM-PUCRio requires more effort to achieve the numerical model for granular assemblies.
desired thickness. Therefore, the process of obtaining Géotechnique, 29(1), 47-65.
thickness using the MPM-PUCRio software is more labour- Crozier, M. J., and Glade, T. (2005). Landslide hazard and
intensive and may involve additional steps when compared risk: issues, concepts and approach. Landslide hazard
to DAN3D. Additionally it is concluded that the coefficient and risk, 1-40.
of friction of the contact method used in the MPM-PUCRio Cruden, D. M., and Varnes, D. J. (1996). Landslides:
can influence the results of particle velocity. Investigation and Mitigation. Chapter 3 Landslides
Regarding the Bingham viscoplastic model and its Types and Processes. Transportation research board
parameters, it can be concluded that they adequately special report, 247.
represent the behaviour in the case of Deschaillons. The Dai, Z., Huang, Y., Cheng, H., and Xu, Q. (2014). 3D
parameters used for this model will be confirmed by numerical modeling using smoothed particle
upcoming laboratory tests. Further analyses are also hydrodynamics of flow-like landslide propagation
needed to confirm which of the methods allows the most triggered by the 2008 Wenchuan earthquake.
accurate impact force estimation. Engineering Geology, 180, 21-33.
Dai, Z., Huang, Y., Cheng, H., and Xu, Q. (2017). SPH Sulsky, D., Chen, Z., and Schreyer, H. L. (1994). A particle
model for fluid–structure interaction and its application method for history-dependent materials. Computer
to debris flow impact estimation. Landslides, 14, 917- methods in applied mechanics and engineering, 118(1-
928. 2), 179-196.
Provost, F., Locat, A., (2022). Estimation de la force Troncone, A., Conte, E., and Pugliese, L. (2020). Back-
d’impact des débris des glissements superficiels Analysis of the Post-Failure Stage of a Landslide Using
(PACC 20-21-02). Ministère de la sécurité publique du the Material Point Method. In Geotechnical Research
Québec. for Land Protection and Development: Proceedings of
Fernández, F., (2020). Modelagem Numérica De CNRIG 2019 7 (pp. 265-272). Springer International
Problemas Geotécnicos De Grandes Deformações Publishing.
Mediante O Método Do Ponto Material. Pontifícia Tan, D. Y., Yin, J. H., Feng, W. Q., Zhu, Z. H., Qin, J. Q.,
Universidade Católica de Rio de and Chen, W. B. (2019). New simple method for
Janeiro. https://doi.org/10.17771/pucrio.acad.52771 calculating impact force on flexible barrier considering
Geertsema, M., Blais-Stevens, A., Kwoll, E., Menounos, B., partial muddy debris flow passing through. Journal of
Venditti, J. G., Grenier, A., and Wiebe, K. (2018). Geotechnical and Geoenvironmental Engineering,
Sensitive clay landslide detection and characterization 145(9), 04019051.
in and around Lakelse Lake, British Columbia, Canada. Wendeler, C. (2016). Debris-flow protection systems for
Sedimentary Geology, 364, 217-227. mountain torrents: basics principles for planning and
Gingold, R. A., and Monaghan, J. J. (1977). Smoothed calculation of flexible barriers.
particle hydrodynamics: theory and application to non- Xu, X., Jin, F., Sun, Q., Soga, K., and Zhou, G. G. (2019).
spherical stars. Monthly notices of the royal Three-dimensional material point method modeling of
astronomical society, 181(3), 375-389. runout behavior of the Hongshiyan landslide. Canadian
Hungr, O., Morgan, G. C., and Kellerhals, R. (1984). Geotechnical Journal, 56(9), 1318-1337.
Quantitative analysis of debris torrent hazards for
design of remedial measures. Canadian Geotechnical
Journal, 21(4), 663-677.
Locat, J., and Demers, D. (1988). Viscosity, yield stress,
remolded strength, and liquidity index relationships for
sensitive clays. Canadian Geotechnical Journal, 25(4),
799-806.
Lucy, L. B. (1977). A numerical approach to the testing of
the fission hypothesis. Astronomical Journal, vol. 82,
Dec. 1977, p. 1013-1024., 82, 1013-1024.
Llano-Serna, M. A., Farias, M. M., and Pedroso, D. M.
(2016). An assessment of the material point method for
modelling large scale run-out processes in landslides.
Landslides, 13, 1057-1066.
McDougall, S., and Hungr, O. (2004). A model for the
analysis of rapid landslide motion across three-
dimensional terrain. Canadian Geotechnical Journal,
41(6), 1084-1097.
Nguyen, T. S., Yang, K. H., Ho, C. C., and Huang, F. C.
(2021). Postfailure characterization of shallow
landslides using the material point method. Geofluids,,
1-25.
Peng, C., Li, S., Wu, W., An, H., Chen, X., Ouyang, C., and
Tang, H. (2022). On three-dimensional SPH modelling
of large-scale landslides. Canadian Geotechnical
Journal, 59(1), 24-39.
Soga, K., Alonso, E., Yerro, A., Kumar, K., and Bandara,
S. (2016). Trends in large-deformation analysis of
landslide mass movements with particular emphasis on
the material point method. Géotechnique, 66(3), 248-
273.
Squarzoni, G., Bayer, B., Franceschini, S., and Simoni, A.
(2020). Pre-and post-failure dynamics of landslides in
the Northern Apennines revealed by space-borne
synthetic aperture radar interferometry (InSAR).
Geomorphology, 369.
Numerical Simulation of Rock Bridge Failure
Using Combined Finite-Discrete Element
Method
Yalin Li, Davide Elmo
Norman B. Keevil Institute of Mining Engineering, The University of
British Columbia, BC, Canada
Omid Mahabadi
Geomechanica Inc., Toronto, Ontario, Canada

ABSTRACT
The Bologna Interpretation of rock bridges by Elmo (2023) compared rock bridges to a spinning coin: you can only tell
whether the coin is tail or head when it stops spinning. The spinning coin represents a stable rock mass, and rock bridges
inside the rock mass can only be observed or measured when the rock mass has failed. Therefore, the rock bridge
percentage approach to relating equivalent rock mass strength should not be used since the value of rock bridge
percentage cannot be directly measured before the failure occurs. It has been discussed that rock bridge failure is a time-
dependent brittle failure process, and the mode of failure is directly related to rock block and joint strengths. This paper
presents the results of predictive analysis for investigating the mechanisms of rock bridge failure using Parus rock (скала́
Па́рус) as a case study. A numerical model implementing the combined Finite-Discrete Element Method (FDEM) is used
to capture the process of rock bridge failure. The simulation results show that the tensile cracks initiate at the point where
the joint terminates. The initiated cracks propagate across rock blocks and coalesce with other induced or natural fractures
forming rock bridges. The rock fragments exposed to the free surface slide along well-coalesced fractures leading to
potential catastrophic failure. This study concludes that 1) rock bridge failure is a time-dependent process which was not
considered in the previously developed mathematical equations for calculating rock bridge strength, and 2) both rock block
and joint strength affect the modes of rock bridge failure.

1 INTRODUCTION explicitly capture the potential fracturing process and rock

bridge failure within a jointed rock mass. This research
Rock bridges are generally defined as the portion of intact uses the Parus rock (Sail rock) as a case study. This
rock between discontinuity surfaces that can fail under study's central objective is to understand how rock block
certain stress conditions. The measurements of rock bridge and joint strength affect rock mass fragmentation and the
percentage with respect to the equivalent rock mass mechanisms of rock bridge failure.
strength have been initially studied for slope stability
analysis since the early 1960s (e.g., Jennings, 1970, 1972; 2 COMBINED FINITE-DISCRETE ELEMENT
Jennings and Steffen, 1967; Terzaghi, 1962). The previous METHOD (FDEM)
studies suggest that the rock bridge percentage can be
calculated first based on the length/area of the intact The combined Finite-Discrete Element Method (FDEM) is
portion relative to the length/area of the rock joint. The an innovative numerical technique that combines the Finite
equivalent rock mass strength parameters (e.g., equivalent Element Method (FEM) and the Discrete Element Method
cohesion and friction angle) are then calculated based on (DEM) techniques, with the incorporation of fracture
the rock bridge percentage and intact rock strength. Those mechanics principles, to explicitly simulate progressive
equivalent rock mass strength parameters can be used for damage and failure process in brittle materials (Klerck
limit equilibrium analysis. 2000, Munjiza 2004, Mahabadi 2012, Elmo et al. 2013,
Recently, Elmo (2022) and Elmo et al. (2023) argued Lisjak and Grasselli 2014). The simulation using FDEM
that the rock bridge percentage approach to relating rock starts with FEM analysis for continuum material, followed
mass strength should not be used, as the rock bridge by DEM when the fractures are generated within the
percentage cannot be directly measured. According to the continuum material and newly formed discrete bodies
Bologna Interpretation (Elmo 2023), a rock bridge exists begin to interact with each other.
only when rock mass fails. Furthermore, Elmo (2023) Figure 1 shows the intra-element and inter-element
pointed out that damage-related rock bridge failure is a fractures that FDEM can generate within a meshed
time-dependent process, and numerical analyses using continuum. The intra-element fracture shown in Figure 1b
discrete element models with the incorporation of fracture is caused by splitting the existing meshed elements
mechanics principles should be conducted for better connected at the same failure point (Figure 1a). The failure
insights into potential rock bridge failure. plane is orthogonal to the direction of minor principal stress
The focus of this research is to study the mechanisms at the failure point (Figure 1b). The mesh coordinate and
of rock bridge failure and its time-dependent behaviour. connectivity matrices are subsequently updated to insert
Predictive numerical analyses using the combined Finite- this type of fracture. Note that the intra-element fracture
Discrete Element Method (FDEM) are conducted to generation scheme usually results in poor mesh quality
leading to smaller model time step size and computational softening behaviour when the local stress exceeds the
instability (Elmo et al. 2013). local crack element strength, while the meshed
The inter-element fracture shown in Figure 1c can be triangular/tetrahedral element remains elastic. The
formed along boundaries of meshed elements in the softening behaviour of the crack element is controlled by
direction most favourably oriented to the failure plane. In the values of Mode I (GfI) and Mode II (GfII) fracture
this case, the pattern of inter-element fractures is highly energies. In the numerical program, a heuristic function by
dependent on the mesh topology. FDEM analysis using the Evans and Marathe (1968) based on laboratory results
inter-element fracture generation scheme usually avoids controls crack element post-peak softening behavior. The
instability when dealing with complex model and mesh residual stress on the yielded crack element gradually
geometries. reduces with increasing opening/slipping displacement
between elements. The broken crack element replaces the
yielded crack element to form an open fracture when the
relative displacement between two meshed elements
reaches critical values, and stress reduces to zero.
Using the crack element approach, the fracture
mechanisms controlling rock bridge failure can be
captured, and failure modes (i.e., Mode I, Mode II and
Mode I-II) of generated fractures can be differentiated to
study the mechanisms.

Figure 1. Fracture generation scheme in FDEM (after

Klerck, 2000; Owen et al., 2005; figure redrawn by Lisjak
and Grasselli, 2014): a) failure point and fracture
orientation; b) intra-element fracture; and c) inter-element
fracture.

The commercially available FDEM code Irazu

(Geomechanica Inc 2019) was used for this study. In Irazu,
a continuum body is discretized by several triangular (2D)
or tetrahedral (3D) elements bonded together at their
contacts using four-noded (2D) or six-noded (3D)
quadrilateral cohesive crack elements (Figure 2; Lisjak and
Grasselli, 2014; Mahabadi, 2012). An open fracture is
formed when two elements are fully separated under stress
and yielded crack elements are replaced with broken
elements. Since only inter-element fractures can be
generated in Irazu code, the mesh topology is critical for
capturing realistic fracture patterns.

Figure 3. Constitutive model of a cohesive element in Irazu:

a) Mode I; and b) Mode II. c) mixed Mode I-II (Tatone and
Grasselli 2015, Geomechanica Inc 2019).

3 NUMERICAL SIMULATIONS OF PARUS ROCK: A

CASE STUDY

The Parus rock (also known as Sail rock) is a natural

sandstone monolith of the late Cretaceous located on the
shore of the Black Sea in Krasnodar Krai, Russian
Federation (Elmo et al. 2022). The Parus rock is about
Figure 2. Representation of triangular/tetrahedral elements 20 m long, 25 m high and 1.5 m wide. A naturally formed
connected by crack elements in Irazu (Lisjak and Grasselli circular opening is located several meters above sea level
2014). near the NW corner. Figure 4 shows two visible joint sets
on the SE face. One joint set is subvertical and persistently
Figure 3 shows the constitutive behaviours of crack dipping 80 degrees toward NE. Rock joints in another joint
elements used by the Irazu code. Following fracture set are shorter than the prior set and dip about 30 degrees
mechanics principles, Mode I, Mode II and Mode I-II are toward SW. These two joint sets are approximately
used to describe yielding/failure modes of the crack perpendicular to each other, forming a blocky rock mass.
element in tension, shear and mixed tension-shear, Furthermore, the interaction of non-persistent joints
respectively. The crack element yields and undergoes perfectly shows the existence of rock bridges. The Parus
rock shown in Figure 4 is used as a case study to step size and increase the model runtime (Tatone and
investigate the potential rock bridge failure within a jointed Grasselli 2015). The element size of 0.1 m was found to
rock mass. balance the model runtime and results resolution. The base
of the Parus rock model was fixed in both horizontal and
vertical directions. In the model, only the gravitational force
was applied.
Table 1 lists three scenarios considered to investigate
how rock block strength affects rock bridge failure. Since
not all defects and natural fractures within rock blocks can
be explicitly simulated in the numerical model, the
conventional Hoek-Brown criterion was used to estimate
the strength of the sandstone block at the scale of 1 m2.
The Geological Strength Index (GSI) values of 70 and 65
(case B and C) were assumed for estimating the defected
sandstone block strength. The equivalent cohesion and
friction angle were used for the input parameters of the
model (Elmo et al., 2022).
Table 2 lists the input parameters for the numerical
model based on estimated rock block strength. Note that
the input parameters for the FDEM model listed in Table 2
are micro-properties (for meshed triangular and crack
elements) and must be adjusted. Geomechanica Inc
(2019) suggests that the elastic properties of triangular
elements and strength properties of crack elements can be
selected to be equal to those of sandstone blocks. The
numerical parameter (i.e., penalties) values for crack
elements need to be chosen as multipliers of the elastic
properties of triangular elements. It is suggested that the
normal and fracture penalty values of 10 times the
triangular element Young’s modulus can be used. The
Figure 4. Parus rock at SE face (photo from Google Earth tangential penalty value as a multiplier of triangular
shared under a creative commons license CC2.5). element Young’s modulus is suggested to be a function of
mesh size as follows:
Figure 5a shows the deterministic Discrete Fracture 𝑛 = 15.29𝑥 −1.6 [1]
Network (DFN) based on trace maps sketched from photos where 𝑛 is the multiplier of Young’s modulus and 𝑥 is the
in Figure 4 and studies from Elmo (2023) and Elmo et al. average mesh size.
(2022). The model was then meshed by 15,958 triangular The unconfined compressive tests on a homogeneous
elements with an average size of 0.1 m (Figure 5b). The sandstone model were simulated. The results showed that
mesh size was selected to ensure optimal mesh quality the macroscopic strength and deformation of the
when discretizing complex DFN geometry. A smaller mesh sandstone model match the properties assigned to the
size can be adopted for a better model resolution. mesh element and crack element by using the
However, this would significantly reduce the model time combinations listed in Table 2.

Figure 5. a) Deterministic DFN of Parus rock; and b) FDEM model mesh topology.
Table 1. Rock block properties for three different scenarios 4 SIMULATION RESULTS
in this analysis.
Figure 6 shows the final failure modes and displacement
Model
GSI
Erm crm σtm
φrm (°)
contours for cases A, B and C. For case A, only the
ID (GPa) (MPa) (MPa) structurally controlled failure is captured. The rock blocks
A 100 20.5 10.4 4.4 62 at SW and NE sides slide along the joints while no
B 70 15.1 1.6 0.8 62 significant intact failure and rock bridge failure are
C 65 13 1.2 0.6 62
captured. As the rock block strength decreases (cases B
and C), more fractures are captured near SW and NE
Table 2. Input parameters for FDEM model. flanks leading to more unstable rock fragments. The
degree of fracturing is related to the rock block strength. In
Input parameters Value Unit
the figure, the failure mode of those fractures is mainly
Density ρ 2600 kg/m3
Meshed Young’s modulus (E) Erm Pa Model I (tensile fracture).
triangular Poisson’s ratio (v) 0.25 - Figure 7 shows the failure process, and rock bridge
element Viscous damping failure for case B. Tensile fractures are initiated within the
1 -
factor rock block at the area where the natural rock joints
Friction angle φ φrm ° terminate, then propagate with increasing the model run
Cohesion c crm Pa time (effective time). Initiated fractures coalesce with
Tensile strength σt σtm Pa adjacent induced fractures and pre-existing joints, resulting
Crack
Mode I fracture in rock bridge failure. The rock bridge failure was found in
element (σt/6.88)2/E* N/m
energy GfI
(cohesive
Mode II fracture
the area near SW and NE flanks, where blocks are
for rock 10GfI* N/m exposed to the free surfaces. The rock fragments near SW
energy GfII
block) and NE flanks slide along well-coalesced fractures with
Normal penalty pn 10Erm Pa▪m
Fracture penalty pf 10Erm Pa effective time, leading to potential catastrophic failure. This
Tangential penalty pt 608.7Erm** Pa/m process was found to be time-dependent.
Crack Friction angle φ 30 ° Sensitivity analyses were conducted to investigate how
element Normal penalty pn 1E+11 Pa▪m joint tensile and shear strength affect rock bridge failure. It
(frictional was found that the joint cohesive strength significantly
for rock Tangential penalty pt 1E+10 Pa/m
affects the failure mode. As shown in Figure 8, the fractures
joint)
* Selection of the values is based on laboratory test results by
are formed once the rock blocks near SW and NE flanks
Zhang (2002) and suggestions by Geomechanica (2019). start to slide along the joints when the joint cohesions are
** Selection of the tangential penalty for the cohesive crack 0 and 0.2 MPa. No rock bridge failures are captured if joint
element is based on the suggestion by Geomechanica (2019) to cohesion is increased to 0.5 MPa. The simulations found
ensure the frictional force can be fully mobilized. that the joint tensile strength has minimal effects on rock
bridge failure.

Figure 6. Final failure modes and vertical displacement contours for simulations of cases A, B and C. Note that figures
share the same legend.
Figure 7. Captured failure process for case B. Failure mode legend for fractures is referred to in Figure 6.

Figure 8. Final failure modes of case B with different joint cohesion values. Failure mode legend for fractures is referred
to in Figure 6.

6 ACKNOWLEDGEMENTS
5 DISCUSSION AND CONCLUSION
The authors wish to acknowledge the MITACS grant in
This paper used a two-dimensional numerical model collaboration with the Centre of Innovation in Mineral
implementing FDEM to study the mechanisms and Resource Engineering (CIMRE) and Newcrest Mining
processes of rock bridge failure. It is found that both rock Limited for the financial support provided to this research.
block and joint cohesive strength affect the degree of brittle
fracturing and formation of rock bridges. In addition, the 7 REFERENCES
simulation results showed that rock bridge failure is a time-
dependent process. Tensile fractures are initiated when the Elmo, D. 2023. The Bologna Interpretation of Rock
non-persistent rock joints terminate, then fractures Bridges. Geosciences, 13(2): 33. MDPI.
propagate and coalesce with the nearest stress-induced or Elmo, D., Stead, D., Eberhardt, E., and Vyazmensky, A.
natural fractures forming rock bridges. This process was 2013. Applications of finite/discrete element
not considered in the previous studies using probabilistic modelling to rock engineering problems.
step path failure predictions using the DFN model and International Journal of Geomechanics, 13(5):
measurements of rock bridge percentage for slope stability 565–580. American Society of Civil Engineers.
analysis. Elmo, D., Tasnim, Z., Borgatti, L., and Marcato, G. 2022.
It is suggested that numerical analysis using discrete The metaphysical nature of rock bridges and the
element/combined finite-discrete methods should be used challenge of measuring their conditional
to understand the stability of fractured rock masses better. existence. Tucson, AZ, US.
We caution against historical design methods presented in Evans, R., and Marathe, M. 1968. Microcracking and
the literature based on measuring the so-called rock bridge stress-strain curves for concrete in tension.
percentage since it is impossible to measure entities that Matériaux et Construction, 1(1): 61–64. Springer.
do not exist until failure has occurred. Geomechanica Inc. 2019. Irazu Geomechanical Simulation
Software. Toronto, Ontario, Canada.
Jennings, J. 1970. A mathematical theory for the
calculation of the stability of slopes in open cast
mines. In Planning open pit mines, proceedings,
Johannesburg. AA Balkema Cape Town. pp. 87–
102.
Jennings, J. 1972. An approach to the stability of rock
slopes based on the theory of limiting equilibrium
with a material exhibiting anisotropic shear
strength. In Stability of Rock slopes. ASCE. pp.
269–302.
Jennings, J., and Steffen, O. 1967. The Analysis of the
stability of slopes in deep opencast mines. Civil
Engineering= Siviele Ingenieurswese, 1967(3):
41–54. South African Institution Of Civil
Engineering (SAICE).
Klerck, P.A. 2000. The finite element modelling of discrete
fracture in quasi-brittle materials. PhD Thesis,
University of Wales Swansea.
Lisjak, A., and Grasselli, G. 2014. A review of discrete
modelling techniques for fracturing processes in
discontinuous rock masses. Journal of Rock
Mechanics and Geotechnical Engineering, 6(4):
301–314. Elsevier.
Mahabadi, O.K. 2012. Investigating the influence of micro-
scale heterogeneity and microstructure on
geomaterials' failure and mechanical behaviour.
PhD Thesis.
Munjiza, A.A. 2004. The combined finite-discrete element
method. John Wiley & Sons.
Owen, D.R.J., Pires, F., de Souza Neto, E., and Feng, Y.
2005. Continuous/discrete strategies for the
modelling of fracturing solids. Nato science series
sub III Computer and systems sciences, 194:
230–266.
Tatone, B.S.A., and Grasselli, G. 2015. A calibration
procedure for two-dimensional laboratory-scale
hybrid finite–discrete element simulations.
International Journal of Rock Mechanics and
Mining Sciences, 75: 56–72.
doi:https://doi.org/10.1016/j.ijrmms.2015.01.011.
Terzaghi, K. 1962. Stability of steep slopes on hard
unweathered rock. Geotechnique, 12(4): 251–
270. Thomas Telford Ltd.
Zhang, Z. 2002. An empirical relation between mode I
fracture toughness and the tensile strength of
rock. International journal of rock mechanics and
mining sciences, 39(3): 401–406. Elsevier.
Comparison of geosynthetic reinforced soil
wall solutions using analytical design
methods and numerical modelling
A. Moncada & S. Olivella
Department of Civil and Environmental Engineering, Universitat
Politècnica de Catalunya·BarcelonaTech (UPC), and International
Centre for Numerical Methods in Engineering (CIMNE), Barcelona,
Spain.
I.P. Damians
Department of Civil and Environmental Engineering, Universitat Politècnica de Catalunya·BarcelonaTech
(UPC), International Centre for Numerical Methods in Engineering (CIMNE), and VSL International, Barcelona,
Spain.
R.J. Bathurst
Department of Civil Engineering, Royal Military College of Canada, Kingston, Ontario, Canada.

ABSTRACT
European design standard (prEN-1997 202x) permits the use of suitably verified numerical models for the design of
reinforced soil walls (RSW). The paper first compares three analytical design method outcomes (Coherent Gravity,
Simplified, and Stiffness method) available in US (AASHTO 2020) and Canadian (CSA 2019) design codes. Polyester
(PET) strap reinforcement arrangements from each method were then simulated using a 2D finite element model (FEM)
which considered construction stages and transient compaction conditions. The models for each arrangement were
analyzed using different material factors applied to the soil frictional strength. Serviceability limit states (SLS) and ultimate
limit states (ULS) were evaluating via horizontal wall deformations, soil shear strains, and maximum reinforcement tensile
loads. Numerical results remained within SLS and ULS criteria. The maximum tensile loads from the numerical models
were close to, but lower than the predicted loads using the Stiffness Method, which in turn were lower than the loads
predicted using the Simplified and Coherent Gravity methods.

RÉSUMÉ
La norme européenne de conception (prEN-1997 202x) autorise l'utilisation de modèles numériques dûment vérifiés pour
la conception des murs en sol renforcé (RSW). L'article compare d'abord trois résultats de méthodes de conception
analytiques (gravité cohérente, méthode simplifiée et méthode de rigidité) disponibles dans les codes de conception
américains (AASHTO 2020) et canadiens (CSA 2019). Les arrangements de renfort de sangle en polyester (PET) de
chaque méthode ont ensuite été simulés à l'aide d'un modèle d'éléments finis (FEM) 2D qui a pris en compte les étapes
de construction et les conditions de compactage transitoires. Les modèles pour chaque disposition ont été analysés en
utilisant différents facteurs de matériaux appliqués à la résistance au frottement du sol. Les états limites de service (SLS)
et les états limites ultimes (ULS) ont été évalués par le biais des déformations de la paroi horizontale, des déformations
de cisaillement du sol et des charges de traction maximales des armatures. Les résultats numériques sont restés dans les
limites des critères SLS et ULS. Les charges de traction maximales des modèles numériques étaient les plus proches,
mais inférieures aux charges prédites en utilisant la méthode de rigidité, qui à leur tour étaient inférieures aux charges
prédites en utilisant les méthodes de gravité simplifiée et cohérente.

1 INTRODUCTION approach for walls that fall within and beyond the scope of
current analytical models. The numerical models must
Numerical models have proven to be useful tools to gain include the wall construction stages and account for
insight into the mechanical behaviour and performance of serviceability limit states (SLS) and ultimate limit states
reinforced soil walls (RSWs) constructed with metallic and (ULS).
geosynthetic reinforcement materials (e.g., Damians et al. A numerical model of a RSW requires careful selection
2014; Huang et al. 2009; Yu et al. 2015a; among others). of a single stiffness value for each reinforcement layer that
All design standards provide analytical methods to quantify accounts for continuous sheet reinforcement materials or
and predict the behaviour of RSWs. Typically, numerical discontinuous (strip) reinforcement depending on the type
models are used to design structures that fall outside the of RSW. For steel (inextensible) reinforcement, the
scope of analytical methods (e.g., difficult foundation selection of a single stiffness value is straightforward.
conditions, unusual wall geometries and very tall walls, and However, the material behaviour of extensible polymeric
difficult or extreme loading conditions). The latest revision reinforcement materials, such as polyolefin and polyester
of the European design standard (prEN1997-3 202x) now (PET) materials, is load-, time-, and temperature-
allows numerical models to be used as a primary design dependent (e.g., Bathurst and Naftchali 2021).
Consequently, great care must be taken when choosing a Three analytical methods were used to compute the
unique stiffness value. An accepted method for both maximum reinforcement load in each reinforcement layer
numerical and analytical approaches is to use the stiffness under operational conditions (Stiffness method, Simplified
modulus corresponding to, for example, 2% strain and method, and Coherent Gravity method). The reinforced soil
1000 hours (AASHTO 2020; Allen and Bathurst 2019). This material was assigned a friction angle of 38º, representing
value can be obtained from isochronous stiffness curves a triaxial peak friction angle, no cohesion, and unit weight
based on creep curves constructed from laboratory of 18.9 kN/m3, for each design method.
product-specific testing, or approximations based on group
data for different reinforcement types (Bathurst and 2.1 Stiffness method
Naftchali 2021).
The present study provides the details of the numerical The Stiffness method (Allen and Bathurst 2018) is
models and performance outcomes for three RSW walls empirically based and formulated for design of RSWs
with reinforcement layer arrangements resulting from three under operational conditions. The Stiffness method is
analytical methods that satisfy the recommendations in applicable for inextensible (steel) and extensible
prEN-1997. The numerical modelling was carried out using (geosynthetic) type reinforcement types. In the latest
the 2D finite element software package CODE_BRIGHT edition of AASTHO LRFD Design Specifications (AASHTO
(Olivella et al. 1996). Staged construction was simulated. 2020), the Stiffness method is restricted to extensible
Numerical outcomes for SLS and ULS were compared with (polymeric) reinforcement products only.
analytical design method predictions. ULS numerical The maximum tensile load (Tmax) in each reinforcement
simulations were carried out using the strength reduction layer is computed as:
method.

Tmax = Sv [H Dtmax + (Href / H) S] Ka fb g local c [1]

2 DESIGN CONSIDERATIONS

The reinforced soil wall in this paper is an idealized 10.5 m- Here Sv is the tributary (vertical) spacing of the
high structure with discrete precast concrete facing panels reinforcement layer,  is soil unit weight, H is wall height,
seated on polymeric bearing pads between the horizontal Dtmax is a load distribution factor with depth below the crest
joints, and attached to PET strap soil reinforcement layers. of the wall, Href = 6 m is a reference height value, S is the
The bottom wall facing panel is seated on a concrete equivalent height of average soil surcharge acting on the
levelling pad. Reinforcement layers have a uniform vertical reinforced soil zone, and Ka is the active earth pressure
spacing of 0.75 m. The wall face is embedded to a depth coefficient (Ka = (1−sin )/(1+sin )), where  is the soil
of 0.525 m (i.e., 5% of wall height). The PET strap friction angle. The remaining parameters, fb, g, local and
reinforcement layers have a length of 7.35 m (i.e., 0.7 of c are dimensionless influence factors for wall facing
wall height) and a coverage ratio of 0.32 (i.e., 9 cm-wide batter, global reinforcement stiffness, local stiffness, and
straps, 2 straps per connection, 4 connections per 2.25 m soil cohesion, respectively.
wide panel). A constant (dead load) uniform surcharge of Connection strength, reinforcement tensile strength
25 kPa was placed over the entire width of the backfill soil. and pullout strength are treated as ultimate limit states in
The properties that were varied were stiffness and North American practice. In order to keep the reinforced
ultimate tensile strength (UTS) of the reinforcement. The soil zone at operational (working stress) conditions the
UTS (i.e., grade) of the reinforcement straps were 25, 30, Stiffness method requires that the soil in the reinforced soil
and 40 kN/strap. No reinforcement straps above grade 40 mass not develop a contiguous shear failure zone through
were required. Reinforcement stiffness values were the height of the wall. This criterion is met by limiting the
estimated at 2% strain and 1000 hours using the power law maximum strains in the reinforcement layers wall to a
formulation proposed by Bathurst and Naftchali (2021) for prescribed value. This so-called “soil failure” limit state is
PET straps and group data for different products falling treated as a serviceability limit state.
within the PET strap category. Stiffness values for grades
25, 30, and 40 were 197, 231, and 316 kN/strap, 2.2 Simplified method
respectively. These values are in accordance with
laboratory data from product-specific creep data made The Simplified method appears in Canadian design
available to the authors (Hang-Won and Je-Goo 2020). practice (CSA 2019) and as a permitted legacy method in
The design of RSWs includes assessment of external the USA code (AASHTO 2020). It is applicable for both
and internal stability limit states. External stability limit extensible and inextensible reinforcement types. The
states include overturning, sliding, foundation bearing maximum tensile load in each reinforcement layer is
capacity, and global and compound stability modes of computed as follows:
failure (as detailed in prEN-1997 202x). For internal
stability, the maximum tensile load that each reinforcement
layer must carry is determined. Next pullout, connection Tmax = Sv Ka v [2]
rupture, and reinforcement rupture or over-stressing limit
states must be examined. The present study is focused on
internal stability limit states only.
Here, v is the vertical earth pressure at the reinforcement are met (i.e., no structural collapse occurs), then the limit
layer elevation including the contribution of any uniform state conditions are satisfied. prEN-1997 (202x) defines a
distributed surcharge. material factor of m = 1.25 for ULS conditions. The present
study also uses m = 2.00 to examine the influence of
2.3 Coherent gravity method weaker soil on numerical model outcomes. The backfill
material was assigned a cohesion of 1 kPa to avoid
The third method used to calculate reinforcement numerical convergence problems in near-surface
maximum tensile load is the Coherent Gravity method elements. Consequently, the material factor was applied to
(CGM) which is used for RSWs constructed with the friction angle only.
inextensible steel reinforcement. The maximum tension in
each reinforcement layer is calculated as follows: 2.5 Analytical model designs

In the analytical models the friction angle of the soil was not
Tmax = K Sv [v L / (L – 2 e)] [3] reduced (i.e., m = 1.00). The lowest grade for
reinforcement considered in this study was 25. The
Stiffness method resulted in a design with all grade 25
Here, K is a dimensionless earth pressure coefficient that reinforcement layers. A better optimized design would
varies linearly with depth, starting at K = K0 = 1−sin  at the require lower grade reinforcement strips which are not
top of the wall, to a fixed value of K = Ka at 6 m of depth available from the product line assumed in this study.
and beyond, L is the reinforcement length, and e is the Reinforcement arrangements are summarized in Table 1.
eccentricity caused by the retained soil mass acting at the
back of the reinforced soil mass.
Current design codes, such as CSA (2019) in Canada, Table 1. Reinforcement grade distribution (from bottom to
AASTHO (2020) and BS8006-1 (2016) limit the CGM to top) for each design method.
walls with inextensible reinforcement. Miyata et al. (2018)
reviewed instrumented and monitored PET strap walls Reinforcement Design
constructed with modern PET strap reinforcement Layer/Grade
arrangement method
materials and demonstrated that these structures fall within
1 2 3 4 5 6-14
the extensible geosynthetic category based on back-
calculated global reinforcement stiffness (Sg), which is Stiffness
A 25 25 25 25 25 25
calculated as the sum of reinforcement layer stiffnesses method
divided by the wall height. Nevertheless, their review Simplified
B 40 30 30 30 25 25
recorded that some designers have used the Coherent method
Gravity method for PET strap walls because of their Coherent
relatively high stiffness compared to polymeric sheet C Gravity 40 40 40 30 30 25
reinforcement products. method

2.4 Limit state criteria verifications

3 NUMERICAL MODEL
In the most recent draft of the prEN-1997 (202x), the use
of numerical models is permitted as a primary design Numerical simulations were carried out using the finite
methodology, rather than simply a complementary method element (FE) program CODE_BRIGHT (Olivella et al.
to verify designs using analytical methods. To use 1996). The FE mesh geometry, material zones and
numerical models for RSW design they must be shown to structural components are shown in Figure 1.
have been used for similar problems (i.e., verified). SLS Soil materials (i.e., backfill, front embedment,
and ULS conditions must be assessed at each construction foundation, and interfaces) were modelled using an elastic-
stage, as well as at end of construction conditions. The plastic constitutive model with dilatancy. Dilatancy angles
model must use material property values adjusted by code- were assumed as  =  – 30º. The reinforced soil friction
specified material factors (i.e., non-reduced values for SLS angle of 38º representing a triaxial peak friction angle value
and strength-reduced properties for ULS). Each stage is could be reasonably increased up to 1.1 to 1.2 times due
deemed safe if displacements and reinforcement loads are to the plane-strain conditions assumed in the numerical
within prescribed limits. According to Allen and Bathurst model (Kulhawy and Mayne 1990). However, for simplicity
(2013) a contiguous internal failure mechanism through the and to keep the friction angle below 40 degrees, as
height of the reinforced soil mass in a geosynthetic RSW is recommended in AASHTO (2020), the friction angle of 38
expected to occur when strains in the reinforcement degrees was used for both numerical and analytical
exceed 2-3%. Tolerable horizontal displacements will vary solutions. To simulate the use of lower compaction energy
between different design guidelines (Bathurst et al. 2010). equipment near the facing panels, as recommended in
In this work, a maximum value of 0.5% of the structure construction practice, a lower elastic modulus value was
height was adopted (BS 8006-1 2016). used for the soil within 1 m of the back of the wall facing
For ULS design, a material factor (m) must be applied (see Table 2). The foundation stiffness was selected to
to the soil strength parameters to reduce the soil strength represent a competent and stiff foundation.
(i.e., a reduction of tan() and c). If equilibrium conditions
 the embedded reinforcement elements, are constructed.
The placement of the next facing panel and soil layer is
repeated until the full wall height is achieved. A transient
compaction stress of 5 kPa was applied uniformly over the
soil layers. The front toe embedment was applied once the
wall reached a height of 4.5 m (i.e., three panel-height).
Concrete elements and bearing pads were modelled
using a linear elastic model. Facing panels and the levelling
pad were assigned Young’s moduli of 32 and 25 GPa,
respectively, and Poisson’s ratio of 0.2. Bearing pads
comprised of two or four units per panel width were placed
at the horizontal joints between panels, resulting in pad
layer stiffness of 3.3 and 6.6 kPa, respectively. Two pads
were placed between the top four panels (i.e., above 6 m
of height), and four pads between the bottom three panels
(Damians et al. 2016).

Table 2. Soil and interface property values used in

Figure 1. 2D finite element mesh geometry. numerical model.

Soil-facing and soil-reinforcement interfaces were

Friction angle,

Poisson’ s
angle,  [º]
Cohesion,

Dilatancy

modulus,
modelled using a continuum element approach. This

E [MPa]

ratio [-]
c [kPa]

Young
 [º]
methodology has been successfully used in the past for Material
RSW finite element models when zero-thickness elements
are not readily available (Damians et al. 2021, 2022).
Interfaces are defined using a reduction factor (R i). The
soil-facing interface was assigned Ri = 0.6 applied to the Foundation soil 36 5 6 100 0.3
strength (i.e., tan() and c) and stiffness parameters (i.e., Backfill soil 38 1 8 10 or 20 0.3
E). For the soil-reinforcement interface, the contact area of Soil-facing interface 25.1 0.6 0 4.02 0.45
the reinforcement strip and soil (Astrip), the contact area of Soil-reinforcement
the soil-soil interaction (Asoil), and the total contact area at 34.8 0.89 8 20 0.45
interface
the reinforcement layer elevation (Alayer) were taken into
account to obtain equivalent interface parameters for the
2D numerical model, hence: 4 RESULTS

Figure 2 shows the horizontal outward displacements

Ri = [(Ri(strip) Astrip) + (Ri(soil) Asoil)] / Alayer [4] during and after the construction process using the
reinforcement arrangement A (Stiffness method) in Table
2). From the beginning of construction, horizontal
Here, the soil-soil surface at each reinforcement layer displacements form an outward curved distribution,
remains unchanged (i.e., reduction factor of Ri(soil) = 1) centred at about the middle of the structure (Fig. 2a, 2b),
while the soil-reinforcement contact reduction factor is set and moving to about the bottom third of the wall height as
to Ri(strip) = 0.67. For the soil-reinforcement interface Ri is the structure is completed (Fig. 2c). The application of a
applied to the strength parameters. surcharge results in further horizontal displacements (Fig.
The calculation details for the elastic modulus interface 2d). Maximum horizontal displacements were within 0.19%
parameters in Table 2 can be found in the papers by of the structure height (i.e., 2 mm) after the application of
Damians et al. (2015, 2022) and Yu et al. (2015b). the 25 kPa surcharge. As noted earlier, a maximum
Reinforcement elements were modelled using a linear horizontal displacement of 0.5% of H was assumed to
elastic model, in which straps are represented by a satisfy the SLS criterion. This criterion was met at all
continuous layer with equivalent stiffness considering the construction stages using the three reinforcement layer
reinforcement coverage ratio and the reinforcement arrangements.
stiffness values used in the analytical designs. Global As the material factor is increased, a failure mechanism
reinforcement stiffness values (Sg) ranged from 0.93 to can be observed based on shear strains within the
1.08 MPa and fall within the range of 0.79 to 1.96 MPa for reinforced soil mass (Fig. 3) starting near the toe of the
typical instrumented PET strap walls reported by Miyata et vertical wall. As the material factor increases, shear strains
al. (2018). increase to approximately 1.2% at the bottom of the
In the numerical model, the simulated construction of reinforced soil zone. In this region a linear failure zone can
the foundation and levelling pad is carried out first. Second, be imagined consistent with expectations using active
the first concrete panel is constructed. Third, the first 1.5 m earth pressure theory applied to extensible reinforced soil
layer of reinforced and retained backfill material, including masses (dashed line in Fig. 3). Failure is expected to occur
 for the three analytical design methods. The Stiffness
method reinforcement arrangement gives much lower
reinforcement tensile loads in all reinforcement layers. The
reinforcement arrangements from the Simplified and
Coherent Gravity methods result in higher predicted loads,
particularly at the base of the structure.
Miyata et al. (2018) compared predicted loads using the
same three design methods with measured values from
full-scale instrumented wall structures. Their results
(a) (b) showed that the Simplified method and the CGM lead to
safe designs by over-predicting reinforcement loads,
particularly for the lower layers, and most noticeable using
the CGM. On the other hand, the Stiffness method gave
more accurate load predicts when compared to measured
values. The numerical model results using the same
reinforcement scheme as the Stiffness method can be
seen to fall below the values from the analytical Stiffness
method. It is interesting to note that the linear trend over
approximately the top half of the wall appears in both the
analytical and numerical results. However, over the bottom
half of the wall the numerical results do not follow the
constant load trend predicted by the analytical model. In
(c) (d) fact, the K-stiffness method (Allen et al. 2003; Bathurst et
al. 2005) which preceded the current Stiffness Method
Figure 2. Horizontal displacements for (a) 3-, (b) 5-, and recommended that the load distribution function (Dtmax in
(c) 7-panel height, and (d) after surcharge is applied using Eq. 1) follow a trilinear distribution consistent with observed
reinforcement arrangement A from the Stiffness method (measured) loads that were attenuated near the bottom of
design and SLS condition (m = 1) as per prEN-1997 the walls due to the contribution of the foundation to carry
(202x). Note: scale legend is [m]. some of the lateral soil load acting on the wall facing. This
leg of the Dtmax distribution was discontinued for the current
Stiffness method to ensure that there is enough
when shear strains greater than 2% form a contiguous reinforcement load capacity over the bottom of the wall
zone through the reinforced fill. For the ULS limit state
condition (m = 1.25), shear strains do not exceed 1.5%.
For a material factor of m = 2.00, higher shear strains in
the range of 1.25 to 1.5% are present but a contiguous
failure mechanism has not developed. While not shown
here, the tensile strains in the reinforcement were also well
below values of 2-3% that are symptomatic of a contiguous
shear zone through the reinforced soil zone as explained
earlier. Not unexpected, the largest shear strains occur in
the bottom third of the wall where horizontal wall
displacements are largest.
Figure 4 shows the horizontal displacement profile of
the wall facing with increasing material factors applied to
tan(). Results are shown using the reinforcement design m = 1.00 m = 1.25
with arrangement A. The ULS numerical simulation using
m = 1.25 results in a slight increase in horizontal
displacements with respect to the base case (m = 1.00),
nevertheless, displacements are within the SLS
displacement criterion. To encourage the development of
possible failure mechanisms in the reinforced soil mass,
the material factor was increased to m = 2.00. The result
was further outward displacement as great as 0.28% of the
wall height. As before, there is a clear bulging shape over
the bottom third of the structure, with the two bottom panels
tilting the most of all panels. Negligible displacements were
m = 2.00
observed at the foot of the wall (i.e., 3 mm), which can be
attributed to the front embedment acting as a horizontal
Figure 3. Numerical model shear strains after surcharge
restraint.
application using different material factor values applied
Figure 5 shows the maximum tensile load in each
to the soil friction angle together with reinforcement
reinforcement layer using the reinforcement arrangements
arrangement A.
Figure 4. Horizontal displacement profiles at the wall Figure 6. Maximum reinforcement tensile loads from
facing for numerical models with different material factor numerical model for (a) SLS (m = 1) and (b) ULS (m =
values applied to the soil friction angle and using the 1.25) cases.
reinforcement arrangement A from the Stiffness method
design. Note: the datum for this plot is the toe of the wall
at the end of construction. Stiffness, Simplified and Coherent Gravity method
designs). As expected, larger differences appear in the
bottom half of the structure where the reinforcement layer
should a compound failure mechanism develop through the grades vary the most between model arrangements.
reinforced soil zone close to the foundation. Higher grade reinforcement layers generate higher loads
Figure 6 shows maximum tensile loads from numerical because they are stiffer. Increasing the material factor to
models considering SLS (m = 1) and ULS (m = 1.25) the value required by prEN-1997 results in a detectable but
conditions. The top seven layers for each model case are small increase in reinforcement tension, particularly in
very similar (i.e., using reinforcement arrangements for layers 2 to 7. The differences are not of practical concern
for design. The lowest reinforcement UTS is approximately
25 kN/strip. This means that less than 10% of the
unfactored rupture capacity of the reinforcement layers is
mobilized and thus the ultimate tensile strength limit state
is not exceeded. Assuming an overall reduction factor RF
= 3 applied to ultimate tensile strength (Tult) to account for
creep, durability and installation strength reduction
mechanisms as in North American practice, the allowable
tensile strength (Tal = Tult/RF) is also not exceeded.
Figure 7a shows the magnitude and distribution of
maximum loads in each reinforcement layer using
reinforcement arrangements A, B and C and material factor
m = 2.00 applied to tan  of the soil. Even for this larger
material factor, reinforcement loads are less than the
computed values using the analytical Stiffness method.
The differences between data sets are less than 1.0
kN/strap, which is judged to be negligible for practical
design purposes.
Figure 7b shows the location of the maximum tensile
load in each reinforcement layer. Superimposed on this
figure are the assumed failure surfaces using the three
analytical methods found in design codes. The Simplified
and Stiffness methods are based on active earth pressure
theory with the failure surface propagating up from the toe
Figure 5. Calculated maximum reinforcement tensile of the wall at 45º +  / 2 from the horizontal. The bilinear
loads using the three analytical design methods and the curve is used for inextensible (steel) reinforcement and the
numerical model results with unfactored soil strength (m Coherent Gravity method. The latter can be argued to be a
= 1.00) and reinforcement arrangement A.
 geotechnical structures. The present study describes an
initial effort to examine whether or not selected limit states
that appear in the latest European design guidelines are
satisfied using the results of a numerical model. The
numerical model uses reinforcement arrangements and
PET strap reinforcement materials determined from
analytical models using the Stiffness method, Simplified
method and the Coherent gravity method (CGM). Staged
construction was used in each simulation including the
influence of transient compaction equipment loading.
Serviceability limit state (SLS) conditions were verified
using unfactored soil shear strength (material factor m =
1.00). Ultimate limit state (ULS) conditions were checked
with a soil strength reduction material factor of m = 1.25
and 2.00. The main findings are as follows:
(a) (b)
• For the reinforcement grade range considered
Figure 7. (a) Maximum tensile load in each (minimum grade of 25), the Stiffness method
reinforcement layer, and (b) maximum load location provided the most optimized design (lowest
versus normalized distance from the wall face using m maximum tensile load in each layer), with all layers
= 2 and the three reinforcement arrangements. of grade 25. The Simplified method and CGM
required higher grade reinforcement for the bottom-
most layers.
better fit for all three reinforcement arrangements using the • Numerical model results were in closest agreement
results of the numerical model. The location of these with the Stiffness method predictions for the
internal failure surfaces is only of interest to analytical maximum reinforcement load at end of construction.
methods because they are used to calculate the • No excessive deformations were observed at any
reinforcement embedment length in the analytical method construction step for the three numerical models,
calculations for the pullout limit state. and thus the numerical model satisfied SLS criteria
In compliance with prEN-1997 guidelines for numerical (i.e., verification). Horizontal displacements
modelling, structure instability was not observed using m = remained less than 0.25% of the structure height at
1.25 for ULS checks for all three cases. Increasing this every construction stage. Maximum reinforcement
material factor to m = 2.00 resulted in greater horizontal tensile loads for all three reinforcement
displacements, shear strains, and reinforcement loads, but arrangements did not exceed the values computed
model stability was not compromised. using the analytical Stiffness method.
It should be noted that global instability and bearing • No structural failure or numerical issues were
capacity or sliding failure were not observed. This is observed using a material factor of 1.25; thus, the
ascribed to the high rigidity of the foundation (E = 100 numerical model outcomes satisfied code-specified
MPa). ULS criteria. Reinforcement loads did not exceed
Horizontal displacements at the base of the wall were 10% of the ultimate tensile strength (UTS), meaning
approximately 3 mm and were similar for all reinforcement that reinforcement rupture is not expected.
arrangements and m cases; thus, the limit state for an Increasing the material to m = 2.00 resulted in
external base sliding failure mechanism was not a concern. incrementally larger horizontal displacements and
This can be attributed to the front foot embedment, which shear strains, but not enough to generate failure
includes the levelling pad and part of the first segmental mechanisms.
panel, whose passive resistance act as a restraint at the • The onset of a potential reinforced soil failure
base of the wall. mechanism in proximity to the toe of the wall was
Overturning about the toe was not observed which is, observed in the numerical model using the highest
regardless, not unexpected for relatively flexible
material factor m = 2.00. However, the distribution
geosynthetic reinforced soil walls. The European standard
of maximum soil shear strains and the peak strain
requires the verification of all failure modes, while, in
in the reinforcement layers were not great enough
AASHTO (2020) and CSA (2019) codes in North America,
to suggest ULS conditions for the internal stability
the external stability overturning limit state is not
limits states investigated.
considered for design. The local horizontal wall
• No evidence of global, compound or external
displacements were small and thus an internal wall face
stability failure mechanisms was observed. This is
bulging failure limit state was not a concern.
largely attributed to the large foundation stiffness
and the front footing embedment.
5 CONCLUSIONS
Reinforced soil wall structures are large deformation
problems, with more than one potential failure mechanism
The most recent revision of prEN-1997 allows the use of a
(limit state). Consequently, all potential limit states must be
verified numerical model to be used for the design of
investigated. Numerical models, as permitted by prEN-
1997 for design, if properly used and verified, facilitate CSA. 2019. Canadian Highway Bridge Design Code.
design for both serviceability and ultimate limit states. CAN/CSA-S6-19. Canadian Standards Association
Likewise, where project conditions fall within the scope of (CSA), Mississauga, Ontario, Canada.
analytical methods, such as the Simplified, Stiffness and Damians, I.P., Bathurst, R.J., Josa, A., and Lloret, A. 2014.
Coherent Gravity methods, numerical models are a useful Numerical study of the influence of foundation
complementary tool to gain insight on wall performance compressibility and reinforcement stiffness on the
features that cannot be easily evaluated using these behavior of reinforced soil walls. International Journal
analytical models (e.g., wall deformations and strains). of Geotechnical Engineering, 8(3), 247-259.
Damians, I. P., Bathurst, R. J., Lima, J., Lloret, A., & Josa,
A. 2015. Numerical study of the use of actively-
ACKNOWLEDGEMENTS tensioned polymeric strips for reinforced soil walls.
Geotechnical Engineering for Infrastructure and
The authors wish to thank Aaron Kim from GECO Industrial Development. January 2015, 3833-3838.
(Korea, Rep) for providing data for polymeric straps Damians, I. P., Bathurst, R. J., Lloret, A., and Josa, A.
(FASTEN products) from reliability assessment testing 2016. Vertical facing panel-joint gap analysis for steel-
records. The authors wish to acknowledge the support of reinforced soil walls. International Journal of
the Department of Civil and Environmental Engineering Geomechanics, 16(4), 04015103.
(DECA) of the Universitat Politècnica de Damians, I.P., Bathurst, R.J., Olivella, S., Lloret, A., and
Catalunya·BarcelonaTech (UPC) and the International Josa, A. 2021. 3D modelling of strip reinforced MSE
Centre for Numerical Methods in Engineering (CIMNE) and walls. Acta Geotechnica, 16(3), 711-730.
the funding received from the Spanish Ministry of Economy Damians, I.P., Olivella, S., Bathurst, R.J., Lloret, A., and
and Competitiveness through the “Severo Ochoa Josa, A. 2022. Modeling soil-facing interface interaction
Programme for Centres of Excellence in R&D” (CEX2018- with continuum element methodology. Frontiers in Built
000797-S-20-4). Environment, 8, 842495-1.
Huang, B., Bathurst, R.J., and Hatami, K., 2009. Numerical
study of reinforced soil segmental walls using three
REFERENCES different constitutive soil models. Journal of
Geotechnical and Geoenvironmental Engineering 135
AASHTO. 2020. LRFD Bridge Design Specifications, 9th (10), 1486–1498.
Ed. American Association of State Highway and Hang-Won, C. and Je-Goo, J. 2020. Reliability assessment
Transportation Officials (AASHTO), Washington, DC, of polymeric straps (FASTEN FS E) for soil
USA. reinforcement. M213-19-15189. GECO Industrial Co.,
Allen, T.M. and Bathurst, R.J. 2013. Comparison of Ltd. FITI Testing and Research Institute, Chungbuk,
working stress and limit equilibrium behavior of Korea.
reinforced soil walls. In Sound geotechnical research to prEN 1997-3. 202x. Eurocode 7: Geotechnical Design —
practice: Honoring Robert D. Holtz II (pp. 499-513). Part 3: Geotechnical Structures. F.E. with agreed CRs
Allen, T. M. and Bathurst, R.J. 2018. Application of the (v2022:5). Technical Committee CEN/TC 250
simplified stiffness method to design of reinforced soil “Structural Eurocodes”.
walls. Journal of Geotechnical and Geoenvironmental Miyata, Y., Bathurst, R.J., and Allen, T. M. 2018. Evaluation
Engineering, 144(5), 04018024. of tensile load model accuracy for PET strap MSE
Allen, T.M. and Bathurst, R.J. 2019. Geosynthetic walls. Geosynthetics International, 25(6), 656-671.
reinforcement stiffness characterization for MSE wall Olivella, S., Gens, A., Carrera, J., and Alonso, E.E. 1996,
design. Geosynthetics International, 26(6), 592-610. 'Numerical Formulation for a Simulator
Allen, T.M., Bathurst, R.J., Walters, D.L., Holtz, R.D. and (CODE_BRIGHT) for the Coupled Analysis of Saline
Lee, W.F. 2003. A new working stress method for Media '' Engineering Computations, 13(7), 87-112.
prediction of reinforcement loads in geosynthetic walls. Kulhawy, F.H. and Mayne, P.W. 1990. Manual on
Canadian Geotechnical Journal, 40(5), 976-994. estimating soil properties for foundation design. Report
Bathurst, R.J., Allen. T.M. and Walters, D.L. 2005. EL-6800, Electric Power Research Institute (EPRI).
Reinforcement loads in geosynthetic walls and the case Palo Alto, California, 306p.
for a new working stress design method. Geotextiles Yu, Y., Bathurst, R.J., and Miyata, Y. 2015a. Numerical
and Geomembranes, 23(4), 287-322. analysis of a mechanically stabilized earth wall
Bathurst, R.J., Miyata, Y., and Allen, T.M. 2010. Facing reinforced with steel strips. Soils and Foundations,
displacements in geosynthetic reinforced soil walls. In 55(3), 536-547.
Earth Retention Conference 3 (pp. 442-459). Yu, Y., Damians, I.P., and Bathurst, R.J. 2015b. Influence
Bathurst, R.J and Naftchali, F.M. 2021. Geosynthetic of choice of FLAC and PLAXIS interface models on
reinforcement stiffness for analytical and numerical reinforced soil-structure interactions. Computers and
modelling of reinforced soil structures. Geotextiles and Geotechnics 65, 164-174.
Geomembranes, 49(4), 921-940.
BS 8006-1: 2010 + A1. 2016. Code of practice for
strengthened/reinforced soils and other fills. BSI, Milton
Keynes, UK.
Instrumentation and Deformation
Measurement of Charlotte County Culvert
No. 2. Using Shape Array Technology
Mohammad Rezania, Othman Nasir, Arun Valsangkar
University of New Brunswick, Fredericton, New Brunswick,
Canada
Joe MacDonald
NBDTI, Fredericton, NB, Canada

ABSTRACT
A reliable technique for the measurements of short and long term deformation response of corrugated steel culvert pipes
(CSCP) is an essential tool to assess the structural integrity and condition of CSCP. Shape Array method is one of the
promising techniques that has the potential for CSCP deformation monitoring. In this work, the load-deformation response
of single CSCP is experimentally investigated with Shape Array system during a full-scale load test performed on an
existing CSP on Route 127 Culvert No.2, in Charlotte County, southwest New Brunswick. The results showed that Shape
Array method was capable of providing accurate and robust monitoring .The deformations were also independently
monitored using LiDAR. The Shape Array measurements agreed with deformation data collected using the LiDAR
technique.

RÉSUMÉ (Change the French version as noted above)

Une technique fiable pour mesurer la réponse aux déformations à court et à long terme des tuyaux de ponceau en acier
ondulé (CSCP) est un outil essentiel pour évaluer l'intégrité structurelle et l'état des CSCP. La méthode de l'array de formes
est l'une des techniques prometteuses qui présente un potentiel pour la surveillance des déformations des CSCP. Dans
cette étude, la réponse charge-déformation d'un seul CSCP est examinée expérimentalement à l'aide du système de
l'array de formes lors d'un essai de charge à grande échelle réalisé sur un CSP existant sur le ponceau n°2 de la route
127, dans le comté de Charlotte, au sud-ouest du Nouveau-Brunswick. Les résultats ont montré que la méthode de l'array
de formes était capable de fournir des capacités de surveillance précises et robustes. Les mesures de l'array de formes
étaient en accord avec les données de déformation collectées à l'aide de la technique LiDAR.

1 INTRODUCTION content, soluble salts, and oxygen content, affect how long
steel pipe will last in soil. The majority of soils have a pH
For more than a century, corrugated steel pipe (CSP) has between 6.0 and 8.0, which is recognized as neutral and is
been the preferred option for engineers and used in many an acceptable range for steel pipe durability, but acidic
projects. With its wide variety of coatings, corrugation soils, which are more frequently found in areas with
patterns, and wall thickness, CSP offers dependable considerable rainfall, have lower pH values and tend to be
durability. The flexibility and value of CSP are unmatched more corrosive. The amount of moisture in the soil can also
by any other pipe material on the market today (Wagener have an impact on CSP's durability. Moreover, Granular
and Leagjeld, 2014). soils that allow the culvert’s backfill to drain quickly
Corrugated steel pipe offers a variety of environmental increase durability, and low moisture content soils are often
advantages, such as lower greenhouse gas emissions, not corrosive to CSP. High clay content soils are more
lower acid rain, and a reduction in the depletion of corrosive than well-drained soils because they tend to hold
resources required in manufacturing, such as water and water for longer periods of time (Wagener and Leagjeld,
electricity. 2014). All these factors cause the CSP to degrade and
CSP is produced in a variety of forms and sizes deform, effecting the stability and integrity of CSP structure
(National Corrugated Steel Pipe Association, 2018), with Wagener and Leagjeld, 2014; García and Moore, 2015).
diameters ranging from 15 centimeters to more than 15 A common technique for determining the structural
meters and up to 30 meters of fill height. CSP is a integrity of CSP is to perform full scale load test of the
reasonably priced product that can be installed quickly culverts using a loaded truck, moving or stationary. The
even during extreme weather conditions because it is not stationary load test requires positioning a truck with a
much susceptible to climatic variability or moisture. known load on top of the culvert at specific spots, and for
A material's durability refers to its ability to resist moving loading tests, it involves moving the truck over the
against deterioration brought on by corrosion, abrasion, culvert at a predetermined speed. The performance of the
applied loads, and installation techniques. A number of culvert in responding to the load test is then measured and
factors, such as soil resistivity, acidity (pH), moisture evaluated using variety of tools, like displacement
transducers, fibre optics, shape array (SAA) or light
detection and ranging laser scanning survey method
(LiDAR) cameras.
The main objective of this research was to determine
the effectiveness of using SAA and LiDAR survey methods
to monitor the performance of CSP. In this project, SAA
and LiDAR were used to record displacements of Culvert
No. 2 on Route 12, southwest New Brunswick.

2 SITE DESCRIPTION

The investigated (CSP) culvert is in Charlotte County,

southwest New Brunswick region, approximately 100 km
southwest of the City of Fredericton as shown in Figure 1.
This culvert was selected for this study by New Brunswick
Department of Transportation and Infrastructure (NBDTI)
as the culvert was under an ongoing rehabilitation using
mortar lining system (NBDTI, 2022).

Culvert No. 2 crosses Route 127 with approximately 45˚ Figure 1. Site Location (Google Earth Image)
degree angle, with a diameter of 1.55 m and an invert
length of approximately 34 m and a slope of 4.1% (north to The streambed to roadbed depth at the road centerline
south). is about 4.4 meters, with a depth of soil above the culvert
of up to 2.8 meters as shown in Figure 2 (NBDTI, 2022).

Figure 2. Plan and Section of Culvert No. 2 on Route 127 (NBDTI, 2022)
 The culvert corrugate steel plate shown in Figure 3, has types of SAA deformation sensors, including wire-based
thickness of 5 mm, and the corrugations are defined by a sensors, fiber optic sensors, and piezoelectric sensors.
pitch of 125 mm (5˝) and a depth of 25 mm (1˝). Figure 3 Each type has its own advantages and disadvantages,
shows that the culvert faces a significant corrosion at the depending on the specific application. One of the main
base of the culvert. limitations of shape array deformation sensors is their
susceptibility to damage or failure under certain conditions,
such as high strain rates or extreme temperatures. In
addition, the accuracy of the measurements may be
affected by the position and orientation of the sensors.

Figure 3. CSP details at the south exit of the culvert. Figure 4. Shape Array (Measurand, 2023)

In this work, the SAA segments (Measurand, 2023) are

installed under the right lane of the road from north to
3 EXPERIMENTAL PROGRAM south, 14.75 meters away from northern entrance of the
culvert as shown in Figure 5 (a) in circular orientation inside
The experimental program methodology included using the the pipe as shown in Figure 5 (b). Each segment of SAA
shape array system in deformation measurements under records coordinate as a signal and the data is transferred
load test, and perform a validation measurements using to the data acquisition system (DAS) shown in Figure 5 (c).
LiDAR scanning. This methodology was achieved by doing
the following main tasks:

- Installation of the Shape Array System

- Performing a Load Test
- Completing a LiDar Scanning

3.1 Shape Array Instrumentation

Shape Array sensors (SAA®, Measurand, 2023), shown in

Figure 4, are a geotechnical sensors can measure
deformations for soil and rock such as short- and long-term
monitoring of the stability of embankments, tunnels, and
foundations monitoring (Measurand, 2023; Dasenbrok et
al, 2011; Yan et al, 2021; Lipscombe, 2014; Pitilakis et al,
2013). SAA composed of a series nodes connected by
flexible elements to track change in shape of the
surrounding material. The use of SAA in geotechnical Figure 5. Shape array instrumentation a) location of SAA
engineering has increased in the recent due to affordability from the north end of the culvert, b) SAA segments, c) data
offered by new manufacturing techniques development in acquisition system (DAS) box.
wireless communication technology. There are several
 All segments and conduit contain extended cable of position 1, NR: North direction, Rear axle, SF: South
SAA, tightened to the culvert by pipe strap and wall screws. direction, Front axle, SR: South direction, Rear axle.
Conduits contain extended cable of Shape Array, tightened
to the crown of the culvert and extended to the data
acquisition system (DAS) box. SAA had twenty-two
segments, with segment 1 to segment 21 shown in Figure5
(b).

3.2 LiDAR Instrumentation

In addition to the SAA monitoring, , a 5-meter section of the

culvert's length was scanned with Laser Scanning survey
method LiDAR by Atlantic Data Acquisition Services Inc.
This 5-metre section was located 2.5-metres on either side
of the Centre of the West (south bound) lane of Route 127.
A 360-degree 2D LiDAR scan was taken along the 5-meter
section in increments of 30 centimeters, producing 18
cross-sectional scans that were used to determine cross
sections along the culvert's central section as shown in
Figure 6 (d and e). The LiDAR scanner was pulled through
the 5-metre section along a sliding rail system as seen in Figure 7. Loading Test Setup a) Loading test truck, b, c and
Figure6 (d and e). The rail system was bolted to a d) location of the SAA.
rectangular aluminum tube and then fastened to two
sawhorses. The two sawhorses had plastic containers with
approximately 30 kilograms of water on each platform
stabilizing them. A tripod was affixed near the inlet to the
CSP (Figure 2d) and used as a reference point to ensure
each full scan was at the same Z axis location. The fibre
glass pull rods were marked and stopped at 30 centre
increments (Figure 2d).

Figure 6. LiDAR instrumentation, including, LiDAR rail

setup, 360-degree LiDAR scanner.

3.3 Loading Test

The static load testing was undertaken on Wednesday

November 3rd, 2021. The test vehicle was a dump truck Figure 8. Truck locations of the loaded truck positions
shown in Figure 7 (a) weighing 14,340 kg ,supplied by at both directions (north and south) (Alderman, 2012).
NBDTI. The truck was loaded with 6,420 kg of gravel,
bringing its total weight to 20,760 kg. The truck was aligned 4 RESULTS AND DISCUSSION
to the marks on the ground showing the location of the
installed SAA as shown in Figure 7 (b, c, and d). 4.1 SAA Results
Twenty static load tests were completed, ten times in
each traffic direction a at the wheel's locations shown in Figure 9 shows the vertical displacement, and Figure 10
Figure 8 (Alderman, 2012). The truck was positioned over shows the horizontal displacement data obtained from
each location for about 20 seconds before moving to the SAA in the culvert resulted from loading test in the south
next. In Figure 8, NF1: North direction, Front axle and bound of the road over the culvert. The most critical
segments regarding vertical displacement in Figure 9 are
the segments that are mounted on the crown of the culvert Figure 10. Horizontal cumulative displacement in south
(segment 6, 14 and 16, refer to Figure 5b). direction before lining application.

4.2 LiDAR Results

VERTICAL CUMULATIVE DISPLACEMENT Figure 11 shows 360 degree 2D LiDAR scan captured
0.035 along the 5 metre section, showing 18 cross-sectional
DISPLACEMENT (mm)

scans and Figure 12 shows a zoomed in details of the

0.025 scanned points coordinates change at NR3, NR5, and NR5
0.015 loading scenarios compared to the reference initial case.
0.005 There is a negligible change in coordinates, reflecting a
negligible deformation. The results agree with the SAA
-0.005
1 3 5 7 9 11 13 15 17 19 21 observation in term of small deformation.
-0.015
-0.025
SEGMENT NUMBER
SF1 SF2 SF3 SF4 SF5
SR1 SR2 SR3 SR4 SR5

Figure 9. Vertical cumulative displacement in south

direction before lining application.

The most important segments for horizontal

displacement shown in Figure 10 are the segments which
are mounted close to the springline along both sides
(segment 11 and 1, refer to Figure 5b). The maximum
vertical displacement in south direction static loading test
were measured at segments 14, 15 16 and 17, while the
maximum horizontal displacements were measured at
segments 9 to 13. The maximum displacement's location
agrees with the expected location (Sezen, 2008). As noted
in Figures 9 and 10, the maximum deflection values in
vertical direction were 0.027 mm and .06 mm in horizontal
direction. These small, measured values are consistent
with the 2.8 m soil cover and the fact that structural integrity Figure 11. 18 LiDAR scanned cross-sections of a 5-metre
of the pipe is not compromised in spite of noticeable section of CSP.
corrosion at the bottom of the pipe.

HORIZONTAL CUMULATIVE
DISPLACEMENT
0.07
DISPLACEMENT (mm)

0.05

0.03

0.01

-0.01 Figure 12. Cross-sectional LiDAR scan – The Green points

1 3 5 7 9 11 13 15 17 19 21
are the unloaded baseline scan while the other coloured
points represent the three statically loaded positions (NR3,
-0.03
SEGMENT NUMBER NR5, and NR5).

5 CONCLUSIONS AND RECOMMENDATIONS

SF1 SF2 SF3 SF4 SF5
The field testing was successfully accomplished in line with
SR1 SR2 SR3 SR4 SR5
the project's criteria, which included measurements of the
culvert deformation using SAA. The following conclusions
of the field testing are presented:
 6 REFERENCES
- The segments positioned near the springline
along both sides are the most important in terms Alderman, A. (2012). Field Testing of the Rehabilitated
of horizontal cumulative displacement (segment Kelly Creek No. 2 Corrugated Steel Culvert (M. Eng
3, 10, 11 and 20). Report). Co-Supervisor Dr. Eldo Hildebrand.
- The maximum vertical displacement for north and Dasenbrock, D., Abdoun, T., & Bennett, V. (2011). Real-
south lane loading is 0.04 mm at NR5 position and time structural health monitoring of landslides and
0.03 mm at SR5 position, respectively. For north geotechnical assets with ShapeAccelArrays. In J. Han
direction loading, the maximum horizontal & D. E. Alzamora (Eds.), Proceedings of Geo-Frontiers
displacement is 0.06 mm at NR5 and NR4 2011: Advances in Geotechnical Engineering (pp.
positions, while it is slightly larger than 0.06 mm 1585-1594). Reston, VA: American Society of Civil
at SR5 and SR4 positions in south direction. Engineers.
Maximum vertical displacement in the García, D. B., & Moore, I. D. (2015). Performance of
aforementioned positions occurred in both deteriorated corrugated steel culverts rehabilitated with
directions at segments 7 and 15. sprayed-on cementitious liners subjected to surface
- The maximum horizontal cumulative loads. Tunnelling and Underground Space Technology,
displacement at those positions happened at 47, 222-232.
segments number 10 and 11 in both directions. Google earth maps.
The key finding of this report was that the SAA Lipscombe, R., Carter, C., Perkins, O., Guerrero, S., &
technology is capable of measuring a continuous Thurlow, P. (2014). The use of Shape Accel Arrays
deformation profile of culverts. (SAAs) for measuring retaining wall deflection. In
- The static loading tests with a loaded truck did not Proceedings of the 39th Annual Conference on Deep
result in any noticeable deformation at any CSP Foundations (15 p.). Hawthorne, New Jersey: Deep
spots, but ShapeArray technology has Foundations Institute.
demonstrated that even the smallest Measurand. (2023). SAAV: Shape Array for Structural
deformations can be continuously recorded by Monitoring. Retrieved from
this device. ShapeArray instrumentation results http://saav.measurand.com/home
are very accurate and reliable by comparing them National Corrugated Steel Pipe Association. (2018).
to LiDAR data, and that there is good agreement Corrugated Steel Pipe Design Manual (2nd ed.).
in deformation measurements between SAA and Dallas, TX: National Corrugated Steel Pipe
LiDAR. Association.
- The SAA was easy to install in the field including New Brunswick Department of Transportation and
the data logger system. The deformations Infrastructure. (2022). Personal communications.
measured by SAA were in close agreement with Pitilakis, K., Anastasiadis, A., Pitilakis, D., & Rovithis, E.
LiDAR survey undertaken. This field case (2013). Full-scale testing of a model structure in
demonstrates the potential for use of SAA Euroseistest to study soil-foundation-structure
technology in future projects. interaction. Geotechnical Research, 50(1), 12-14.
Sezen, H. (2008). In-Situ Load Testing of Corrugated Steel
Utilizing SHAPEARRAY geotechnical instruments Pipe-Arch Culverts.
inside culverts offers numerous advantages for measuring Wagener, B. D., & Leagjeld, E. E. (2014). Culvert Repair
strains. SHAPEARRAY's high-precision deformation Best Practices, Specifications and Special Provisions −
monitoring capabilities enable detecting and analyzing Best Practices Guidelines. Minnesota Department of
sensitive changes in the culvert's structural behaviour, Transportation.
such as minor settlements, displacements, or Yan, R., Take, W. A., Hoult, N. A., Meehan, J., & Levesque,
deformations. This level of sensitivity is particularly C. (2021). Evaluation of shape array sensors to quantify
valuable in assessing the long-term performance and the spatial distribution and seasonal rate of track
stability of culverts, especially in challenging environments settlement. Transportation Geotechnics, 27, 100487.
or regions prone to ground movements. By accurately
capturing even small strains, engineers and geotechnical
experts can gain deeper insights into the culvert's
behaviour under various loading conditions, ensuring its
design meets structural integrity, safety standards and
optimizing maintenance efforts. In future works,
SHAPEARRAY's integration in culverts will prove
invaluable in enhancing the resilience of infrastructure
projects, enabling real-time monitoring and timely
intervention to prevent potential failures, and ultimately
contributing to safer and more durable geotechnical
engineering structures.
Numerical analysis of temperature
development around the storage rooms of a
DGR, and the influence of facility geometry.

Caleb Coulson, Othman Nasir

Department of Civil Engineering – University of New Brunswick,
Fredericton, New Brunswick, Canada

ABSTRACT
The deep geological repository (DGR) is an underground facility, constructed in a stable host rock formation to permanently
host nuclear waste, including heat generating spent nuclear fuel. This heat will radiate into the surrounding environment
and increase the temperature in the host rock. This temperature rise can lead to safety and serviceability issues. To
minimize this concern, geometry of the facility should be optimized such that temperature rise within the host rock from the
spent fuel source is minimized, while minimizing the total footprint of the facility. To investigate this, numerical models of a
DGR were constructed, and analyzed utilizing FLAC3D. Temperature development in the immediate vicinity around the
storage rooms was investigated by the use of a near field model in which the storage room spacing was varied to find the
optimum design.

RÉSUMÉ
Le dépôt géologique profond (DGR) est une installation souterraine construite dans une formation rocheuse stable pour
accueillir de manière permanente les déchets nucléaires, y compris les combustibles nucléaires usés générant de la
chaleur. Cette chaleur se propage dans l'environnement environnant et augmente la température dans la roche hôte.
L'élévation de la température peut entraîner des problèmes de sécurité et de fonctionnalité. Afin de minimiser cette
préoccupation, la géométrie de l'installation doit être optimisée de manière à réduire au minimum l'élévation de la
température dans la roche hôte due à la source de combustible usé, tout en minimisant l'emprise totale de l'installation.
Pour étudier cela, des modèles numériques d'un DGR ont été construits et analysés en utilisant FLAC3D. Le
développement de la température dans les environs immédiats des salles de stockage a été étudié à l'aide d'un modèle
de champ proche dans lequel l'espacement des salles de stockage a été modifié pour trouver la conception optimale.

1 INTRODUCTION 2022). A hypothetical DGR can be seen illustrated in Figure

1. These storage rooms are built to host steel walled, used
With an increased social desire for cleaner energy fuel containers (UFC’s). UFCs are a key part of the DGR
production, nuclear power acts as a viable alternative to design and play an integral role in containing and isolating
coal and petroleum-based energy sources. While Nuclear the spent fuel, designed such that they can withstand
energy does not produce CO2, it does produce another massive amounts of mechanical force in the case of rock
potentially harmful bi-product; spent fuel bundles. These failure, to ensure the waste is not released into the
fuel bundles, while no longer viable for energy production, surrounding environment (NWMO, 2016). The used fuel
will remain radioactive, as well as produce heat for containers are encased in bentonite clay in order to further
hundreds of thousands of years. isolate the used fuel from the environment (see Figure 1).
In Canada, the agency responsible for the
management of spent nuclear fuel is the National Waste
Management organization (NWMO). Currently the NWMO
utilizes temporary storage techniques, Dry Cask Storage,
to host Canada’s spent fuel reserve. This is a safe
technology, however this technique requires repackaging
every 40 to 50 years, and with a total current reserve of
3,200,000 used fuel bundles (approximately 11,800m 3, or
72,000,000kg) of nuclear waste (Reilly, 2022) (NWMO,
2016), the need for a long term, permanent storage regime
is prompted.
The deep geological repository is proposed as the
best long-term solution to the storage of spent nuclear fuel
by the NWMO and many equivalent national organizations
(NWMO, 2016). A DGR is a deep underground facility, in
the magnitude of 500-700 meters below ground surface, Figure 1: Deep Geological Repository concept (NWMO,
comprised of a network of underground tunnels and 2016)
placement rooms in an acceptable host rock (NWMO,
 To ensure the used fuel is properly isolated, the
temperature in the host rock must be managed such that
the temperatures do not become excessive. Heat output
from the spent fuel will propagate into the host rock via
conduction, this increase in temperature will lead to thermal
expansion, and due to the confined nature of the rock, this
will lead to stress field changes which can cause thermal
crack generation and/or propagation. As a general rule, the
maximum temperature should remain below 100 °C to
ensure vapor pressures don’t form that can propagate and
worsen cracks within the host rock, as well as increase
hydraulic gradients (Guo R. , 2018).
Temperature development around the storage
rooms of a DGR has been numerically investigated by
previous researchers. Guo in 2016, analyzed the thermal
performance of a hypothetical DGR in crystalline rock
utilizing COSMOL, a finite element analysis software to
model and investigate this phenomenon by use of a near
and far field model (Guo R. , 2016). Nasir et al. conducted
similar works in 2019 utilizing FLAC3D in which a near field
model with very similar geometry to Guo’s was used to
investigate the generation of thermally induced stress in
the rock around the storage room (Nasir, Brennan, & Taek
Oh, 2019). Temperature development within the DGR were
found to reach a max of approximately 84 °C by Both Guo
and Nasir at a time of 45 years after placement (Nasir,
Brennan, & Taek Oh, 2019).
Previous investigations have investigated the
temperature development around the storage rooms of a
DGR with fixed geometries with 20 metre centre to centre
room spacings. Currently the NWMO recommends a 25-
metre center to centre spacing for the storage rooms, but
states this is not an absolute set value. To investigate the
sensitivity of this parameter, a sensitivity analysis was
undergone in which the centre-to-centre room spacing was
varied from 10 to 25 metres to find the DGR design that
optimized facility size and temperature development

Figure 2: Numerical modelling methodology

2 METHODOLOGY

The general methodology for this investigation consisted of 2.1 Model Geometry
generating a near field DGR model geometry, applying
material properties, a heat source, as well as boundary and To Begin investigating the temperature development
initial conditions, and simulating temperature development around the storage rooms of a DGR, a near field model was
for 150 years as shown in Figure 2. The initial model was generated. The near field model was modeled after
created with 20 metre centre to centre spacing to match the hypothetical DGR layout in crystalline rock, shown in
conditions used by Guo (2016) and Nasir et al. (2019) for Figure . The near field model is representative of a thin slice
validation purposes. With the temperature development in through a storage room of a DGR, to demonstrate how the
the model validated the geometry was altered to have near field model relates to the entire facility, a simple
centre to centre room spacings of 10, 15, and 25 metre schematic has been produced, shown in Figure . Figure a
room spacings. shows a simplified rendering of the DGR in Figure .
Following the flow through Figure b, c and d, it can be seen
that the near field model is representative of a thin slice
through a DGR storge room. By utilizing the natural
symmetry of the facility, the near field model allows for
precises storage room contents to be modeled, while also
keeping the relative complexity and computational power
necessary for long term analysis relatively low.
 spacers are 0.5m in width and are used to space the
containers and their buffer box. The remainder of the
rooms are back filled with bentonite back fill pellets.

Figure 5: Near field model geometry

Figure 3: Hypothetical DGR Layout in Crystalline Rock
(Guo R. , 2016) 2.2 Material Properties

To investigate the temperature development within the

model, thermal properties were required for each material
component. The thermal conductivity, specific heat
capacity, and density for each material can be seen in
Table 1.

Table 1: Material properties

Th. Spec.
Density
Group Conductivity Heat
(kg/m3)
(W/m∙K) (kJ/m3K)
Granite Rock 3.0 845 1995
Buffer Boxes 2.0 1060 2276
Spacer Blocks 1.0 1280 1960
Gap Fill 0.4 870 1439
Container 60.5 434 7800

2.3 Initial and Boundary Conditions

For initial conditions, a natural geothermal gradient was

Figure 4: Relation of near field model to full DGR facility applied to the model, with the temperature being set to 5
°C at the ground surface with a 0.012 °C/m gradient
The full geometry of the near field model is 0.75m downwards through the model, resulting in a temperature
in thickness, 40m in width, and 10,000m in height with the of 11 °C at the elevation of the room bottom, and 125 °C at
storage room bottom set 500m below ground surface. the bottom of the model at a depth of 10,000m BGS. In
There are 5 materials in the model, granite rock (light blue), using the near field model, some assumptions have been
dense back fill spacer blocks (green), bentonite buffer made to utilize symmetry in the model. For example, the
boxes (magenta), containers (blue), and bentonite back fill four outside vertical boundaries have an adiabatic thermal
pellets (red) as shown in Figure . The rooms of the deep condition. Meaning it is assumed that the storage rooms
geological repository are 2.2m in height, and 3.2m in width. extend infinitely in the horizontal directions, and therefore,
For validation purposes, the center-to-center room spacing no thermal gradient exists at these surfaces.
was set to 20m. The modeled mark II UFCs are bisected
longitudinally, to a width of 0.25m, representing half the 2.4 Heat Source
width of a full, ~0.5m diameter container. The containers
have also been bisected transversely for both of the two The heat source in a DGR is the used fuel. This spent fuel
half rooms on either side of the model. The dense back fill will produce heat for hundreds of thousands of years. The
volumetric heat output of the spent fuel containers can be
seen in Table 2.

Table 2: UFC volumetric heat output

Volumetric Volumetric
Time after Time after
heat heat
placement placement
output output
(years) (years)
(Q/m3) (Q/m3)
0 290.98 120 79.34
5 267.13 130 75.85
10 244.87 170 66.62
15 225.79 270 56.45
20 209.89 470 46.27
25 193.99 970 32.11
30 181.27 1970 21.94
40 157.57 4970 15.9
45 147.88 9970 11.43
50 139.13 19970 6.61
60 124.34 34970 3.61
70 112.42 49970 2.27
80 102.88 99970 0.65
105 86.02 999970 0.24
Figure 6: Points used for temperature development
validation:
2.5 Temperature development validation and sensitivity
analysis The temperature development at each point of
analysis from this research can be seen compared to the
The temperature development in the near field model was temperature developments found by Guo (2016) and Nasir
simulated for 150 years post placement of the UFCs using et al. (2019) in Figure . In the near field model utilized for
20 metre centre to centre spacing. The results will be this research, point A, the location at the center of the
compared to previous findings by Guo (2016) and Nasir et placement room reached a peak temperature of 81.8 °C at
al. (2019) to validate the temperature development in this a time of 48.2 years after placement, and point B, the
research. Once validated, the centre to centre spacing will location just outside of the room reached a max temp of
be varied to values of 10, 15, and 25 metres, to investigate 76.2 °C at a time of 66.4 years post placement. From
how this affects the temperature development around the Figure it can be seen that at point A, Nasir et al. found a
storage rooms. maximum temperature of 84.7 °C at a time of 45.8 years
after placement, and Guo found a max temperature of
83.0°C.
3 RESULTS AND DISCUSSIONS The small differences in the temperature
development between the model in this research, Guo’s,
3.1 Temperature development validation and Nasir et al.’s could be due to the slight differences in
model geometry. The model used in this research also
the results of the temperature development in and around included more storage rooms than previous works by Nasir
the storage rooms was compared to previous works et al. and Guo. In addition to this, differences in grid
performed by Guo (2016) and Nasir et al. (2019), who discretization between the models could be another
conducted similar research on near field models of explanation for the small differences in temperature
hypothetical DGRs. The model geometries used by Guo developments. However, the peak temperature found in
and Nasir are very similar to the geometry analyzed in this this research, is within a relatively small margin of less than
research. The temperature development was plotted for 5 3°C of the temperatures found previously, and the
different locations shown in Figure . Location A is at the temperature development at each of the 5 points monitored
middle of a storage room, location B is at the top of the were very similar to the temperature development found by
storage room, location C, D and E are 1, 5, and 10 metres previous researches. Therefore, the results can be
above the top of the room respectively. These are the considered accurate for use in further analysis
same locations monitored by Guo (2016) and Nasir et al.
(2019), and were selected because the temperature
development within the model in this research could be
directly compared to the temperature developments of Guo
and Nasir et al. for validation.
 Figure 7: Temperature development comparison for this research and Guo’s (2016) and Nasir et al.’s (2019) previous
findings

3.2 Temperature Development room spacing

sensitivity analysis

The near field model geometry was adjusted to have center

to center spacings of 25, 15, and 10 metres to investigate
the sensitivity of the temperature development in and
around the DGR storage rooms on the geometry of the
facility. To monitor the temperature development in the
model, additional observation points were added between
the rooms. These can be seen in Figure ; Points A through
E remained the same as the temperature development
validation however, four additional points were added,
point F and H are positioned directly next to the storage
rooms at 1.1m above room bottom, point G is positioned at
the midpoint between rooms at 1.1m above room bottom,
and point I is positioned at the center of the middle storage
room. The model was again run for 150 years with the
same material properties, boundary and initial conditions,
for room spacing of 25, 20, 15, and 10 metres
 However, the maximum temperature within the rooms
increased by ~33 ºC relative to the 20-metre spacing when
room spacing was lowered to 15 metres, and increased
78°C relative to the 20-metre spacing when room spacing
was lowered to 10 metres.

Table 3: Maximum temperature at each point for various

room spacings

Room Spacing (m) 25 20 15 10

Max Temp at Point A (°C) 81.2 81.8 115 159
Max Temp at Point B (°C) 74.4 76.2 109 155
Max Temp at Point C (°C) 70.9 73.5 107 152
Max Temp at Point D (°C) 67.6 70.7 103 148
Max Temp at Point E (°C) 64 67.3 98.6 141
Max Temp at Point F (°C) 72.7 74.9 108 154
Max Temp at Point G (°C) 66.6 70.5 104 151
Max Temp at Point H (°C) 72.7 74.9 108 154
Max Temp at Point I (°C) 81.2 81.8 115 159

While no substantial difference is noted between the max

temperature for 20 and 25m spacings, a considerable rise
in temperature is observed when lowering the room
spacing to 15 metres and 10 metres at all points. This leads
to the conclusion that a 20-metre center to center spacing
of the storage rooms may be the optimized design for
management of temperature development. Increasing the
room spacing to 25 metres does not appreciably lower the
temperature development around the storage rooms, and
leads to a 25 percent increase in area required to store the
same amount of fuel.
Figure 8: Points used in temperature development It also leads to the conclusion that decreasing the
sensitivity analysis room size to 15 or 10 meters is likely not appropriate, as it
increases the temperature in the host rock above the
allowable threshold of 100 °C. Therefore, the NWMO’s
A summary of the temperature development for recommended storage room spacing of 25 metres appears
each room spacing at the points extending vertically away to not be the optimized design to maximize storage while
from the rooms can be seen in Figure . limiting temperature rise in the host rock. However, this
The results of the temperature development at the points may not be the only rational for selecting a 25-metre
added for this sensitivity analysis, that extend horizontally spacing, as temperature development is highly site specific
between the rooms (Points A, F, G, H and I in Figure ) can and modeling would have to be undergone with properties
be seen in Figure . directly taken from the selected host rock to confirm this
Table 3 summarizes the maximum temperature at conclusion.
each point of analysis for the 4 various room spacings.
From Table 3, it can be seen that the difference in
maximum temperature for room spacings of 25 metres and
20 metres is nearly negligible. Within the rooms (points A
and I) the difference in max temperature is less than 1 °C.
 Figure 9: Temperature development at points (A, B, C, D and E) For 10, 15, 20, and 25 storage metre room spacing

Figure 10: Temperature development at points between rooms (A, F, G, H, I) for 10, 15, 20 and 25 metre room spacing

4 CONCLUSIONS A near field model was created to investigate the

temperature development in and around the storage rooms
of a DGR due to the thermal output of the spent fuel. The
temperature development in the near field model was
simulated with 20 metre centre to centre room spacings for
150 years post placement of the spent fuel. The
temperature development within the model was then
compared to previous research conducted by Guo and
Nasir et al. It was found that the temperature development
within the model used in this research matched very closely
with previous temperature developments found in similar
models (Guo R. , 2016), (Nasir, Brennan, & Taek Oh,
2019), in which maximum temperatures observed were
within 3 °C of the temperatures found in the model used in
this research.
Using the validated model, A sensitivity analysis was
conducted to investigate the effects of storage room
spacing on the temperature’s development in and around
the storage rooms of a DGR. The NWMO currently
recommends a 25-metre center to center room spacing. It
was found that no substantial difference is noted between
the maximum temperature in the host rock for 20 metre and
25 metre spacings, but a considerable rise in temperature
is observed when lowering the room spacing to 15 metres
and 10 metres at all points, generating temperatures above
the maximum allowable temperature of 100 °C. This leads
to the conclusion that the 20-metre center to center spacing
of the storage rooms may be the optimized design for
management of temperature development, while
minimizing the footprint of the facility. It also leads to the
conclusion that decreasing the room size further to 15 or
10 meters is likely not appropriate, as it considerably
increases the temperature in the host rock.

5 REFRENCES

Guo, R. (2016). Thermal Response of a Mark II Conceptual

Deep Geological Repository in Crystalline Rock.
Toronto: NWMO.
Guo, R. (2018). Thermal Response of a Conceptual Deep
Gological Repository in Sedimentary Rock.
Toronto: NWMO.
Nasir, O., Brennan, G., & Taek Oh, W. (2019). Near-Field
Thermo-Mechanical Coupled Processes in Host
Rocks of High-Level Waste Deep Geological
Repistories. St. Johns: Canadian Geotechnical
Society.
NWMO. (2016). Deep Geological Repository. Retrieved
from nwmo.ca: https://www.nwmo.ca/en/A-Safe-
Approach/Facilities/Deep-Geological-Repository
NWMO. (2016). What is Used Nuclear Fuel? Toronto:
NWMO.
NWMO. (2022). Deep Geological Repository. Retrieved
from NWMO: https://www.nwmo.ca/en/A-Safe-
Approach/Facilities/Deep-Geological-Repository
NWMO. (2022). What other countries are doing. Retrieved
from NWMO: https://www.nwmo.ca/en/Canadas-
Plan/What-Other-Countries-Are-Doing
Reilly, T. (2022). Nuclear Fuel Waste Projections in
Canada - 2022 Update. Ottawa: NWMO.
Tuesday, October 3, 2023

ROCK MECHANICS I
EFFECT OF THE DAMAGE ZONE AROUND
MINING EXCAVATIONS IN ROCK SALT
BASED ON INTERNAL STATE VARIABLE
TIME-DEPENDENT MODELLING

Jonathan D. Aubertin1, Michel Aubertin2, Abtin Jahanbakhshzadeh2

1École de technologie supérieure (ETS),

Université du Québec, Montréal, QC, Canada

2Polytechnique Montréal,

Montréal, QC, Canada

ABSTRACT
Evaluation of the stress and strain distributions around underground openings in mines of evaporites (rock salt, potash) is
essential to determine optimal room and pillar dimensions, minimal distance between levels, and crown pillar thickness.
The behavior of such low porosity soft rocks exhibits a strong time- and history-dependency when subjected to deviatoric
loading. Observations indicate that the time-dependent inelastic (creep) deformations away from the walls (i.e. inside large
pillars) can be ductile, while the rock near the openings often show a semi-brittle behavior with the emergence of a damage
zone. This article focuses on the response of the damage zone around mining excavations in rock salt. The analysis is
based on the effect of damage, which induces a reduction in stiffness and acceleration of the inelastic strain rate. The
numerical investigation evaluates the geometry of the damage zone around underground excavations based on
experimental data collected on rock salt. A time-dependent model with a strain-hardening (SH) internal state variable (ISV)
and a damage component is used to simulate the response of mining excavations induced by the evolving stress state
and related displacements.

RÉSUMÉ
L’évaluation de la distribution des contraintes et des déformations autour des ouvertures souterraines dans les mines
d'évaporites (sel gemme, potasse) est essentielle pour déterminer la dimension optimale des ouvertures et leur
espacement. Le comportement de ces roches tendres montre une forte dépendance au temps et à l'historique de
chargement. Les observations indiquent que les déformations inélastiques différées (fluage) loin des parois sont de nature
ductile, alors que la roche près des parois peut présenter un comportement semi-fragile avec l'apparition d’une zone
d’endommagement. Le présent article analyse la réponse de cette zone autour des excavations minières dans le sel
gemme. L'analyse tient compte de l'effet de l’endommagement qui induit une réduction de la rigidité et une accélération
du taux de déformation inélastique. Des simulations numériques évaluent la géométrie de la zone d’endommagement
autour des excavations minières souterraines sur la base d’observations expérimentales sur le sel gemme. Un modèle de
comportement inélastique avec une variable d'état interne d’écrouissage et un paramètre d’endommagement est utilisé
pour simuler le comportement différé (fluage) induit par l'évolution de l'état de contrainte et des déplacements autour des
excavations.

1 INTRODUCTION intracrystalline plasticity and micro fracturing (Aubertin et

al. 1998a).
The geomechanical behavior of low porosity rocks can Studying the semi-brittle regime (SBR) is of interest for
range from a brittle response, as is the case for fine grained many applications in rock mechanics, such as the
plutonic rocks at relatively low confining pressure, to time- evaluation of frictional stresses and seismic response in
dependent ductile (fully plastic) time-dependant straining the earth crust, the creation and evolution of fault-zones,
for evaporites under commonly encountered loading and the extraction of geothermal energy at depth (e.g.
conditions in the rock mass. (Carter and Tsenn 1987; Price 2013; Jacquey and Cacace
The response of rocks depends on the properties of the 2020). The SBR is also relevant for low porosity soft rocks
constituent grains and on the characteristics of the various when assessing the effect of the damage zone around
defects that control their non-linear behavior. Brittle underground openings used as waste disposal chambers
mechanisms are essentially related to the creation and or for mining operations (Chan et al. 1996; Thorel and
propagation of microcracks that produce pressure Ghoreychi 1996; Cristescu and Hunsche 1997; Aubertin et
dependent non-isovolumetric straining. Ductile processes al. 1998a; Roberts 2015).
are mainly associated with dislocations generation and A semi-brittle behavior is commonly observed in
motion within grains (crystals), leading to fully plastic, time- evaporites, such as rock salt and potash, when the loading
dependent isovolumetric strains that are (almost) pressure conditions involve a relatively high deviatoric stress (above
insensitive. Under specific loading conditions, some 0.3 to 0.5 of the peak strength) at low confining pressure
materials, including various rock types, can also show a (Aubertin et al. 1993, 1998a; Thorel 1994). These
semi-brittle response. This transitional inelastic behavior conditions often prevail near the walls of underground mine
regime usually involves the simultaneous action of openings, where a damage zone may develop. In such
instances, the time-dependent inelastic response of the the damage initiation threshold (DIT) favors brittle
rock tends to be ductile (fully plastic) away from the walls processes associated with damage due to microcracking.
(i.e. inside large pillars), while it depends on the amount of Decreasing the loading rate (in a constant strain rate, CSR,
damage induced by the stress state near the excavation test) tends to promote ductile processes by reducing the
boundaries. Such semi-brittle response should be taken deviatoric stress (at a given strain), which can then remain
into account when analyzing the behavior of mine closer to (and even under) the DIT. Below this threshold,
openings, but it is not commonly the case in practice. The strains are ductile (fully plastic), isovolumetric and (almost)
resulting damage zone and superficial conditions are independent of the mean stress. Straining becomes
particularly important for safe and reliable operation of pressure sensitive above the DIT, but volumetric strains
large underground excavations in salt mining operations. are usually much smaller in the SBR than for a brittle
This article investigates the semi-brittle behavior of rock behavior. Semi-brittle microcracking in rock salt generally
salt (and other low porosity soft rocks) around underground remains diffuse and uniformly distributed until the peak
mining excavations. A time-dependent, strain hardening strength is reached (and even beyond). The apparent SBR
(SH) constitutive model with an internal state variable (ISV) saturation (peak) stress in a CSR test is lower than the
was implemented in FLAC (Itasca Consulting Group 2019) corresponding stress for a fully plastic behavior because
to simulate the ductile behavior of rock salt (Aubertin et al. microcracking weakens the rock (Aubertin et al. 1998a).
2023). A damage component is added here to the Damage weakening also tends to reduce Young’s
constitutive model, and the new parameters are evaluated modulus.
explicitly based on laboratory test results. The model In creep tests, the transient (hardening) phase in the
formulation, referred to as ISV-SH-D is deemed to capture SBR is usually followed by a pseudo steady-state and/or
the geomechanical response in the ductile and semi-brittle by an accelerating (tertiary) creep phase. The minimum
regimes. creep rate in the SBR for a given deviatoric stress is larger
The proposed formulation is first used to simulate than the stationary creep rate for a comparable ductile
constant stress (creep) tests at different confining condition, due to the weakening effect of microcracking
pressures. The ISV-SH-D model, also implemented in the (Sgaoula 1997; Aubertin et al. 1998b). General strain
numerical code FLAC, is then applied to simulate the localization (leading to macrofracturing) progresses slowly
damage zone and resulting convergence rate near typical near and beyond the peak strength, so the rock may be
mine openings. Results illustrate the influence of semi- considered as a continuum even in the early part of the
brittle behavior on deformation near the walls. The tertiary creep phase or along the softening portion of the
geometry of the damage zone around a high bench room stress-strain curve. Small scale strain localization may
configuration is also evaluated with the proposed model occur but microcracks coalescence is restricted by the
and compared with empirical evidence in room and pillar plastic deformation processes. Also, while microcracking is
mine workings. generally considered an irreversible process, microcracks
closure under pressure may lead to (partial) healing in rock
2 SEMI-BRITTLE RESPONSE OF SOFT ROCKS salt, which tends to reduce the effect of past loading on the
degradation of the rock strength and elastic parameters
The semi-brittle behavior has been investigated on a (Brodsky and Munson, 1994).
number of engineering materials, including various
crystalline rocks submitted to high pressure and 3 CONSTITUTIVE MODEL FORMULATION
temperature (Carter and Kirby 1978; Hirth and Tullis 1989;
Fredrich et al. 1989; Pec et al. 2016; Nicolas et al. 2016). 3.1 Time-dependent Inelastic Modelling
In practice however, only a few types of rock show a semi-
brittle response under commonly encountered conditions The following introduces an elastic-viscoplastic model with
in mines. This can be the case with rock salt (and other a damage component to represent the ductile and semi-
evaporites), for which intracrystalline plasticity due to brittle regimes of rock salt inelastic behavior under various
dislocations motion is relatively easy to initiate, but which loading conditions, up to failure.
is partially restricted by the limited number of active slip The proposed constitutive model formulation is
systems that in turn favors concomitant microfracturing. expressed in terms of partial differential equations for the
This type of transitional behavior is not easy to investigate strain rate 𝜖̇ (s −1 ). Modelled strains include elastic (𝜖 𝑒̇ ) and
and characterize, and its analysis for designing mine inelastic (𝜖 𝑖̇ ) components, with the latter arising from
openings involves many challenges related to the ductile (𝜖𝑣𝑖̇ ) and damage induced (𝜖𝑑𝑖̇ ) behavior. The total
constitutive equations formulation and their application strain rate can thus be expressed using the following
through numerical modelling. The constitutive and equation (Aubertin et al. 1998b):
numerical models must then consider that rock salt semi-
brittle behavior is intrinsically time-dependent, with
𝜖̇ = 𝜖 𝑒̇ + 𝜖𝑣𝑖̇ + 𝜖𝑑𝑖̇ [1]
inelastic straining occurring through distributed stable
microcracking (cataclastic damage and flow) combined to
mechanisms of intracrystalline (visco) plasticity.
3.2 Ductile regime
Despite the difficulties involved, important observations
have been made through specialized experimental testing
The following kinetic law is the basis for ductile response
in the semi-brittle regime (SBR) of rock salt. It has been
of the model (Aubertin et al. 1991, Yahya et al. 2000):
shown, for instance, that a deviatoric stress in excess of
𝜖̇𝑣𝑖 = 𝐴 〈
𝜎
̅ − 𝜎𝑖 𝑁
〉 [2] strain and strain rate, cumulated damage and failure
𝐾
(ultimate state). This relationship can be represented by an
evolution law of the form (Kachanov 1986; Askes et al.
where 𝜎̅ (MPa) is the equivalent (von Mises) deviatoric
2020):
stress expressed from the deviatoric stress tensor (𝜎̅ =
√3𝐽2 = √3/2 ∙ 𝑆𝑖𝑗 ∙ 𝑆𝑖𝑗 ); A (s−1 ) and 𝑁 are material 𝐷̇ = 𝐹1 (𝐷) ∙ 𝐹2 (𝜎, 𝑆𝐷 ) ∙ 𝐹3 (𝜖𝑖𝑗 , 𝜖𝑖𝑗̇ , 𝑇) [8]
parameters; 𝐾 (MPa) represents the drag stress and 𝜎𝑖
includes internal stress components (MPa). where 𝑇 is temperature (considered constant here), 𝑆𝐷 is
The internal state variable strain-hardening (ISV-SH) the damage initiation surface (or DIT), and 𝐹1 , 𝐹2 , 𝐹3 are
model was developed from Eq. 2 to represent the inelastic functions capturing the influence of cumulative damage,
behavior of rock salt under loading conditions associated stress conditions with respect to the damage surface, and
with typical room and pillar mining sequences (Aubertin strain rate and cumulative strain. Sgaoula (1997) (see also
and Aubertin 2021, 2022; Aubertin et al. 2023). The main Aubertin et al. 1998b) adapted Eq. 8, based on
kinetic law for ISV-SH simplifies Eq. 2 by neglecting the experimental evidence, to represent rock salt semi-brittle
internal stress component 𝜎𝑖 ; it can thus be written as: response; the corresponding SUVIC-D model equation
serves below to formulate the damage component included
̅ 𝑁
𝜖̇v𝑖 = 𝐴 ( )
𝜎
[3] in the proposed ISV-SH-D model.
𝐾
The non-linear yield/failure surface of rock salt is
The drag stress variable K in Eq. 3, which represents the approximated here by the Mises-Schleicher (MS) criterion
viscoplastic kinetic law without a yield surface, serves to (Aubertin 2020; Aubertin et al. 2021):
model the transient and steady state inelastic flow,
considering the time and loading-history dependent √3𝐽2 = √(𝐶0 − 𝑇0 )𝐼1 + 𝐶0 𝑇0 [9]
response. The evolution of variable 𝐾 converges toward its
saturated value 𝐾′ (MPa) at steady state (i.e. 𝐾 < 𝐾′ where 𝐶0 and 𝑇0 are the uniaxial compressive and tensile
implies transient state). The value of variable 𝐾 changes strength (MPa) of the rock; 𝐼1 (= 𝜎11 + 𝜎22 + 𝜎33 , MPa) is
according to the following evolution law: the first invariant of the stress tensor. The MS criterion can
also be applied to define the damage initiation threshold
𝐾̇ =
𝑑𝐾
= 𝐴5 (1 −
𝐾
) 𝜖̇𝑣𝑖 [4] (DIT) of rock salt (Aubertin and Simon 1997).
𝑑𝑡 𝐾′
4 NUMERICAL IMPLEMENTATION OF ISV-SH-D
where 𝐴5 is a material parameter. The saturated value 𝐾′
is obtained from Eq. 3 by replacing the actual deviatoric 4.1 Numerical tools
stress component 𝜎̅ by the value 𝜎̅′ associated with the
actual inelastic state strain rate 𝜖̇𝑖 (𝑠 −1 ). The numerical computations and simulations presented
here were performed following an explicit implementation
𝜎
̅′
𝐾′ = 1 ≥ 1 MPa [5] of the constitutive equations via Python scripting (Van
𝜖̇ 𝑖 𝑁
( 𝑣) Rossum and Drake 2009), and then adapted for the 2D
finite differences model (FDM) FLAC. Numerical
𝐴

implementation of the ISV-SH constitutive equations, used

The strain rate is then linked with the steady state
for the ductile regime, is presented in Aubertin et al. (2023).
(stationary) condition, with the corresponding strain rate
The next subsections provide additional details for the
represented by the following hyperbolic sine function,
implementation of the updated model ISV-SH-D applied to
which captures the inelastic flow at low and high deviatoric
simulate the semi-brittle regime or rock salt.
stresses (Aubertin et al. 1998a; Aubertin and Julien 2015):

𝜎
̅ 4.2 Net stress, net modulus and MS criterion
𝜖̇ 𝑠𝑠 = 𝜖0̇ ∙ sinh𝑛 ( ) [6]
𝜎0
With the ISV-SH-D model, the damage variable is
where 𝜖0̇ (𝑠 −1 ), 𝑛, and 𝜎0 (MPa) are material parameters. evaluated explicitly, based on a specific version of Eq. 8,
The saturated (stationary) value 𝜎̅′ in Eq. 6 is evaluated presented by Aubertin et al. (1998b); the proposed
by isolating 𝜎̅ in Eq. 7 with the evolving inelastic strain rate formulation can be expressed as:
𝜖̇𝑣𝑖 (for 𝜖̇ 𝑠𝑠 ):
1 √𝐽2 −𝐹𝑖 1.5
𝑖
𝜖̇ 𝑛 𝐷= 〈 〉 ∙ 𝐷𝑐 [10]
𝜎̅′ = 𝜎0 ∙ sinh−1 ( 𝑣̇ ) [7] 𝐹0 −𝐹𝑖
𝜖0

where and 𝐹0 and 𝐹𝑖 are the relative positions along the

3.3 Semi-brittle regime failure (peak) and damage initiation surfaces. The critical
damage value 𝐷𝑐 (= 0.5) and that of the exponent (= 1.5)
Following the commonly used approach in Continuum are based on experimental results on rock salt (presented
Damage Mechanics, the effect of damage is represented in Sgaoula, 1997).
by a state parameter 𝐷, which varies from 0 (no damage) More specifically, the failure (𝐹0 ) and damage initiation
to Dc at the onset of failure (or localisation). The evolution (𝐹𝑖 ) surfaces, expressed in the 𝐽2 − 𝐼1 plane, are
of 𝐷 may depend on the stress conditions (above the DIT), computed as (based on Eq. 9):
 Table 1: Failure and damage parameters for Eq. 11, 12.
𝐹0 = √1/3[(𝐶0 − 𝑇0 )(𝐼1 ) + 𝐶0 𝑇0 ] [11]
𝑪𝟎 𝑻𝟎 𝝈𝒄𝒊 𝑻𝒊
𝐹𝑖 = √1/3[(𝜎𝑐𝑖 − 𝑇𝑖 )(𝐼1 ) + 𝜎𝑐𝑖 𝑇𝑖 ] [12] 25 MPa 1 MPa 10 MPa 0.8 MPa

4.3 Semi-Brittle Experimental Response

where 𝐶𝑖 (< 𝐶0 ) and 𝑇𝑖 (< 𝑇0 ) are the uniaxial compressive
and tensile initiation strength threshold.
The magnitude of damage induced strains was evaluated
The effect of damage modifies the stress and Young’s
by Sgaoula (1997) as an extension of the SUVIC model;
modulus of the material; this effect is expressed using the
the corresponding formulation of SUVIC-D is also
net (deviatoric) stress 𝜎̃ (MPa) and net elastic modulus 𝐸̃
presented in Aubertin et al. 1998b. Figure 2 shows
(GPa):
experimental results for constant stress (creep) tests
𝜎
̅ conducted on Waste Isolation Pilot Plant (WIPP) salt
𝜎̃ = [13] specimens. Simulations with the SUVIC-D model are also
1−𝐷
included in the figures. The viscoplastic (𝜖𝑣𝑖̇ ) and damage
𝐸̃ = 𝐸(1 − 𝐷) [14] induced (𝜖𝑑𝑖̇ ) components of cumulative strains are also
plotted separately to show their respective contributions for
Material parameters for the failure and damage initiation two confining stresses ( 𝜎3 = 2 MPa in Figure 2 (a) and 1
surfaces were determined from triaxial compression tests MPa in Figure 2 (b)).
performed on specimens from the Weeks Island and Avery
Island mines (Louisiana, USA) (taken from Sgaoula, 1997 (a)
and Aubertin 2020). The two high purity salt deposits are 25
characterized by large crystals and high-grade orebodies. WIPP (experimental) σ3 = 2 MPa
Figure 1 shows experimental data for the damage SUVIC-D
20
initiation threshold (DIT) and failure (Peak) strength Visco-elastic
Axial Strain (%)
conditions from the Weeks Island (1500 and 1400 levels) Brittle
15
and Avery Island (AI) mines (Sgaoula 1997; Aubertin 2020;
Aubertin et al. 2021). The curved surfaces (shown in 2D
here) for the damage initiation and peak stress are fitted to 10
the experimental data points, as shown in the graph.The
corresponding MS criterion (Eqs. 11, 12) parameters used 5
to plot the curves are listed in Table 1.
35 0
0.0E+00 5.0E+05 1.0E+06 1.5E+06 2.0E+06
30
Time (s)
25
(b)
𝜎1 − 𝜎3

20 0.16
3

15
0.14
𝐽2 =

WIPP (Experimental)
10 SUVID-D
0.12
5 Visco-elastic
Axial Strain (%)

0.1 Brittle
0
0 20 40 60 80 100 120 0.08
𝐼1 = 𝜎1 + 𝜎2 + 𝜎3 (MPa)
0.06
1500-Peak 1400-Peak
1500-DIT 1400-DIT 0.04
AI-DIT Brittle surface
DIT surface σ3 = 1 MPa
0.02
Figure 1: Experimental results from triaxial compression
tests on Weeks Island (levels 1400 and 1500) and Avery 0
Island (AI) salt specimens. The two surfaces for damage 0.00E+00 1.00E+05 2.00E+05 3.00E+05
initiation and peak stress are plotted using Eq. 11, 12 and
Time (s)
parameters from Table 1.
Figure 2: Constant stress creep test results on rock salt
from the WIPP and SUVIC-D simulations, (a) confining
pressures 𝜎3 = 2 MPa; (b) confining pressures 𝜎3 = 1 MPa
(adapted from Sgaoula, 1997)
For the ISV-SH-D model presented here, the ratio of Figure 4 shows experimental results for three constant
damage to viscoplastic (creep) strains is evaluated stress (creep) tests performed on Avery Island rock salt
explicitly as a function of 𝜎3, , which tends to suppress (with 𝜎3 = 15 MPa), as reported by Senseny et al. (1993).
microfracturing (damage) when approaching 3 to 5 MPa This figure also shows simulations of the creep tests were
(Aubertin et al. 1998a, 1999). A step-wise linear also performed using the ISV-SH model with parameters
approximation is given by the followings: from Table 2.

𝜎3 ≥ 1, 𝐶 = −0.5𝜎3 + 1.5 ISV-SH Experimental points AI salt

𝜖𝐷̇
≈ 𝐶(𝜎3 ) { 𝜎3 < 1, 𝐶 = 1 [15] 0.08
𝜖𝑣̇
𝜎3 ≥ 3, 𝐶 = 0 𝜎3 = 15 MPa
𝜎̅ = 15 → 16 MPa
The relationship between ratio 𝐶 and 𝜎3 is plotted in Figure
0.06
3. It is assumed here (based on experimental data shown
in Sgaoula 1997 and others) that rock salt response is
ductile (i.e. without damage) for 𝜎3 ≥ 3 MPa, for the

Strain
deviatoric stress of interest. 0.04

1.5 𝜎̅ = 10 → 12.5 MPa

0.02

1 𝜎̅ = 5 MPa
0
𝐶
Ductile regime 0.0E+00 2.0E+07 4.0E+07
(𝜎3 ≥ 3 𝑀𝑃𝑎) Time (s)
0.5
Semi-Britle
(𝜎3 < 3 𝑀𝑃𝑎) Figure 4: Experimental results for three constant stress
creep tests performed on Avery Island rock salt (data taken
0 from Senseny et al. 1993), and simulated response given
by the ISV-SH model (Eq. 3-8) using parameters from
0 2 4
Table 2 (adapted from Aubertin et al. 2023)
𝜎3 (MPa)
One of the tests in Figure 4 was used to conduct new
Figure 3: Linear approximation applied to define the strain simulations with ISV-SH-D for 2 different (much lower)
𝜖̇
rate ratio 𝐶 = 𝐷̇ as a function of the minimum principal confining pressures. Figure 5 (a) shows the simulated
𝜖𝑣
stress 𝜎3 . curves for this incremental creep test (initial deviatoric
stress 𝜎̅ = 10 MPa, increased to 12.5 MPa after 260 days),
𝜖̇
for 𝜎3 = 2 MPa, which yields a ratio 𝐶 = 𝐷̇ = 0.5. The figure
𝜖𝑣
5 COMPARISON OF ISV-SH AND ISV-SH-D plots the simulated total strain given by ISV-SH-D, and the
previously simulated results given by ISV-SH; it also
The influence of the added damage parameter 𝐷 on shows the respective strain contributions from the
simulated material response is first evaluated through viscoplastic (𝜖𝑣𝑖 ) and damage induced (𝜖𝐷𝑖 ) components of
explicit comparison for conventional creep tests. The the model formulation (Eq.1). Figure 5 (b) shows the same
ductile behavior of the material is based on experimental 𝜖̇
testing (Senseny et al. 1993), carried on Avery Island rock type of results for 𝜎3 = 1 MPa, which yields ratio 𝐶 = 𝐷̇ =
𝜖𝑣
salt at high confinement stress (𝜎3 = 15 𝑀𝑃𝑎). Material 1 ; the curves in this figure for 𝜖𝑣𝑖 and 𝜖𝐷𝑖 are thus the same.
parameters for Eq. 3 – 8 are listed in Table 2. It is seen in these figures that the viscoplastic response
with some damage (SBR) leads to larger strains then the
Table 2: ISV-SH model parameters calibrated on creep ISV-SH associated with the ductile regime. This is a
tests performed on Avery Island rock salt samples consequence, in part, of the deviatoric stress (net stress)
(Adapted from Aubertin et al. 2023). amplification due to damage which increases the strains
and strain rates (Eq. 13-14).
Parameter Value The results shown above highlight the increased strain
A 1.6 E-14 expected under low confining pressure, when a semi-brittle
𝐴5 3.2E+8 behavior may occur. Under such conditions, damage
𝜎0 14 MPa increases the total strain rate by adding a damage induced
N 7 strain component and by increasing the deviatoric (net)
n 4 stress acting on the viscoplastic component. Such type of
𝐸0 9E-10 loading conditions may prevail near underground
𝐸 17.06 GPa excavations, as is investigated next.
𝜈 0.33
 Figure 6 shows a conceptual representation of the
(a) panel layout from a vertical section view. The excavation
6E-02
layout is representative of a typical panel in room and pillar
𝜎̅ = 10 → 12.5 MPa salt mining operations. In the simulations, the natural stress
𝜎3 = 2 MPa 𝜖𝑣𝑖
state is isotropic (set at 15 MPa). The excavation sequence
was carried from left to right with 30-day intervals between
4E-02 ISV-SH-D
each room. Detailed information on model construction,
mesh density and other modelling conditions are provided
Strain

in Aubertin et al. (2023).

ISV-SH
Figure 7 shows the simulated roof vertical (downward)
2E-02
deformation (at opening center) for the five rooms. The
simulations were carried with the ISV-SH-D (semi-brittle
𝜖𝐷𝑖 response) and ISV-SH (ductile) models. The comparison
illustrates the accentuated deformation (and rates) given
0E+00
by the semi-brittle response under low confining stress
0E+00 1E+07 2E+07 3E+07
near the wall excavation. Although the differences between
Time (s) the 2 models are not negligible, the results indicate that the
magnitude of the displacements induced by the damage
(b) component (developing near the walls) is relatively small
2E-01 compared with the global time-dependent viscoplastic
𝜎̅ = 10 → 12.5 MPa straining that occurs in the entire rock mass surrounding
𝜎3 = 1 MPa
the mine openings. The situation is however different when
1E-01
ISv-SH-D considering what specifically happens near the walls, as
shown below.
Strain

8E-02 𝜖𝑣𝑖 = 𝜖𝐷𝑖 1.4E+00

ISV-SH-D
1.2E+00 ISV-SH
Vertical displacement (m)

4E-02
ISv-SH 1.0E+00
R1
0E+00 8.0E-01
0E+00 1E+07 2E+07 3E+07
Time (s) 6.0E-01
Figure 5: Simulated constant stress (creep) test results
obtained with the ISV-SH-D model for (a) 𝜎̅ initially set at 4.0E-01
10 MPa, then increased to 12.5 Mpa after 2.25E+07 R3 R5
seconds; minimum principal stress 𝜎3 = 2 Mpa (which R4
yields 𝐶 = 0.5) (b) 𝜎̅ initially set at 10 MPa, then increased
2.0E-01 R2
to 12.5 MPa after 2.25E+07 seconds; minimum principal
0.0E+00
stress 𝜎3 = 1 MPa which yields 𝐶 = 1. The different
components of the ISV-SH-D model are shown, and 0.00E+00 1.00E+08 2.00E+08
compared with the ISV-SH simulated curve. Time (s)
Figure 7: Roof vertical convergence (downward) over time
6 SEMI-BRITTLE BEHAVIOR AROUND for rooms 1 to 5 (R1, R2, R3, R4, R5) simulated with ISV-
UNDERGROUND MINING EXCAVATIONS SH-D and ISV-SH (see text for details).

The influence of a semi-brittle response of rock salt around

underground mining excavation is investigated through The damage profile developing around a mining opening is
numerical simulations using the FDM software FLAC. Both further investigated by simulating the excavation of a tall 30
ISV-SH and ISV-SH-D models were implemented in the meters room (15 m in width) associated with vertical
code using the built-in scripting language fish. benching operation. Figure 6 shows the conceptual
A conventional 5-room panel with openings created in representation of the simulated model analysed with FLAC.
sequence is simulated here to investigate the stress The natural (horizontal and vertical) stresses were set at
redistribution (not shown here) and amplitude of room 20 MPa, which is large enough to induce a significant semi-
convergence. The simulations are based on a similar brittle response of the rock salt near the opening. The
investigation previously conducted with the ISV-SH model excavation model response was simulated over a period of
(Aubertin et al. 2023). 1 year.
 15 MPa

15 MPa
15 MPa

Figure 6: Conceptual representation (not to scale, vertical section) of a five-room mine panel with openings created in
sequence in the simulations with FLAC. The rooms were excavated from left to right at 30-day intervals.
(a) (b)
Symmetry 20 MPa
axis

75 m

180 days
7.5 m excavation

300 m 30 m 20 MPa

0 19 30 𝜎 (MPa) 0
𝜎̅ (MPa) 3

0.5 (c) (d)

Figure 6: Conceptual representation (not to scale, vertical

section) of a 30 meters high (half-width of 7.5 m) benched
room in conventional room and pillar mine settings. 𝐷
Dimensions taken from Hoentszch et al. (2019) and
Aubertin (2020).

Figure 7 presents the simulated contours of the deviatoric

and minor principal stresses after 180 days from the
opening excavation; the damage variable 𝐷 around the
benched excavation is also, after 1 day and after 180 days,
is also shown. Damage contour plots exhibit a small 0
concentrated damage zone near corners of the excavation 1 day 180 days
shortly after the excavation. The later stage exhibits excavation excavation
progressive damage envelopes extending deeper inside Figure 7: Contours of the (a) deviatoric (von Mises) stress
the rock mass. and (b) minor principal stress after 180 days; and contour
of the damage variable D after (c) 1 day and (d) 180 days.
These simulation results exhibit the characteristic adopted here means that some aspects of the semi-brittle
damaged zones which are commonly observed in mine response rock salt are not taken into account, including for
workings with tall rooms and near pillars (e.g. Van instance the directional nature of hardening and damage,
Sambeek et al. 1993, Van Sambeek 1996). Such damage which would require tensorial variables instead of the
tends to become negligible deeper inside the rock mass scalar (isotropic) formulation applied above. Also, the
(away from the walls), where the mean (confining) stress is additional strains due to damage could lead to macroscopic
large enough to prevent microcracking. failure (with spalling), as is sometimes observed in the field.
The resulting failure, not considered here, would lead to
7 DISCUSSION AND CONDLUDING REMARKS different material (discontinuous) properties (e.g. 𝐸̃ = 0)
that would further influence the stress redistribution and
The term cataclastic flow (or damage) used above to convergence pattern; additional work is planned to assess
characterize the semi-brittle behavior of rock salt generally the implications of such conditions. The influence of ground
refers to permanent straining of rocks due to distributed support systems is also of interest (Aubertin and Aubertin
fracturing. It is also used for relatively brittle rocks 2021).
exhibiting strains that follow the relative movement of Despite these limitations, the ISV-SH model and newly
broken fragments induced by micro and macrofractures proposed ISV-SH-D model, which incorporates the semi-
(without intracrystalline process contribution). In low brittle regime, offer an accessible and practical solution to
porosity soft rocks, observations indicate that ductile evaluate the complex time-dependent behavior of low
processes occur at the same time as distributed porosity soft rocks such as rock salt and potash in
microcraking develops, leading to a different type of semi- conventional mine settings. The ISV-SH-D model is a
brittle behavior. useful and relatively simple solution now implemented in
It is also noted here that a brittle-ductile transition is FLAC, which is deemed capable of representing key
sometimes identified from the experimental stress-strain features of the characteristic response of rock salt
curves (mostly on hard rocks), with the(so-called) ductile encountered in conventional salt mining layouts.
behavior attained at relatively high confining pressure
when deviatoric strains occur without softening, despite 8 REFERENCES
general rock fracturing (e.g. Paterson and Wong, 2005;
Davarpanah et al. 2023). This type of (apparent) ductile Askes, H., Hartikainen, J., Kolari, K., Kouhia, R., Saksala,
behavior is however different than the fully plastic and T., and Vilppo, J. 2020. On the Kachanov-Rabotnov
semi-brilltle regimes considered here to assess and model continuum damage model. Rakenteiden Mekaniikka,
the inelastic response of rock salt and other low porosity 53(2): 125–144. doi:10.23998/rm.82528.
soft rocks. Aubertin, J.D. 2020. Characterization of rock salt response
The work presented in this article extends a strain- to blasting using terrestrial laser scanning technology.
hardening constitutive formulation with an internal state Queen’s University.Aubertin, J.D., and Aubertin, M.
variable, developed for a ductile time-dependent behavior, 2021. Simulating the influence of rigid support on the
to account for material response in the semi-brittle regime. time-dependent response (creep) of soft rock around
The proposed ISV-SH-D model is applied to the semi-brittle underground openings. In GeoNiagara 2021.
behavior of rock salt by adding a damage component, with Canadian Geotechnicl Society, Niagara Falls.
related strains and damage weakening effect. The damage Aubertin, J.D. Aubertin, M. 2022. Comparative
strain rate is evaluated explicitly as a function of the assessment of transient creep models to analyze
minimum principal stress and computed as a ratio of the openings in low porosity soft rocks. Proc.
viscoplastic (ductile) strain rate. Damage weakening is GeoCalgary, 11 pages.
implemented by incorporating the net (deviatoric) stress Aubertin, J.D., Aubertin, M., and Jahanbakhshzadeh, A.
and net (Young’s) elastic modulus in the calculation of the 2023. Numerical implementation and application of
inelastic strain and strain rate. an internal state variable model to analyze the time-
Numerical simulations are conducted to illustrate the dependent behavior of mining excavations in rock
influence of damage on room convergence in conventional salt. Canadian Geotechnical Journal,.
salt mining conditions. The proposed model captures the doi:10.1139/cgj-2021-0593.
characteristic damage zone features observed in mine Aubertin, J.D., Hashemi, A.S., Diederichs, M., and
workings. It is also shown that the damage variable induces Hutchinson, D.J. 2021. Elliptical Blast Cratering in
non-negligible changes to the convergence rate and overall Low Porosity Soft Rock Due to Emitted and Reflected
pattern of the time-dependent inelastic deformation. The Pressure Waves Interaction. Rock Mechanics and
stress redistribution is also altered, with the resulting Rock Engineering : 1–16. doi:10.1007/s00603-021-
convergence rate becoming more pronounced. 02572-2.
The damage variable in the ISV-SH-D model is deemed Aubertin, M., Gill, D.E., and Ladanyi, B. 1991. A unified
irreversible over the (relatively small) time period viscoplastic model for the inelastic flow of alkali
considered here, but some healing could be taken into halides. Mechanics of Materials, 11(1): 63–82.
account for other applications. Elsevier. doi:10.1016/0167-6636(91)90039-3.
The ISV-SH-D model formulation is loosely based on Aubertin, M., Julien, M., and Li, L. 1998a. The semi-brittle
the more elaborate SUVIC-D model (Aubertin et al. 1998b), behavior of low porosity rocks. Keynote Lecture. In
which is also much more complex and challenging to apply 3rd North Americal Rock Mechanics Symposium. p.
for the analysis of mining excavations. The simplifications 6590.
Aubertin, M., Julien, M.R., Servant, S., and Gill, D.E. 1999. Jacquey, A.B., and Cacace, M. 2020. Multiphysics
A rate dependent model for the ductile behavior of salt Modeling of a Brittle‑Ductile Lithosphere: 2. Semi‑
rocks. Canadian Geotechnical Journal, 36: 660–674. brittle, Semi‑ductile Deformation and Damage
doi:10.1139/t99-033. Rheology. Journal of Geophysical Research: Solid
Aubertin, M., and Julien, M.R. 2015. Analysis of the Earth, 125(1). doi:10.1029/2019JB018475.
Viscoplastic Behaviour of Underground Openings in Kachanov, L.M. 1986. Introduction to continuum damage
Rocksalt. In International Symposium on Rock mechanics. Springer Netherlands, Dordrecht.
Mechanics. ISRM, CIM, Montreal. p. 11. Nicolas, A., Fortin, J., Regnet, J.B., Dimanov, A., and
Aubertin, M., Sgaoula, J., and Gill, D.E. 1993. A damage Guéguen, Y. 2016. Brittle and semi-brittle behaviours
model for rocksalt: Application to tertiary creep. In of a carbonate rock: influence of water and
Proc. 7th Symp. on Salt. Kyoto, Japan. pp. 117–125. temperature. Geophysical Journal International,
Aubertin, M., Sgaoula, J., Servant, S., Julien, M., and Gill, 206(1): 438–456. doi:10.1093/gji/ggw154.
D.E. 1998b. An up-to-date version of SUVIC-D for Paterson, M. S., and Wong, T.-F. (2005). Experimental
modeling the behavior of salt. In 4th Conf. Rock Deformation—The Brittle Field. Springer
Mechanical Behavior of Salt. Edited by M. Aubertin Pec, M., Stünitz, H., Heilbronner, R., and Drury, M. 2016.
and H.R. Hardy. Jr. Trans Tech. Publications. pp. Semi-brittle flow of granitoid fault rocks in
205–220. experiments. Journal of Geophysical Research: Solid
Aubertin, M., and Simon, R. 1997. A damage initiation Earth, 121(3): 1677–1705.
criterion for low porosity rocks. International journal of doi:10.1002/2015JB012513.
rock mechanics and mining sciences & Price, J.J. 2013. Fault and Joint Development in Brittle and
geomechanics abstracts, 34(3–4): 554. Pergamon. Semi-Brittle Rock. 1st Edition. Edited by F.H.T.
doi:10.1016/S1365-1609(97)00145-7. Rhodes. Elsevier.
Brodsky, N.S., Munson, D.E. 1994, Thermomechanical Roberts, L. 2015. Proceedings of Mechanical behaviour of
Damage Recovery Parameters for Rocksalt from the salt VIII. CRC Press, Rapid City.
WIPP, Proc. 1st North American Rock Mechanics Van Rossum, G., and Drake, F.L. 2009. Python 3
Symposium, Balkema, 731-740. Reference Manual. CreateSpace, Scotts Valley, CA.
Carter, N.L., and Kirby, S.H. 1978. Transient creep and Van Sambeek, L.L. 1996. Salt Pillar Design Equation. In
semibrittle behavior of crystalline rocks. Pure and Mechanical Behavior of Salt IV.
Applied Geophysics PAGEOPH, 116(4–5): 807–839. Van Sambeek, L.L., Ratigan, J.L., and Hansen, F.D. 1993.
doi:10.1007/BF00876540. Dilatancy of rock salt in laboratory tests. International
Carter, N.L., and Tsenn, M.C. 1987. Flow properties of Journal of Rock Mechanics and Mining Sciences &
continental lithosphere. Tectonophysics, 136(1–2): Geomechanics Abstracts, 30(7): 735–738.
27–63. doi:10.1016/0040-1951(87)90333-7. doi:10.1016/0148-9062(93)90015-6.
Chan, K.S., Munson, D.E., Bodner, S.R., and Fossum, A.F. Senseny, P.E., Brodsky, N.S., and DeVries, K.L. 1993.
1996. Cleavage and creep fracture of rock salt. Acta Parameter Evaluation for a Unified Constitutive
Materialia, 44(9): 3553–3565. doi:10.1016/1359- Model. Journal of Engineering Materials and
6454(96)00004-3. Technology, 115(2): 157–162. American Society of
Cristescu, N., and Hunsche, U. 1997. Time effects in rock Mechanical Engineers Digital Collection.
mechanics. New York: John Wiley & Sons, New York. doi:10.1115/1.2904201.
Davarpanah, S. M., Sharghi, M., Narimani, S., Török, A., Sgaoula, J. 1997. Extension d’un modèle viscoplastique au
Vásárhelyi, B. (2023) Brittle‑ductile transition stress of comportement semi-fragile du sel gemme. Doctoral
different rock types and its relationship with uniaxial thesis, Polytechnique Montréal, Montréal.
compressive strength and Hoek–Brown material Thorel, L. 1994. Plasticité et endommagement des roches
constant. Vol.:(0123456789) Scientific Reports, ductiles. Application au sel gemme. Doctoral Thesis,
Nature, https://doi.org/10.1038/s41598-023-28513-3 École Nationale des Ponts et Chaussées.
Fredrich, J.T., Evans, B., and Wong, T.-F. 1989. Thorel, L., and Ghoreychi, M. 1996. Plasticité et
Micromechanics of the brittle to plastic transition in endommagement du sel gemme. Revue Française
Carrara marble. Journal of Geophysical Research: de Géotechnique, (77): 3–17.
Solid Earth, 94(B4): 4129–4145. doi:10.1051/geotech/1996077003.
doi:10.1029/JB094iB04p04129. Yahya, O.M.L., Aubertin, M., and Julien, M.R. 2000. A
Hirth, G., and Tullis, J. 1989. The effects of pressure and unified representation of the plasticity, creep and
porosity on the micromechanics of the brittle-ductile relaxation behavior of rocksalt. International Journal
transition in quartzite. Journal of Geophysical of Rock Mechanics and Mining Sciences, 37(5): 787–
Research, 94(B12): 17825. 800. Elsevier Science Ltd. doi:10.1016/s1365-
doi:10.1029/JB094iB12p17825. 1609(00)00016-2.
Hoentzsch, S., Aubertin, J.D., and Voigt, J. 2019. Geology
of the Weeks Island Salt Dome with a focus on the
current 1,500 ft. level of the Morton Salt Mine. In
Proceedings of Solution Mining Research Institute.
New Orleans: SMRI, 2019, New Orleans.
Itasca Consulting Group. 2019. FLAC - Fast Lagrangian
Analysis of Continua. Itasca, Minneapolis.
Representation of progressive
shear stress of rock joints using the
updated complete stress – displacement
surface model, CSDS
Akram Deiminiat1 and Jonathan D. Aubertin1
1Department of Construction Engineering,

École de Technologie Supérieure (ÉTS), Montréal, Canada

ABSTRACT
Rock masses are composed of discontinuities and intact rocks. Previous studies have shown that rock mass deformations
include not only macroscopic behaviour, but also microcracks that initiate, propagate, connect macrocracks and finally
induce rock failure. It is thus vital to incorporate geo-mechanical characteristics of rock joints and normal loading into rock
joint deformation model. The complete stress-displacement surface model, CSDS, describes the full shear behavior of
rock joints subjected to differential loads. In this study, the model is updated by proposing a procedural stepwise method
to calibrate CSDS model and effectively represent full shear stress of loaded rock joints by considering geo-mechanical
properties. In addition, the proposed method is used with the plane failure analysis to describe progressive shear stress of
large-scale rock slides using monitoring deformation data. The updated calibration method is then validated by existing
experimental data and a real case study. The results show trustworthy of the proposed model.

RÉSUMÉ
Les masses rocheuses sont composées de discontinuités et de roches intactes. Des études antérieures ont montré que
les déformations des masses rocheuses comprennent non seulement le comportement macroscopique, mais aussi les
microfissures qui s'initient, se propagent, connectent les macrofissures et finalement induisent la rupture de la roche. Il est
donc essentiel d'intégrer les caractéristiques géomécaniques des joints rocheux et la charge normale dans le modèle de
déformation des joints rocheux. Le modèle de surface complète de contrainte-déplacement, CSDS, décrit le comportement
de cisaillement complet des joints rocheux soumis à des charges différentielles. Dans cette étude, le modèle est mis à jour
en proposant une méthode procédurale par étapes pour calibrer le modèle CSDS et représenter efficacement la contrainte
de cisaillement complète des joints rocheux chargés en tenant compte des propriétés géo-mécaniques. En outre, la
méthode proposée est utilisée avec l'analyse de défaillance plane pour décrire la contrainte de cisaillement progressive
des glissements rocheux à grande échelle en utilisant les données de surveillance de la déformation. La méthode
d'étalonnage mise à jour est validée par des données expérimentales existantes et une étude de cas réelle. Les résultats
montrent que le modèle proposé est digne de confiance.

1 INTRODUCTION Fukuzono (1985) probably was the first, who introduced

a method to monitor rock slope deformation. The method
Rock masses consist of structural fractures and joints that addressed the problem with the prediction of rock face
separate intact rock components. The mechanical failures, and it was later used by other researchers (Crosta
properties of these interfaces are important to determine and Agliardi, 2003; Sharon et al., 2005; Rose and Hungr,
overall behavior and assess stability around construction 2007). Nevertheless, periodic variation in shear stress has
and excavations. The present article investigates a never been studied due to the limitation of the available
numerical formulation to describe the behavior of loaded shear strength criteria. Numerous criteria have been
rock interfaces with practical implications for field proposed to predict peak and residual shear strength of
monitoring. rock fractures (Fortin et al., 1988; Maksimović, 1992; Saeb
The potential for failure of rock slopes is primarily due and Amadei, 1992; Homand et al., 2001; Grasselli and
to the shear strength of joint surfaces under loading Egger, 2003). It is emphasized, based on field observations
(Ladanyi and Archambault, 1969; Barton, 1982; Bandis et of the continuous and progressive movement of rock
al., 1983). Understanding the shear behavior of structural instabilities at large scale, that the pre and post peak shear
interfaces is thus crucial for safe and stable rock stress profile is necessary for studies in various rock
slopes.Slope stability evaluation methods can be engineering problems (Martin and Chandler, 1994;
categorized into four groups: kinematic analysis, limit Fairhurst and Hudson, 1999; Simon, 1999; Simon et al.,
equilibrium method, numerical modeling, and empirical 2003).
methods (Stead et al., 2006; Ataei and Bodaghabadi, 2008; A complete shear stress – shear displacement surface
Basahel and Mitri, 2017; Amagu et al., 2021). These model, named CSDS, was proposed by Simon (1999). The
techniques help to assess the stability of slopes and their model describes post-peak shear stress curve using model
potential for failures. parametrization and laboratory shear test data. The original
formulation of CSDS required extensive iterative curve

1
fitting which limited its applicability and relatability to 2.2 Updated CSDS Model
physical phenomenon and parameters. Recent work has
been undertaken by the authors of this article to update this Some modifications are suggested to make the CSDS
model and correlate it with parameters of physical model calibration easier. The update version incorporates
meanings, to offer to rock engineering community a proven formulas into CSDS model for several model
complete shear stress-displacement numerical tool. parameters that may otherwise it needs extensive curve
The present work describes an updated version of the fittings and trial and error. The formulas are shortly
CSDS model with laboratory measurement-based presented in the following section.
calibrations and parameters corresponding to physical
components. These changes improve the model 2.2.1 Normal Closure Model (Bandis et al. 1983)
parameter determination which otherwise requires
extensive curve fitting. The updated model was used to Bandis et al. (1983) proposed an exponential relationship
showcase preliminary applicability to large scale rock slope to describe maximum closure and initial normal stiffness by
instability problems though a simplified plane failure case considering the geophysical properties of joints (i.e., joint
study. aperture (aj) and JRC) as follow.

𝐽𝐶𝑆 −0.68
𝑉𝑚 = 8.57( ) ( 11)
2 COMPLETE SHEAR STRESS – DISPLACEMENT 𝑎𝑗
SURFACE MODEL (CSDS) 𝐽𝐶𝑆
𝐾𝑛𝑖 = 0.02 ( ) + 1.75𝐽𝑅𝐶𝑝 − 7.15 (12)
𝑎𝑗
2.1 The CSDS Model

The CSDS model proposes an exponential equation to where aj is initial joint aperture (mm) and it can be obtained
describe the shear stress – shear displacement surface by Eq. 13.
profile (Simon, 1999).
𝐽𝑅𝐶𝑝 𝜎𝐶
𝑎𝑗 = (0.2 − 0.1) (13)
𝐹(𝑢) = 𝜏 = 𝑎 + 𝑏𝑒𝑥𝑝(−𝑐𝑢) − 𝑑𝑒𝑥𝑝(−𝑒𝑢) (1) 5 𝐽𝐶𝑆

where a, b, c, d and e are model parameters with the where c is uniaxial compressive strength (MPa) and JCS
condition of a, b, c, d, e > 0 and c < e. is joint compressive strength (MPa).
These equations are incorporated to CSDS model to
𝑎 ≈ 𝜏𝑟 (2) facilitate the calculation of 1 and 2 in Eqs. 8 and 9, and to
𝑐 ≈ 5/𝑢𝑟 (3) ensure the influence of joint deformation is considered in
5𝑢𝑝 the analyses.
𝜏𝑝 − 𝜏𝑟 [1 − exp (− )]
𝑢𝑟
𝑑= (4) 2.2.2 Peak shear strength criterion after Barton and
5𝑢𝑝
exp (−
𝑢𝑟
) − exp (−𝑒𝑢𝑝 ) Choubey (1977)
𝑏 = 𝑑−𝑎 (5)
𝑑𝑒𝑢𝑟 5 Barton and Choubey (1977) proposed the commonly
− exp [𝑢𝑝 (𝑒 − )] = 0 (6) known peak shear strength criterion by considering the
5(𝑑 − 𝑡𝑟 ) 𝑢𝑟
joints roughness at the peak state and joint compressive
strength as follow.
where p and r are peak and residual strengths (MPa),
respectively. The up and ur are displacements at the peak 𝐽𝐶𝑆
and residual shear stress, respectively (mm). These 𝜏𝑝 = 𝜎𝑛 tan [𝐽𝑅𝐶𝑝 𝑙𝑜𝑔 ( ) + 𝜑𝑏 ] (14)
parameters are usually measured by conventional 𝜎𝑛
experiments.
The CSDS model also describes normal displacement In the updated CSDS model, the peak shear stress
– shear displacement profile based on the content behind obtained from laboratory tests and/or obtained from Eq. 1
the maximum normal closure, Eq. 7 is considered equal to the p obtained by Eq. 14. The
Barton model is used in the present study due to easy
𝑉 = 𝛽1 − 𝛽2 exp(−𝛽3 𝑢) (7) estimation of JRC and its worldwide application in the
engineering field.
where 𝛽1 , 𝛽2 , and 𝛽3 are model parameters that can be
determined by following equations. 2.2.3 Mobilized Shear Strength after Barton (1982)
and Asadollahi (2009)
𝜎𝑛 𝑘 𝜎𝑛 𝑉𝑚
𝛽1 = 𝑢𝑟 (1 − ) 2 tan𝑖0 + (8) Barton (1982) proposed mobilized shear strength model
𝜎𝑐 𝐾𝑛𝑖 𝑉𝑚 − 𝜎𝑛
𝜎𝑛 𝑉𝑚 that is originally taken from Barton peak shear strength
𝛽2 = 𝛽1 − (9) criterion with the replacement of p with m and JRCp with
𝐾𝑛𝑖 𝑉𝑚 − 𝜎𝑛
JRCm. This model considers the progressive degradation
1.5
𝛽3 ≅ (10) of joint roughness during the shear process as follow.
𝑢𝑟

2
 𝐽𝐶𝑆 - For all the data points on the post – peak curve, the u/up
𝜏 = 𝜎𝑛 tan (𝐽𝑅𝐶𝑚 𝑙𝑜𝑔 ( ) + 𝜑𝑟 ) (15)
𝜎𝑛 (or ε/εp) is calculated and added into Eq. 17 to determine
𝜏𝑚
arctan ( ) − 𝜑𝑟 JRCm.
𝜎𝑛 - To calculate the CSDS model parameters, 1, 2 and 3
𝐽𝑅𝐶𝑚 = (16)
𝑙𝑜𝑔 (
𝐽𝐶𝑆
) using Eq. 19 from the updated model and Eqs. 9 and 10
𝜎𝑛 from Section 2.1.
- The axial strain on the post peak stress – strain curve is
where  is the mobilized shear stress (MPa) and r is a function of several parameters, Eq. 20 (Simon et al.
residual friction angle (°). 2003):
Since Eq. 15 calculates the mobilized shear stress ∆𝑢𝑐𝑜𝑠𝛽 ∆𝑉𝑠𝑖𝑛𝛽 ∆𝜎1
without taking the displacement into account, Barton 𝜀 = 𝜀𝑝 + − + (20)
𝐿 𝐿 𝐸
(1982) proposed a simple relationship between JRCm/JRCp
and u/up that is defined through a data series. Using the where ε is the axial strain, εp is peak strain, Δu is difference
table and the up, JRCp and u values taken from in shear displacement (mm), ΔV is difference in normal
experimental data, JRCm can be obtained. displacement (mm), Δ1 is difference in main axial stress
Asadollahi (2009) also modified Barton’s model as (MPa), L is the initial sample length (mm), E is the elastic
follow to obtain more complete shear stress – displacement modulus of rock (MPa) and  is the shear plane angle (°).
curve. The model was later validated by Asadollahi and If one subtracts the peak and elastic strains from Eq. 20
Tonon (2010). and rewrite the equation in terms of Δu1, Eq. 21 can be
obtained for each data on the post peak stress - strain
𝐽𝑅𝐶𝑚 𝑢 curve.
= ( )−0.381 (17)
𝐽𝑅𝐶𝑝 𝑢𝑝
𝜎𝑛 0.34 𝐽𝐶𝑆 At 1st point, ΔV1 = V1 = 0. Eq. 21 becomes:
𝑢𝑝 = 0.0077𝐿0.45 ( ) cos(𝐽𝑅𝐶𝑝 𝑙𝑜𝑔( )) (18) ∆𝜎1,1
𝐽𝐶𝑆 𝜎𝑛
∆𝑢1 = (𝜀1 − 𝜀𝑝 − ) ∗ (𝐿/𝑐𝑜𝑠𝛽) (21)
𝐸
where L is the specimen length (m). Δu1 = u1

Analysis on the comparison between these models and At 2nd point, ΔV2 = V2 – V1
the shear stress – displacement curve obtained from From Eq. 20,
experimental tests showed that both models can predict ∆𝜎1,2 ∆𝑉2 𝑠𝑖𝑛𝛽
∆𝑢2 = (𝜀2 − 𝜀𝑝 − + ) ∗ (𝐿/𝑐𝑜𝑠𝛽) (22)
well. So, Eqs. 15, 16 and 18 are incorporated into the 𝐸 𝐿
updated CSDS model, and 1 can be obtained easier and u2 = Δu2 + u1
more precise through Eq. 19
This procedure is repeated for all the data points.
𝜎𝑐 - The obtained values for u1 to un (n: number of the data
𝛽1 = 𝑢𝑟 tan (𝐽𝑅𝐶𝑚 log ( )) (19) point on the post peak curve) are used with Eq. 1 to obtain
𝜎𝑛
the progressive shear stress.
- The shear stress versus u profile obtained by these data
3 UPDATED CSDS MODEL CALIBRATION is comparable to that obtained by direct shear test.
In the absence of direct shear test data, an alternative
Since CSDS model can be calibrated via conventional method given in the following steps can be used to
laboratory tests, the updated calibration method can be determine model properties and CSDS model parameters.
used to describe full shear stress-displacement profile, full
normal displacement – shear displacement curve and full - The JRCp obtained by triaxial test data, is used with the
axial stress – strain profile. Below are special cases back calculation of Eq. 14 to obtain p.
reflecting typical laboratory testing programs, and the - Since the residual friction angle is almost equal to that
corresponding calibration procedure. obtained by direct shear tests (Khosravi 2016), the r is
used in Mohr-Coulomb equation with cohesion equal to
3.1 Full Shear Stress – Displacement Curve with zero to obtain r.
Triaxial Compression Test, with/without Direct - Since JRCp is known from Step 1 and JCS is known from
Shear Test uniaxial tests, up can be calculated by Eq. 18.
- With applying curve fitting and back calculation of Eq. 1,
When direct shear test data are available, CSDS model the displacement corresponding r can be considered as
properties such as p, r, up and ur can be directly extracted residual displacement.
from the test data and used to calculate the model
parameters (i.e., a, b, c, d and e). Having the model 3.2 Full Shear Stress – Displacement Curve with
parameters, the following steps can be carried out with the Direct Shear Test
use of triaxial and axial test data.
When direct shear test data is available, model properties
- To obtain JRCp using the back calculation of Eq. 14. are directly extracted from measurements. Consequently,
- The aj, Vm and Kni values are determined from Eqs. 11, CSDS model parameters can be calculated. Adding the
12 and 13, respectively.

3
displacement values into Eq. 1, the progressive shear The application of new CSDS model calibration method is
stress can be determined. exemplified in this section for the full shear stress-
For the cases when some of the model properties such displacement profile. Direct shear and triaxial compression
as r, ur and etc. are not available from experimental data, test data are taken from Khosravi. (2016) and Wang et al.
the proposed equations that are incorporated to the CSDS (2016).
model, mechanical properties of rock sample such as b,
JCS (or c) and E obtained by uniaxial compressive test, 4.1 Full Shear Stress – Displacement Curve with
and curve fitting technique can be used to obtain model Triaxial Compression Test, with/without Direct
parameters. Shear Test

3.3 Full Axial Stress – Strain Curve with Triaxial Triaxial and uniaxial compression test data of a type of rock
Compression Test obtained by Khosravi (2016) and direct shear test results
of the same rock obtained by Khosravi and Simon (2018)
The method for estimation of post peak stress – strain at normal loads of 5 and 8 MPa are used in this section.
curve proposed by Simon (1999) is used in this section with Tables 1 and 2 show the rock characteristics and direct
some modifications. shear test data used for the analysis, whereas Table 3
In this method, once CSDS model parameters are presents the model properties obtained in this study using
obtained by triaxial and uniaxial compressive test data, the proposed model.
following steps can be taken:
Table 1 Rock properties taken from khosravi (2016)
- For the data points on the pre and post peak stress –
strain curves, shear stress and normal stress are first 3 B E T So b i
determined by using Eqs. 23 and 24. (MPa) (°) (GPa) (MPa) (MPa) (°) (°)
24 17 45.4 280 33 53 33
1 1
𝜎𝑛 = (𝜎1 + 𝜎3 ) − (𝜎1 − 𝜎3 )𝑐𝑜𝑠2𝛽 (23)
2 2
1
𝜏 = (𝜎1 − 𝜎3 )𝑠𝑖𝑛2𝛽 (24) Table 2 Direct shear test results taken from Khosravi and
2 Simon (2018) for two normal stress values
where 1 is the principal stress (MPa); 3 is the minor stress
n up ur p r r
(MPa). 0 (°)
(MPa) (mm) (mm) (MPa) (MPa) (°)
- With the application of linear solver in Excel and Eq. 1,
shear displacement corresponding to the obtained shear 5 0.39 6.0 6.3 3.5
48 46
stress can be calculated. For each shear stress, there are 8 0.44 6.0 9.9 6.9
two values for displacement. Depending on the pre or post
peak curve, the displacement corresponding to that part of Table 3 Model properties obtained in this study by CSDS
the curve must be used. model
- For post peak curve, the displacement larger than up must
be used. n up ur p r 0 r
- Model parameters 1, 2 and 3 are calculated by Eq. 19 (MPa) (mm) (mm) (MPa) (MPa) (°) (°)
of the updated model and Eqs. 9 and 10 from Section 2.1.
5 0.51 6.7 5.8 3.54
- Using the predicted u values, 1, 2 and 3 in Eq. 7, the 46.6 43.5
normal displacement (V) is calculated. 8 0.5 6.0 8.9 6.8
- The axial strain for post peak is then determined by Eq.
20. Figure 1 shows a compression between the full shear
- For pre peak curve, the predicted displacement smaller stress – displacement curve obtained by direct shear test
than up must be used. Step 4 is then repeated for the data and those obtained by the modified CSDS model for
points on the pre peak curve. Since the volume change in normal loads of 5 and 8 MPa.
pre peak zone is positive, Eq. 25 must be used to obtain
normal displacement.
8 n = 5 MPa
Shear stress (MPa)

𝑉 = −( 𝛽1 − 𝛽2 exp(−𝛽3 𝑢)) (25)

6
- The Axial strain is then calculated using Eq. 26.
∆𝑢𝑐𝑜𝑠𝛽 ∆𝑉𝑠𝑖𝑛𝛽 ∆𝜎1 4
𝜀𝑝𝑟𝑒−𝑝𝑒𝑎𝑘 = 𝜀 + − + (26)
𝐿 𝐿 𝐸
2 Direct shear test
where εpre-peak is the axial strain before peak.
Modified CSDS
0
0 2 4 6 8 10 12
4 VERIFICATION OF THE UPDATED CSDS MODEL Shear displacement (mm)

4
 12 Figure 2 A comparison on the full shear stress –
n = 8 MPa displacement curves obtained by direct shear test (DST)
10 and modified CSDS model without DST; experimental data
are taken from Khosravi and Simon (2018) for normal loads
Shear stress (MPa)

8
of 5 MPa and 8 MPa
6
As shown in Figure 2, the full shear stress profiles
4 obtained for two normal loads are precisely predicted by
2 Direct shear test the proposed method using the alternative. Nevertheless,
Modified CSDS the predicted curve for larger normal load is better than that
0 obtained for normal load of 5 MPa. The results tend to
5 0 10 indicate that when direct shear test data is not available,
Shear displacement (mm) the alternative method can be used to describe the full
Figure 1 Comparisons between full shear stress – shear profile. However, further verification with using the
displacement curves obtained by direct shear test (DST) experimental data of various materials under different
and those obtained by modified CSDS model; experimental normal loads is required to check if the conclusion given
data are taken from Khosravi and Simon (2018) for normal here remains consistent.
loads of 5 MPa and 8 MPa
4.2 Full Shear Stress – Displacement Curve with
Figure 1 reveals that the model can predict the original Direct Shear Test
curve precisely for the two normal loads. A comparison
between Tables 2 and 3 lead to the same conclusion. The application of CSDS model is exemplified in this
The data given in Tables 1 and 2 are used here again section with the use of direct shear test results obtained by
to further validate the model without using direct shear test Wang et al. (2016) for a rock joint with JRC of 19.05. Table
data. Table 4 shows the model properties obtained by the 5 shows mechanical properties of the rock sample obtained
updated CSDS model and Figure 2 indicates a comparison under the normal load of 2 MPa and deformation rate of 6
on the shear stress curve obtained by direct shear test and mm/s. Based on the available data, appropriate method
SCSD model with using the alternative method. among those described in Section 3 was used here to
obtain the full shear stress curve. Table 6 shows the data
Table 4 The model parameters obtained in this study by obtained by the new calibration method of CSDS model
using the modified CSDS method without direct shear test and Figure 3 exhibits comparison between the laboratory-
data based shear profile and that obtained by the proposed
calibration method.
n up ur p r r
0 (°) Table 5 Mechanical characteristics of the rock sample,
(MPa) (mm) (mm) (MPa) (MPa) (°)
5 0.51 5.4 5.8 4.5 taken from Wang et al. (2016)
42
53.0
8 0.5 5.5 10.7 7.2 .0
Triaxial compression
Young's Poisson's UCS stress b
modulus ratio (MPa) (ᵒ)
8 n = 5 MPa C (MPa) Øpeak (ᵒ)
Shear stress (MPa)

6 10.35 0.166 41.8 8.78 44.01 46

4 Table 6 The CSDS model properties obtained in this study

2 Direct shear test
n up r ur JRCc-
Modified CSDS without DST… p (MPa)
(MPa) (mm) (MPa) (mm) mob
0
0 2 4 6 8 10 12 2 5.0 4.1 4.04 15.04 6.7
Shear displacement (mm)
6
15
5
n = 8 MPa
Shear stress (MPa)
Shear stress (MPa)

4
10
3

5 2
Direct shear test Experimental shear test
1
Modified CSDS without DST data CSDS model
0 0
0 2 4 6 8 10 12 0 5 10 15 20
Shear displacement (mm) Shear displacement (mm)

5
Figure 3 Full shear stress – displacement profiles obtained To obtain the shear behavior of rock slide at or close to
by the experimental shear tests taken from Wang et al. the point 4 by the CSDS model, mechanical properties of
(2016) and the new calibration method proposed in this the rock and normal load acting on the rock slide should be
study. obtained if they are not available. In this case study, the
mechanical characteristics of the rock were obtained by the
The accuracy of the proposed method for the full shear triaxial and uniaxial shear tests that are used in this study
stress – displacement curve can be seen again in Figure 3. and presented in Table 7.
The full shear behavior of rock joints obtained in this study
exhibit the same trend as that of the original curve. Table 7 Mechanical properties of rock (Kundu et al., 2017)

Triaxial Direct shear

5 APPLICATION OF UPDATED CSDS MODEL FOR Uniaxial compressive test
Unit
LARGE SCALE ROCK SLIDES compressive strength (rock (Rock joint)
weight face)
strength
(kg/m3) (MPa)
The verified updated CSDS calibration method is used in C C Ø
Ø (ᵒ)
this section along with the plane failure analysis to describe (MPa) (MPa) (ᵒ)
progressive shear behavior of large-scale rock slides. The 2700 90 0.4 42 0.05 34
application of suggested method is exemplified here using
deformation data that were monitored and reported by To determine normal loading, the plane failure analysis
Kundu et al. (2017) for one of the joints occurred along the proposed by Hoek and Bray (2005) is used in this section.
rock slope face. The slope has 70 m height with a slope Table 8 shows geometry of the sliding plane that is
angle of 60ᵒ, while the joint has a dip angle of 55ᵒ. Figure extracted from Figure 4, and the plane failure analysis for
4 shows geometry of the rock slide. As seen, there are 4 monitoring point 2.
monitoring points along the rock face. The deformation rate
obtained by monitoring point 2 is used for the analysis, Table 8 Geometry of the rock slide at monitoring point 2
which is shown in Figure 5. and the plane failure analysis
Upper slope
n
H (m) Z (m) A (m2) p (ᵒ) f (ᵒ) W (N)
(MPa)
23 22.14 1.05 55 60 84.7 0.0008

Using the mechanical properties of rock and normal

loading with the updated CSDS model presented in Section
2.2, p, JRCp, up, r and ur are estimated and presented in
Table 9.

Table 9 CSDS model properties obtained by CSDs model

Monitoring p up r ur
JRCp
point (MPa) (cm) (MPa) (cm)
Figure 4 Geometry of the studied rock face (Kundu et al. 2 0.052 0.015 0.032 0.0044 9.3
2017), the monitoring point 2 located almost in the upper
slope Since the model properties are known from Table 9, the
model parameters (i.e., a, b, c, d, e and d) are calculated
0.2 and used with the monitoring deformation to determine
progressive shear stress for the monitoring period.
Monitoring point 2 In this case study, the shear stress curve at points 5 to
Displacement (cm)

0.15 8 (see Figure 4) were determined using DEM simulation.

The curve obtained at Point 6, which is located close to the
0.1 monitoring point 2, is used here for the verification of shear
stress curve. Figure 6 shows comparisons on the shear
stress curve taken from Kundu et al., (2017) and that
0.05 obtained by the proposed method.
As seen, the methodology used in this study results in
0 a shear stress profile that is fit to that taken from Kundu et
al. (2017).
0 0.5 Time (s)1 1.5 2
Figure 5 Displacement versus time monitored at point 2,
data taken from Kundu et al., (2017)

6
 0.06
0.05 7 ACKNOWLEDGEMENT
Shear stress (MPa)

0.04
The authors acknowledge the financial support from
0.03 Natural Sciences and Engineering Research Council of
Canada through its Discovery Grant program (RGPIN -
0.02
Kundua et al. (2017)
2022-03893), and École de Technologie Supérieure (ÉTS)
0.01 CSDS model
construction engineering research funding.
0
0 0.5 1 1.5 2
Time (s) 8 REFERENCES

Amagu, C.A., Zhang, C., Kodama, J.I., Shioya, K.,

Figure 6 A comparison between shear stress profile taken Yamaguchi, T., Sainoki, A., Sharifzadeh, M. 2021.
from Kundu et al. (2017) and that obtained by CSDS model
Displacement measurements and numerical analysis of
versus time for monitoring point 2. Long-Term rock slope deformation at Higashi-Shikagoe
limestone quarry, Japan. Advances in Civil
Engineering, 2021: 1-15.
6 DISCUSSION AND CONCLUSIONS
Asadollahi, P. 2009. Stability analysis of a single three-
dimensional rock block: effect of dilatancy and high-
The present work’s aim was to contribute to rock velocity water jet impact. PhD Thesis. University of
engineering problems with rock slope stability. To this Texas at Austin, USA.
purpose, the complete shear stress – displacement surface
Asadollahi, P., Tonon, F. 2010. Constitutive model for rock
model, called CSDS, was first modified to improve the fractures: Revisiting Barton's empirical model. Eng.
calibration method. Using the proposed method, certain Geol. 113(1-4): 11-32
model parameters were obtained by the proven formulas Ataei, M., Bodaghabadi, S. 2008. Comprehensive analysis
that may otherwise require extensive curve fitting. Then,
of slope stability and determination of stable slopes in
the updated CSDS model was used along with the plane
the Chador-Malu iron ore mine using numerical and
failure analysis and monitored deformation rate to describe limit equilibrium methods. Journal of China University
progressive shear stress for large scale rock slides. of Mining and Technology, 18(4): 488-493.
Validation of the new CSDS model calibration method
Bandis, S.C., Lumsden, A.C., Barton, N.R. 1983.
showed that the proposed method could precisely describe
Fundamentals of rock joint deformation. International
the full shear stress profile of rock joints with the use of Journal of Rock Mechanics and Mine Science &
triaxial compression tests with/without direct shear test Geomechanical Abstract., 20(6): 249-268.
data. In addition, the incorporation of mobilized shear
Barton, N., Choubey, V. 1977. The shear strength of rock
strength criterion and JRC model to the updated CSDS
joints in theory and practice. Rock Mechanic. 10(1): 1-
model allowed further development of the model with plane 54.
failure analysis, and its application for large scale rock Barton, N. 1982. Modelling rock joint behavior from in situ
slides. Analyses on a real case study demonstrated that
block tests: implications for nuclear waste repository
the integrated CSDS model can predict the progressive
design (Vol. 308). Office of Nuclear Waste Isolation,
shear stress for the monitoring deformation of a certain Battelle Project Management Division, Columbus,
period. Ohio: 96.
Besides the achievements of the current study, there
Basahel, H., and Mitri, H. 2017. Application of rock mass
are some limitations. The model was verified by using few
classification systems to rock slope stability
experimental data series. Using more laboratory shear test assessment: A case study. Journal of rock mechanics
data of different materials obtained under different testing and geotechnical engineering, 9(6): 993-1009.
conditions are required to further validate the proposed Crosta, G.B., and Agliardi, F. 2003. Failure forecast for
calibration method and its application for the prediction of
large rock slides by surface displacement
shear behavior of rock slides in large scale. measurements. Canadian Geotechnical Journal, 40(1):
Furthermore, the present study only focuses on the 176-191.
plane failure with simple geometry while there are more Fairhurst, C., and Hudson J. 1999. Draft ISRM suggested
rock slope failures in the literature such as wedge failure,
method for the complete stress-strain curve for intact
toppling and son on (Hoek & Bray 2005). Additional work rock in uniaxial compression. International Journal of
is necessary to develop application of the proposed Rock Mechanics and Mine Science. 36(3): 279-289.
method for other types of rock slope failures. Fortin, M., Archambault, G., Aubertin, M., Gill, D.E. 1988.
The normal loading considered in all the analysis is the
An algorithm for predicting the effect of a variable
static load. The dynamic load has not been considered in normal stiffness on shear strength of discontinuities.
the proposed method. Since the monitoring displacement In Proceeding of the 15th Canadian Rock Mechanic
rate used in this study was for a short monitoring time, it is Symposium, Toronto, Canada, pp. 109-117.
necessary to further validate the proposed method using
Fukuzono, T. 1985. A new method for predicting the failure
the data of longer monitoring time.
time of a slope. In Proceedings of 4th International

7
 Conference and Field Workshop on Landslide., 1985:
145-150.
Grasselli, G., and Egger, P. 2003. Constitutive law for the
shear strength of rock joints based on three-
dimensional surface parameters. International Journal
of Rock Mechanics and Mine Science. 40(1): 25-40.
Hoek, E., and Bray, J.D. 2005, 4th edition. Rock slope
engineering. CRC press.
Homand, F., Belem, T., Souley, M. 2001. Friction and
degradation of rock joint surfaces under shear
loads. International Journal of Numerical Analysis
Methods Geo-mechanic. 25(10): 973-999.
Khosravi, A. 2016. Estimation and validation of post-peak
behavior of hard rocks. PhD Thesis. Ecole
Polytechnique, Montreal, Canada.
Khosravi, A., Simon, R. 2018. Verification of the CSDS
model in estimating the postpeak behavior of hard
rocks. International Journal of Geomechanic. 18(3),
04017166.
Kundu, J., Sarkar, K., Singh, T.N. 2017. Static and dynamic
analysis of rock slope–a case study. In ISRM European
Rock Mechanics Symposium-EUROCK 2017.
OnePetro.
Ladanyi, B., and Archambault, G. 1969. Simulation of
shear behavior of a jointed rock mass. In The 11th US
Symposium on Rock Mechanics (USRMS). OnePetro.
Martin, C.D., Chandler, N.A. 1994. The progressive
fracture of Lac du Bonnet granite. International Journal
of Rock Mechanics and Mine Science &
Geomechanical Abstract. 31(6): 643-659.
Rose, N. D., and Hungr, O. 2007. Forecasting potential
rock slope failure in open pit mines using the inverse-
velocity method. International Journal of Rock
Mechanics and Mining Sciences, 44(2): 308-320.
Saeb, S., and Amadei, B. 1992. Modelling rock joints under
shear and normal loading International Journal of Rock
Mechanics and Mine Science & Geomechanical
Abstract., 29: 267–78.
Sharon, R., Rose, N., Rantapaa, M. 2005. Design and
development of the Northeast layback of the Betze-post
open pit. In Proc. of the SME Annual Meeting, Salt Lake
City, US. Pre-print: 05-09.
Simon, R. 1999. Analysis of fault-slip mechanisms in hard
rock mining. PhD Thesis. McGill University, Montreal,
Canada.
Simon, R., Aubertin, M., Deng, D. 2003. Estimation of post-
peak behaviour of brittle rocks using a constitutive
model for rock joints. In 56th Canadian Geotechnic
Conference. Winnipeg, Canada.
Stead, D., Eberhardt, E., Coggan, J.S. 2006.
Developments in the characterization of complex rock
slope deformation and failure using numerical
modelling techniques. Engineering geology, 83(1-3):
217-235.
Wang, G., Zhang, X., Jiang, Y., Wu, X., Wang, S. 2016.
Rate-dependent mechanical behavior of rough rock
joints. International Journal of Rock Mechanics and
Mining Sciences, 83: 231-240.

8
Calf Robe Bridge Abutment Stabilization:
Case Study
Daniel Ferg, Ajay Sharma & Amit Garewal
GeoStabilization International Inc, Vancouver, British Columbia,
Canada

ABSTRACT
This technical paper provides a comprehensive account of the abutment
stabilization project undertaken by GeoStabilization International (GSI) at the Calf Robe Bridge in Calgary, Alberta. The
project involved installation of a discharge conduit through the toe of the bridge abutment to facilitate a downstream
discharge from the Bonnybrook Wastewater Treatment Plant (BBWWTP). Before the wastewater infrastructure could be
installed, the non-cohesive soils at the abutment needed a bespoke stabilization solution, including the excavation of a
75° slope. Core samples of the underlying bedrock suggested the presence of a possible shear band within the
mudstone that interlayered with more competent sandstone. The design-build solution chosen to stabilize the abutment
to allow for the installation of conduits is presented in this paper. The solution consisted of the installation of 280 hollow
bar soil nails with a 150 mm reinforced shotcrete facing. To mitigate against the potential shear band, the design called
for the installation of 118 drilled reinforced concrete shafts. This paper presents the solution and addresses the
challenges associated with excavation, installation of soil nails and drilled shafts, load testing methods, and QA/QC
testing. Overall, this project is an impressive feat of engineering and an excellent example of GSI's expertise in providing
innovative stabilization solutions.

RÉSUMÉ
Ce document technique fournit un compte rendu complet du projet de stabilisation des culées entrepris par
GeoStabilization International Inc. (GSI) au pont Calf Robe à Calgary, en Alberta. Le projet impliquait l'installation d'un
conduit de décharge à travers le pied de la culée du pont pour faciliter une décharge en aval de l'usine de traitement des
eaux usées de Bonnybrook (BBWWTP). L'excavation d'une pente à 75° dans des sols non cohésifs au niveau de la
culée était nécessaire avant l'installation de l'infrastructure d'assainissement et nécessitait une solution de stabilisation
sur mesure. Des échantillons de carottes du substrat rocheux sous-jacent ont suggéré une possible bande de
cisaillement dans le mudstone qui est intercalé avec du grès plus compétent. La solution de conception-construction
choisie pour stabiliser la culée afin de permettre l'installation de conduits est présentée dans cet article. La solution a
consisté en la pose de 280 clous de sol en barres creuses avec un parement en béton projeté armé de 150 mm. Pour
atténuer la bande de cisaillement potentielle, la conception a nécessité l'installation de 118 puits forés en béton armé .
Les défis associés à l'excavation, à l'installation de clous de sol et de puits forés, aux méthodes d'essai de charge, aux
tests QA/QC sont également abordés.

1 INTRODUCTION The paper also outlines the operation procedures,

including drilling and grouting, verification testing, and
Located beneath the Calf Robe Bridge in Calgary, construction planning. Throughout the project, engineers
Alberta, the project site encompasses the West abutment closely monitored the site conditions and made
and Pier 4. The bridge is situated in close proximity to the modifications to the stabilization plan to ensure a safe and
Bonnybrook Wastewater Treatment Plant (BBWWTP) and reliable solution. This paper aims to provide a
serves as a vital access point for Deerfoot Trail, the comprehensive understanding of the project's technical
freeway portion of Alberta Highway 2, and the primary aspects, challenges faced, and the solution implemented
north-south thoroughfare for Calgary. to address them.
This technical paper provides a detailed account
of the slope stabilization undertaken by GSI at the project 2 ABUTMENT STABILIZATION DESIGN
location. The client requested precise excavation at a 75°
slope to accommodate the installation of a 2,400 mm GSI received a geotechnical evaluation report from the
diameter concrete storm drain and two side-by-side box prime contractor, completed by the city’s engineering
culverts measuring 3,000 mm in height and 3,600 mm in consultant. which provided a comprehensive background
width. The conduits were to be placed in a single trench to the project, geotechnical investigation data, sensitivity
aligned between the west bridge abutment and pier #4 at analysis and geotechnical parameters to be used in
the bridge. detailed design. GSI used the parameters provided by the
The scope of work included a detailed design consultant including a Mohr-Coulomb soil model and a
package, supply and installation of soil nails, reinforced generalized Hoek-Brown bedrock model with an inclusion
shotcrete, and steel-reinforced concrete drilled shafts. of a weak layer in the bedrock, susceptible to strain
This paper discusses the geotechnical conditions and softening. Modelling the bedrock with conservative values
design criteria outlining the limitations and assumptions. and the inclusion of a slip susceptible layer was justified
given the importance of the infrastructure asset being Table 1. Geotechnical Design Material Properties
protected.
Reference Parameter Granular Native Overall Weak
2.1 Design Criteria Fill Sand and Bedrock Laminated
Gravel Formation Zone
Global stability of the site and abutment/pier soil Unit Weight (kN/m3) 22 22 23 23
displacements resulting from the proposed construction Peak Friction Angle,
32 34 N/A 25
sequencing were analyzed using the finite element φ’ (degrees)
analysis software PLAXIS 2D 2020. The PLAXIS model Residual Friction
N/A N/A N/A 12
created for this project was focused on capturing the Angle, φ’ (degrees)
behavior of the temporary shoring elements and Effective Cohesion, C
0 0 N/A 0
(kN/m2)
associated construction stages to understand how the
Reference Elastic
proposed excavation may result in settlement or lateral Modulus at 50% 26.3E3 31.1E3 - 1.0E6
displacement in the west abutment and pier #4, and soil Strain (kN/m2)
around the abutment and pier. The minimum static global Reference Elastic
slope stability factor of safety (FS) targeted in the design Modulus Constrained 26.3E3 31.1E3 - -
is 1.25 for short-term, temporary construction loading (kN/m2)
conditions. Reference Elastic
Modulus During
Pier 4 loading of 210 kPa (includes weight of Unload/Reload
79.0E3 93.4E3 - -
concrete) along with the peak bedrock condition; reflective (kN/m2)
of the characteristic loading condition, and Pier 4 loading Reference Small
of 300 kPa (includes weight of concrete) along with the Strain Shear Modulus 79.0E3 85.6E3 - -
residual bedrock condition; reflective of the foreseeable (kN/m2)
worse-case condition were considered for the design Unconfined
Compressive - - 1.0 -
analysis.
Strength (MPa)
Geological Strength
2.2 Soil and Rock Elements Index
- - 30 -

Material Constant - - 5 -
The subsurface profile used in the modeling and analysis
was based on the field and laboratory data provided in the
geotechnical report by the client’s consultant, and our Nominal (ultimate) bond stress values for subsurface
experience in similar soil units. The granular fill and native materials were estimated based on the material types and
sand and gravel unit identified at the site were modeled tables in the Federal Highway Administration (FHWA)
using the Hardening Soil small (HSsmall) constitutive GEC Circular No.4 (1999), FHWA GEC Circular No.5
model. A key benefit of the HSsmall model is that it (2002), and FHWA GEC Circular No.7 (2015), as well as
provides an adjustable shear modulus degradation curve our experience with similar material types. In PLAXIS, a
based on the Hardin-Drnevich relationship (PLAXIS linear and material dependent function was used to
2014). The HSsmall model includes a stress-dependent account for the bond reduction in the weak laminated
stiffness formulation, as well as shear hardening and zone within the bedrock formation. Granular Fill – 73 kPa,
compaction (cap) hardening in primary loading. The Native Sand and Gravel – 73 kPa and Bedrock – 200
reference stress used to initialize stress conditions was kPa.
calibrated based on available field data, typical values in
the literature, and GSI’s experience in similar materials. 2.3 Structural Elements
The overall sedimentary rock formation identified
at the site was modeled using the Hoek-Brown The slope stabilization plan consisted of soil nails, drilled
constitutive model. A benefit of the Hoek-Brown model is shafts, and reinforced shotcrete facing at the west
that it provides a better non-linear failure criterion for the abutment and drilled shafts at the pier #4. The soil nails at
strength of rocks as opposed to the linear Mohr-Coulomb the west abutment consisted of injection drilled T40N
failure criterion. In order to analyze the impact of the threaded hollow bar steel declined at 40 degrees off the
stress state within the rock mass on the effectiveness of horizontal and installed with a 100 mm diameter drill bit.
the existing west abutment and pier #4, as well as the Embedment length of soil nails was 5.5 m in the upper
proposed stabilization elements, the conventional Mohr- three rows and 11.6 m in the bottom two rows. The
Coulomb constitutive model was employed to model the spacing for soil nails is 1.5 m in the upper three rows and
weak laminated mudstone zone found at the site between 1 m in the bottom two rows. The hollow bar reinforcement
Elevation 1018.6 and 1018.1 meters (m). The shear selected for the soil nails was sized according to the
strength profiles for the site soils and bedrock formation structural demands observed in the PLAXIS model.
were defined based on the material properties identified in The drilled shafts installed at the abutment and pier #4
the consultant’s Geotechnical Report. Table 1 below had a diameter of 0.3 m and were installed vertically. The
summarizes the estimated engineering properties for shafts were reinforced with No.9 rebar and embedment
each unit identified at the site. lengths will be 8.5 m. Spacing of the shafts were set at
0.45 m and 1.5 m for the abutment and pier#4,
respectively. The solid bar reinforcement selected for the
drilled shafts was sized according to the structural existing bridge abutment and pier. The full model
demands observed in the PLAXIS model. sequential construction stages in GSI’s PLAXIS model are
The soil nail and drilled shaft elements were modelled outlined below. The model deformations were reset to
in PLAXIS as embedded beam elements, with axial and zero following application of the abutment and pier loads.
bending stiffness values based on the structural Therefore, the calculated deformation values are relative
capacities. The embedded beam elements were modeled to the point in time following application of the loads at the
using soil nail pullout resistances based on the available west abutment and pier #4.
subsurface data, and in similar materials. The stiffness of
the embedded beams was defined based on the elastic Stage 1 – Generate initial at-rest (k0) stress conditions.
properties of the steel bars and up to 50% contribution of Stage 2 – Begin building up fill and abutment/pier layers.
the grout stiffness surrounding the bars. The reinforced This sequence was divided into three stages in the
shotcrete facing was modelled in PLAXIS as a 2D plate PLAXIS model to account for stress history in the granular
element with axial and bending stiffness values based on fill and native sand and gravel layers.
the dimensions of the facing. Stage 3 – Apply existing structural loads at the west
We note that the existing west abutment and approach abutment and pier #4.
slab piles were modelled as embedded beam elements. Stage 4 – The existing rip rap along the slope at the west
Based on conversations with the project team, spacing of abutment will be removed and installation of first row of
the piles at the west abutment was selected as soil nails will be completed. A global stability FS
approximately 1.4 m and spacing of the piles at the west calculation was performed in PLAXIS following installation
abutment approach slab was selected as approximately of the first row of nails.
2.6 m. The modelled embedment depths of the piles at Stage 5 – Excavation will continue up to approximately
the west abutment and approach slab were Elevation 0.3m below the second row of soil nails and installation of
1022.2 m. The modelled embedment depths of the piles second row of soil nails at the west abutment will be
at the trailing west abutment were up to approximately completed. A global stability FS calculation was
9 m below ground surface. performed in PLAXIS following installation of the second
To model existing loading conditions at the west row of nails and first lift of shotcrete.
approach slab and abutment, surcharges of 16 and 100 Stage 6 – The sequence of excavation and soil nail
kilonewtons per square meter (kN/m2) were applied at the installation will continue down to approximately 0.3 m
approach slab and abutment, respectively. A surcharge of below the third row of soil nails at the west abutment. A
660 kN/m2 was applied to model existing loading global stability FS calculation was performed in PLAXIS
conditions at pier #4. following installation of the third row of nails and second
Table 2 below outlines the structural properties of the lift of shotcrete.
structural elements used in our PLAXIS 2D model. Stage 7 – The sequence of excavation and soil nail
Properties used for structural elements were based on installation will continue down to approximately 0.3 m
established values from the manufacturer. below the fourth row of soil nails at the west abutment.
The drilled shaft array proposed at pier #4 will also be
Table 2. Properties for Structural Elements installed in this sequence. A global stability FS calculation
was performed in PLAXIS following installation of the
Structural Area, A Unit Moment Elastic fourth row of nails, third lift of shotcrete, and drilled shafts.
Element (m2) Weight of Inertia, Modulus, Stage 8 – The sequence of excavation and soil nail
(kN/m3) Ix (m4) E (kN/m2) installation will continue down to approximately 0.3 m
T40N Soil Nail 8.1E-3 3.0 5.2E-6 29.1E6 below the fifth row of soil nails at the west abutment. The
drilled shafts will also be installed at the bottom of this
Drilled Shaft 7.3E-2 3.0 4.2E-4 47.3E6
excavation lift. A global stability FS calculation was
Shotcrete performed in PLAXIS following installation of the fifth row
4.6E-2 0.33 3.0E-4 25.0E6
Facing of nails, final lift of shotcrete, and array of drilled shafts.
Stage 9 – The sequence of excavation will continue down
2.4 Approach to Staged Construction to the proposed bottom of excavation.

Global stability was analyzed using the finite element 2.5 PLAXIS 2D Finite Element Model Results
analysis software PLAXIS 2D 2020. In PLAXIS, a factor of
safety stability analysis uses a so-called phi/c reduction Tables 3 and 4 below provide the estimated soil,
method in which the strength of materials is reduced with abutment, and pier displacements (represented as a
an MSF factor until failure is reached for a stable value of vector), and factor of safety value, for each construction
MSF. The MSF factor obtained at failure represents the phase in PLAXIS for peak and residual strength
global minimum factor of safety for the model. GSI conditions in the laminated zone, respectively. (The first
completed a short-term (temporary construction) stability proposed construction sequence represents Stage 4 in
assessment, of the temporary slope stabilization system. the PLAXIS model.)
The PLAXIS model had several sequential As indicated in Table 3, the minimum required FS
construction stages to take into consideration the effects value of 1.25 is achieved up to the bedrock elevation.
of duration, stress history, and construction sequence on When considering a residual friction angle of 12 degrees
the behavior of the temporary stabilization system and in the weak laminated bedrock formation, the FS value
calculated at the bottom of the excavation is 1.17, as abutment/pier themselves. The abutment monitoring plan
indicated in Table 4 (See Stage 9). Under the peak was part of a robust shoring instrumentation program
strength condition, the maximum calculated soil including ShapeArrarys and development of appropriate
displacement at the bottom of excavation is approximately actions in response to trigger alarms.
14.5 millimeters (mm) and occurs in the foreslope of the
west abutment. Under the same peak condition, the 3 CONSTRUCTION OPERATION
maximum calculated displacement of the west abutment
is approximately 7.5 mm when reaching the bottom of In preparation for the drilling operation, verification tests
excavation. Further, the maximum displacement were completed to ensure that the design grout-ground
calculated at pier #4 when reaching the bottom of bond assumptions were achievable. During the course of
excavation is 0.5 mm under the peak strength condition. the project, four verification tests were conducted using
When considering a residual strength condition within the sacrificial nails. These nails were loaded incrementally up
laminated zone, the maximum calculated soil to the maximum test load. The tests were performed at
displacement at the bottom of excavation is 18.0 mm, as various locations on the site to ensure that the field-tested
shown in Table 4 (See Stage 9). Additionally, the bond strength exceeded the assumed strength. This
maximum calculated abutment and pier displacement at method ensured that the strength of the installed nails
the bottom of excavation is 11.5 and 10.0 mm, was accurately determined and confirmed that the design
respectively, when considering the residual strength specifications were met.
condition. During the verification testing, the maximum test
load (200% of the design load) was incrementally applied
Table 3. PLAXIS Results – Peak Strengths to sacrificial soil nails. The bond strength was assumed to
be 11.5 kN/m for soil layers and 31.5 kN/m for bedrock
Stage Max. Soil Max. Max. Pier Global layers, and the design load was calculated by multiplying
Displacement Abutment Displacement FoS the bond strength by the bond length. At 1.5 times the
(mm) Displacement (mm) design load, the nails underwent a creep period of 10
(mm) minutes to verify that the material could sustain a constant
Stage 4 1.0 <0.5 0.0 1.87 stress over time. The load was then increased to twice the
Stage 5 3.0 1.5 <0.5 1.73 design load. All soil nail tests performed as expected
confirming the design assumptions.
Stage 6 6.0 3.0 <0.5 1.50
Upon confirming the design, a construction
Stage 7 12.5 6.0 <0.5 1.40 execution plan was prepared incorporating the designed
Stage 8 13.5 6.5 <0.5 1.31 stage approach for the excavation and installation of the
Stage 9 14.5 7.5 0.5 1.27 soil nails as per the PLAXIS results to minimize the
displacements at the abutment and pier.
During the drilling process, engineers closely
Table 4. PLAXIS Results – Residual Strengths
monitored the grout intake of each hole, and observed
significant grout loss in several of the soil nails, which
Stage Max. Soil Max. Max. Pier Global indicated the presence of loose soil material or open joints
Displacement Abutment Displacement FoS within the bedrock layer. These observations were
(mm) Displacement (mm)
carefully documented and factored into subsequent
(mm)
decisions and modifications to the stabilization plan, to
Stage 4 1.0 <0.5 0.0 1.73 ensure that the final solution was effective, reliable, and
Stage 5 3.0 1.5 <0.5 1.61 optimized for the site's unique geotechnical
characteristics.
Stage 6 6.0 3.0 <0.5 1.38
Several alternate drilling techniques were adapted
Stage 7 12.5 6.0 <0.5 1.29 to minimize the grout loss. An alternate grout mix with
Stage 8 14.5 7.5 <0.5 1.27 higher viscosity was used to drill the holes, and once the
target depth was reached, the design mix was injected
Stage 9 18.0 11.5 10.0 1.17
until fresh grout was visible at the collar. The holes were
then topped off with grout at the end of each day. By
To address the lower FS associated with a residual implementing these measures, the drill anulus was
strength condition developing, a modified stepped effectively filled with grout and grout overage contributing
approach was considered along the out-of-plane length of to the overall ground stability due to permeation, ensuring
the excavation to mitigate potential destabilizing the integrity of the slope stabilization system.
mechanisms within the bedrock formation. Additionally, a Once the installation of each soil nail was
monitoring plan was implemented for the ground surface complete, the slope was prepared for shotcrete
settlement and abutment and pier movement during installation by installing reinforcements including welded
construction to evaluate performance of the temporary wire mesh and horizontal and vertical rebar walers. Strip
stability system and the west abutment and pier #4. The drains were installed behind the welded wire mesh from
project team also developed a response and contingency the crest of the slope and extended beyond the toe to
plan with alarm levels tied to tolerable deflections of the allow water drainage behind the shotcrete face. After
soil around the west abutment and pier #4, and the slope preparations were complete, a 150 mm thick layer
of shotcrete was applied to the reinforced soil nail face
while hexagonal nuts and square plates were wet set on
each soil nail into the shotcrete. In total, 110 m3 of
shotcrete was installed on the site.
The shotcrete was used to tie the soil nails
together and was applied to a single row of soil nails at a
time. During the operation, there was use of a flash coat
of grout to mitigate sloughing issues, as discussed below.
After the shotcrete application at each stage,
excavation of the next stage was undertaken. However
excess sloughing started to arise after stage 6 resulting in
large voids (up to 600 mm) behind the already applied
shotcrete face. To mitigate this issue, a grout flash coat
was applied on the face after excavation was completed,
as depicted in Figure 3. This solution was implemented as
a temporary measure while a permanent solution was Figure 3. Permeation Grouting Operation
devised and implemented.
The permeation grouting process involved drilling 2-
meter-long permeation rods at a 75-degree angle, using a
water to cement ratio of 0.6. After the permeation grouting
was completed, no further instances of sloughing were
observed, indicating that the permeation grouting process
had effectively reinforced the soil and improved the
stability of the slope.

3.1 Drilled Shafts with Reinforced Concrete

A total of 92 drilled shafts were installed at the toe of the

slope cut, the drilled shafts comprised of a 300 mm
diameter, 8.5-meter-deep drilled hole. The shafts were
backfilled with 42 MPa self-consolidating concrete (SCC).
The welded rebar cage was carefully constructed with six
30M vertical rebars in a circular pattern, tied together with
10M rebar at 305mm vertical spacing. The prefabricated
Figure 1. Application of a grout flash coat. rebar cages were delivered to the site and installed after
the drilled shaft alignment was excavated to the design
This method proved to be insufficient for all stages, as elevation, followed by the completion of shaft drilling using
excessive sloughing occurred during the excavation of the following procedure:
stage 7 due to lack of fines in the observed material. In
order to address this issue, permeation grouting was 1. Position the drill mast over the marked shaft
carried out following the installation of soil nails and prior location.
to the excavation stage below the soil nails. This involved 2. Using a level on the mast, verify the drill mast is
the injection of a low viscosity grout mixture into the positioned perpendicular to the ground.
surrounding soil, which served to fill any gaps or voids 3. Slowly spin the 300 mm drill bit to start the shaft.
and improve soil stability. By utilizing this approach, the 4. Drill shaft to 8.60 m depth
necessary ground support was achieved to facilitate safe a. Attach additional drill steel as required.
and effective excavation, thereby ensuring the overall b. Suppress dust with water using the
success of the project. built-in dust suppression system.
c. Lift drill cutting by adding foam to the
dust suppression system.
5. Confirm the shaft depth using a water level
meter.
6. Move the drill to the next shaft location and
repeat steps 1 through 5.

During the installation of the drilled shafts, it was

observed that there was air communication between
holes. Although concrete communication was not visible,
it is possible that the high-pressure air used during the
drilling operation caused this behavior, potentially
affecting the quality of the drilled shafts. To mitigate this
issue, the drilling operation was carried out with every fifth
Figure 2. Section view showing permeation grouted zone.
pile being drilled to ensure adequate distance between DS415. 1984. The Danish code of practice for foundation
the drilled shafts. engineering. Danish Society of Civil Engineering.
Federal Highway Administration (FHWA), 1999.
3.2 Quality Assurance and Control Geotechnical Engineering Circular No.4- Soil Nail
Walls. United States Department of Transportation.
Quality assurance and control tests were conducted Federal Highway Administration (FHWA), 2002.
throughout the project to ensure the production met the Geotechnical Engineering Circular No.5- Drilled
required standards. A total of 14 proof tests were Shafts. United States Department of Transportation.
performed to verify the load capacity of the soil nails met Federal Highway Administration (FHWA), 2005. Micropile
the design requirements. The load applied during proof Design and Construction Reference Manual,
tests was 1.5 times the design load with creep period, and Publication No. FHWA NHI-05-039, United States
all the tested soil nails met the design load requirement. Department of Transportation.
To ensure the grout's quality, the grout mix was Federal Highway Administration (FHWA), 2015.
tested with a mud balance on the site, and it was verified Geotechnical Engineering Circular No.7- Soil Nail
that it met the required water-to-cement ratio. Additionally, Walls and Slope Stability. United States Department of
10 cubes were casted per grouting day and sent to a Transportation
third-party laboratory to conduct compressive strength Post-Tensioning Institute (PTI), 2014. Recommendations
tests in accordance with CCIL and CSA standards. All for Prestressed Rock and Soil Anchors, 5th ed., PTI
tests passed the design requirement. DC35.1-14, Phoenix, Arizona, USA.
The compressive strength of the shotcrete was PLAXIS. PLAXIS 2D 2020. User Manual. Delft, The
also tested by a third-party laboratory. Ten concrete Netherlands: Bentley Systems, 2020.
cylinders were casted per day of shotcrete operation for Terzaghi, K. and Peck, R.B. 1987. Soil Mechanics in
testing, and all the tests passed the design requirement. Engineering Practice, 2nd ed., McGraw Hill, New
Furthermore, two shotcrete panels were prepared during York, NY, USA.
shotcrete installation, and three cylinders were cored from .
each panel for testing.
The shotcrete face's movement was monitored by
a full-time survey multiple times per day, categorized into
three zones: Green (continue), Yellow (caution), and Red
(stop). During the project, Yellow and Red zones were
observed due to several factors, possibly including the
variation in ambient temperature during curing and slight
movement of the shotcrete wall due to sloughing during
drilling operation. To ensure safety, as there was no
visible cracking of soil nail popping out, thorough daily
visual inspections were maintained.

4 CONCLUSIONS

The success of this project was in large part due to a

collaborative effort between stakeholders. GSI prepared a
design package which was the result of teamwork from
GSI and the consultant. The importance of the asset to
the public safety was at the forefront throughout the pre-
project and construction phases and resulted in a safety-
first approach to providing a safe corridor for the required
infrastructure.
Unforeseen construction issues did arise and were
mitigated with solutions that upheld the safety of the
public and maintained project schedule. The
comprehensive understanding of the technical aspects
presented in this paper serves as a valuable resource for
future projects involving similar stabilization challenges.

5 REFERENCES

American Petroleum Institute (API). 1990. 13B-1

Recommended standard procedure for field testing
water-based drilling fluids, API, Dallas.
Canadian Geotechnical Society (CGS) 2006. Canadian
Foundation Engineering Manual, 4th ed., BiTech
Publishers Ltd, Richmond, BC, Canada
Challenges and solutions of 2D Digital
Image Correlation for Brazilian Tensile
Strength testing on hard rock
Émélie Gagnon, Timothy R. Packulak, Samuel K. Woodland,
Mark S. Diederichs, & Jennifer J. Day
Department of Geological Sciences and Geological Engineering,
Queen’s University, Kingston, ON, Canada

ABSTRACT
Brazilian tensile strength (BTS) testing is crucial for hard rock engineering projects. Despite being uncommon, there is
growing interest in monitoring strain during the test to assess true tensile strength, analyze geological effects, and study
fracture development. Given the heterogeneous strain fields in BTS tests, obtaining full-field strain maps using affordable
Digital Image Correlation (DIC) is preferred over limited-coverage foil strain gauge measurements (SG). However, applying
DIC in BTS testing is challenging due to the small strain expected in hard, low porosity rock (below 0.1% or 0.005 mm in
a 50-mm diameter specimen). This paper offers technical guidance on employing two-dimensional (2D) DIC in BTS
laboratory testing of hard rocks, covering hardware selection and speckle pattern application. It presents validation results
from a testing program that equipped BTS specimens with a lateral strain gauge on one side and DIC on the other.

RÉSUMÉ
Les essais de résistance à la traction brésilienne (BTS) sont essentiels pour les projets d'ingénierie de roches dures. Bien
que rare, surveiller la déformation pendant l'essai suscite un intérêt croissant afin d’étudier l’influence de l’hétérogénéité
géologique sur le développement des fractures. Il est souhaitable de mesurer la déformation sur la surface entière d’un
échantillon, une tâche possible avec la corrélation d'images numériques (DIC). Cependant, l'application de la DIC aux
essais de BTS est complexe en raison de la faible ampleur des déformations attendues, inférieures à 0,1 %, ou 0,005 mm
dans une carotte de roche de 50 mm de diamètre. Cet article conseille l'utilisation de la DIC bidimensionnelle (2D) dans
les essais BTS en laboratoire sur les roches dures. La sélection du matériel et les méthodes d'application du motif à pois
sont couvertes. Les résultats de validation consistant à équiper chaque spécimen de BTS d'une jauge de contrainte d'un
côté et de DIC de l'autre sont présentés.

1 INTRODUCTION Image Correlation (DIC) offers an alternative by measuring

strain over the entire specimen surface, but it requires
There are many motivations for measuring strain during expertise in optics, image processing, and computer vision;
Brazilian Indirect Tensile Strength (BTS) testing of rock. areas that may not fall within the realm of knowledge of
Firstly, it allows observing how specimen geometry and geotechnical engineering practitioners (Sutton et al. 2009).
platen contact influence load distribution, providing insights DIC involves capturing images of a surface undergoing
into the validity of theoretical assumptions and analytical deformation and correlating the position of local areas on
solutions (Hondros 1959, Claesson and Bohloli 2002). the surface through an optimization process based on the
Secondly, it enables the study of intact rock heterogeneity grey-level intensity values. DIC can be performed using a
and its impacts on fracture development within the single camera (2D-DIC) or multiple cameras (3D- or stereo-
specimen, aiding in understanding the variability in DIC). 2D-DIC is suitable for immobile, planar surfaces
emergent strengths across a range of tests and deriving strictly undergoing in-plane deformation oriented
representative tensile strength properties for complex orthogonal to the camera axis (Zhao et al. 2019). However,
lithologies. Finally, strain measurements can help identify any deviation from these conditions may introduce errors
the occurrence of the first crack at the center of the in displacement or strain measurements. This risk is
specimen, which is considered analogous to the true significant enough that the International Digital Image
tensile strength of the rock, as crack propagation is Correlation Society (iDICs) strongly recommends using
constrained by the BTS specimen geometry, leading to 3D-DIC, which allows for the dimensional decomposition of
peak strengths that exceed direct tensile strength (DTS) displacements, over 2D-DIC even when the
results (Perras and Diederichs 2014, Packulak et al. 2023). aforementioned conditions are met (2018). Nonetheless,
Although strain measurement during BTS testing is not for many rock mechanics practitioners, 2D-DIC presents an
required by the ISRM suggested method (ISRM 1978) or initial opportunity to enter into the use of this method,
ASTM standard (ASTM 2020), traditional foil strain gauges primarily due to the cost and knowledge barriers
(SG) have been commonly used (Patel and Martin 2018, Li associated with 3D-DIC. While comprehensive DIC
et al. 2019, Packulak and Day 2023). Lateral SGs have systems that include hardware, software, and training are
limitations as they only cover a central portion of the available, their cost poses a barrier for many research and
specimen (typically 10 to 20 mm), resulting in a biased view industry organizations. As a result, building a DIY DIC
of strain that fails to capture complex patterns. Digital system with separate hardware (camera, light, mounting)
and open-source algorithms has become a popular option Region of interest (ROI): The area over which strain
(Aliabadian et al. 2019, Arza-García et al. 2022). It is also is to be measured. In this case, the entire BTS specimen
much simpler to use a single camera, than to start cross-sectional area. Hence the ROI is a circular area with
synchronizing two cameras, especially considering the a diameter of 47.6 mm, 63 mm for common NQ and HQ
already steep learning curve for multidisciplinary users. core, respectively.
The objective of this study was to assess the degree of Field of view (FOV): The area captured in a photo
error and the practical viability of using 2D-DIC for BTS should be small enough to fill most of the frame with the
testing of low-porosity rock. The aim was to develop a specimen while ensuring the specimen remains in the
standardized procedure that serves as an alternative to frame throughout the test. In BTS testing, the FOV can be
strain gauges and ensures repeatability across multiple maximized (i.e. filled by the ROI) because the specimen
users. To make the adoption of DIC more accessible to can be treated as stationary during the test.
other practitioners, open-source DIC software was used, Stand-off distance (SOD): The distance of the camera
and reports on the hardware selection process are and the test piece, which affects the FOV. The FOV can be
included. The final DIC system design selected here is not estimated using Equation 1, considering the width of the
intended to be universally applicable. However, sharing the camera sensor (𝐿𝐶𝑆 ) and the focal length of the lens (FL):
process undertaken for its selection aims to facilitate
implementation and validation of DIC systems in other 𝐹𝑂𝑉 = 𝐿𝐶𝑆 (
𝑆𝑂𝐷−𝐹𝐿
) [1]
laboratories interested in using DIC for BTS testing of rock. 𝐹𝐿

Reducing the SOD in 2D-DIC increases errors arising

2 FUNDAMENTALS OF DIC from out-of-plane motion (Arza-García et al. 2022).
Depth of field (DOF): The distance, parallel to the
This section provides an ad hoc overview of DIC and its camera axis, between the nearest and furthest objects that
application in BTS testing, specifically focusing on are in focus in a photo. This is a minor consideration for
monitoring strain. Prior to testing, a speckle pattern 2D-DIC and BTS testing because no out-of-plane
consisting of black and white stochastic features is applied movement is expected. Nonetheless, having sufficient
to the surface of the specimen. This pattern ensures that DOF helps ensure focus is maintained, reducing sensitivity
local subsets of pixels will have unique and distinguishable to alignment and unexpected out-of-plane motion.
grey level intensity distributions, ranging from 0 (black) to Spatial Gradients: The variations in strain across the
255 (white). The characteristics of this speckle pattern will ROI. Experience in the Queen’s Advanced Geomechanics
be discussed in more detail in Section 4. During the test, Testing Laboratory (QAGTL) has shown that maximum
photos are captured using a stationary camera positioned strains of 1E-03 are expected during BTS testing on low-
perpendicular to the surface of the BTS specimen. This porosity hard rock. A minimal sensitivity of 1E-04 is
camera must remain immobile relative to the specimen required to capture most of the elastic & plastic strain
throughout the testing process. After the test, the acquired development throughout the test. This is estimated to be
photos undergo processing using a DIC algorithm. This equivalent to minimum displacements of 0.005 mm.
algorithm utilizes matching criteria such as the sum of least Noise floor: The maximum level of random noise in the
squares or cross-correlation to locate the subsets in each DIC measurement that allows for extraction of meaningful
photo (Sutton et al. 2009). It assumes homogeneous strain information from the image data. Unloaded specimen
deformation within each subset, the size of which the user photos taken in the testing configuration are used to
selects based on camera resolution and the desired scale estimate the noise floor. As no deformation is expected in
of the deformation of interest. To describe the movement these photos, any computed displacement is considered
of each subset, a simple 6-component deformation vector noise. A target noise floor of 5E-04 mm was selected, as
is utilized. The first two components represent rigid body an order of magnitude lower than the desired sensitivity.
translation in the horizontal and vertical directions, while Frame rate: The temporal resolution of the
the remaining four components describe various types of measurement. A higher frame rate signifies more data and
pure and simple shear. By the end of the DIC process, the processing time. A minor consideration is that the amount
user is provided with a comprehensive matrix of of displacement between frames should be kept to
displacement and strain values that cover the entire region approximately less than a subset size to ensure proper
of interest (ROI), allowing for a detailed analysis of the subset matching (iDICs 2018). BTS specimens are loaded
specimen's behavior during the test. at a rate of 0.06 kN/s in the QAGTL, to be in accordance
with the ISRM suggested method and ASTM standard.
Applications interested in monitoring crack propagation
3 2D-DIC SYSTEM DESIGN FOR BTS TESTING require frame rates on the order of 1000-200k Hz (Xie et
al. 2022). In this case, the goal was simply to observe the
Before conducting a DIC measurement, it is essential to development and localization of strain throughout the test
establish the expectations and requirements of the and a frame rate varying from 1-22 Hz was sufficient.
mechanical test to ensure that the chosen DIC system can Exposure Time: The duration for which the camera
meet those demands without exceeding its limits (iDICs sensor is exposed to light and records a picture. For quasi-
2018). The specific requirements for BTS testing are static situations like BTS testing, motion blur from
outlined below. These requirements guided the selection of increased exposure time is a minor consideration.
hardware and optical and post-processing parameters. Increased exposure time simply reduces the required
object illumination and increases contrast. In general, the
exposure time should be no greater than the inverse of the 3.2 Software selection
frame rate (iDICs 2018).
Synchronization and Triggering: The method used to The selection of a DIC algorithm prioritized open-source
synchronize the strain measurement with the loading frame options, considering the availability of support and
data and other strain measurement methods employed. In documentation, as well as the software's track record in
this case, data acquisition was manually initiated at the terms of accuracy, reliability, and validation. The open-
initiation of loading, and each photo was recorded with a source package Ncorr was chosen due to its cost-free
precise time stamp for subsequent manual synchronization nature, integration within the MATLAB software platform,
with machine data. and previously published verification, providing confidence
in its accuracy and reliability (Blaber et al. 2015, Mathworks
3.1 Hardware selection Inc. 2022).

The suitability of three different cameras for strain Table 1. Parameters for the three cameras considered.
monitoring during BTS testing was evaluated. The cameras
considered were two machine vision GigE cameras, with Lucid Vision Labs Fujifilm
resolutions of 1.5 MP and 5 MP, and a 26.1 MP mirrorless Camera
Phoenix Triton X-T4
camera. Parameters and characteristics of each Resolution 1.6 MP 5 MP 26.1 MP
camera/lens pair are presented in Table 1. Practical (pixels) 1440x1080 2448x2048 6240x4160
considerations guided the selection of these imaging Edmund Optics FUJI XF
systems. It is worth noting that each camera has a Lens
C Series Macro
monochromatic option, square pixels, and a global shutter, Focal L* 12 mm 25 mm 80mm
as recommended by iDICs (2018). Machine vision GigE
Aperture F/1.8 F/1.4-F/16 F/2.8
cameras are commonly used in industrial applications due
LD* (%) < 2.5 < 0.7 0
to their robustness, high frame rates, and compatibility with
FOV** 158 mm 135 mm 126 mm
standard software for image processing. The inclusion of
cameras with different resolutions aimed to evaluate the (pixel/mm) 9 19 49.5
influence of resolution on the accuracy and precision of (mm/pixel) 0.11 0.055 0.02
strain measurements. Additionally, the mirrorless camera SOD (m) 45 cm 45 cm 60 cm
with a higher megapixel count was included to explore the DOF (mm) 32 14 11
potential benefits of utilizing consumer-grade cameras, FR* (FPS) 71.6 22 1
which often feature advanced sensor technology and ET* (ms) 14 24 77
enhanced image quality. Aperture
f/2.8 f/4.0 f/5.6
Appropriate lenses were selected for each camera to used
match the required FOV, SOD, and DOF. Lens distortion, Speckle
0.2-0.4 0.2-0.3 0.06-0.1
if uncorrected, is one of the main causes of error in DIC Size (mm)
measurements (Zhao et al. 2019). For 2D-DIC, the use of *Focal L = Focal Length. LD = Lens Distortion. FR = Frame
bilateral telecentric lenses is recommended as they provide rate. ET = Exposure Time.
constant magnification and can minimize errors resulting **The maximum dimension (width) of FOV is presented.
from out-of-plane translation (Sutton et al. 2009). If
standard lenses are preferred, due to cost considerations,
availability, or space constraints, it is advisable to maximize 4 SPECIMEN PREPARATION: SPECKLE PATTERN
the focal length and SOD (iDICs 2018). Only standard
lenses were considered in this study and the selection To provide unique features for tracking in the ROI, a black
process prioritized lenses with low distortion and low cost. and white stochastic speckle pattern is applied to the
To ensure accurate DIC measurements, it is crucial to specimen's surface prior to testing. The size of the
have a camera/lens mounting system that provides precise speckles significantly impacts measurement resolution. If
adjustability, stability during testing, and minimizes camera the speckles are too small, aliasing, reduced contrast, and
vibrations (iDICs 2018). For this purpose, a sturdy tripod image blurring can occur. Conversely, if the speckles are
equipped with vibration suppression pads was used to too large, measurement precision is compromised. It is
mount the camera perpendicular to the BTS specimen. ideal to have speckles sized at 3-6 pixels to ensure
Consistent and uniform lighting across the ROI during adequate grey level variability (Sutton et al. 2009).
testing was maintained by employing additional lighting in Achieving the sub-millimeter speckle sizes required
the QAGTL. For an 8-bit camera, photos with a contrast of when working with low strains in rock testing is challenging.
approximately 50% (equivalent to 130 grey-level counts Conventional methods such as spray painting and
between light and dark features) should be taken to attain airbrushing often create larger features. Patel and Martin
a displacement accuracy of around 0.005 pixels (Reu (2018) faced this challenge when trying to perform DIC on
2015). A diffuse or cross-polarized light source is Lac du Bonnet Granite BTS specimens. They required a
recommended, and accordingly a 5500 K diffused light density of 250 speckles/cm2 (speckle size of 0.5 mm) and
source placed directly behind the camera in this work. ultimately used an ultra-fine permanent marker to
individually draw speckles. This difficulty in preparing
suitable speckle patterns has hindered the widespread
adoption of DIC in practical rock testing laboratories, where
spending hours on specimen preparation is impractical.
In response to this limitation, research conducted at the
QAGTL focused on optimizing speckle pattern preparation
methods. Four methods, namely spray paint, airbrush,
stamp, and laser, were carefully refined to serve as viable
options for BTS specimen preparation. The first three
method involve a similar approach, with the application of
a white layer of paint covered by black features. On the
other hand, the laser method entails applying a thin layer
of black paint, overlaying it with white paint, and then laser-
engraving a computer-generated speckle pattern to
expose the underlying black paint. These optimized
methods were each tested on five BTS specimens of
Westerly Granite, where DIC was applied on one side and
a strain gauge (SG) on the other. An overview of each
technique is provided in Figure 1 and detailed procedures
are provided in Table 1. Further results and discussion can
be found in Woodland et al. (2023).

5 EXPERIMENTAL VALIDATION

Because rock is inherently heterogeneous and its true

strain cannot be verified, experimental validation was
conducted as an alternative. In this validation process, DIC
strain measurements obtained from one side of a BTS
specimen were compared to stress-strain curves derived
from a lateral strain gauge (SG) placed at the center of the
other side of the specimen, which served as the known true
strain value. To facilitate this comparison, virtual
extensometers or strain gauges were incorporated into the
DIC software. These virtual gauges allow users to locate
specific points and calculate the strain differences between
them. A virtual SG was placed in the same location as the
physical SG for each test, and a DIC stress-strain curve
was extracted. The systematic and random errors of the
DIC measurements were evaluated by calculating the
average and standard deviation of the differences between
the SG and DIC strain values for each 0.01 MPa increment
in stress.
Because this paper discusses two main optimization
studies: camera selection and speckle pattern preparation,
the studies are interconnected and influenced each other.
In the camera selection study, laser speckle patterns were
prepared and optimized for each camera, taking into
consideration the desired feature sizes listed in Table 1.
Various rock types, including Westerly granite (W), Norite
(N), Key Anacon (KA), and Pointe-du-Bois granite (PDB),
were tested based on their availability in the QAGTL. The
objective was to find specimens with homogeneous
mineralogy to ensure that the theoretical strains developing
on each side of the specimen were identical. Despite the
variation in rock types, it is expected that the comparison
of camera performance will not be significantly impacted.
This is because the DIC strain measurements were
individually compared to the SG measurements, providing
a normalized error value that can be compared
independently of the rock type, ensuring a fair evaluation.
The experimental setup for each camera is described in Figure 1. An overview of the four speckle pattern
Table 1 and Section 3. application methods optimized for BTS specimens.
Table 2. Detailed procedures for the preparation of speckle patterns on BTS specimens (from Woodland et al. 2023)

Method Speckle Pattern Preparation Procedures for BTS Specimens

i. Position the specimens perpendicular to the spray paint nozzle, approximately 25 cm away.
ii. Apply thin, even coats of white matte spray paint until opaque (typically two (2) passes).
Spray iii. Allow the white paint to dry for approximately 15-30 minutes.
Paint iv. Position the specimens about 40 cm from the paint can for the desired 0.2 mm speckle size.
i. Apply black paint in short bursts of less than a second, around 5 bursts per specimen, until a
satisfactory speckle coverage and density are achieved.
ii. Align specimens in an airbrush spray booth with flat faces perpendicular to and about 15 cm away
from the airbrush nozzle.
iii. Apply thin, even coats of white acrylic paint by quickly moving the gravity fed airbrush nozzle across
Airbrush the specimen faces, alternating direction until coverage is achieved (typically six (6) passes).
iv. Allow the white paint to dry for approximately 5-10 minutes.
v. Create the black speckles using black acrylic paint in a 0.3 mm airbrush nozzle set to 30-35 PSI
(~207-241 kPa) and a tongue depressor as shown in Figure 1b.
i. Generate a 60 x 60 mm speckle pattern with 0.2 mm speckles using Correlated Solutions Speckle
Generator (2022) (75% density, 99% variation).
ii. Use a 3D laser engraving machine to burn the pattern into a rubber stamp base and create a stamp.
Stamp iii. Apply an even base white coat to the specimen surface using spray paint or an airbrush.
iv. Apply black ink speckles using a water-based pigment ink pad and the stamp. Rotate the stamp for
multiple applications if necessary (5-6 was found to be optimal in this work).
v. Allow several days (at least 48 hours) for drying before testing.
i. Apply an even base black coat to the specimen surface using an airbrush.
ii. Allow the black paint to dry for approximately 5-10 minutes.
iii. Apply thin and even layers of white paint using the airbrush until opaque. This is important to ensure
a homogeneous burning effect during the laser process.
iv. Import a computer-generated speckle pattern (same as for the stamp) onto a Glowforge laser cutter.
Laser
Adjust the laser intensity (~20%) and engraving speed (~40%) for optimal burn. Calibrated
parameters can be used for multiple specimens if the painting thickness is consistent.
v. Position the specimen inside the laser engraving machine using custom 3D printed specimen holders
to ensure a level surface within the laser focus range (up to 12.7 mm).
vi. Focus the laser head onto the specimen surface and engrave (approximately 15 minutes).

For the speckle pattern optimization study, only deformation measurement is 0.01 mm, which is too large
Westerly granite specimens were used. The 5MP machine for accurate strain measurements.
vision camera with constant optical and post-processing The DIC strain readings obtained using the 26.1 MP
parameters was used for all specimens. camera exhibited a maximum error that is equivalent to the
mean strains expected during a test. Most of the DIC
stress-strain curves obtained using this camera exhibited a
6 RESULTS & DISCUSSION regular sinusoidal pattern, which is not expected from a
geological standpoint. Upon troubleshooting, it was
First, the results of the comparison between different discovered that the error in strain readings is likely
cameras for 2D-DIC are discussed in this section. Mean attributed to camera heating. Thermistors placed on the
strain differences calculated between the SG and DIC camera body recorded a temperature increase of 7°C over
measurements are summarized in Figure 2. Each camera a 1.5-hour shooting period. This increase in temperature
was tested in four to six experiments. can introduce errors into the DIC measurements due to
The DIC strain readings obtained using the 1.6 MP thermal expansion of the cameras and lenses, as well as
camera showed significant errors. The mean error value is drift induced by thermal expansion of the camera mounts
approximately equivalent to the mean strains expected (iDICs 2018). The camera's temperature had not yet
during a test, indicating that the measurements are highly stabilized even after the battery died and had to be
inaccurate. This suggests that the resolution of the camera recharged. To mitigate these errors, it is usually
was too low to accurately capture the small strains typically recommended to allow cameras to reach a stable operating
observed in BTS testing. DIC algorithms are generally temperature before shooting. Unfortunately, this was not
capable of solving displacements with a precision of 0.1 to possible given the battery life of the 26 MP camera. While
0.01 pixel (Sutton et al. 2009), but with a resolution of 0.11 machine vision cameras are often optimized for long-term
mm/pixel, the conservative estimate for the precision of shooting and temperature stability, it should be noted that
not all consumer-grade cameras possess these features.
 The 5 MP machine vision camera demonstrated an resulting in more consistent splatter sizes. The laser
error below 1E-04, which falls within an acceptable range speckle pattern method demonstrates relatively small
for the tests conducted. Based on these results, the 5 MP differences between DIC and SG data, indicating high
camera was chosen as the preferred option for 2D-DIC repeatability and quality of speckle patterns. The use of a
testing in the QAGTL laboratory. These findings emphasize laser engraving machine ensures consistent speckle sizes,
the successful application of 2D-DIC for BTS testing when shapes, and densities, with the only variability coming from
a camera/lens system that meets the spatial and temporal different paint layer thicknesses. This method provides
resolution requirements is selected. It is crucial to carefully more accurate strain measurements compared to the
consider the capabilities of the camera and lens to ensure stamp and spray paint methods. Ultimately, the choice of a
accurate and reliable strain measurements. The speckle pattern preparation method for 2D DIC strain
performance of the 5 MP machine vision camera highlights measurements comes down to cost, repeatability, and
the importance of selecting appropriate equipment for accuracy. However, the laser speckle pattern method
achieving reliable results in DIC analysis for BTS testing. stands out as a streamlined and reliable approach, making
it an ideal choice for laboratory and industrial
environments.
When comparing the accuracy of airbrush and spray
paint speckle patterns, the spray paint method shows
higher median errors and greater standard deviation of
systematic error. Difficulties in controlling spray paint
splatter during pattern preparation led to inconsistency, as
subtle variations in finger pressure can significantly impact
the size of the painted speckles. In contrast, the airbrush
method offers better control and consistency by allowing
the setting and maintenance of specific pressures,
resulting in more consistent splatter sizes.

Figure 2. The average systematic error in DIC strain

measurement obtained using each camera. The specimen
lithologies tested are included above each box.

The results of the comparison between different

speckling methods are summarised in Figure 3. The stamp
method for speckle pattern preparation demonstrates the
poorest performance in terms of systematic error
magnitude and standard deviation. The 2D-DIC Figure 3. The average systematic error in DIC strain
measurements of stamped specimens do not accurately measurement obtained in 5 tests using each speckle
represent the expected strains, possibly due to aliasing pattern preparation method.
caused by small speckles produced by the stamp. As a
result, the stamp application method is discouraged for 2D- The optimized 2D-DIC setup in the QAGTL has been
DIC strain measurement in BTS testing on low porosity successfully employed for testing over 50 specimens of
rocks. When comparing the accuracy of airbrush and spray heterogeneous rock, exemplifying its capabilities. A
paint speckle patterns, the spray paint method shows representative example is shown in Figure 4, featuring a
higher median errors and greater standard deviation of sample from the Nepisiguit Falls Formation of the
systematic error. Difficulties in controlling spray paint Tetamagouche Group in the Bathurst Mining Camp, New
splatter during pattern preparation led to inconsistency, as Brunswick. This specific specimen is a coarse-grained,
subtle variations in finger pressure can significantly impact crystal-rich, tuffaceous sedimentary rock with moderate
the size of the painted speckles. In contrast, the airbrush foliation, composed of quartz, plagioclase feldspar,
method offers better control and consistency by allowing muscovite, epidote, and chlorite.
the setting and maintenance of specific pressures,
Figure 4. An overview of the SG and DIC strain measurements for a foliated Meta-Sediment BTS specimen.

The strain maps obtained through DIC analysis reveal rock heterogeneity and fracture development. This paper
that fractures primarily initiated along weak foliation layers focused on the practical viability of using 2D Digital Image
at the platen contacts. Notably, the presence of a low Ψ Correlation (DIC) for BTS testing of low-porosity rock. By
angle (dip with respect to the core axis) resulted in certain selecting appropriate hardware, and using a novel laser
foliation layers exhibiting a tendency to flake off, indicated speckle pattern application method, a standardized
by higher strain concentrations at the top center of the procedure was developed to make DIC more accessible to
specimen. Minor discrepancies between strain gauge (SG) practitioners. Overall, the findings showed that 2D-DIC can
and DIC measurements are likely due to the be used to measure strain during BTS testing of rock.
heterogeneous distribution of fractures across the
specimen, rather than measurement error.
The true tensile strength (TTS) of the specimen can be 8 ACKNOWLEDGEMENTS
determined from SG data (Packulak et al. 2023), but is not
clearly evident in the DIC strain map or the extracted This research was financially supported by Canada’s
stress-strain curve, which is presented in its raw, unedited Nuclear Waste Management Organization (NWMO), the
form. Strain maps play a crucial role in interpreting a series Department of Geological Sciences and Geological
of BTS test results on heterogeneous specimens, providing Engineering at Queen’s University, and the Natural
valuable insights into crack initiation that may deviate from Sciences and Engineering Research Council of Canada
the assumption of initiation at the center of the specimen. (NSERC) through the Undergraduate Student Research
These findings demonstrate the significance of strain maps Award (held by S.K. Woodland), the Canada Graduate
in understanding fracture behavior and highlight the Scholarship (Master’s) Program (held by É. Gagnon), the
capabilities of the optimized 2D-DIC setup in capturing Alexander Graham Bell Canada Doctoral Graduate
complex deformation patterns in heterogeneous rocks. Scholarship Program (held by T.R.M. Packulak, CGS-D
One limitation of 2D-DIC appears to be its ability to 535289-2019), and the Discovery Grants Program (held by
observe strain localization early in the test. In studies J.J. Day).
utilizing expensive commercial 3D-DIC systems, strain
localization has been successfully observed early in the
test (Patel and Martin 2018). In contrast, the optimized 2D- 9 REFERENCES
DIC setup discussed here primarily observes strain
localization towards the later stages of tests (shown in Aliabadian, Z., Zhao, G.F., and Russell, A.R. 2019. Crack
Figure 4), although the average strains are still relatively development in transversely isotropic sandstone discs
well captured earlier in the tests. Depending on the specific subjected to Brazilian tests observed using digital
goals and requirements of the application, practitioners image correlation. Int. J. Rock Mech. Min. Sci., 119:
may find that the additional cost of a 3D-DIC system is 211-221.
warranted. However, for many applications, the capabilities Arza-García, M. et al. 2022. Evaluation of a low-cost
of 2D-DIC may be sufficient. approach to 2-D digital image correlation vs. a
commercial stereo-DIC system in Brazilian testing of
soil specimens. Arch. Civ. Mech., 22: 13 pages.
7 CONCLUSIONS ASTM 2020. D2936-20: standard test method for direct
tensile strength of intact rock core specimens. ASTM
In conclusion, measuring strain during Brazilian Tensile International, West Conshohocken
Strength (BTS) testing of rock offers several benefits,
including validating theoretical assumptions, and studying
Blaber, J., Adair, D., and Antoniou, A. 2015. Ncorr: Open-
Source 2D Digital Image Correlation Matlab Software.
Exp. Mech., 55(6): 1105-1122.
Claesson, J. and Bohloli, B. 2002. Brazilian test: stress field
and tensile strength of anisotropic rocks using an
analytical solution. Int. J. Rock Mech. Min. Sci., 39:
991-1004.
Correlated Solutions 2022. Speckle Generator. Correlated
Solutions. https://correlated.kayako.com/article/7-
speckle-generator. Accessed 26 January 2023.
Hondros, G. 1959. The evaluation of Poisson’s ratio and
modulus of materials of a low tensile resistance by the
Brazilian (indirect tensile) test with particular reference
to concrete. Aust. J. Appl. Sci., 10: 243-268.
iDICs (International Digital Image Correlation Society),
Jones, E.M.C., and Iadicola, M.A. (ed) 2018. A Good
Practice Guide for Digital Image Correlation. DOI:
10.32720/idics/gpg.ed1/print.format
ISRM 1978. Suggested methods for determining tensile
strength of rock materials. Int. J. Rock Mech. Min. Sci.
Geomech. Abstr., 15(3): 99–103.
Li, H., Lai, B., Liu, H. 2019. Determination of Tensile Elastic
Parameters from Brazilian Tensile Test: Theory and
Experiments. Rock Mech. Rock Eng., 52: 2551-2568.
Mathworks Inc. 2022. MATLAB version 9.13.0 (R2022b),
Natick, Massachusetts: https://www.mathworks.com
Packulak, T.R.M. et al. 2022. Determining true tensile
strength from Brazilian tensile strength laboratory
testing. In the proceedings of: GeoCalgary 2022
Reflection on Resources. 6 pages.
Packulak, T.R.M., and Day, J.J. 2023. The Impact of Axial
Load Distribution on Brazilian Tensile Testing on Rock
Rock Mech. Rock Eng. (Paper Accepted).
Packulak, T.R.M. et al. 2023. Measurement of true tensile
strength from Brazilian tensile strength laboratory tests.
Submitted to: Can. Geotech. J. April 2023. [Under
review]
Patel, S., Martin, C.D. 2018. Evaluation of Tensile Young’s
Modulus and Poisson’s Ratio of a Bi-modular Rock
from the Displacement Measurements in a Brazilian
Test. Rock Mech. Rock Eng., 51: 361-373
Perras, M.A., and Diederichs, M.S. 2014. A review of the
tensile strength of rocks: Concepts and testing.
Geotech. Geol. Eng., 32: 525-546.
Reu, P. 2015. All about speckles: Contrast. Exp. Tech.
39(1): 1-2.
Sutton, M.A., Orteu J.J., and Schreier H.W. 2009. Image
Correlation for Shape, Motion and Deformation
Measurements, Springer, New York, NY, USA.
Woodland, S.K. et al. 2023. Paint speckle application
recommendations for digital image correlation analysis
of Brazilian tensile strength tests on low porosity rocks.
Submitted to: Rock Mech. Rock Eng., April 2023.
[Under review]
Xie, F., Xing, H., and Wang, M. 2022. Evaluation of
processing parameters in high-speed digital image
correlation for strain measurement in rock testing. Rock
Mech. Rock Eng., 55: 2205-2220.
Zhao, J., Sang, Y., and Duan, F. 2019. The state of the art
of two-dimensional digital image correlation
computational method. Eng. Rep., 1: 15 pages.
The Effects of Temperature on the Axial
Response of Fully Grouted Rock Bolts
Utilizing Distributed Fiber Optic Sensors
Chuyue (Chelsey) Guo, Kieran Moore & Nicholas Vlachopoulos
Department of Civil Engineering – Royal Military College of
Canada (RMCC), Kingston, Ontario, Canada

ABSTRACT
Fully grouted rock bolts (FGRB) are adaptable to a wide range of ground conditions and are employed in diverse tunnelling
environments from permafrost to geothermally active sites and from hard rocks to weak rock masses. As underground
operations become more complex, temperatures at the extremes will become more prevalent which can potentially affect
the bonding and performance of the FGRB. Existing literature exploring effects of temperature on these systems is limited.
A detailed laboratory investigation is currently being conducted, leveraging the distributed fiber optic sensors capability
developed by the research group of the authors. This paper summarizes historical research and innovations of the
methodology and its expansion into temperature effects. This research aims to bridge selected gaps in the existing
literature and determine the true performance of these systems within specific temperature regimes.

RÉSUMÉ
Les boulons entièrement scellés (BES) sont adaptables à une large gamme de conditions de sol et sont utilisés dans des
environnements divers, allant du pergélisol aux sites géothermiquement actifs et des roches dures aux masses rocheuses
faibles. À mesure que les opérations souterraines deviennent plus complexes, les températures aux extrêmes deviendront
plus fréquentes, ce qui peut potentiellement affecter l'adhérence et les performances des BES. La littérature existante
explorant les effets de la température sur ces systèmes est limitée. Une enquête de laboratoire détaillée sera menée en
tirant parti de la capacité des capteurs à fibres optiques distribuées développée par le groupe de recherche des auteurs.
Cet article résumera la recherche historique et les innovations de la méthodologie et de l'expansion des effets de la
température. Cette recherche vise à combler l'écart dans la littérature existante et à déterminer les performances de ces
systèmes dans des régimes de température spécifiques.

1 INTRODUCTION support is to maintain confinement in the rock mass so that

it remains self-supported (Vlachopoulos et al. 2013).
Underground excavations and tunnelling are major
engineering endeavours and have been critical capabilities
for the advancement of human civilizations. They had an
important role in Ancient Egyptian burial and religious
construction works (El Salam 2002) and used for military
operations, water distribution, and transportation of Ancient
Romans (Castellani and Dragoni 1997). In the modern era,
they have vital infrastructure roles such as transportation,
nuclear waste management, underground research
facilities, communications, mining operations, utilities
distribution, hydroelectric power generation, and flood
control, among others. Figure 1. Example of support and reinforcement elements
Over the millenniums, methods and technologies in used in underground excavations (from Forbes et al. 2018)
underground construction and engineering design have
experienced significant innovations resulting in projects Fully grouted rock bolts (FGRB) surged in popularity
increasing in size and complexity. One of the main since their introduction as they are versatile and adaptable
challenges continues to be rock stability. Removal of rock to a wide range of ground conditions. In order to effectively
during excavation causes a disturbance in the stress design support systems, in situ geological conditions and
equilibrium resulting in rock displacement (i.e. radial support element behaviours need to be well understood. A
convergence). In an underground construction project, a multitude of global research programs have been
combination of the different support and reinforcement undertaken, however, a full-range of mechanical behaviour
types will be utilized as seen in Figure 1. The types chosen and interactions are still not completely understood. This is
are dependent on the extent of fractured or loosened due in part to past technological monitoring and sensing
material around the excavation and the ground limitations and a lack of spatial resolution in conventional
type/conditions (Hoek and Wood 1987). The purpose of instrumentation during experimental testing. An innovative
strain monitoring technique was proposed by Vlachopoulos
et al. (2014) utilizing distributed fiber optic sensors (DOS). below freezing over a period of years (Jenness 1949). The
This method enabled continuous strain profiling and minimum frozen period is generally considered two
capturing of behaviour at the micro-scale. The technique consecutive winters. The coloured regions in Figure 2
has been developed with corporate sponsors and improved represent discontinuous and continuous permafrost areas.
through a series of studies into FGRB by the research Deep profiles of mean ground temperatures at various
group of the authors providing an unprecedented spatial monitoring stations have sub-zero temperatures up to 400-
resolution of 0.65 mm (Forbes 2015; Cruz 2017; O’Connor 650 m below ground. The most northern CAF site is CFS
2020; Moore 2021). Alert located on Ellesmere Island NV where the mean
Mining and civil engineering operations have pushed ground temperature up to 15 m below surface is -15°C.
deeper underground and expanded to more extreme
climatic areas leading to a rise of the demands and risks of
these projects. FGRB are employed in a diverse range of
ground environments from permafrost to geothermally
active tunnels. Research into the effects of temperature on
the performance of these systems is limited in existing
literature. This paper details the historical research on this
subject, the background and existing monitoring
methodology developed by the authors, and recently
proposed improvements with results from testing the initial
specimens within the conceptual design.

2 BACKGROUND

There is a large range of ground conditions that

underground projects encounter and must adapt to in order
to maintain safe operations. Two operational temperature Figure 2. Map of permafrost in Canada (modified after
extreme scenarios in Canada were considered during the Natural Resources Canada 2016)
conceptualization of this investigation: 1. Canadian
interests in the North and 2. Nuclear waste repositories. As activities in the North increase, understanding
Canada’s northern region is abundant in natural frozen ground behaviour and monitoring the interactions of
resources. Increasing temperatures in the Arctic has support systems in these remote regions will be essential
improved accessibility and demand for resources has for infrastructure support and future regional
increased global interest in the region. The Canadian developments; especially within the context of climate
Armed Forces (CAF) has a key role in demonstrating change.
Canada’s sovereignty in the North through permanent
presence and patrol activities. The CAF has several sites 2.2 Nuclear Waste Repository Environment
across the North including Yellowknife NWT, Iqaluit NU,
and Whitehorse YT. These support training, exercises, and Deep geological repositories are generally agreed to be
operations; the infrastructure is located on permafrost. one of the more efficient and safe long-term storage
(Government of Canada 2022) solutions for HLRW, which will require isolation periods
As well, Canada has generated electrical power for between 105 to 106 years. They provide a significant
over half a century using the CANDU nuclear reactors and geological barrier for the waste and decrease the risks of
a by-product of the process is spent nuclear fuel. CANDU future condition changes as the effects of changes
fuel, which is the main nuclear fuel in Canada, is a solid decrease with depth (Gibb 1999). Many countries including
that consists of uranium dioxide. Existing used nuclear fuel Canada, China, Finland, France, Germany, India, Italy, and
is temporarily stored at or near power generation sites. The Japan have decided to use this method. The total used
Adaptive Phased Management (APM) is Canada’s plan for nuclear fuel bundles at the end of operations of existing
the long-term management of used nuclear fuel, which will reactors are estimated at 5.5 million. The estimates will
be contained and isolated in a deep geological repository change depending on if reactor operations are terminated
between 500-1000 m below ground. This approach is in- early or extended longer. When the used nuclear fuel
line with best practices of the international community for bundle is removed from the reactors, they are placed in
high-level radioactive waste (HLRW). (Nuclear Waste water pools to decrease temperature and radioactivity for
Management Organization 2023) seven to ten years before being placed in dry storage
(Nuclear Waste Management Organization 2023).
2.1 Permafrost Environment The general design of deep geological repositories
include a series of shafts and tunnels with the containment
Almost half of Canada’s landmass and areas of the of the HLRW canisters inside a buffer material (commonly
western Arctic seabed is underlain with permafrost. bentonite) and surrounded by the natural geological barrier
Permafrost is any soil, or even bedrock, irrespective of its (i.e. surrounding rock) (King et al. 2017; Xue et al. 2021).
texture, degree of induration, water content, or geological The typical thermal criterion for peak temperature at the
character, in which the temperature has been continuously surface of the canister or interior of the buffer material is
100°C in order to control thermal stress and maintain
mechanical stability of the repository. Xue et al. (2021) 2019; Huang 2021). At higher temperatures attention is
presented mathematical models as solutions for required to ensure water does not evaporate during grout
temperature development and distribution during the curing (Li et al. 2017b).
exothermic process of nuclear waste. As seen in Figure 3, There has been limited studies investigating the effects
the temperature decays exponentially with distance from of surrounding rock mass temperatures on FGRB. Zhang
the canister centre in the local solution and slowly et al. (2021) studied the effects of freeze-thaw cycles and
decreases at the global scale. The average temperature sulfate attacks on cement grouted short embedment length
first increases with disposal time and then decreases. FGRB. The compressive strength and shear strength of the
cement grout was found to decrease with increasing
freeze-thaw cycles as the pores in the grout were enlarged
by ice stress during freezing.
When rock bolts are installed in geothermally active
tunnels, physical and chemical reactions may occur
resulting in bond behaviour degradation and structural
damage. The studies tested short embedment lengths with
cement grouts from 20°C to 70°C. The results found there
to be an initial strength increase at the lower test
temperatures but after 35°C the ultimate pull-out strength
decreased with increasing temperature (Li et al. 2017b;
Figure 3. Temperature change with distance from Wang et al. 2018). Li et al. (2017a) through regression
repository centre: a. Local; b. Global (modified after Xue et analysis proposed the relationship between temperature
al. 2021) and ultimate strength followed a cubic polynomial function.
Li et al. (2017b) found that the rate of exponential decay of
2.3 Temperature and Fully Grouted Rock Bolts axial force and shear stress decreases along the
embedment length as temperature increases, with the
The first use of rock bolts in tunnel support dates as early slowest being at 50°C. However, with an increase in
as 1913 gaining significant widespread use in the 1970s temperature, the maximum shear stress on the bolt-grout
(Bobet and Einstein 2011). Rock bolts have advanced interface decreased and its distribution along the
significantly in the past 40 years with increased embedment length became more uniform thus, concluding
understanding of load transfer mechanisms and that FGRB are not greatly impacted by geothermal activity.
innovations in the technology. The basic assembly of a Effects of temperature has yet to be extensively
rock bolt consists of a plain steel rod that is mechanically investigated and existing studies have focused on the
or chemically anchored into the rock mass on one end and higher temperature range. The studies used short
a faceplate and nut on the other (Vlachopoulos et al. 2018). embedment lengths and conventional monitoring
The fundamental concept is to create a zone of instrumentation thus could not account for potential micro-
reinforcement in the rock mass thereby improving mechanisms. The two scenarios previously discussed
confinement and utilizing its inherent mass strength to be were used to select the experimental temperature range for
self-supportive. The reinforcement system enables the this investigation.
transfer of load from the unstable regions near the
boundary of the excavation to the stable rock mass at 2.4 DOS Technology
depth (Thompson and Villaescusa 2014). FGRB are
classified as continuously mechanically coupled (CMC), The majority of past research into rock bolts has utilized
providing an extended shear surface for load transfer and conventional instrumentation to monitor pull-out test
development of higher anchorage capacities than setups. The common approach was to measure applied
mechanically anchored rock bolts (Signer 1990). Historical load and displacement at the borehole collar, however, this
research has found that the effects of adhesion are technique provided little insight into the stress distribution
negligible and that mechanical interlock is the predominant along the length of the bolt. Improvements were made
component for load transfer in FGRB (Yazici and Kaiser through the application of discrete monitoring techniques
1992; Aziz et al. 2006), which transfers load via contact (typically foil strain gauges) to the bolt (Farmer 1975; Li et
surfaces over its embedment length. The distribution of al. 2017b; Zhang et al. 2021) in order to better understand
axial stresses in the bolt and bolt-grout interface resemble the fundamental behaviour of rock bolts. Strain gauges are
an exponential decay curve (Serbousek and Signer 1987). installed on the surface of the bolt but they may interfere
Temperatures at the extremes can potentially affect the with bolt-grout interaction. Also, the strain profile between
bonding and performance of rock bolts. For cement based measuring points have to be interpolated so localized
grouts, the concerns in cold temperature environments are phenomena will be lost where a gauge is not present. The
slow strength development and early-age frost damage. typical spatial resolution was greater than 50 mm thus, the
Regular Portland Cement (PC) will not properly hydrate in resultant profile may not accurately represent the
sub-zero conditions thus will not likely gain sufficient behaviour (Cruz et al. 2016).
strength. Research into cement mixtures for permafrost New and innovative methods were explored in order to
have found different mixtures and additives that improve improve the spatial resolution and achieve a continuous
performance with the key characteristics being high early strain profile along the entire length of the bolt. One of the
strength and high hydration rate (Morris 1970; Reny et al. new methodologies utilizing DOS was developed by the
research program led by Dr. Vlachopoulos (Vlachopoulos length fails earlier than the rest of the system thus the DOS
et al. 2014; Vlachopoulos et al. 2018) in combination with was unable to capture the strain profile over the entire
an industrial partner. The method leverages the Rayleigh loading sequence. O’Connor (2020) mounted DOS to the
Optical Frequency Domain Reflectometry (ROFDR) to exterior of the confinement pipe in order to monitor radial
measure strain. When light is sent through an optical fiber, strains and observe changes past the serviceability of bolt
an index profile is created. Rayleigh backscattering results DOS. A 25 mm de-bonded zone for the DOS into the
from random variations in the profile with the scatter embedment length on the loaded end was introduced by
amplitude being a function of the distance along the fiber. Moore (2021). This addressed the premature failure and
External factors, such as strain and temperature, cause a also unsupported loading as the grout often sheared on the
spectral shift that can then be calibrated for use as a loaded end during failure. The concept of the de-bonded
distributed sensor (Gifford et al. 2005). This technique has zone has been further developed in this investigation.
resulted in an unprecedented spatial resolution of 0.65mm
for this research group. For temperature monitoring spatial 3.1 Instrumentation Conceptual Design
resolution of 0.1°C was found in Gifford et al. (2005). The
fiber is embedded into the rock bolt to not affect the bolt- In this investigation, the DOS technique will be used to
grout interface and capture geo-mechanisms associated monitor both strain and temperature of the grouted rock
with axial loading. The technique has been improved in bolt specimen. The prototype tested as proof of concept for
each testing iteration and implemented in laboratory and the new DOS configuration is shown in Figure 5. The strain
in-situ investigations into confining medium, embedment DOS (ε DOS) will be applied following the methodology
length, grouting materials, rib spacing, and grout annulus developed by the research group. An improvement in
(Cruz et al. 2016; Forbes et al. 2018; O’Connor et al. 2020; creating the de-bonded zone is the use of stainless steel
Moore and Vlachopoulos 2021). tubes, which the fiber will be fed through in order to prevent
adherence to the bolt. The stainless steel tubes will be
extended outside the embedment length as additional
3 METHODOLOGY protection for the sensor in the higher stress location.

In order to study the effects of temperature on FGRB, a

series of pull-out tests are in the process of being
conducted on specimens cured at specific temperatures for
different durations of time. The most recent preparation
and testing procedure is a summation of the iterative
development process from the research group. The
methodology has been modified in order to monitor strain
and temperature changes of the specimen during testing.
Lessons learned and recommendations from previous
group members have been explored during conceptual
design and proof of concept testing.
Rebar rock bolts with diametrically opposing machined
grooves are used by the research group. Moore (2021)
moved the location of the grooves to the longitudinal ribs
rather than at the transverse ribs (Cruz 2017; O’Connor
2020) to enhance conformance by decreasing impacts on
mechanical interlock, the new groove location has been
used in this investigation. The DOS is placed inside the
grooves (Figure 4) and embedded using proprietary metal
bonding adhesive. The DOS is comprised of the sensing
fiber that is spliced to an LC connector and a coreless
termination fiber. The positioning of the DOS allows for any Figure 5. Instrumentation conceptual design for prototype
bending due to axial loading to be averaged out.
This investigation is the first time this research group
will use the DOS technique in order to monitor temperature
so thoroughly. The pull-out tests will be conducted in a
laboratory at room temperature. Thus any changes from
the curing temperature of the specimen will need to be
monitored in order to identify the impact on behaviour. The
Figure 4. DOS configuration in diametrically opposing
temperature DOS (T DOS) will be a single sensing line that
grooves (from Cruz et al. 2016)
is fully de-bonded from the bolt, other than a single
One of the limitations of previous set ups was the anchorage point. The freedom of movement is to isolate
premature failure of the DOS. The fiber was bonded to the the sensor from bolt loading so that strains in the T DOS
bolt before the embedment length, which was problematic are resulting from temperature change. The entire length
as the sensor fails when the rebar yields. During the pull- of the fiber will be fed through a stainless steel tube. To
out test, the portion of the rebar outside of the embedment fully bond the ε DOS to the bolt and avoid air voids and
damage from the overlapping stainless steel tube of the T The selected temperature regime will consist of four
DOS, a layer of metal bonding adhesive will be laid over curing environments. The temperatures were selected
the ε DOS prior to embedment of the T DOS. The groove based on the background scenarios of ground condition
dimensions are also increased from previous setups of 2.5 extremes from permafrost to nuclear waste repository
mm x 2.5 mm (Forbes 2015; Cruz 2017; O’Connor 2020) (geothermally active). For the control and geothermally
and 3 mm x 3 mm (Moore 2021) to 3 mm x 4 mm. This is active conditions, regular PC grout will be used. For the
to accommodate the extra DOS and ensure the stainless permafrost conditions, Nordic Cable CT will be used as it
steel tube is fully embedded inside the groove and not is specifically formulated for underground cold weather
adversely affecting the bolt-grout interface. applications for anchoring cables and bolts. The simulated
conditions are found in Table 1. The specimens will be
cured in RMC laboratory freezers and conditioning rooms.
4 TESTING SCHEME
Table 1. Simulated operational environments for testing
In this study, the effects of temperature on FGRB
performance will be explored. As the steel bolt experiences Simulated Condition Curing Temp (°C) Grout Type
more extreme temperatures and cement grout cures at
Permafrost -20 Permafrost Grout
different temperatures, the physical and chemical reactions
will change thus, potentially affecting bonding and capacity. Permafrost -5 Permafrost Grout
Past research has found temperature affecting stress Control 20 PC
distribution and ultimate pull-out strength. There have been Geothermally Active 50 PC
very limited studies and the few focus on the higher
temperature range. The gaps in existing literature include Each curing environment will have six specimens. Each
lack of investigations over a complete temperature regime specimen will cure for different durations as detailed in
encompassing all operational environments, experiments Table 2. The curing times will examine short-term and long-
conducted on short embedment lengths which is not how term behaviours. This will add to the understanding of
FGRB are installed in-situ, testing does not include micro- capacity development in FGRB respective to time and
mechanisms as conventional instrumentation were used, environment. During preparation, tools and materials will
and correlations have mainly been through regression be conditioned to testing temperatures prior to grouting.
analysis. This study aims to address the identified For testing, extra care will be made to minimize changes in
limitations and help bridge these gaps in literature. specimen temperature. Insulation will be wrapped around
A series of pull-out tests will be conducted at the RMC the specimen prior to removal from curing rooms.
structures laboratory on 24 instrumented FGRB
specimens. The tests will be performed with a 322.41 Table 2. Testing scheme
Material Testing System (MTS) with a capacity of 500 kN
following the typical setup that has been used in this Serial Cure Number of Name of Specimens
research series as shown in Figure 6. The test specimen Time Specimens
will be using 20M 400W rebar at 1.6 m length and (Days)
embedment length of 1.3 m. Previous group members
1 3 4 P20-3, P5-3, C20-3, G50-3
utilized No. 6 Grade 60 (20M 420W) rebar, however, due
to local availability the bar was changed. The confining 2 7 4 P20-7, P5-7, C20-7, G50-7
medium will be schedule 80 steel 1¼” nominal pipe. The 3 28 4 P20-28, P5-28, C20-28, G50-28
size of the confining pipe will create a thin grout annulus so 4 35 4 P20-35, P5-35, C20-35, G50-35
that shear stress at the bolt-grout interface is 5 60 4 P20-60, P5-60, C20-60, G50-60
representative of the shear stress in the annulus itself 6 90 4 P20-90, P5-90, C20-90, G50-90
(Farmer 1975). 1each serial will include one specimen from each curing
temperature

Materials testing will be conducted to confirm material

properties with literature values. Two grout types are used
due to the broad temperature regime. A baseline
comparison between the grouts is being conducted for
compressive strength and split tensile strength with testing
cylinders cured at the same temperature. Properties of the
rebar will be confirmed with tensile test. A new single mode
polyimide coated fiber was proposed for this investigation
as the previous acrylate coated fiber utilized by the
research group had slippage between the layers resulting
in a 50 mm transition zone for strain pickup. A prototype
was tested as proof of concept of the proposed DOS
configuration and testing of the polyimide fiber.

Figure 6. Instrumentation location for monitoring program

5 RESULTS OF PROTOTYPE TESTING The data collection rate was set to 2 Hz and the MTS
loading rate was 1 mm/min. The prototype was loaded to
The prototype, seen in Figure 5, was tested using an 810 0.45Fy (54.5 kN) in order to remain within the elastic zone.
MTS at the RMC structures lab. The rebar was 600 mm in Figure 10 shows a snapshot of the strains captured by the
order to fit into the testing machine with four strain gauges T DOS during the different stages of loading in the sensing
attached to the surface. The new polyimide fiber was tested length. The average strain at start test was 0.0με, at mid
to determine the relationship between the strain and test 1.2με, and at end test 1.7με, which are well within the
temperature. The calibration set up, Figure 7, included a ±25με uncertainty of the data collector. The test confirmed
glycol bath and the fiber (TCal Sensor) built in the same that the embedment method using stainless steel tubing to
manner as the T DOS that will be embedded in the FGRB de-bond the fiber was successful as strain was not picked
specimen. The testing covered the range of curing up during prototype loading. After the loading was stopped,
temperatures at 5°C increments. The results provided in a heat gun was applied to the prototype and observations
Figure 8 have good linear correlation with an average slope were made in real-time of the strain changes in the DOS.
of 8.5με/°C. The test confirmed that the configuration isolated changes
in strain as a response to temperature. The prototype was
unloaded after the T DOS testing ended.

Figure 7. Temperature fiber calibration setup

Figure 10. Strain profile of T DOS during loading

5.2 Strain DOS testing

The data collection rate was set to 2 Hz and the MTS

loading rate was set to 1 mm/min and later increased to 2
mm/min. The initial objective was to load the rebar until
fracture in order to confirm the ultimate strength, however,
the max machine stroke was reached at 162.4 kN thus, the
test was ended. Figure 11 shows the strain profile along
the entire fiber (i.e. both sides of the rebar) at various loads.
The strains were fairly symmetrical about the loop
Figure 8. Relationship of strain and temperature indicating that axial loading during testing did not result in
any significant bending. The results found the transition
The prototype testing, Figure 9, happened in two section for strain pickup in the polyimide fiber as 10-20 mm,
serials. The first test was to confirm the isolation of the T which is less than the 50 mm of the previous acrylate fiber.
DOS from rebar loading and its response to temperature Results from Weisbrich et al. (2020) concluded that
change when embedded. The second test was the test of polyimide and ORMOCER® had immediate strain pickup
the ε DOS and material properties of the steel rebar. and did not require consideration of a transition section,
only the acrylate coating needed a 40 mm consideration.
This is most likely due to the stronger bond between the
layers as the polyimide coating cannot be mechanically
removed. However, this resulted in the disadvantage of
increased brittleness. The polyimide fiber had a minimum
bend radius that was larger than the looping groove
machined into the end of the rebar, which is intended to
protect the fiber during specimen preparation and
installation within the MTS. A larger loop was created that
was free hanging with heat shrink wrapped around the
Figure 9. Monitoring layout of prototype testing fiber, as seen in Figure 5. The brittleness and external loop
has caused reconsideration in using the polyimide fiber for
5.1 Temperature DOS testing strain sensing in this investigation as it would be a weak
point during handling and setup.
 Table 4. Modulus of Elasticity results

SG 1 SG 2 SG 3 SG 4 DOS
E (GPa) E (GPa) E (GPa) E (GPa) E (GPa)
200.00 160.00 170.83 200.00 200.00
Avg (SG 1/3) E = 185.42 Avg (SG 2/4) E = 180.00
1averages of strain gauges at same distance on opposite
sides of rebar

6 SUMMARY

The existing literature investigating the effects of

Figure 11. Strain profile of ε DOS at different loads along temperature on FGRB is limited. As underground projects
entire fiber length are becoming more complex and found in diverse ground
conditions, understanding performance changes due to
The rebar bolt yielded at 120 kN at which the ε DOS failed. these extreme environments will be vital for safe design of
The yield strength was found to be 460 MPa, which is within ground support. The authors of this paper aim to utilize the
literature range for 400W rebar and comparative to results state of the art technique noted herein to determine the
of 470-480 MPa from Moore (2021). The Modulus of behaviours and micro-mechanisms in the specific
Elasticity (E) was found to be 200.00 GPa using results of temperature regime outlined. Testing of a prototype for
the ε DOS, which is comparable to literature values for mild proof of concept has found the new DOS configuration will
steel. The measurements were verified by comparing the be able to monitor specimen temperature changes and
DOS measurements at the location of the strain gauges, monitor strain profile during testing. This research will
found in Figure 12 and Table 3. The percent difference of provide greater understanding into the complexities of axial
DOS strains were small when compared to SG 1 and SG loading of FGRB with respect to temperature variations.
4, with larger differences with SG 2 and SG 3. This could
be a result of misalignment as the comparison of the
Modulus of Elasticity found that SG 2 and SG 3 results 7 ACKOWNLEDGEMENTS
were less accurate, as seen in Table 4.
The authors would like to acknowledge support from the
Royal Military College of Canada, the RMC Green Team,
Sika Canada Inc., the Department of National Defense,
and the Natural Sciences and Engineering Research
Council.

8 REFERENCES

Aziz, N., Jalaifar, H., & Conclaves, J. 2006. Bolt Surface

Configurations and Load Transfer Mechanism. 2006
Coal Operators’ Conference, Wollongong, Australia.
Bobet, A. and Einstein, H.H. 2011. Tunnel Reinforcement
with Rock Bolts, Tunnelling and Underground Space
Technology, 26(1): 100-123.
Cruz, D.F. 2017. The Geomechanical Response of Axially
Loaded Fully Grouted Rock Bolts Utilizing Fibre Optics
Figure 12. Strain profile of ε DOS along sensing length Technology [Master’s Thesis, Royal Military College of
Canada]. Kingston: Department of Civil Engineering.
Cruz, D.F., Vlachopoulos, N., & Forbes, B.J. 2016. The
Table 3. Comparison of strain measurements of ε DOS Geo-Mechanical Response of Axially Loaded Rock
Bolts using Fiber Optic Technology. XV Colombian
Load % Difference from ε DOS Geotechnical Congress & II International Specialized
(kN) SG 1 SG 2 SG 3 SG 4 Conference of Soft Rocks, Cartagena, Colombia.
Castellani, V. and Dragoni, W. 1997. Ancient tunnels: from
5 10.20% 2.46% 6.85% 13.90%
Roman outlets back to the early Greek civilization. 12th
40 2.74% 16.49% 7.50% 1.32% International Congress of Speology, Switzerland, 3:
70 0.64% 15.57% 8.45% 1.41% 265-268.
90 3.99% 20.14% 13.46% 7.21% El Salam, M.E.A. 2002. Construction of underground works
110 4.23% 21.70% 13.05% 8.85% and tunnels in ancient Egypt, Tunnelling and
Underground Space Technology, 17(3): 295-304.
Farmer, I.W. 1975. Stress Distribution along a Resin https://open.canada.ca/data/en/dataset/d1e2048b-
Grouted Rock Anchor, International Journal of Rock ccff-5852-aaa5-b861bd55c367
Mechanics and Mining Sciences & Geomechanics Nuclear Waste Management Organization. 2023.
Abstracts, 12(11): 347-35. Canada’s Plan – Adaptive Phased Management.
Forbes, B.J. 2015. The Application of Distributed Optical https://www.nwmo.ca/en/Canadas-Plan/About-
Sensing for Monitoring Support in Underground Adaptive-Phased-Management-APM
Excavations [Master’s Thesis, Queen’s University]. O’Connor, T.P. 2020. The Geo-Mechanical Response of
Kingston: Department of Geological Sciences & Axially Loaded Fully Grouted Rock Bolts Utilizing
Geological Engineering. Distributed Fiber Optic Technology [Master’s Thesis,
Forbes, B., Vlachopoulos, N., & Hyett, A.J. 2018. The Royal Military College of Canada]. Kingston:
application of distributed optical strain sensing to Department of Civil Engineering.
measure the strain distribution of ground support O’Connor, T.P., Vlachopoulos, N., Forbes, B., &
members, Facets, 3(1). Diederichs, M. 2020. Grout Comparison for Fully
Gibb, F.G.F. 1999. High-temperature, very deep, Grouted Rock Bolts Using Distributed Fiber Optic
geological disposal: a safer alternative for high-level Sensors. 54th U.S. Rock Mechanics/Geomechanics
radioactive waste?, Waste Management, 19: 207-211. Symposium.
Government of Canada. 2022. Arctic and Northern Policy Reny, S., Cruz, J.P., & Clements, W. 2019. Developments
Framework. of New Cementitious Grout for Permafrost Conditions,
https://www.rcaanc- Ground Support.
cirnac.gc.ca/eng/1560523306861/1560523330587 Signer, S.P. 1990. Field Verification of Load Transfer
Gifford, D.K., Soller, B.J., Wolfe, M.S., & Froggatt, M.E. Mechanics of Fully Grouted Roof Bolts, US Bureau of
2005. Distributed Fiber-Optic Temperature Sensing Mines (1903).
using Rayleigh Backscatter. ECOC 2005. Serbousek, M.O., & Signer, S.P. 1987. Linear Load-
Hoek, E. and Wood, D.F. 1987. Support in Underground Transfer Mechanics of Fully Grouted Roof Bolts,
Hard Rock Mines, Underground Support Systems, 35: Bureau of Mines Report of Investigations 9135. United
1-6. States Department of the Interior.
Huang, G. 2021. Performance of Calcium Sulfoaluminate Thompson, A.G. & Villaescusa, E. 2014. Case Studies of
Cement-based Mixtures Used for Permafrost Regions Rock Reinforcement Components and Systems
[PhD Thesis, University of Alberta]. Edmonton: Testing, Rock Mechanics and Rock Engineering, 47:
Department of Civil and Environmental Engineering. 1589-1602.
Jenness, J.L. 1949. Permafrost in Canada, Arctic, 2(1): 13- Vlachopouos, N., Cruz, D., and Forbes, B. 2018. Utilizing
27. a novel fiber optic technology to capture the axial
King, F., Hall, S.D. & Keech, P.G. 2017. Nature of the near- responses of fully grouted rock bolts, Journal of Rock
field environment in a deep geological repository and Mechanics and Geotechnical Engineering, 10(2): 222-
the implications for the corrosion behaviour of the 235.
container, The International Journal of Corrosion Vlachopoulos, N., Diederichs, M.S., & Marinos, P. 2013.
Processes and Corrosion Control, 52: 25-30. Tunnel behaviour associated with weak Alpine rock
Li, F., Jin, H., Hu, D., Wang, B., & Jia, Y. 2017a. Influence masses of the Driskos Twin Tunnel system, Egnatia
of Temperature and Roughness of Surrounding Rocks Odos Highway, Canadian Geotechnical Journal, 50(1):
on Mechanical Behavior of Rock Bolts, Soil Dynamics 91-120.
and Earthquake Engineering, 103: 55-63. Vlachopoulos, N., Forbes, B., Oke, J., & Hyett, A. 2014.
Li, F., Quan, X., Jia, Y., Wang, B., Zhang, G., & Chen S. The Development of a Technique in order to Measure
2017b. The Experimental Study of the Temperature the Distributed Optical Sensing of a Forepole
Effect on the Interfacial Properties of Fully Grouted temporary support element employed within an
Rock Bolt, Applied Science, 7: 327. Umbrella Arch System. EUROCK, International
Moore, K.S. 2021. The Effects of Rib Spacing and Grout Symposium, Vigo, Spain.
Annulus on the Geomechanical Response of Axially Wang, B., Guo, X., Li, F., Hu, D., & Jia, Y. 2018.
Loaded Fully Grouted Rock Bolts Utilizing Distributed Mechanical Behavior of Rock Bolts under a High
Fiber Optic Technology [Master’s Thesis, Royal Military Temperature Environment, International Journal of
College of Canada]. Kingston: Department of Civil Rock Mechanics and Mining Sciences, 104: 124-130
Engineering. Xue, Y., Sun, D., & Wang, L. 2021. An analytical solution
Moore, K.S. & Vlachopoulos, N. 2021. Utilizing a Fiber of temperature in a nuclear waste repository, Annals of
Optic Methodology to Capture the Axial Response of Nuclear Energy, 163.
Rib Spacing Effects on Grouted Rock Bolts. Yazici, S. & Kaiser, P.K. 1992. Bond Strength of Grouted
GeoNiagara 2021, Niagara Falls, Ontario, Canada. Cables, International Journal of Rock Mechanics and
Morris, E.F. 1970. Evaluation of Cement Systems for Mining Sciences & Geomechanics Abstracts, 29(3):
Permafrost. 99th Annual Meeting of American Institute 279-292.
of Mining, Metallurgical, and Petroleum Engineers, Zhang, W., Wang, L., Song, Y., & Tan, Y. 2021.
Denver, Colorado, USA. Experimental Study on the Mechanical Properties of
Natural Resources Canada. 2016. Permafrost, Atlas of Grouted Rock Bolts Subjected to Sulfate Attack and
Canada, 5th Edition. Freeze-thaw Cycling, Construction and Building
Materials, 291.
Title – Underground Instrumentation
Program – Interpretation and Benefits
Douglas Milne
University of Saskatchewan, Saskatoon, Saskatchewan, Canada

ABSTRACT
The interpretation of field instrumentation can be challenging, however, a great deal of information and valuable insight
into rock mass behaviour can be gained. This paper summarizes an instrumentation program from a New Brunswick
underground mine conducted 3 decades ago. The instrumentation installed included microseismic strain and deformation
monitoring, as well as visual inspections.
A major benefit of an instrumentation program, and its’ interpretation, is the opportunity it provides to geotechnical and
rock mechanics engineers to better understand the complex response of rock masses to engineering structures. None of
the insights highlighted in this paper relied upon computer modelling, but instead were based on the application of relatively
basic rock mechanics principals, field observations and discussions with colleagues. Field programs such as these are key
to advancing rock mechanics for individuals and for the profession.

RÉSUMÉ
L'interprétation d'instrumentation de terrain peut être difficile, cependant, une grande quantité d'informations peut être
recueillie, ainsi qu'un degré de compréhension gratifiant du comportement des masses rocheuses. Cet article résume un
programme d'instrumentation utilisé dans une mine souterraine du Nouveau-Brunswick réalisé il y a 3 décennies.
L'instrumentation installée comprenait une surveillance microsismique, des déformations et ainsi que des inspections
visuelles.
Un avantage majeur du programme d'instrumentation et de son interprétation est l'opportunité qu’elle offre aux
ingénieurs en géotechnique et en mécanique des roches de mieux comprendre la réponse complexe des masses
rocheuses aux structures d'ingénierie. Aucune des idées mises en évidence dans cet article ne reposait sur la modélisation
informatique. Ils sont plutôt fondé sur l'application des principes relativement élémentaires de la mécanique des roches,
les 'observations sur le terrain et les discussions avec des collègues. Les programmes d'instrumentation de terrain comme
ceux-ci sont essentiels pour l’avancement de la mécanique des roches pour les individus et pour la profession.

1 INTRODUCTION closure stations and the Canmet strain cells. The

interpretation of instrumentation was based on basic rock
This paper summarizes the results of an instrumentation mechanics principles, underground observations, as well
program conducted 3 decades ago at the Brunswick Mine as observed rock mass properties.
near Bathurst New Brunswick. The purpose of the study
was to better understand the microseismic activity at the 2 BASIC MINING METHOD AND GEOLOGY
mine and focused on estimating induced stresses during
the mining of a stressed pillar. The research was At the time of this study, the Brunswick Mine was producing
sponsored by Brunswick Mining, the Noranda Technology about 10,000 tonnes of ore per day. It was in operation for
Centre and the Mining Research Directorate. Papers have nearly 50 years producing lead, zinc, copper and silver ore
been written on this project, but they have mainly focused until it closed in 2013. The instrumented sill pillar area was
on the microseismic activity and the computer modelling between the 575 and 725 levels, at an average depth
conducted (Grant, D., 1993; Hudyma et al., 1994; Hudyma below surface of 600 metres and the instrumentation
et al., 1994). program was conducted from the summer of 1992 to the
The instrumentation program was conducted to gain spring of 1993. Open stope mining was used in the study
information on rock bursting in a highly stressed sill pillar, area.
at a depth of approximately 600 metres. The The ore body consisted of massive sulphide located
instrumentation installed included: between weaker layers of chlorite schist on both the
• Microseismic monitoring hanging wall and footwall. The ore zone was 5 to 10 metres
• 10 Canmet strain cells thick in the study area with a dip between 70 to 75 degrees
• Ground Movement Monitors (GMM’s) to the west. Select rock and rock mass properties are given
• 11 Extensometers in Table 1.
• Borehole camera holes
• Drift closure stations 2.1 Mining in the Instrumented Stoping area.
Mining in this study area was conducted in the 725 sill pillar
The instrumentation was augmented with frequent (Figure 1) with blast holes initially drilled down from the 6-
underground inspections. This paper looks in detail at sub level level and then with up and down holes, as mining
some of the results from field observations, extensometers, progressed to the south (Figure 2).
Figure 2. Stope instrumentation for the 725 sill pillar, highlighted in yellow, looking east (After Hudyma et al.,
1994).

Figure 3 Stope isometric showing the stope blasting sequence and stope dimension (After, Milne and Snell,
2018).
Table 1. Select Rock and Rock Mass Properties observed on the steel installing wedges, as well as some
corrosion on the cell body. After consulting with the UBC
Property Massive Chloritic Dyke Metallurgical Engineering Department, it was found that the
Sulphide Schist movement measured by the strain cell could easily
correspond with the observed corrosion. Stress change
Young’s 155 GPa 63 GPa
estimates were still obtained from the instruments by
Modulus looking at the readings immediately before and after each
UCS 220 MPa blast. The supplier changed the grade of steel used in the
Q’ 33 4 10 manufacture of the cells.
Classification There were a few goals for the strain cells:
RQD 100 90 90 1. To help determine if the rock mass was yielding or
behaving elastically
Jn 9 9 9
2. To determine the zone of influence and likely zone
Jr 2 1.5 2 seismic response
Ja 1 4 2 3. To estimate the change in stress due to blasting
and
4. To help calibrate computer models.
As shown in Figure 2, the stopes were dipping 70 o to 75o
to the west. The stope blasted below the 6-sub had an Some of these goals require the assistance of computer
average up dip length of 32 metres and blasting above the modelling, but some goals can be met by applying some
6-sub averaged an up dip length of about 14 metres. basic rock mechanics theory.
Mining progressed to the south, towards the centre of the There is a simple equation for estimating stress around
sill pillar area. a spherical opening in a hydrostatic stress field (eq. 1).
The stope raise was taken in 7 blasts and blast 8 to 19 𝑎 3
𝜎𝜃𝜃 = 𝑃(1 + 0.5 ( ) ) (Equation 1)
corresponded to stoping in the instrumented area. No 𝑟
backfill was used in this area. After blast 19, on November Where: P = hydrostatic field stress (or stress change)
9th, 1992, a significant hanging wall and abutment failure a = radius of the spherical cavity
occurred, and a pillar was left before mining resumed. Most r = the distance from the centre of the cavity
of the instrumentation was lost after blast 19, however 𝜎𝜃𝜃 = induced tangential stress
significant information was still gathered.
To further simplify the interpretation of the instruments,
it’s assumed that the equation for the change in tangential
3 INSTALLED INSTRUMENTATION strain would follow the same format and would vary simply
as a function of opening radius and the distance from the
The majority of the instruments were installed from the 575 centre of the opening all cubed (eq. 2).
𝑎 3
Level and the 6-sub (Figures 3 and 4). As shown, 𝜀𝜃𝜃 ~k ( ) (Equation 2)
𝑟
instrumentation consisted of 10 Canmet Strain Cells (soft Where: εθθ = tangential strain
inclusion cells), 11 Extensometers (MPBX), ground a = radius of the spherical cavity
movement monitors (GMM’s), borehole camera holes and r = the distance from the centre of the cavity
closure stations. Instruments were installed prior to local k = a constant calibrated with field data
mining activity to obtain initial background readings.
Prior to the start of mining in the instrumentation zone, Figure 5 shows a long section with the strain cell locations
but after creation of the sill pillar, borehole slotter tests were and readings following blast 12. Blast 12 had the following
conducted to estimate stresses in the area on the 575 dimensions: 8m wide, 10 metres long and 25 metres up
Level (Figure 3). East/west stresses of 35 MPa approx- dip. For purposes of assessing strain readings due to the
imately normal to the orebody, north/south stresses of 26 blast 12, the stope was represented by a spherical cavity
MPa along the orebody strike and ~ 8 MPa near vertical with a diameter of 8 metres. Most strain cells read
stress were estimated from the borehole slotter tests. tangential strain, or close to it. Using equation 2 and
calibrating to readings from two of the strain cells located
3.1 Canmet Strain Cells close together, a series of concentric circles is shown
representing approximated microstrain contours induced
Canmet strain cells were developed by Canmet and are by blast 12. Based on this interpretation, an average strain
unfortunately not currently available. The cells are of less than 5 µε could be expected at distance exceeding
designed to deform with the borehole without appreciably 60 metres from the centre of blast 12. Data suggests the
resisting any borehole squeezing. The cells were installed rock is not yielding where the strain cells are installed. It
with steel wedges, in a fashion similar to the Geokon stiff also seems unlikely that microseismic activity due to a
inclusion stress cells. change in compressive stresses would occur at a greater
After installation, a gradual increase in stresses were distance than this. An estimate of stress change could be
observed prior to any local mining activity. It was suggested obtained if the Kirsch (1898) equation was applied to take
that creep behaviour was occurring in the massive into account the influence of the borehole the stress cells
sulphide. After removing one of the cells that had exceeded are installed in. This is not presented here in the interest of
it’s measurement range, significant corrosion was space (and time). A very rough estimate of initial stresses
 Figure 4 575 Level showing instrumentation locations (After Hudyma et al., 1994)

Figure 4. 6-Sub showing instrument locations (From Hudyma et al., 1994)

Figure 5. Long section showing strain measured after blast 12. Strain contours are based on tangential stresses around a
spherical opening (After Milne., 1993)

could also be obtained by applying equation 1 to the 3.3 Closure Stations

available data.
Closure stations were installed in the 575 Sub-level. They
3.2 Extensometers (MPBX) were read several times between each stope blast. They
were read with a manual tape extensometer with a
Extensometers were installed to measure deformation resolution down to 0.01 mm. Prior to a significant stope
and strain in both the hanging wall and footwall of the hanging wall and abutment failure after blast 19, taken on
stoping area. They were designed to measure increasing November 9th, 1992, significant stress increases between
strain in advance of the mining front and relaxation and 10 and 20 MPa were estimated from closure deformations
deformation adjacent to the open stope. between 0.5 and 1.5 mm and an estimated rock mass
After the raise had been created and stope blasting was modulus of approximately 50 GPa. A simple program call
started, alternating zones of compression and tension were Closure was developed to estimate stress change from
measured in the ore zone and in the immediate hanging closure measurements. The program was developed
wall and footwall. Tensile and compressive deformations based on strain gauge rosette theory and the Kirsch
were recorded with less than 0.1 mm of movement equation presented in Goodman, 1989. This program
between anchors and it was initially believed to be forms the basis of a lab in the 4th year Geological
instrument error. Open fractures were soon reported in the Engineering class at the University of Saskatchewan.
stope abutment within the area of expected high After blast 19, significant hanging wall and abutment
compressive stresses (Figure 6). These open fractures failure occurred, and the strain recorded was well in excess
were found to line up with the recorded zones of tensile of a possible elastic response. The direction of maximum
deformation (Figure 7). compressive deformation became parallel to the estimated
There were small scale undulations along the hanging failure extent.
wall foliation. The observed zones of tensile deformation
were interpreted as shear deformation initiated by stresses
flowing around the open stope and interacting to bends in 4 CONCLUSIONS
the ore – hanging wall contact (Figure 8). This mechanism
was believed to be key to deteriorating ground conditions A great deal of information was obtained from the
and subsequent hanging wall instability. It is worth noting instrumentation program, however, in some cases it was
that the small stope opening was inducing shearing in the not the information that was expected.
hanging wall and in the competent massive sulphide at Canmet strain cells were used to help calibrate a
distances exceeding 30 metres. It was felt that this numerical model. The cells gave the zone of stress
shearing mechanism could have contributed to seismic influence around the stope blasts. This zone of influence
activity occurring much further from the mining zone then dropped off away from the stoping blocks as a function of
the zone influenced by increasing compressive stresses. 1 over the distance away cubed. They also highlighted
 Figure 6. Open fracture observed in the back of the stope abutment in the ore horizon.

Figure 7. Plan view of 6-sub showing open fractures and tensile zones measured with the hanging wall extensometers.
(From Hudyma et al., 1994.)
 augmented the rock mechanics understanding of the
practicing engineers involved in the study. Much of this
understanding may have been hindered by the early
application of computer modelling.

4.1 References

Grant, D., 1993. BEAP modelling results of the Brunswick

Mines 725 sill pillar mining of the 13N and 16N stopes.
Internal Noranda Technology Centre report.

Hudyma, M., Grant, D., Potvin, Y., Brummer, R., & Milne,
D. (1994). Geomechanics of sill pillar mining in rock
burst prone conditions. Phase 1 final report to the
Mining Research Directorate.

Hudyma, M., Potvin, Y., Milne, D., Brummer, R., Grant, D.,
& Board, M. (1994). Geomechanics of sill pillar mining.
in A. A. Balkema (Ed.), Proceedings of the 1st North
American Rock Mechanics Symposium, Austin, Texas,
969–976.
Figure 8. Stress induced shearing around open stopes,
(From Milne, 1993). Milne, D. (1993). Interpretation of the Brunswick mine
instrumentation program. Unpublished internal report,
Noranda Technology Centre.
areas of elastic strain and provided some of the data that
numerical modelling would have given. Milne, D. (1997). Underground design and deformation
The extensometers were installed to provide further based on surface geometry. (Doctoral dissertation) The
data, and better coverage, on zones of compressive and University of British Columbia.
tensile stress. The extensometers instead provided
information on a shearing mechanism that was driving Milne, D and Snell, G (2018). New failure prediction
deformation at significant distances from the stope blast. method for underground excavations. CIM Journal, Vol.
They also showed a mode of both hanging wall and 9, No. 1, 2018. https://doi.org/10.15834/cimj.2018.4
abutment failure.
The closure stations also provided an estimate of stress Goodman, R.E., 1989. Introduction to Rock Mechanics,
change as well as an indication of the transition of elastic John Wiley and Sons, Toronto, 562 pp..
to post peak deformation. The interpretation of data
provided by this instrumentation provided both insight into
the rock mass behaviour at the Brunswick Mine, but also
 Tuesday, October 3, 2023

ADVANCED TESTING I
Laboratory investigation on the effect of
roughness anisotropy on shear strength of
rock fractures
William R. Hoyle, Nicholas R. MacDonald, Jennifer J. Day,
Mark S. Diederichs
Department of Geological Sciences and Geological Engineering –
Queen’s University, Kingston, ON, Canada

ABSTRACT
In laboratory direct shear testing of rock fractures, it is current practice to shear an ellipsoidal specimen in the direction of
its longest axis, as this allows for the greatest amount of data collection and is assumed to be the direction of shear
displacement for unoriented boreholes with inclined, planar joints. However, it is unclear where the shear resistance at this
orientation ranks when compared to alternative shear directions, or whether it is the realistic direction of shear encountered
in a rockmass. This paper uses 3-Dimensional roughness analysis based on point clouds of fracture surfaces in rock core
created using Structure from Motion photogrammetry to calculate the range of potential roughness values. This is then
correlated to laboratory testing data on grout replicas of rock fracture surfaces. It is shown that roughness anisotropy has
different degrees of impact on shear strength depending on the normal confinement stress and the intact rock strength.

RÉSUMÉ
La pratique courante des essais de cisaillement de fractures rocheuses en laboratoire consiste à cisailler un échantillon
ellipsoïdal dans la direction de son axe le plus long. Celle-ci est supposé être la direction du déplacement par cisaillement
pour les trous de forage non orientés avec des joints inclinés et planaires et permet de recueillir le plus grand nombre de
données. Cependant, on ne sait pas exactement où se situe la résistance au cisaillement dans cette orientation par rapport
à d'autres directions de cisaillement, ni s'il s'agit de la direction réaliste de cisaillement rencontrée dans une masse
rocheuse. Cet article utilise une analyse de la rugosité en 3D basée sur des nuages de points de surfaces de fissures
dans des carottes de roche créées à l'aide de photogrammétrie pour calculer la gamme des valeurs de rugosité
potentielles. Ces valeurs sont ensuite mises en corrélation avec les données d'essais en laboratoire sur des répliques de
fractures rocheuses en mortier de ciment. Il est démontré que l'anisotropie de la rugosité a différents degrés d'influences
selon la contrainte de confinement normale et la résistance de la roche.

1 INTRODUCTION joint in meta-sediment, selected because of its low

roughness and visual roughness anisotropy. The second,
The shear strength of rock joints and other fractures is an JA2 is in a mudstone, and is characterized by medium
important parameter for numerically modelling deformation roughness and visual anisotropy. The final joint JI1, is
behaviour in underground excavations or slopes in jointed composed of pink granite and was chosen as it represents
rockmasses. These large-scale design projects frequently low-medium roughness and visually isotropic
use laboratory testing to predict geomechanical characteristics.
parameters from drill core. Often this drill core is not To isolate the effect of the joint roughness from the
oriented so the strikes (or dip directions) of discontinuities strength and basic friction angle of the material and the
that intersect the drill core are unknown. Current guidelines normal confinement on the rock joint, replica samples of
for direct shear testing from the International Society for each joint were created using and 3D printed molds of each
Rock Mechanics and Rock Engineering (ISRM), note that joint following the procedures developed by MacDonald
a shear orientation can be chosen based on the estimated (2022).
direction of displacement (Muralha et al. 2014); however,
this is not always known. 2 REPLICA MODEL CREATION
This paper aims to demonstrate the importance of
shear direction on rock core samples in laboratory direct The rock joint specimens were placed on a turntable with
shear testing and provide a method to obtain a range of ground control points and photographed at 15° horizontal
estimates for shear strength when in situ shear direction is increments using a Nikon D7200 DSLR camera.
unknown. Shear direction is especially important to Metashape was then used to generate scaled point clouds
consider in joints that exhibit roughness anisotropy of each specimen using SfM photogrammetry (Agisoft,
(meaning the roughness is different in different directions). 2023). These were meshed, imported into Blender
This is distinct from roughness heterogeneity which refers (Blender Foundation, 2023), rotated upside down, and
to joints having different roughness along the same profile fixed to a blank mold created in CAD (Siemens, 2023). The
or in different regions of the same joint. resulting mold was sliced in PreForm (Formlabs, 2023) and
Structure from Motion (SfM) photogrammetry was used 3D-printed in a Form3 printer using stereolithography
in this study to create point clouds of three unique rock (SLA) printing for high precision (layer thickness = 25 μm).
joints as shown in Figure 1. The first joint, JA1, is a striated
 Sika 300 PT was used as grout material to cast replica predict which direction the shear strength will be lower in
specimens as it has a high dry strength. This was mixed at by assessing asperity angles. θ*max is the maximum
a cement to water ratio (C:W) of 4:1 and cured for 28 days. apparent dip angle for a given profile and C is a fitting
It should be noted that mineral oil was required to prevent parameter.
damage to the replica specimens upon removal of the mold
and this may affect the basic friction angle of the material. JA1
Chryssanthakis and Barton (1990) found that using silicone
oil-based releasing agents can cause the friction angle to
change by up to 10%. Further work is planned to
investigate and minimize this effect.

3 ROUGHNESS QUANTIFICATION

The resistance to sliding, or, shear strength (τ) of an object

sliding along a plane is considered to be a function of the
cohesion (or apparent cohesion) between the object and
the plane (c), the normal confining stress, σn, and the
friction angle (ɸ) (Coulomb, 1776), as follows:

𝜏 = 𝑐 + 𝜎𝑛 tan⁡(𝜙) [1]

For rough surfaces, such as those found in natural rock JA2

fractures, Patton (1966) proposed that the friction angle is
a combination of the average inclination of asperities along
the plane (i) and the basic friction angle of the material (ɸb).
This gives:

𝜏 = 𝑐 + 𝜎𝑛 tan⁡(𝑖 + 𝜙𝑏 ) [2]

This model is limited as it assumes the asperities do not

degrade during shearing; however, Barton (1971) showed
that the shear strength of the rock surface degrades as the
asperities are worn and introduced the joint compressive
strength (JCS) of the asperities. Barton (1973) made
improvements to incorporate roughness of the
discontinuity through the joint roughness coefficient (JRC).
This parameter is a scale from 0-20 that describes the
roughness of a joint surface. Later, Barton & Choubey
(1977) also found that the basic friction angle may not be
accurate depending on the condition of the discontinuity,
JI1
which was later confirmed by Grasselli (2003) and
MacDonald (2022). This leads to the equation:

𝐽𝐶𝑆
𝜏 = 𝜎𝑛 tan [𝐽𝑅𝐶𝑙𝑜𝑔10 ( ) − 𝜙𝑟 ] [3]
𝜎𝑛

where ɸr is the residual friction angle of the discontinuity.

A limitation of JRC is in order to forward predict the
shear strength of a rock joint, a user must compare several
profiles along a fracture surface to a series of 10
characteristic profiles from Barton & Choubey (1977). This
can cause a significant degree of error if the user is
inexperienced (Beer et al., 2002). Another limitation of JRC
is that it is a 2-dimensional (2D) parameter and does not
consider the direction of shear along a given profile (i.e.,
forwards, or backwards). This means it is unable to
incorporate roughness anisotropy in its original state. To
incorporate anisotropy into this model, JRC can be taken
at many orientations around the joint to attain a pseudo 3D- 1 cm
roughness value. Additionally, to incorporate shear Figure 1. Photographs of the surfaces JA1 (striated
direction in each 2D profile, the directional roughness metasediment), JA2 (anisotropic mudstone), and JI1
metric (θ*max/C+1) (Grasselli et al. 2002) can be used to (isotropic granite).
 A MATLAB (MathWorks INC, 2022) code was used to 𝐽𝑅𝐶 = 95𝑃𝑅𝐼10 [5]
calculate the roughness of the fracture surfaces which
assesses the JRC and mean apparent asperity angle, This process is then repeated across all sections of the
(θ*avg.) of the joint. This code was first developed by sample with a length greater than ⅓rd of the maximum
Packulak et al. (2019) and Bonneau (2022) and modified length of the sample and a sampling interval (SI) of 1% of
for this study. To account for anisotropy and heterogeneity the maximum sample length. The JRC measurements of
across the core sample, this code assesses JRC and θ*avg. these profiles are exported as a gaussian distribution. This
in 10° increments from 0°-350° and analyzes 100 profiles accounts for roughness heterogeneity across the sample.
in each direction. For each of these profiles, a After this is completed, the sample is rotated 10° and the
representative combination of sine waves is fit to the data. process is repeated until all 36 directions have been
A fast Fourier transform (FFT) is then completed on each assessed. The full data is then exported as a polar
of these profiles, which is used to calculate PRI10 according histogram which shows the angular distribution of JRC
to (Pickering, 2015) as follows: values (Figure 2). A summary of the maximum and
minimum JRC values for each sample is provided in
𝑃𝑅𝐼10 = ∑10
𝑗 𝐴∙𝑓 [4] Table 1. This data can then be used to predict the shear
strength of a fracture surface for any shearing direction.
Where A is the amplitude of a waveform that fits the
roughness profile and f is the respective frequency. This is
then used to calculate JRC using Equation 5 (Pickering,
2015).

JA1 JA2

JI1 Bottom JI1 Top

Figure 2. Polar histograms of JRC for an anisotropic striated joint surface (JA1), an anisotropic mudstone joint surface
(JA2), and two isotropic granite surfaces (JA1 Top and JA1 Bottom).
Table 1. Summary of roughness ranges. =

JRC ΔJRC
Sample ID ΔJRC
max. min. (%)
JA1 3.26 2.04 -1.22 -37.4
JA2 7.94 5.78 -2.16 -27.2
JI1 Top 4.08 2.96 -1.12 -27.5
JI1 Bottom 4.02 2.94 -1.08 -26.9

4 SHEAR STRENGTH ANISOTROPY

There have been many studies investigating 3D roughness

analysis or even specifically rock fracture roughness
anisotropy (Riss et al., 1997; Gentier et al., 2000; Grasselli
et al., 2002; Tatone and Grasselli, 2009); however, these
were mainly focused on asperity damage evolution with =
shearing, and few have investigated the effect of
roughness anisotropy on shear strength anisotropy.
Grasselli and Egger (2003) briefly note shear strength
anisotropy, which attempts to create a relationship
between the 3D roughness analysis method by Grasselli et
al. (2002) and shear strength. While Grasselli et al. (2002)
give a detailed analysis of roughness and its effect on
shear strength, it requires the measurement of many
parameters of the rock and a fitting parameter (C), which
makes it a more complex version of JRC until C can be
found to have a measurable value. Additionally, Grasselli
et al. (2002) did not investigate how changing JCS:σn
affects the relative drop in shear strength.
Using the data provided by the MATLAB code
mentioned in Section 3, Microsoft Excel was used to
predict the shear strength around each core sample at
varying confinements using Equation 3 (per Barton, 1973).
The JRC used for the calculation was taken as the
harmonic mean of all JRC values from the 100 profiles in
each direction. From this, the maximum estimated shear =
strength (τmax), the minimum estimated shear strength
(τmin), and their respective azimuth angles can be found.
Interestingly, the shear strength is not simply a
reflection of the roughness in every direction but also the
strength of the rock and the normal confinement. This is of
course a predictable relationship based on Equation 3, but
its effect can easily be overlooked. Figure 3 shows the
estimated shear strengths on the same rough surface
under different levels of confinement using Equation 3. It is
evident that shear strength anisotropy has a negative
correlation with normal stress. Furthermore, shear strength
anisotropy is more prevalent in stronger rocks and lower
confinements.
In extreme cases of very low JCS:σn ratios, (such as a
mudstone situated deep underground) a fracture surface
with a large degree of anisotropy in surface roughness may
not have a large degree of anisotropy in shear strength.
Conversely, a fracture in a hard rock with low confinement
may exhibit anisotropic behaviour even if it has isotropic
roughness. Figure 3. Polar plots of shear strength for an anisotropic
mudstone fracture (sample JA2) at various confinements
ranging from 0.0001*JCS to approaching JCS (45 MPa).
 To compare the effect of the roughness anisotropy of
different fracture surfaces, the authors propose evaluation
of the relative reduction in shear strength from the direction
of maximum sliding resistance to the direction of minimum
sliding resistance, per Equation 6. This approach
demonstrates the potential degree to which current
laboratory tests can overpredict shear strength on rock
joints if the shearing direction is nearer the direction of
maximum sliding resistance.
𝜏𝑚𝑎𝑥 −𝜏𝑚𝑖𝑛
𝜏∆ ⁡(%) = × 100⁡% [6]
𝜏𝑚𝑎𝑥

Where τΔ is the relative reduction in in shear strength from

the direction of maximum shear resistance to the direction
of minimum shear resistance.
As shown in Figure 4, the degree to which shear
strength can drop for a given strength of rock and normal
confinement behaves logarithmically. This demonstrates
that a sample with visible roughness anisotropy (JA2) can
decrease by up to 18% for a JCS:σn ratio of 400 based on
the chosen shearing direction. This ratio is representative
of a granite (JCS = 200 MPa) under low normal Figure 5. Predicted relative drop in shear strength vs the
confinement (σn = 0.5 MPa). common logarithm of the joint compressive strength to
normal confinement ratio.

Notably, sample JA1 (striated metasediment) exhibits

only slightly more anisotropic behaviour than the isotropic
granite sample (JI1) with respect to shear strength despite
having a larger change in JRC (15%). This is because
larger JRC values more greatly affect the shear strength.
In other words, rougher samples will likely have more shear
strength anisotropy than smoother ones.

5 LABORATORY TESTING

Laboratory direct shear testing was conducted using a

servo-controlled rock direct shear system (RDS-200) by
GCTS in the Queen’s University Advanced Geomechanics
Testing Laboratory. Test results from 8 replica samples of
discontinuity JA2 are presented in this paper.
Tests were performed under constant normal stress
boundary conditions at two different normal confinements
(σn = 1 MPa and 3 MPa) and four different shear directions
Figure 4. Predicted relative drop in shear strength vs the (α = 0°, 330°, 300°, and 270°). The normal confinements
joint compressive strength to normal confinement ratio. were chosen to generate a Mohr-Coulomb strength
envelope and the shearing directions were chosen as this
range includes the predicted directions of maximum (270°)
The same relationship is shown in Figure 5 on a and minimum (330°) shear strengths of the sample. The
common logarithmic scale with the intercept of each results (shown in Figure 6) indicate that the sample
trendline set to zero, because a material with no displays strength anisotropy as predicted by calculations
compressive strength will have no shear strength per following the method by Pickering (2015) applied to SfM
Equation 3. The relationships provided from Figure 5 are photogrammetry models of the samples. The estimated
simpler than those in Figure 4, which provides a simple directions of maximum and minimum shear strength were
method of quantifying shear strength anisotropy of a rough adequately predicted (especially at larger confinements),
surface. Like other fitting parameters in fracture surface but further research is underway to improve the accuracy
roughness analysis, more research needs to be conducted of shear strength prediction based on roughness and the
to investigate what measurable values control this range of shear strengths in anisotropic specimens.
coefficient.
Figure 6. Polar plot of sample JA2 (striated metasediment) at various normal confinements showing predicted shear
strengths (per PRI method by Pickering (2015) applied to SfM photogrammetry models of the samples) compared to
measured shear strengths from direct shear testing on synthetic replica grout specimens in different shearing directions.

Possible causes for precision errors in the SfM 7. In these test specimens, the largest shearing resistance
photogrammetry to PRI method of predicting shear is at 270° and the lowest is at 330°, which agrees with the
strength could be: predictive calculations based on roughness. This indicates
that using SfM photogrammetry to aid in choosing a
1. The accuracy of the reconstructed 3D meshed shearing direction may be an effective method although
SfM photogrammetry model of the fracture surface more tests are planned to confirm this and increase the
compared to the original fracture surface, accuracy of this method. Figure 7 also shows that the
2. The PRI method (Pickering, 2015) used to calculated cohesion (at zero applied normal stress) will be
calculate JRC was only calibrated on the 10 different depending on the shearing direction. This is
characteristic profiles by Barton & Choubey because the “cohesion” being calculated in this open
(1977), fracture is the component of friction acting in the direction
3. The method of using harmonic means of 100 perpendicular to the mean fracture surface. It is predicted
profiles JRC across the joint surface in each that this apparent friction angle will be larger for rougher
direction to calculate shear strength, and samples as rougher samples have a higher mean asperity
4. The range of moisture conditions the grout angle and therefore a greater component of friction in the
replicas were cured in (RH = 35%-55%) affecting direction perpendicular to the mean plane of the fracture
their compressive strength. surface.
From Figure 7, the peak shear strength in each
Shear strength envelopes sorted by shear direction shearing direction were calculated and summarized in
from the direct shear tests in this study are shown in Figure Table 2. These show that there is a difference of 7° and
0.54 MPa in peak friction angle and cohesion, respectively, resistance (270°) and the predicted direction of minimum
between the predicted direction of maximum shearing shearing resistance (330°).

Shear Strength, τ (MPa)

Shear Strength, τ (MPa)

Figure 7. Barton-Bandis and Mohr-Coulomb peak shear strength envelopes of sample JA2 in different shearing directions.

shear strength anisotropy of the rock and may make

Table 2. Summary of Mohr-Coulomb shear strength anisotropically rough surfaces behave more isotropically
properties for various shearing directions. as normal stresses approach the UCS of the rock.
This work shows that 3D and pseudo 3D roughness
Shear Peak friction Cohesion, c quantification methods are important tools for correctly
direction, α (°) angle, ɸpeak (°) (MPa) choosing a representative direction a specimen should be
000 48 0 sheared in when the in situ shear displacement direction is
unknown. Furthermore, the use of SfM photogrammetry
270 46 0.55 point cloud data and statistical methods of determining
300 45 0.1 JRC (or other roughness metrics) are critical for accurately
and reliably estimating the shear strength of rock fractures
330 39 0.5 until fitting parameters can be measured in analytical
models.
The variability in peak friction angle and the change in It is important to consider directional variability in shear
apparent cohesion between different shearing directions strength parameters. Additional work is underway to
on the same sample is shown to be significant (even at low- produce a relative shear strength reduction model to assist
moderate roughness values) and therefore important to practitioners with choosing shear direction for direct shear
consider for applications of direct shear test data in rock tests.
engineering design.

ACKNOWLEDGMENTS
6 CONCLUSIONS AND RECCOMENDATIONS
This research was financially supported by the Nuclear
This study demonstrates that rock fracture roughness Waste Management Organization of Canada and the
anisotropy is important when selecting a shearing direction Natural Sciences and Engineering Research Council of
for laboratory direct shear testing. Additionally, the relative Canada (Discovery Grants held by Diederichs and Day).
rock strength to applied normal stress ratio will affect the
 Sciences & Geomechanics, 5: 329-343.
REFERENCES https://doi.org/10.1016/0148-9062(90)92708-M
MathWorks Inc. 2022. MATLAB version: 9.13.2126072
Agisoft. 2023. Metashape Professional Photogrammetry (R2022b), Accessed: May 01, 2023. Available from:
Model Software, version. 1.7.3, Accessed: May 01, https://www.mathworks.com
2023. Available: https://www.agisoft.com/ Muralha, J. Grasselli, G. Tatone, B. Blϋmel, M.
Barton, N. 1971. A relationship between joint roughness Chryssanthakis, P. Yujing, J. 2014. ISRM Suggested
and joint shear strength. Symposium Société de la Method for Laboratory Determination of the Shear
mécanique des roches. Nancy. Strength of Rock Joints: Revised Version, Rock
Barton, N. 1973. Review of a New Shear-Strength Criterion Mechanics and Rock Engineering, 47: 291-302.
for Rock Joints, Engineering Geology, 7: 287-332. https://doi.org/10.1007/s00603-013-0519-z
https://doi.org/10.1016/0013-7952(73)90013-6 Packulak, T.R. Bonneau, D.A. Dossett, W.C. Rudderham,
Barton, N. Choubey, V. 1977. The Shear Strength of Rock G.A. Day, J.J. 2019. Evaluating Characterization
Joints in Theory and Practice, Rock Mechanics, 10: 1- Methods for Joint Surface Roughness. In Proceedings
54. https://doi.org/10.1007/BF01261801 of the 53rd US Rock Mechanics/ geomechanics
Beer, A.J. Stead, D. Coggan, J.S. 2002. Estimation of the Symposium, New York, NY, USA. 19-1994
Joint Roughness Coefficient (JRC) by Visual Pickering, C. 2015. Modeling Roughness of Rock
Comparison, Rock Mechanics and Rock Engineering Discontinuity Surfaces, Master’s Thesis, The University
35(1): 65-74. https://doi.org/10.1007/s006030200009 of Mississippi. Available from:
Blender Foundation, 2023. Blender version: 3.5.1, https://egrove.olemiss.edu/etd/1316/.
Accessed: May 01, 2023. Available from: Riss, J. Gentier, S. Archambault, G. Flamand, R. 1997.
https://www.blender.org Sheared Rock Joints: Dependence of Damage Zones
Bonneau, D.A. 2022. Applications of Remote Sensing for on Morphological Anisotropy, International Journal of
Characterizing Debris Transfer Processes. Doctoral Rock Mechanics and Mining Sciences 34:3-4.
thesis. Queen’s University. Available from https://doi.org/10.1016/S1365-1609(97)00151-2
https://qspace.library.queensu.ca/handle/1974/29193 Siemens, 2023. SolidEdge version: Student Edition 2023.
Chryssanthakis, P. Barton, N. 1990. Joint Roughness Accessed: May 01, 2023. Available from:
(JRC) characterization of a rock joint and joint replica at https://resources.sw.siemens.com
1 m scale. Rock Joints: Proceedings of the International Tatone, B. Grasselli, G. 2009. A method to evaluate the
Symposium on Rock Joints (Eds. Barton N, three-dimensional roughness of fracture surfaces in
Stephansson O), Loen, Norway, 4-6 June, pp 27-33. brittle geomaterials. Review of Scientific Instruments.
Coulomb, C.A. 1776. Essai sur une application des règles 80: 125110. https://doi.org/10.1063/1.3266964
de maximis et minimis quelques problèmes de statique, Tatone, B. Grasselli, G. 2010. A new discontinuity
relatis a l’architecture. Mémoires de Mathematique de roughness parameter and its correlation with JRC,
l’Academie Royale de Science 7 Paris. International Journal of Rock Mechanics and Mining
FormLabs. 2023. PreForm version: 2.15 Accessed: May Sciences, 47: 1391-1400.
01, 2023. Available: https://support.formlabs.com https://doi.org/10.1016/j.ijrmms.2010.06.006
Gentier, S. Riss, J. Archambault, G. Flamand, R. Hopkins, Tse, R. Creuden, D.M. 1979, Estimating Joint Roughness
D. 2000. Influence of fracture geometry on shear Coefficients. International. Journal of Rock Mechanics
behaviour. International Journal of Rock Mechanics Mining Sciences & Geomechanics Abstracts. 16: 303-
and Mining Sciences. 37: 161-174. 307. https://doi.org/10.1016/0148-9062(79)90241-9
https://doi.org/10.1016/S1365-1609(99)00096-9 Yu, X. Vayssade, B. 1991. Joint Profiles and their
Grasselli, G. Wirth, J. Egger, P. 2002. Quantitative three- Roughness Parameters, International Journal of Rock
dimensional description of a rough surface and Mechanics, Mining Sciences & Geomechanics, 4: 333-
parameter evolution with shearing, International 336. https://doi.org/10.1016/0148-9062(91)90598-G
Journal of Rock Mechanics, and Mining Sciences, 39:
789-800. https://doi.org/10.1016/S1365-
1609(02)00070-9
Grasselli, G. Egger, P. 2003. Constitutive law for the shear
strength of rock joints based on three-dimensional
surface parameters. International Journal of Rock
Mechanics and Mining Sciences. 40: 25-40.
https://doi.org/10.1016/S1365-1609(02)00101-6
MacDonald, N. 2022. A Critical Laboratory Investigation of
Multi-Stage Direct Shear testing procedures on Rock
Joints using Synthetic Replica Specimens. Master’s
Thesis. Queen’s University. Available from:
https://qspace.library.queensu.ca/handle/1974/30277
Maerz, N.H. Franklin, J.A. Bennett, C.P. 1990. Joint
Roughness Measurement Using Shadow Profilometry,
International Journal of Rock Mechanics, Mining
Behaviour of Two Poorly-Graded Sand in
Constant Height Direct Simple Shear Test
Ankan Mohajan, & Bipul Hawlader
Memorial University of Newfoundland, St. John’s, NL, Canada

ABSTRACT
Soil behaviour could be investigated using different types of laboratory testing systems. Among them, triaxial
compression and direct shear tests are commonly used for practical engineering. However, in many field problems, the
stress state and loading conditions are similar to simple shear conditions. It is also well-known that the response under
loading depends on soil type. In the present study, behavior of two sands, namely silica sand and Ottawa sand, is
investigated by conducting a series of constant height direct simple shear tests. Tests were conducted on dry sands
specimens at loose and medium dense conditions. The samples were sheared by monotonic loading. It is shown that the
undrained response of sand depends on relative density, and size and shape of soil grains.

RÉSUMÉ
Le comportement du sol pourrait être étudié à l'aide de différents types de systèmes d'essais en laboratoire. Parmi eux,
les essais de compression triaxiale et de cisaillement direct sont couramment utilisés pour l'ingénierie pratique.
Cependant, dans de nombreux problèmes de terrain, l'état de contrainte et les conditions de chargement sont similaires
aux conditions de cisaillement simple. Il est également bien connu que la réponse sous chargement dépend du type de
sol. Dans la présente étude, le comportement de deux sables, à savoir le sable de silice et le sable d'Ottawa, est étudié
en effectuant une série d'essais de cisaillement simple direct à hauteur constante. Des essais ont été effectués sur des
échantillons de sables secs dans des conditions lâches et moyennement denses. Les échantillons ont été cisaillés par
chargement monotone. On montre que la réponse non drainée du sable dépend de la densité relative, de la taille et de la
forme des grains de sol.

1 INTRODUCTION provide a better understanding of these types of

problems.
Several factors affect the behavior of sand, such as grain The objective of the present study is to investigate the
size and shape, relative density of soil sample, mode of behaviour of two poorly graded sands (silica sand and
shearing, arrangement of particles (fabric), and drainage Ottawa sand) which are not only different in particle size
conditions. The effects of particle size, shape and but also in particle shape and angularity. As mentioned
angularity on stress–strain behaviour of granular material before, several factors (e.g., sample preparation and
has been reported by several researchers (Lambe and fabric) could affect the response; however, the effects of
Whitman, 1979; Shahu and Yudhbir, 1998; Guo and Su, those factors are not investigated in the present study.
2007). Conducting triaxial compression test on two sands
at different confining pressure and initial void ratio, Guo
and Su (2007) stated that, interparticle locking due to 2 EXPERIMENTAL SETUP AND MATERIALS
angularity of particle tends to increase the peak friction
angle and also affects the dilatancy characteristics of Tests were conducted using the direct simple shear
sand. Based on cyclic simple shear tests on sand of the apparatus (ADVDCSS) built by GDS Instruments Ltd. for
same relative density, Vaid et al. (1985) showed that the Memorial University of Newfoundland, Canada (Fig. 1(a)).
resistance to liquefaction is higher for angular sand than The apparatus can be used to conduct monotonic and
rounded sand. No volume change occurs in an undrained cyclic tests at drained and undrained loading conditions
test, which can be ensured by maintaining constant height (Tarhouni and Hawlader 2021 & 2023). In this system, the
in DSS tests, as the specimen is laterally confined. lateral displacement of the soil specimen is restrained by
Therefore, the undrained DSS test is also known as using a stack of 1-mm-thick rigid Teflon-coated steel
constant height test. rings. No volume change occurs if the specimen height is
The undrained behavior of sand is stress path kept constant during shearing, which represents an
dependent—that is, significantly different response would undrained test (also known as constant height or constant
be expected if it is sheared in triaxial compression, triaxial volume DSS test). The present DSS apparatus can be
extension or simple shear mode (e.g., Vaid et al. 1990). used to conduct tests in saturated conditions; however, in
While direct shear and triaxial tests are commonly the present study, tests are performed on dry sand.
performed for geotechnical characterization, many A dual-axis 5-kN load cell controls and measures the axial
practical geotechnical problems are in simple shear mode and lateral loads with an accuracy better than 0.1% and a
(e.g., infinite slope failure, displacement of soil around pile resolution of 0.2 N. An LVDT of 2.5 mm range is
shaft). Therefore, direct simple shear (DSS) tests would mounted close to the DSS specimen to measure the axial
displacement with an accuracy better than 0.1% and a The shape of the sand particles is evaluated from the
resolution of 0.1 micron. In addition, an axial displacement scanning electron microscope (SEM) image, as shown in
transducer and motor encoder are used to crosscheck the Fig. 3. The silica sand is mostly sub-rounded and sub-
measured vertical displacements. A shear displacement angular to angular. The Ottawa sand is also a silica sand
transducer of 10 mm range with an accuracy better than manufactured by US Silica (Ottawa, Illinois) which has
0.1% and a resolution of 0.1 micron is located under the rounded to sub-rounded grains of larger size than the
specimen base. The shear displacement is also silica sand used in the present study. The other physical
measured using an encoder. A high-precision feedback properties of these two types of sand are summarized in
system controls forces and displacements while Table 1.
maintaining a high level of accuracy. Further details of the
apparatus specifications and capabilities are available in
Tarhouni and Hawlader (2021). (a) (b)

a)

b)

Tamping Figure 3: Scanning electron microscope (SEM) image: (a)

rod silica sand, (b) Ottawa sand

Table 1: Physical property of sands

Silica sand Ottawa sand

c) Mean grain size, D50 (mm) 0.2 0.39
Coefficient of uniformity, Cu 2.3 1.71
Figure 1: (a) Direct simple shear testing apparatus; (b)
Sample preparation; (c) Sample after test Coefficient of curvature, Cc 1.25 1.17
Specific gravity, Gs 2.65 2.65
Maximum void ratio, emax 1.048* 0.767**
Direct simple shear tests were conducted on two poorly Minimum void ratio, emin 0.606* 0.495**
graded sands: (i) silica sand and (ii) Ottawa sand. The
*Tarhouni & Hawlader (2021), **Tastan et al. (2022)
particle size distribution of the sand is shown in Fig. 2.
The basic properties of the sand are shown in Table 1. All
the tests were conducted according to ASTM standards.
2.1 Sample preparation
Also, some of these properties are obtained from previous
studies (Tarhouni and Hawlader, 2021; Tastan et al.
Two types of sand at two levels of density (loose and
2022)
medium dense) were tested. The soil specimens at loose
conditions (relative density, Dr = 31%–39%) were
prepared by funnel raining method, while the medium
dense specimens (Dr = 54%–64%) were prepared by dry
100
compaction after depositing the sand by funnel raining
methods as in loose sand specimen preparation. To
80 prepare the soil specimens, first a latex membrane
(0.012-inch thickness) was attached to the bottom
pedestal using two O-rings (Fig. 1(b)). The Teflon coated
% Finer

60 steel rings were then placed around the membrane

holding its upper part. Now, folding the upper part of the
40 membrane over the mould, a suction is applied that
attached the membrane inside the steel rings. Measuring
the weight of the sand required for specimen preparation,
20 Ottawa Sand the sand particles are poured inside the mold evenly
Silica Sand using a 6.8-mm diameter funnel. For loose specimens,
0
the drop height was almost zero. However, for medium
10 1 0.1 0.01
dense sand specimens, the drop height was ~10 mm. In
addition, the sand was compacted in three layers using a
Particle size, mm 68-mm diameter tamping rod (Fig. 1(c)). The initial height
Figure 2: Grain size distribution of silica and Ottawa sand of the specimens was 21.2–24.8 mm.
The mould with specimen was then placed on the direct are not shown here as the boundary effects affect the
simple shear apparatus. The top cap was slowly lowered response at large strains.
until it came in contact with the top surface of the soil
specimen, and a small normal stress (z) of 2.5 kPa
(seating load) was applied. The upper part of the
membrane was then folded back and then fixed with the
top cap with an O-ring. After that, the suction mold was
removed, and the bottom pedestal was attached with the
direct simple shear apparatus with four screws. A LVDT of
2.5 mm range was then placed next to the soil specimen
to measure vertical displacement. Note that the vertical
encoder located above the top cap also provides the
vertical displacement. It is to be noted here that, unlike
some DSS apparatus where the vertical displacement is
prevented simply by locking the vertical movement of the
top cap for constant height tests, the vertical displacement
in the present ADVDCSS is maintained zero (i.e.,
constant height) through computer controlling of the
vertical displacement using the encoder. The initial height
of the specimen measured before was verified using the
encoder axial displacement reading. At that time, the
specimen was ready for subsequent loading. The
specimen was then consolidated by applying normal
stress z0 of 80–300 kPa and then sheared at a shearing
rate of 0.16%/min while maintaining constant height
(undrained conditions). Here consolidation represents the
vertical compression of the dry sand specimen. The
shearing rate used in this study is similar to previous
studies (e.g.,Tarhouni and Hawlader, 2021).
The initial height of the soil specimen prior to
consolidation is 21–25 mm, and diameter of the specimen
is 70 mm. Based on the properties listed in Table 1, the
relative density of the loose and medium dense sand
specimens were 31%–39% and 54%–64%, respectively.
Figure 4: Stress–strain behaviour: (a) Silica sand; (b)
Ottawa sand
3 RESULTS AND DISCUSSION

3.1 Stress-strain behavior The overall response of Ottawa sand (Fig. 4(b)) is similar
to that of silica sand (Fig. 4(a)). However, the shear stress
Sieve analyses show that Ottawa sand is coarser than the at a given shear strain is significantly lower in Ottawa
silica sand (Fig. 2). In addition, a significant difference sand than in silica sand. As will be discussed further in
between these two sands is also found in the shape of the the later section, the slope of the ultimate steady state line
sand gains—the silica sand is sub-angular. The size and in the stress path or critical state friction angle of these
shape of the soil particles and relative density affects the two sand are not significantly different. Therefore, such a
stress–strain behaviour. large difference in shear stress is due to the higher
Figures 4(a) and 4(b) show the variation of shear stress dilation of silica sand, potentially due to sub-angular to
on the horizontal plane (zx) with shear strain (γ) for silica angular grain size, compared to that of Ottawa sand.
and Ottawa sands, respectively. In the following figures, Stress–strain behaviour can be better evaluated by
the thin and thick lines are used for loose and medium plotting shear stress in normalized form (/z0) (Fig. 5).
sand, respectively. Initially, the shear stress increases For loose silica sand, the normalized shear stress is
rapidly with shear strain within γ <1% and then the rate of significantly influenced by consolidation pressure prior to
increase slowed down or even decreased for a shearing; for example, the rate of increase of /z0 is
considerable increase in γ, especially for low higher for higher consolidation pressure (thin lines in Fig.
consolidation pressures. After that, the shear stress 5(a)). However, the effects of consolidation pressure on
increases with an increase in shear strain. For medium /z0 is less significant for medium sand (thick lines in Fig.
silica sand, the zx remains almost constant after γ ~ 12%, 5(a)). In Ottawa sand, the effects of consolidation
which implied that these specimens reached the ultimate pressure and density on /z0–γ are less significant than
steady state (Fig. 4(a)). On the other hand, loose silica that observed in silica sand (Fig. 5(b)).
sand specimens did not reach such ultimate steady state
within the shear strain of 20%. Note that the tests were
continued up to 30% shear strain; however, the results
 between the response of these two sand is potentially due
to higher dilation of silica sand during shearing.

Figure 5: Stress–strain response: (a) Silica sand; (b)

Ottawa sand
Figure 6: Stress paths: (a) Silica sand; (b) Ottawa sand

3.2 Stress paths Again for a better comparison of the response for different
densities and different consolidation pressures, the stress
Figure 6 shows the stress path (z–zx) for both sands. path is plotted by normalizing the normal and shear
For loose sand under low initial consolidation pressure, a stresses by the initial consolidation pressure (Fig. 7.).
significant strain softening occurs before the phase Both sands show varying initial contractive behaviour
transformation and then strain hardening. However, for (curves move to the left) which depends on the density
medium sand, the softening is relatively smaller than that and initial consolidation pressure. The maximum
of loose sands. These specimens had phase reduction of z/z0 occurs for low consolidation pressure
transformation (i.e., contraction to dilation) after a
and loose sand—for example, z/z0 reduced from 1.0 to
relatively smaller reduction of normal stress. After that soil
0.15 for loose silica sand before the phase change. The
moves towards the steady-state line at large shear
reduction of z/z0 is more for Ottawa sand (compare Fig.
strains.
7(b) with 7(a)). Another important observation is that,
Figure 6(b) shows the stress paths for Ottawa sand.
Compared to the response of silica sand (Fig. 6(a)), these while the slope of the zx/z0 vs. z/z0 lines after phase
sand specimens experience lower shear stress prior to transformation is different, they tend to merge to the same
phase transformation. line at large shear strains, except for the test on loose
For medium sand, the phase transformation occurs silica sand under 80 kPa consolidation pressure.
approximately at a constant angle in z–zx plot. The
phase transformation of medium silica sand occurs earlier
(lower angle ~22) while it occurs at ~26 for the Ottawa
sand. Also, for the Ottawa sand, the phase transformation
angle is closer to the steady state line. The difference
 phase transformation occurred at a lower angle
in silica sand.

5 ACKNOWLEDGEMENTS

The work presented in this paper has been supported by

the Natural Sciences and Engineering Research Council
of Canada (NSERC) and Equinor Research Chair grants.

6 REFERENCES

Al Tarhouni, M.A. 2020. Behaviour of Sand in Monotonic

and Cyclic Simple Shear Loading at Low-Stress Level,
Ph.D. Thesis, Memorial University of Newfoundland,
Canada.
Al Tarhouni, M.A. and Hawlader, B. 2021. Monotonic and
Cyclic Behaviour of Sand in Direct Simple Shear Test
Conditions Considering Low Stresses, Soil Dynamics
and Earthquake Engineering, 150: 106931.
Al Tarhouni, M.A. and Hawlader, B. 2023. Drained Cyclic
Behaviour and State-Dependent Stress–Dilatancy
Relationship of Sand in Direct Simple Shear Tests,
Soil Dynamics and Earthquake Engineering, 168:
107801
Been, K., Jefferiest, M.G. and Hachey, J. 1991. The
Critical State of Sands, Geotechnique, 41(3): 365-381.
Belkhatir, M., Missoum, H., Arab, A., Della, N. and
Schanz, T. 2011. Undrained Shear Strength of Sand-
silt Mixture: Effect of Intergranular Void Ratio and
Other Parameters, KSCE Journal of Civil Engineering,
Springer, 15(8): 1335-1342.
Bopp, P.A. and Lade, P.V. 2005. Relative Density Effects
on Undrained Sand Behavior at High Pressures, Soils
and Foundations, 45(1): 15-2.
Castro, G., Poulos, S.J., France, J.W. and Enos, J.L.
1982. Liquefaction Induced by Cyclic Loading, Report
to National science Foundation, Washington, DC.
Guo, P. and Su, X. 2007. Shear strength, Interparticle
Locking, and Dilatancy of Granular Materials,
Figure 7: Normalized stress paths: (a) Silica sand; (b) Canadian Geotechnical Journal, 44: 579–591.
Ottawa sand Kang, X. and Kang, G. 2015. Modified Monotonic Simple
Shear Tests on Silica Sand, Journal of Marine
Georesources & Geotechnology, 33: 122–126.
4 CONCLUSIONS Ladd, C.C. 1991. Stability Evaluation During Staged
Construction, ASCE Journal of Geotechnical
A series of undrained (constant height) monotonic direct Engineering, 117(4): 540-615.
simple shear tests was conducted on silica and Ottawa Lambe, T.W. and Whitman, R.V. 1979. Soil Mechanics,
sand. Tests were conducted on dry loose and medium John Wiley & Sons, Inc., New York.
dense sand after consolidating the specimens to 80–300 Miura, K., Maeda, K., Furukawa, M. and Toki, S. 1997.
kPa. The following conclusions can be drawn from this Physical Characteristics of Sands with Different
experimental study. Primary Properties, Soils and Foundations, 37(3): 53-
1) Angular silica sand shows more dilative 64.
behaviour than that of Ottawa sand. Negussey, D., Wijewickremwe, K.D. and Vaid, Y.P. 1988.
2) Medium dense sand specimens reached the Constant-Volume Friction Angle of Granular Materials,
ultimate steady state approximately at 12% Canadian Geotechnical Journal, 25: 50-55
shear strain; however, stress increase with shear Rouse, P.C. 2018. Relation between the Critical State
strain continued in loose sand, especially for Friction Angle of Sands and Low Vertical Stresses in
loose silica sand at lower consolidation stress. the Direct Shear Test, Soils and Foundations, 58:
3) Phase transformation in medium dense sand 1282–1287.
occurred approximately at a constant angle in
the shear stress vs normal stress plot. Also, the
Shahu, J.T. and Yudbhir. 1998. Model Tests on Sands
with Different Angularity and Mineralogy, Soils and
Foundations, 38(4): 151-158.
Shinohara, K., Oida, M. and Golman, B. 2000. Effect of
Particle Shape on Angle of Internal Friction by Triaxial
Compression Test, Powder Technology, Elsevier, 107:
131–136.
Sivathayalan, S. and Ha, D. 2011. Effect of Static Shear
Stress on the Cyclic Resistance of Sands in Simple
Shear Loading, Canadian Geotechnical Journal, 48:
1471–1484.
Souza Junior, P.L., Santos Junior, O.F., Fontoura, T.B.
and Freitas Neto, O. 2020. Drained and Undrained
Behavior of an Aeolian Sand from Natal, Brazil, Soils
and Rocks, São Paulo, 43(2): 263-270.
Tastan, E.O. and Carraro, J.A.H. 2022. The Effect of
Principal Stress Rotation and Intermediate Principal
Stress Changes on the Liquefaction Resistance and
Undrained Cyclic Response of Ottawa Sand, Journal
of Geotechnical and Geoenvironmental Engineering,
ASCE, 148(5).
Taylor, D.W. 1948. Fundamentals of Soil Mechanics.
John Wiley & Sons Inc., New York.
Tsomokos, A. and Georgiannou, V.N. 2010. Effect of
Grain Shape and Angularity on the Undrained
Response of fine Sand, Canadian Geotechnical
Journal, 47: 539-551.
Vaid, Y.P., Chern, J.C., and Tumi, H. 1985. Confining
Pressure, Grain Angularity, and Liquefaction, Journal
of Geotechnical Engineering, 111(10): 1229-1235.
Vaid, Y.P., Chung, E.K.F. and Kuerbis, R.H. 1990. Stress
Path and Steady State, Canadian Geotechnical
Journal, 27.
Winterkorn, H.F. and Fang, H.Y. 1986. Foundation
Engineering Handbook, Galgotia Book Source
Publishers, New Delhi.
Yoshimine, M., Ishihara, K. and Vargas, W. 1998. Effects
of Principal Stress Direction and Intermediate Principal
Stress on Undrained Shear Behavior of Sand, Soils
and Foundations, 38(3): 179-188.
Yoshimine, M., Ishihara, K. and Vargas, W. 1998. Flow
Potential of Sand During Liquefaction, Soils and
Foundations, 38(3): 189-198.
Yoshimine, M., Robertson, P.K. and Fear Wride, C.E.
1999. Undrained Shear Strength of Clean Sands to
Trigger Flow Liquefaction, Canadian Geotechnical
Journal, 36: 891–906.
Imaging a Sealed Shale Sample Using
Industrial Computed Tomography Scanning
Mahya Roustaei1,2, Joel Pumple1, Jordan Harvey1 & Duane Froese1
1Department of Earth and Atmospheric Sciences – University of

Alberta, Edmonton, Canada

2Department of Civil Engineering, Ghent University, Ghent, Belgium

Colin Dreger3
SNC Lavalin Inc., Vancouver, British Columbia, Canada
Derek Martin3
Klohn Crippen Berger, Vancouver, British Columbia, Canada
Ian Kerr & Andrew Watson
BC Hydro, Vancouver, British Columbia, Canada
3Seconded

ABSTRACT
Shale is typically a heterogeneous geomaterial that can undergo rapid transformation once exposed to the atmosphere.
Various methods have been used to minimize the effects of this transformation on the physical and mechanical properties.
For this study, 61.1 mm (HQ3) diameter shale core samples were investigated. Immediately following core retrieval these
cores were coated in canola oil and sealed in plastic wrap. The samples were then suspended in a PVC pipe and filled
with wax. The end caps were also PVC and partially filled with wax to ensure the ends of the samples were confined.
These tubes were used to maintain the integrity of the samples during storage and transportation. Non-destructive high-
resolution X-ray Computed Tomography (CT) scanning was used to investigate the mesostructure of the shale cores while
within the PVC container. The 3D composition of the core was imaged by industrial CT scanning. Images resolve the
distribution of burrow features (trace fossils) and fissures throughout the core. We conclude that this study provides
evidence for the value of non-destructive methods prior to removing sample containers.

RÉSUMÉ
Le schiste est un géomatériau hétérogène qui peut subir une transformation rapide une fois exposé à l'atmosphère.
Diverses méthodes ont été utilisées pour minimiser les effets de cette transformation sur les propriétés physiques et
mécaniques du schiste. Pour cette étude, des échantillons de schistes d'un diamètre de 61,1 mm (HQ3) ont été prélevés
dès que la carotte a été récupérée du tube de prélèvement. Les échantillons ont été enduits d'huile de canola et scellés
en utilisant du film plastique. Les échantillons ont ensuite été suspendus dans un tuyau en PVC et remplis complètement
de cire. Les bouchons d'extrémité étaient également en PVC et partiellement remplis de cire pour assurer l’étanchéité du
tube. Ces «tubes» ont été utilisés pour maintenir l’intégrité des échantillons, puis entreposer et transporter les échantillons.
Une analyse par tomographie axiale calculée par ordinateur (TACO) à haute-résolution non destructive a été utilisée pour
étudier la mésostructure des carottes de schiste à l'intérieur des tubes en PVC. La composition tridimensionnelle des
carottes a été imagée à l’aide du TACO. Les images permettent de visualiser la répartition des traces de fossiles et des
fissures dans toute la carotte. Nous concluons que cette recherche démontre la pertinence des méthodes non destructives
et permet de mieux comprendre les applications techniques.

1 INTRODUCTION CO2 storage and are proposed to host nuclear waste

repositories.
Estimating the in-situ properties of argillaceous shales is Procedures for sampling and testing of argillaceous
challenging because of the material disturbance that shales have evolved, however, there are still no
occurs when sampling for characterization purposes. recommended standards or guidelines. Giger et al. (2018)
These shales are relatively weak in terms of rock strength, provided procedures for sampling and testing of drill cores
brittle, fine-grained, and easily eroded sedimentary rocks of the Opalinus Clay in Switzerland needed to achieve
formed from mineral-rich silt, clay and/or mud that good practice in commercial testing laboratories. As noted
degrades and weathers into thin layers when exposed to by Giger et al (2018) medical X-ray tomography was used
the atmosphere. Argillaceous shales are typically to check the internal variability of the core for the selection
Cretaceous and Tertiary age and constitute as much as of samples, the conditions at the time of sampling and the
50% of the global sedimentary rock mass and outcrop in conditions at the time of testing. Keller & Giger (2019) also
about one third of the emerged earth surface (Gens 2015). used medical and high-resolution X-ray computed
These shales form the foundation of major civil engineering tomography (XCT) to assess the bulk density of the core
projects, host shale-gas reservoirs, form the cap-rock for material and when coupled with XRD analyses, were able
to characterize the mineral constituents with a reasonable
degree of certainty.
The development of computed tomography (CT)
scanning technology is evolving as a diagnostic tool for
assessing the variability of shales for testing. The objective
is to better understand the physical properties of shales
using fast and reliable characterization methods that
minimize sample exposure to atmosphere, therefore
limiting relaxation and deterioration which adversely impact
subsequent advanced laboratory testing outcomes. In this
study 61.1 mm diameter (HQ-sized) Shaftesbury shale
cores were retrieved during a drilling campaign at the Site
C Hydroelectric Project in British Columbia, Canada. The
samples were sealed and transported in PVC pipe Figure 1. Plan view of the site
containers to the Permafrost Archives Science (PACS)
Laboratory at the University of Alberta. There they were
scanned by a high-resolution X-ray commercial CT
scanner to image the 3D internal mesostructure geometry
of the core. The scanning was carried out while the sample
remained in the PVC container. In this paper we describe
the sampling procedure and the CT scanning methodology.

2 SITE DESCRIPTION

The Site C Clean Energy Project is a 1,100 MW

Hydroelectric project currently being constructed on the
Peace River approximately 7 km south of Fort St John,
British Columbia (Figure 1). Watson et al (2022) provides
an overview of the project and its geological setting. The Figure 2. Stratigraphy of the site and the location of the
project is located on the western edge of the Interior Plains extracted samples
and is part of the Western Canadian Sedimentary Basin.
The modern Peace River was formed during the retreat of
the most recent glaciation. Drainage of glacial Lake Peace 3 METHODS
caused incision through the Quaternary overburden
sequences and the Lower Cretaceous sedimentary 3.1 Sample Preparation
bedrock in the lower portion of the valley (Hartman and
Clague, 2008). Three samples were selected for this study, which were
The bedrock at Site C consists of the Lower Cretaceous extracted from two different boreholes, namely IC_RCC-L5
shales, siltstones, and silty sandstones of the Shaftesbury and IC_RCC-M7, during the investigation in 2020. As the
Formation, part of the Fort St John Group (Figure 2). These rock core was recovered, samples were selected and
rocks formed in a shallow marine depositional environment immediately prepared for sample preservation using the
and have a regional dip of 1° to 2° to the northwest. The following procedure:
rock at the base of the valley, where the samples were
collected, has an average unconfined compressive • Thoroughly wash the core in water to remove mud
strength of ranges from 5 to 15 MPa (Watson et al 2022), and debris.
is poorly indurated, and has up to 60% clay mineral content • Brush the sample liberally with canola oil.
(Sargent & Cornish 1985; Cornish & Moore 1985). The rock • Wrap in canola oil-saturated cheesecloth.
is prone to swelling and slaking, and undergoes rapid • Wrap in plastic wrap.
weathering when exposed, making sample recovery and • Wrap in clean cheesecloth.
preservation challenging (Sargent & Cornish 1985). • Wrap in fiberglass tape to create a “harness” to
Rock core samples of this study were collected and support and suspend the sample.
preserved during an investigation and testing program that • Hang and centre the sample in a Schedule 80 4” ID
took place in the Summer and Fall of 2020 on the south PVC tube, cut approximately 50 mm longer than the
bank of the Peace River. Samples were extracted from the sample, and with a bottom cap glued in place with
shales and silty shales that make up the lower stratigraphy PVC glue.
of the dam site (Figure 2). The shales and siltstones are • Fill the tube with beeswax just above melting
typically dark grey to black, soft to medium hard with local temperature (approximately 65° to 75° Celsius).
silty to sandy lenses of laminae. Samples were selected • Once the sample has cooled completely, a cap
from the intact portions of the core and ranged in length partially filled with beeswax is glued and tapped in
from 15 cm to 45 cm with a typical sample size of 20 cm to place using a rubber mallet to establish
30 cm. confinement.
 Samples were saturated with canola oil within 15 The four individual scans were reconstructed into three-
minutes of core retrieval and encased in beeswax within 1 dimensional gray-scale volumes using CT pro 3D software.
hour. Once the lids were glued in place the samples were Beam-hardening corrections and reconstruction were
effectively sealed from the atmosphere. applied to the scans also using Nikon XT 3D Pro software.
Reconstructed images were stitched together, visualized,
3.2 Scanning Methodology compared, and segmented using ORS Dragonfly software
(ver. 2021.1) to identify higher-density features throughout
Sealed core samples were scanned at the PACS lab at the the core.
University of Alberta using a Nikon XT H 225 ST Industrial Once run through CT scanner, encased cores were
CT scanner. The system uses an electronically adjustable removed from their PVC casings. PVC was split in half
225Kv 225W power source (Figure 3). This system is using a small circular saw, making sure not to cut too deep.
equipped with a tungsten rotating reflection target head Once split, the thick layer of beeswax was separated from
coupled with a 16-bit 2000x2000 pixel detector capable of the core using hot water to melt the majority of the wax.
a focal spot size range of 3-121 µm depending on the size The remaining beeswax was shaved off from the core and
of the area of interest and the size of the object being the core was wrapped in plastic wrap to reduce moisture
scanned. The core was scanned as four circular scans, loss.
with 10% overlap between scans. Circular scans were
used to minimize ring artifacts, with optimized projections
(3141 projections per rotation) averaged at 4 frames per 4 RESULTS
projection, resulting in a high-quality scan at 47 μm voxel
(three-dimensional pixel) resolution. A copper filter of 0.25 The L5-032 core which was selected to be scanned
mm was added to reduce artifacts and allow for increased consists of shale and interbedded silty sandstone of the
X-ray penetration/power through the dense shale core Shaftesbury Formation. Figure 4 shows the result of a CT
while avoiding detector saturation. scan that was completed on the sealed core sample as a
2D slice. This figure illustrates that the CT method provided
clear high-resolution images of the core within the PVC
housing (Figure 4). A combination of scan quality and
density contrast within the material allows for visual
differentiation of the materials based on the variation of
textures and grey value intensity. Overall, the PVC, wax,
and cheesecloth are similar in terms of gray values relative
to the core. Within the core itself, linear high-intensity
features are observed as the brightest or highest-density
material relative to the surrounding material. This clear
resolution and contrast suggest that software-based image
segmentation should isolate the shale core material from
the surrounding core housing materials and further
differentiate the internal components of the core itself.

Figure 4: 2D slice of the scanned core

Figure 3. (A) The CT scanner of the PACS Lab. (B) The
internal setup for the core scan
 The voxel gray values of each scan can be plotted Note that where the additional layer of PVC at the
versus the voxel population (population of volume pixels) bottom capped end of the tube is visible, the core material
and shown as a histogram (Figure 5). The shape of the and burrow features appear slightly darker in grey values.
histogram is affected by the proportions of component This increases the complexity of the segmentation process,
materials and can be taken as a representation of the as burrow features in this region appear less dense than
volumetric content in a sample (Calmels et al., 2010). The those above. This is a drawback of scanning the core while
bright spots shown in Figure 4 are interpreted as higher in the tube and could be remedied in the future by
density inclusions where infilling of burrow-like features positioning the sample length between the end caps during
occurred (red interval in the histogram of Figure 5). sample preparation, or by keeping the thickness of PVC
consistent for the whole length.

Figure 5: Voxel gray values versus voxel population of the

scan

High-resolution CT allowed the reconstruction of a 3D

view of these features in 2 different scales shown in Figure
6. This 3D illustration clearly reveals a complex network of
burrow features and fissures throughout the core’s profile.
These features are visible at the high end of the histogram
suggesting a high density relative to the host material and
lower volume as seen in Figure 5. The intensity and
configuration of the network varied throughout the core and
did not follow any discernible structure within the core itself.
Keller (2023) discovered a network of similar features while
conducting high-resolution XCT scanning of Opalinus Clay
samples recovered during investigations at the Mont Terri
Underground Research Laboratory in Switzerland.
Once scanning was completed the cores were split for
visual inspection. Figure 7 shows the scanned sample
within the PVC container and immediately following
removal, along with a high-resolution image of the split core
and the CT results of the encased core. The burrow-like
structures are not obvious when observing the high-
resolution split core, highlighting the utility of the CT
method to differentiate meso-scale features which may
otherwise be overlooked, using a non-destructive Figure 6: Illustration of the higher-density material in 2D
approach. and 3D
Figure 7: (A) Shale sample within PVC housing, (B) Shale sample after removal from the tube, (C) Split core image, (D)
2D CT core slice, (E) 3D CT image of the trace fossil features

5 CONCLUSIONS This non-destructive method can aid in the planning

process for future destructive work, and using their results
In this study, we non-destructively investigated the along with the CT imaging, provide useful details on the
microstructure of a sample of Shaftesbury Formation shale core with minimum destruction. In addition, this method
preserved in a PVC container. The investigation technique does not require the removal of the core from the casing,
used an industrial high-resolution CT scanner. thereby maintaining the integrity of the core for future
The results showed that the CT method can provide investigations.
clear, high-resolution images of the core within the PVC
housing, eliminating the need to remove the sample from a
confined condition where it’s protected from further 6 ACKNOWLEDGEMENTS
moisture loss and deterioration. This becomes increasingly
important when subsequent testing of the shale samples is The authors acknowledge the contributions to this project
required, where the results rely heavily on moisture content by BC Hydro, Klohn Crippen Berger, SNC-Lavalin, the Site
and the extent of sample damage. The method also turned C Resident Geological team, Cambridge Insitu Ltd.,
out to be particularly useful to image a complex network of Conetec Investigations Ltd., and Foundex Explorations.
burrow features (trace fossils) throughout the core’s profile. The authors would also like to thank BC Hydro for providing
These features are visible in the high end of the histogram the samples for the study and for permission to publish our
suggesting a high density relative to the host material. The findings.
intensity of the network varied throughout the core and did
not follow any visible core stratigraphy or structure.
It should be noted that the burrow-like structures are 7 REFERENCES
not obvious when observing the split core image,
highlighting the utility of the CT method as a non- Calmels, F., Clavano, W. R., & Froese, D. G. (2010).
destructive approach. Progress on X-ray computed tomography (CT)
Sealed core samples present a challenge for scanning in permafrost studies. GeoCalgary 2010 : The
researchers when considering the planning required to 63. Canadian Geotechnical Conference and 6.
quickly image and analyze the core once it has been Canadian Permafrost Conference, Calgary, AB
removed from the casing and exposed to the atmosphere. (Canada), 12-15 Sep , 1353–1358.
 https://doi.org/10.1016/j.egyr.2022.05.126
Cornish, L. J., & Moore, D. P. (1985). Dam Foundation
Investigations for a Project on Soft Shale. Proceedings
of 38th Canadian Geotechnical Conference,
Edmonton, pp171-178.
Gens, A. (2013). On the Hydromechanical Behaviour of
Argillaceous Hard Soils-Weak Rocks. Proceedings
15th European Conference on Soil Mechanics and
Geotechnical Engineering - Geotechnics of Hard Soils
- Weak Rocks (Part 4), (Editors: A. Anagnostopoulos,
M. Pachakis, & C. Tsatsanifos). 71-118.
Giger, S. B., Ewy, R. T., Favero, V., Stankovic, R., & Keller,
L. M. (2018). Consolidated-Undrained Triaxial Testing
of Opalinus Clay: Results and Method Validation.
Geomechanics for Energy and the Environment, 14,
16-28.
Hartman, G. M. D., & Clague, J. J. (2008). Quaternary
stratigraphy and glacial history of the Peace River
valley, northeast British Columbia. This article is one of
a selection of papers published in this Special Issue on
the theme Geology of northeastern British Columbia
and northwestern Alberta: diamonds, shallow gas,
gravel, and glaciers . Canadian Journal of Earth
Sciences, 45(5), 549–564. https://doi.org/10.1139/E07-
069
Keller, L. (2023). Investigation of a clay rock drill core using
X-ray computed tomography (XCT) with particular
focus on fossilized tube systems. Unpublished
manuscript. Zürich University of Applied Sciences,
Winterthur, Switzerland.
Keller, L. M., & Giger, S. B. (2019). Petrophysical
Properties of Opalinus Clay Drill Cores Determined
from Med-XCT Images. Geotechnical and Geological
Engineering, 37(4), 3507–3522.
https://doi.org/10.1007/s10706-019-00815-2
Sargent, D. W., & Cornish, L. J. (1985). Water
Susceptibility of Shaftesbury Shale. Proceedings of
38th Canadian Geotechnical Conference, Edmonton,
pp: 197-206.
Watson, A. D., Martin, D., Bruce, J., Van Gassen, W.,
Benabdellah, B., Douglas, S. et al. (2022). Site C Clean
Energy Project, Design and Construction Overview.
Proceedings from CDA 2022 Annual Conference,
October 17-19, 15p.
Considerations for Internal Erosion and
Observations of Particle Migration on
Bennett South Moraine Sand
Amy Swerid
Stantec, Vancouver, BC, Canada
Jonathan Fannin
Department of Civil Engineering – University of British Columbia,
Vancouver, BC, Canada

ABSTRACT
Dam owners have a responsibility to manage their dams to a standard where the risks associated with dam failure are as
low as reasonably practicable; this includes identifying and assessing the risk associated with internal erosion. Internal
erosion is responsible for approximately 35% of all earth embankment dam failures worldwide. In current practice, empirical
screening tools are available to assess material susceptibility to internal erosion. Current research now focuses on
development of a mechanics-based understanding of such erosion processes, to advance the management of internal
erosion risks.

Experimental testing was carried out on granular materials obtained from the W.A.C. Bennett Dam south moraine borrow
source to develop a mechanics-based understanding of internal instability. An index void ratio test program was carried
out on gap-graded BSM sand to characterize the sand in loose and dense states and assess the effect of changing finer
fraction content on the void ratio of a gap graded mixture. During testing, quantifiable particle migration was observed and
measured. This paper presents a review of experimental findings and discusses the implications relating to internal erosion
and considerations for index test methodologies for gap-graded soils.

RÉSUMÉ
Les propriétaires de barrages ont la responsabilité de gérer leurs barrages de manière à ce que les risques associés à la
rupture du barrage soient aussi bas que raisonnablement possible; cela nécessite donc de procéder à l’identification et à
l’évaluation du risque associé à l’érosion interne. L’érosion interne est responsable d’environ 35 % de toutes les ruptures
de barrages en remblai de terre dans le monde. Dans la pratique actuelle, il existe des outils d’évaluation empiriques
permettant d’évaluer la vulnérabilité des matériaux à l’érosion interne. La recherche actuelle vise à mieux comprendre ces
processus d’érosion sur le plan mécanique afin de faire progresser la gestion des risques liés à l’érosion interne.

Des essais expérimentaux ont été effectués sur des matériaux granulaires provenant de la source d’emprunt de la moraine
sud du barrage W.A.C. Bennett afin de comprendre l’instabilité interne sur le plan mécanique. Un programme d’essais en
fonction de l’indice des vides a été réalisé sur du sable à granulométrie variable provenant de la moraine sud du barrage
afin de caractériser le sable dans des états lâches et denses, et pour évaluer l’effet de la modification du contenu en une
fraction plus fine de l’indice des vides d’agrégats à granulométrie discontinue. Au cours des essais, une migration
quantifiable des particules a été observée et mesurée. Cet article présente une revue des résultats expérimentaux et
aborde les implications relatives à l’érosion interne et des considérations liées aux méthodes d’essai d’indices pour les
sols à granulométrie discontinue.

1 INTRODUCTION instability (ICOLD 2017; USBR & USACE 2019). These

screening tools are primarily based on particle size
Internal erosion is responsible for approximately 35% of all distribution. Currently no mechanics-based analytical
earth embankment failures (ICOLD 2019). Internal erosion models have been developed to assess material
occurs when seepage flow erodes and transports soil susceptibility to internal instability.
particles within the dam embankment or foundation. There In 2017 the UBC internal erosion research group
are four mechanisms for the initiation of internal erosion embarked on a multi-year research project supported by
including internal instability, concentrated leak erosion, BC Hydro and the Natural Sciences and Engineering
backward erosion, and contact erosion. The current study Research Council of Canada (NSERC). The objective of
focuses on the mechanism of internal instability. Internal the project is to develop a mechanics-based understanding
instability occurs when the finer particles within a soil mass of internal instability using laboratory testing to inform a
are transported through it by seepage flow and the coarse critical state type model to quantitatively predict the effect
grains remain in place (ICOLD 2017). of internal erosion in embankment dams. Materials for
In current practice, empirical screening tools are laboratory testing were obtained from the Bennett South
available to assess material susceptibility to internal
Moraine (BSM) borrow source, used in construction of the maximum and minimum densities for six different test
W.A.C. Bennett Dam and referred to as BSM sand herein. standards using six different uniformly graded sands. The
Index testing was carried out on uniformly graded and study found for the same sand, the index properties varied
binary mixtures of BSM sand to characterize the sand in from acceptable to unacceptable tolerances and the
loose and dense states and assess the effect of changing measured index properties relied heavily on the methods
finer fraction content on the void ratio of a gap-graded used.
mixture. During testing, quantifiable particle migration was
observed and measured. This paper presents a review of 2.1.1 Index Void Ratios for Gap-Graded Sands
experimental findings and discusses the implications as
they relate to internal erosion and considerations for index The effect of changing fines content on the emax and emin
test methodologies for gap-graded soils. of gap-graded mixtures has been well studied (see for
example, Pitman et al., 1994; Lade et al., 1998;
Cubrinovski & Ishihara, 2002). Lade et al. (1998)
2 BACKGROUND conducted a study with a systematic experimental test
program to evaluate emax and emin of binary mixtures with
Soil structure is a fundamental engineering property that changing Sf. Uniform sands were blended to form binary
affects the density of a material and subsequently the mixtures with 1.8 ≤ D′50/d′50 ≤ 30. The study found both Sf
mechanical response of granular soils. Structure is the and mean particle diameter ratio (D′50/d′50) have a
“combined effect of the soil fabric, composition and significant effect on the void ratio.
interparticle forces”, in which fabric refers to the Cubrinovski & Ishihara (2002) conducted a
“arrangement of particles, particle groups and pore spaces comprehensive study on the characteristics of emax and emin
in a soil” (Mitchell & Soga, 2005). using data from over 300 sandy soils including clean
In binary and gap-graded materials the particle packing sands, binary mixtures of sands, sands and silts, and
arrangement and resultant microstructure can be highly sands and clays. The void ratio range (emax-emin) was used
complex. A binary mixture consists of a soil mixture with a as an index property to aid in the soil characterization.
coarse fraction of diameter (D′) and a finer fraction of a Particle shape was found to influence emax and emin. An
diameter (d′). It is widely accepted the particle packing increase in particle roundness resulted in a higher void
arrangement in gap-graded mixtures is affected by the size ratio. The fines content in a binary mixture largely affects
ratio of the coarse and fine fractions (D′/d′) and the the particle structure, and thus emax and emin. In the gap-
percentage finer fraction content (Sf) in the total mixture graded mixtures assessed, the typical fines threshold
(Lade et al., 1998; Cubrinovski & Ishihara, 2002; content was found to be Sf(th) ≈ 30%. Below Sf(th) the void
Thevanayagam et al., 2002; Prasomsri & Takahashi, ratio decreased with increasing Sf; above Sf(th) the void ratio
2020). At Sf = 0%, the soil structure is that of the coarse increased with increasing Sf. It was also concluded that
fraction alone. As a finer fraction is added to the coarse “wide gaps in the grain-size distribution may significantly
fraction the fine particles infill the voids of the coarse affect the particle structure of sands and these effects
fraction structure. If the voids of the coarse fraction are should be carefully considered when producing composite
sufficiently large enough, at relatively low S f the finer soils” (Cubrinovski & Ishihara, 2002).
particles may not contribute to the particle structure and the
soil structure remains fully clast supported. A theoretical
threshold fines content (Sf(th)) has been identified, at which 3 INDEX VOID RATIO TESTING OF BSM SAND
the finer fraction cannot be fully contained within the coarse
fraction and the fines must contribute to the soil structure. In the current study, an index test program was carried out
on uniformly graded and binary mixtures of BSM sand to
2.1 Index Void Ratios characterize the sand in loose and dense states, assess
the effect of changing finer fraction content on the void ratio
The index void ratios emax and emin are used to represent of a gap-graded mixture, and compare the indices of BSM
the respective loosest and densest states of a given soil. sand with other similar mixtures.
Standardized test procedures have been developed to
determine the emax and emin values for a soil. In North 3.1 Materials and Gradations
America standardized tests for emax and emin include ASTM
D4254 (ASTM, 2016a) for emax, and ASTM D4253 (ASTM, Tests were performed on BSM sand sourced from the
2016b) for emin. Non-standardized methods have also been W.A.C. Bennett Dam south moraine borrow pit. The
used to determine index void ratios (see for example, location of the borrow pit is shown on Figure 1. The south
McGeary, 1961; Lade et al., 1998). They are generally moraine borrow pit was the primary material source for the
used if the material to be tested is outside the gradation W.A.C. Bennett Dam and was described as a glacial-fluvial
limits specified in standardized test procedures. well-bedded gravel, sand and silt mixture (Morgan & Harris,
The index test values of emax and emin are non-unique 1967) and subsequently reinterpreted as a kame moraine
and can have a high degree of variability. The void ratios deposit (Mathews, 1980).
determined using these test methods are not necessarily Physical properties of BSM sand were assessed in a
the absolute maximum and minimum void ratios for a previous study by the UBC internal erosion research group
particular soil (ASTM, 2016a; ASTM, 2016b). Lunne et al. (Antunes, 2021). The minerology in the size range of
(2019) conducted a study to assess differences between interest is comprised of approximately 65-70% silicates
 and 25-30% carbonates with the remainder being were carried out using (i) standardized, (ii) minor variations
miscellaneous minerals. The particle shape is subangular of standardized, and (iii) non-standardized test methods.
to angular where angularity increases with decreasing The trials were carried out on BSM 10:15, the gradation
grain size. speculated to be most susceptible to segregation.
Three uniformly graded components were used in the Following each trial test segregation was assessed both
study to form idealized gap-graded mixtures. Each mixture qualitatively and quantitatively. The test procedure was
consists of a coarse fraction and a finer fraction. Two selected based on the method found to result in the most
different coarse fractions were examined in testing, with homogenous test specimen and repeatable extreme void
particles passing sieve sizes from the No. 6 to 14 (BSM ratio.
6/14) and No. 10 to 18 (BSM 10/18); the finer fraction
consisted of particles passing from the No. 70 to 140 sieve
size (BSM 70/140). The idealized mixtures have a nominal
gap ratio (D′15/d′85) of 7 and 10, respectively for mixtures of
BSM 10/18 with BSM 70/140 (termed BSM 7) and BSM
6/14 with BSM 70/140 (termed BSM 10). The finer fraction
content (Sf) for both BSM 7 and BSM 10 was varied in 5%
intervals from 15 to 35%, to yield the test gradations BSM
7:15 to BSM 7:35 and BSM 10:15 to BSM 10:35 (see
Figure 2).

Bennett South Moraine

Burrow Source

Figure 1. Location of the Bennett South Moraine Burrow

Source (Extracted from Google Earth on May 5, 2023)

3.2 Test Apparatus Figure 2. Gradations used to test for emax and emin

The equipment used for minimum and maximum index

densities complied with ASTM test standards (ASTM 3.3.1 Procedure for Maximum Void Ratio
2016a; ASTM 2016b). The apparatus to conduct the emax
and emin testing included an ASTM standard mold with a The maximum void ratio was determined using Method A
volume of 2,830 cm3, digital scale, and ASTM pouring of ASTM (2016a) with a minor variation. Specifically, the
device. Additionally, a vibratory table, standard mold procedure to fill the mold was slightly modified from that of
assembly, surcharge base plate, 25.6 kg surcharge weight, ASTM. Rather than filling the mold with the pouring device
and dial-indicator gauge were used (see ASTM, 2016b) to in one steady stream, the mold was filled with 500 g
carry out emin testing. The vibrating table used was a floor- “batches” of the desired gradation which was mixed within
mounted vibrating table with vertical vibrating deck the pouring device then poured into the mold following ASTM
calibrated to have a double amplitude displacement of (2016a) procedures.
0.279 mm at a frequency of 60 Hz in accordance with
ASTM (2016b). 3.3.2 Procedure for Maximum Void Ratio

3.3 Test Procedures The minimum void ratio was determined using Method 1A
of ASTM (2016b). A slight modification to the procedure
The relatively large gap ratio of the proposed test was made where the vibrating time was doubled from 8
gradations raised concern for potential segregation of the minutes to 16 minutes. It was found that doubling the
finer fraction during testing. To mitigate this concern, vibrating time provided more repeatable results. The
measures were taken to minimize segregation during procedure is briefly summarized below. Additionally, the
testing. Prior to initiating the test program, preliminary trials
mold was filled in 500 g “batches” of the desired test 80/200 mixture has a D′50/d′50 = 12.5. At Sf = 0% the emax =
gradation. 0.77, and with increasing Sf it decreases to emax = 0.57 at
Sf = 35% and then increases to emax = 0.94 at Sf = 100%.
3.3.3 Specimen Homogeneity In comparison: BSM 7 has a D′50/d′50 = 11, and at Sf = 0%
it has an emax = 0.82 that decreases to emax = 0.54 at Sf =
Quality assurance checks were carried out to assess the 35%, and then increases to emax = 1.01 at Sf = 100%.
specimen homogeneity following completion of emax and
emin tests. The material in the mold was extracted in 6
layers of approximate equal thickness. The coarse and fine
fractions in each layer were separated using a No. 70
sieve. The coarse and fine fractions were weighed and the
Sf in each extracted layer was calculated. The post-test Sf
of each layer was compared with the target Sf for the test
gradation.

3.4 Results

3.4.1 Effect of Finer Fraction Content on Void Ratio

The individual uniformly graded components represent the

coarse fraction (BSM 10/18 or BSM 6/14) and the finer
fraction (BSM 70/140) of the gap-graded mixtures. The
index void ratios of the two coarse fractions were found to
be similar. BSM 6/14, yields emax = 0.78 and emin = 0.58.
BSM 10/18, yields emax = 0.82 and emin = 0.59 (see Figure
3). The values for the finer fraction are larger than those of
the two coarse fractions: BSM 70/140, yields emax = 1.01
and emin = 0.68 (see Figure 3).
The finer fraction was blended with each of the coarser
fraction to form gap-graded mixtures of BSM 7 and BSM
10 with finer fraction contents of Sf = 15, 20, 25, 30, and
35%. Overall, for both the BSM 7 and BSM 10 mixtures the
emax and emin decrease with increasing Sf to approximately
Sf = 30% (see Figure 3). The emax of both gradations is
nearly constant for 30 ≤ Sf ≤ 35%, whereas the emin
increases from Sf = 30% to Sf = 35%. For Sf ≥ 35% the void
ratio is expected to increase to the void ratio of BSM
70/140, representative of Sf = 100%.
Figure 3. Comparison of (a) emax and (b) emin tests of BSM
The results of the index tests on BSM sand were
compared with the emax and emin values reported for similar
binary Figure 3sand and other gap-graded sands mixtures BSM 10 has a D′50/d′50 =14, and at Sf = 0% it has
of two select research sands (Pitman et al., 1994; Lade et an emax = 0.78 that decreases to emax = 0.49 at Sf = 35%,
al., 1998; Ouyang & Takahashi, 2016; Prasomsri & and then increases to emax = 1.01 at Sf = 100%.
Takahashi, 2020), see Figure 3. The objective of the Accordingly, the BSM sand mixtures and the Cambria and
comparison is to benchmark the packing behavior of BSM Nevada 80/200 mixture exhibit a very similar variation of
sand against similar sands of similar gap-gradation, based emax with changing Sf. The finding supports the claim that
on particle size and particle size ratio (D′50/d′50). It is noted BSM sand exhibits a similar packing behavior to other gap-
that the index void ratios reported for the other sands were graded sands. The other mixtures compared had a similar
determined using different test methods. Thus, the trend as BSM sand with increasing Sf but the absolute emax
comparison is carried out under the assumption that whilst values are not as close to BSM sand and the Cambria and
different test procedures may result in different absolute Nevada mixtures. The emax of two silica sand mixtures
values of index void ratio, any relative change with S f is a compared (Ouyang & Takahashi, 2016; Prasomsri &
product of the soil material properties and not the test Takahashi, 2020) have a minimum emax at Sf ≈ 30% but the
method itself. emax values are approximately 0.1 to 0.2 larger than that of
The variation of emax with changing Sf for BSM sand the BSM sand mixtures.
exhibits a trend that is similar to all of the other mixtures. The variation of emin for BSM sand with changing Sf is
Specifically, emax decreases with increasing Sf to a compared to other gap-graded sands for purposes of
minimum value in the range of 30 ≤ Sf ≤ 35%, and completeness (see Figure 3(b)), but not assessed
subsequently increases to the emax at Sf = 100%. The quantitatively because of the adverse segregation noted
Cambria and Nevada 80/200 mixture is the most previously. In general, the emin of BSM sand with changing
comparable to BSM sand. The Cambria and Nevada Sf trends similarly to the other sand mixtures compared.
Table 1. Index void ratio test program: post-test specimen homogeneity

Difference between measured finer fraction content and target finer fraction content
Maximum void ratio
BSM BSM BSM BSM BSM BSM BSM
Layer 7:15 7:20 10:15 10:20 10:25 10:30 10:35
1 1.9 -2.1 -8.7 -1.1 0.2 -1.2 -0.6
2 -0.6 -1.0 -4.4 -0.9 -0.2 -0.9 0.1
3 -2.0 1.0 -4.0 0.8 -0.2 -0.5 -0.2
4 -0.6 -0.4 3.7 -0.9 0.1 -0.2 0.1
5 -1.6 0.7 3.5 0.6 -1.6 1.5 -0.2
6 2.8 1.8 9.9 1.5 1.7 1.3 1.0
Minimum void ratio
BSM BSM BSM BSM BSM BSM BSM BSM BSM BSM
Layer 7:15 7:20 7:25 7:30 7:35 10:15 10:20 10:25 10:30 10:35
1 -15.0 -20.0 -18.2 -4.8 -3.2 -15.0 -18.0 -19.5 -9.5 -4.1
2 -15.0 -10.7 0.4 0.2 1.5 -15.0 -13.0 0.0 0.7 0.7
3 -7.7 2.1 3.2 -0.6 0.5 -12.4 -1.4 3.7 0.8 -0.4
4 8.3 6.6 3.6 0.7 1.2 9.8 8.8 4.2 1.9 1.9
5 12.4 9.6 4.0 1.6 -0.4 15.2 10.5 4.9 1.9 1.0
6 17.1 12.4 7.0 2.8 0.5 17.4 13.2 6.7 4.2 0.9

In summary, the purpose of comparing the index void ratios content is relatively constant 32 ≤ Sf ≤ 36% at the end of
of BSM sand to other mixtures was to benchmark the densification by vibration. Post-test specimen homogeneity
properties found for two gap-graded BSM sands with other checks established that segregation ranged from negligible
similar sand gradations. Overall, the trends and index void for some gap gradations to very considerable for others. A
ratios found for BSM sand are similar to the other binary subjective limit-value approach was used to assess if
mixtures. Of the mixtures assessed, BSM sand is most segregation was likely adverse: a specimen was assumed
similar to a Cambria and Nevada sand mixture. representative of the target gradation if it had a post-test
finer fraction content ±2 < ∆Sf < ±5% in any single layer and
3.6.3 Specimen Homogeneity all other layers had -2 < ∆Sf < 2%.
For emax tests, only BSM 10:15 is deemed to be
Specimen homogeneity was checked following completion potentially affected by adverse segregation. All other
of most index tests. The specimen was extracted in six gradations are deemed representative of the target
approximately uniform layers and the finer fraction content gradation. The emax = 0.60 for BSM 10:15 (see Figure 3) is
in each layer then determined. The results of the post-test in good agreement with the emax of other tests on BSM 10
specimen homogeneity are summarized in Table 1 in terms and the resulting emax agrees well with that of BSM 7. Since
of the difference between the measured and target finer there is good agreement with gradations that are not
fraction content (∆Sf). The results given are for one test per adversely affected by segregation, the BSM 10:15 gap-
gradation. gradation is not considered adversely affected by
All emax test gradations with the exception of BSM segregation in index testing at the loosest state.
10:15 had ∆Sf ±3% in all layers and most layers were within For emin tests, all gap-gradations except BSM 7:35
∆Sf ±1%. BSM 10:15 experienced the greatest amount of and BSM 10:35 are considered to be potentially adversely
segregation of all emax tests: the top half of the specimen affected by segregation. Accordingly, the emin tests with 15
(top three layers) had -9 ≤ ∆Sf ≤ -4% while the bottom half ≤ Sf ≤ 30% are not believed appropriate to use in an
of the specimen (bottom three layers) had 4 ≤ ∆S f ≤ 10%. assessment of microstructure since there is clear evidence
Post-test homogeneity was not assessed for the BSM 7 of specimen inhomogeneity as a consequence of adverse
material with 25 ≤ Sf ≤ 35% since minimal segregation was segregation at the densest state.
found in the BSM 10 material with 25 ≤ Sf ≤ 35%.
Post-test homogeneity measurements for the emin
tests indicate relatively greater amounts of segregation 4 KEY FINDINGS AND CONSIDERATIONS
compared to the emax tests. For 15 ≤ Sf ≤ 25%, some of the
top layers lose up to all of the finer fraction content. For 30 The emax and emin values found for BSM sand were similar
≤ Sf ≤ 35% the segregation is less pronounced in the top to those found for similar sand mixtures. Index testing on
and layers. In all emin tests, the bottom layer finer fraction the uniformly graded composites identified emax and emin of
BSM sand increase with decreasing particle size. A general Cubrinovski, M. & Ishihara, K. (2002). Maxiumum and
trend was identified in the gap-graded mixtures where the minimum void ratio characteristics of sands . Soils
index void ratios decrease with changing Sf between 0 ≤ Sf and Foundations, 42, 65-78.
≤ Sf(th) and then increase between Sf(th) ≤ Sf ≤ 100%. The ICOLD (2019a). Bulletin 99 Update Statistical Analysis of
trend is generally consistent with expectations from particle Dam Failures (Draft). Dam Incident Database.
packing theory. Overall, BSM sand has similar packing ICOLD (2017). Internal Erosion of Existing Dams, Levees
properties to other research sands. and Dikes, and their Foundations. Bulletin 164.
The test program included quality assurance where Lade, P. V., Liggo, C. D. & Yamamuro, J. A. (1998). Effects
tests were checked, as appropriate, for segregation at the of Non Plastic Fines on Minimum and Maximum Void
end of test. It was found that all emax tests were not Ratios of Sand. Geotechnical Testing Journal, Vol.
adversely affected by segregation whereas, in contrast, the 21, No. 4, 336-347.
emin tests suffered from adverse segregation for 15 ≤ Sf ≤ Lunne, T., Knudsen, S., Blaker, O., Vestgarden, T., Powell,
30%. Thus, BSM 7 and BSM 10 were adversely affected J., Wallace, C., . . . Ghanekar, R. (2019). Methods
by segregation during emin testing below Sf(th). The used to determine maxiumum and minimum dry
segregation is attributed to finer fraction particle migration unit weight of sand: Is there a need for a new
in a dry state. standard? Canadian Geotechnical Journal, 56,
The segregation observed in the test program 536-553.
illustrates standardized test methodologies for the Mathews, W.H. (1980). Retreat of the last ice sheets in
minimum void ratio may not be suitable for soils with large northeastern British Columbia and adjacent Alberta.
gap-gradations, such as the BSM7 and BSM 10. Care Geological Survey of Canada, 331, 22p.
should be taken when determining the minimum void ratio McClelland, V. (2020). On the Critical State of Gap-Graded
for gap-graded sands, as segregation may adversely affect Sandy Soil. The University of British Columbia,
the test results. M.A.Sc. thesis.
McGeary, R. K. (1961). Mechanical Packing of Spherical
Particles. Journal of the American Ceramic Society,
ACKNOWLEDGMENT Vol. 44, No. 10, 513-512.
Mitchell, J. & Soga, K. (2005). Fundementals of Soil
This work is part of an industry-funded research project on Behavior third edition. Hoboken: John Wiley & Sons,
modelling the internal erosion of earth dams, with support Inc.
from BC Hydro Dam Safety and the NSERC CRD program. Morgan, G. C. & Harris, M. (1967). Portage Mountain Dam.
Canadian Geotechnical Journal, IV(2), 142-166.
Ouyang, M. & Takahashi, A. (2016). Influence of initial fines
5 REFERENCES content on fabric of soils subjected to internal
erosion. Canadian Geotechnical Journal, 53, 299-
Antunes, A. (2021). On the Critical State Locus of the 313.
Bennett South Morain Sand. The University of British Pitman, T. D., Robertson, P. & Sego, D. (1994). Influence
Columbia, M.A.Sc. thesis. of fines on the collapse of loose sands. Canadian
ASTM (2016a). Standard Test Methods for Minimum Index Geotechnical Journal, 31, 728-739.
Density and Unit Weight of Soils and Calculation of Prasomsri, J. & Takahashi, A. (2020). The role of fines on
Relative Density. Designation D4254-16. internal instability and its impact on undrained
ASTM (2016b). Standard Test Methods for Maximum Index mechanical response of gap-graded soils. Soils and
Density and Unit Weight of Soils Using a Vibratory Foundations, 60, 1468-1488.
Table. USBR & USACE (2019). Internal Erosion Risks for
Cubrinovski, M. & Ishihara, K. (1999). Empirical correlation Embankments and Foundations. In Best Practices in
between SPT N-value and relative density for sandy Dam and Levee Safety Risk Analysis.
soils. Soils and Foundations, 39(5), 61-71.
Utilizing a microfiber-optic pressure sensor
in triaxial testing for local measurements in
clay samples
Kevin Wallin, Dr. Derek Martin & Dr. Michael Hendry
University of Alberta, Edmonton, Alberta, Canada

ABSTRACT
This paper presents early results from the adaptation of a triaxial device that utilizes a micro-fiber optic pressure sensor
(MFOPS) to measure local (in specimen) pore pressures generated during shear. This includes an overview of calibration
and a brief discussion on the results. During the triaxial testing, the system measures the pore pressure applied at the top
and bottom of the specimen with standard external pore pressure sensors, which are then compared to an internal
measurement captured by the MFOPS. This triaxial testing program is being utilized to confirm the applicability of the
MFOPS technology.

RÉSUMÉ
Cet article présente les premiers résultats de l'adaptation d'un dispositif triaxial qui utilise un capteur de pression à
microfibre optique (MFOPS) pour mesurer les pressions interstitielles locales (dans l'échantillon) générées pendant le
cisaillement. Cela comprend un aperçu de l'étalonnage et une brève discussion sur les résultats. Pendant le test triaxial,
le système mesure la pression interstitielle appliquée en haut et en bas de l'échantillon avec des capteurs de pression
interstitielle externes standard, qui sont ensuite comparés à une mesure interne capturée par le MFOPS. Ce programme
d'essais triaxiaux est utilisé pour confirmer l'applicabilité de la technologie MFOPS.

1 INTRODUCTION parallel to the direction of isotropy. Restricting dilation

during shear creates the potential for rapidly increasing
Countries with argillaceous (clay-rich) formations such as pore pressures. This boundary condition that suppresses
Canada and Switzerland are assessing this class of dilation is similar to a spring controlled by the rock mass
material as potential host rock for nuclear waste disposal stiffness and is generally referred to as a strain-controlled
or as barrier rock above a repository. The current boundary (Button & Blümel, 2002; Di Donna et al., 2016;
international regulatory framework for nuclear waste Poturovic et al., 2015).
disposal requires unprecedented predictive design Triaxial tests cannot simulate this boundary condition
capacity, with Canada considering timeframes as far out as and pore pressures are typically measured using global
one million years (NWMO, 2013). The site selection monitoring on the external boundary of the specimen. To
process in Canada is being conducted by The Nuclear investigate the impact of the boundary conditions on the
Waste Management Organization (NWMO) who has pore pressure response in heavily overconsolidated shales
narrowed down their site selection from 22 locations to only two issues needed to be solved: 1) develop a technique
two sites, with a final selection expected in 2024. A Middle where a pore pressure sensor could be mounted internally
Ordovician aged argillaceous limestone that is part of the in the material so that local shearing-induced pore
Cobourg Formation in southern Ontario is one of the pressures can be monitored, and 2) develop a laboratory
deposits being considered for nuclear waste storage test system that can accommodate the internal pore
(Intera, 2011). In Switzerland, the Opalinus Clay (OPA) has pressure sensors while maintaining a strain controlled
already been chosen as the host rock and the site selection (zero dilation) boundary condition during shear loading.
process is complete with the repository location announced This paper describes the development of the internal pore
in September 2022 to be Nördlich Lägern. To meet the pressure sensor and its calibration and testing.
required design criteria for safe storage, the behavior of the
geological material surrounding these nuclear waste
repositories needs to be comprehensively understood to 2 STRAIN LOCALIZATION AND THE HISTORY OF
proceed with engineering design (Gens, 2013). INTERNAL PORE PRESSURE MEASUREMENTS
It is well known that the pore pressure response in clay-
rich soils is dominated by the stress history of the material. The concept of strain localizations and shear zones is well
The fundamental characteristic pore pressure response understood both in coarse and fine grained materials,
observed in overconsolidated clays is that during shear- where strain localization has been described as a
induced yielding the material will dilate, which then leads to precursor to the formation of a shear zone (Giger et al.,
a decrease in pore pressure. Around the face of 2015; Su, 2005; Winhausen et al., 2022).
underground tunnels, dilation in transversely isotropic To supplement the testing that has been completed, an
materials like OPA or the argillaceous deposits in the assessment of the localized pore pressure response along
Cobourg can be suppressed if the driving of the tunnel is these discrete shear surfaces under undrained conditions
would provide a more complete understanding of the pass through for the MFOPS cabling, and a new sample
effective stress path. Excess pore water pressures inside pedestal (Figure 1 and 2). The triaxial system has a total of
of shear zones has been theorized and studied numerically five pressure sensors: two sensors located at the top of the
in clays and sands (Schrefler et al., 1998; Shuttle & Smith, sample, one at the base, a global cell pressure sensor, and
1990; Thakur, 2007, 2011), however, experimental testing the MFOPS inside of the sample. The MFOPS extends
has primarily been limited to plane strain testing or triaxial approximately to the midpoint of the sample (33 mm) to
testing with external sensors (Amann et al., 2011; Cheng, attempt to be within the predicted failure zone.
2004; Thakur et al., 2018; Viggiani et al., 1994; Wild &
Amann, 2018; Winhausen et al., 2022; Yuan et al., 2013).
For these tests, pore pressure sensors were placed on the
external boundary of the specimen at varying distances
from the shear zone. As expected, the highest pore
pressure response occurred at the sensors located closest
to the shear band, with the cause being attributed to
contraction inside of the failure zone.
An attempt at internal pore pressure measurements on
a shear zone of a fine-grained material was trialed in the
laboratory by Su (2005) on weak rocks from the McMurray
formation in Alberta, Canada. Su (2005) employed a large
scale triaxial testing program on 100 mm diameter by
200 mm high samples with a pore pressure transducer
placed along the bedding plane that had a diameter of
about 14 mm. The large diameter of the sensor restricted
the practicality of the approach as the specimens needed
to be bored from the base to the pre-determined slip
surface causing significant damage to the specimens. This
program had limitations due to the available sensor
technology at the time but did demonstrate the importance
of measuring the pore pressure within the shear zone using Figure 1. Modified triaxial cell assembly.
internal readings.
Another triaxial program with internal measurements
was trialed in 2022 on specimens of saturated loose and
dense sand by Liang et al. (2022). This program placed a
series of three metal tubes of 1.6 mm diameters inside of
the specimens that were connected to external MFOPS.
The metal tubes were then filled with low compressibility
piezometric fluid (mineral oil) to attempt to minimize
reading delay. This program showed that MFOPS provided
reliable measurements throughout shearing and that they
had a faster response rate than traditional external
sensors. It also demonstrated clear variations of the
incremental pore water pressure throughout testing which
showed that local stress variations can occur rapidly. The Figure 2. Triaxial base assembly. a) Base assembly with
main shortcoming of Liangs application is that the MFOPS specimen pedestal mounted. b) Specimen pedestal in two
are still a long distance away from the shear zone of the pieces. c) Specimen pedestal following epoxying of fiber
specimen and are reliant on a transfer of pressure from the sensor into assembly.
tip of the metal tubes to the sensors located at the outside
of the triaxial cell. Nonetheless, Liang et al. (2022) did
demonstrate that the advances in fiber-optic technology The MFOPS being utilized was originally developed by
can be used to further the methodology developed by Su OpSens for use in the medical field, specially to measure
(2005) to measure the internal pore pressure distribution pressure changes in coronary arteries. The MFOPS is an
during shearing. extrinsic single point diaphragm sensor based on micro-
electromechanical systems technology employing the
Fabry-Perot principles that measure a phase shift of white
3 TESTING APPARATUS light (Figure 3). The OpSens sensors were chosen based
on their robustness, ability to be sealed inside of a
For this program an existing triaxial assembly was specimen, commercial availability, reasonable cost, small
modified. The modifications involved designing a new base size (3.2 mm diameter), their advertised accuracy, and the
assembly that allowed a MFOPS to pass through the measurement range (0 to 6050 kPa).
specimen pedestal and extend from the bottom of the Testing was originally undertaken with a MFOPS that
specimen to the center. This included the design and did not have a pore stone protecting the diaphragm of the
construction of a new base plate, a mount that provided a sensor. This resulted in the diaphragm being displaced by
mechanical loading (soil pushing into the sensor) during
the consolidation and shearing phases of testing. This was OPA. As such, cylindrical heavily overconsolidated
resolved by working with OpSens to produce a new sensor specimens of kaolinite that had a diameter and height of
that has a casing and pore stone to protect the diaphragm 70 mm were used, which will also be the size of the
from any mechanical loading. To confirm this, an specimens used for the direct shear testing program.
unconfined compression test was completed on a lightly To prepare the specimens, dry kaolin powder is mixed
overconsolidated specimen of kaolinite. This test showed with deaired water until the predetermined liquid limit is
that the axial load being applied to the new sensor caused exceeded (water content ~45%). The prepared kaolin is
no mechanical response (Figure 4). then placed inside of a modified oedometer and is
consolidated in general accordance with ASTM
D2435/D2435M-11. The modified oedometer utilizes
building pressure and a difference in areas to apply a load
of up to 20 MPa to the specimen (Figure 5). Following the
final loading step at 20 MPa, the specimen is unloaded to
between 300 kPa and 400 kPa and left for 24 hours to allow
negative pore pressures to dissipate before removing the
specimen from the testing apparatus. This step is
completed to mitigate damage of the specimen from
excessive suction between removing the specimen from
Figure 3. MFOPS shown against pen for scale. the oedometer and loading it into the triaxial cell. Following
the 24 hour wait period, the specimen is removed from the
oedometer, taped with duct tape and two hose clamps are
used to apply a radial pressure to the specimen to protect
it from damage during the cutting and drilling process. The
specimen is then cut to length using a band saw and finally
a drill press is used to create a cavity for the 3.2 mm
diameter MFOPS to be placed inside. The cavity is drilled
35 mm into the specimen, allowing an approximate 2 mm
gap between the tip of the MFOPS and the depth the cavity
is drilled. An example specimen is shown in Figure 6.

Figure 4. Unconfined compression calibration test

examining mechanical response of MFOPS. a) Modified
OpSens sensor with casing and pore stone. b) Test results
showing change in pore pressure and deviator stress
versus axial strain.

4 SAMPLE PREPARATION

The goal of this portion of the research program is to

demonstrate that it is possible to gather local pore pressure
measurements at or near the shear zone of heavily
overconsolidated clay samples using MFOPS. Heavily Figure 5. High pressure oedometer cell.
overconsolidated clay samples are being used as a proxy
material for OPA during the initial proof of concept stage
which has an overconsolidation ratio of about 5 and is the Once the cavity is drilled the MFOPS is either placed
final target material. This will then allow the authors to directly inside the specimen and then loaded into the
apply the same testing principles inside of a novel direct triaxial cell or a bentonite grout mixture is used to fill the
shear testing apparatus that will eventually be used to test cavity before inserting the MFOPS inside. The grout
mixture acts as a seal like the method used for grouted-in load to the specimen, flooding the cell with water, applying
vibrating wire piezometers. a confining pressure, and finally increasing the cell
The kaolinite powder used has a specific gravity of 2.65 pressure in 50 kPa or 100 kPa increments until the upper
and a mineral content that is 97% kaolinite with primary limit of the cell was reached. Between each step, the back
chemical components composed of SiO2 (46%) and AL2O3 pressure line is kept open and a backpressure that is
(37%). At the completion of consolidation specimens 100 kPa less than the cell pressure is applied to avoid
typically have a void ratio between 0.55 and 0.58 (porosity damaging the specimen. Skempton’s pore pressure
between 35% and 37%) with a moisture content of about coefficient B checks were also completed prior to each
20%. In comparison, OPA from Mont Terri has a porosity increase in pressure to determine the saturation of the
between 12% and 18%, and a moisture content between specimen. The resultant Skempton’s B values are shown
5% and 9% (Minardi et al., 2019). on Figure 8.

Figure 6. Example of a heavily overconsolidated test

specimen. a) Specimen wrapped in duct tape with hose
clamps. b) Cavity in specimen after drilling. c) Example
specimen. d) Failure in specimen after shearing. e) Face of
failure plane post shearing

Figure 7. Example Calibration Plot of MFOPS

5 TRIAXIAL TESTING PROCEDURE AND RESULTS

In this paper, three separate tests on heavily

overconsolidated kaolinite specimens will be shown and
discussed. All triaxial tests consisted of consolidated
undrained tests that generally follow ASTM D4767-11. The
first test (a) does not have grout inside of the MFOPS cavity
and was sheared at a “normal” shearing rate. The second
test (b) had no grout inside of the MFOPS cavity and was
sheared at a “fast” strain rate. The final test (c) had
bentonite grout inside of the MFOPS cavity and was once
again sheared using a “fast” strain rate.

5.1 Calibration

Pore pressure sensor calibration was also completed

between each triaxial test to confirm the operability and
repeatability of results. Calibration testing consisted of
filling the triaxial cell with water (no specimen) and
increasing the cell pressure in a stepwise manner. The Figure 8. Skempton's pore pressure coefficient
response of the external sensors and MFOPS were then
recorded, compared, and a linear fitting method was used
to calibrate the MFOPS to the external sensors. An In general, the MFOPS is not saturated at the beginning
example plot from a calibration test is shown on Figure 7. of testing leading to a slower response to pore pressure
increases when compared to the external sensors. As the
5.2 Saturation MFOPS becomes saturated, faster responses and better
correlations to the external sensors are observed. The
The modified triaxial cell used for this testing program has results of the saturation stage for the three specimens are
a maximum cell pressure capacity of about 1 MPa, thus all also shown on Figure 9 in terms of pressure versus time.
testing completed is constrained to this upper limit. Due to the pressure limitation of the triaxial cell, B values
Saturation was typically completed by applying a seating
were generally between 0.6 and 0.7 prior to completing the Note that on Figure 9 a) the MFOPS is elevated
consolidation and shearing phases of testing. compared to the external sensors until a cell pressure of
about 800 kPa is applied to the specimen. A dissipation
type event then occurred, and the readings come in line
with the external sensors. The authors believe that a piece
of dry soil was lodged at the tip of the MFOPS and it wasn’t
until this elevated cell/back pressure was reached that the
pore pressures could properly circumvent this blockage.

5.3 Consolidation

For all specimens consolidation was completed with a cell

pressure of 1 MPa and a decrease in back pressure from
900 kPa to 800 kPa, resulting in an effective stress of
200 kPa. The triaxial cell has drainage at both ends of the
specimen and the volume change was measured based on
the change in fluid volume as measure by the back
pressure pump.
The results of the consolidation stage are shown on
Figure 10 in terms of pore water pressure and volume
change versus the square root of time. It can be observed
that the for the specimens with no grout (Figure 10 a) and
b)) that the pressure at the MFOPS immediately drops to
the back pressure applied indicating that a global reading
(same as the external sensors) is being measured. This is
likely due to the strength of the specimens where the cell
pressure load being applied is not sufficient to create a seal
between the MFOPS and the specimen.
The test with a bentonitic grout mixture (Figure 10 c))
shows that a local pore water pressure is being measured
by the MFOPS. The pore pressure near the center of the
specimen undergoes a dissipation event, where the
MFOPS has a higher pore pressure that eventually aligns
with the external sensors over time. This pore pressure
dissipation closely mimics the initial volume change
measured by the back pressure pump.

5.4 Shearing

As mentioned, two different strain rates were used as part

of this program. For consolidated undrained triaxial tests
strain rates slow enough to allow pore pressure
equalization across the specimen are typically chosen. The
value is determined based on the time to 100%
consolidation from the volumetric strain versus square root
of time plot. This was completed for the “normal” tests and
a strain rate of 0.0018 mm/min was used. For this program,
variations in pore pressures across the specimen are
required to observe if local pore pressures are being
measured by the MFOPS. As such, a “fast” strain rate
based on the time to 50% consolidation from the volumetric
strain versus square root of time plot was also used. The
“fast” strain rate used was 0.0091 mm/min.
The results of the shearing stage are shown in Figures
11 and 12 in terms of change in pore water pressure and
deviator stress versus axial strain. From the figures, it can
be observed for the specimens with no grout (a and b) that
the MFOPS generally follows the response of the external
Figure 9. Saturation plots. a) no grout and "normal" strain
sensors during the initial elastic loading phase. Once
rate specimen. b) no grout and "fast" strain rate specimen.
plastic damage begins to occur at around 0.1% axial strain,
c) grout and "fast" strain rate specimen.
the MFOPS is slightly elevated compared to the external immediately once the deviator stress is applied. The
sensors (in the range of 5 kPa) and slowly dissipates this MFOPS shows a higher pore pressure through the initial
excess pore pressure throughout shearing. elastic loading phase (ending at approximately 0.4% axial
strain), then begins to taper and reach peak at a lower axial
strain than the external sensors, potentially indicating the
presence damage in the form of micro-cracks.

Figure 10. Consolidation plots. a) no grout and "normal"

strain rate specimen. b) no grout and "fast" strain rate
specimen. c) grout and "fast" strain rate specimen.
Figure 11. Shearing plots. a) no grout and "normal" strain
rate specimen. b) no grout and "fast" strain rate specimen.
When examining the specimen with grout and a “fast”
c) grout and "fast" strain rate specimen.
strain rate (c in Figures 11 and 12), it can be observed that
the MFOPS and the external sensors deviate almost
 Button, E. a, & Blümel, M. (2002). Servo-Controlled Direct
Following the peak change of pore pressure, a rapid Shear Tests on Phyllites. 5th North American Rock
decrease at the location of the MFOPS occurs which is Mechanics Symposium.
expected based on the fundamental model for heavily https://online.tugraz.at/tug_online/voe_main2.getVol
overconsolidated specimens which indicates that internal lText?pDocumentNr=112165&pCurrPk=8593
dilation of the sample is taking place through microcracking Cheng, X. (2004). Localization in dutch dune sand and
and crack extension. The pore pressure near the center of organic clay. Delft.
the specimen remains below the external sensors until
about 1.8% axial strain which correlates well with the
beginning of the peak of the deviator stress. This is the
point where crack extension has reached a critical point
and specimen failure is about to occur.
After the peak deviator stress occurs, two new peaks in
pore pressure are observed by the MFOPS that are not
seen with the external sensors. These peaks potentially
represent dilation and contraction that is occurring during
shear along the failure surface. These results would be like
capturing the pore pressure response along the failure
plane of a landslide with a vibrating wire piezometer. With
further refining of the testing technique, these results could
provide valuable information about the pore pressure
response in the post peak environment.

6 CONCLUSIONS

The results of the initial triaxial testing on heavily

overconsolidated specimens of kaolinite showed that the
internal MFOPS correlate well with the external sensors
through calibration and testing. This demonstrates
repeatability and robustness. The testing completed with
a “fast” strain rate and grout also showed that the MFOPS
can obtain local pore pressure measurements near the
shear plane of the specimen accomplishing the primary
goal of the triaxial testing program.
Once refined, this information could potentially provide
additional insight into the initial elastic loading and post
peak effective stress path of heavily overconsolidated
argillaceous materials which could help better characterize
not only tunneling problems, but any problem where
shearing is occurring under saturated conditions.

7 ACKNOWLEDGEMENTS

This research is made possible through the support of

NWMO, NAGRA, the Natural Sciences and Engineering
Research Council of Canada, and the Geotechnical Center
at the University of Alberta. Special thanks to Hossein
Kasani (NWMO), Silvio Giger (NAGRA), Alessio Ferrari
(EPFL), and the team at the University of Alberta: Lucas
Duerksen, Vivian Giang, Jennifer Stogowski, Nathan
Deisman, and Jakob Brandl for their time and technical
support.

8 REFERENCES

Amann, F., Button, E. A., Evans, K. F., Gischig, V. S., &

Blümel, M. (2011). Experimental study of the brittle
behavior of clay shale in rapid unconfined
compression. Rock Mechanics and Rock Figure 12. Zoomed in Shearing plots. a) no grout and
Engineering, 44(4), 415–430. "normal" strain rate specimen. b) no grout and "fast" strain
https://doi.org/10.1007/s00603-011-0156-3 rate specimen. c) grout and "fast" strain rate specimen.
Di Donna, A., Ferrari, A., & Laloui, L. (2016). Experimental geomechanical properties under consideration of
investigations of the soil–concrete interface: Physical confinement and anisotropy. Engineering Geology,
mechanisms, cyclic mobilization, and behaviour at 237(January), 32–41.
different temperatures. Canadian Geotechnical https://doi.org/10.1016/j.enggeo.2018.02.012
Journal, 53(4), 659–672. https://doi.org/10.1139/cgj- Winhausen, L., Khaledi, K., Jalali, M., Urai, J. L., & Amann,
2015-0294 F. (2022). Failure mode transition in Opalinus Clay:
Gens, A. (2013). On the hydromechanical behaviour of a hydro-mechanical and microstructural perspective.
argillaceous hard soils-weak rocks. 15th European Solid Earth, 13(5), 901–915.
Conference on Soil Mechanics and Geotechnical https://doi.org/10.5194/se-13-901-2022
Engineering - Geotechnics of Hard Soils - Weak Yuan, J., Zhang, Q., Li, B., & Zhao, X. (2013). Experimental
Rocks (Part 4), 71–118. https://doi.org/10.3233/978- analysis of shear band formation in plane strain tests
1-61499-199-1-71 on Shanghai silty clay. Bulletin of Engineering
Giger, S. B., Marschall, P., Lanyon, B., & Martin, C. D. Geology and the Environment, 72(1), 107–114.
(2015). Hydro-mechanical response of Opalinus https://doi.org/10.1007/s10064-012-0453-4
Clay during excavation works - a synopsis from the
Mont Terri URL. Geomechanik Und Tunnelbau, 8(5),
421–425. https://doi.org/10.1002/geot.201500021
Liang, Y., Beier, N., & Sego, D. C. (2022). New method for
internal pore-water pressure measurements.
Geotechnical Testing Journal, 45(2), 490–502.
https://doi.org/10.1520/GTJ20210008
Minardi, A., Ferrari, A., & Laloui, L. (2019). Arbeitsbericht
NAB 19-18.
National Waste Management Organization. (2013).
Adaptive Phased Management.
Poturovic, S., Schubert, W., & Blümel, M. (2015).
Comparison of constant normal load (CNL) and
constant normal stiffness (CNS) direct shear tests.
EUROCK 2015 & 6th Geomechanical Colloquim,
445–450.
Schrefler, B. A., Zhang, H. W., Pastor, M., & Zienkiewicz,
O. C. (1998). Strain localisation modelling and pore
pressure in saturated sand samples. Computational
Mechanics, 22(3), 266–280.
https://doi.org/10.1007/s004660050359
Shuttle, D. A., & Smith, I. M. (1990). LOCALISATION IN
THE PRESENCE OF EXCESS POREWATER
PRESSURE. Computers and Geotechnics, 9, 87–
99.
Su, X. (2005). Pore Pressure Response and Shear
Behavior of Shear Zones in Weak Rocks Near Fort
McMurray. University of Alberta.
Thakur, V. (2007). Strain localization in sensitive soft clays.
Norwegian University of Science and Technology,
Trondheim.
Thakur, V. (2011). Numerically observed shear bands in
soft sensitive clays. Geomechanics and
Geoengineering, 6(2), 131–146.
https://doi.org/10.1080/17486025.2010.546434
Thakur, V., Nordal, S., Viggiani, G., & Charrier, P. (2018).
Shear bands in undrained plane strain compression
of Norwegian quick clays. Canadian Geotechnical
Journal, 55(1), 45–56. https://doi.org/10.1139/cgj-
2016-0443
Viggiani, G., Finno, R. J., & Harris, W. W. (1994).
Experimental observations of strain localisation in
plane strain compression of a stiff clay. Localisation
and Bifurcation Theory for Soils and Rocks, July,
189–198.
Wild, K. M., & Amann, F. (2018). Experimental study of the
hydro-mechanical response of Opalinus Clay – Part
1: Pore pressure response and effective
Simulation of solid-fluid interaction in
granular flow of irregularly shaped particles
via coupled LBM-DEM method
Mohammad Hassan Ahmadian, Wenbo Zheng
School of Engineering – University of Northern British Columbia,
Prince George, British Columbia, Canada

ABSTRACT
Simulating granular flow in geotechnical engineering is mostly associated with adopting a coupling method to simulate the
interaction of solid grains and seeping fluid, such as using computation fluid dynamics to couple with the discrete element
method. While this approach has enhanced our knowledge about the underlying mechanisms that trigger geohazards like
landslides and internal erosion, most studies simplified soil particles as spheres and the morphological impact of particle’s
shape is neglected. In this study, the lattice Boltzmann method (LBM) and the discrete element modeling (DEM) are
coupled to simulate the interaction between irregular soil grains and fluid by using two open-source codes. For this purpose,
the irregular shaped particles are represented by multi-sphere clumps in DEM, and LBM is used to account for the fluid
part interaction at meso-scale. The accuracy of the coupled method is demonstrated by simulating several benchmark
cases and comparing them with the results from literature. The coupled LBM-DEM method is then used to simulate the
collapse and transport of submerged granular particles to demonstrate the influence of grain morphology in the solid-fluid
interaction.

RÉSUMÉ
La simulation de l'écoulement granulaire en génie géotechnique est principalement associée à l'adoption d'une méthode
de couplage pour simuler l'interaction entre les grains solides et le fluide qui s'infiltre, comme l'utilisation de la dynamique
des fluides numérique couplée à la méthode des éléments discrets. Bien que cette approche ait amélioré notre
compréhension des mécanismes sous-jacents qui déclenchent des géorisques tels que les glissements de terrain et
l'érosion interne, la plupart des études simplifient les particules du sol en tant que sphères et l'impact morphologique de la
forme des particules est négligé. Dans cette étude, la méthode de Boltzmann sur réseau (LBM) et la modélisation par
éléments discrets (DEM) sont couplées pour simuler l'interaction entre des grains de sol irréguliers et un fluide en utilisant
deux codes open source. À cet effet, les particules de forme irrégulière sont représentées par des agrégats de sphères
multiples dans le DEM, et le LBM est utilisé pour tenir compte de l'interaction avec la partie fluide à l'échelle mésoscopique.
L'exactitude de la méthode couplée est démontrée en simulant plusieurs cas de référence et en les comparant aux
résultats de la littérature. La méthode couplée LBM-DEM est ensuite utilisée pour simuler l'effondrement et le transport de
particules granulaires immergées afin de mettre en évidence l'influence de la morphologie des grains dans l'interaction
solide-fluide.

1 INTRODUCTION numerical methods that accounts for both phases of solid

and fluid. For the solid phase, discrete element modeling
Understanding fluid-solid interaction is essential in (DEM) is the particle-based method that is often used to
geotechnical engineering since it provides insight into simulate a system of granular material. However, for the
some geological phenomena such as landslides, internal fluid phase there are a number of fluid flow simulation
erosion, or debris flow. Such disastrous events usually methods that can be categorized as unresolved and
occur due to fast transport of densely packed granular resolved computational fluid dynamics (CFD). Within the
particles that was initiated by a seepage flow within resolved computational fluid dynamics, lattice Boltzmann
granular system (Yang et al. 2019). Therefore, studying the method (LBM) is one of the most accurate ones. While a
behavior of such a system requires considering both LBM simulation is computationally expensive, LBM
phases of solid particles as well as fluid phase. The benefits from simplified governing equations that lead to
fundamental and well-known analytical equations in excellent performance in parallel computing in addition to
studying granular flows are Darcy’s law and Ergun its fine-grid accurate results (Yang et al., 2019). Therefore,
equation (Yang et al., 2019). Although both equations are compare to unresolved CFD, LBM is an efficient technique
validated by numerous experiments, the macroscopic as long as the focus of the study is on simulating a
approach provides no information about the underlying reasonably large-scale flow with high accuracy (Yang et al.,
mechanisms involved in granular flows. In that sense, it is 2019).
important to look at granular flow from a microscopic In this paper, we developed an LBM-DEM coupling
perspective. framework to simulate flow containing irregularly shaped
Literature contains several numerical methods that are particles (ISP). For that purpose, we utilized a MATLAB
capable of simulating granular flow at microscale (Shen et code that fills the volume of an ISP with overlapping sub-
al. 2022). The modelling usually involves coupling two particles to form a clump (Angelidakis et al. 2021). Then
three benchmark cases were considered to validate our forces are also calculated based on a spring-dashpot
coupling models with clumps. Finally, to study the impact model as follows (Yang et al., 2019):
of real particle morphology in granular flow simulations, we
modeled a granular column of ISP immersed in a fluid.
Afterward, the evolution of column collapse within fluid is 𝐅𝑛 = 𝑘𝑛 𝛿𝑛 + 𝑐𝑛 ∆𝐮𝑛 [5]
simulated and compared to the same case with regular
shape particles (sphere). The comparison shows the
importance of particle shape in granular flow simulation. 𝑡
𝐅𝑡 = −𝑘𝑡 ∫𝑡 𝑐,0 ∆𝐮𝑡 𝑑𝑡 − 𝑐𝑡 ∆𝐮𝑡 [6]
We consider this research to be a step towards studying 𝑐

more realistic simulations of granular flows in geotechnical

In the above equations 𝐅 is the force vector, 𝑘 is the
engineering.
spring stiffness, c is damping coefficient, δ is the amount
that two interacting particles overlap each other (measured
2 METHODOLOGY based on distance of two interacting particles from each
other). Also the index n and t refer to normal and tangential
LBM is a mesoscopic scale fluid flow simulation method directions. It should be mentioned that in this study, the
based on kinetic theory. In this method the computational Hertz tangential model is adopted for particle-particle and
domain divides into structured Cartesian lattice nodes and particle-wall interaction. By considering all forces (contact
then a probability distribution function (PDF) is defined on 𝐅c , gravity 𝐆, fluid drag forces 𝐅f ) acting on the particle, the
velocity of particles can be obtained using second Law of
each node of the lattice. In other words, fi (𝐱, t) is the PDF Newton (Zhang et al. 2021):
of the point 𝐱 at time 𝑡 at direction 𝑖 . By applying BGK
approximation in Boltzmann equation, the governing
equation of fluid flow at mesoscale can be written as 𝑚𝐚 = 𝐅𝑐 + 𝐆 + 𝐅𝑓 [7]
follows (Yang et al., 2019):

1 eq 𝐼𝛚̇ = 𝐓𝑐 + 𝐓𝑓 [8]
fi (𝐱 + 𝐜i δt , t + δt ) − fi (𝐱, t) = − [fi (𝐱, t) − fi (𝐱, t)] [1]
τ

To simulate solid particles with irregular shapes, a

eq 𝐜i .𝐮f (𝐜i .𝐮f )2 uf 2 multi-sphere approach is used. In this approach, each
fi = wi ρf [1 + + − ] [2]
c2s 2c4s 2c2s individual solid particle is represented as a collection of
smaller spheres, which are bonded together (Shen et al.
2022). As a result, each ISP also known as clump consists
where in these equations, 𝑤𝑖 is weighting factor of the of a number of sub-particles as spheres. In this paper, we
eq
lattice structure, τ is the relaxation time, fi is equilibrium utilized a MATLAB code that fills the volume of an ISP with
several sub-particles bonded together to form a clump
distribution function, and cs = 1⁄√3 is the speed of sound (Angelidakis et al. 2021). Therefore, the outcome of the
in lattice unit. After calculating fi at each node, the fluid code is a list of coordinates and radii of sub-particles that
density ρf and velocity 𝐮f can be reconstructed using DEM gets to simulate the clump using the Liggghts multi-
sphere model.
following equations (Yang et al., 2019):

𝜌𝑓 = ∑18
𝑖=0 𝑓𝑖 [3]

𝜌𝑓 𝐮𝑓 = ∑18
𝑖=0 𝒄𝑖 𝑓𝑖 [4]

The evolution of the PDF is then modeled using two

main steps: collision and propagation. By iterating the
collision and propagation steps, the LBM can simulate at a
mesoscopic scale, allowing for the simulation of complex
fluid flows like granular flows.
Moreover, DEM is adopted to simulate solid particles. Figure 1. The way IMB accounts for solid-fluid interface
According to the model, the solid particles are regarded as
rigid bodies following Newton’s second law and the
particles interact with one another through a set of contact For LBM and DEM to work together, it is essential to
forces (𝐅c = 𝐅n + 𝐅t ), which depend on the properties of the use a coupling framework and the immersed moving
particles and the nature of their interactions. The contact boundary (IMB) method is used in this paper. According to
the IMB method, each computational cell divides to some
sub-cells (see Figure 1) and then depend on whether a 3.1 Settling of single clump
sub-cell is solid, partially-, or fully-saturated fluid, the
collision operator on the right-hand side of Equation 1 will For the first benchmark case, settling of a heavy single
be determined (Noble and Torczynski 1998). For further particle in an ambient fluid is simulated (Yang et al. 2019).
information, Figure 2 illustrates the flowchart of the LBM- As mentioned before, the single spherical particle is
IMB-DEM coupling framework. regenerated using 100 sub-particles as clump. The
simulation domain as shown in Figure 3a. The diameter of
the clump is 15 mm and its density is 1120 𝑘𝑔⁄𝑚3 .
For this simulation, the lattice resolution and relaxation
times are 20 and 0.56 respectively. By considering
diameter of particle as length scale and particle’s terminal
velocity as characteristic velocity, the simulation was
performed for four Reynolds numbers (Re = 1.48, 4.1, 11.6,
31.9). Then the results were compared to experimental
data (ten Cate, Nieuwstad, Derksen, & Van den Akker,
2002) as shown below.

-0.02

Setteling velocity (m/s)

-0.04
Figure 2. Flowchart of LBM-IMB-DEM coupling framework, Re = 1.48 Sim
modified from Yang et al. (2019) -0.06 Re = 1.48 Exp
Re = 4.1 Sim
-0.08 Re = 4.1 Exp
3 VALIDATION CASES Re = 11.6 Sim
-0.1
Re = 11.6 Exp
This study was conducted to simulate granular flow using Re = 31.9 Sim
-0.12
LBM-DEM. For this purpose, a C++ code is used. The code Re = 31.9 Exp
uses Palabos (LBM open source library) for LBM -0.14
simulation, and LIGGGHTS as an external library for DEM 0 1 2 3 4 5
simulation (Seil and Pirker 2017). To validate the current Time, t (s)
code, three benchmark cases (Figure 3) were studied to Figure 4. Comparison of current simulation and
demonstrate the accuracy of current LBM-DEM model experimental results of a single clump settling in ambient
using clumps. The three benchmark cases are 1) settling fluid
of single spherical clump, 2) settling of two spherical
clumps and 3) settling of a disk-shaped clump all in an
ambient fluid. It is worth noting that all 3D LBM simulations
conducted in this study used D3Q19 lattice structure.

(a) t = 0.72 s (b) t = 1.48 s

Figure 5. Z velocity contour of a clump at different times
(Re = 11.6)

(a) (b) (c)

Figure 3. Schematic of benchmark case domains a) single According to Figure 4, it can be seen that for low
clump b) two clumps c) disk shape clump. All boundaries Reynolds number, the velocity of particle first increases
are considered as no slip wall. until reached its terminal velocity where the buoyancy force
is equal to gravity. Then as the particle gets closer to the
bottom, the velocity gradually decreases. However, for the
higher Reynolds number the same process only happens
faster. In addition, Figure 4 demonstrates that even though
the spherical particle is simulated by clump, the settling
velocity at each Reynolds number is in good agreement
with the experimental results. Figure 5 also provides the Z
velocity contour of the x-plane that is located at the middle
of the tank. The figure depicts that as the particle gains
more speed, the wake that occurs behind the particle also
gets bigger.

3.2 Settling of two clumps

In the second benchmark validation, the interaction

between two clumps falling from a certain height is
investigated. Figure 3b shows the dimensions of the
domain. In this case, the density of particles is 1350 𝑘𝑔⁄𝑚3 (a) t = 0 s (b) t = 1.2 s
with the diameter equal to 12.7 mm. For this simulation, the
lattice resolution and relaxation times are 10 and 0.52
respectively. In addition, fluid density and viscosity are
1195 𝑘𝑔⁄𝑚3 and 0.0305 𝑃𝑎. 𝑠 respectively. Figure 6 shows
our simulation results compare to another simulation by
Dash and Lee (Dash and Lee 2015). The following
equations are used to non-dimensionalize the time and
velocity in Figure 6.

t c = √dp /|ρb ⁄ρf − 1|g [9]

Uc = √|ρb ⁄ρf − 1|gdp [10] (c) t = 1.44 s (d) t = 2.24 s

Figure 7. The location of two clumps at different times

As shown in figure 6, when two particles start to fall

3.3 Settling of disk-shape clump
(Figure 7a), they initially have same speed, but after some
time the trailing particle gets more speed since it is located
The last simulated benchmark case is settling of a disk
at the wake of the leading particle also called drafting
shape clump. Same as before, the shape of the disk is
(Figure 7b). Then the particles start to contact each other
generated using a clump of 100 sub-particles. The lattice
(kissing, Figure 7c) until they begin to tumble and separate
resolution and relaxation times are 15 and 0.55
at the later stage (Figure 7d). While there is slight
respectively. Furthermore, the initial orientation of the disk
discrepancy between the two results, figure 6 shows that
our simulation of clump particles can follow all three stages is set to 45° as shown in figure 3c. According to literature,
of drafting, kissing, and tumbling which are associated with the settling process of disk-shaped particle depends on the
this benchmark. dimension of the disk (Willmarth et al. 1964). Therefore, we
used this point to bring another validation to our code.
There are two non-dimensional numbers that
characterize the settling process of a disk-shaped particle,
0
Leading particle (current simulation) Reynolds number and non-dimensional moment of inertia
-0.2
with following definition.
Trailing particle (current simulation)

-0.4
Trailing particle (Dash & Lee)
πρp t
Leading particle (Dash & Lee) I∗ = [11]
U/Uc

64ρf d
-0.6

-0.8
where t is the thickness of disk. (Willmarth et al., 1964)
-1 presents that the falling of disc in fluid depends on the value
of the two numbers (𝐼∗ and Re) could have either stable or
-1.2 unstable falling patterns. Therefore, we specified two cases
0 5 10 15 20 25 30 35 where in each case the property of fluid as well as disk are
t/tc set in a way that, based on (Willmarth et al., 1964) in one
Figure 6. Comparison of settling velocity of two particle case (case 1), the disc should have stable falling, while in
between our simulation and (Dash & Lee 2015) another case (case 2) the falling should undergo a periodic
oscillation. Table 1 provides the value of the non-
dimensional numbers for each case.
Table 1. Non-dimensional number values for two cases
related to disk-shaped particle benchmark.

𝐼∗ 𝑈𝑧,𝑚𝑒𝑎𝑛 (𝑚/𝑠) 𝑅𝑒𝑝

Case 1 0.0191 0.021 21

Case 2 0.0398 0.1087 267

In addition, Table 2 provides the characteristics of the

domain for two cases. It should be mentioned that in both (a) t = 0.14 s (b) t = 0.28 s (c) t = 0.38 s
cases, the thickness of disk is 0.3 times the diameter. Figure 9. Z-velocity contour of settling disk-shaped clump
at different times (Case 2).
Table 2. Characteristics of two cases related to disk-
shaped particle benchmark. 4 IMMERSED GRANULAR COLLPASE

𝜌𝑝 (𝑘𝑔 ⁄𝑚3 ) d (mm) Box size (mm) In this section, we presented the simulation of column of
irregular shape particles collapsing in an ambient fluid. The
Case 1 1300 1 mm 6×6×15 dimension of domain is given in Figure 10. The properties
Case 2 2700 2 mm 12×12×48 of fluid and solid particles are listed in Table 3. The ISP
used in this simulation are regenerated using a clump of
100 sub-particles. The lattice resolution and relaxation
times are 10 and 0.502 respectively. To generate the
Looking at Figure 8 shows that our simulation results column of 150 particles, first particles stacked at one side
generates the same pattern predicted by Willmarth et al. of the domain using only DEM method (𝐿𝑖 region).
(1964). In other words, our results correctly simulate the
falling of disk-shape particles in two different cases.
Moreover, Figure 9 provides further illustration of a disk Table 3. Characteristics of fluid and solid
falling with oscillating pattern in the fluid domain.
Parameter Value
Particle irregular shape particle diameter, 𝑑𝑝 1 mm
0

-0.02 regular shape particle diameter, 𝑑𝑝 0.8 mm

Settling velocity (m/s)

-0.04 Density, 𝜌𝑝 2468 𝑘𝑔 ⁄𝑚3

Fluid Density, 𝜌𝑓 1000 𝑘𝑔 ⁄𝑚3
-0.06
Case 1 Viscosity, 𝜇𝑓 0.001 𝑃𝑎. 𝑠
-0.08
Case 2
-0.1

-0.12

-0.14
0 0.2 0.4 0.6
Time (s)
Figure 8. Settling velocity of disk-shaped particle in two
different cases.

Figure 10. Domain dimensions for the case of immersed

column collapse with 150 particles
 By removing the wall that holds particles to one side, As a demonstration of the influence of particle shape on
the particles start to collapse to the other side of the domain the simulated results, the simulation is performed for the
(Figure 11). Collapsing of particles changed the fluid same number of particles but using spherical shape
velocity field and higher velocities were found near the particles. By comparing two simulations together (Figure
surface of the granular flow, especially near the toe of the 12a & 12b), the impact of particles’ shape on the simulation
granular when the particles originally in the upper right of outcome can be appreciated. A few clump particles
the model collapsed to the bottom boundary. travelled a greater distance than those in the spherical
particle model. Also, the slope and surface topography of
the clump particle model seems to be more realistic, while
the spherical particle model has a very high column of
particle near the wall boundary in the back. Currently, we
are systematically investigating the influence of particle
shapes on travel distance of granular flows and the
(a) resulting slope profile, for which we plan to present in
another manuscript.

5 CONCLUSION

In this paper we have studied the impact of particle shape

in granular flow using LBM-DEM coupling method. First the
irregular particle shapes are generated as clumps. Then
the simulation method is validated using three benchmark
(b) cases from literature. Finally, the process of collapsing a
column of irregularly shaped particles was simulated and
the simulation outcome was compared with the same case
but with particles with spherical shape. In conclusion, our
results show that there is good potential in LBM-DEM to
Figure 11. Column collapse at different times a) t = 0.06 s simulate particles with more realistic shape, and that the
b) t = 0.12 s shape of particles has considerable impact on simulation
outcome.

6 REFERENCES

Angelidakis, V., Nadimi, S., Otsubo, M. and Utili, S. 2021.

CLUMP: A Code Library to generate Universal
Multi-sphere Particles. SoftwareX, 15: 100735.
(a) Dash, S.M. and Lee, T.S. 2015. Two spheres
sedimentation dynamics in a viscous liquid
column. Computers & Fluids, 123: 218-234.
Noble, D.R. and Torczynski, J.R. 1998. A Lattice-
Boltzmann Method for Partially Saturated
Computational Cells. International Journal of
Modern Physics C, 09(08): 1189-1201.
Seil, P. and Pirker, S. 2017. LBDEMcoupling: Open-Source
Power for Fluid-Particle Systems. Proceedings of
the 7th International Conference on Discrete
Element Methods, Singapore.
Shen, Z., Wang, G., Huang, D. and Jin, F. 2022. A resolved
CFD-DEM coupling model for modeling two-
phase fluids interaction with irregularly shaped
particles. Journal of Computational Physics, 448:
(b)
110695.
ten Cate, A., Nieuwstad, C.H., Derksen, J.J. and Van den
Akker, H.E.A. 2002. Particle imaging velocimetry
experiments and lattice-Boltzmann simulations on
a single sphere settling under gravity. Physics of
Fluids, 14(11): 4012-4025.
Willmarth, W.W., Hawk, N.E. and Harvey, R.L. 1964.
Steady and Unsteady Motions and Wakes of
Figure 12. Final deposition of column of particles a)
irregularly shaped particles, b) spherical particles
 Freely Falling Disks. The Physics of Fluids, 7(2):
197-208.
Yang, G.C., Jing, L., Kwok, C.Y. and Sobral, Y.D. 2019. A
comprehensive parametric study of LBM-DEM for
immersed granular flows. Computers and
Geotechnics, 114: 103100.
Zhang, R., Kim, H.J. and Dinoy, P.R. 2021. Particle Flow
Simulation Based on Hybrid IMB-DEM-LBM
Approach with New Solid Fraction Calculation
Scheme. Applied Sciences, 11(8): 3436.
Tuesday, October 3, 2023

CONTAMINANT
TRANSPORT
Linking soil-freezing characteristics to
bioremediation for contaminated cold-
climate soil: Review of recent progress
Wonjae Chang, Jihun Kim and Tasnim Nayeema
Department of Civil, Geological and Environmental Engineering –
University of Saskatchewan, Saskatoon, SK, Canada

ABSTRACT

Soil is a complex material in which physical, chemical, and biological processes are occurring. Linking physical and
chemical soil data to biological data can reveal the underlying interactive mechanisms that drive soil microbial responses.
Seasonal soil freezing is an excellent example of such soil processes interacting closely. This study reviews the
accumulated outcomes of recent bioremediation research into how soil-freezing characteristic curves (SFCCs) are related
to soil biological data in hydrocarbon-contaminated soils in cold climates. Soil biostimulation treatments change the SFCCs
of petroleum hydrocarbon-contaminated site soils and the conventional physical soil parameters of the SFCC (θ = αTβ)
can be effectively related to the feasibility of cold-tolerant bioremediation and its optimization. This finding enabled us to
optimize unfrozen water retention in favour of cold-tolerant passive bioremediation extended to seasonal freezing
conditions, beyond the short active treatment season, resulting in significant hydrocarbon biodegradation in freezing and
frozen soil. Furthermore, linking soil-freezing data and microbial data led to the development of an integrated soil
respiration and thermal modelling tool that is useful for optimizing soil treatments for cold-tolerant bioremediation in remote
cold regions, prior to site remediation.

RÉSUMÉ

Les sols contaminés sont des matériaux complexes dans lesquels se produisent des réactions physiques, chimiques et
biologiques. Établir un lien entre les propriétés physiques et chimiques d'un sol et les processus biologiques permet
d'élucider les mécanismes fondamentaux et interactifs qui déterminent les changements dans l'activité microbienne dans
le sol. Le gel saisonnier d'un sol est un excellent exemple d'une étroite interaction entre ces différents processus. Cette
étude passe en révision les conclusions de récentes études en bioremédiation sur la connexion entre la courbe
caractéristique de gel du sol (CCGS) et les caractéristiques biologiques des sols contaminés par les hydrocarbures
pétroliers dans les climats froids. Les traitements visant la biostimulation de tels sols contaminés modifient les CCGS (θ =
αTβ) et les simples paramètres physiques formant cette courbe peuvent être liés à la faisabilité et à l'optimisation des
stratégies de bioremédiation tolérante au froid pour ces sols. Cette astuce nous a permis d'optimiser la rétention d'eau
non congelée pendant le gel saisonnier d'un sol contaminé, en faveur d'une biodégradation passive d'hydrocarbures plus
active et tolérante au froid, étendue aux conditions hivernales. Cette optimisation de la rétention d'eau non congelée a
entraîné une biodégradation importante d'hydrocarbures au cours de l'hiver dans le sol traité. En outre, l'établissement
d'un lien entre la courbe caractéristique de gel du sol et les données microbiennes nous a permis de mettre au point un
outil intégré de modélisation thermique et de respiration microbienne qui est utile pour optimiser les traitements de
biostimulation pour les sols dans les régions froides éloignées, sans devoir débuter la bioremédiation sur site.

1 INTRODUCTION degrading bacteria). Changes in SFCCs affected by

biostimulation treatments for microbial enhancement have
The soil-freezing characteristic curve (SFCC) is not been extensively studied. It is important to accumulate
considered a useful tool for understanding soil freezing data and knowledge about shifting SFCCs of various
behaviour, predominately in geotechnical engineering for contaminated soils and treated soils to cost-effectively
designing and managing infrastructure in cold regions. manage and remediate them at remote cold sites in
Recent research indicates that the SFCC is also useful for Canada. Thus far, the relationship between SFCCs and
developing bioremediation strategies for cold-climate sites. soil biological data has not received sufficient attention in
Briefly, the soil thermal phase change from unfrozen to the literature. This study reports a link between the SFCC
frozen causes the redistribution of ions in soil pores, as and soil biological data by reviewing the accumulated
some are excluded from pore ice and become outcomes of recent bioremediation research for cold-
concentrated in pore water of reduced volume (i.e., climate site soils.
unfrozen water films). Meanwhile, seasonal shifts in
microbial community compositions occur. Changes in the 2 SOIL MICROENVIRONMENT DURING FREEZING
natural setting at the pore scale are related to the
emergence of winter microbial populations. This provides When soil freezes, the availability of bulk unfrozen
insight on how to leverage cold-loving microorganisms in water to microorganisms decreases, which drastically
cold climates (e.g., freezing-tolerant hydrocarbon- influences intrinsic microbial activity (Öquist et al. 2009;
Sparrman et al. 2004). Unfrozen water content is a critical dependent on soil surface area, particle size, and pore
requirement for cold-adapted microorganisms to survive in size, and it decreases with decreasing temperature (Wen
cold-climate soils. Change in unfrozen water content is et al. 2012). A decreasing matric potential increases the
correlated to microbial metabolic activity (Öquist et al. tendency for water to adsorb onto the soil surface and
2009; Panikov et al. 2006; Rivkina et al. 2000). In frozen move into small pores and capillaries in soil particles (Wen
soils, unfrozen water is commonly present in form of thin et al. 2012). Clays with high surface areas and small
water films in the soil matrix. It was reported that unfrozen particle sizes retain more unfrozen water content than
water film thickness is related to surface area, adsorptive sandy soil (Ma et al. 2017; Ren et al. 2017).
and capillary forces (Sparrman et al. 2004), and the
presence of organic matter and petroleum hydrocarbons 3 MICROBIAL ACTIVITY IN FREEZING SOILS
(Drotz et al. 2010; Siciliano et al. 2008).
Changes in unfrozen water content influence Liquid unfrozen water is a prerequisite for the
osmotic and matric potential in soil microenvironment transport of substrates, nutrients, oxygen, and by-products
(Drotz et al. 2009). Osmotic potential is related to soil (Harvey et al. 2012; Öquist et al. 2009; Tucker 2014).
freezing and the associated formation of unfrozen water Eutectophiles have been observed at the interface
films. It is also related to dissolved solute concentrations between water and ice (Deming 2002). Cold-adapted
(Bing and Ma 2011). When solute concentrations increase bacteria survive and remain active in freezing and frozen
in soil pore water, the freezing point of a soil solution soils due to an anti-freezing adaptation, which inhibits
decreases (Bing and Ma 2011; Han et al. 2018), and soil intracellular ice formation with anti-freeze proteins (Bore et
freezing is delayed due to elevated solute concentrations al. 2017). Previously reported evaluations of microbial
(Ma et al. 2017; Wu et al. 2015). During soil freezing at less activity in freezing/frozen soils are summarized in Table 1.
than -3 °C/day, which is slow enough to mimic seasonal
freezing, solutes are excluded from pore ice and Table 1. Reported microbial activity measurements in
redistributed to unfrozen water films (Konrad and freezing and frozen soils (from Kim et al. 2021).
McCammon 1990). Figure 1 illustrates a conceptual
microenvironment within hydrocarbon-contaminated soils
subjected to seasonal freezing, where unfrozen liquid Temperatures Substrate Evaluation Reference
water and pore ice coexist.
Dynamic from C-labeled
14
Rivkina et al.
C-CO2
14
5 to -20 °C acetate (2000)

Dynamic from C-labeled

14
Panikov et al.
C-CO2
14
10 to -40 °C glucose (2006)

Soil organic Öquist et al.

Fixed at -4 °C CO2 production
matter (2009)

Glucose
From 10 to -8 °C Tilston et al.
and CO2 production
(2010))
sucrose

Figure 1. Thin unfrozen water film in seasonally freezing From 0, -3, -6, -9, C-labeled
13 13
C-CO2 Tuorto et al.
and frozen hydrocarbon-contaminated soil where -12 to -20 °C acetate DNA replication (2014)
hydrocarbon-degrading bacteria survive. Seasonal
freezing leads to the expulsion of solutes from pore ice 13
C-CO2
and their accumulation in the unfrozen water film (from Fixed at 5, -5, C-labeled
13
Bore et al.
Metabolic
Kim et al. 2018. Reproduced by permission of Sci. Total and -20 °C glucose
transformation
(2017)
Environ., Elsevier)

In freezing hydrocarbon-contaminated soils, low C-labeled

13 13
C-CO2 Segura et al.
Fixed at -4 °C
temperatures increase the viscosity of petroleum cellulose Cell growth (2017)
hydrocarbons and decrease their volatility and solubility
(Atlas 1981). Dissolved hydrocarbons can also be
excluded from ice during slow freezing and become
13
C-CO2
Fixed at 5, -5, C-labeled
13
Bore et al.
Metabolic
concentrated in the remaining unfrozen water (Barnes et and -20 °C glucose
transformation
(2019)
al. 2004). When hydrocarbon concentrations reach
solubility, excess hydrocarbons are excluded from water as 13
C-CO2
non-aqueous liquids (NAPLs) (Barnes et al. 2004). The Fixed at -3 C-labeled
13
Anabolic and Segura et al.
formation of NAPLs with high toxicity may affect microbial and -5 °C glucose catabolic (2019)
activity and retard the onset of winter microbial activity activity
(Margesin and Schinner 2001). Matric potential is
 Various substrates that support microbial activity Chang et al. (2011) reported the emergence in
in freezing and frozen soils have been examined. Tuorto et partially frozen soils of Corynebacterineae-related bacterial
al. (2014) showed DNA replication at temperatures down populations, which are hydrocarbon degraders, indicated
to -20 °C using 13C-labeled acetate. Microbial metabolic by an increase in the copy numbers of alkB genes
transformation has been observed at -5 and -20 °C using responsible for hydrocarbon biodegradation in the areas
13C-labeled glucose (Bore et al. 2019; Bore et al. 2017). where unfrozen water was still available. Biodegradation
Cell growth has been detected at a fixed temperature rate constants at sub-zero temperatures were also
of -4 °C using 13C-labeled cellulose (Segura et al. 2017). investigated. Chang et al. (2011) applied seasonal freezing
Catabolic and anabolic activity were also detected at fixed and thawing temperatures to hydrocarbon-contaminated
temperatures of -3 and -5 °C using 13C-labeled glucose soils and reported TPH degradation rate constants of
(Segura et al. 2019). 0.0046 to 0.017 d-1 when soil temperatures ranged
Changes in microbial activity are related to from -5 to 10 °C. At a fixed temperature of -5 °C, Karppinen
changes in unfrozen water content. Rivkina et al. (2000) et al. (2017b) showed that the degradation rate constant
reported that the temporal decrease in unfrozen water film for F3 hydrocarbons (C16–C34) was 0.0036 d-1. As
thickness during soil freezing is significantly correlated to summarized in Table 2, hydrocarbon degradation in
the gradual depression of microbial metabolic activity as freezing and frozen soils is still comparable to respiration
tracked by 14C-radiolabeled acetate. Panikov et al. (2006) at positive temperatures.
described how soil respiration (CO2 production) is related
to unfrozen water content in thin unfrozen water films at 4 MODIFIED SFCC OF HYDROCARBON-
temperatures down to -39 °C. Öquist et al. (2009) also CONTAMINATED SOILS TREATED FOR
demonstrated that CO2 production was related to unfrozen BIOREMEDIATION
water content when the temperature was fixed at -4 °C.
Tilston et al. (2010) reported that the inflection point (rapid The soil freezing characteristic curve (SFCC)
change) in soil respiration occurs at around -2 °C, below describes the relationship between unfrozen water content
0 °C, and might be more closely correlated with changes in and sub-zero temperature (Anderson and Tice, 1972;
unfrozen water content than in soil temperature. Azmatch et al., 2012; Koopmans and Miller, 1966; Spaans
and Baker, 1996; Watanabe and Osada, 2017). It is
empirically expressed by Equation 1, where θ is unfrozen
Table 2. Previously reported hydrocarbon biodegradation water content (m3/m3), α and β are curve-fitting parameters
rates (from Kim et al. 2021). obtained from non-linear regression, and T is the absolute
value of soil temperatures below 0 °C (Andersland and
Ladanyi, 2003).
Study
Temperature Rate constant Reference
type
𝜃 = 𝛼𝑇𝛽 [1]
Variable from - 12.9 mg Eriksson et al.
Lab
5 to 7 °C TPH/kg/day (2001)
Equation 1 predicts unfrozen water content in
TPH: 0.022 to Zytner et al.
freezing soils to a fair degree of accuracy considering that
Field Variable overburden pressure and osmotic potential are taken to be
0.0043 d-1 (2001)
constant (Anderson and Tice, 1972). The SFCC has also
been formulated as an exponential function (McKenzie et
Variable from 90 mg
Thomassin- al., 2007) and non-linear function (Kozlowski, 2007). The
Field Lacroix et al. SFCC describes the soil ice content at specific soil
0 to 10 °C TPH/kg/day
(2002) temperatures for a particular soil type (Devoie et al., 2022)
and is a core tool in cold-region engineering (civil
TPH: 0.015 to Paudyn et al. infrastructure, pipelines, foundations, pavements, etc.)
Field Variable (Ren et al., 2018).
0.026 d-1 (2008)
Ice and unfrozen water coexist in frozen soils and
their relative proportions are dependent on temperature
Variable from - Chang et al. (Azmatch et al., 2012; Ren and Vanapalli, 2019). Changes
Lab TPH: 0.0046 d-1
5 to 14 °C (2011)
in unfrozen water content lead to changes in the physical,
mechanical and thermodynamic properties of frozen soils
2.6 to 60 mg
TPH/kg/day (Ren and Vanapalli, 2019; Watanabe and Osada, 2017),
(< 100 days) affecting heat transport, groundwater flow and biological
Variable from - McWatters et al. activity (McKenzie et al., 2007; Watanabe and Osada,
Field
30 to 10°C (2016)
13 to 26 mg 2017). Seasonal change in unfrozen water content
TPH/kg/day regulates the migration of salt in freezing and frozen soil
(> 100 days) (Bing and Ma, 2011; Kozlowski, 2009) and influences
seasonal shifts in microbial communities (Chang et al.,
Karppinen et al. 2018).
Lab Fixed at -5 °C F3: 0.0036 d-1
(2017b) The various biostimulation amendments that have
been applied in the context of bioremediation for petroleum
hydrocarbon-contaminated soils in cold climates are Mature
Gomez and Sartaj,
summarized in Table 3, which shows that the most organic -3.5 to -24 °C Field
(2014, 2013)
commonly studied amendment for biostimulation is compost
inorganic N, P and K nutrients (Børresen et al. 2007; Delille Fish meal,
et al. 2004; Dias et al. 2015; Gomez and Sartaj 2013, 2014; -5.5 to 7.7 °C Field Dias et al. (2015)
fertilizer
Karppinen et al. 2017a; Karppinen et al. 2017b; Karppinen
et al. 2019; Kauppi et al. 2011; Kim et al. 2018, 2021; McWatters et al.
Martínez Álvarez et al. 2017; Martínez Álvarez et al. 2020; Nutrients -30 to 10 °C Field
(2016)
McCarthy et al. 2004; McWatters et al. 2016; Mohn and
Stewart 2000; Namkoong et al. 2002; Nayeema 2023; Karppinen et al.
Biochar,
Ruberto et al. 2003; Sanscartier et al. 2009; Walworth et al. -5 °C Lab (2019, 2017b,
fertilizer
2017a)
2001; Whyte et al. 2001; Zhang et al. 2019). The
application of plant-derived biochar and clay minerals (e.g., Nutrients,
zeolite) has also been studied. -30 to 30 °C Field Kim et al., (2018)
humate

Table 3. Soil amendments applied in previous Biochar,

30 °C Lab Zhang et al. (2019)
bioremediation studies for petroleum hydrocarbon- nutrients
contaminated soils at low temperatures (from Nayeema
2023). Álvarez et al. (2020,
Nutrients 0 to 10 °C Field
2017)
Amendment Experiment conditions References
Nutrient,
Mohn and Stewart, -10 to 4 °C Lab Kim et al. (2021)
Nutrients 7 °C Lab zeolite
(2000)

Nutrients 5 °C Lab Whyte et al. (2001)

Figure 2 shows the SFCCs of various successfully
biostimulated petroleum hydrocarbon-contaminated soils.
Peat moss, The soil amendments, inorganic nutrient (N, P, and K), peat
Variable Field Zytner et al. (2001)
nutrients
moss and zeolite, caused an upward shift in the SFCCs,
corresponding to greater unfrozen water retention in the
Eriksson et al.
Nutrients -5 to 7 °C Lab treated contaminated soils during freezing, which in turn
(2001)
can promote biodegradation.
Walworth et al.
Nutrients 5 to 20 °C Lab
(2001)

Sewage
Namkoong et al.
sludge, 20 °C Lab
(2002)
compost

Ruberto et al.
Nutrients -2 to 10 °C Field
(2003)

Nutrients 0 to 20 °C Field Delille et al. (2004)

McCarthy et al.
Nutrients 1.3 to 4.9 °C Field
(2004)

Børresen et al.
Nutrients -5 to 5 °C Lab
(2007)

Paudyn et al.
Nutrients Variable Field
(2008)

Fertilizer,
Sanscartier et al.
bulking Variable Field
(2009)
agents
Figure 2. Upward SFCC shifts caused by soil biostimulation
amendments to freezing hydrocarbon-contaminated soil,
Nutrients -5 to 14 °C Lab Chang et al. (2011)
indicating greater unfrozen water retention (from Kim et al.,
2021, Reprinted with permission from Kim J., Lee, A.H.,
Wood chips, Lab, and Chang, W., Environ. Sci. Technol., 55(13): 9172-9180.
16 to 18 °C Kauppi et al. (2011)
nutrients Field Copyright 2021 American Chemical Society).
5 SFCC INDEX (α) FOR BIOREMEDIATION
As shown in Figure 3, biodegradation activity in
Kim et al. (2018) reported that notable petroleum freezing petroleum hydrocarbon-contaminated soils was
hydrocarbon biodegradation occurred during the freezing correlated with an increasing SFCC index, or α value
months in nutrient- and humate-amended, outdoor pilot- (Eq. 1). The α values ranging from 0.08 to 0.1 were
scale biopiles that exhibited greater unfrozen water correlated to enhanced biodegradation (Fig. 3). Very low α
retention over a winter season in Saskatchewan. A values (sandy soils) suggest that the soils may not retain
controlled laboratory study by Kim et al. (2021) further unfrozen water content enough to extend cold-adapted
reported that unfrozen water can be manipulated to microbial activity to freezing and frozen conditions due to
achieve greater hydrocarbon biodegradation at site- the rapid drop in unfrozen water content. Higher clay
representative freezing temperatures in soils that received contents in the soils result in higher α values (α > 0.1).
nutrient and clay mineral amendments. Figure 3 shows that However, at the same time, clay aggregates may increase
clay mineral (zeolite) and nutrient amendments increased nano- or micro-sized pores that are less accessible to
surface area and unfrozen water retention, respectively, bacteria. Therefore, more datasets need to be collected for
which in turn increased hydrocarbon degradation. The various soil types and treatments in future research on
same study indicated that, during the early phase of soil extended, cold-tolerant bioremediation.
freezing (4 to -5 °C), the role of nutrients is significant in
alleviating the abrupt freezing stress near the depressed
freezing point of the treated contaminated soil. The clay 6 PREDICTING THE SFCC OF BIOSTIMULATED
mineral amendment enhanced unfrozen water retention in SOILS
the subsequent deeply frozen phase (-5 to -10 °C) by
increasing surface area in the treated soil. Simulating the SFCCs of petroleum hydrocarbon-
contaminated soils amended with a variety of soil
amendments for bioremediation is beneficial for developing
cold-tolerant bioremediation strategies for remote cold
regions. Error! Reference source not found. illustrates
that SFCCs simulated using TEMP/W modelling were well
fitted to the experimental SFCCs of a site soil amended
with clay minerals (zeolite) under site-representative
freezing conditions. The experimental and simulated
SFCCs were similar (R2 = 0.98 to 0.99), which validates
that the TEMP/W model was properly calibrated for the site
soil. A predictive SFCC modelling tool could be put to
practical use if it were able to make SFCC predictions
under different scenarios.

Figure 4. Experimental and simulated SFCCs of a site soil

that received a clay mineral amendment (from Nayeema,
2023).

Figure 3. Biostimulation using nutrient and zeolite Figure 5 shows SFCCs that were produced by
amendments to increase unfrozen water content and varying the initial water content parameter within the
surface area, respectively, in frozen hydrocarbon- TEMP/W model calibrated for the same site soils, which
contaminated soil. Greater unfrozen water retention had an initial water content of 15% in the soil freezing
correlated to greater hydrocarbon biodegradation (from experiment. SFCC predictions were made by changing the
Kim et al., 2021; (from Kim et al., 2021, Reprinted with water content parameter to 7.5% and 20%. It was observed
permission from Kim J., Lee, A.H., and Chang, W., Environ. that initial water content can induce a shift in the SFCC
Sci. Technol., 55(13): 9172-9180. Copyright 2021 associated with a change in α value. Increasing the water
American Chemical Society).
content (20%) shifts the SFCC upward, whereas Bore, E.K., Apostel, C., Halicki, S., Kuzyakov, Y., and
decreasing the water content (7.5%) shifts the SFCC Dippold, M.A. 2017. Microbial metabolism in soil
downward. An upward shift towards the SFCC range of at subzero temperatures: adaptation mechanisms
clay may mean the soil will exhibit clay-like properties such revealed by position-specific 13 °C labeling. Front
as slower freezing, whereas a downward shift may indicate Microbiol., 8: 1-10.
faster freezing, similar to sand. This shift in the simulated Børresen, M.H., Barnes, D.L., and Rike, A.G. 2007.
SFCC caused by the initial water content is consistent with Repeated freeze–thaw cycles and their effects on
the findings of Teng et al. (2020), which state that different mineralization of hexadecane and phenanthrene
initial water contents greatly influence SFCCs. in cold climate soils. Cold Reg. Sci. Technol.,
49(3): 215-225.
Chang, W., Klemm, S., Beaulieu, C., Hawari, J., Whyte, L.,
and Ghoshal, S. 2011. Petroleum hydrocarbon
biodegradation under seasonal freeze−thaw soil
temperature regimes in contaminated soils from a
sub-Arctic Site. Environ. Sci. Technol., 45(3):
1061-1066.
Delille, D., Coulon, F., and Pelletier, E. 2004. Effects of
temperature warming during a bioremediation
study of natural and nutrient-amended
hydrocarbon-contaminated sub-Antarctic soils.
Cold Reg. Sci. Technol., 40(1): 61-70.
Deming, J.W. 2002. Psychrophiles and polar regions. Curr.
Opin. Microbiol., 5(3): 301-309.
Dias, R.L., Ruberto, L., Calabró, A., Balbo, A.L., Del
Panno, M.T., and Mac Cormack, W.P. 2015.
Hydrocarbon removal and bacterial community
Figure 5. Experimental SFCC of a site soil that received a
structure in on-site biostimulated biopile systems
clay mineral amendment compared to SFCCs simulated
designed for bioremediation of diesel-
by varying the initial water content in TEMP/W (from
contaminated Antarctic soil. Polar Biol., 38(5):
Nayeema, 2023). 677-687.
Drotz, S.H., Sparrman, T., Schleucher, J., Nilsson, M., and
Öquist, M.G. 2010. Effects of soil organic matter
7 CONCLUSIONS
composition on unfrozen water content and
heterotrophic CO2 production of frozen soils.
Linking soil-freezing characteristics to biological data is
Geochim. Cosmochim. Acta, 74(8): 2281-2290.
feasible and useful for managing and designing the
Drotz, S.H., Tilston, E.L., Sparrman, T., Schleucher, J.,
bioremediation of petroleum hydrocarbon-contaminated
Nilsson, M., and Öquist, M.G. 2009. Contributions
soils, especially for remote cold regions. The SFCC index, of matric and osmotic potentials to the unfrozen
or α value, is correlated to unfrozen water content and
water content of frozen soils. Geoderma, 148(3-
hydrocarbon biodegradation. Simulating SFCCs for
4): 392-398.
contaminated soils treated with soil amendments for
Eriksson, M., Ka, J.-O., and Mohn, W.W. 2001. Effects of
bioremediation is feasible. This review suggests that the
low temperature and freeze-thaw cycles on
SFCC is a promising tool for optimizing and managing cold-
hydrocarbon biodegradation in Arctic tundra soil.
tolerant bioremediation strategies for remote cold sites.
Appl. Environ. Microbiol., 67(11): 5107-5112.
Gomez, F., and Sartaj, M. 2013. Field scale ex-situ
References
bioremediation of petroleum contaminated soil
under cold climate conditions. Int. Biodeterior.
Atlas, R.M. 1981. Microbial degradation of petroleum
Biodegradation, 85: 375-382.
hydrocarbons: an environmental perspective.
Gomez, F., and Sartaj, M. 2014. Optimization of field scale
Microbiol. Rev., 45(1): 180.
biopiles for bioremediation of petroleum
Barnes, D.L., Wolfe, S.M., and Filler, D.M. 2004. hydrocarbon contaminated soil at low
Equilibrium distribution of petroleum
temperature conditions by response surface
hydrocarbons in freezing ground. Polar Rec.,
methodology (RSM). Int. Biodeterior.
40(3): 245.
Biodegradation, 89: 103-109.
Bing, H., and Ma, W. 2011. Laboratory investigation of the Han, Y., Wang, Q., Kong, Y., Cheng, S., Wang, J., Zhang,
freezing point of saline soil. Cold Reg. Sci.
X., and Wang, N. 2018. Experiments on the initial
Technol., 67(1-2): 79-88.
freezing point of dispersive saline soil. Catena,
Bore, E.K., Halicki, S., Kuzyakov, Y., and Dippold, M.A.
171: 681-690.
2019. Structural and physiological adaptations of Harvey, A.N., Snape, I., and Siciliano, S.D. 2012. Changes
soil microorganisms to freezing revealed by
in liquid water alter nutrient bioavailability and gas
position-specific labeling and compound-specific
13C analysis. Biogeochemistry, 143(2): 207-219. diffusion in frozen antarctic soils contaminated
 with petroleum hydrocarbons. Environ. Toxicol. McCarthy, K., Walker, L., Vigoren, L., and Bartel, J. 2004.
Chem., 31(2): 395-401. Remediation of spilled petroleum hydrocarbons
Karppinen, E.M., Siciliano, S.D., and Stewart, K.J. 2017a. by in situ landfarming at an arctic site. Cold Reg.
Application method and biochar type affect Sci. Technol., 40(1): 31-39.
petroleum hydrocarbon degradation in northern McWatters, R., Wilkins, D., Spedding, T., Hince, G.,
landfarms. J. Environ. Qual., 46(4): 751-759. Raymond, B., Lagerewskij, G., Terry, D., Wise, L.,
Karppinen, E.M., Stewart, K.J., Farrell, R.E., and Siciliano, and Snape, I. 2016. On site remediation of a fuel
S.D. 2017b. Petroleum hydrocarbon remediation spill and soil reuse in Antarctica. Sci. Total
in frozen soil using a meat and bonemeal biochar Environ., 571: 963-973.
plus fertilizer. Chemosphere, 173: 330-339. Mohn, W.W., and Stewart, G.R. 2000. Limiting factors for
Karppinen, E.M., Mamet, S.D., Stewart, K.J., and Siciliano, hydrocarbon biodegradation at low temperature in
S.D. 2019. The Charosphere Promotes Arctic soils. Soil Biol. Biochem., 32(8): 1161-1172.
Mineralization of 13C-Phenanthrene by Namkoong, W., Hwang, E.Y., Park, J.S., and Choi, J.Y.
Psychrotrophic Microorganisms in Greenland 2002. Bioremediation of diesel-contaminated soil
Soils. J Environ Qual, 48(3): 559-567. with composting. Environ Pollut, 119(1): 23-31.
Kauppi, S., Sinkkonen, A., and Romantschuk, M. 2011. Nayeema, T.F.K. 2023. Zeolite-Assisted Bioremediation of
Enhancing bioremediation of diesel-fuel- Petroleum Hydrocarbon-Contaminated Soils in
contaminated soil in a boreal climate: Comparison Cold Climates: A Treatment Strategy and CO2
of biostimulation and bioaugmentation. Int. Respiration Modelling for Seasonal Freezing.
Biodeterior. Biodegradation, 65(2): 359-368. M.Sc. Thesis, Department of Civil, Geological and
Kim, J., 2021, Stress-tolerant bioremediation strategy for Environmental Engineering, University of
cold-climate sites: laboratory, modelling and field Saskatchewan.
studies for extending petroleum hydrocarbon Öquist, M.G., Sparrman, T., Klemedtsson, L., Drotz, S.H.,
biodegradation to sesonally freezing and frozen Grip, H., Schleucher, J., and Nilsson, M. 2009.
contaminated soil phases. Ph.D. Thesis, Water availability controls microbial temperature
Department of Civil, Geological and responses in frozen soil CO2 production. Global
Environmental Engineering, University of Change Biol., 15(11): 2715-2722.
Saskatchewan. Panikov, N., Flanagan, P., Oechel, W., Mastepanov, M.,
Kim, J., Lee, A.H., and Chang, W. 2018. Enhanced and Christensen, T. 2006. Microbial activity in
bioremediation of nutrient-amended, petroleum soils frozen to below -39 °C. Soil Biol. Biochem.,
hydrocarbon-contaminated soils over a cold- 38(4): 785-794.
climate winter: The rate and extent of Paudyn, K., Rutter, A., Rowe, R.K., and Poland, J.S. 2008.
hydrocarbon biodegradation and microbial Remediation of hydrocarbon contaminated soils
response in a pilot-scale biopile subjected to in the Canadian Arctic by landfarming. Cold Reg.
natural seasonal freeze-thaw temperatures. Sci. Sci. Technol., 53(1): 102-114.
Total Environ., 612: 903-913. Ren, J., Vanapalli, S.K., and Han, Z. 2017. Soil freezing
Kim, J., Lee, A.H., and Chang, W. 2021. Manipulation of process and different expressions for the soil-
Unfrozen Water Retention for Enhancing freezing characteristic curve. Sci. Cold. Arid.
Petroleum Hydrocarbon Biodegradation in Reg., 9(3): 221-228.
Seasonally Freezing and Frozen Soil. Environ. Rivkina, E., Friedmann, E., McKay, C., and Gilichinsky, D.
Sci. Technol., 55(13): 9172-9180. 2000. Metabolic activity of permafrost bacteria
Konrad, J., and McCammon, A. 1990. Solute partitioning in below the freezing point. Appl. Environ.
freezing soils. Can. Geotech. J., 27(6): 726-736. Microbiol., 66(8): 3230-3233.
Ma, T., Wei, C., Xia, X., Zhou, J., and Chen, P. 2017. Soil Ruberto, L., Vazquez, S.C., and Mac Cormack, W.P. 2003.
freezing and soil water retention characteristics: Effectiveness of the natural bacterial flora,
connection and solute effects. J. Perform. Constr. biostimulation and bioaugmentation on the
Fac., 31(1): D4015001. bioremediation of a hydrocarbon contaminated
Margesin, R., and Schinner, F. 2001. Biodegradation and Antarctic soil. Int. Biodeterior. Biodegradation,
bioremediation of hydrocarbons in extreme 52(2): 115-125.
environments. Appl. Microbiol. Biotechnol., 56(5): Sanscartier, D., Zeeb, B., Koch, I., and Reimer, K. 2009.
650-663. Bioremediation of diesel-contaminated soil by
Martínez Álvarez, L., Ruberto, L., Lo Balbo, A., and Mac heated and humidified biopile system in cold
Cormack, W. 2017. Bioremediation of climates. Cold Reg. Sci. Technol.,55(1): 167-173.
hydrocarbon-contaminated soils in cold regions: Segura, J.H., Nilsson, M.B., Haei, M., Sparrman, T.,
Development of a pre-optimized biostimulation Mikkola, J.-P., Gräsvik, J., Schleucher, J., and
biopile-scale field assay in Antarctica. Sci. Total Öquist, M.G. 2017. Microbial mineralization of
Environ., 590: 194-203. cellulose in frozen soils. Nat. Commun., 8(1):
Martínez Álvarez, L.M., Ruberto, L.A.M., Gurevich, J.M., 1154.
and Mac Cormack, W.P. 2020. Environmental Segura, J.H., Nilsson, M.B., Schleucher, J., Haei, M.,
factors affecting reproducibility of bioremediation Sparrman, T., Székely, A., Bertilsson, S., and
field assays in Antarctica. Cold Reg. Sci. Öquist, M.G. 2019. Microbial utilization of simple
Technol., 169: 102915. carbon substrates in boreal peat soils at low
 temperatures. Soil Biol. Biochem. 135(2019):
438-448.
Siciliano, S.D., Schafer, A.N., Forgeron, M.A., and Snape,
I. 2008. Hydrocarbon contamination increases the
liquid water content of frozen Antarctic soils.
Environ. Sci. Technol., 42(22): 8324-8329.
Sparrman, T., Öquist, M., Klemedtsson, L., Schleucher, J.,
and Nilsson, M. 2004. Quantifying unfrozen water
in frozen soil by high-field 2H NMR. Environ. Sci.
Technol., 38(20): 5420-5425.
Thomassin-Lacroix, E., Eriksson, M., Reimer, K., and
Mohn, W. 2002. Biostimulation and
bioaugmentation for on-site treatment of
weathered diesel fuel in Arctic soil. Appl.
Microbiol. Biotechnol., 59(4-5): 551-556.
Tilston, E., Sparrman, T., and Öquist, M. 2010. Unfrozen
water content moderates temperature
dependence of sub-zero microbial respiration.
Soil Biol. Biochem., 42(9): 1396-1407.
Tucker, C. 2014. Reduction of air-and liquid water-filled soil
pore space with freezing explains high
temperature sensitivity of soil respiration below 0
°C. Soil Biol. Biochem., 78: 90-96.
Tuorto, S.J., Darias, P., McGuinness, L.R., Panikov, N.,
Zhang, T., Häggblom, M.M., and Kerkhof, L.J.
2014. Bacterial genome replication at subzero
temperatures in permafrost. ISME J., 8(1): 139-
149.
Walworth, J., Braddock, J., and Woolard, C. 2001. Nutrient
and temperature interactions in bioremediation of
cryic soils. Cold Reg. Sci. Technol., 32(2): 85-91.
Wen, Z., Ma, W., Feng, W., Deng, Y., Wang, D., Fan, Z.,
and Zhou, C. 2012. Experimental study on
unfrozen water content and soil matric potential of
Qinghai-Tibetan silty clay. Environ. Earth Sci.,
66(5): 1467-1476.
Whyte, L., Goalen, B., Hawari, J., Labbé, D., Greer, C., and
Nahir, M. 2001. Bioremediation treatability
assessment of hydrocarbon-contaminated soils
from Eureka, Nunavut. Cold Reg. Sci. Technol.,
32(2): 121-132.
Wu, M., Tan, X., Huang, J., Wu, J., and Jansson, P.-E.
2015. Solute and water effects on soil freezing
characteristics based on laboratory experiments.
Cold Reg. Sci. Technol., 115: 22-29.
Zhang, B., Zhang, L., and Zhang, X. 2019. Bioremediation
of petroleum hydrocarbon-contaminated soil by
petroleum-degrading bacteria immobilized on
biochar. RSC Advances, 9(60): 35304-35311.
Zytner, R., Salb, A., Brook, T., Leunissen, M., and Stiver,
W. 2001. Bioremediation of diesel fuel
contaminated soil. Can. J. Civ. Eng., 28(S1): 131-
140.
Assessing toxicity risk to aquatic life
from discharging groundwater
contaminant plumes using in situ
benthic caging
James W. Roy & Lee Grapentine
Water Science and Technology Directorate,
Environment and Climate Change Canada,
Burlington, ON, Canada

ABSTRACT

Risks from the discharge of groundwater contaminants to aquatic ecosystems have typically been determined by
comparing measured concentrations to aquatic life guidelines. Ideally, in-field measures could provide a clearer picture of
potential toxicity given the commonly complex site hydrological and geochemical conditions. For surface water
contaminants, in situ toxicity testing with caged organisms is a common field measure of such risk, but this approach has
rarely been applied to groundwater contaminants. The objective of this research was to demonstrate and evaluate in situ
caging for assessing the toxicity risk to benthic organisms from discharging groundwater contaminant plumes. The caging
tests (1-4 week deployments) were performed at surface waters (one river, one stream, one pond) receiving plumes of
organics (chlorinated solvents and petroleum hydrocarbons) or landfill leachate. Different cage designs and orientations
(horizontal, vertical, mix of both), and various toxicity metrics including organism survival and growth with two test
organisms (an amphipod (Hyalella azteca) and midge larvae (Chironomus riparius; stream site only), were examined. All
sites showed probable toxicity impact related to the discharging groundwater plumes, though other site conditions
confounded the results in some cases. The vertical cages showed a greater toxicity response to groundwater contaminants
and harmful conditions than horizontal cages, while a new hybrid design showed promise in its first and only test. Organism
survival gave a clearer metric of groundwater contaminant toxicity than did organism growth. The study findings indicate
that in situ toxicity cage testing holds promise for providing additional field-based information for aquatic ecosystem toxicity
assessments at groundwater-contamination sites and provide some guidance on future deployments.
Radon, atmospheric pressure, and
temperature monitoring of a ‘breathing well’
in Alberta
T.A. Morais1, S. Verma1,2, E.L. Eldridge1, and M.C. Ryan1
1Geoscience, University of Calgary, Calgary, AB, Canada
2Advisian, Calgary, AB, Canada

ABSTRACT

Radon in indoor air is a leading cause of lung cancer, but geogenic sources and transport pathways are poorly understood.
An 11-day monitoring program in a 'breathing' water well near a building with elevated indoor Rn concentrations ([Rn])
showed consistent variations between downhole [Rn] and atmospheric pressure over four cycles. Peak [Rn] of > 60,000
Bq/m3 (compared to an indoor air health guideline of 200 Bq/m 3) were observed during periods with sustained decreases
in PATM. The major gas species composition of daily grab samples showed two endmembers, with a high [Rn] end member
associated with elevated CO2 and Ar (and decreased O2 and CH4) concentrations.

Fugitive gas migration is commonly recognized in oil and gas wells, where deep, free phase methane gas migrates upwards
along pathways outside the well (i.e., in the well annulus). Water well drilling and completion may similarly provide a gas
migration pathway for deeper sources of radon gas.

RÉSUMÉ

Radon dans l'air intérieur est l'une des principales causes de cancer du poumon, mais les sources géogéniques et les
voies de transport sont mal comprises. Un programme de surveillance de 11 jours dans un puits d'eau «respiratoire» près
d’un bâtiment avec des concentrations intérieures élevées de Rn ([Rn]) a montré quatre cycles des variations entre [Rn]
dans le puit et la pression atmosphérique. Des pics [Rn] de > 60 000 Bq/m 3 (par rapport à une recommandation pour la
santé de l'air intérieur de 200 Bq/m3) ont été observés pendant les périodes de diminution soutenue des PATM. La
composition des principales espèces de gaz dans échantillons quotidiens a montré deux membres terminaux, avec un
membre terminal élevé [Rn] associé à des concentrations élevées de CO 2 et d'Ar (et une diminution de l'O2 et du CH4).

La migration fugitive de gaz est communément reconnue dans les puits de pétrole et de gaz, où le gaz méthane profond
en phase libre migre vers le haut le long de voies à l'extérieur du puits (c'est-à-dire dans l'espace annulaire du puits). Le
forage et l'achèvement de puits d'eau peuvent également fournir une voie de migration de gaz pour des sources plus
profondes de gaz radon.

1 INTRODUCTION Efforts to map radon risk based on geology have not

been successful, with a major European Union initiative
Radon-222 (henceforth annotated as Rn) gas inhalation in designated ‘neither a straightforward nor a technically easy
indoor air is a leading cause of lung cancer among task’ (Bossew et al., 2020). This suggests that main source
Canadians who have never smoked tobacco and is of geogenic Rn source is not occurring in the shallow
responsible for 16% of lung cancer cases in Canada (Kim subsurface (as assumed in mapping efforts). Rather, the
et al., 2016). Formed as an intermediate daughter product lack of success in mapping indoor Rn to geology suggests
(and the only gaseous daughter product) in the U-Th decay that its sources and transport pathways are not well
chain, Rn concentrations are typically low in the understood.
atmosphere (~5 to 15 Bq/m3) but much higher in the soil Significantly higher radon concentrations are observed
and unsaturated zone, where they routinely reach tens and in homes using groundwater as a water supply than homes
hundreds of thousands of Bq/m3 (Chau et al., 2011). that rely on surface water (Casey et al., 2015; Chen, 2021),
The ubiquitous presence of Rn in both the shallow including Albertan homes (Eldridge, 2022). This suggests
subsurface and indoor air is notable given Rn’s half-life of a unique pathway exists in homes that rely on well water
3.8 days. This implies that Rn is i) continually generated in (i.e., rural residents and smaller communities) relative to
the subsurface, and ii) transported either by diffusion those that are supplied by surface water (i.e., most
and/or by advection in as dissolved or free phase gas. If residences in urban communities like cities and large
free phase gas migration occurs, it would travel along towns). The single largest differences in construction of
connected fracture and large pore throat pathways under homes supplied by ground- vs. surface water are the use
buoyancy in the saturated zone (Chen et al., 2023). Once of a water well for household water supply, and an onsite
it reaches the water table, radon transport into the shallow septic system for wastewater disposal.
subsurface prior to entering building areas in contact with At first glance, pumping groundwater into homes for
the subsurface would need to occur before decay. domestic use seems a logical radon pathway that could be
responsible for higher [Rn] in indoor air of homes on well Although no record was identified for the water well in
waters but this is not borne out in surveys of dissolved [Rn] Alberta’s Groundwater Information Centre, nearby well
in well water which consistently find that groundwater records report glacial materials to ~41.5 m below ground
concentrations are low relative to the recommended action surface (m bgs) underlain by alternating shale and
level for dissolved Rn in water supplies of 2,000 Bq/L sandstone that is typical of the Paskapoo bedrock aquifer
(Health Canada, 2009). For example, dissolved [Rn] in the region (Grasby et al., 2009). The well was sounded
ranged from 7 to 22 Bq/L (average 32 Bq/L) in 20 well water to a depth of at least 71 m and had a static water level of
samples in southern Alberta (Olson and Anderson, 2007), about 59.5 m. A downhole camera identified a change in
and dissolved Rn in 40 Alberta provincial groundwater casing diameter from a deeper 10.2 cm (4 inch) casing to
monitoring wells ranged from 0.20 to 75.1 Bq/L, with a a shallower 15.2 cm (6 inch) surface casing at ~ 5.6 m bgs
mean of 8.1 Bq/L (Eldridge, 2022). (shown schematically in Figure 1). No completion details
Free phase gas emanations of radon are known to were available for the well, but the change in casing
occur in geologic features like volcanoes, faults, and diameter likely provided a connection between free phase
springs (Ball et al., 1991), which are starting to be gas inside the well casing and in the subsurface outside the
considered in the context of radon in indoor air (Drolet and casing.
Martel, 2016). Although fugitive methane gas migration is
widely identified to occur around petroleum wells (Abboud 3 RESULTS AND DISCUSSION
et al., 2020), it has not been considered in the context of
geogenic radon pathways contributing to indoor air. 3.1 Time series of radon and air pressure
Breathing wells have been identified in the scientific
literature for more than 130 years (Willard, 1982). They Three major and one minor cycles in [Rn] and PATM
cycle between ‘inhalation’ of air into and ‘exhalation’ of (Figure 2A and B) were identified using 9-hour rolling PATM
subsurface gas out of the well during sustained periods of average values (not shown). Four PATM minima values
increasing and decreasing PATM, respectively. During were identified that separated periods of decreasing and
exhalation, subsurface air enters the well casing via a increasing PATM over the 11-day monitoring period
connection between the well casing and the unsaturated (between ~87.1 and 86.7 kPa; indicated by circled
zone, for example by a casing ‘break’ and/or by having a numbers on Figure 2A and B). The four PATM minima were
screen that is at least partially completed in the unsaturated each associated with elevated [Rn] (> 54,000 Bq/m 3 for the
zone (Levintal et al., 2020). In essence, breathing wells three major PATM cycles, (Oct 10, 14, 18th) and >18,500
undergo unsaturated zone barometric pumping as Bq/m3 for the minor PATM cycle (Oct 15th).
observed in fractures (Massman, 2006) and other After each period of decreasing PATM, [Rn] continued to
unsaturated zone features. Reports of breathing wells are increase past each PATM minima even though PATM was
relatively uncommon (Hill, 2004), although regionally increasing (Figure 2A). During this ‘lag time’, increasing
common breathing wells have been identified (Freckelton, PATM was not yet reflected by decreasing [Rn]. After this lag
2019). time however, [Rn] decreased rapidly (over three to six
A breathing water well that was anecdotally identified hours) during well ‘inhalation’ under continually increasing
near a Calgary region residence with elevated [Rn] in PATM to background concentrations.
indoor air provided an opportunity to evaluate whether A similar lag time was observed after the initiation of
radon gases were complicit in well breathing. An 11-day decreasing PATM (i.e., after PATM maxima). Background
monitoring program combined hourly measurement of concentrations (interpreted visually as >35 Bq/m3)
radon in free phase gas samples from inside the water well continued to be observed for about four to ten hours after
casing (2.5 m bgs) and from outside air at an ‘at-grade’ PATM maxima were reached for a four-to-ten-hour ‘lag time’
location about three meters distance from the water the [Rn] began to increase again.
wellhead. Grab samples collected on a near-daily basis The lag times between changes in [Rn] with time and
from inside the well casing were analyzed for major gas the inflections between PATM cycles are interpreted as the
composition. Vertical downhole temperature profile time it takes either for i) air to move down the well to reach
monitoring was conducted to evaluate the depth at which the “well casing” sampling point at 2.5 m bgs (during
subsurface gases were entering into the well casing. In periods of inhalation, or decreasing [Rn] and increasing
addition to observing whether radon was complicit in well PATM), or ii) for subsurface gas to move up the casing from
breathing, radon geogenic source and pathways were the casing break to the ‘casing’ sampling point during
considered. periods of exhalation, or increasing [Rn] and decreasing
PATM. The length of the lag period depends on the relative
2 BREATHING WELL DESCRIPTION depths of the casing break and casing gas sampling. For
example, if the Rn sampling point were higher up in the
The breathing water well was anecdotally identified near a casing, [Rn] would decrease sooner after PATM started
Calgary-region residence with elevated [Rn] in indoor air increasing, and [Rn] would increasing later after the start of
radon concentrations by a homeowner who reported a periods of decreasing PATM.
‘whistling’ sound in the annulus of the pipe conduit between During sustained decreases in PATM, [Rn] increased
the well and its water cistern. The homeowner graciously over relatively long periods (i.e., more than 48 hours in the
provided access for well casing gas monitoring to evaluate first cycle between on October 8-10th) and never
whether Rn gases were involved in well breathing. plateaued. This suggests [Rn] would have continued to
increase had the decreasing PATM been sustained. Thus,
the gas concentrations sampled did not approach the
‘endmember’ composition of the subsurface gas source(s).

Figure 1. A) Approximate location of breathing well in Calgary, Alberta region. B) Schematic diagram of breathing well
studied showing observed ‘casing break’ at 5.6 m bgs. The yellow lines represent copper tubing for hourly radon
concentration monitoring and daily grab sample collection from i) 2.5 m bgs in the ‘wellhead’ and ii) an ‘at-grade’ location
about three meters from the wellhead. Filled yellow circles indicate the downhole temperature monitoring locations
(wellhead, 2.5, 12, and 25 m bgs) and the at-grade monitoring location. The arrows indicate interpreted gas transport
pathways during well ‘exhalation’ (red) ‘inhalation’ (blue). The dashed box shows where the blow-up schematic of the well
casing with an uncemented annulus that could be acting as a gas migration pathway for free phase gas transport, as
shown in C).

3.2 Downhole well temperatures and depth of the occurring, since when subsurface gases entering the well
“casing break” connection to the unsaturated zone are not prone to diel air temperature variations.
The observation of diel air temperature changes during
Air temperature (sampled at grade) increased in the inhalation in the shallow well above (but not below) the
daytime periods to maxima between about 10 and 20oC observed casing break at 5.6 m bgs (Figure 1) suggests
and decreased overnight to minima ranging between about the casing break served as the primary connection with the
3 and -5oC (Figure 2C). Muted diel air temperature unsaturated zone, where subsurface gases entered the
fluctuations were observed in the shallower monitoring casing at during periods of sustained PATM decrease.
depths inside the well casing (i.e., wellhead (0.25 m bgs)
and 2.5 m bgs; Figure 2D) for parts of the 11-day 3.3 Water table and temperature fluctuations
monitoring program but did not occur in deeper monitoring
points (as shown by relatively small standard deviations in No significant fluctuations in the temperature (Table 1) or
temperatures at 12 and 25 m bgs; Table 1). water table elevation (data not shown) were observed
Visual comparison shows that shallow downhole during the 11-day monitoring period.
temperature fluctuations mimics air temperatures (Fig. 2C
and 2D), confirming these periods are when the well is 3.4 At-grade radon concentrations
“inhaling” (i.e., periods of increasing PATM and decreasing
or background [Rn]; Figure 2A). Conversely, the periods The [Rn] in air sampled at-grade at a location ~3 m from
when the observed shallow well casing temperatures are the wellhead (Figure 1) showed a muted refection of the
relatively constant are observed when well exhalation is [Rn] at 2.5 m bgs inside the well casing (Figure 2A). This
suggests that radon was advected to the outside air at outdoor air were up to four times Health Canada’s guideline
ground surface either from the top of casing and/or in the for indoor air.
soil around the well. The observed concentrations in

Figure 2. Time series for 11-day monitoring showing A) radon concentrations ([Rn]) in the well casing (on a vertical log
scale). Vertical dashed lines indicate inflections between periods of increasing and decreasing [Rn]. Open symbols show
[Rn] sampled at 2.5 m bgs in the wellhead, where black symbols have low concentrations (<35 Bq/m3), red symbols show
periods with increasing [Rn], blue open symbols indicate decreasing [Rn]). Filled, black symbols [Rn] measured in air
sampled ~ five meters from the wellhead. The dashed vertical bars indicate well breathing activity (as indicted by
temperature; see text) and the light grey shading show night-time. The circled numbers denote the PATM minima in each
of four major PATM cycles; B) atmospheric pressure (PATM; kPa), and C) temperature (o C) measured for air, and D)
Wellhead (0.25 m bgs), and downhole (2.5, 12, and 25 m bgs) temperatures. Note: different vertical scales are used for C
and D; red and blue vertical arrows are drawn according to downhole temperatures.

Nitrogen was the dominant gas species in the grab

3.5 Grab sample composition and radon gas source samples collected, constituting between 77.4 and 78.2%
and pathway (Table 2). The strongest correlation between Rn and the
gas species analyzed was with CO2 (r = 0.96; Table 2).
Both Rn and CO2 were also positively correlated with Ar (r in the subsurface. Since Rn is rarely produced at
= 0.60 and 0.77, respectively) and negatively correlated sufficiently high concentrations to form free phase gas, it
with O2 (r = -0.71 and -0.82, respectively) and CH4 (r = - often partitions into free phase CO2, with the latter known
0.66 and -0.65, respectively). Thus, the gas composition at as a ‘carrier gas’ (Etiope and Martinelli, 2002). The source
2.5 m depth in the well casing can be characterized as two of free phase CO2 is neither clear, with potential candidates
endmembers, with Rn-, CO2-, and Ar-rich gas in one end being CH4 oxidation (Cesar et al., 2021), or ‘deeper’ gases
member, and O2 and CH4 in the other end member. that reach the shallow subsurface via geologic structures,
Oxygen is the only gas species whose presence can be which are often associated with Rn (Sukanya et al., 2022).
definitively associated with air and shallow subsurface gas. The Rn and CO2 were clearly transported as a free phase
The negative association with O2 suggests Rn and CO2 gas in the subsurface, entering the casing in the
have a deeper source. A negative correlation between Rn subsurface. Although free phase gases are known to
and methane observed in Alberta groundwater was exsolve from groundwater (Amos et al., 2005), the lack of
attributed to free phase gas removal (or ‘gas stripping’) any indication of free phase gas in the groundwater
from groundwater in methane-rich groundwater (Eldridge, monitoring (i.e., fluctuations in water pressure or EC;
2022), which is common in the Calgary region (Humez et Morais et al., 2021) suggest the groundwater is not a
al., 2016). significant source of free phase gas.
Fugitive (free phase) methane gas migration is widely
Table 1. Descriptive statistics (avg., median, std. dev., and recognized to occur around oil and gas wells (Kang et al.,
95% CI) for temperature in air (~3m from the wellhead), 2014), with depths that are beyond the depth of fresh
inside the wellhead (WH; 0.25 m bgs), and downhole 2.5, groundwater (Fleming et al., 2022). Fugitive gas migration
12, and 25 m bgs; and gauge water pressure in occurs in a significant fraction of oil and gas wells in Alberta
groundwater at ~5.5 m below the water table (GW; kPa). (5 to 15%; Abboud et al., 2021) despite well completion
For each parameter, n = 15,841. regulations that are considerably more rigorous than
required in the water welling industry. If fugitive gas
Loc- Depth migration is occurring at the breathing well site, it could be
Median Avg. Std Dev 95 % CI a mechanism by which deeper CO2- and Rn-rich gases
ation (m bgs)
reach the 5.6 m casing break and are exchanged with the
Temperature (oC) casing gas and shallow subsurface.
Air 0 4.13 5.48 5.69 0.089
4 CONCLUSIONS
WH 0.25 8.65 7.28 2.89 0.045
2.5 2.5 9.70 9.51 0.63 0.010 Four cycles of alternately increasing and decreasing
atmospheric pressure over the 11-day monitoring program
12 12 7.73 7.79 0.18 0.003
were associated with lagged, but systematic fluctuations in
25 25 7.19 7.22 0.11 0.002 radon concentrations inside the wellhead. Radon
concentrations remained at background levels (<35 Bq/m3)
GW 65 6.68 6.68 0.00 n.d.
during sustained periods of increasing PATM and increased
Pressure (kPa) during sustained decreasing PATM. Similar fluctuations
were observed in [Rn] in out-of-doors air sampled at-grade
GW 65 54.14 54.13 0.09 0.001
about three meters from the wellhead.
Maximum observed [Rn] radon concentrations
(>60,000 Bq/m3 inside the well casing and >800 Bq/m3 in
Table 2. Pearson’s correlation coefficients for major gases the out-of-doors air) suggest the subsurface gas
in grab samples collected from well casing at 2.5 m bgs (n concentrations are high enough to be significant source of
= 9). indoor air Rn.
Wellhead grab samples showed a two endmember with
Rn Ar O2 N2 CO2 CH4 the high [Rn] endmember associated with CO2 and Ar, and
the low [Rn] endmember with increased O2 and CH4.The
Rn 1.00 - - - - - association between Rn and CO2 in the high [Rn]
Ar 0.60 1.00 - - - -
endmember could either be because they have common
source(s), or because Rn partitions into free phase CO2
O2 -0.71 -0.68 1.00 - - - during its transport to surface.
The time series of vertical temperature profile suggests
N2 -0.08 -0.19 0.72 1.00 - - the subsurface gas entered the casing (and air exited the
CO2 0.96 0.77 -0.82 -0.21 1.00 - casing) between 2.5 and 12 m bgs, likely at the 5.6 m deep
casing ‘break’.
CH4 -0.66 -0.33 0.59 0.18 -0.65 1.00 The geogenic source and transport pathways of the
subsurface gas with elevated [Rn] are not clear, but gas
migration may be a relevant pathway.
The strong association between Rn and CO2 could
reflect a common geogenic endmember. Alternatively, it
could be an artifact of Rn partitioning into free phase CO2
5 REFERENCES Grasby, S.E., Chen, Z., Hamblin, A.P., Wozniak, P.R.J,
and Sweet, A.R., 2009. Regional characterization of the
Abboud, J., Watson, T. L., and Ryan, M.C., 2020. Fugitive Paskapoo bedrock aquifer system, southern Alberta.
methane gas migration around Alberta’s petroleum Can J Earth Sci, 45: 1501-1516.
wells. Greenhouse gases: Science and technology, 11: Health Canada, 2009. Guidelines for Canadian drinking
37-51. water quality. Guideline technical document:
Amos, R.T., Mayer, K.U., Bekins, B.A., Delin, G.N., Radiological parameters. Health Canada, H128-1/10-
Williams, R.L., 2005. Use of dissolved and vapor‐ 614E-PDF.
phase gases to investigate methanogenic degradation Hill, S.R., 2004. The physical and geochemical
of petroleum hydrocarbon contamination in the characterization of oxygen-depleted breathing wells in
subsurface. Water Resources Research, 41: W02001. central Alberta. MSc. Thesis, University of Alberta,
Ball, T.K., D.G. Cameron, T.B. Colman, and P.D. Roberts, Edmonton, AB.
1991. Behaviour of radon in the geologic environment: Humez, P., Mayer, B., Ing, J., Nightingale, M., Becker, V.,
A review. Quarterly J. Engineering Geology, 24: 169- Kingston, A., Akbilgic, O., and Taylor, S., 2016.
182. Occurrence and origin of methane in groundwater in
Bossew, P., Cinelli, G., Ciotoli, G., Crowley, Q.G., De Cort, Alberta (Canada): Gas geochemical and isotopic
M., Medina, J.E., Gruber, V., Petermann, E., and approaches. Science of the Total Environment 541:
Tollefsen, T. 2020. Development of a Geogenic Radon 1253-1268.
Hazard Index—Concept, History, Experiences. Int J Kang, M., Kanno, C.M., Reid, M.C., and Onstott, T.C.,
Environ Res Public Health, 17: 4134. 2014. Direct measurements of methane emissions from
Casey, J.A., Ogburn, E.L., Rasmussen, S.G., Irving, J.K., abandoned oil and gas wells in Pennsylvania. PNAS,
Pollak, J., Locke, P.A., Schwartz, B.S., 2015. 111: 18173-18177/
Predictors of indoor radon concentrations in Kim, S.-H., Hwang, W.J., Cho, J.S., and Kang, D.R., 2016.
Pennsylvania, 1989-2013. Environ. Health Attributable risk of lung cancer deaths due to indoor
Perspectives, 123: 1130-1137. radon exposure. Ann Occup Environ Med., 28: 8.
Cesar, J., Mayer, B., Humez, P., 2021. A novel isotopic Levintal, E., Dragila, M.I., Zafrir, H., and Weisbrod, N.,
approach to distinguish primary microbial and 2020. The role of atmospheric conditions in CO2 and
thermogenic gases in shallow subsurface radon emissions from an abandoned water well.
environments. Applied Geochemistry, 131: 105048. Science of the Total Environment, 137857.
Chau, N.D., Dulinski, M, Jodlowski, P., Nowak, J., and Olson, R.A., and Anderson, S., 2007. Preliminary water
Rozanski, K., 2011. Natural radioactivity in well sampling to assess the uranium potential in the
groundwater – A review. Isotopes in Env and Health Whiskey Gap area (NTS 82H/2,3), Alberta Energy and
Studies, 47: 415-437. Utilities Board, EUB/AGS Earth Sciences Report 2007-
Chen, J., 2021. A summary of residential radon surveys 08.
and the influence of housing characteristics on indoor Massman, W. J., 2006. Advective transport of CO2 in
radon levels in Canada. Health Phys., 121: 574-580. permeable media induced by atmospheric pressure
Chen, X., Liu, Y., Jiang, Y., Feng, S., 2023. Radon fluctuations: 1. An analytical model. Journal of
transport carried by geogas: prediction model. Env. Geophysical Research, 111: G03004
Sci. and Pollution Research. Morais, T.A., B.L. Ladd, N.A. Fleming, and M.C. Ryan,
https://doi.org/10.1007/s11356-023-28616-4 2021. Free-Phase Gas Detection in Groundwater Wells
Drolet, J.P., Martel, R., 2016. Distance to faults as a proxy via Water Pressure and Continuous Field Parameters.
for radon gas concentration in dwellings. J. Env. Groundwater, 60:262.
Radioactivity, 152: 8-15. Sukanya, S., Noble, J., and Joseph, S., 2022. Application
Eldridge, E.L., 2022. Radon in Alberta Groundwater. of radon as an environmental tracer in hydrogeological
Radon in Alberta groundwater (Unpublished MSc and geological investigations: An overview.
thesis). Geoscience, University of Calgary, Calgary, Chemoshere. 303: 135141.
AB. Willard, J.T., 1892. A breathing well in Logan County,
Etiope, G. and Martinelli, G., 2002. Migration of carrier and Kansas. Science, 20: 337.
trace gases in the geosphere. Phys. of the Earth and
Planetary Interiors, 129:185-204.
Fleming, N.A., Morais, T.A., Mayer, K.U., and Ryan, M.C.,
2021. Spatiotemporal variability of fugitive gas
migration emissions around a petroleum well.
Atmospheric and Pollution Research, 12(2021) 101094
Freckelton, C., 2019. A physical and
geochemical characterization of southwestern
Ontario's breathing well region. PhD Thesis, University
of Western Ontario, London, ON. Electronic thesis and
dissertation repository, Paper 1105.
Tuesday, October 3, 2023

DAMS AND
EMBANKMENTS I
Toward greater confidence in numerical
analysis results in geotechnical engineering
Vincent Castonguay
Seequent, Québec, QC, Canada

ABSTRACT
Numerical analyses are routinely performed in geotechnical engineering to
study the stability of slopes, to determine the pore-water pressure
conditions in a domain, or to calculate the resulting stresses and strains following a perturbation. These three types of
analysis are among the most reported analyses in geotechnical reports and yet, convergence information is seldom
displayed with the results. Convergence is inherent to numerical processes; it dictates whether the results of an analysis
can be trusted or not. Hence, geotechnical engineers should strive to diligently report the convergence status of any
numerical analysis they perform. Convergence is discussed in this paper in the context of the three numerical analysis
types aforementioned. Propositions to improve convergence disclosure are presented.

RÉSUMÉ
Les analyses numériques sont couramment effectués en ingénierie géotechnique pour étudier la stabilité des pentes,
déterminer les conditions de pression d'eau interstitielle dans un domaine, ou calculer les contraintes et déformations
résultant d'une perturbation. Ces trois types d'analyse font partie des analyses les plus effectuées pour les rapports
géotechniques et pourtant, les informations relatives à la convergence qui s’y rattachent sont rarement affichées avec les
résultats des modélisations. La convergence est inhérente aux processus numériques ; elle détermine si les résultats d'une
analyse sont fiables ou non. Par conséquent, les ingénieurs en géotechnique doivent s'efforcer de communiquer avec
diligence l'état de convergence de toute analyse numérique qu'ils effectuent. La convergence est examinée dans cet article
dans le contexte des trois types d'analyse numérique susmentionnés. Des propositions pour une meilleure divulgation de
la convergence sont présentées.

1 INTRODUCTION desired convergence criterion and declares itself “done”. At

worst, it reaches the maximum number of iterations
Numerical analyses have been part of the geotechnical allowed and is forced to call it quits, no matter the state of
engineering landscape for decades. These are performed the solution.
at various stages of different types of engineering projects The reader of a geotechnical report containing
to inform engineers on physical systems they must deal numerical analysis results has, a priori, no way of telling
with through their practice: slopes, embankments, what the state of a numerical analysis was when the
foundations, etc. Numerical analysis refers to a large array geotechnical engineer produced the values or plots they
of analysis types, which all have in common the use of a are reviewing. Were the results correctly converged, or
computer to solve nonlinear equations, in most cases; were they ill-converged? In the vast majority of
equations which would otherwise prove impossible to solve geotechnical reports, there are no indications regarding
by hand, within reasonable time constraints. this; engineers simply report the results and readers must
Geotechnical engineering projects incorporating trust them that convergence was appropriately handled.
numerical analyses generally use these tools in This paper discusses convergence in three of the most
conjunction with a variety of other tools, ranging from hand commonly performed numerical analysis types in
calculations to problem-dependent spreadsheet solutions. geotechnical engineering: limit equilibrium slope stability,
Once the engineer is through with their calculations, they seepage, and stress-strain analyses. Each of these brings
usually write a report, in which they detail their convergence challenges. Geotechnical reports containing
methodology. A reader of such reports quickly realizes not such analyses should provide more than simply the end
all calculation methods are born equal when it comes to results; they should include clear indications for the reader
transparency. While one might be able to follow the as to the state of convergence in their models to build
reasoning and hypotheses posed in reported hand confidence in the results they are viewing. In addition, this
calculations, it is often quite different for numerical transparency exercise would also lead the geotechnical
analyses. engineers performing numerical analyses to deliberately
All numerical analyses share a common challenging interrogate the models they are building, review
denominator: they are never exact. By nature, numerical convergence results, and ensure greater quality for the
processes are approximations. They are of course work they are producing.
approximations of a physical system, that is inherent to the Throughout the paper, figures are shown to
idea. But they are also, and most importantly, demonstrate ideas and concepts. These figures were
approximations in the sense that a numerical calculation is produced with GeoStudio (Seequent, 2023), a multi-
never “finished”. At best, a numerical calculation reaches a physics numerical simulation software suite specialized in
geotechnical analyses. The concepts discussed herein are factor of safety affecting the column, and by definition,
not specific to GeoStudio and could have been given the method, the entire slip surface. Since the factor
demonstrated via any other similar software suite. of safety is also an unknown initially, as it is precisely the
value the engineer wishes to calculate, therein lies the
1.1 Terminology conundrum: the normal force depends on the factor of
safety, but the factor of safety also depends on the normal
There are several expressions a geotechnical engineer force. One must know one to know the other.
venturing into the numerical world should know. Here are To resolve this issue, a first guess of the factor of safety
a few important definitions pertaining to the discussion at must be provided, then this first guess can be used to
hand: calculate a first estimate of the normal force at the base of
Converged solution: the results of a calculation which the column. With the normal force in hand, an updated
satisfied a given convergence criterion. factor of safety can be calculated. If the difference between
Convergence criterion: a user-specified (generally) the updated and previous factors of safety is larger than the
tolerance that must be met between results from two tolerance set (i.e. the convergence criterion), then the
successive iterations for the numerical process to come to updated factor of safety is used to calculate an updated
a halt. normal force at the base of the column. This allows once
Iteration: a single trial at solving the numerical again to calculate a new value for the factor of safety that
equations in an attempt to find a satisfying answer. can be compared to the previously calculated value,
Maximum number of iterations: a safety exit allowing comparing the difference between the two to the
the numerical calculation to stop when the convergence convergence criterion. This feedback loop is called
criterion fails to be met. iterating.
The iteration process will proceed until one of two
things happens: either the calculated value for the factor of
2 LIMIT EQUILIBRIUM SLOPE STABILITY safety stops changing significantly between two
ANALYSES successive iterations and the convergence criterion set by
the user is met. Or, when the convergence criterion fails to
Slope stability analyses using the limit equilibrium be met, the maximum number of iterations allowed will be
technique are among the most commonly performed reached, concluding the process. The degree of
numerical analyses in geotechnical engineering. confidence one can entrust into the results of this process
Nevertheless, engineers sometimes still fail to appreciate entirely depends on which of these outcomes prevails. A
this type of analysis is a numerical process. Most converged solution signifies success (if the convergence
geotechnical engineers will have performed hand criterion sought after was appropriately sized in the first
calculations of simple slope stability problems at university, place). Conversely, a solution which was stopped in its
for which the Fellenius (1936) or Ordinary method is used. tracks after having reached the maximum number of
In this case, there is no numerical process involved, since iterations allowed could mean anything between relatively
the intercolumn normal and shear forces are ignored, successful to completely wrong.
which brings the number of unknowns neatly equal to the
number of available equations from static equilibrium 2.2 Iterating for the lambda value
(summation of forces and moment equal to zero). This
yields a mathematically determinate problem, solvable by The second numerical process being solved when using a
hand, perfect for teaching purposes. limit equilibrium method which properly accounts for
However, the Ordinary method is not used in common intercolumn forces relates to the calculation of the
geotechnical practice, since neglecting the intercolumn intercolumn shear forces. These forces are unknowns,
forces amounts to an important oversimplification of the treated as functions of the intercolumn normal forces. The
physical system the numerical model is trying to represent. shape of the function relating both sets of forces depends
As such, more robust limit equilibrium methods are on the chosen limit equilibrium method. The Spencer
routinely used nowadays to solve slope stability problems, method simply states that the intercolumn shear force is a
such as the Spencer (1967) and the Morgenstern-Price fraction (lambda value) of the intercolumn normal force.
(1965) methods. By accounting for intercolumn normal and The Morgenstern-Price method goes one step further and
shear forces, these methods bring improved rigour, but at allows users to specify a function that will describe how the
a cost: iteration processes. Two processes, to be precise: intercolumn shear force varies along the height of the soil
iterating for the factor of safety, and iterations to find a column as a function of the intercolumn normal force. In
suitable lambda value. general, but not obligatorily, this function is assumed to be
a half-sine. Still, as for the Spencer method, the chosen
2.1 Iterating for the factor of safety function is multiplied by a scaling factor, the lambda value,
to properly scale the intercolumn shear and normal forces.
The first of two numerical processes being solved when For both methods, the lambda value is the unknown
using a limit equilibrium method which properly accounts that will unlock the calculation of the interslice shear forces.
for intercolumn forces relates to the normal force that With these forces in hand, the solver has everything it
develops at the base of each column. The equation for the needs to calculate the factors of safety. Note “factors”,
summation of forces in the vertical direction is used to plural, not “factor”, singular. The equations for the
calculate that force. It depends, among other things, on the summation of forces in the horizontal direction and the
summation of moments (both of which must equal zero) A factor of safety vs lambda plot where the cross-over point
provide independent expressions for the factor of safety. A between the force and moment factor of safety curves is
satisfactory solution to the limit equilibrium problem will ambiguous (i.e. if the curves almost become parallel) or
thus require the resulting factors of safety from both where more than one possible solution could be obtained
expressions to be equal. By varying the lambda value used if the researched lambda range was extended (such as
to calculate the intercolumn shear force, a series of factors shown in the right-hand side of Figure 2), would be deemed
of safety can be calculated for both expressions. A poorly converged.
satisfactory solution is found when a clear unambiguous
cross-over point can be found between both functions, as 2.3 Proposed improvements for stability analyses
shown in Figure 1. The cross-over point defines the lambda
value for which the factors of safety calculated by both the Given the two numerical processes occurring at solve time
moment and force equilibrium expressions are equal. This in limit equilibrium slope stability analyses that properly
becomes the final value for the factor of safety affecting the account for interslice forces, how can reviewers of
slope (1.46 in Figure 1 for example). geotechnical report judge the validity and rigour of a given
analysis? For example, by simply looking at the left-hand
side of Figure 2, which is typically the only figure reported
Factor of Safety vs. Lambda in reports, how could a reviewer make sure the maximum
1.7 number of iterations was not reached when calculating the
factor of safety and normal force? Or how could they make
sure the factor of safety vs lambda plot is unambiguous?
They simply could not.
Moment To remove this uncertainty, a better practice would be
1.6 Slip 25
to always report limit equilibrium slope stability analysis
Factor of Safety

results together with the factor of safety vs lambda plot, the

Force convergence criterion and the final iterations count (versus
Slip 25
the maximum number of iterations allowed).
1.5 Figure 2 shows an example of this good practice. In this
F of S
Slip 25
case, the reviewer can immediately notice that there are
most probably two lambda values where a cross-over point
could exist: the point found by the software, and a second
1.4
point just outside the positive range of lambda values
-1 0 1 explored during the calculations. Even worse, the second
Lambda cross-over point would correspond to a lower factor of
Figure 1. Lambda plot for limit equilibrium stability analysis safety (around 0.9) than that of the reported value (around
1.0). The reviewer could thus conclude that they should not
trust the reported factor of safety.

1.002 Factor of Safety vs. Lambda

1.02
24
1

0.98 Moment
Slip 20
Factor of Safety

19
0.96

Force
0.94 Slip 20
Elevation [m]

14 0.92 F of S
Slip 20

0.9

0.88
-2 -1 0 1 2
9
Lambda

Convergence
4
Convergence criterion: 0.001
0 5 10 15 20 25 30 Iterations: 30/100
Distance [m]
Figure 2. Slope stability results plot including convergence-related information
3 SEEPAGE ANALYSES For geotechnical engineering applications, unsaturation
means lower water flow in certain areas. The engineer will
Seepage analyses are very common in geotechnical want to account for this, in most cases. To do so, the
engineering, as they are often conducted alongside slope engineer will use a saturated/unsaturated flow model in the
stability or strain-stress analyses to calculate (or estimate) finite element software they use to simulate seepage. In
the pore-water pressure conditions in the domain. Water doing so, the engineer will once again face challenges that
flow in soils is driven by total head gradients and is scaled pertain to numerical processes.
by the hydraulic conductivity of the various soils water
percolates through. Hydraulic conductivity denotes how 3.1 Iterating for the pore-water pressure
quickly soil lets water flow through its pores. Its value
depends, among other things, on the soil’s coarseness and Equation 1 shows the finite element equation that needs to
the suction that might exist in the pore space, if be solved at every node in a domain. The equation is
unsaturated. Soils become unsaturated when the suction inherently very simple: it states that the product of the
that develops above the phreatic line becomes too great element characteristic matrix [𝐾] with the vector of nodal
for the soils to bear, and it releases some of the water unknowns {𝐻} (total head at each node) will result in the
contained in its pores. It is crucial for the geotechnical nodal forcing vector {𝑄} (the flow rate at each node).
engineer to recognize that in the vast majority of use cases
for geotechnical engineering applications, the soils through
which water flows are unsaturated. At greater depths, a [𝐾]{𝐻} = {𝑄} Equation 1
saturated zone might exist, but closer to the surface, where
most engineering projects live, soils will be either saturated
or unsaturated, depending on the location. The element characteristic matrix [𝐾] contains the known
As opposed to saturated soils, where all the soil’s pores values describing the soils needed to solve the finite
are filled with water, unsaturated soils contain a proportion element equation. Among these is the hydraulic
of air within their pores. This air restricts the flow of water, conductivity of the soil at the calculation node. If the soil is
as there isn’t a continuous saturated pathway for water to unsaturated, the value of hydraulic conductivity will depend
travel through rapidly. Instead, a water droplet must avoid on the matric suction (which is generally taken to be the
air bubbles, which slows down its progression through the negative of the pore-water pressure in gage conditions). In
soil. As a result, the hydraulic conductivity of unsaturated the vector of nodal unknowns {𝐻}, the total head is the sum
soils is lower than that of the very same soil at saturation. of the elevation charge and the pressure head, which can
And the more unsaturated a soil becomes (as matric be written in terms of pore-water pressure by dividing by
suction increases), the lower its hydraulic conductivity is the unit weight of water.
going to be. A typical set of volumetric water content and Therein lies the problem: the pore-water pressure is
hydraulic conductivity for unsaturated soils is shown in what the finite element equation tries to calculate (the
Figure 3 for reference, where the volumetric water content vector of nodal unknowns {𝐻}), but it is also an input to the
at saturation is 0.45 and the horizontal hydraulic calculation, via the hydraulic conductivity that informs the
conductivity at saturation is 1𝑥10−6 𝑚/𝑠. element characteristic matrix [𝐾]. It is the same conundrum
as for limit equilibrium slope stability analyses: it takes one
to know the other. To calculate the unsaturated hydraulic
conductivity, an estimation of the matric suction (or
negative pore-water pressure) is used. Then, the finite
element equation is solved and we obtain the pore-water
pressure. Are these pore-water pressure values the same
as those used to calculate the hydraulic conductivity?
Probably not. To ensure the validity of the solution
obtained, we must then use these new pore-water pressure
values to calculate the hydraulic conductivity once again,
solving the finite element equations one more time. Yet
again, once done, the resulting pore-water pressure values
will be compared to those from the previous step. If the
values from both steps are close enough to each other (i.e.
if the difference between the two is smaller than the
convergence criterion set by the user), then the solution is
accepted as final. This process can only go on for so long
though, and a maximum number of iterations needs to be
set to avoid eternal computation.

3.2 Iterating for the seepage face position

Figure 3. Typical unsaturated flow characteristics for soils
Another numerical process at work for seepage analyses
in some, but not all, numerical packages, relates to the
calculation of the position of seepage faces. There are
many geotechnical applications where the position of
seepage faces is unknown a priori, and the engineer will let Potential seepage
face review
the software calculate this for them. To do so, the software
will allow certain nodes in the domain, chosen by the
Boundary condition Boundary condition
engineer, to become either no-flow, or zero pressure head
boundary conditions. In a seepage analysis, a no-flow No flow Zero pressure head
boundary condition surrounds the domain and is the 𝑄=0 =0
condition that restricts water from exiting the domain. A
zero pressure head boundary condition is the exact Switch boundary
opposite: it allows water to flow freely in and out of the condition type
domain, until the pressure head is exactly zero at the +
location of the boundary condition. Trigger new
review
When the potential seepage face location is sought
after on a series of nodes on the edge of the domain (like Is Is
the downstream side of a dam for example), the solver can 0? 𝑄 0?
alternate between no-flow and zero pressure head
boundary conditions at will until satisfactory results are
obtained. For the sake of demonstration, let us assume the
solver randomly attributes each node where the review is Review
allowed to either one of the two possible boundary completed
conditions. Once the finite element equations are
satisfactorily solved and the numerical process detailed in Figure 4. Potential seepage face review procedure
the previous section is through, the solver will inspect all
the nodes where the potential seepage face position is
being reviewed. At these nodes, water can flow out of the 3.3 Proposed improvements for seepage analyses
domain (if the zero pressure head boundary condition was
applied by the solver), or the pore-water pressure can be Just as was the case for limit equilibrium slope stability
smaller or equal to zero (if the no-flow boundary condition analyses, the validity of the calculated pore-water pressure
was applied by the solver). Any node where these values from seepage analyses reported by geotechnical
conditions are not respected will be flagged by the solver. engineers depends on the state of convergence at the end
The flagged nodes will have their boundary condition of the calculation process. Whenever large hydraulic
switched to the opposite type and a new review will be conductivity contrasts exist between adjacent materials,
triggered. This procedure is illustrated in Figure 4. such as is often the case in embankment dams, it is not
As for any other numerical process, limits must be set uncommon that a higher number of iterations will be
by the user to control the behaviour of the potential needed to reach a converged solution. If the engineer is not
seepage face review. In this case, there are no vigilant, they might report erroneous calculation results.
convergence criteria to respect from one iteration to the Figure 5 shows such a situation. Reported on its own,
other: the solver instead performs “smaller or equal” without any indication about the state of convergence, it
checks. However, a maximum number of reviews allowed might very well be difficult for the reviewer to assess the
must be set so that the solver cannot enter an infinite validity of the analysis results. The suspicious rise of the
revision loop. piezometric surface as the dam’s downstream face is
nearing might be a giveaway in this specific case, but the
clues are not always that obvious.

70

60
Convergence Water Pressure
Convergence criterion ( ): 0.005m -100 - 0 kPa
Iterations: 500/500 0 - 100 kPa
50 Unconverged nodes: 5,166 100 - 200 kPa
Reviews: 3/3 200 - 300 kPa
300 - 400 kPa
40 400 - 500 kPa
Elevation [m]

30

20

10

-10
-5 15 35 55 75 95 115 135 155 175 195 215 235 255 275

Distance [m]

Figure 5. Seepage analysis result plot including convergence-related information

In Figure 5, the maximum number of iterations was impossible to describe each of these in detail in the context
reached on the last potential seepage face revision of this paper. Instead, general examples are considered to
completed. This means the pore-water pressure calculated favour discussing ideas rather than specifics.
in the domain might be slightly wrong, or completely wrong,
there is no way to easily tell without displaying the 4.1 Iterating for the nodal loads and displacements
unconverged nodes, or at the very least the number of
unconverged nodes. The maximum number of allowed Fundamentally, most of the resolution algorithms for
revisions for the positioning of the potential seepage face stress-strain finite element modelling will strive to resolve
was also reached in this case. This could spell no trouble, Equation 2, which takes the same form as Equation 1 for
as maybe the last revision done was conclusive, but seepage analyses. The element characteristic matrix [𝐾],
chances are more reviews might have been needed. As which includes the soil’s stiffness matrix, multiplies a vector
such, there is no way for the reviewer to know if the of nodal displacements {𝑑}, resulting in a vector of nodal
piezometric surface daylights where it should be on the loads {𝑅}. Just as for seepage where either head or flow
seepage face. could be specified at a node to calculate the counterpart
Together, these two deficiencies in convergence render resultant, the same goes for stress-strain: either
the results of the seepage analysis shown in Figure 5 displacements or loads can be applied to solve for the
untrustworthy. By failing to disclose the state of counterpart resultant.
convergence at the end of the analysis, the engineer who
reports these results doesn’t provide the necessary tools to
the reviewer to properly assess the quality and validity of [𝐾]{𝑑} = {𝑅} Equation 2
their work.
Fortunately, by simply displaying information relating to
convergence on figures reporting seepage simulation A typical solution scheme to solve Equation 2 for stress-
results, uncertainty can be avoided. The right-hand side of strain analyses is the Modified Newton-Raphson algorithm,
Figure 5 shows how this can be done. The chosen schematically shown in Figure 6.
convergence criterion, number of iterations performed and
allowed, number of unconverged nodes, and number of
reviews performed and allowed are displayed directly in the
figure. Similar to how a reviewer was easily able to judge
the stability results shown in Figure 2 as untrustworthy, the
same reviewer can now also make the same judgement
just as easily when it comes to Figure 5 and the
convergence information it displays.

4 STRESS-STRAIN ANALYSES

Stress-strain analyses are becoming more prevalent in

geotechnical engineering as the computing power of
computers continues to increase. The goal of such
analyses is to simulate the stress-strain behaviour of soils Figure 6. Modified Newton-Raphson algorithm (modified
in response to certain perturbations like loadings, pore- from Potts and Zdravković, 1999)
water pressure changes, excavations, and many others.
Just as seepage analyses require constitutive relationships
to describe the variation of hydraulic conductivity with pore- A load increment (Δ𝑅𝑖 ) is applied on a given calculation
water pressure changes and desaturation, stress-strain step, and the displacement increment (Δ𝑑𝑖 ) is sought after.
analyses require constitutive relationships to describe how The global stiffness matrix (𝐾0 ) is considered constant over
soils deform in response to stress changes. These the increment, even though the soil response is known to
analyses can become fairly complex since the stress-strain be non-linear (as the “true solution” line in Figure 6
behaviour of soils is a nuanced topic, where highlights). The resulting incremental displacement for the
compartmentalization of soil behaviour types is almost iteration (Δ𝑑1 ) is used to evaluate the strains at each
unavoidable, creating a plethora of constitutive models to integration point. The constitutive model uses these
choose from for the engineer. Most of these soil models will incremental strains to calculate the resulting stress
acknowledge the non-linear nature of soils’ stress-strain changes. The stress changes can finally be resolved into
behaviour in some capacity. This non-linear behaviour equivalent nodal forces and be compared to the externally
hinders our ability to solve the finite element equations applied loads. Both quantities will most probably not be
easily; iterative processes will once again occur during equal, within tolerance, on the first try (unless a linear
solving. elastic model is used) and a residual load will exist (Ψ1 ).
Many numerical simulation packages focus on On the subsequent iteration, the residual load vector is
simulating the stress-strain behaviour of soils. Each used as the driver of deformation, instead of the original
package employs its own choices of integration schemes vector of nodal loads (𝑅). The same process is then
to deal with iterative processes. Of course, it would be
repeated: calculating the incremental displacement (Δ𝑑2 ),
calculating the resulting strains at each node and
Relative Unbalanced Energy Error vs. Iteration
integrating the constitutive models for the stress changes,
1
resolving into nodal forces and comparing to the residual
loads (Ψ1 ). In the end, a new residual load vector is
obtained (Ψ 2 ), and the process will repeat itself until a 0.8

Relative Unbalanced Energy Error

satisfactory solution is obtained.
As for all the other numerical processes discussed in 0.6
this paper, there are two possible outcomes for the iterative
process described above: respecting a convergence
0.4
criterion (i.e. the residual load at the end of an iteration is
small enough to be disregarded); or reaching the maximum
number of iterations allowed. Once again, the validity of the 0.2

calculated results entirely depends on which of the two

outcomes prevailed. 0
The integration of the constitutive model along the 0 10 20 30 40 50

strain path that is occurring as part of the Modified Newton a) Iteration

Raphson scheme is a numerical process nested into the

original process. The non-linear nature of the stress-strain Relative Unbalanced Energy Error vs. Iteration
response requires the use of what is called a stress point 12

algorithm. A dedicated convergence criterion must then be

set for the stress update occurring during this phase of the 10

Relative Unbalanced Energy Error

resolution.
Figure 7 shows typical results one might encounter 8

when performing stress-strain analyses. In this figure,

convergence performance is reported via the relative 6

unbalanced energy error (RUEE): the product of the

residual load vector (Ψ) and the displacement increments 4

vector (Δ𝑑) on any given iteration, normalized by the same

product for the first iteration. Naturally, on the first iteration, 2

the RUEE is 1. On the subsequent iterations, the RUEE

should decrease, as the solution should tend toward a 0
0 10 20 30 40 50
complete redistribution of the unbalanced loads. Figure 7 Iteration
b)
a) portrays an analysis where the expected performance is
obtained: the RUEE continuously decreases and the Figure 7. Relative unbalanced energy error: a) well-
convergence criterion is met after 50 iterations. On the behaved convergence; b) ill-behaved convergence
other hand, Figure 7 b) shows problematic convergence
results: the RUEE initially decreases, but somehow
increases dramatically at various calculation points, even A more transparent way of reporting stress-strain analysis
reaching as high as 18 on the 35th iteration (meaning that results would look like Figure 8, where a contour map is
the unbalanced energy was momentarily 18 times higher accompanied by a convergence plot where the relative
than after the first iteration). After 50 iterations, the process unbalanced energy error is shown for all calculation steps.
stops, as the maximum number of iterations was reached. The convergence settings for the analysis are also
At that point, the RUEE was still 1.3, which means the final indicated in the figure so that the reviewer can appreciate
solution reported by the software is worst at the end of the that the maximum number of iterations reached on any
iterative process than at the first iteration. This would be a calculation step never exceeded the maximum allowed.
great source of concern for the engineer that reports the The convergence criteria set for each part of the resolution
stress-strain simulation results and should be adequately scheme are also indicated for transparency. The reviewer
displayed in the report for the benefit of the reviewer. would then be provided with the opportunity to judge if
these criteria were set purposely large (to favour easier
4.2 Proposed improvements for stress-strain analyses convergence, at the cost of precision).

As for the slope stability and seepage analyses, the results

of stress-strain numerical simulation analyses should 5 THE ROLE AND RESPONSIBILITIES OF
ideally be reported together with clear indications of SOFTWARE PROVIDERS
convergence performance to allow a reviewer to trust the
results they are interpreting. On their own, contour plots of Engineers are the sole responsible for the quality and
deformations or stresses don’t provide any indication validity of the work they perform. The tools they use in their
regarding the validity of the solution. An unsuspecting work, like numerical analyses, are merely that: tools. A tool
engineer might overlook convergence results such as can be used correctly or incorrectly, it all depends on the
those shown in Figure 7 b) and report results they believe user. Incorrect use of a tool doesn’t mean the tool is flawed.
are correct, because this was the output provided by the However, some tools are better than others to perform
software. certain tasks, and tools can be enhanced to simplify said
tasks.
 X-Displacement Convergence
≤ -1.2 - -0.9 m 1
-0.9 - -0.6 m

Relative Unbalanced Energy Error

14
-0.6 - -0.3 m
0.8
-0.3 - 0 m
12 0 - 0.3 m 1 days
0.3 - 0.6 m 0.6
10 ≥ 0.6 m
2 days
0.4
Elevation [m]

0.2 3 days
6

4 0
0 20 40 60 80 100 120

2 Iteration

0 Convergence
Scheme: Unbalanced energy
-2 Maximum number of iterations: 300
-3 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42
Tolerable relative error: 0.0001
Distance [m] Tolerable error for stress update: 0.005

Figure 8. Stress-strain analysis result plot including convergence-related information

The role of software providers when it comes to numerical 6 CONCLUSION

analyses is to provide the best tools possible so engineers
can perform their work swiftly and rigorously. Their main Through the lenses of limit equilibrium slope stability,
responsibility on the other hand is to ensure the validity of seepage and stress-strain analyses, convergence was
the solutions they implement in their software. There is a discussed in this paper. Each of these types of numerical
tacit agreement between the engineer and the software analyses includes some form of iterative process where,
provider: the software should be exempt from faults. ideally, the convergence criteria are met within the
In the specific context of this paper, another maximum number of iterations allowed. For many reasons,
responsibility of software providers is manifest: enhance this is however not always the case. Which scenario
the tools they provide based on the evolution of the prevailed matters greatly to the reviewer of a geotechnical
geotechnical practice. As new trends are set by practicing report including numerical analyses, or at least so it should.
engineers in the way they perform their work, software To promote greater confidence in the results of
providers must adapt and shape their tools accordingly. It numerical analyses shown in geotechnical reports, it was
is a question of survival for their business. proposed that convergence results should always be
The main theme of this paper is convergence; disclosed together with the results figures. Depending on
specifically, how convergence results should ideally be the type of analysis considered, this disclosure can take
reported to promote transparency. The question begs itself various forms and may vary depending on the software
then: should software providers enhance their tools to package used, since the solution schemes might differ.
promote better convergence communication? In an ideal
world, they would, which could potentially greatly benefit
the geotechnical profession. However, engineers must 7 ACKNOWLEDGEMENTS
realize that software providers are businesses; while they
continuously try to offer the best tools possible to their The author wishes to acknowledge the contribution of
users, their primary goal is still profitability. If a feature is Pierre-Alain Konrad, who accepted to review this paper
seldom requested by their users, they might not be willing and provided useful suggestions to enhance its content.
to allocate the time and resources necessary to develop
the feature. Conversely, if engineers start to give
convergence the attention it warrants, software providers 8 REFERENCES
will have no choice but to develop better solutions to report
convergence results. Otherwise, they run the risk of Fellenius W. 1936, Calculation of the Stability of Earth
becoming obsolete and being replaced by more agile Dams. Second Congress on Large Dams, Washington,
competitors. DC, 14: 445-463.
Such desirable improvements could take many forms. Morgenstern, N.R. and Price, V.E. 1965, The Analysis of
For example, a functionality to overlay convergence plots the Stability of General Slip Surfaces, Géotechnique,
(or any plot for that matter) together with contour drawings. 15(1): 79-93.
Or a functionality to display the user-defined convergence Potts, D.M. and Zdravković, L. 1999. Finite Element
criteria versus the performance of the analyses (i.e.: 200 Analysis in Geotechnical Engineering: Theory. Thomas
iterations necessary out of the 300 allowed). The options Telford Publishing, London, UK.
are plentiful, it is up to the engineers to demand these Seequent (2023). GeoStudio, version 2023.1. Retrieved
changes from software providers. from http:www.geoslope.com.
Spencer, E. 1967, A Method of Analysis of Embankments
Assuming Parallel Inter-Slice Forces, Géotechnique,
17(1): 11-26
Incorporation of measured pore-water
pressure readings from Gardiner Dam for
3D deformation modelling
Kevin Gelleta, Vicki Nguyen, Joseph Quinn & Daniel G. Bertrand
Klohn Crippen Berger Ltd., Calgary, Alberta, Canada
Kevin Mathison & Keisha Henry
Water Security Agency, Regina, Saskatchewan, Canada

ABSTRACT
A study is currently underway to develop 3D (three-dimensional) finite-volume stress-strain numerical models to enhance
the understanding of the Gardiner Dam’s historic and future geotechnical performance. A key input for calibrating such
models to the field performance is the accurate replication of historic pore-water pressures (PWP). To do this, PWP
distributions were modelled directly from piezometer measurements collected during the construction and operation of the
dam. The measured PWPs were inputted at the location of the various piezometer tips with interpolation done between
each tip according to the technique developed by De Alencar et al. (1992). Given the observed relationship between PWP
spikes with fill placement, PWPs were interpolated between the tips in terms of their 𝐵̅ or 𝑟𝑢 response, rather than absolute
pressure. With this methodology, the resulting PWP regime became reflective of the site conditions, and the overall
calibration of the model deformations was found to improve.

RÉSUMÉ
Une étude est en cours pour développer des modèles numériques 3D (tridimensionnels) de contrainte-déformation par
volume-fini et ce, afin d’améliorer la compréhension des performances géotechniques historiques et futures du barrage
Gardiner. Un élément clé pour calibrer ces modèles aux performances sur le terrain est la reproduction précise des
pressions interstitielles historiques (PWP). Pour ce faire, les distributions PWP ont été modélisées directement à partir des
mesures piézométriques recueillies lors de la construction et de l’exploitation du barrage. Les PWP mesurées ont été
saisies à l’emplacement des piézomètres, et une interpolation a été effectuée entre chaque instrument selon la technique
développée par De Alencar et al. (1992). Compte tenu de la relation observée entre les pressions interstitielles mesurées
et le remblai mis en place, les pressions interstitielles ont été modélisées au moyen de valeurs 𝐵̅ ou 𝑟𝑢 , plutôt qu'en termes
de pressions absolues. Avec cette méthodologie, les pressions résultantes reflètent les conditions changeantes du site, et
une amélioration de l'étalonnage global des déformations des modèles numériques a été observée.

1 PROJECT BACKGROUND geotechnical engineering, the Gardiner Dam was one of

the earliest large dams constructed according to the
The Gardiner dam is located on the South Saskatchewan general principles of the observational method before this
River approximately 100 km south of the city of Saskatoon, technique was more clearly exemplified by Peck (1969). As
as shown in Figure 1. The dam is situated in a semi-arid, a result of this approach, an extensive instrumentation
drought-prone region of the North American Great Plains. network consisting of piezometers, slope inclinometers,
In 1935, the federal government of Canada established the settlement gauges, and survey pins was established during
Prairie Farm Rehabilitation Administration (PFRA) with a the construction of the dam which unveiled unexpectedly
mandate to address soil erosion and water resources high pore-water pressures and large shear displacements
required for agricultural development in the drought- in the clay shale foundation.
affected areas of Manitoba, Saskatchewan, and Alberta Despite the challenges, the dam was successfully
(Agriculture and Agri-Food Canada 2010). In pursuit of this completed in 1967 with revised slopes as shallow as
goal, between 1943 and 1946 the PFRA began site 85H:1V in areas, resulting in it being one of the largest
investigations for a dam at various locations along the hydro dams (ranked by volume of earthfill) that exist in the
South Saskatchewan River, and started construction on world today. The displacements which were first observed
Gardiner Dam in 1959, on what would eventually become during construction have continued to occur to the present
its largest project (PFRA 1980). day, although at a greatly attenuated rate that has not
The construction of the Gardiner dam was a compromised the operation of the facility to-date. However,
monumental undertaking for its time and the PFRA the team of the aforementioned engineers indicated that
retained a team of world-renowned geotechnical more information and research were needed to fully
engineering consultants during the investigation, design, understand the long-term behaviour of Gardiner Dam.
construction, and operation stages including A. Today, the Gardiner Dam is operated by the Water
Casagrande, W. Johnson, K. Terzaghi, and later N. Security Agency (WSA), previously known as the
Morgenstern, R. Peck, S. Wilson, and E. Klohn (Nguyen et Saskatchewan Watershed Authority, who took over
al. 2022). Constructed during an era of rapid advances in responsibility for the structure from the PFRA in 1997
(Agriculture and Agri-Food Canada 2010). The Gardiner
dam remains a vital asset for the province of
Saskatchewan. The presence of the dam stabilizes the
flows of the South Saskatchewan River and generates
hydroelectric power, while the reservoir (Lake Diefenbaker)
provides a reliable irrigation source for agriculture and
creates a recreational area for nearby residents.
Given the importance of the facility, a study initiated by
WSA is currently underway to complete a comprehensive
stability evaluation, deformation analysis, and risk
assessment of Gardiner Dam to help understand the long-
term behaviour of the dam.

Figure 2. Gardiner Dam aerial photo

2.1 Model development and geology

The Gardiner dam is unique among other dams from its era
due to its extensive and well-kept records. Key references
such as the PFRA “Red Book” (1980), publications by
Jaspar and Peters (1979), Morgenstern and Simmons
(1980), Rahman and Kilgour (2000), Powell (2010), and
Scammell (2013) were used to gain an appreciation of the
site geology and suitable material parameters. As per the
process described by Nguyen et al. (2022), the 3D
Figure 1. Geographic location of the Gardiner Dam 100 km geometries of the models were constructed based on a
south of Saskatoon, Saskatchewan, Canada. review of hard-copy as-built drawings, construction
records, contour maps, photographs, topographic surveys,
and borehole logs. The location and thickness of weak
2 3D DEFORMATION MODELS shear zones were inferred from the slope inclinometer
data.
Two 3D finite-volume stress-strain numerical models have A simplified geologic cross section, taken from PFRA
been developed using the computer software FLAC3D (1980), along the Gardiner Dam centreline is shown in
(Itasca 2022). The purpose of the models is to enhance the Figure 3, while the location of the cross-section is shown
current understanding of the dam’s historical geotechnical by the dashed yellow line on Figure 2. The Coteau model
performance and serve as a predictive tool to help assess is centred at the area of a former creek valley and the
the potential impacts of future displacements and/or embankment in this area is built atop deposits of
changes to the operation of the facility. The models were eroded/slumped clay shale of Bearpaw Formation origin.
constructed at the two primary areas of observed Similarly, at the East Embankment of the River model there
deformations to-date. The first model encompasses both is also a large deposit of slumped Bearpaw clay shale
the Coteau Embankment and Coteau Abutment (referred which the embankment is built over. At the area of the River
to hereafter as the ‘Coteau Model’) while the second (Main) Embankment is the location of the former free-
encompasses both the Main (River) Embankment and East flowing South Saskatchewan river and the embankment is
Abutment (the ‘River model’) as shown in Figure 2. built atop a large deposit of river sand, under which intact
bearpaw clay shale is present.
Figure 3. Simplified Geologic Cross-Section along the Gardiner Dam Centreline (Perspective looking downstream) (PFRA
1980)

2.2 Model calibration and the impact of pore-water on theoretical or empirical relationships for modelling the
pressures historic pore-water pressures, they were inputted directly,
as measured by the numerous piezometers. This direct
Once an initial set of material boundaries and parameters method of modelling the pore-water pressures is the same
were defined, the models were analyzed and allowed to approach adopted as part of the previously successful 3D
deform. The strength(s) of the shear zone(s) were then finite difference deformation modelling of the Grant Devine
further calibrated by comparing the predicted deformations Dam, formerly known as the Alameda Dam (Quinn et al.
with the observed values. This process was repeated until 2015).
a reasonable match was met. The models were run as
effective stress analyses with the initial pore-water 2.3 Historical performance
pressures assumed to be hydrostatic with the water table
at the original ground surface. Analysis of the model was As mentioned, the understanding of the performance of the
executed in successive ‘stages’ which corresponded to the Gardiner Dam is informed by a large network of
conditions at different dates during the construction and piezometers, slope inclinometers, settlement gauges, and
operation of the dam. The pore-water pressures were then several sets of survey pins that are aligned together to form
varied across the stages as the construction and operation ‘movement lines’ (ML). Figure 4 illustrates the location of
of the dam progressed. the piezometers installed in the clay shale and located
It was realized early in the calibration process that the within the domains of the Coteau and River models, along
predicted displacements and settlements were highly with the deformation monitoring instruments.
dependent on the adopted pore-water pressures, In general, the performance of the Gardiner Dam has
particularly the pore-water pressures within the shale been marked by high excess pore-water pressures,
foundation where the largest displacements were focused. generated during the construction, that are slow to
As such, a high degree of importance was given to the task dissipate, and large horizontal displacements concentrated
of accurately modelling the pore-water pressures in the along discrete shear zones within the clay shale
model. As the modelling results were also dependent on foundation. At the Coteau model and the East
several other inputs (i.e., dam geometry, material strength, Embankment of the River model, these shear zones are
material stiffness), reducing uncertainty in the pore-water focused within the slumped shale. At the Main
pressure assumption improved the confidence of the Embankment of the River model, the shearing is focused
calibrations for the other inputs. in the intact shale at a lower elevation below the river sand.
Fortunately, and thanks to its observational approach- In all cases, above the shear zones, the incremental
based design, the Gardiner Dam is remarkable among displacements are generally minimal, which suggests that
facilities of its age for its extensive, over half-century old the dam and foundation are sliding along the shear planes
piezometer network which has been largely maintained to as a block, and little to no straining is occurring in the dam
the current day. As such, there is a relatively good fill. Displacement magnitudes are typically greatest just
understanding of the historic pore-water pressure downstream of the dam crest and then attenuate
response at the site with time. Therefore, instead of relying downstream to zero displacements. The displacements
observed during construction have continued to progress
to the present day. However, the rates of the on-going 2.4 Piezometer data review
displacements throughout the Gardiner Dam site have
greatly attenuated since they were first triggered (or Given the large impact the adopted pore-water pressures
reactivated) by construction activity and now the had on the modelling results, significant effort was
fluctuations in the movement rates seem to be related to expended to evaluate the appropriateness of the
the fluctuating reservoir level, with small, incremental piezometric data used as inputs in the model. It must be
displacements occurring over each year as the water level acknowledged that there were numerous difficulties in
is raised to its seasonal high. working with instrumentation data from the 1960s to the
The excess pore-water pressures in the shale were 1990s and collected by engineers and technicians who are
generally highest below the crest and gradually reduced no longer available. Upon a review of the piezometer
towards the downstream toe, as expected in relation to the records, many instances of unexplained and sudden
dam loading (i.e., fill height). Piezometers installed within decreases or increases in the pore water pressure could
the lower elevations of the fill showed excess pore-water be seen in the data from many of the piezometers. These
pressures generated during construction, whereas the instances were investigated, and it was concluded that the
piezometers installed within the higher elevations of the fill readings were most likely erroneous, rather than a genuine
did not. In addition, post reservoir filling data indicate that measurement of rapidly changing field conditions. The
steady-state seepage conditions have not yet been cause of such errors is likely due to various issues
reached as a large portion of the upstream fill (i.e., including, but not limited to, staff turnover (e.g., variations
upstream of the downstream pervious zone) is not yet in data interpretation and variations in procedures used
saturated. Within the river sand, pore-water pressures making the readings), instrument maintenance (e.g.,
appeared to rise slightly with the reservoir upon the first flushing of air bubbles within the lines, replacement of
filling. Since the first filling, pore-water pressure parts), and possible pumping tests in the vicinity of the
measurements in the river sand have remained low piezometer tips. In these instances, judgement was
(relative to the reservoir) and generally consistent, showing exercised to adjust the pore-water pressures at these
no or relatively small changes in pore-water pressure instruments for these anomalous periods.
compared to the magnitude of the changing reservoir. The As an example of this data adjustment, Figure 5 shows
impermeable blanket, constructed on the upstream slope nearly 60 years of readings collected for piezometer
of the dam slope is thought to be responsible for the overall ‘BP10’, installed in the shale foundation, and located
muted response of the river sand to the reservoir. downstream of the crest at the River embankment (see
Figure 4). The pore-water pressure response seen from
this instrument is very typical of the shale piezometers
seen across the site, with large spikes in pore-water
pressure over the period of fill placement, followed by a
long-term trend of dissipation. In 2006, outlier piezometric
elevation readings with no apparent explanation were
taken for this instrument. Furthermore, nearby tips did not
measure a similar dip in pore-water pressure. To avoid
modelling uncharacteristically low pore-pressures in the
model at this area, the inputted pore-water pressure values
were corrected to match the long-term trendline over this
period.

570

560

550
Elevation (m)

540

530

520

510
Model PWP made to
follow the overall trend
500

490
Jan, 1959 Dec, 1973 Dec, 1988 Jan, 2004 Jan, 2019
Date
Piezometric Elevation (m) Reservoir El. (m)
Tip El. (m) Ground Elevation (m)
FLAC PWP at Tip
Figure 4. Gardiner Dam instrumentation network at the
Figure 5. Piezometer BP10
Coteau (top) and River (bottom) models
2.5 Pore-water pressure interpolation 𝑈𝑗∗
𝐹𝑗∗ = 𝑢𝑗∗
[3]
Once the piezometric data was screened, pore-water
pressures were inputted directly into the models at the
location of the readings (i.e., at the piezometer tips).
Where 𝐹𝑗∗ is the correction factor (or error) at
Between the piezometer tips, pore-water pressure values
were interpolated according to the technique developed by piezometer 𝑗, 𝑈𝑗∗ is the measured pore-water pressure at
De Alencar et al. (1992) which is summarized below: piezometer 𝑗, and 𝑢𝑗∗ is the initial pore-water pressure
assumption at piezometer 𝑗. The pore-water pressure was
• An initial assumed pore-water pressure then adjusted at the model grid points by first calculating a
distribution is assigned throughout the model. In weighted ‘interpolation coefficient’ based on the distance
the case of finite element/volume/difference of the nearby piezometer readings. The weighted impact
analysis, this means that an initial set of values of these piezometers to a given grid point was based on
are assigned for each integration grid point (or the following formula:
node) of the mesh.
• The error of this initial assumption with respect to
the measured pore-water pressure value is 1
calculated at each piezometer location. 𝑑𝑖𝑗
• The assumed (pore-water pressure) distribution λ𝑖𝑗 = 1 [4]
∑𝑚=4
𝑗=1 𝑑𝑛
(at the mesh grid points) is then corrected as a 𝑖𝑗
function of these calculated errors through an
iterative process until a certain tolerance is
reached (i.e., until a measure of the errors Where 𝑚 is the number of piezometers being used in
becomes smaller than a specified value). The the interpolation (equal to the 4 nearest piezometers with
closer the initial guess assumed is to the real readings in this study), λ𝑖𝑗 is the interpolation coefficient at
distribution, the less a correction is needed. grid point 𝑖 based on one of the four piezometers 𝑗, 𝑑𝑖𝑗 is
the distance between the grid point 𝑖 and one of the four
With modern advances in computing power, the
piezometer tips 𝑗, and 𝑛 is an exponent factor which effects
implementation of the above method into numerical models
is relatively fast, even for large models like the ones the relative impact of the piezometers with respect to their
created for Gardiner Dam. distance (𝑛 was taken as 5 in this study and higher
The De Alencar et al. (1992) technique is intentionally exponents result in nearer piezometers having higher
flexible and can be tailored as necessary for specific relative impacts). The interpolation coefficient, λ𝑖𝑗 , is
applications. Given that the key soil units within and below expressed as a percentage (or decimal) and is calculated
Gardiner Dam have significantly different pore water for each of the 𝑚 piezometers used in the interpolation. The
pressure responses, the interpolation for each soil unit was sum of the interpolation coefficients at a grid point is equal
performed separately based on only the piezometers to 100% (or 1.0). Once the four interpolation coefficients
founded within that soil unit (e.g., shale, river sand, dam fill, have been calculated, a correction factor for the pore-water
etc.). Furthermore, as the pore-water pressures at pressure at a given grid point was calculated based on sum
Gardiner Dam are clearly correlated to the timing and of the product of the interpolation coefficients and the
quantity of dam fill placed (Figure 5), the pore-water piezometer errors:
pressures were interpolated between the piezometer tips
in terms of their 𝐵̅ (for the foundation materials) or 𝑟𝑢 (for
the fills) as defined below, rather than absolute pressure. 𝐹𝑖 = ∑𝑚=4
𝑗=1 λ𝑖𝑗 𝐹𝑗
∗
[5]
The ‘initial guess’ pore-water pressure assumption was the
average 𝐵̅ or 𝑟𝑢 value as measured by all the piezometers
with readings in each of the respective material units.
𝑢𝑖 = 𝐹𝑖 𝑢𝑖−𝑖𝑛𝑖𝑡𝑖𝑎𝑙 [6]

∆𝒖
𝐵̅ = ∆𝝈 [1] Where 𝐹𝑖 is the pore-water pressure correction factor at
𝒗 grid point 𝑖, 𝑢𝑖−𝑖𝑛𝑖𝑡𝑖𝑎𝑙 is the initial pore-water pressure at the
grid point as calculated from the average 𝐵̅ or 𝑟𝑢 value, and
𝒖 𝑢𝑖 is the final adjusted pore-water pressure at the grid point.
𝑟𝑢 = 𝝈 [2] After the pore-pressures at the grid points were adjusted,
𝒗
the correction factor at each piezometer, 𝐹𝑗∗ , was
recalculated based on the updated pore-water pressure
Once the initial pore-water pressure distribution is regime, and if the correction factor exceeded a certain
defined for each material unit in the model, the error at each tolerance, the process was repeated.
piezometer tip was calculated by: The pore-water pressure regime using this
methodology results in a very close match to the measured
data at the location of the piezometer tips with readings, measured data density of the pore-water pressure regime
and interpolated values between these piezometers tips as is relatively lower during the earliest construction stages,
shown by Figure 6. As such, the reliability of the pore-water compared to at the end of construction. A plot showing the
pressure regime in the model is high at the location of number of available piezometer readings (for all materials)
piezometer readings and become less certain farther from within the domain of the River model is given in Figure 8. A
the piezometers. Therefore, the confidence in the pore- similar trend was also noted for the Coteau model.
pressure regime is higher in areas with a higher density of In some cases, it was found that piezometers were
available piezometer readings such as just downstream of installed in high pore-water pressure regions well after the
the crest and through the center of the embankments fill placement had occurred. This meant that the high pore-
(Figure 4). Fortunately, at Gardiner Dam, the areas of the water pressures in the model were not incorporated until
largest movements and highest pore-water pressures are the date of the first readings (some time after the fill
also the most highly instrumented areas. Figure 7 shows a placement). As these high pore-water pressure zones were
plan view of the pore-water pressure measurements in the found to significantly impact the model deformations, in
shale at the end of construction in the domain of the River these cases it was found that calibration was improved by
model. Of note on Figure 7 is the high pore-water ‘back-projecting’ the high pore-water pressure
pressures (𝐵̅ > 1) present at the centre of the model at the measurements to the timing of the fill placement. An
location of the main embankment (Figure 2). example of this back-projection is illustrated in Figure 9 for
Unsurprisingly, it is at this area in the clay shale under the piezometer ‘CP557’, located downstream of the Coteau
riversand deposit (Figure 3) where the largest shear embankment (Figure 4) in the Coteau model.
displacements have been observed to date at Gardiner
Dam.
250

# of Available Piezometer Readings

Blue node is a piezometer without a
reading at the shown modelling stage.
Piezometer is not used as part of the 200
interpolation.
150

100

50

0
1959 1973 1988 2004 2019
Figure 6. 2D cross section showing the interpolated pore- Year
water pressure contours in the 3D model
# of available piezometer readings at River and
East (All material types)
End of Construction

Figure 8. Number of available piezometer readings (for all

materials) within the domain of the River model

590
580
570
Elevation (m)

560
550
540
530 High FLAC PWPs
back-projected to
520 timing of fill
510 placement
Figure 7. Plan view of measured shale pore-water 500
pressures at the end of construction at the River model. Jan, 1959 Jan, 1964 Jan, 1969 Jan, 1974 Jan, 1979
Date
2.6 Pore-water pressure back-projection CP557 Reservoir El. (m)
Tip El. (m) Ground Elevation (m)
It is important to note that at Gardiner Dam, the FLAC PWP at Tip
piezometers were initialized at various times, including Figure 9. Back-projected high FLAC pore-water pressures
throughout construction, which lasted from 1959 to 1967, at piezometer CP557 of the Coteau model
as well as during the operational phase of the facility. In
general, the overall piezometer density was lowest during
the early period of construction and increased greatly
towards the end of construction. This means that the
3 RESULTS AND MODELLING STATUS simulated deformations and the measured performance.
The efficacy of this type of analysis relies on the
The overall objective of the calibration process is to establishment and maintenance of a large instrumentation
achieve a state where the model’s predicted deformations network, along with good record-keeping during and after
can reasonably match the measured deformations using construction by the facility operator.
values for inputs that are reasonable and consistent with
the measured field and laboratory data. This state has
been termed ‘model equivalence’ by the external review 5 ACKNOWLEDGEMENTS
board overseeing the study.
Using the described piezometer data screening The authors of this paper would like to thank WSA for their
process, interpolation technique, and selective use of back- support, as well as the members of Gardiner Dam review
projections, the calibration of the models was found to panels for their valuable insights throughout the course of
significantly improve and an acceptable model equivalence this project, based on their extensive experience. The
of the Coteau and River models was obtained. Figure 10 Klohn Crippen Berger Ltd. internal review panel consists of
shows an example of the results comparing the model Bill Chin and Bryan Watts, and the WSA Engineering
predicted shear zone displacements against the measured external review panel members include Dr. Norbert
displacements at slope inclinometer 3944 located Morgenstern, Anthony Rattue, and Stephen Rigbey.
downstream of the Coteau embankment in the Coteau
model (Figure 4).
As noted in the introduction, the early engineers 6 REFERENCES
involved with the design and analysis of the Gardiner dam
indicated that additional research was needed to fully Agriculture and Agri-Food Canada. 2010. PFRA 1935 -
understand its long-term behaviour. It is in pursuit of this AESB 2010: 75 years. Agri-Environment Services
goal that geotechnical models of the Gardiner Dam have Branch, Ottawa, Ontario, ON, Canada.
been developed and analyzed. The calibrated models are De Alencar, J.A. Chan, D.H. and Morgenstern, N.R. 1992.
now being used to assess the future expected performance Incorporation of Measured Pore Pressure in the Finite
of the dam. The models are also being used in a predictive Element Analysis. Pressures in the Finite Element
fashion, such as gauging the impact of different reservoir Analysis. Proceedings, 45th Canadian Geotechnical
raising/lowering rates and reservoir elevations on the Conference, Toronto, Ontario
performance of the dam. The conclusions of the study are Itasca Consulting Group, Inc. 2018. FLAC3D Version
ultimately intended to inform a risk assessment and 7.00.137 [computer software]
planning for future operation and maintenance of the Jaspar, J.L. and N. Peters. 1979. Foundation Performance
facility. of Gardiner Dam. Canadian Geotechnical Journal,
16(4): 758-88.
1000 560 Morgenstern, N.R. and J.V. Simmons. 1980. A
900 Deformation Analysis of Gardiner Dam, South
800
550 Saskatchewan River Project, Report to Chief Engineer.
Prairie Farm Rehabilitation Administration, Regina,
Displacement (mm)

700
540 Saskatchewan, University of Alberta.
Elevation (m)

600
Nguyen, V. Quinn, J. Gelleta, K. Bertrand, D.G. and
500 530 Bhuiyan, I. Attaining Model Equivalence – Calibrating
400 Material Parameters for Gardiner Dam with 3D
520
300 Numerical Modelling. Proceedings, GeoCalgary 2022.
200 Calgary, Alberta, AB, Canada.
510
100 Peck, R.B. 1969. Advantages and Limitations of the
0 500
Observational Method in Applied Soil Mechanics.
1959 1974 1989 2004 2019 Géotechnique, 19(2): 171-187.
Date Prairie Farm Rehabilitation Administration. 1980. The
Resultant (Post-Con. Meas.)
Design and Construction of Gardiner Dam and
Resultant (Post-Con. FLAC) Associated Works. Canadian Government Publishing
Reservoir Centre. Hull, Quebec, QC, Canada.
End of Construction
Powell, J.S. 2010. Geotechnical Characterization of the
Figure 10. Measured and calibrated post-construction Bearspaw Shale. Queen’s University.
model displacements at SI 3944 of Coteau Quinn, J. Chin, B. Pernito, M. and Scammell, J. 2015.
Geotechnical Assessment of the Alameda
Dam. Canadian Dam Association Bulletin, Spring
4 CONCLUSION 2015, 26(2): 10-23.
Rahman, M.G. and D.A. Kilgour. 2000. Gardiner Dam -
The direct implementation of a large and well-screened Three Decades of Performance Monitoring. Canadian
database of measured pore-water pressure readings as Dam Association 3rd Annual Conference, Regina,
inputs into numerical deformation models, using the Saskatchewan, SK, Canada. 1: 212-224.
interpolation method developed by De Alencar et al.
(1992), resulted in an improved match between the
Scammell, J.W. 2013. Evaluation of Gardiner Dam's
Ongoing Movement. University of Saskatchewan.
West Val Marie Dam Spillway –
Uplift Mitigation Improvements

Jorge Gutierrez, P.Eng., P.E., Michael Grisdale, P.Eng.

Klohn Crippen Berger Ltd, Saskatoon, Saskatchewan, Canada
Shifullah Khaled, P.Eng., PMP
Water Security Agency, Regina, Saskatchewan, Canada

ABSTRACT
The concrete spillway (the spillway) to the West Val Marie Dam (the dam) was built in the 1950s and was designed with
a heavy reliance on the ability of its under-slab drainage system (UDS) to mitigate hydraulic jump uplift forces. The
existing UDS was constructed with materials now exceeding their intended service life and needed repairs, posing a
high risk of failure due to uplift forces during high-flow events. Various measures were explored to mitigate these
vulnerabilities and associated risks, with a preference for passive measures. The selected improvements included a
partial decommissioning of the existing UDS, installing 44 new shallow point drains, and installing 138 soil anchors with
a bond length of 7 m. This paper discusses processes for weighing long-term costs against short-term risks, selecting,
and designing mitigation measures, and assessing construction risks.

RÉSUMÉ
Le fonctionnement du déversoir en béton du Barrage Ouest Val Marie, construit dans les années 1950, dépend
fortement de l’efficacité de son système de drainage sous dalle (DSD) à atténuer les sous-pressions hydrauliques. Vu
que les matériaux du DSD dépassent leur durée vie utile, ceci présente des risques de soulèvement en cas de
déversements importants. Diverses mesures ont été explorées pour remédier à cette situation, avec une préférence
marquée pour des solutions passives. Les améliorations retenues comprennent le démantèlement partiel du DSD
existant, la mise en place de 44 nouveaux drains à faible profondeur, et l'installation de 138 ancrages au sol d’une
longueur effective de 7 m. Cette publication traite des processus utilisés pour évaluer les coûts à long terme par rapport
aux risques à court terme, de choisir et concevoir les solutions de réhabilitation et d’évaluer les risques liés à la
construction.

1 INTRODUCTION also functions as a balancing reservoir to meet

international apportionment requirements on the
1.1 General Frenchman River watershed. The reservoir capacity is
3,578 dam3 at full supply level (FSL) of 815.72 m. The dam
This document focuses in two general themes of the discharge facilities include an emergency earth-cut
related work including the overall strategy developed in spillway, a gated riparian low-level outlet, and a gated
collaboration with the Saskatchewan Water Security irrigation outlet in addition to the spillway. Based on the
Agency (WSA) to balance risks and priorities; and certain Canadian Dam Association (CDA) guidelines, and the
elements of the geotechnical, structural and hydrotechnical WSA dam consequence classification guideline, the
engineering design work. recommended consequence classification for the dam is
‘VERY HIGH’ for the evaluation of all infrastructure and
1.2 Project Background overall management oversight for the project, and ‘HIGH’
for inflow design flood (IDF) selection. Accordingly, a flood
The WSA became the owner of the dam as part of the event of 1/3 between the 1000-yr flood and the probable
Federal Dam Transfer between WSA and Agriculture Agri- maximum flood (PMF) has been adopted as the project’s
Food Canada (AAFC). The dam is located on the IDF with a rate of 1,185 m3/s (Hatch 2021).
Frenchman River, approximately 23 km northwest of the
Town of Val Marie and 13 km upstream of the Val Marie Previous dam safety reviews have indicated the
Dam. The homogenous earthfill dam with an original height effectiveness of the UDS to be questionable due to
and length of 8.2 m and 915 m respectively, was originally missing, collapsed, or disconnected piping. Based on
constructed in 1939 by the Prairie Farm Rehabilitation these recommendations, WSA assessed that a
Administration (PFRA) – now AAFC. Modifications to the replacement system was required. A target factor of safety
embankment were made in 1949 and 1959 by raising the (FoS) against uplift was to be established and a FoS was
original crest to an elevation of 817.4 m (geodetic, typical), to be estimated. WSA commissioned Klohn Crippen
resulting in a current maximum height of 9.4 m. The Berger Ltd. (KCB) to analyze the condition of the spillway
spillway was built in 1939 and rebuilt in 1959 after it was and to engineer the required upgrades. The engineering of
significantly damaged in the c. 1952 flood. the project began in June 2021 with substantial
construction completion achieved in February 2023. This
The reservoir supplies irrigation water, supplements paper discusses the design of uplift mitigation measures
downstream irrigation demand, is a source of domestic needed to attain the target uplift stability FoS for spillway
water, is used for summer and winter recreational uses. It design discharge of 217 m3/s.
2 ANALYSIS AND DESIGN PARAMETERS and long-term needs, with allocation of funds and
implementation timelines. The tailwater level is a
2.1 Spillway Design Capacity necessary parameter for the development of the hydraulic
jump profile, and the associated level of 812.86 m was
The spillway consists of a 5-bay control section with radial used for this assessment. Other discharge combinations
gates seated on an ogee weir discharging into a sloped were studied, as discussed in Table 2, but this was used
4H:1V chute which terminates without a stilling basin. The as the reasonably extreme condition for the project
spillway is a continuous sloping chute and directly objectives.
discharges into the outlet channel. The lower portion of the
chute is equipped with chute blocks and an end sill to assist 2.5 Target Factors of Safety
in energy dissipation. The spillway original design
discharge capacity of 217 m3/s was selected as the KCB recommended that a target Factor of Safety (FoS) be
required discharge for uplift stability assessment; however, established (Table 1). These FoS were based on the return
if the reservoir level reaches the crest of the dam, the period of the related loading combination.
control section can pass up to 445 m3/s. The emergency
spillway can pass flows up to 300 m 3/s but it will only Table 1. Loading Combinations and Target FoS
operate when the reservoir levels rise to 815.95 m or
Loading Target
above. Return Period (Tr)
Combinations FoS
2.2 Project Design Life Usual Tr ≤ 50 years ≥ 1.5
Unusual 50 < Tr ≤ 1000 years ≥ 1.3
Consideration was given to requirements to be able to
safely pass the IDF of which approximately 400 m3/s Extreme Tr > 1000 years ≥ 1.1
would need to be passed by the spillway. The required
upgrades to enable the spillway to pass this amount were
considered unfeasible in the short term, and it was
It is noted that the Canadian Dam Association (CDA)
decided that mitigation of short-term risks to the spillway
guidelines (CDA 2013) provide FoS for sliding and not
was a driving priority. A design life of 20 years was
specifically for uplift; however, they were considered
selected for the uplift mitigation measures, to afford WSA
applicable based on the return period of the event causing
time to plan for and implement long-term upgrades to
the uplift loading combination.
enable the dam to pass the IDF.

2.3 UDS Assumed Efficiency 3 EXISTING STRUCTURE ASSESSMENT

The UDS serves two primary functions to a spillway: 3.1 Existing Conditions Assessment
Function 1) it provides an outlet to water under the floor
slab, which if not relieved could result in a hydrostatic uplift A site inspection was conducted on July 19, 2021, to
force when the hydraulic jump forms on the spillway; and assess conditions of the spillway, identify new damage to
Function 2) it provides an outlet to water under the slab to the spillway, assess viable repair options, review general
relieve high groundwater levels and to reduce the potential site conditions including presence of groundwater and
for frost heave by helping to reduce the amount of moisture seepage areas, possible access issues and existing public
of the underlaying soils. An assumption of efficacy of the safety measures at the dam. Of particular interest was the
UDS is required to assess uplift stability, which is seepage that was observed at the chute wall weep drain
represented as a direct reduction of hydrostatic uplift force outlets which indicated the presence of water on the earth
(e.g., an efficiency of 50% indicates that the UDS is able to side of the spillway walls. Additionally, several other issues
reduce the uplift force by 50%). such as spillway operation, maintenance and surveillance
were documented. Based on the findings of the site
Although frost heaving has not been identified as a concern inspection, several repair recommendations were provided
at this site, improvements to the UDS were provided as which included superficial concrete repairs, coating repairs
part of this project primarily to provide Function 1 related to for the gates and hoists shafts and additional public safety
high groundwater level. For Function 2, an UDS efficiency measures upgrades.
of 0% was assumed. Both of these are considered
conservative, but commensurate to the level of risk Available construction and post-construction records
associated with this structure and dam. indicated that high groundwater levels existed at the site
and ongoing seepage flows were being collected and
2.4 Tailwater Conditions discharged through the UDS. In 2009, increased drain
flows, suspected to be a result of backflow through the
A total discharge of 450 m3/s (slightly larger than the weep drains, were observed, and in 2011, a testing
1:1000 discharge) was selected as the combined design program found increased drain flows during operation of
event, which was intended to be passed though the the gates. “Cloudy” flows were also observed during
spillway and the emergency spillway. This is not the IDF, periods of changing drain flows, and during a 2019 camera
but WSA selected this amount to balance short-term risks inspection, a pipe elbow between a transverse drain and a
collector drain was found to be missing (KCB 2020). In leakage through the joints and weep pipes at the chute
addition to these anomalies, the condition of the spillway sections were estimated and compared to the capacity of
joints was identified as an area of concern, specifically, the the drainage system. It was determined that inflows due to
deteriorating condition of the metal waterstops within the leakage to under the slab cannot be handled by the
longitudinal joints (KCB 2020). drainage system.

KCB reviewed the available reports and photographs from 3.2 Uplift Assessment
the 1959 spillway construction to understand the
challenges experienced with high groundwater levels and This section discussed the hydrotechnical work completed
to identify potential seepage sources. Details of the to estimate uplift FoS and the uplift forces to achieve the
structure cutoff walls and curtain walls were examined to target FoS.
determine how the original design was meant to control
and manage seepage under the spillway. In a 2020 report KCB (KCB 2020) assessed the hydraulic
performance of the spillway under various flow scenarios.
Existing condition of the UDS pipes was assessed through A stability assessment was completed to compute FoS
a review of video and observations made during drain against uplift for the spillway slab monoliths. A 2D hydraulic
cleanings and inspections. During an inspection conducted model was created from LiDAR data and a tailwater rating
as part of a dam safety review (DSR) completed in 2020 curve was generated. The water surface profile down the
(Wood 2020), the ends of the weep drains were found to spillway chute and the hydraulic jump geometry was
be damaged and possibly redirecting high energy flow determined for each considered scenario. FoS against
behind the walls, and thereby creating opportunity to uplift were computed assuming varying drainage
surcharge the UDS, i.e. causing uplift pressures to efficiencies. From this stability analysis, it was determined
develop. The elbow intended to connect the left sidewall that the spillway would be unable to safely pass the
drainpipe to the second transverse UDS drain line was spillway design discharge in its current state (i.e., before
disconnected or missing and a void developed in the proposed upgrades).
surrounding soils. Although the inspections did not identify
any issues with the portion of the UDS comprised of “no- Load cases used for analysis are shown in Table 2, and
co-rode” pipes, the long-term performance of these pipes they fall into the Unusual category with a minimum target
has been known to be poor. FoS of 1.3 (LC1 and LC2,) or the Extreme category with a
minimum target FoS of 1.1 (LC3). A tailwater level for the
Instances of seepage flows through slab weep drains, combined flows over the spillway and emergency spillway
sidewall drainpipes, and joints were reviewed. Based on was estimated considering previous studies (KCB 2020)
measurements of the depths of the weep drains, it is (Hatch 2021) (PFRA n.d.) and was used to estimate
suspected that the second transverse UDS drain line hydraulic jump profiles.
slopes from right to left (looking downstream), which would
result in more flow at the left sidewall drain than the right. With the formation of the hydraulic jump the sloping chute
Sediment-laden water from the left sidewall drain for the monoliths become subjected to hydrostatic uplift forces.
second transverse drain line was observed during a 2019 Resistance to the uplift forces is afforded by the weight of
drain investigation program, which may be related to the the 457 mm thick concrete slab and the weight of water
disconnected/missing elbow. Seepage flows and slumping flowing on top of the slab. FoS against uplift under current
of the right slope adjacent to the chute was observed conditions for each of the lower and upper monoliths were
during and after construction. On the left slope adjacent to determined for the load cases considered and ranged from
the spillway, minor seepage spots have been observed. 0.86 to 1.10, necessitating uplift mitigation measures to
safely pass the design discharge.
Possible sources of infiltration from the spillway into the
underlaying gravel were assessed. Reports from Two different methods were used to determine the FoS.
investigations completed in 2011 (Author Unknown) and The first method (M1) being the total resisting loads (i.e.,
2019 (AAFC 2019) were reviewed to determine the effects, the weight of water plus the weight of the slab) divided by
if any, of gate operation on water infiltration into the under- the total uplift load; the second method (M2) being the
slab gravel layer and its subsequent drainage. It was found weight of the concrete slab divided by the total of the
that with Gates 1 or 2 (leftmost gates) operating and weep weight of water on the slab subtracted from the total uplift.
drains plugged, flows at the left sidewall drain increased. The first method typically yields a result closer to one with
This suggests that infiltration is occurring not only through both methods being equal when the FoS equals one. It
the weep drains as suspected, but also through the joints should be noted that in some cases M2 results in a
and possibly through cracks in the slab. significant decrease in the total additional hold-down force
required; M1 was used in determining the required hold-
Measurements taken at the reference markers from 1959 down force to achieve the required FoS.
to 2016 were reviewed to determine the incremental
opening and closing joint gap values and also to determine The additional hold-down force required to provide the
the vertical displacement of the reference marks over time. minimum target FoS for chute monoliths 2 and 3 are shown
There is concern that the metal waterstops in all of the in Table 2 considering 0% UDS efficiency (i.e., the drains
longitudinal joints may be damaged, which would allow the as plugged.)
infiltration of water under the spillway slab. The amount of
Table 2. Additional Resistance Force Required 5.1 Skin Friction and Bonding Length
Upper Lower A review of the available borehole information was carried
Loading Target Monolith Monolith out to determine the approximate stratigraphy of the soils
Combination FoS M1 M2 M1 M2 underlying the chute monoliths and thereby estimate the
(kN) (kN) (kN) (kN) grout/soil bond strengths that could be used for the design
LC1 1.3 459 172 1577 425 of the anchors. The grout/soil bond strengths were
estimated using the recommendations presented in
LC2 1.3 1,177 860 2,148 662 Recommendations for Prestressed Rock and Soil Anchors
by the Post Tensioning Institute (PTI 2014). The review
LC3 1.1 840 705 704 144 indicated the presence of sand with clay lenses under the
*The governing load combination (LC2) for both monoliths is slab. An ultimate bond strength of 30 kPa for clay soils and
shown in boldened text. 80 kPa for sandy soil was assumed. Further analysis was
completed to define an average blended value based on
the expected depths and thicknesses of clay and sand
4 EVALUATION OF OPTIONS layers in which an anchor would be installed. An average
blended grout/soil bond strength of 50 kPa was used for
WSA requested KCB to provide options for ‘passive’ uplift design. An anchor diameter of 150 mm was also used for
mitigation, which means that they require as low as design, considering constructability and potential number
possible inspection, operation and maintenance. Possible of anchors required.
measures for providing additional passive uplift mitigation
explored for this site included: 5.2 Group Failure of Anchors
▪ Option 1: Adding soil anchors, considered viable as no Group failure mechanism was also considered. Figure 1
evidence of frost heave observed at the site. describes the general method used for checking group
▪ Option 2: Adding a concrete overlay on the chute slab, failure of the soil anchors.
which may be viable but was discarded as an option
due to the conflicts with the existing and proposed
UDS. Costs for this option would be higher than for
Option 1. For the overlay option, it was determined
that the addition of approximately a 750 mm thick
concrete overlay would be required (governed by LC2,
monolith 3). This option was considered impractical
and was discarded.
▪ Option 3: Structural upgrades to achieve a monolithic
U-shaped section. This option would require
significantly more work, cost and time to achieve and
was discarded.
▪ Option 4: A combination of soil anchors and a concrete
slab overlay was considered and deemed impractical
for the following reasons: 1) required changes to baffle
blocks; 2) higher costs and time for construction; 3)
conflicts with other recent upgrades.
The most cost-effective solution considered was adding
soil anchors. Soil anchors were considered rather than Figure 1. Group Failure of Anchors from (FHA 1999)
larger concrete piles because soil anchors would be small
in size, require less reinforcement, and could be installed The key for adequate resistance to group failure is the
with a simpler connection to the existing slab. availability of enough ‘enclosed’ soil mass to resist uplift
forces between anchors. The 1.6m free-stress and 7m
bond lengths were thus chosen to provide a large enough
5 DESIGN OF UPLIFT MITIGATION MEASURES enclosed soil volume.

Soil anchors were chosen as the preferred uplift mitigation 5.3 Anchor Head Assembly Design
measure. This section describes the geotechnical and
structural analyses performed to design the hold-down A design for a recessed anchor head was completed and
anchors. is shown in Figure 2. A free stressing length of
approximately the depth of the bedding gravel was used. It
was specified that the anchors’ primary grout be pressure
grouted up to the bottom of bedding gravel elevation to
provide improved soil/grout bond strength. A load testing
program was developed to determine a final bond length Anchor failure mechanisms were checked including
during construction, because the bond length was highly soil/anchor skin friction failure, anchor bar steel yielding,
dependent on the assumed skin friction and the load anchor head assembly failure (see paragraph below), local
testing program was also used to confirm that anchors concrete crushing, and concrete punching shear.
were capable of resisting the design loads. The final bond
length used was 7000 mm, which was the original The anchor head assembly was designed to be recessed
estimated required bond length. When determining a bond within the existing concrete to reduce adverse effects to
length a FoS of 2 was used as recommended by Post the spillway hydraulics (Figure 4) and tripping hazards to
Tensioning Institute (PTI 2014). WSA staff. The steel design incorporated complete joint
penetration welds and considered tension, bending, and
For determining the spacing of the anchors, the structural shear failure modes.
capacity of the existing slab was considered. The
construction report (PFRA 1960) indicated that the
average concrete compressive strength was 26.49 MPa 6 RISK ASSESSMENT SUMMARY
(3842 psi) with the standard deviation being 5.14 MPa (745
psi). The specified yield strength for the reinforcing steel The project construction window was selected as
was 276 MPa (40,000 psi). Reinforcing steel drawings September 15 to December 15, 2022, with a buffer to
were not available at the time for design purposes; February 15, 2023 by meeting the operational and
however, hand notes in the Design Data Book (PFRA n.d.) regulatory requirements. WSA temporarily lowered the
suggested that the chute slab reinforcing included top and reservoir by 2.44m to 813.28m to mitigate the following
bottom mats consisting of #6 bars at 12-inch spacing and concerns during construction:
#4 bars at 12-inch spacing, respectively, in both directions.
▪ High water table under the spillway, which
Existing rebar spacing was confirmed by rebar scan prior
increases difficulties during the installation the
to coring holes in the existing slab.
soil anchors.
Approximately 19 soil anchors with a design capacity of 80 ▪ Leakage of water through the gates resulting in
kN per anchor were determined to be required at each wet and icy/slippery working surfaces during
upper monolith and 27 anchors for the lower monolith. The winter.
layout of the anchors is shown in Figure 2 and Figure 3.
Given that there were 3 rows of monoliths, a total of 138 ▪ Decreased work efficiency due to wet conditions
soil anchors were required. The number and distribution of that could extend the work over two seasons and
anchors was determined by dividing each monolith into 6 significantly increase mobilization costs.
sections and analyzing the forces acting at each section.
Sections that only required a single anchor to achieve the
7 CONCLUSION
target FoS were provided with 2 anchors to avoid the
potential for unbalanced moments in the slab. If a section The design team faced significant technical challenges in
required a fraction of an 80 kN hold down anchor, the key engineering disciplines namely geotechnical,
number of anchors for that section was rounded up to the structural, hydrotechnical, risk management and
nearest whole number. The final design of anchor locations construction planning. This included building an
were also chosen to avoid existing drains, baffle blocks, understanding of the site's hydrotechnical and
and other existing features; however, anchor locations geotechnical context based on original records; reconciling
were also adjusted slightly in the field to avoid steel a lower original design capacity against the recently
reinforcement. Anchors were installed perpendicular to the established much larger IDF needs. Before selecting the
sloped slab to help avoid slab reinforcing when drilling and final design, the project team explored various
to simplify the fabrication of mounting hardware for configurations including structure replacement, a
recessing the anchor heads in the slab. comprehensive rehabilitation, installation of soil anchors,
and upgrades to develop a monolithic U-shape section with
The selected soil anchors consisted of 25 mm diameter
a concrete overlay. Tailwater reviews and hydraulic jump
steel threadbars with a yield strength of 517 MPa. This bar
calculations were completed. Determination of uplift
size was larger than required from a strength perspective;
forces, grout-soil bond strength was also completed, and
this was decided to provide extra cross section that may
the associated FoS estimated. Soil anchor heads were
corrode over the 20-year design life. Despite the additional
designed to minimize effects to the chute hydraulics.
cross-section, corrosion protection for the bond length was
Distribution of soil anchor locations were chosen based on
provided by grout cover, the free-stress length was
several considered hydraulic jump scenarios and to avoid
protected with anti-corrosion compound contained within a
existing spillway features. The construction timeline
PVC sheet shrink wrapped at both ends, and the anchor
included a mobilization in September 2022 and substantial
head assembly was filled with anti-corrosion compound
completion in February 2023.
and sealed.
Figure 2. Plan View of Spillway Showing Anchor Layout (not to scale)
Figure 3. Section View of Spillway Showing Soil Anchor (not to scale)

Figure 4. Anchor Head Assembly Design Detail (not to scale)

8 ACKNOWLEDGEMENTS 9 REFERENCES

We thank the following: AAFC. 2019. "West Val Marie Drain Investigation."
Author Unknown. 2011. "West Val Marie Spillway Drain
WSA and KCB for allowing publication of this information. Test."
CDA. 2013. 2007 Dam Safety Guidelines - Revised 2013.
Peter Roy, P.Eng., for performing the soil/grout bond Canadian Dam Association.
strength assessment and reviewing this document. FHA. 1999. Geotechnical Engineering Circular No.4
Ground Anchors and Anchored Systems.
Nayeem Uddin, M.Sc., P.Eng., and Kimberly Kusch, FWHA-IF-99-015, Washington: Office of Bridge
P.Eng., for reviewing this document. Technology Federal Highway Administration -
Sabatini, P.J.; Pass, D.G.; Bachus, R.C.
Joel Hilderman, M.Sc., P.Eng., David Mack, P.Eng., Terry
Hatch. 2021. "Frenchman River System - Dam Breach
Barkway, P.Eng., Glenn McLaughlin, P.Eng. for providing
Inundation Study - H/361920-000-228-226-
technical advice during the performance of the work.
0002." Draft.
Daniel Bertrand, P.Eng., for the translating to French. KCB. 2020. West Val Marie Spillway Drainage
Assessment. Report with dated March 31, 2020,
Gabrielle Bristo, M.Eng., Engineer-In-Training, for her Klohn Crippen Berger Ltd.
support coordinating the preparation of this document. PFRA. 1960. "Construction Report, West Val Marie
Spillway 1959."
The extended team of engineering, technical and PFRA. n.d. Design Data Book. Compilation of design
administrative staff at KCB and WSA who helped with the data. Prairie Farm Rehabilitation Administration.
delivery of this project. PTI. 2014. Recommendations for Prestressed Rock and
Soil Anchors. PTI DC35.1-14, Post-Tensioning
The team of engineers at KGS Group for providing Institute.
technical review as WSA’s Owner’s Engineer during the Wood. 2020. "West Val Marie 2019 Dam Safety Review."
performance of the work.
Delayed Slope Instabilities in Earth Fill
Dams Due to Creep
Marvin Renzo B. Malonzo & Marolo Alfaro
Department of Civil Engineering – University of Manitoba,
Winnipeg, Manitoba, Canada

ABSTRACT
An earth fill dam exhibited significant deformation in the upstream side after 50 years of satisfactory operation. Studies in
this dam revealed that this was brought by creep movement. However, the studies were unable to simulate the strain
softening of the soil coupled with time-dependent creep deformation. To address this, the use of Time-dependent Model
for Structured soils (TMS) was proposed. It is an incremental plasticity time-dependent constitutive model based on the
Modified Cam Clay. It incorporates the principles of critical state soil plasticity which can systematically emulate the primary,
secondary and tertiary creep phases. The research would allow the assessment of the long-term slope stability of water-
retaining structures and the development of remedial measures for existing earth fill dams that do not satisfy modern long-
term slope stability dam safety requirements.

RÉSUMÉ
Un barrage en terre a montré une déformation importante du côté amont après 50 ans de fonctionnement satisfaisant.
Des études dans ce barrage ont révélé que cela était dû au mouvement de fluage. Cependant, les études n'ont pas été
en mesure de simuler l'adoucissement du sol couplé à la déformation par fluage en fonction du temps. Pour résoudre ce
problème, l'utilisation du modèle dépendant du temps pour les sols structurés (TMS) a été proposée. Il s'agit d'un modèle
constitutif de plasticité incrémentale dépendant du temps basé sur la Modified Cam Clay. Il intègre les principes de la
plasticité du sol à l'état critique qui permet d'émuler systématiquement les phases de fluage primaire, secondaire et
tertiaire. La recherche permettrait l'évaluation de la stabilité des pentes à long terme des structures de retenue d'eau et le
développement de mesures correctives pour les barrages en terre existants qui ne satisfont pas aux exigences modernes
de sécurité des barrages de stabilité des pentes à long terme.

1 INTRODUCTION
2 CONSTITUTIVE MODEL
This research work is a continuation of the previous study
conducted by Ubay (2020) on seven earth fill water- The shortcoming of the SSC constitutive model in
retaining dams in Canada as part of her dissertation. The analyzing creep effects over the service life of the structure,
7-earth fill water-retaining dams will be referred to as WD, as mentioned by Ubay (2020), can be potentially
EF, MFLED, MFRED, CBMD, CBBD2, and CBBD4 as per addressed by TMS since it can perform a coupled analysis
request by the dam operators and owners to maintain the by systematically emulating the different creep phases
confidentiality of these facilities. CBBD2 exhibited under increasing stress levels both in drained and
significant deformations in the upstream side, despite undrained conditions. TMS is based on the principles of
operating satisfactorily for more than 50 years, which Modified Cam Clay (MCC) and the overstress postulate
prompted immediate repairs, as well as investigation to proposed by Perzyna in his work in 1963 and 1966 (Kalos,
determine the potential cause of the slope movement. 2014). Originally, Perzyna’s theory was intended to
Among the different potential causes of the delayed describe the time-dependent behavior of metals. The
slope movement discussed in the work of Ubay, creeping mathematical formulation of the theory, as well as the
of the clay material seemed to be the likely reason. Ubay resemblance of metal behavior to soil, allowed its
performed a time-dependent creep deformation analysis application to be extended to geomaterials.
using Soft Soil Creep (SSC) model with clay strength
parameters between the post-peak and residual shear 2.1 Mechanical Behavior of Geomaterials
strengths. The results were able to reproduce the observed
deformations, proving the validity of the hypothesis. Geomaterials can be classified as either structured or
However, it was noted that SSC lacks the capability of a structureless based on the mechanical behavior (Kalos,
fully coupled time-rate-dependent analysis. 2014). Structured soils are natural soils which has
The Time-dependent Model for Structured soils (TMS) undergone cementation, aging, or preconsolidation
was developed by Kalos (2014) as part of his dissertation. (Belokas & Kavvadas, 2011). The structure existing within
The constitutive model incorporates the principles of critical the soil matrix increases the strength and size of the stress
state soil plasticity to simulate the incremental time- domain in which the soil exhibits stiff behavior (Leroueil &
dependent plasticity behavior of the soil. With this, the TMS Vaughan, 1990). Stress history and bonding induces the
is proposed to be used in the analysis to model the soil creation of structure within the soil. The stress history gives
behavior leading to the delayed slope movement of the “memory” of the past loads that the soil experienced,
CBBD2. which in turn governs its mechanical response. Bonding
induces the growth of interparticle resistance which makes 1. Intrinsic Strength Envelope (ISE) – contains all
the soil stronger (Kalos, 2014). When a structured soil is the structureless states of the soil after substantial
subjected to sufficient strains, it is reduced to a strains have accumulated. It is used in the
structureless state (Leroueil & Vaughan, 1990). Therefore, constitutive model as a reference for the
constitutive modelling of structured soils requires the destructuring mechanism of the soil. Within this
knowledge of the structureless state which is the reference space, the anisotropy is considered to be non-
state of the material (Belokas & Kavvadas, 2011). existent due to the chaotic distribution and
Soils in a structureless state can be described by only orientation of the soil particles at the start of failure
using the current specific volume and current effective (Kalos, 2014).
stress (Belokas & Kavvadas, 2011). The strength of the soil
in this state is solely dependent on the interlocking of the 2. Structure Strength Envelope (SSE) – contains the
soil particles (Kalos, 2014). Burland (1990) described the available strength of the soil due to bonding
structureless state of the soil as intrinsic which indicates formation. Only the secondary anisotropy will be
that the material has no bonding, and all of the stress accounted for in this space which is defined by the
history memory is removed. The locus of the structureless position of the plastic yield envelope within SSE
states represents a limiting state of the structured soils (i.e., (bond anisotropy is not considered).
minimum point to which the soil strength can decrease)
which provides significant insight in the constitutive 3. Plastic Yield Envelope (PYE) - encloses all purely
modelling of structured soils (Belokas & Kavvadas, 2011). elastic stress states, including viscoelastic stress
field, within the SSE. Its size is a fraction of SSE,
2.2 Perzyna’s Theory usually 0.01 to 0.05 of SSE, and its center is
controlled by a kinematic hardening law which
This framework is a combination of the elastoplasticity portrays the recent stress history through the
theory and time-dependent behavior of metals. In this secondary anisotropy tensor (Kalos, 2014).
theory (Figure 1), a static yield surface is assumed in the
effective stress space, and the stress state is allowed to These surfaces are defined in the mean effective stress
cross this boundary during a loading increment which (m) and deviatoric stress tensor (s) hyperplane to give a
results to the accumulation of deformations (Perzyna, 1963 generalized framework which eliminates the back-and-
& 1966). The space within the yield surface is the elastic forth rotation of the stress and strain tensors to the principal
region which is coincidental to the elastic field in stress space (Kalos, 2014).
elastoplasticity; wherein the deformations are purely elastic
and can be defined by the generalized Hooke’s Law.
s
Beyond this boundary is the elastoviscoplastic region
wherein the deformation is characterized by the elastic and
viscoplastic components. It should be noted that the static
yield surface can change its shape and position due to the PYE
accumulation of viscoplastic strains (hardening).
ISE
 Elastoviscoplastic Equipotential m
Region Surface

Static Yield SSE

Surface
Figure 2. Graphical Representation of the Characteristic
Surfaces (from Kalos, 2014)
Overstress

Elastic Region 2.4 Soil Material Behavior

  The material behavior during loading is bounded within the

characteristic surfaces. The space within the PYE is
Figure 1. Graphical Representation Perzyna’s Theory occupied by the viscoelastic field where the strain
(from Kalos, 2014) increment is described by the elastic and viscous
components. The viscoelastoplastic field, wherein plastic
2.3 Characteristic Surfaces deformation is present in addition to the elastic and viscous
strains, exists within the SSE that is outside the ISE and
Three characteristic surfaces, presented in Figure 2, are PYE and on the boundary of the characteristic surfaces.
incorporated in the formulation of TMS to describe the
strength degradation of the material. These are: 2.4.1 Elastic Component

The elastic component is defined by associating the

incremental stress tensor (𝑑𝝈) with the incremental elastic
strain tensor (𝑑𝜺𝑒 ) as expressed in Eq. 1. Ce is the elastic stress intensity effect on creep rate, and m defines the
stiffness tensor defined by poroelasticity which is based on speed of strain rate increase with time. Parameter D is the
the stress-strain paths of soil subjected to isotropic and stress ratio which measures the overstress distance as
uniaxial loading, unloading, and reloading. The governing defined in Eq. 8 and Figure 3. The parameter r is a material
equation of poroelasticity is related by the elastic bulk constant employed to achieve faster creep shear strain
modulus (Ke) to the recompression slope (κ) and specific rates at high shear stress levels (Kavvadas & Kalos, 2019).
volume (v) as presented by Eq. 2. The formula tends to
predict non-zero values along a full loading cycle which 2𝐴 sinh(𝑎̅𝐷𝑟 )
makes it a conservative approach. The elastic shear 𝜀̇𝑞𝑣 = 𝑚 ;𝑚 = 1
(1 − 𝑚)𝜀𝑞𝑣 1−𝑚 [6]
modulus (Ge) can be related to the elastic bulk modulus by [1 + ]
𝑟 )
2𝐴 sinh(𝑎̅𝐷 ∙ 𝑡0
Poisson’s ratio (Eq. 3).

𝑑𝝈 = 𝑪𝑒 ∶ 𝑑𝜺𝑒 [1]
2𝐴 sinh(𝑎̅𝐷𝑟 )
𝜀̇𝑞𝑣 = ;𝑚 ≠ 1
v 𝜀𝑞𝑣 [7]
𝐾𝑒 = ∙𝝈 [2] exp [ ]
𝜅 2𝐴 sinh(𝑎̅𝐷𝑟 ) ∙ 𝑡0

2 ∙ 𝐺𝑒 9∙𝑣
= 3− [3] 𝑞 √3 (𝑠: 𝑠)
𝐾𝑒 1+𝑣 2
𝐷=
𝑞𝑓𝑎𝑖𝑙
= ≤1 [8]
3 3
𝑚𝑎𝑥 {√ (𝑠𝐹 : 𝑠𝐹 ), √ (𝑠𝐶 : 𝑠𝐶 )}
2.4.2 Plastic Component 2 2

The plastic component of the strain is defined by an s

associated plastic flow rule and is described through Eq. 4, Initial-CSC
cin
where P is the plastic potential tensor which controls the clim
Limit-CSC
size of the plastic strain tensor and dΛ is the scalar quantity SSE
expressing the magnitude of the plastic strain tensor.
E sF
𝑑𝜺𝑝 = 𝑑𝛬 ∙ 𝑷 [4] sc
PYE
m
2.4.3 Viscous Component
F
The viscous strain is composed of volumetric and d
deviatoric deformations which can be observed in standard
oedometer and triaxial tests, respectively. Viscous Figure 3. Graphical Representation of the Stress States
volumetric strains (𝜀̇ 𝑣 ) tend to increase the strength of soil (from Kalos, 2014)
by developing the pre-consolidation condition (Kavvadas &
Kalos, 2019). However, this causes structure degradation
in highly expansive soils and some weak rocks. The 2.5 Hardening Rules
evolution of 𝜀̇ 𝑣 is described by the semi-logarithmic creep
formula using the specific volume (v) and secondary The position and shape of SSE and PYE is controlled by
compression coefficient (ψ=Cae/ln10). The formula (Eq. 5) hardening rules (Kalos, 2014). Isotropic hardening law is
is valid for t > t0, where t0 is the reference time for the start applied to define the size of SSE, and subsequently, the
of creep. TMS also assumes that 𝜀̇ 𝑣 accumulates at all PYE size through the proportionality ratio ξ. Kinematic
stress fields by simultaneously laying on the SSE and PYE hardening rules define the position of these characteristic
(Kalos, 2014). surfaces, as well as the slope (c) of the Critical State Cone
(CSC) in the stress hyperplane (Kalos, 2014).
𝜓 v
𝜀̇𝑣𝑣 = exp (− 𝜀 𝑣 ) [5] 2.5.1 Isotropic Hardening
v𝑡0 𝜓

The viscous deviatoric component (𝜀̇𝑞𝑣 ) on the other The half-size a of the SSE is treated as a hardening
hand, tend to degrade the material structure similar to the parameter which evolves with the accumulation of plastic
deformations caused by plastic strains in inviscid and viscous strains (Kalos, 2014). The SSE is only allowed
structured soils (Kavvadas & Kalos, 2019). Once high to degrade until the ISE. As such, the half-size of SSE can
shear stress levels are applied to the soil specimen, be related to the half-size of the ISE (a*) through the
deviatoric strains accumulate until creep failure is achieved structure ratio (B). In this sense, the parameter B is similar
(Kalos, 2014). The primary and secondary deviatoric creep to the over-consolidation ratio (Kavvadas & Kalos, 2019).
strains are described by the Singh-Mitchell formula in Eq. 6
and Eq. 7. Singh-Mitchell parameter A controls the 𝑎 = 𝐵 ∙ 𝑎∗ [9]
measure of the viscous deviatoric strains, 𝑎̅ portrays the
The structure ratio undergoes gradual degradation from an tensile strength, is represented by parameter d as
initial value (B0) to a residual value (Bres) when the material illustrated in Figure 3 (Kalos, 2014). This
is subjected to large strains, resulting to the complete parameter degrades exponentially from an initial
destructuring of the soil (Kalos, 2014). Bres can be set to a state (din) to a value of zero at critical state,
value of 1 which indicates that SSE has degraded to the reducing the SSE to the MCC bounding surface
ISE, or it can have a value of slightly above 1 to account as presented in Eq. 12 (Kavvadas & Kalos, 2019).
for any chemical, biological, or thixotropic bonding. In TMS,
parameter B moves between B0 and Bres following a Cam- 𝜕𝑑 𝜕𝑑
𝑑(𝑑) = : 𝑑𝜀 𝑝 + 𝑣 : 𝑑𝜀 𝑣 [11]
Clay evolution pattern at large irreversible strains 𝜕𝜀 𝑝 𝜕𝜀
(Kavvadas & Kalos, 2019). This is achieved by using the
plastic and viscous strain increments to define the size of 3. PYE Location
SSE as shown in the following equation after Kavvadas &
Kalos (2019). Further loading on a stress state that has reached
the surface of the PYE would cause PYE to move
𝜕𝑎 𝜕𝑎 along the path of evolution of the stress state
𝑑𝑎 = : 𝑑𝜀 𝑝 + 𝑣 : 𝑑𝜀 𝑣 [10]
𝜕𝜀 𝑝 𝜕𝜀 (Kavvadas & Kalos, 2019). In TMS, when stress
state M lying on PYE is subjected to additional
stress, PYE moves towards the conjugate point
2.5.2 Kinematic Hardening M’ on the SSE as illustrated in Figure 4. As such,
the center point (L) of PYE moves according to
The Critical State of the soil is reached when the specimen the kinematic hardening rule (Eq. 12) proposed by
undergoes large strains leading to the structureless state Kavvadas & Amorosi (2015). The hardening rule
of the material. Due to its uniqueness, it is an important defines the evolution and movement of PYE by
parameter in the constitutive model. TMS predicts unique Eq. 13.
critical states by ensuring that the soil state falls onto a
unique CSL in the v – ln(m) plot which corresponds to a 𝝈 − 𝝈𝑳
𝑳𝑴 = 𝜉𝑲𝑴′ → 𝜎𝑀′ = (𝑎 − 𝑑)𝑰 + [12]
Critical State Cone (CSC) in the s – m plane (Kavvadas & 𝜉
Kalos, 2019). Kinematic hardening rules are implemented
to achieve this which define the translation of SSE along 𝑑𝑎
𝑑𝝈𝑳 = 𝝈 + 𝑑𝜇 ∙ 𝜷 [13]
the hydrostatic axis, evolution of the location of the PYE, 𝑎 𝑳
and the slope of the CSC in the stress hyperplane which
subsequently affects the eccentricity of SSE.
s
1. Critical State Cone (CSC) Evolution M’

The slope of the CSC (c), which is related to the M
PYE
MCC parameter M (𝑐 = 𝑀√2/3), is allowed to
transition between an initial and final value. The
L m
Initial-CSC represents high-speed strain rates, O K
while the Limit-CSC correspond to quasi-static
ISE
strain rates with an order of 10-7 (Kalos, 2014).
Failure of the specimen is achieved when the SSE
CSC passes through point E on the PYE wherein
the stress is constant (Figure 3). The slope of the
CSC at failure does not necessarily coincide with
Limit-CSC which allows the material to fail even Figure 4. Graphical Representation of PYE
before reaching its weakest state. The Movement (from Kalos, 2014)
degradation of the CSC inclination simulates the
tertiary creep failure under drained conditions 3 METHODOLOGY
(Kalos & Kavvadas, 2018). The evolution of the
slope of CSC from an initial value (c in) to a lower Pertinent laboratory tests shall be conducted to determine
value (clim) is described by an exponential decay the necessary parameters for the analysis. In total, TMS
function (Eq. 11) making use of the deviatoric uses 21 parameters in its formulation wherein 9 can be set
plastic and viscous strain increments. empirically and the remaining factors can be determined
through testing as presented in Table 1 and Table 2,
∂c ∂c respectively. The Commercial Finite Element Code
𝑑𝑐 = 𝑝
: 𝑑𝜀 𝑝 + 𝑣 : 𝑑𝜀 𝑣 [11] SIMULIA ABAQUS shall then be used to assess the
∂𝜀 ∂𝜀
delayed deformation observed in CBBD2 using the UMAT
2. SSE Translation Subroutine function (TMS was constructed using Fortran).

The effect of bonding of soil particles due to

isotropic compression, which is the buildup of
Table 1. Empirical TMS Parameters (from Kavvadas & was only induced prior to the calculation of the factor of
Kalos, 2019) safety of the state of the structure at the time. As such, a
Symbol Parameter Typical Value more rigorous solution is needed to address the
shortcoming of the SSC model.
ν Poisson’s ratio 1/3
TMS (Kalos, 2014) is a constitutive model based on the
t0 Reference time at start of creep Arbitrary theories of Modified Cam Clay and overstress by Perzyna
Ratio of SSE and PYE ellipsoidal 0.01 – 0.05 (1963 & 1966). It uses characteristic surfaces (SSE, PYE,
ξ
shapes and ISE) defined in the stress hyperplane to simulate the
Bres Ratio of residual structure 1 material degradation under loading. Kalos (2014) did note
Deviatoric destructuring variable for d 1.5 – 2 times in his work that further study is needed regarding the real-
ϑqp
due to plastic strain accumulation of θqp world performance of the model.
Deviatoric destructuring variable for d 1.5 – 2 times Therefore, a unique opportunity is available to
a2v
due to creep of a1v implement TMS in analyzing the delayed instability
Constant employed in the interpolation 3 – 7 observed in CBBD2. The research will assess the long-
δ term slope stability of water-retaining structures and help in
for the PYE plastic modulus
Exponent employed in the interpolation 1 developing remedial measures for existing earth fill dams
y that do not satisfy modern long-term slope stability dam
for the PYE plastic modulus
r Singh-Mitchell parameter; modification 1 safety requirements.
exponent
5 ACKNOWLEDGEMENT
Table 2. TMS Parameters to be Determined in the
We would like to thank Dr. Alexandros Kalos and Dr.
Laboratory (from Kavvadas & Kalos, 2019)
Michael J. Kavvadas of the National Technical University
Symbol Parameter Laboratory of Athens for their support and guidance in using TMS for
Test
this research.
c Slope of CSC in the stress hyperplane Triaxial Test
Drained 6 REFERENCES
ηqp Deviatoric destructuring variable
Triaxial Test
Oedometer Belokas, G., & Kavvadas, M. (2011). An Intrinsic
η vp Volumetric destructuring variable Compressibility Framework for Clayey Soils.
Test
Niso* Intrinsic specific volume at p=1kPa Geotechnical and Geological Engineering, 855 - 871.
Cam clay parameter; slope of
κ Oedometer Burland, J. (1990). On the Compressibility and Shear
recompression line
Test Strength of Natural Clays. Geotechnique, 40(3), 329 -
Cam clay parameter; slope of
λ
compression line
378.
Creep Kalos, A. (2014). Investigation of the Nonlinear Behavior of
ψ Standard logarithm of creep rate Oedometer
Test
Soil Materials by Creep Simulation. [Doctoral
dissertation, National Technical University of Athens].
Deviatoric destructuring variable for c
θqp National Technical University of Athens Institutional
due to plastic strain accumulation Drained Creep
Repository. Retrieved from
Deviatoric destructuring variable for c Triaxial Test http://dx.doi.org/10.26240/heal.ntua.1597
a1v
due to creep
Singh-Mitchell parameter; strain rate at Kalos, A., & Kavvadas, M. (2018). Slope Instabilities
A reference time (t) when deviator stress
Triggered by Creep Induced Strength Degradation.
is zero
Numerical Methods in Geotechnical Engineering IX,
Singh-Mitchell parameter; slope of Undrained 1097 - 1104.
𝑎̅ linear segment in the logarithmic strain Creep Triaxial
rate – shear stress plot Test
Kavvadas, M., & Amorosi, A. (2015). A Constitutive Model
Singh-Mitchell parameter; slope of for Structured Soils. Geotechnique, 50(3), 263 - 273.
m linear segment in the logarithmic of
strain rate and time plot
Kavvadas, M., & Kalos, A. (2019). A Time-dependent
Plasticity Model for Structured Soils (TMS) Simulating
Drained Tertiary Creep. Computers and Geotechnics,
4 CONCLUSION 130 - 143.
The delayed instability in CBBD2 was attributed to the
Leroueil, S., & Vaughan, P. (1990). The General and
creeping of the clay core and blanket as indicated by the Congruent Effects of Structure in Natural Soils and
analyses of Ubay (2020). The Soft Soil Creep (SSC) Weak Rocks. Geotechnique, 40(3), 467 - 488.
constitutive model was used to simulate creep behavior
which yielded promising results. However, the analysis was
limited due to the sequential assessment wherein the
degradation of the shear strength of the soil due to creep
Perzyna, P. (1963). The Constitutive Equations for Rate
Sensitive Plastic Materials. Quarterly of Applied
Mathematics, 20(4), 321 - 332.

Perzyna, P. (1966). Fundamental Problems in

Viscoplasticity. Advances in Applied Mechanics, 9, 243
- 377.

Ubay, I. (2020). Stability Assessment of Aging Water-

Retaining Earth Fill Dams. [Doctoral dissertation,
University of Manitoba]. University of Manitoba
Institutional Repository. Retrieved from
http://hdl.handle.net/1993/35140
An experimental study on the influence of soil
structure on erosion behavior of sensitive clays
Amir-Hossein Daneshi-Sadr, François Duhaime & Yannic Ethier
Construction Engineering Department – École de technologie
supérieure, Montréal, Québec, Canada

ABSTRACT
The present research attempts to evaluate the influence of the aggregated structure of sensitive marine clays on their
erosion characteristics. Hole erosion tests were carried on five pairs of remolded and undisturbed specimens, each of
which derived from a distinctive parent soil. Hole erosion test results corroborated the high erosion resistance of the intact
structured clays. The erodibility of remolded samples, which were subjected to unidimensional compression, was
significantly greater due to the destructuration of interparticle bonding between soil fabrics. In addition, a decrease in soil
erosion resistance was observed concurrently with an increase in sensitivity. A perceptible interrelation was recognized
between the critical shear stress and the plasticity index, wherein soil erodibility decreases with the increase of the plasticity
index.

RÉSUMÉ
Cette recherche a tenté d'évaluer l'influence de la structure agrégée des argiles marines sensibles sur leurs
caractéristiques d'érosion. Des essais d'érosion de trou ont été réalisés sur cinq paires d'échantillons remaniés et intactes,
chacun provenant d'un sol parent distinctif. Les résultats d’essai d'érosion de trou ont corroboré la résistance élevée à
l'érosion des argiles structurées intacts. En raison de la déstructuration de la liaison entre les particules du sol, les
échantillons remaniés qui ont été soumis à une compression unidimensionnelle ont montré une érodabilité
comparativement plus élevée. De plus, une diminution de la résistance à l’érosion du sol a été observée lorsque la
sensibilité est plus élevée. Une interrelation perceptible a été constatée entre la contrainte de cisaillement critique et l'indice
de plasticité, dans lequel la résistance à l’érosion du sol s'améliore avec l'augmentation de l'indice de plasticité.

1 INTRODUCTION enlargement of the conduit diameter, it is feasible to

determine the erosion characteristics.
Internal erosion due to concentrated leaks is deemed a The erosion behavior of cohesive soils is governed by
significant threat to the stability and efficient operation of their geotechnical properties, such as mineralogy of clay
hydro-technical structures, e.g., levees, dikes, and particles, saturation, and void ratio, and by hydraulic and
embankments. Seepage through fissures and defects environmental factors, such as pore fluid chemistry,
results in the detachment of soil particles and the transfer temperature, and pH. In the case of natural clays, the
of eroded particles downstream, thereby causing the erodibility is influenced by the distinctive soil structure and
defect to enlarge and the structure to become unstable. the strength of planes of weakness (Lefebvre et al. 1985).
The literature has presented a large number of Numerous geological incidents of both regional and local
experimental and numerical methods aimed at enhancing significance, such as landslides and erosion, have been
comprehension of soil erosion mechanisms and documented in the literature (Bjerrum et al. 1969, La
establishing the relationship between soil properties and Rochelle et al. 1970, Brzezinski 1971, Eden et al. 1971,
erosion characteristics, namely critical shear stress, Tavenas et al. 1971, Eden 1972, Mitchell and Klugman
erosion coefficient, and erosion rate index. The hole 1979, Williams et al. 1979, Lévy et al. 2012). These
erosion test (HET), developed by Wan and Fell (2002, occurrences underscore the significance of the thorough
2004a, b), has been extensively employed in studying the comprehension of the behavior of sensitive clays
erosion behavior of soils owing to its straight-forward Understanding the particular characteristics of the
methodology and reliable results. Several scholars have sensitive clays and the impact of mechanical and
made modifications to the testing and interpretation environmental parameters on the geotechnical properties
methodology in order to enhance the accuracy and efficacy is of great importance for the construction and
of soil erosion determination. After placing the sample maintenance of hydro-geotechnical structures.
inside the test cell, a preformed conduit in the center of the The mineralogy, fabric structure, stress-strain behavior,
sample is subjected to an eroding water flow in order to compressibility, and shear strength of the natural clays
simulate concentrated leak erosion in earth structures. The have been comprehensively examined in the literature
exerted tangential seepage force stimulates detachment of (Crawford 1961, Kenney 1968, LaRochelle and Lefebvre
soil particles from the conduit walls, which are then carried 1971, Mitchell 1976, Bentley and Smalley 1978, Baudet
downstream via the flow. Through monitoring the hydraulic and Stallebrass 2004). Notwithstanding, the erosion
state of the eroding flow and assessment of the characteristics of sensitive clays and the influence of
interparticle bonds and stratification on erosion parameters
have yet to be comprehensively and experimentally where 𝑓𝐿 is the friction factor for laminar flow (kg/m2s), 𝑓𝑇
investigated. The existing literature on the erosion is the friction factor for turbulent flow (kg/m 3), and Q is the
properties of sensitive clays is limited by the small number flow rate (m3/s).
of tested samples and by testing methodologies that differ
from current methods (Lefebvre et al. 1985, 1986, Chapuis 3 MATERIAL TESTED
1986). The primary aim of the present study is to appraise
the erosion behavior of structured clays and to gauge the A substantial area of eastern Canada is draped by thick
impact of interparticle bonding on erosion characteristics. accumulations of post-glacial marine clays known for their
high sensitivity. These sensitive clays are known as
2 DETERMINATION OF THE SOIL EROSION Champlain clays or Leda clays. They consist of fine
CHARACTERISTICS sediments eroded from the igneous and metamorphic
rocks of the Canadian Shield. These clays were initially
The principal equation which describes soil erosion in HET deposited in saline water but later leached by freshwater
can be written as during post-glacial rebound (Mitchell and Klugman 1979,
Torrance 1979, 1983). Geological history, sedimentation,
𝜀̇𝑡 = Ce (τt - τc) [1] and post-sedimentation processes all contributed to the
distinct structure of sensitive clays. Sedimentation factors
where, 𝜀̇𝑡 = erosion rate per unit surface area at time t such as mineralogical composition and deposition
(kg/s/m2), Ce= coefficient of soil erosion (s/m), τt = hydraulic environment promoted flocculation, whilst post-
shear stress exerted by eroding flow at time t (N/m 2), and sedimentation factors such as slow consolidation and
τc = critical shear stress for erosion initiation (N/m 2). Also, contemporaneous cementation of particles facilitated the
the erosion rate index (IHET) is defined as formation of the aggregated soil structure. Natural clays in
which the formation of cemented interparticle bonds was
𝐼𝐻𝐸𝑇 = −log⁡(𝐶𝑒 ) [2] initiated during the incipient phase of consolidation exhibit
greater porosity in comparison to clays with bonds that
Wan and Fell (2002, 2004a, b) introduced an formed subsequent to overburdening by pursuant deposits
interpretation methodology based on Equation 2 in order to (Quigley and Ogunbadejo 1972, Quigley 1980).
quantitatively study the soil erosion mechanisms due to Champlain clays can present fissures, zones of
concentrated leakage in embankments. This methodology weakness, and silt or sand lenses (Gaskin et al. 2003). The
is based on the principle that for erosion to start, the shear aggregated structure of the natural clay is formed by card-
stress applied to the surface of the conduit should be house oriented particles which are connected to each other
greater than τc. Given that the hydraulic gradient through by rigid cementation bonds (Gillott 1979). The Champlain
the conduit is measured by i=Δh/L, the exerted shear stress clay deposits exhibit a spectrum of consolidation states,
to the hole periphery can be defined by ranging from normally consolidated to overconsolidated
across different geographical regions, with variable
𝜙𝑡 mineralogy and particulate composition. Undisturbed non-
𝜏𝑡 = 𝜌𝑤 𝑔𝑖 [3]
4 remolded natural clays can have a high shear strength,
nonetheless, destructuration substantially reduces their
where ρw is the density of the eroding fluid (kg/m3), g is the strength.
gravitational acceleration (9.81 m/s2), and ϕt is the average Few investigations have been conducted to assess the
diameter of the erosion hole at time t (m). Hereupon, the surface or internal erosivity of undisturbed sensitive clays.
erosion rate per unit surface area of the preformed erosion Research findings indicate that the erosion mechanism of
conduit is determined by intact structured clays is more intricate compared to
conventional cohesive soils. While undisturbed non-
𝜌𝑑 𝑑𝜙𝑡
𝜀̇𝑡 = [4] weathered structured clays are strong enough to resist
2 𝑑𝑡
shear stresses as high as 400 Pa, the resistance to water
Where ρd= soil dry density (kg/m3), and dϕt/dt= change in erosion is considerably reduced for remolded specimens.
hole diameter with time (m/s). The initial hole diameter, ϕ i, Furthermore, for samples presenting erosion, it mostly
is 6 mm, and the final diameter of the eroded hole, ϕ f, can takes place at the level of either small lumps of clayey
be calculated from the volume of the paraffin candle which material or silt and sand particles along planes of
is casted in the hole after test termination. The diameter ϕt weakness and inherent heterogeneities (Lefebvre et al.
of the hole at time t can be calculated as follows for laminar 1985, 1986).
and turbulent flow conditions, respectively.
4 HET APPARATUS AND TEST PROCEDURE
1
16𝑄𝐿𝑓𝐿
𝜙𝑡 = ( )
3 [5] The HET apparatus at ÉTS is illustrated in Figure 1. The
𝜋𝜌𝑤 𝑔∆ℎ
upstream/downstream flow chambers and the main cell are
fabricated from transparent PVC, thus enabling qualitative
𝜙𝑡 = (
64𝑄2 𝐿𝑓𝑇
)5
1
[6] assessment of emerging effluent turbidity as well as the soil
𝜋2 𝜌𝑤 𝑔∆ℎ specimen condition. The main cell of the apparatus,
accommodating the soil during the experiment, is affixed
between the upstream and downstream flow chambers.
 (a)

(b)

Figure 1. (a) Photograph, and (b) schematic illustration of the HET apparatus;
notes: 1- inlet pipe, 2- flowmeter, 3- upstream valve for the regulation of the water flow and the initiation of test, 4- upstream
port for the connection of the differential pressure sensor, 5- HET cell, 6- downstream port for the connection of the
differential pressure sensor, 7- valve for the regulation of the water flow, 8- upstream flow chamber, 9- downstream flow
chamber, 10- turbidimeter, 11- outlet pipe, 12- data acquisition card, and 13- computer.

HETs are performed on both remolded and intact entailed the unidimensional compaction of four soil layers
specimens of sensitive clays with the objective to evaluate on top of a plastic disc, while preserving the initial water
their erosion behavior and the impact of destructuration on content. The purpose of utilizing the disc downstream of
their stability. The intact specimens were collected using the specimen is to prevent specimen slaking and excessive
thin-walled samplers in order to preserve the intact clay reduction of the length of the conduit (Lim 2006). A 6 mm
mechanical properties and structure. After collecting the axial hole was created by pushing a sharp metal rod at the
samples, they were sealed with paraffin wax, and center of the HET specimen. The aim is to induce erosion
preserved in a humid room to avoid the formation of tension only in the preformed hole in order to simulate the surface
cracks caused by drying. Hole erosion test specimens were erosion phenomenon in pre-existing defaults. Before test
introduced in the HET apparatus using the sharp edge of commencement and after affixing the cell, the upstream
the HET cell. The preparation of remolded samples and downstream flow chambers were simultaneously filled
with water and saturated. Hole erosion tests are performed sensitivity (CFEM; Canadian Geotechnical Society 2006).
based on the constant head loss testing method. The Samples F1 and F6 exhibited the lowest sensitivity values,
adjustable upstream reservoir is capable of providing 3.5 while F36 was regarded as the most sensitive sample
m pressure head. Each test head was maintained for 12 to among those tested.
15 minutes. If no sign of erosion was detected, the
upstream pressure head was increased until progressive 800
erosion and critical conditions were initiated.
In order to determine a correlation between erosion 700

characteristics and geotechnical properties of samples, 600

Atterberg limits were measured for the five sensitive clays

τc (N/m2)
500
(BNQ 2501-092). In addition, to demonstrate the influence
400
of the soil structure on the resistance to water erosion,
sensitivity of the natural clays was determined using the fall 300

cone penetrometer (BNQ 2501-110). 200

100

0
5 RESULTS AND DISCUSSIONS 0 0.5 1 1.5 2 2.5 3 3.5
Ce (×10-5 s/m)
Hole erosion tests were carried on five sensitive clay
sample. The erodibility of each sample was evaluated in
Figure 2. Correlation between the critical shear stress and
two different conditions, namely intact and remolded.
the erosion coefficient
Atterberg limits, sensitivity, remolded and undisturbed
strengths, and moisture content of the tested specimens
Photographs of some of the paraffin wax candles are
are provided in Table 1. Hole erosion tests specifications
presented in Figure 3. It is important to note that due to
are provided in Table 2. On the basis of HET results, the
intense floc erosion and slaking at the downstream end,
improvement of the soil erodibility is attained by the
preparation of the paraffin candle was unfeasible for some
increment of the critical shear stress and reduction of the
samples (Figure 4). The existence of weak planes and
erosion rate (Figure 2).
silt/fine sand lenses adjacent to the erosion hole facilitates
The most frequently employed parameter for
the separation of clay lumps, which accounts for the non-
comparing shear strength of intact and remolded soil is
uniform expansion of the conduit diameter in some
referred to as sensitivity. According to the sensitivity
specimens.
measurements obtained with the fall cone test, soil
specimens demonstrated generally low to medium

Table 1. Geotechnical parameters of natural clays used in the research program

Soil w (%) Cu (kPa) 1 Cur (kPa) 1 St 1 LL (%) 2 PL (%) PI (%)

F1 45.9 22 3.9 6 50 18 32
F2 28.9 120 12.1 10 45 18 27
F6 46.4 52 9.9 5 44 19 25
F6A 29.7 23 2.9 8 41 16 25
F36 41.0 137 6.8 20 44 21 23
1
Undisturbed strength, Remolded strength, and Sensitivity determined by the cone penetration test (BNQ 2501-110).
2
Liquid limit determined by the cone penetration test (BNQ 2501-092).

Table 2. Erosion characteristics of undisturbed and remolded specimens

Soil τc (Pa) IHET Ce (× 10-5 s/m) Qualitative description 1

F1 722.7 5.1 0.79 Very slow
F1-Remolded 227.4 4.5 2.95 Moderately slow
F2 377.8 5.1 0.8 Very slow
F2-Remolded 205.4 4.7 1.93 Moderately slow
F6 334.4 4.9 1.31 Moderately slow
F6-Remolded 225.0 4.7 1.9 Moderately slow
F6A 218.6 4.9 1.4 Moderately slow
F6A-Remolded 189.1 4.7 2.03 Moderately slow
F36 126.1 4.9 1.3 Moderately slow
F36-Remolded 91.7 4.6 2.74 Moderately slow
1
Qualitative description of soil erosivity in accordance with Wan and Fell (2004b).
 Noteworthy information pertaining to the natural
800
sedimentary formation and stratified structure of the
700 Undisturbed
samples can be gleaned from a meticulous examination of
600
the candles. At the interface between two layers with Remolded

distinct properties, soil erosion intensifies which causes the 500

τc (N/m2)
eroded section to expand. 400
300

Flow Direction 200

100
0
F1 F2 F6 F6A F36
Soil Sample

Figure 5. Influence of soil remolding on the critical shear

stress of the used structured clays
Figure 3. Photographs of the paraffin wax candles from Figure 6 shows that the critical shear stress tends to
post-test eroded conduits decrease with an increase in sensitivity. The observed
tendency is more pronounced in the case of undisturbed
samples in comparison to remolded samples.
Inconsistencies can be ascribed to variations in the
characteristics of intact samples, such as the water content
and the depth from which the samples were collected.
The results reveal that, despite the fact that brittle
interparticle links have been altered in remolded clays,
other parameters, including mineralogy and
physicochemical characteristics of particles, strongly
impact the erosion behavior of sensitive samples.

(a) (b) 800

Undisturbed
700
Remolded
Figure 4. Pictures of the (a) downstream and (b) upstream 600
sides of the eroded HET specimen.
500
τc (N/m2)

Based on HET results and as depicted in Figure 5, the 400

influence of the structure degradation on the erosion
300
resistance of the tested cemented clays is substantial. The
shear resistance of the soil is a function of the coefficient 200
of friction, which is in turn influenced by the structure and 100
stratification of the soil matrix. When structured clays are
subjected to one-dimensional compression, the brittle 0
0 5 10 15 20 25
intra-aggregate bonds within the soil structure are St
disintegrated, resulting in particles to behave like their Figure 6. Variation of the critical shear stress for different
original pre-cementation essence in which the shear degrees of the sensitivity of the tested natural clays
strength is controlled by the interparticle friction (Quigley
and Thompson 1966). The compression induces a A correlation can be established between the critical
significant alteration to the porous structure of the clay, shear stress and Atterberg limits for undisturbed and
without complete obliteration, whereby the randomly remolded specimens (Figure 7). The divergence of some
oriented particles tend to reorient orthogonally to the samples from the general trend can be explained by the
applied force direction (Delage and Lefebvre 1984). The influence of several geotechnical and environmental
increased parallelism of the particles results in a significant parameters, such as particle mineralogy, sedimentation
decrease in the coefficient of friction (Kenney, 1968). At conditions, and particle size (Locat et al. 1984).
this stage, soil erosion behavior is significantly influenced The present study corroborates the findings of earlier
by both consolidation state and the chemistry of pore water studies (e.g., Wan and Fell 2002, Soroush et al. 2019,
(Arulanandan et al. 1973, Sherard et al. 1976). Shourijeh et al. 2020) indicating that soils with higher
plasticity index are more resistant to water erosion.
 results of the hole erosion tests, undisturbed marine clays
exhibit a capacity to withstand applied shear stresses of up
800 to 700 Pa. On the contrary, it has been observed that the
Undisturbed (a)
700 process of remolding and structural degradation leads to
600 Remolded an increase in erodibility.
Through applying a unidirectional compression, brittle
500
links between soil fabrics break. As a result, the porous
τc (N/m2)

400 structure of the soil is altered and soil particles that were
300 randomly oriented tend to form parallel layers. In view of
200 the fact that destructured specimens possess lower
coefficients of friction compared to undisturbed ones,
100
remolded clays have lower shear resistance, thus are more
0 erosive compared to intact samples.
0 10 20 30 40
It was discussed that for the majority of the tested clays,
PI (%)
the critical shear stress reduces as the sensitivity value
800
increases. In addition, a correlation was observed between
Undisturbed (b) the critical shear stress and the plasticity index.
700
Remolded
600 7 REFERENCES
500
τc (N/m2)

400 Arulanandan, K., Sargunam, A., Loganathan, P., and

300
Krone, R. 1973. Application of chemical and electrical
parameters to prediction of erodibility. Soil Erosion:
200
Causes and Mechanisms, Prevention and Control: 42-
100 51.
0
0 20 40 60 80 Baudet, B., and Stallebrass, S. 2004. A constitutive model
LL (%) for structured clays, Géotechnique, 54(4): 269-278.

800 Bentley, S. P., and Smalley, I. J. 1978. Inter‐particle

700 Undisturbed (c) cementation in Canadian post‐glacial clays and the
Remolded problem of high sensitivity (St > 50), Sedimentology,
600
25(2): 297-302.
500
τc (N/m2)

400 Bjerrum, L., Loken, T., Heiberg, S. and Foster, R. 1969. A

field study of factors responsible for quick clay slides,
300
Proceedings of the 7th International Conference on Soil
200 Mechanics and Foundation Engineering, Mexico City,
100 Mexico, 2: 531-540.
0
0 5 10 15 20 25 30 Brzezinski, L. S. 1971. A review of the 1924 Kenogami
PL (%) landslide, Canadian Geotechnical Journal, 8(1): 1-6.
Figure 7. Correlation between the critical shear stress and Bureau de normalisation du Québec (BNQ). 2014a. Soils –
(a) plasticity index, (b) liquid limit, and (c) plastic limit for Determination of liquid limit by the fall cone
tested samples penetrometer and determination of plastic limit,
Canadian Standards Association and Bureau de
normalisation du Québec, CAN/BNQ 2501-092.
6 SUMMARY AND CONCLUSIONS
Bureau de normalisation du Québec (BNQ). 2014b. Soils –
The distinct structure of sensitive eastern Canadian clays Determination of undrained shear strength and
is attributed to the gradual accumulation of the sedimentary determination of sensitivity of cohesive soils using the
particles, which occurred concurrently with the formation of fall cone penetrometer, Canadian Standards
cementitious interparticle bonding across the Champlain Association and Bureau de normalisation du Québec,
Sea basin. Remolding and structural degradation break CAN/BNQ 2501-110.
down the interparticle bonds, resulting in a substantial
decrease in shear strength. This article aimed to Canadian Geotechnical Society. 2006. Canadian
demonstrate the influence of the structure of sensitive foundation engineering manual, 4th ed. BiTech
Champlain Sea clays on their erosion characteristics. Publishers Ltd.
The newly designed HET apparatus at ÉTS along with
the interpretation methodology were presented. Both Chapuis, R. P. 1986. Quantitative measurement of the
remolded and undisturbed specimens were used in the scour resistance of natural solid clays, Canadian
course of this experimental research. With regard to the Geotechnical Journal, 23(2): 132–141.
Crawford, C. 1961. Engineering studies of Leda clay, In Lévy, S., Jaboyedoff, M., Locat, J., and Demers, D. 2012.
Soils in Canada, University of Toronto Press, Toronto, Erosion and channel change as factors of landslides
Canada, 3: 200-217. and valley formation in Champlain Sea Clays: The
Chacoura River, Quebec, Canada, Geomorphology,
Crawford, C. B. 1968. Quick clays of eastern Canada. 145: 12-18.
Engineering Geology, 2(4): 239-265.
Lim, S. S. 2006. Experimental Investigation of Erosion in
Daneshi-Sadr, A.-H. 2018. An Experimental Parametric Variably Saturated Clay Soils.Ph.D. thesis, The
Study of the Effects of Physical and Chemical Soil University of New South Wales, NSW, Australia.
Properties on Erosion of Fine Grained Soils. M.Sc.
Thesis, Amirkabir University of Technology, Tehran, Locat, J., Lefebvre, G., and Ballivy, G. 1984. Mineralogy,
Iran. chemistry, and physical properties interrelationships of
some sensitive clays from Eastern Canada, Canadian
Delage, P., and Lefebvre, G. 1984. Study of the structure Geotechnical Journal, 21(3): 530-540.
of a sensitive Champlain clay and of its evolution during
consolidation, Canadian Geotechnical Journal, 21(1): Mitchell, R. J., and Klugman, M. A. 1979. Mass instabilities
21-35. in sensitive Canadian soils, Engineering Geology, 14(2-
3): 109-134.
Eden, W.J. 1972. Some observations at Le Coteau
landslide, Gatineau, Québec, Canadian Geotechnical Quigley, R. M. 1980. Geology, mineralogy, and
Journal, 9: 508-514. geochemistry of Canadian soft soils: a geotechnical
perspective, Canadian Geotechnical Journal, 17(2):
Eden, W.J., Fletcher, E.B. and Mitchell, R.J. 1971. South 261-285.
Nation River landslide, 16 May 1971, Canadian
Geotechnical Journal, 8(3): 446-451. Quigley, R. M., and Ogunbadejo, T. A. 1972. Clay layer
fabric and oedometer consolidation of a soft varved
Fell, R., Wan, C. F., Cyganiewicz, J., and Foster, M. 2003. clay, Canadian Geotechnical Journal, 9(2): 165-175.
Time for development of internal erosion and piping in
embankment dams, Journal of Geotechnical and Quigley, R., and Thompson, C. 1966. The fabric of
Geoenvironmental Engineering, 129(4): 307-314. anisotropically consolidated sensitive marine clay,
Canadian Geotechnical Journal, 3(2): 61-73.
Gaskin, S., Pieterse, J., Shafie, A. A., and Lepage, S. 2003.
Erosion of undisturbed clay samples from the banks of Sherard, J. L., Dunnigan, L. P., and Decker, R. S. 1976.
the St. Lawrence River, Canadian Journal of Civil Identification and nature of dispersive soils. Journal of
Engineering, 30(3): 585-595. the Geotechnical Engineering Division, 102(4): 287-
301.
Gillott, J. 1979. Fabric, composition and properties of
sensitive soils from Canada, Alaska and Norway, Shourijeh, P. T., Soroush, A., and Daneshi-Sadr, A.-H.
Engineering Geology, 14(2-3): 149-172. 2020. The effects of lime, bentonite and nano-clay on
erosion characteristics of clay soils, European Journal
Kenney, T. C. 1968. A review of recent research on of Environmental and Civil Engineering, 26(9): 3762-
strength and consolidation of soft sensitive clays, 3787.
Canadian Geotechnical Journal, 5(2): 97-119.
Soroush, A., Shourijeh, P. T., and Fouladi, S. R. 2018. The
Kenney, T. C., and Ali, M. S. 1968. Discussion of “Stability effects of soil erosion characteristics on critical filter
of Natural Slopes in Sensitive Clay”, Journal of the Soil design in embankment dams, Geotechnical Testing
Mechanics and Foundations Division, 94(5): 1185- Journal, 42(3): 789-816.
1190.
Tavenas, F., Chagnon, J.Y. and La Rochelle, P. 1971. The
La Rochelle, P., Chagnon, J.Y. and Lefebvre, G. 1970. Saint-Jean-Vianneylandslide: Observations and
Regional geology and landslides in the marine clay eyewitness accounts, Canadian Geotechnical Journal,
deposits of Eastern Canada, Canadian Geotechnical 8(3): 463-478.
Journal, 7(2): 145-156.
Torrance, J. K. 1979. Post-depositional changes in the
La Rochelle, P., and Lefebvre, G. 1971. Sampling pore-water chemistry of the sensitive marine clays of
disturbance in Champlain clays, ASTM International, the Ottawa area, eastern Canada, Engineering
STP 483: 143—163. Geology, 14(2-3): 135-147.
Lefebvre, G., Rohan, K., and Douville, S. 1985. Erosivity of Torrance, J. K. 1983. Towards a general model of quick
natural intact structured clay: evaluation, Canadian clay development, Sedimentology, 30(4): 547-555.
Geotechnical Journal, 22(4): 508-517.
Wan, C.F., and Fell, R. 2002. Investigation of internal
Lefebvre, G., Rohan, K., and Milette, J.-P. 1986. Erosivity erosion and piping of soils in embankment dams by the
of intact clay: Influence of the natural structure, slot erosion test and the hole erosion test, UNICIV
Canadian Geotechnical Journal, 23(4): 427-434. Report No. R-412, ISBN 85841379 5, School of Civil
 and Environmental Engineering, The University of New
South Wales, Sydney, Australia.

Wan, C. F., and Fell, R. 2004a. Investigation of rate of

erosion of soils in embankment dams, Journal of
Geotechnical and Geoenvironmental Engineering,
130(4): 373-380.

Wan, C. F., and Fell, R. 2004b. Laboratory tests on the rate

of piping erosion of soils in embankment dams,
Geotechnical testing journal, 27(3): 295-303.

Wan, C. F., and Fell, R. 2008. Assessing the potential of

internal instability and suffusion in embankment dams
and their foundations, Journal of Geotechnical and
Geoenvironmental Engineering, 134(3): 401-407.

Williams, D. R., Romeril, P. M., and Mitchell, R. J. 1979.

Riverbank erosion and recession in the Ottawa area,
Canadian Geotechnical Journal, 16(4): 641-650.
Tuesday, October 3, 2023

FOUNDATIONS II
Continuous Flight Augercast (CFA) Piles as
Deep Foundation Elements for a High-Rise
Building
Alexandre Almeida, Naresh Gurpersaud & Hiu Lee
Keller, Acton, ON, Canada

ABSTRACT
The construction of a 34-storey residential building required a deep foundation system competent enough to support the
applied loads, and, at the same time, mitigate excessive settlements. Continuous Flight Augercast (CFA) Piles were
selected as deep foundation elements for the building. A CFA pile is a cast-in-place reinforced concrete pile, constructed
in a single pass process. The main difference between a CFA pile and conventional bored piles is that the concrete is
pumped through the hollow core auger during auger withdrawal, maintaining the drill hole stable throughout the whole
drilling process. The CFA pile simple and quick installation methodology are key benefits when applied to deep soil profiles.
The CFA piles were connected to the structure through a central raft and surrounding pile caps. An overview of the project,
its geotechnical conditions, foundation design considerations, results of a full-scale static compression load test, quality
control program, and the construction results are highlighted in this paper.

RÉSUMÉ
La construction d'un immeuble résidentiel de 34 étages nécessitait d’un système de fondation suffisamment compétent
pour supporter les charges appliquées et, en même temps, atténuer les tassements excessifs. Les pieux à la tarière
continue ont été sélectionnés comme éléments de fondation profonde pour le bâtiment. Un pieux à la tarière continue est
un pieu en béton armé coulé sur place, construit en un seul passage. La principale différence entre les pieux à la tarière
continue et les pieux forés est que le béton est pompé à travers d’un noyau creux de la tarière pendant le retrait de la
tarière, maintenant le trou de forage stable tout au long du processus de forage. La méthodologie d'installation simple et
rapide des pieux à la tarière continue est un avantage clé pour les profils de dépôt de sol profond. Les pieux à la tarière
continue ont été reliés à la structure par un radier central et les têtes des pieux. L'ensemble du projet, les conditions
géotechniques, les considérations relatives à la conception des fondations, les résultats d'un essai de charge statique en
grandeur réelle, le contrôle de la qualité et les résultats de la construction sont présenté dans ce document.

1 INTRODUCTION robust serviceability analysis of the foundation system is

extremely important to avoid any excessive total and
Continuous Flight Auger (CFA) piles mainly differ from differential settlements of the structure. The concept of a
conventional bored piles due to the installation process. piled raft approach is often used in CFA pile design for
CFA piles are constructed by drilling and concreting in a high-rise buildings. Piled rafts consider not only the
single-pass operation (see Figures 1 and 2). The stability contribution of the pile stiffness to the settlement mitigation
of the drill hole is provided during auger withdraw by of the structure but also the contribution of the raft direct
maintaining a positive concrete pressure. The central contact with the bearing layer.
reinforcement cage installation is the last stage of the pile
construction. This simple installation makes CFA piles a
competitive alternative to traditional large-diameter shafts
or driven piles. Commonly, pile diameters of up to 1,200
mm can be achieved with depths of up to 45 m.
Quality Control (QC) of CFA piles assists in
minimizing any installation defects which can potentially
result in pile structural issues. The main component of the
QC of CFA piles is the Data Acquisition (DAQ) system. The
DAQ system provides real-time information to the driller
and inspector regarding the drilling rate, rotation, torque,
pouring pressure, pouring flow, and withdrawn speed. The
information of each pile is stored and can be easily
accessible through cloud-based software.
CFA pile design is especially critical for high-rise
buildings. CFA piles are inherently slender than a typical
large-diameter shaft. Also, the application of more
aggressive failure load determination methods means a
less stiff foundation element. In these circumstances, a
Figure 1. CFA Pile installation
Figure 2. CFA Pile installation process (from Almeida et al. 2021)

2 PROJECT OVERVIEW structure, on the other hand, has some rigidity which helps
compensate for this stiffness difference.
A 34-storey building is currently under construction and will Combining the analysis of the 2 mechanisms above,
house 273 no. residential units and 273 no. vehicular the experienced foundation designer can determine the
parking spots. The building required a substantially stiff suitable pile layout. For the job presented in this paper, the
foundation to provide both support for the applied loads final pile layout is shown in Figure 5. The Typical CFA pile
and to minimize the settlements. profile is shown in Figure 6. A 50 MPa concrete was used
Understanding how the applied loads from the structure in the piles. The CFA pile was also reinforced with a top
transfer to the foundation is a key element for the cage made of 6-20M vertical rebars coupled with 10M ties.
foundation layout determination. Typically, high-rise
buildings will have a central rigid core, where lateral loads
applied to the structure (i.e., earthquake or wind) will be
resisted. This configuration can be compared to the human
body’s spinal cord. Figure 3 illustrates this typical building
arrangement.
A load redistribution within the foundation system is
observed during a lateral load event. The axial
compression forces are increased on the side of the
building across the lateral load application plane, whereas
the axial compression forces will be decreased (or
potentially tension will be generated) on the side closest to
the lateral force application plane. The central core of the
building will as a result attract the greatest portion of this
load imbalance due to its stiffer nature. When CFA piles
are used as the solution for the foundation, this mechanism
usually requires a central piled raft with surrounding
smaller pile caps.
In addition, many interaction mechanisms within the
soil occur due to the applied loads, such as the pile-to-pile,
the pile-to-raft, raft-to-pile, and raft-to-raft interactions, as
shown in Figure 4. Typically, the central elements within a
raft will experience the highest magnitude of induced
settlements. Thus, its stiffness is reduced. The raft Figure 3. Interaction between applied wind forces and
building structure/foundation system
Figure 4. Pile and raft interactions: pile-to-raft and pile-to-
pile (top); and raft-to-raft and raft-to-pile (bottom) (Hain and
Lee 1978)

Figure 6. CFA pile structural profile and cross sections as

per Keller (2022a)

3 SUBSURFACE CONDITIONS

The Geotechnical Investigation included Standard

Penetration Tests (SPT), Pressuremeter Tests (PMT),
water content measurements (w), Grain Size Analysis, and
Organic Matter. A very competent silt and sand layer is
present below 17 m. The geotechnical conditions are
shown in Figure 7.

Figure 5. CFA pile layout for the building foundation (from

Keller 2022a)
 Figure 8. Load distribution onto the central piled raft (from
Keller 2022b)
Figure 7. Results from the geotechnical investigation as per
McClymont & Rak (2021)

4 CENTRAL PILED RAFT

The CFA pile design was developed with special attention

given to the piled raft. A 3D Finite Element Method was
used to simulate its behaviour.
As all forces, elements, and soil conditions are run
within the same model, the software automatically
considers the effects of the interactions that occur between
the foundation elements, as shown in Figure 4.
The mechanism of load distribution within the building
under a lateral load shown in Figure 3 can also be used to
understand the piled raft behaviour under the same lateral
load. Therefore, it is important to identify the worst-case
conditions of load distribution within the raft and the piles
to design for the Ultimate Limit State condition. This case
is shown in Figure 8 for the central piled raft.
The Serviceability Limit State is usually the governing
criterion for the piled raft design. Settlement is evaluated
as per the Structural Commentaries of the National
Building Code of Canada (2015) recommendations.
According to Fellenius (2023), if the settlement condition is
satisfactory within a pile group, the ultimate condition is
also usually satisfactory. The 3D FEM provides the benefit
to also generate settlement plots. The project criterion was
to maintain the total expected settlements below 25 mm.
The maximum estimated settlement as per the 3D FEM Figure 9. Settlement distribution across the central piled
model is equal to 14 mm and minimum settlement of 7 mm. raft for the serviceability case (from Keller 2022b)
The estimated settlement plot is shown in Figure 9.
Figure 10 shows the installed CFA piles during its
detailing process for the raft installation.
 failure load than the latter. In Canadian practice,
geotechnical consultants have often preferred the former
over the later due to its conservative nature. However, a
serviceability analysis is not regularly done, and by doing
so, there is a higher degree of certainty regarding the
building stability and behaviour. It can be argued that with
a reliable serviceability analysis, a more aggressive
determination method of the failure load of a single pile can
be used, such as the Hansen’s 80% method.

1st Load Test Stage

Q (kN)
0 1000 2000 3000 4000 5000 6000
0

Settlement(mm)
5
Figure 10. CFA piles of the central piled raft 10
15
20
5 FULL-SCALE STATIC LOAD TEST
25
5.1 Load Test Results 30
35
A full-scale static compression pre-production load test
2nd Load Test Stage
was conducted on a sacrificial pile to validate the design
parameters (see Figure 11). The loading was carried out Q (kN)
with general reference to ASTM D1143 (2020) Quick Test 0 1000 2000 3000 4000 5000 6000
(Procedure A) method. The test pile was a 500-mm 0
diameter, 22.3-m long CFA pile. Harsh winter conditions
Settlement (mm)

prevented the test pile to achieve its full structural capacity 20

by the test day. The load test was conducted in 2 days in
order to allow drilling to start for lightly loaded piles. Both 40
tests load-settlement plots are shown in Figure 12.
60

80
Figure 12. Load test results (from Keller 2022c)

A clear plunging failure was observed during testing at

approximately 5,000 kN, where additional load could not be
sustained by the pile.
Hansen’s 80% method tries to identify the point which
80% of the ultimate load would yield 25% of the ultimate
measured settlement. When this point is achieved during
testing, the test has failed geotechnically. This is done by
plotting √𝑠/𝑄 vs 𝑠, where 𝑄 is the applied load and 𝑠 is the
measured settlement of a pile load test. The linear portion
of this plot would indicate the ultimate load based on
equation 1 below. The ultimate settlement is calculated
according to equation 2. The Hansen’s plot for the load test
Figure 11. CFA pile static load test (Keller 2022c) is shown in Figure 13.
1
𝑄𝑢 = [1]
2√𝐶1 𝐶2
5.2 Single Pile Failure Criterion
𝐶2
The Canadian Foundation Engineering Manual (2004) 𝑠𝑢 = [2]
𝐶1
provides a literature review of 2 determination methods for
the failure load of a single pile: Davisson’s Offset Limit Where 𝐶1 and 𝐶2 are, respectively, the slope and y-
method and the Hansen’s 80% Method. According to intercept of the line.
Almeida et al (2018), the former yields a more conservative
 1st Load Test Stage
√𝑠/𝑄 [√mm/kN] 0.003

0.002

0.001

0.000
0 10 20 30 40 50
s (mm)
2nd Load Test Stage
0.003
√𝑠/𝑄 [√mm/kN]

0.002

0.001

0.000
0
40 20
60 80 100
s (mm)
Figure 13. Hansen’s 80% plot (Keller 2022c)

Thus, the test pile failure load was estimated as

approximately 4,900 kN as per the Hansen’s 80% method.

6 QA&QC PROGRAM

The main components of a reliable QA&QC program for

CFA piles are good field notes taken by the field personnel
and the Data Acquisition (DAQ) System, showing all
pertinent installation parameters, such as auger rotation
rate, depth of auger, torque and crowd forces, installation
date, volume of concrete, concrete pressure, depth of the
injection point, and auger withdrawal rate. DAQ records are
available in a real-time basis for the contractor’s design
engineer and/or operations personnel to review and
provide any required guidance. Anomalies can be detected
through this installation review process. Concrete strength
lab testing should also be part of the QA&QC program to
confirm that the required design concrete strength is
achieved.
Pile Integrity Tests (PIT) were also conducted as part
of the QA&QC program. The PIT is performed by
Figure 14. CFA pile head preparation before PIT: company
positioning one or two accelerometers on the pile head and
A (top) and company B (bottom)
using a hand-held hammer to impact the pile head. The PIT
equipment collects the acceleration data and provides a
plot which the aim is to identify the pattern of wave
The results submitted by company A condemned 5 out
propagation and reflection along the pile depth. The intent
6 tested piles. The results submitted by company B raised
is to identify any significant changes in the pile cross
questions regarding 3 out of 16 tested piles (all piles by
section.
company A were also tested by company B). Out of the 3
Two different companies were retained to perform PITs
questioned piles by company B, 2 of them were concluded
in this job. One critical part of the test is the pile head
to have bulged at a certain elevation, when comparing the
preparation. If the pile head is not correctly smoothened
PIT with the DAQ reports. Thus, no remediation was
before the test, the test results may be significantly
necessary for these piles. The other questionable pile by
compromised. Figure 14 shows how the pile heads were
company B was suspected to have its top section
prepared by the two companies.
compromised. This pile had its top section exhumed since
it was one of the piles with deep cut-off elevations for the
central piled raft. No structural issues were found at its top
section. During the time of PIT, this pile, however, had a Canadian Commission on Building and Fire Codes. 2015.
presence of a bar debonder pipe to facilitate its Structural Commentaries (User’s Guide – NBC 2015:
subsequential detailing process. It is believed that the gap Part 4 of Division B). National Research Council of
intentionally created inside the debonder was raised as a Canada.
structural issue of the pile. This difference in test results
shows that PIT is extremely dependent on the testing Federal Highway Administration (FHWA). 2007.
company’s experience. One simple and critical step that Geotechnical Engineering Circular No. 8 Design and
was overseen, the pile head preparation, and the Construction of Continuous Flight Auger (CFA) Piles,
overinterpretation of the test results could have cost the Report no. FHWA-HIF-07-03, Washington, DC, USA
owner a substantial loss both in financial and schedule
terms. Fellenius, B.H., 2023. Basics of foundation design—a
As recommended by the Federal Highway textbook. Electronic Edition, www.Fellenius.net, 548 p.
Administration (2007), an automated monitoring for quality
control of CFA piles should not be solely relied upon. Field Hain, S.J. and Lee, I.K. 1978. The Analysis of Flexible Raft-
observations and documentation are important aspects for Pile Systems, Geotechnique, 28(1), 65-83
QA&QC. A simple monitoring of pile connections during
pile installation can greatly assist on the quality assurance Keller. 2022a. CFA Piles Shop Drawings, Acton, ON
of the installed piles.
Keller. 2022b. CFA Piles Design Report – R01, Acton, ON

7 SUMMARY & CONCLUSIONS Keller. 2022c. Continuous Flight Augercast (CFA) Piles
Load Test Report, Acton, ON
CFA piles were successfully installed to support a 34-
storey high building. A detailed design process was McClymont & Rak. 2021. Supplementary Soil Investigation
followed to guarantee the building stability and Proposed CFA installation, Toronto, ON
serviceability conditions. The understanding of the
structure-foundation interaction is important when defining The Canadian Geotechnical Society. Foundations
the foundation layout of the building. A full-scale static pile Committee. 2004. Canadian Foundation Engineering
load test was conducted to confirm the pile design Manual, Canadian Geotechnical Society
parameters. A stringent QA&QC program was put in place
to validate the CFA production piles. PIT results can
complement the QA&QC program with careful review and
corroborated with as-built records/field observations.

ACKNOLEDGEMENTS

The authors would like to acknowledge the work performed

by the Field and Operations personnel from Keller and the
project team.

8 REFERENCES

Almeida, A., Gurpersaud, N. and Lee, H. 2021. Application

of high-capacity Continuous Flight Augercast (CFA)
piles in the Greater Toronto Area. Proceedings of
GeoNiagara 2021: 74th Canadian Geotechnical
Conference, Niagara, ON.

Almeida, A., Jesswein, M., Liu, J., Gurpersaud, N. and

Bruce, J. 2018. Evaluation of Determination Methods
for Ultimate Axial Capacity of Micropiles in Ontario
soils. Proceedings of GeoEdmonton 2018: 71st
Canadian Geotechnical Conference, Edmonton, AB.

ASTM D1143 / D1143M-20. 2020. Standard Test Methods

for Deep Foundation Elements Under Static Axial
Compressive Load, ASTM International, West
Conshohocken, PA, USA
Driven Pile Resistance Gain with Time –
Evaluation of Pile Load Tests: A Case Study
Darcy Hansen, P.Eng., Anastasia Poliacik, P.Eng., Lisa Coyne,
P.Eng., WSP Canada Inc.
Tony Sangiuliano, Andrew DeSira, Ministry of Transportation
Ontario

ABSTRACT
The Ministry of Transportation of Ontario (MTO) invested in a static pile load test during the detailed design of the Highway
400/Essa Rod overpass replacement in Barrie, Ontario, to determine the axial resistance of the test pile and to evaluate
the strength gain with time. The static pile load test supported optimization of the foundation design, which ultimately
included 312 HP 310x110 piles of lengths ranging from 30 to 35 metres, as well as development of acceptance criteria to
control the installation of production piles. The test pile achieved a relatively low ultimate geotechnical resistance on initial
driving, with negligible increase during restrike. However, the static pile load test demonstrated an appreciable
improvement over a multi-week period, with the ultimate geotechnical resistance increasing by approximately 150% and
210% after 38 days and 70 days, respectively. Significant cost and schedule savings were realized in the use of higher
geotechnical resistances in the pile design based on the load test data.

RÉSUMÉ

Le Ministère des Transports de l'Ontario (MTO) a investi dans un essai de charge statique sur pieux au cours de la
conception détaillée du remplacement du pont de l'autoroute 400/Essa Road à Barrie, en Ontario, afin de déterminer la
résistance axiale du pieu d'essai et d'évaluer l'augmentation de la résistance avec le temps. L'essai de charge statique a
permis d'optimiser la conception de la fondation, qui comprenait finalement 312 pieux HP 310x110 d'une longueur de 30
à 35 mètres, et de développer des critères d'acceptation pour contrôler l'installation des pieux de production. Le pieu
d'essai a atteint une résistance géotechnique ultime relativement faible lors de l'enfoncement initial. Cependant, l'essai
de charge statique du pieu a démontré une amélioration appréciable sur une période de plusieurs semaines, la résistance
géotechnique ultime augmentant d'environ 150 % et 210 % après 38 jours et 70 jours, respectivement. L'utilisation de
résistances géotechniques plus élevées dans la conception des pieux sur la base des données de l'essai de charge a
permis de réaliser des économies significatives en termes de coûts et de calendrier.

1 INTRODUCTION production piles. Discussion is provided on the gain in

geotechnical resistance with time, construction
The Highway 400/Essa Road Overpass Replacement in specifications, cost and schedule savings, and results
Barrie, Ontario involves the replacement of the existing obtained to date. Finally, “lessons learned” are provided for
50 m wide, six-lane, single-span overpass with an 80 m consideration in future pile load tests.
wide, ten-lane, two-span structure under MTO Contract
2022-2008. The new abutments were designed to be
supported on driven steel H-piles and the centre pier on
drilled shaft foundations (locally termed caissons). Static 2 PROJECT AND SITE CONDITIONS
pile load testing (SPLT) was completed during the design
phase to optimize the pile design, given the significant The existing Essa Road Overpass is located on Highway
number and length of piles required for support of the 400 in Barrie, Ontario. To accommodate the local road
abutments. Approximately 12 km to 14 km of driven piles widening, the proposed replacement is planned to be a
were estimated to be required to support the new two-span bridge with semi-integral abutments with a total
abutments. span length of about 94 m. Essa Road will require lowering
Typically, the capacity of a driven pile is predicted by as much as 1.5 m in places so that the new overpass
using static empirical methods during detailed design. meets vertical clearance requirements.
Dynamic analysis conducted during driving is a method to The location of the static pile load test was determined
verify the pile capacity during the pile installation. The most based on site access and proximity to the proposed
accurate method of pile capacity determined is the Static structure. The test was located approximately 50 m west of
Pile Load Test. the north abutment of the existing Essa Road Overpass.
This paper provides a comparison of the ultimate
geotechnical resistances assessed for driven steel H-piles
by static methods of analysis, static pile load testing
including both the ASTM quick and maintained tests, and
high-strain dynamic pile testing on the test pile and
3 SITE INVESTIGATION AND SUBSURFACE The subsurface conditions encountered in Borehole
CONDITIONS PLT-1 advanced at the pile load test site are generally
consistent with those encountered in Boreholes ERO-1 to

Figure 1: Stratigraphic cross-section along north abutment of Essa Road Overpass

ERO-12.
Two boreholes and one Dynamic Cone Penetration
Test (DCPT), advanced to depths of about 20 m to 30 m,
were completed as part of preliminary design. As part of
the detail foundation investigation, a total of twelve
boreholes (designated as ERO-1 to ERO-12) were 4 TEST PILE INSTALLATION AND DYNAMIC
advanced to provide coverage for the proposed overpass TESTING PROCEDURES
foundation elements and one borehole (designated as
PLT-1) was advanced at the proposed static pile load 4.1 Test Pile and Micropile Installation
testing location. In-situ Standard Penetration Testing (SPT)
was carried out in all boreholes and five standpipe The test pile (designated as TP-1) and associated reaction
piezometers were installed to monitor the stabilized micropiles were installed between November 4 and 22,
groundwater level at the site. 2019, approximately 5 m west of Borehole PLT-1. A
In general, the subsurface conditions at the Essa Road schematic of the test arrangement is shown in Figure 2.
Overpass consist of fill and surficial deposits underlain by The steel HP 310x110 test pile, equipped with an
an extensive non-cohesive deposit of sand to silty sand Ontario Provincial Standard Drawing (OPSD) 3000.100
approximately 30 m to 35 m in thickness, which has a Type I driving shoe, was driven to a tip elevation of
variable state of compactness ranging from loose to very Elevation 216.1 m (about 31.6 m below ground surface)
dense; however, the piles were installed predominantly in using a Liebherr H40/7 hydraulic hammer with a maximum
dense to very dense sand to silty sand with an estimated rated energy of about 55 kJ. The test pile consisted of two
internal angle of friction of 35°. This deposit is interlayered sections of HP 310x110 welded together, for a total test pile
with various fine-grained deposits and is underlain by very length of approximately 33 m.
dense or hard deposits of silts, sands and clays, which
extend to at least 40 m to 45 m below ground surface. A 4.2 Hiley and High-Strain Dynamic Testing
stratigraphic cross-section through the North Abutment of
the Essa Road Overpass is shown in Figure 1; Borehole A Hiley formula plot was completed in accordance with
PLT-1 at the static pile load testing site is projected onto MTO’s Standard Drawing SS103-11. High Strain Dynamic
this cross-section. Testing (more commonly known as Pile Dynamic Analyzer
The groundwater level is generally about 3 m below the (PDA) testing) was carried out in accordance with ASTM
natural ground surface. Pressurized groundwater D4945. Both Hiley and PDA testing were carried out at the
conditions (confined by discontinuous clayey sand End of Initial Drive (EOID) on November 4, 2019 and at the
interlayers) are present within the lower portion of the non- Beginning of Restrike (BOR) on November 11, 2019.
cohesive deposit.
 Figure 2: Static pile load test arrangement

purpose of this test, the load increments were to be added

5 STATIC PILE LOAD TEST PROCEDURES until either the estimated failure load (i.e., 3,600 kN) was
achieved or 30 mm (10% of the pile diameter) of
5.1 Pile Load Test Arrangement cumulative pile displacement was observed. The pile was
not to be loaded to failure, to allow for later testing to failure.
The load test arrangement was constructed between As shown in Figure 3, 29 mm of cumulative pile
November 19 and 22, 2019. As shown in Figure 2, the load displacement was observed after applying a load of about
test arrangement consisted of two W920x420 steel 2350 kN (or at the 13th load increment) approximately 3
reaction beams, two timber cribs, and four reaction hours after the quick test began. Therefore, upon
micropiles (46 mm diameter rebar grouted within 115 mm completion of the 13th load increment, the loads were
diameter holes to 30 m depth) to counteract the jacking removed in six decrements of approximately 390 kN each.
load. A hydraulic cylinder jack was used to transfer the load Applied Load (kN)
between the top of the test pile and the reaction beam.
A load cell was used to monitor the applied loads on the 4000 3000 2000 1000 0
test pile. Four dial gauges were set up radially on a 0.0
reference frame to measure the vertical movements of the
top of the pile as the test progressed. The accuracy of the 5.0
dial gauges is approximately ±0.13 mm. The dial gauge
readings were used as the primary measurement system 10.0
for pile axial movements and a wire line, comprised of a
15.0
horizontal medal rod welded to the reference beam and
vertical scale welded to the test pile plate, was used as the 20.0
Movement (mm)

secondary system for pile axial movements. The pile load

test arrangement was in general conformance with ASTM 25.0
D1143M.
30.0
5.2 Procedure A – Quick Test
35.0
On December 10, 2019, 38 days after the installation and
restrike of the test pile, a static pile load test was carried 40.0
out in general accordance with Procedure A – Quick Test Applied Load (kN) vs. Movement 45.0
method of ASTM D1143M. Based on an anticipated failure
load of 3,600 kN (equal to the estimated ultimate Pile Shortening
50.0
geotechnical resistance from static methods of analysis), Offset Limit Load
test loads were applied in increments of approximately 55.0
180 kN (about 5% of the anticipated failure load). Each Figure 3. Applied load vs. movement for Procedure A –
load increment was held for about 15 minutes. For the Quick Test (December 10, 2019 - 3 hours)
 of the test pile and led to the localized deformation/yielding
5.3 Procedure B – Maintained Test of the flanges at the top of the test pile.
As evident from the measured increased rate of
On January 13 and 14, 2020, 70 days after the installation displacement, it appears the test pile was approaching
and restrike of the test pile, a static load test was carried axial geotechnical failure. Based on this information and
out in general accordance with Procedure B – Maintained considering the penultimate load of 3150 kN was
Test method of ASTM D1143M. As the pile load test was approaching the estimated ultimate bearing resistance for
completed early in the design phase and the design load the HP 310x110 pile, the design team concurred that it is
was unknown at the time of testing, a design load of 1,800 unlikely that a higher bearing resistance could be justified
kN was assumed. Based on the assumed design, test for use in the design. As such, in consultation with the
loads were applied in increments of about 450 kN (or 25% owners (MTO), it was determined that there would not be
of the assumed design load). Each load increment was a significant benefit to repair the test pile and reaction
held for a minimum of 20 minutes and maximum of 2 hours, micropiles to complete additional testing for an anticipated
until the rate of axial movement did not exceed 0.25 mm nominal gain in ultimate axial geotechnical resistance.
per hour. For the purpose of this test, the load increments
were to be added until either 200% of the design load (i.e.,
3,600 kN) was achieved or progressive movement greater
than 45 mm (i.e., 15% of the pile diameter) was observed.
As shown in Figure 4, about 40 mm of cumulative
pile displacement was observed prior to applying the last
load increment approximately 8 hours after the test began.
Upon applying the eighth and final load increment, the
flanges on the test pile yielded and the jack shifted at a load
of about 3,500 kN, and no further loading or unloading
could be carried out. A photo of the yielded upper portion
of the test pile is shown in Figure 5.

Applied Load (kN)

4000 3000 2000 1000 0
0.0

5.0
Figure 5. Test pile after localized yielding of top of pile
during Procedure B – Maintained Test.
10.0

15.0 Lessons from this pile yielding failure are briefly

summarized in the “Lessons Learned” section of this paper.
20.0
Movement (mm)

6 GEOTECHNICAL RESISTANCE ASSESSMENT

25.0
6.1 Static Analysis During Detail Design (Pre-SPLT)
30.0
The ultimate (unfactored) axial geotechnical resistances
35.0 were assessed using various methods for steel HP
310x110 piles installed to depths ranging from about 25 m
40.0
to 35 m below ground surface (i.e., pile tip elevations
45.0 ranging from Elevation 226.5 m to 210.8 m) at the static
Applied Load (kN) vs. Movement pile load testing site based on Borehole PLT-1 and at the
Pile Shortening 50.0 proposed north and south abutments for the overpass
Offset Limit Load replacement based on Boreholes ERO-1 to ERO-12. The
55.0 methods included both β and Ks effective stress methods
Figure 4. Applied load vs. movement for Procedure B – outlined in CFEM, Poulos and Davis (1980), Meyerhof
Maintained Test (January 13, 2020 - 8.75 hours) (1976) and Decourt (1995), using an estimated initial
effective friction angle of 35° in the dense to very dense
silty sand to sand. The ultimate geotechnical resistances
5.4 Failure of Test Pile and Reaction System were estimated to range from 3,000 kN to 3,800 kN
including a combination of shaft resistance and end-
On the basis of site observations and monitoring results, bearing.
the cause of the failure was attributed to uplifting (up to
about 20 mm) of two of the four reaction micropiles which 6.2 Hiley Formula on Test Pile
caused an eccentric / inclined loading condition to develop;
this resulted in stress concentrations in a portion of the top
The Hiley formula calculations are shown in Table 1 and 6.5 Static Pile Load Test Procedure B – Maintained
were completed in accordance with the Standard Drawing Test
SS103-11. It should be noted that Standard Drawing
SS103-11 does not provide a value for the efficiency factor The pile movement measurement shown on Figure 4 is
for a hydraulic hammer; however, the ratio between the based on the average of the four dial gauge readings. The
measured energy and the energy input by the pile driving applied load on the test pile was measured from the load
operator (i.e., a measure of the efficiency of the pile driving cell output. Based on the results of Procedure B without
hammer) during PDA testing carried out at EOID and BOR extrapolation, the ultimate (unfactored) geotechnical
is slightly above 1.0. Therefore, for the purpose of the Hiley resistance of the pile after 70 days is assessed to be at
formula calculation, the efficiency factor has been taken as least 3,150 kN, which is equal to the last maintained load
equal to 1.0. increment prior to the termination of the test. Based on the
results of Procedure B with linear extrapolation, the
Table 1. Ultimate Geotechnical Resistance of Test Pile ultimate (unfactored) geotechnical resistance of the test
Based on Hiley Formula pile was further assessed, and the results are summarized
in Table 4.
Ultimate Pile
Pile Depth (m) Test Condition 1
Resistance (kN) 2 Table 4. Geotechnical Resistance from Maintained Test
31.6 to 31.8 EOID 1,675
31.8 BOR 1,625
Assessment Method Ultimate Pile Resistance (kN)
1
EOID denotes End of Initial Drive, and BOR denotes Beginning of
Restrike. Davisson Offset Method1 3,300
2
The ultimate geotechnical resistances presented are unfactored
values. 10 per cent of Pile Diameter 2,750
1
Davisson Offset Method based on linear extrapolation of the test
6.3 High-Strain Dynamic Testing (PDA Testing) on Test data.
Pile
Although Procedure B was not completed in its
The results of the PDA testing on the test pile at end of entirety, the test reached the penultimate load increment
initial driving (EOID) and beginning of restrike (BOR) are and the data obtained up to that point are considered valid.
provided in Table 2. The test was approaching the failure load as defined by
ASTM D1143M, which is the load at which the total axial
Table 2. Ultimate Geotechnical Resistance of Test Pile movement exceeds 15% of the pile diameter/width (or
from PDA 45 mm for an HP 310x110 pile). Based on the trend of the
data, it is anticipated that the test would have reached 45
Ultimate Pile mm of pile movement prior to reaching and maintaining
Pile Depth (m) Test Condition 1
Resistance (kN) 2 200% of design load at 3,600 kN, at which time the
31.6 to 31.8 EOID 1,500 unloading portion of Procedure B would have commenced.
31.8 BOR 1,550
1
EOID denotes End of Initial Drive, and BOR denotes Beginning of
Restrike.
2
The ultimate geotechnical resistances presented are unfactored 7 DISCUSSION
values.
7.1 Geotechnical Resistance Recommendation
6.4 Static Pile Load Test Procedure A – Quick Test
A comparison of the ultimate (unfactored) geotechnical
The pile movement shown on Figure 3 is based on the resistances and the factored ultimate geotechnical
average of the four dial gauge readings, and the applied resistances from the Hiley testing, PDA testing, and static
load was measured from the load cell output. Based on the pile load testing (SPLT) Procedure A and Procedure B is
results of Procedure A, the ultimate (unfactored) presented in Table 5.
geotechnical resistance of the test pile after 38 days was
assessed, and the results are summarized below. Table 5. Comparison of Geotechnical Resistance

Table 3. Ultimate Geotechnical Resistance from Quick Test Factored

Ultimate
Ultimate
Assessment Geotechnical
Test Geotechnical
Assessment Method Ultimate Pile Resistance (kN) Method Resistance
Resistance
(kN)
Davisson Offset Method1 2,600 (kN)1
Pile PDA 1,500 750
10 per cent of Pile Diameter 2,300 Driving
EOID Hiley 1,675 840
1
Davisson Offset Method based on linear extrapolation of the test
data. Pile PDA 1,550 775
Driving
BOR Hiley 1,625 810
 SPLT Davisson 2,600 1,560
8 IMPACT ON DESIGN AND CONSTRUCTION
Proc. A 10% of Pile 2,300 1,380

Davisson 3,150 1,890

8.1 Pile Acceptance Criteria
SPLT
Proc. B 10% of Pile 2,750 1,650 It was anticipated that the geotechnical resistance of the
1
Geotechnical resistance factor is based on Table 6.2 of CHBDC
production piles would experience a similar gain in capacity
CSA-S6:19 for a typical degree of understanding. A resistance over time, provided the piles were constructed similarly and
factor of 0.5 is used for dynamic / pile driving testing and 0.6 for driven to or below Elevation 216 m and into very dense,
SPLT. native silty sand to sand deposit.
As there was not an appreciable increase in
It is noted that the Davisson Method is commonly used geotechnical resistance measured at 7 days after EOID via
to estimate the geotechnical resistance of HP 310x110 Hiley and PDA, the design team decided there was little
piles, based on pile load test data (CFEM, 2004), whereas value in prolonging the restrike timing beyond the minimum
the 10% of Pile Diameter method is better suited to 24-hour period specified in OPSS.PROV 903. For a design
estimate the geotechnical resistance of larger piles. As factored ultimate axial geotechnical resistance of 1,890 kN,
such, based on the SPLT Procedure B test, it was the acceptance criteria for production piles shown in Table
recommended that a factored ultimate geotechnical 7 were included in the contract. PDA testing was specified
resistance of 1,890 kN be used for detail design of HP to be completed on at least 10% of piles.
310x110 piles that are driven to or below Elevation 216 m
and into the very dense, native silty sand to sand deposit. Table 7. Acceptance Criteria for Production Piles
Based on the applied load versus movement data from
the Procedure B static pile load test, as plotted on Figure 4, Minimum Ultimate
Test
it is estimated that the unfactored serviceability Test Method Geotechnical Resistance
Condition
geotechnical resistance for 25 mm of movement is (kN)
approximately 2,500 kN. The factored serviceability High-Strain EOID 1,500
geotechnical resistance for 25 mm of movement is Dynamic (PDA)
therefore 2,250 kN based on a geotechnical resistance Testing BOR 1,550
factor, ϕgu, of 0.9 for a static test and a consequence factor, EOID 1,625
ψ, of 1.0 for a typical degree of site understanding (CHBDC Hiley Testing
CSA-S6:19). BOR 1,625

7.2 Estimated Strength Gain with Time 8.2 Contract Specifications

The results of the PDA and Hiley tests performed at EOID The target criteria rely on an observed gain in geotechnical
were compared with the results of the PDA and Hiley tests resistance with time based on the SPLT Procedure B test
performed at BOR, and the results of Static Pile Load results measured approximately 70 days after test pile
Testing ASTM D1143M Procedure A at 38 days and installation. The target criteria are appropriate for design
Procedure B at 70 days, as summarized in Table 6. An provided that superstructure construction will not occur
increase of up to 210% was observed in the estimated until after the piles have experienced the majority of the
ultimate geotechnical resistance over a period of 70 days. anticipated strength gain. An Operational Constraint was
included in the contract that specified the superstructure
Table 6. Summary of Increase in Geotechnical Resistance construction could not occur for at least two months
with Time following production piling. Commencement of
substructure construction could be permitted prior to this
Ultimate Average two-month period.
Set-up
Geotechnical Geotechnical
Test Period Given the potential variability of the subsurface
Resistance Resistance
(Days)
(kN) Gain3 (%) conditions at this site, it was recommended that any test
Pile Driving results below the acceptance criteria be assessed by the
0 1,500 to 1,6751 - Foundation Engineering Specialist (FES) in conjunction
EOID
Pile Driving Approximately with the Design Team, and consideration should be given
7 1,550 to 1,625 1
BOR 0 (±3) to the measured results from PDA and Hiley testing for
SPLT Proc. A 38 2,6002 155 to 175 nearby piles.

SPLT Proc. B 70 3,150 2

190 to 210 8.3 Production Piling
1
The range of ultimate geotechnical resistance is obtained from
the Hiley formula and PDA testing. At the time of writing this paper, pile installation for the
2
The ultimate geotechnical resistance is obtained using Davisson portions of the new structure which are to accommodate
method, as this method considered to be better suited for HP the outside widening of Highway 400 (adjacent to the
310x110 piles. existing highway) has been completed. A total of 54 PDA
3
The approximate average gain in geotechnical resistance has tests were carried out, at EOID and BOR for various soil
been compared to the PDA and Hiley test results at EOID.
set periods, on 54 of the production piles. The results of
the PDA testing showed that about 80% of the piles impactful, as the fine grained non-cohesive and cohesive
achieved estimated ultimate (unfactored) resistances deposits found at this site proved challenging ground
greater than 1500 kN at BOR, as specified in the Contract conditions for pile driving in some locations.
Documents. In general, the piles saw an increase in
geotechnical resistance between the EOID and
subsequent BOR testing. The PDA test results from the
production piles are compared to the target BOR value and 9 LESSONS LEARNED
the results of the static pile test in Figure 5.
The following “lessons learned” are provided for awareness
and consideration for static pile load testing. Where
applicable, these recommendations should be addressed
in the terms of reference or specifications for static pile load
tests.

1. The installation of the reaction micropiles and the

setup of the reaction frame should be supervised
and signed off by the structural engineer who
designs the reaction frame, prior to commencement
of the static pile load test(s). An on-site inspection
of the reaction frame should be carried out by the
reaction frame designer prior to each pile load test
if and where multiple load test procedures are
completed. As part of this inspection, the reaction
frame designer/contractor should confirm and
document that the center of the reaction frame is
plumb with the center of the test pile, and that the
hemispherical bearing, jack, and load cell are
centered prior to commencing each pile load test.
Figure 5. Comparison of production pile PDA results with Where relatively high ultimate test loads are
target BOR value and static pile load test results. applied, the upper portions of the test pile flanges
should be reinforced to minimize the potential for
In cases where the estimated BOR values were lower localized deformation or buckling that may be
than 1,500 kN, the results were evaluated relative to the associated with non-concentric loading on the top of
depth of the pile, Hiley results (corrected for hammer the pile.
efficiencies measured during PDA testing), the blows per 2. The grout mixture (i.e., the cement, bentonite, and
last 0.2 m of pile penetration, field observations and the water ratio) used for reaction micropile installation
results of other PDA tests within the group of piles. The should be recorded and grout samples should be
strength gain observed in the production piles through PDA obtained at the time of installation and submitted for
testing was generally similar to or greater than that laboratory testing to confirm adequate grout
observed in the test pile. strength. Confirmation of grout strength should be
Based on review by the Foundation Engineering provided prior to commencement of any pile load
Specialist, the designers and MTO, it was decided based tests.
on the results of the PDA and the observed resistance gain 3. Each reaction micropile should be proof-tested to
that sufficient production piles in each quadrant had confirm it can withstand the planned maximum test
achieved the target BOR resistance and that the pile group load.
as a whole was anticipated to achieve the required design 4. It is recommended that the contract documents
loads given the two-month period of setup prior to specify surveying of the reaction frame be carried
superstructure construction. out using optical level shooting of fixed points on
each reaction pile at a specified frequency (e.g.,
8.4 Design Efficiency every hour), to provide more accurate observations
of movement. It is further recommended that review
The results of the pile load test demonstrating significant and/or alert levels for differential movement of dial
resistance gain with time allowed for optimization of the readings be established as a safety precaution
abutment pile design, decreasing the total number of piles related to the potential failure of the reaction system
required for the contract. The results of PDA testing during and/or the test pile.
production piling provided the contract administration and 5. It is recommended that data acquisition technology
design team with appropriate information to make informed be used in place of manual dial gauges to collect
decisions with regards to pile acceptance, in consideration consistent, real-time data throughout the static pile
of the resistance gain with time. This allowed for minimal load tests. Implementation of data acquisition
impact to the construction schedule, which may have had systems allows for field staff to be at a safe distance
to be extended if further pile driving was required to obtain from the hydraulic jack and reaction frame should
the target geotechnical resistances. This was particularly
 any sudden shifts/movements (failure) of the load test results also contributed to significant time and
reaction system or test pile occur. cost savings during construction.
6. Consideration should be given to installing vibrating The static pile load testing also provided lessons
wire piezometers to measure porewater pressures learned that should be considered for future pile load tests
prior to, during, and after test pile installation, which in similar conditions. Further assessment based on the
will allow for an improved understanding of initial results of longer-term static pile load tests at multiple sites
porewater pressure development and dissipation is recommended to permit further calibration of static
over time, and in turn could be used to correlate analyses with such results.
strength gain or relaxation over time.
7. If a hydraulic hammer is used for installation of piles,
OPSS.PROV 903 (Deep Foundations) requires the 11 REFERENCES
contractor to submit information on the hammer
energy, rated energy and operating efficiency; if this ASTM International D4945 Standard Test Method for High-
cannot be demonstrated in advance for a hydraulic Strain Dynamic Testing of Deep Foundations
hammer, the efficiency of the pile driving hammer
should be verified during production pile installation ASTM International D1143M Standard Test Methods for
using PDA testing. It is recommended that more Deep Foundations Under Static Axial Compressive
research / understanding be carried out to Load
determine whether the Hiley test method (which
remains MTO’s standard test on production piles) is Canadian Geotechnical Society. 2006. Canadian
still applicable and can be appropriately modified for Foundation Engineering Manual, 4th Edition. BiTech
use with a hydraulic hammer (as opposed to Publisher Ltd., British Columbia, Canada.
conventional diesel hammer). This could include
PDA testing to confirm the efficiency of the hydraulic Canadian Standard Association (CSA) Group. Canadian
hammer, as well as continued comparison of Hiley Highway Bridge Design Code (CHBDC 2019) and
and PDA testing on production piles. It is noted that Commentary on CAD.CSA S6:19.
some in the deep foundation industry are migrating
toward the use of PDA testing and elimination of Chapman, L.J. and Putnam, D.F. 1984. The Physiography
Hiley testing, which has safety implications related of Southern Ontario. Ontario Geological Society,
to placing personnel in or near to the “line of force”. Special Volume 2, 3rd Edition. Accompanied by Map
Based on the preliminary findings from the P2715.
production piles installed at this site to date, the
hammer efficiencies were found to vary between Golder Associates Ltd., 2021. Foundation Investigation
piles and depths. In general, the Hiley tests were and Design Report, Essa Road Overpass, Highway
not consistent with the results of the PDA, even 400/Essa Road Interchange Reconstruction, Barrie,
when corrected for the energy efficiency range Ontario, G.W.P. 2337-16-00, Assignment No. (MTO
observed during PDA testing. However, these GEOCRES No. 31D-767)
observations may be due to the specific conditions
at this site; hence, site-specific comparisons MTO’s Standard Drawing SS103-11 Pile Driving Control
between Hiley and PDA are recommended if and
where any reliance is to be placed on Hiley testing Ontario Provincial Standard Specifications, OPSS.PROV
results. 903 Construction Specification for Deep Foundations

Ontario Provincial Standard Drawings OPSD 3000.100

10 CONCLUSIONS Foundation Piles, Steel H-Pile Driving Shoe

The static pile load test carried out for the detail design at
the Highway 400-Essa Road Overpass site provided the
design team with information that allowed them to develop
an efficient design, which led to significant savings in the
number and length of piles, and to savings in time during
construction associated with production piling. Specifically,
a higher factored ultimate geotechnical resistance was
used in design than was suggested by static analysis or by
the results of Hiley and PDA testing on initial driving.
The results also allowed for the construction team
and foundation engineering specialist to assess pile
acceptance during construction where there were
deviations from the target estimated ultimate resistance, to
make informed assessments of the acceptance of the pile
groups during construction. In this regard, the static pile
Centrifuge Modeling for Cyclic Axial
Performance of Helix Piles in Sand

Naveel Islam, Lijun Deng, Rick Chalaturnyk

Department of Civil & Environmental Engineering-University of
Alberta, Edmonton, Alberta, Canada
& Luke Penner
Reaction Piling Inc., Nisku, Alberta, Canada

ABSTRACT
The article shows the axial load responses of five instrumented model aluminum piles subjected to incremental cycles of
pseudo-static loads through centrifuge model testing in dry sand. Average axial cyclic displacements of 1%, 2%, 5%, and
10% of the shaft or helix diameter were adopted in four packages. Later, a post-cyclic monotonic compressive
displacement was implemented. All load tests were conducted inflight at 20 g scale units. It was observed that, with
increased cycle intensity, the load-bearing capacity of the piles increased. However, a loss of strength occurs due to a
possible gap formed at the base and around helices. Moreover, the theoretical equations over predicted the ultimate axial
capacity of the piles. The piles with double helices and larger inter-helix spacing generally provided a higher load-bearing
capacity. Last, the distribution of the ultimate axial loads, the maximum loads from cyclic sequences, and the ultimate unit
post-cyclic shaft resistances are shown.

RÉSUMÉ
L'article montre les réponses aux charges axiales de cinq pieux en aluminium modèles instrumentés soumis à des cycles
incrémentiels de charges pseudo-statiques par le biais d'essais sur modèle de centrifugeuse dans du sable sec. Des
déplacements cycliques axiaux moyens de 1 %, 2 %, 5 % et 10 % du diamètre de l'arbre ou de l'hélice ont été adoptés
dans quatre ensembles. Plus tard, un déplacement compressif monotone post-cyclique a été implémenté. Tous les tests
de charge ont été effectués en vol à des unités d'échelle de 20 g. Il a été observé qu'avec l'augmentation de l'intensité du
cycle, la capacité portante des pieux augmentait. Cependant, une perte de résistance se produit en raison d'un éventuel
espace formé à la base et autour des hélices. De plus, les équations théoriques sur prédisent la capacité axiale ultime des
pieux. Les pieux à double hélices et à espacement inter-hélices plus important ont généralement fourni une capacité
portante plus élevée. Enfin, la distribution des charges axiales ultimes, les charges maximales des séquences cycliques
et les résistances post-cycliques unitaires ultimes de l'arbre sont présentées.

1 INTRODUCTION observed a continuous accumulative displacement even at

the low cycles. The higher the load amplitude cycle, the
Helical piles, an innovative deep foundation type, are larger the displacement rate. Khidri and Deng (2022)
popular in the construction industries, predominantly in performed field axial cyclic load-controlled tests on
Western Canada. These piles are used to support instrumented screw piles in the sand to measure the load
transmission towers, solar panels, wind turbines, and transfer within the pile segment. However, field tests are
building systems that might exhibit cyclic motions during affected by the enhanced cost of testing at the site.
their operational stage or earthquakes (Urabe et al., 2015; Centrifuge modeling of helix pile-soil interaction was
Al-Baghdadi et al., 2016; Schiavon et al., 2019; Brown et proved effective for monotonic testing (Levesque et al.,
al. 2019). Such cyclic motion in the axial direction of the 2003; Tsuha et al., 2012). Additional strain gauges as
piles might lead to an excessive settlement or axial pile instrumentations could provide useful insights into the axial
failure. load redistribution between the shaft and helices (Li et al.,
The design of helical piles subjected to vertical cyclic load 2022). However, studies on the cyclic nature of helix or
or seismic motions is usually overlooked in practice (Tsuha similar pile types in sands using centrifuge are very limited.
et al., 2012). Moreover, the current guideline includes Most research was focused only on the effect of pile
estimations for the axial load-bearing capacity of the piles geometries and the number of cycles over the load-bearing
for static conditions only (e.g., Mitsch and Clemence 1985; capacity of the piles. Centrifuge testing by Urabe et al.
Ghaly and Hanna 1991, CFEM 2006, Perko 2009). (2015) showed the degradation of shaft friction by about
A cyclic nature exhibits complex loading that demands 20-50% when subjected to cycles on both smooth and wing
appropriate experimental facilities. Some researchers piles in sand. Schiavon et al. (2019) performed similar tests
conducted load-controlled, axial cyclic field and model tests on helix anchors to ascertain the post-cyclic monotonic
on the sand to study the axial load response behavior of performance in the sand. Post-cyclic tests on such piles
piles (Jardine and Standing, 2012; Tsuha et al., 2012). showed a reduction in the ultimate uplift capacity of 4-9 %.
Through a continuous axial load in 15 cycles, El Naggar In contrast, there is a gap in appropriate knowledge on the
and Abdelghany (2007) showed a reduction of about 10% axial cyclic load response mechanism and capacity of helix
was observed in the helix piles. Petereit (1987), through a piles in sand.
cyclic load test on 1 g model helix anchors in the sand,
The study is intended to present the axial load responses Schiavon et al. (2016) showed through centrifuge tests
of smooth, single, and double helix piles in sand subject to that, for the model sand to impose no scale effect on the
incremental pseudo-static cycles through centrifuge tests. piles, the sand should consist of uniformly distributed grain
The developed centrifuge modeling technique (see Islam sizes, should not contain a lot of fine sand dust, and
et al. 2022) is adapted to accustom the cyclic nature of requires an effective helical radius of the model helix pile
displacements at stages. First, the axial load-displacement (i.e., rh) to the average grain size (d50) greater than 58.
relationship of a smooth and selected helix pile is shown at From the particle size distribution of Sil 3 sand, the
definite displacement stages and pile depths. Second, the estimated d50 = 0.29 mm and rh =19.05 mm; thus, rh/ d50 =
axial responses of all five model piles at the post-cyclic 65.69. Moreover, the estimated Cu = 1.5 mm and Cc = 1.08
monotonic compressive displacement stage are shown for mm for the model sand indicates a uniform distribution of
each strain gauge location. Moreover, the measured and grain sizes.
the estimated ultimate capacity from the post-cyclic and no-
cyclic stages are compared. Last, the internal axial ultimate y
(a)
load transfer and the ultimate shaft resistances are shown 811
over the pile depth. 51 709

2 CENTRIFUGE MODELING

112.7
A 183.1 171.4 171.2 A
P0CyC P2CyC P4CyC
The centrifuge model testing was conducted under the
designation of test series NI02. Islam et al. (2022) detail the B B

77.3
379
303
9 .7
4. Ø12
results of the test series NI01. In NI02, five model piles 11
Ø3
were tested in a model silica sandbox. Fig 1a shows the 268.8 P1CyC 171.5 P3CyC 8.1 268.8

113
test layout plan of the model piles. Fig. 1b-c shows the

38
x
cross-sectional profiles of the model piles arranged in two
NI02 (a)
separate carriageway sections on the model box. z
The piles are designated as PX L, where P stands for pile, A-A (b)
X indicates the pile number, and L stands for loading type.
The smooth pile is denoted as P0, and the helix piles, as P0CyC P2CyC P4CyC

100
P1 to P4. In the following sections, the L is designated
either as M or C to separate the non-cyclic monotonic (M) SG2 SG4 1D SG4
test (see Islam et al. 2022) from the post-cyclic monotonic 2D SG3

228.6
3.5D
(C) tests (current test). 1D SG3
2D
3.5D
300

All the model piles were installed at 1g but later tested in SG1 SG1 SG2 0.5D SG2SG1
flight at the 20 g scale units for each displacement 1.5D
0.3D

sequence. Table 1 summarizes the scale factors of the D d

71.4
Silica Sand
30

analyzed parameters from the prototype to the model scale x

of the pile. After completing each test pile (one at a time), NI02 (b)
the model piles were left in place until the end of the test of z

B-B (c)
the last pile. The minimum center-to-center pile spacing in
rows was kept four times the helix diameter to minimize test P1CyC P3CyC
zone influence (Li et al., 2022).
100

Table 2 summarizes the dimensions of the model piles in

model and prototype scale units. All tested five aluminum SG3 1D SG4
piles, i.e., one smooth (P0), one single-helix (P1), and three 3D 1D SG3
228.6

SG2
double-helix piles with helix spacing ratio (Sr) of 1.5 (P2), 0.5D SG2 1.5D 2.5D
300

2.5 (P3), and 3.5 (P4) had a prototype shaft diameter, (d) 0.3D
SG1 SG1

254 mm, and for the helix piles, the helix diameter, (D) 762
71.4

Silica Sand
mm. Common to practice, the pitch of the helices (P),
30

x
defined as the opening size of the helix, was set to 254 mm NI02 (c)
in prototype scale, equal to the shaft diameter of the pile. Vertical Smooth pile Vertical Helix pile
Axial Strain
Axial strain gauges (SGs) are installed at designated Gauges D=38.1mm d=12.7mm Unit: mm
locations along the pile shaft inside a grooved trench (see 0 12 24 36 inch
Fig 1b-c for the SG locations). Marked as SG1 from the tip
Figure
0 1.
.1 (a).2Layout
.3 plan
.4 of.5test .6
NI02;.7(b) Cross-sectional
.8 .9 1 meter
to the SG4 at the top end of the pile, these SGs are placed
profile at A-A section; and (b) Cross-sectional profile at B-
at a minimum distance of 10 mm from the helix blades to
B section. All dimensions are in model scale.
avoid any possible helix bending stresses. The setup and
calibration of the strain gauges are detailed in Islam et al.
Table 1. Selected scale factors (prototype/model) of
(2022).
centrifuge modeling tests
Table 3 summarizes the properties of the model silica
sand. The silica sand (Sil3) was procured from Sil Industrial
Minerals in Edmonton, Alberta. Sil 3 was selected as the Term Force Dimension Stress
effective model sand considering a reduced scale effect. Scale Factor 400 20 1
Table 2. Summary of Pile Dimensions in Model and Prototype Scale Unit

Type No. of Shaft dia. Helix Dia. Pile Length Helix Spacing Lower Helix Embedment Sr = (S⁄D)
helices d (mm) D (mm) L (mm) S (mm) E (mm)
Model Scale P0 0 12.7 38.1 271.8 NA. 150 NA.
Unit P1 1 12.7 38.1 271.8 NA. 150 NA.
P2 2 12.7 38.1 271.8 57.2 150 1.5
P3 2 12.7 38.1 271.8 95.2 150 2.5
P4 2 12.7 38.1 271.8 133.4 150 3.5
Prototype P0 0 254 762 5436 N.A. 3000 N.A.
Scale Unit P1 1 254 762 5436 N.A. 3000 N.A.
P2 2 254 762 5436 1144 3000 1.5
P3 2 254 762 5436 1905 3000 2.5
P4 2 254 762 5436 2668 3000 3.5

Table 3. Geotechnical properties of model silica sand constant rpm gear motor mounted to the actuator and the
head of the pile. The electric actuator assists in the axial
Term Value movement of the piles. It is controlled through computer 1,
Specific Gravity (Gs) (Correlated) 2.65 located in the control room. The soil model box,
1.61 constructed as an aluminum container, measures internally
Average Model density (), Mg/m3
709.2 mm (length) x 300 mm (width) x 400 mm (height) with
Constant Volume Friction angle (ϕcv), °.* 31.7
wall thickness ≥ 30 mm. Appropriate sand Pluviation
Relative density (%) (As-built) 49.7
technique was adopted to control the height and rate of
Void Ratio (As-built) 0.646
deposition of sand to ensure an optimum relative density
Note: *Measured from the direct shear test at the throughout the model box. The as-built relative density and
confinement stress level 50-400 kPa. void ratio of sand for centrifuge testing are shown in Table
3. Details of the adopted Pluviation technique and
As shown in Fig. 2, the centrifuge model assembly consists calibrations have been mentioned in Islam et al. (2022).
of three major parts: the dual-axis electric actuator affixed Last, the gear motor exerted a driving torque at a constant
at the top of the container, the soil model box, and the
Normalized Displacement (%)

22.5
20.0 300 600
180 400
17.5
P0 PL1 L1 SG2 P0 PL1 L2 SG2 P2 PL2 L1 SG2 P0 PL2 L2 SG2 P0 PL3 L1 SG2 P0 PL3 L2 SG2 P0 PL4 L1 SG2 P0 PL4 L2 SG2
15.0
P0 PL1 L3 SG2 P0 PL1 L4 SG2 P0 PL2 L3 SG2 P0 PL2 L4 SG2 P0 PL3 L3 SG2 P0 PL3 L4 SG2 P0 PL4 L3 SG2 P0 PL4 L4 SG2
120 12.5 225 300 450
10.0
60 7.5 150 200 300
5.0
0 2.5 75 100 150
0.0
Axial Load (kN)

-60 -2.5 0 0 0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50
Time (Hrs)
-120 -75 -100 -150
0.0 0.5 1.0 1.5 2.0 0 1 2 3 4 0 2 4 6 8 10 0 4 8 12 16 20
40 80 160 320
P0 PL1 L1 SG1 P0 PL1 L2 SG1 P0 PL2 L1 SG1 P2 PL2 L2 SG1 P0 PL3 L1 SG1 P0 PL3 L2 SG1 P0 PL4 L1 SG1 P0 PL4 L2 SG1
P0 PL1 L3 SG1 P0 PL1 L4 SG1 P0 PL2 L3 SG1 P0 PL2 L4 SG1 P0 PL3 L3 SG1 P0 PL3 L4 SG1 P0 PL4 L3 SG1 P0 PL4 L4 SG1
20 60 120 240

0 40 80 160

-20 20 40 80

-40 0 0 0

-60 -20 -40 -80

0.0 0.5 1.0 1.5 2.0 0 1 2 3 4 0 2 4 6 8 10 0 4 8 12 16 20

Normalized Displacement (%)

Figure 2. Centrifuge Model Assembly.

rotation rate of 23 revolutions per min (rpm)—a 12-volt DC Simultaneously, an axial displacement rate of 292.2 mm/
power supply assisted in maintaining the required torque. min (= 23 × P/min) was maintained using the electric
actuator. Such dual action of the actuator and the motor controlled cycles has been implemented in the literature
was to maintain the standard practice of installing the piles (e.g., Schivon et al., 2019; Khidri and Deng, 2022). The
at a penetration rate of one pitch of the helix per revolution; cyclic load sequences denoted by PL1 to PL4 are at a
to keep the soil disturbance minimal (Perko, 2009; Tsuha mean normalized displacement of 1%, 2%, 5%, and 10%
et al., 2012). Axial SGs are connected to the data logger of the shaft (for P0) and helix diameter (for P1-P4). Later,
channel at the side and later to the data logging computer a monotonic compression test (CL) was performed at the
in the control room. post-cyclic stage by pushing the actuator down at a
Figure 3 presents the sequences of axial loading constant 0.333 mm/min rate up to a normalized
incorporated to implement the cyclic and post-cyclic displacement of 20% of the shaft (for P0) and helix
monotonic compression on the piles. First, four sequences diameter (for P1-P4). It is to note that the constant rate test
of one-way compressive-tensile displacements are has also been used in past studies on centrifuge tests on
implemented in four loops at a constant 1 mm/min helical piles (Wang et al. 2013; Li et al. 2022).
displacement rate. A similar sequence of one-way load-
Normalized Displacement (%)

22.5
20.0 PL4 CL
17.5
15.0
12.5 PL3
10.0
7.5
PL2
5.0 PL1
2.5 Compressive
0.0 Pseudostatic Cyclic Displacements Displacement
-2.5
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50
Time (Hrs)
Figure 3. Stages of subsequent pseudo-static cyclic and compressive normalized axial displacements (w/d for P0 and w/D
for P1-P4) on the pile against time.

3 AXIAL BEHAVIOR OF PILES to SG1 (at the bottom). Along the rows, the load capacity
at the higher cycle intensity was considerably larger for a
The axial load vs. normalized displacements of selected particular SG. The area under the curve degrades
model piles are shown in Fig. 4 and 5. The relationships gradually with increased cycles and intensity. Such as, at
are established from the strain gauges located along the PL1, the first loop, L1, depicts the general trend of the load
pile depths. For each sequence of cycles (PL1 to PL4), the response for a typical helix pile in sand. However, with
curves L1 to L4 indicates each of the four loops. From Fig. additional loops, the strength degrades gradually. Later, at
4, the smooth pile (P0), it was observed that the axial load- PL2, the trend is different that PL1; there is a gradual
bearing capacity increased with higher cycle intensity. A strength decrease even during the loading stage, at about
similar loading-unloading trend was observed at each loop 3% normalized displacement. Such nature might be
at a particular sequence. At a definite stage, the first loop because of the gap of sand surrounding the helix location
cycle pushes a relatively undistributed sand layer, depicts caused by the previous loading sequence. Like the P0 pile,
the maximum load resistance, and later, the cycles a gradual horizontal shift of possible strength increase was
degrade in an area with repetitions. Moreover, a rapid observed from L1 to L4 at each PLs.
strength degradation was observed at the end stage of Figure 6 shows the maximum axial cyclic capacity
displacement (nearly about 0% at PL1 to about 5% at PL4). distribution along the pile depth for P0 to P4 (left to right)
Such nature illustrates the general gap effect at the pile for each cyclic load stage. It can be seen that, with
base, mostly observed during frequent loading-unloading increased cycle intensity, the capacity of the piles
and reloading sequences of motion. increased linearly with depth. The piles P2 to P4 had a
From Fig. 5, the helix pile (P3), it was noted that the cyclic double helix. It can be observed that the maximum capacity
load vs. normalized displacement relationship for a was higher in pile P3 with larger spacing in comparison to
particular load sequence (along the columns) is similar at pile P2.
the definite SG location. The load-bearing responses at the
SG4 (at the pile top) were considerably higher compared
 150
PL1 240
PL2 320
PL3 480
PL4
150 240 320 480
L1 L1
100
100
180
180
240
240
360
360
L2 L2
L3
50 120 160 240 L3
50 120 160 240 L4
0 60 80 120 SG2 L4
0 60 80 120
Axial Load (kN)
Axial Load (kN)

-50 0 0 0
-50 0 0 0
-100 -60 -80 -120
-100 0.0 0.5 1.0 1.5 -60 2.0 0 1 2 3 -804 0 2 4 6 8 -120
10 0 4 8 12 16 20
0.0 0.5
30 1.0 1.5 2.0 0 601 2 3 4 0 1202 4 6 8 10 0280 4 8 12 16 20
30 60 120 280
15 90 210
40
15 90 210
0 40 60 140
20
0 -15 60 140
30 70
20
0
SG1
-15 -30 30 0 70 0
0
-30 -45 -20 0 -30 0 -70
0.0 0.5 1.0 1.5 2.0 0 1 2 3 4 0 2 4 6 8 10 0 4 8 12 16 20
-45 -20 -30 -70
0.0 0.5 1.0 1.5 2.0 0 1 2 3 Normalized
4 0 Displacement
2 4 6 8 (%)
10 0 4 8 12 16 20

Normalized Displacement (%)

Figure 4. Axial Pseudostatic Cyclic Load vs. Normalized Displacement for Smooth Pile (P0).

150 PL1 240 PL2 320 PL3 480 PL4

600 1200 3600 6000 L1
450
100
800
180 240
4800
360 L2
2700
50 120 160 240
L3
3600
300 400
1800 L4
0 60 80 2400 120
150 0
Axial Load (kN)

-50 0
900
0 1200 0
SG4
0 -400
0 0
-100 -60 -80 -120
-150
0.0 0.5 0.0
1.0 0.5
1.5 1.0
2.0 1.5-8002.0 0 1 2 3 4 0 2 4 6 8 10 0 5
0
10
4 8
15
12 16
20
20
0 1 2 3 4 0.0 2.5 5.0 7.5 10.0
30 60 120 280
600 900 4800
2400
15 90 210
450 600 40 3600
1800
0 300 60 140
300 2400
20 1200
-15 30 70
150
0
600 1200
SG3
0
Axial Load (kN)

-30 -300 0 0
0
0 0
-45 -600 -20 -30 -70
0.0 0.5 0.0
1.0 0.5
1.5 1.0
2.0 1.5 2.0 0 1 02 13 2 4 3 4 0.0 2.5 0 5.02 4
7.5 6
10.0 8 10 0 5 0 104 15
8 20 16
12 20
300 600 2400
Normalized
1400 Displacement (%)
400 1800
200 1050
200
700 1200
100
0
0
350 600 SG2
-200
0 0
-100 -400
0.0 0.5 1.0 1.5 2.0 0 1 2 3 4 0.0 2.5 5.0 7.5 10.0 0 5 10 15 20

100 300 1000 1800

200 750 1350

50
100 500 900

0 0 250 450 SG1

-100 0 0
-50
0.0 0.5 1.0 1.5 2.0 0 1 2 3 4 0.0 2.5 5.0 7.5 10.0 0 5 10 15 20
Normalized Displacement (%)

Figure 5. Axial Pseudostatic Cyclic Load vs. Normalized Displacement for Helix Pile (P3).
 Qmax (kN)
0 100 200 300 400 500 0 1000 2000 3000 4000 0 1000 2000 3000 4000 0 1000 2000 3000 4000 5000 0 100 200 300 400

0.0

0.5

1.0
Depth in pile (m)

1.5

2.0

2.5
PL1
PL2
3.0 PL3
P0 P1 P2 PL4 P3 P4

Figure 6. Axial Maximum Cyclic Load Distribution along the pile length.

Figure 7 shows the post-cyclic monotonic 500compression

SG2
5000

load vs. normalized displacement relationship (a) for all five (d)
SG4
400
model piles from the SGs over the pile depth. All piles were 4000

displaced up to a normalized displacement300 of 20% of the 3000

SG3

shaft (for P0 only) or helix diameter. There is a nonlinear

increase in load resistances on all the piles200
with increased SG1 2000 SG2

displacement with maximum load observed for SG4

Axial Load (kN)

SG1
100
compared to SG1. The helix piles' bearing capacity was
1000

significantly higher than the smooth pile. However,

0 pile P4 0
did not represent the true load-bearing 5000 resistances,
0 5 10 15 20 0 5 10 15 20
200

possibly due to SG data logging errors. (b) (e)

Axial Load (kN)

4000 SG3
150
500 5000
SG2 3000
(a) (d) SG4
SG2 SG4
100
400 4000 SG3
2000

300 SG3 SG2

3000 SG1 50
1000

200 SG1 2000 SG2

0 0
Axial Load (kN)

0 5 10 15 20 0 5 10 15 20
SG1
100 4000 1000
(c) Normalized Displacement (%)
0 SG4
0 5 10 15 20 3000
0
0 5 Figure
10 7. Axial
15 20 Post Cyclic Compression vs. Normalized
Displacement: (a) P0; (b) P1; (c) P2; (d) P3 & (e) P4.
5000 200 SG3
(b) (e)
Axial Load (kN)

4000 2000
SG3
150
ForSG2the smooth pile, according to the pile load testing
guideline (ASTM D 1143), the failure load is stated by the
3000
1000 SG4
SG2 100 SG1

2000 stage
SG3
where the axial displacement reaches 15% of the
SG1
0
0 5 50 10 15
shaft
20
diameter.
SG2 For helix piles, however, there are no
1000 standard guidelines; however, in practice, it is common to
0
consider
Normalized Displacement
0
(%)ultimate capacity at 5% of the normalized
0 5 10 15 20 0 5 displacement
10 15 20 for sand (Sakr, 2011; Elsherbiny et al., 2013).
4000
(c) NormalizedThe theoretical(%)
Displacement estimated ultimate capacity was also
SG4 calculated for the smooth and helix piles following the
Canadian Foundation Engineering Manual (CFEM 2006):
3000
SG3

2000

SG2
for smooth piles:
1000
SG1
𝑄𝑢 = 𝜋𝑑𝐸𝜎𝑣 𝐾𝑠 𝑡𝑎𝑛𝛿 + 𝑁𝑡 𝜎𝑣 𝐴𝑠 [1]
0

and for helix piles:

0 5 10 15 20

Normalized Displacement (%)

 Eq
Pl
W
In
S
1 1500 R
𝑄𝑢 = 𝛾𝐻𝑏 𝐴𝑏 𝑁𝑞 + 𝛾𝐷𝐴𝑏 𝑁𝛾 + 𝜋𝑑𝐿𝜎𝑣 𝐾𝑠 𝑡𝑎𝑛𝛿 + 𝛾𝐻𝑡 𝐴𝑡 𝑁𝑞 + Pe
2
1 R
𝛾𝐷𝐴𝑡 𝑁𝛾 [2] 1250
R2= 0.834
A
2
Post-Cyclic Monotonic

Estimated Qu (kN)
1000 R2= 0.896
where Hb, Ht, E, and L stand for the depth to bottom helix, E
depth to top helix, lower helix embedment, and total length P
750
of pile, respectively; σv is the overburden stress; Ks states W
the coefficient of lateral earth pressure in compression In
500 S
loading; δ refers to the interface friction angle between soil
R
and pile material; As, Ab, and At indicates the surface area
250 P
of the smooth, bottom helix, and top helix of the piles, P0 M
P2 M
P0 C
P2 C
P1 M
P3 M
P1 C
P3 C R
respectively and Nq, Nγ are the dimensionless bearing P4 M P4 C
A
0
capacity factors for the local shear condition and N t refers 0 250 500 750 1000 1250 1500
to toe bearing capacity factor of the smooth pile (refer to Measured Qu (kN)
CFEM 2006).
Figure 8 shows the plot for the estimated ultimate capacity Figure 8. Comparison of Measured and Estimated Ultimate
from CFEM (2006) to the measured post-cyclic ultimate Axial Capacities from Monotonic and Post-Cyclic
capacity from centrifuge test results for the model piles (as Compression test.
in Figure 7). Further, measured values from non-cyclic load
tests (presented by Islam et al. 2022) were shown. It can The ultimate unit shaft resistance (qsU) on the smooth
be seen that; the post-cyclic capacity is reasonably lower portions of the pile segment between two adjacent SGs
compared to the monotonic loading sequences only. In was plotted along the pile depth in Fig. 9b. to study the
contrast, the methods in the current guideline might lead to nature of shaft resistance after loading. The unit shaft
overestimating the piles' capacity subjected to cyclic resistance (qs) was calculated as follows:
loading. The coefficient of determination (R2) changes from
0.896 to 0.834 from a monotonic only to a post-cyclic 𝑞𝑠 =
𝑄𝑡𝑜𝑝 −𝑄𝑏𝑜𝑡
[3]
monotonic loading stage. 𝐴𝑠
Figure 9a shows the axial ultimate compressive load
distribution along the depth in a pile from the SGs. The where Qtop and Qbot are the measured axial loads at the top
capacity of the piles with helices was comparatively higher and bottom of a definite shaft segment, and As is the outer
compared to smooth piles. With the increase in depth, the surface area of the segment. With the addition of helices,
ultimate load decreased. the ultimate unit shaft resistance increased over the length
of the pile.
Qult (kN) qsU (kPa)
0 50 100 150 200 250 0 500 2000 2500

P0 C
0.0 0.0 P1 C
P2 C
P3 C
0.5 0.5 P4 C
Depth in pile (m)

1.0 1.0

1.5 1.5

2.0 2.0

2.5 2.5

3.0 3.0
(a) (b)

Figure 9. (a) Axial Ultimate Load and (b) Ultimate Unit Shaft Resistance distribution along the pile length.

4 CONCLUSIONS increased depths. The following major conclusions can be

drawn:
The article overviews the axial cyclic behavior of 1. The implementation of the sequence of cyclic load
instrumented smooth and helix piles through centrifuge causes a reduction of load bearing capacity of the piles.
model tests at a 20 g scale to simulate the axial cyclic With increased intensity, a possible gap at the base or
operational/ seismic loading nature. The study also aims around the helices occurs due to the loading sequence
to provide the internal load transfer within the pile shaft with at the previous step.
2. Along the pile depth, the maximum axial cyclic load Khidri, M., and Deng, L. 2022. Field axial loading tests of
changes; the highest load occurs at the first loop of the screw micropiles in sand. Canadian Geotechnical
cycle and later degrades with subsequent cycles at a Journal, 59(3): 458-472.
particular load intensity. Levesque, C.L., Wheaton, D.E. and Valsangkar, A.J. 2003.
3. Comparing a pseudo-static monotonic and a post-cyclic Centrifuge Modeling of Helical Anchors in Sand,
monotonic stage with the estimated theoretical capacity Proceedings of the 12th Panamerican Conf. on Soil
Mechanics and Foundation Engineering, 2: 1859-1863.
shows a significant difference. Even though the
Li, W., Deng, L. and Chalaturnyk, R. 2022. Centrifuge
monotonic loading at the non-cyclic stage was
modeling of the behavior of helical piles in cohesive soils
comparable; however, the theoretical equations over from installation and axial loading. Soils and
predicts the capacity measured at the post-cyclic Foundations, 62(3):101141.
loading stages. Mitsch, M.P., Clemence, S.P., 1985. The uplift capacity of
helix anchors in sand: Uplift behavior of anchor
5 ACKNOWLEDGEMENTS foundations in soil. In Proc. ASCE, New York, pp. 26–47
Perko, H.A. 2009. Helical piles: a practical guide to design
This project is funded by the Natural Sciences and and installation. John Wiley & Sons.
Engineering Research Council of Canada (NSERC) and Petereit, R. 1987. The static and cyclic pullout behavior of
Reaction Piling Inc. The authors appreciate the plate anchors in fine saturated sand. MSc
experimental assistance of Yazhao Wang and Dmytro Thesis, Oregon State University, Corvallis, U S A.
Pantov from the University of Alberta. The Canadian Sakr, M. 2011. Installation and performance characteristics
Foundation for Innovation supported the development of of high capacity helical piles in cohesionless soils. DFI
the Centrifuge (GeoCERF) facility. Journal-The Journal of the Deep Foundations
Institute, 5(1): 39–57.
REFERENCES Schiavon, J. A., Tsuha, C. D. H. C., and Thorel, L. 2016.
Scale effect in centrifuge tests of helical anchors in
Al-Baghdadi, T.A., Brown, M.J. and Knappett, J.A. 2016. sand, International Journal of Physical Modeling in
Development of an inflight centrifuge screw pile Geotechnics, 16(4):185-196.
installation and loading system. In 3rd European Conf. Schiavon, J. A., Tsuha, C. D. H. C., Neel, A., and Thorel,
on Physical Modelling in Geotechnics. 239–244. L. 2019. Centrifuge modeling of a helical anchor under
ASTM D 1143/D 1143M-20. Standard Test Methods for different cyclic loading conditions in sand. International
Deep Foundation Element Under Static Axial Journal of Physical Modelling in Geotechnics, 19(2):72-
Compressive Load. West Conshohocken, VA. 88.
Brown, M., Davidson, C., Brennan, A., Knappett, J., Tsuha, C. D. H. C., Aoki, N., Rault, G., Thorel, L., and
Cerfontaine, B., and Sharif, Y. 2019. Physical modeling Garnier, J. 2012. Evaluation of the efficiencies of helical
of screw piles for offshore wind energy foundations. In anchor plates in sand by centrifuge model
1st International Symposium on Screw Piles for Energy tests, Canadian Geotechnical Journal, 49(9):1102-1114.
Applications. Urabe, K., Tokimatsu, K., Suzuki, H., and Asaka, Y. 2015.
CFEM. 2006. Canadian Foundation Engineering Manual. Bearing capacity and pullout resistance of wing piles
4th ed. Canadian Geotechnical Society, BiTech during cyclic vertical loading. In Proceedings of the 6th
Publisher Ltd., Canada. International Conference on Earthquake Geotechnical
El Naggar, M.H., and Abdelghany, Y. 2007.Helical screw Engineering, 358-367.
piles (HSP) capacity for axial cyclic loadings in cohesive Wang, D., Merifield, R.S., and Gaudin, C. 2013. Uplift
soils. Proceedings of the 4th international conference on behavior of helical anchors in clay. Canadian
earthquake Geotechnical engineering, Thessaloniki- Geotechnical Journal, 50(6): 575–584.
Greece, June, 25–8.
Elsherbiny, Z.H., and El Naggar, M.H. 2013. Axial
compressive capacity of helical piles from field tests and
numerical study. Canadian Geotechnical.
Journal, 50(12):1191-1203.
Ghaly, A. and Hanna, A. 1991. Experimental and
theoretical studies on installation torque of screw
anchors. Canadian Geotech. Journal, 28 (3): 353–364
Islam, N., Deng, L., Chalaturnyk, R., Penner, L. 2022.
Centrifuge modeling technique for axial compressive
loading on helical piles in sand. The 75th Canadian
Geotechnical Conference-GeoCalgary2022, Calgary,
Alberta, Canada, Oct 2-5
Jardine, R. J., and Standing, J. R. 2012. Field axial cyclic
loading experiments on piles driven in sand. Soils and
Foundations, 52(4):723–36.
Bi-Directional Static Load Test for Caissons
in Georgian Bay Shale: A Case Study in
Greater Toronto Area, Ontario
Madison Kennedy, Joe Carvalho & Lisa Coyne
WSP Canada Inc., Mississauga, Ontario, Canada
Tony J. Sangiuliano & Minkyung Kwak
Ministry of Transportation Ontario, Toronto, Ontario, Canada

ABSTRACT
Rock-socketed drilled shaft foundations for Ministry of Transportation Ontario (MTO) projects have traditionally involved
relatively long sockets relying on a significant proportion of sidewall resistance, which may not take full advantage of base
resistance. A shorter, more efficient rock socket was designed for bridge piers in weak to medium strong Georgian Bay
shale as part of the twinning of QEW/Credit River bridge in Mississauga, Ontario. An axial bi-directional static load test,
also called an Osterberg cell or “O-cell” test, was completed to measure the ultimate sidewall and base geotechnical
resistances of the rock socket on the test shaft, for comparison to and verification of the factored geotechnical resistance
used in the design of the production drilled shafts. These results expand the knowledge base for caisson foundation design
in the Georgian Bay rock formation, and demonstrate the effectiveness of SQUID, CSL and TIP in interpreting O-cell test
results and in quality control during construction.

RÉSUMÉ
Les fondations de puits forés dans la roche pour les projets du Ministère des Transports de l'Ontario (MTO) ont
traditionnellement impliqué des socles relativement longs reposant sur une proportion significative de la résistance des
parois latérales, ce qui ne prendre pas compte de la résistance de la base. Un caisson plus court et plus efficace a été
conçu pour les piliers de pont dans les schistes argileux de la Baie Georgienne de résistance faible à moyenne dans le
cadre de l'élargissement du pont QEW/Credit River à Mississauga, en Ontario. Un essai de charge statique bidirectionnelle
axiale a été réalisé pour mesurer les résistances géotechniques ultimes de la paroi latérale et de la base du fondation, à
des fins de comparaison et de vérification de la résistance géotechnique utilisée dans la conception. Ces résultats
élargissent la base de connaissances pour la conception des fondations de caissons dans cette formation rocheuse, et
démontrent l'efficacité du SQUID, du CSL et du TIP dans l'interprétation des résultats des essais et dans le contrôle de la
qualité pendant la construction.

1 INTRODUCTION approximately 1.7H:1V and 3.5H:1V on the west and east

sides of the river respectively.
A twin bridge has been designed and is under construction The existing bridge is an approximately 256 m long and
to carry the Queen Elizabeth Way (QEW) over the Credit 29 m wide, seven-span spandrel arch structure, with
River valley in Mississauga, Ontario, as part of a public- concrete arches at the piers. It was constructed as a four-
private partnership project administered by Infrastructure lane bridge in 1934 and was widened to six lanes in 1960.
Ontario (IO) and the Ministry of Transportation (MTO). The The twin bridge is being constructed immediately to the
original Reference Concept Design in bid stage north of the existing QEW Credit River bridge with a
incorporated twelve 1.8 m diameter drilled shafts (locally separation distance of ±2.0 m between the old and new
termed caissons) at each pier, with rock sockets extending structures. The new bridge consists of a three-span
approximately 8 m to 10.5 m into the Georgian Bay shale structure with caisson-supported piers located in the
bedrock. The pier foundation design developed in the bid floodplain on each side of the Credit River, and abutment
and execution stages incorporated ten 1.5 m diameter footings founded on bedrock near the crests of the valley.
caissons, with 2 m to 2.5 m long rock sockets. The
approach to geotechnical design and the results of bi-
directional static load testing are presented in the following 3 SITE INVESTIGATION AND SUBSURFACE
sections. CONDITIONS

A series of geotechnical investigations were carried out at

2 SITE DESCRIPTION – EXISTING AND NEW this site to determine the subsurface conditions and obtain
BRIDGE the data required for the reference concept design included
in the Design Build package. The geotechnical
The QEW/Credit River crossing connects Toronto, investigation consisted of geotechnical boreholes including
Mississauga and the Niagara Peninsula. The Credit River in situ tests such as optical and acoustic televiewer to
valley is approximately 20 m deep at the site and incised confirm the bedrock condition and packer testing to
into shale bedrock, with valley slopes oriented at
determine the hydraulic conductivity of the rock. Figure 1 this structure (CHBDC, 2019). For the design based on
shows the borehole locations and generalized stratigraphy. analysis, geotechnical resistance factors of 0.4 and 0.8
were applied for the ultimate limit state (ULS) and
serviceability state (SLS), respectively.

4.1 Calculation Methods

The geotechnical resistance of the caissons was estimated

based on a combination of resistance from the base of the
drilled shaft, and from sidewall resistance along the potion
of the shaft imbedded in competent bedrock.

4.1.1 Ultimate Geotechnical Resistance – Drilled Shaft

Base

The ultimate geotechnical resistance for the drilled shaft

based was evaluated using the Hoek Brown strength
model. The rock mass strength parameters were based on
the interpretation of the laboratory test results and an
estimation of the rock mass rating (RMR) based on the
expected mode of failure of the rock mass.
The relationships between the intact strength and the
field strength as a function of Geological Strength Index
(GSI, or RMR′76) are given in Equation 1 to 3 (Hoek et al.
2002).

GSI−100
( )
𝑚𝑏 = 𝑚𝑖 × 𝑒 28−14𝐷 [1]

GSI−100
Figure 1. Borehole location and soil strata at the QEW/ (
9−3𝐷
)
Credit River Bridge pier locations 𝑠=𝑒 [2]

1 1 −GSI⁄15
In general, the Credit River valley has been incised into 𝑎= + (𝑒 − 𝑒 −20⁄3 ) [3]
the shale bedrock with varying thicknesses and 2 6
composition/consistency of soil at each of the proposed
foundation elements. At the west pier, the subsurface
conditions within the floodplain consist of fill and clayey silt where 𝑚𝑖 is the intact rock Hoek-Brown m parameter;
to silty clay with organics underlain by sand and gravel and GSI is the geological strength index; and D is the
clayey silt (residual soil). At the east pier, the subsurface disturbance factor. The selection of the Hoek Brown
conditions consist of fill underlain by clayey silt or silty parameters is further discussed below.
sand, organic soils, and silty sand containing organics and The geotechnical resistance calculation was based on
clayey silt pockets. Shale bedrock underlies these soils at a two-wedge failure, below and adjacent to the foundation,
depths of approximately 6 m and 7 m below floodplain level as shown in Figure 2. In the case of a caisson the rock
at the west and east piers, respectively. below and adjacent to the caisson base is confined by the
The bedrock consists of thinly laminated to medium- rock and overburden surrounding it and, therefore, can be
bedded grey shale of the Georgian Bay Formation, the considered to be in a triaxial stress state.
upper portion of which is moderately to slightly weathered, The ultimate geotechnical resistance of the caisson
becoming slightly weathered to fresh with depth. The shale base was calculated using Equation 4 (Hoek et al. 2002).
is very weak to weak with an average unconfined
compressive strength of 15 MPa, and it contains stronger 𝑎
𝑚 ∙ 𝜎3𝐴
limestone layers. 𝑞𝑢𝑙𝑡 = 𝐶𝑓1 [𝜎3𝐴 + 𝜎𝑐𝑖 ( + 𝑠) ] [4]
𝜎𝑐𝑖

4 ANALYSIS BASED FOUNDATION DESIGN

where, 𝜎3𝐴 is given by Equation 5,
The foundation design was initially carried out based on
analysis and was subsequently confirmed with in situ 𝑎
testing. The QEW/Credit River bridge is considered a 𝑚 ∙ 𝑞𝑠
𝜎3𝐴 = 𝑞𝑠 + 𝜎𝑐𝑖 ( + 𝑠) [5]
lifeline structure, and as such a consequence factor of 𝜎𝑐𝑖
Ψ = 0.9 was used based on a high consequence level for
 carried out at each pier location based on the different
loading conditions provided by the structural engineer to
determine the settlement of the foundation.

4.2 Parameter Selection

The parameters required to complete the analysis of the

geotechnical resistance were obtained from information
and laboratory testing carried out during the previous site
investigations which included core logging, unconfined
compressive strength tests, point load tests and downhole
optical televiewer logging.

4.2.1 Rock Mass Quality (RMR76 / GSI)

To quantify the rock mass quality the Rock Mass Rating

(RMR) estimates using Bieniawski’s 1976 ratings (typically
5 points less than the 1989 ratings) were used for
estimates of strength. The 1976 rather than 1989 ratings
Figure 2. Illustration of the model for the calculation of the have been used because the original relationships
caisson base ultimate geotechnical resistance (modified between rock mass Hoek-Brown parameters and intact
from Wyllie, 1999). Hoek-Brown parameters were based on the 1976 ratings.
In the estimation of RMR, the spacing rating was based
on the most adverse joint set, or the joint set of most
where 𝑞𝑠 , is the vertical confining stress of the wedge significance to the failure mode. Due to the nature
adjacent to the caisson base and 𝐶𝑓1 is a shape factor (orientation of bedding, typical frequency of non-sub-
horizontal joints) of the Georgian Bay Shale the joint
which is 1.2 for a circular foundation shape (Sowers, 1970).
spacing rating was evaluated in two different ways
depending on the analysis being carried out.
4.1.2 Ultimate Geotechnical Resistance – Drilled Shaft
For the ultimate geotechnical resistance of the
Side Wall
Georgian Bay Shale, the vertical and sub vertical cross-
jointing is more significant than bedding, as failure of the
In practice, socket side resistance capacity is calculated by
foundation across the bedding would require rupture
assuming a single average value of unit side resistance
across the intact rock. Vertical cross joints typically have
acting along the concrete-rock interface, for each rock
spacing in the range of meters, and as such the sub vertical
layer. This value is multiplied by the surface area of the
joint spacing was used to calculate the RMR 76 for
shaft to obtain a total side resistance, given by Equation 6
geotechnical resistance.
(CGS, 2006).
When considering serviceability geotechnical
resistance, or settlement of the foundation, the bedding in
the Georgian Bay Shale plays a major role in the stiffness
𝑄𝑠 = 𝑓𝑠𝑢 × 𝜋𝐷𝐿 [6]
of the rock mass because the load acts perpendicularly to
it. In the vertical direction, the stiffness of the rock mass is
mostly controlled by the compression of the bedding,
The parameter 𝑓𝑠𝑢 was estimated using the relationship
especially if clay seams are logged in the rock core.
proposed by Horvath and Kenney (1979), Equation 7.
Therefore, the rating for spacing in the calculation of RMR 76
for settlement purposes should be based on the bedding
spacing, or its fracture frequency; also, the joint condition
𝑓𝑠𝑢 = 𝑏√𝑞𝑢 [7] rating should consider the continuity and the
roughness/infilling of the bedding.
The RMR76 values obtained for the strength and
which correlates the uniaxial compressive strength (𝑞𝑢 ) modulus parameters at this stie are presented relative to
of the weaker material (rock or concrete) and the shaft elevation in Figure 3.
roughness. For the caissons, a shaft roughness of R1 for
straight, smooth-sided socket, grooves or indentations less 4.2.2 Unconfined Compressive Strength and Intact
than 1 mm deep (after Pells et al. 1980) was selected and Modulus
a roughness value (b) of 0.45 was selected based on data
summarized by Rowe and Armitage (1987). The unconfined compressive strength and the elastic
modulus of the intact rock was obtained from uniaxial
4.1.3 Serviceability Geotechnical Resistance compression tests instrumented with strain gauges, as well
as correlations with Point Load Tests (PLT). The UCS from
The settlement of the pier foundations was estimated using the laboratory testing and point load testing, as well as the
the rock mass properties and the commercially available
program RSPile by Rocscience Inc. A group analysis was
intact modulus results for the pier foundations are bedding is consistently greater (RQD = 100) or smaller
summarized in Figure 4. (RQD = 0) than 100 mm. Therefore, the spacing should be
used directly in the estimation of the vertical modulus
according to Equation 8 (Goodman, 1989).

1 1 1
= + [8]
𝐸𝑛 𝐸𝑖 𝑘𝑛 𝑠

Where En is the normal modulus, where stress is

applied normal to the joints, Ei is the intact modulus of the
rock, kn is the stiffness of the contact, and s is the joint
spacing.
The spacing is readily available from drillhole records;
however, the stiffness of the bedding will depend on how
tight the bedding is, or whether there are softer clay seams
and how thick they are. The best way to assess the
tightness of the bedding is using optical televiewer images,
where available.
In a horizontally layered rock mass, the lateral
(horizontal) modulus is mostly controlled by the intact rock
strength and not affected by the bedding. If vertical or sub-
vertical jointing is widely spaced, the lateral modulus of the
rock mass should approach intact values.

Figure 3. Rock Mass Rating vs. Elevation 4.2.4 Hoek-Brown Parameter mi

The estimation of the intact Hoek-Brown parameter 𝑚𝑖 is

typically obtained from curve fitting laboratory results of
tensile, uniaxial and triaxial testing. In the absence of
triaxial testing, suggested values by Hoek and Brown
(reported in Hoek and Marinos, 2000) can be used, or
simple relationships between tensile strength and uniaxial
compressive strengths can be used as reasonable
estimates.
The suggested values for 𝑚𝑖 for shales by Hoek and
Brown (Hoek and Marinos, 2000), when no other data is
available, are 6±2. While no direct tensile or Brazilian tests
were undertaken, a large number of axial and diametral
Point Load Tests (PLT) were done on the recovered rock
core samples. The ISRM suggested method for
determining point load strength suggests that the 𝐼𝑠(50)
value is approximately 0.8 times the Brazilian tensile
strength (ISRM 1985).
Given an average UCS of 15 MPa and an average
tensile strength of 0.72 MPa, the 𝑚𝑖 is estimated to be 17.4
using the more conservative Hoek and Martin approach.
Experience in the Georgian Bay shale suggests that values
of 𝑚𝑖 may be lower. Therefore, a value of 8 representing
the higher end of the range suggested by Hoek and Brown
was adopted for the design.

Figure 4. Unconfined Compressive Strength vs Elevation 4.3 Analysis-Based Results

The factored ultimate geotechnical resistance at ultimate

4.2.3 Rock Mass Modulus limit state was calculated to be 10,880 kPa for the base
resistance and 630 kPa for the sidewall resistance. The
To obtain the rock mass modulus the RMR can be used to design value for the factored ultimate geotechnical
account for the discontinuities of the rock mass. However, resistance based on the analysis was 17,500 kN and
using RMR for calculating the vertical modulus (stiffness) 19,000 kN for the approximately 2.0 m and 2.5 m minimum
of the rock mass can be inaccurate, especially if the rock socketed lengths, respectively. The factored
serviceability geotechnical resistance for 10 mm of 1L-10 because the ultimate capacity had been reached on
settlement was 17,500 kN and 18,500 kN for the 2.0 m and the sidewall of the rock socket above the Osterberg cell
2.5 m rock socket length, respectively. assembly, and higher loads could not be sustained. The
test shaft was then unloaded in five decrements.

5 AXIAL BI-DIRECTIONAL STATIC LOAD TEST

77. 12 m
AS-BUILT
An axial bi-directional static load test (BDSLT, also referred ~76.06 m 1.06 m

to as an Osterberg cell or “O-cell” test) was completed on GROUND SURFACE

a test shaft that was drilled near the west pier of the Twin
Bridge. The purpose of the BDSLT was to measure the TOP OF
1.44 m
74. 62 m

ultimate base and sidewall geotechnical resistances of the

CONCRETE

3.8 m
rock socket, for comparison to and verification of the TELLTALE

factored geotechnical resistances used in the design of the S.G. 73.94 m

production drilled shafts for the piers.

5.1 Test Methodology 72.26 m (May 13, 2021– 8:05AM) 1525 mm O.D.
CASING
OVERBURDEN
The test shaft was constructed to match the production
drilled shafts and included a permanent steel casing with a S.G. 71.44 m

1524 mm outside diameter and 1500 mm inside diameter, 70. 5 m

with a 1500 mm diameter rock socket.

The BDSLT was conducted using a 27 MN-rated RIVER STONE ~69.46 m

Osterberg cell, having a diameter of 860 mm and using a

steel base plate of 940 mm diameter in the “Chicago EL. 68.86 m

method” of testing, as described in FHWA Publication No. WEATHERED ROCK 68.15 m S.G. 68.44 m
FHWA-NHI-10-016 (Drilled Shafts: Construction S.G. 67.94
Procedures and LRFD Design Methods). The nominal SLIGHTLY m

1500 mm diameter rock socket was designed with a WEATHERED TO

FRESH ROCK
minimum 3.0 m rock socket length in the slightly weathered ~3.0 m
S.G. 66.94
m
to fresh bedrock, plus an additional drilling depth of about
0.6 m to accommodate the O-cell assembly (load cell and S.G. 65.94
m
steel plates, plus concrete levelling layer). Based on the 65.01m

results from Golder’s boreholes within the west pier 64.5 m

footprint, the rock socket was designed with its base at
0.505 m
0m 1m 2m 3m

Elevation 64.5 m and top at about Elevation 68.2 m; the

Figure 5. Profile of the test shaft showing elevations of
permanent casing was required to extend a minimum of
relevant components
0.5 m into the weathered bedrock, and not to extend below
Elevation 68.2 m. A profile illustrating the test setup is
provided in Figure 5.
Throughout the test, the top of shaft displacement was
Inspection of the test shaft construction was completed
monitored using a pair of automated digital survey levels
as follows: SONICaliper measurement of the rock socket
from an average distance of approximately 16 m. The
to confirm the diameter and integrity; rock socket base
upper compression displacement was measured using
inspection via SID (shaft inspection device) and SQUID
6 mm telltale rods positioned inside casings embedded
(Shaft Quantitative Inspection Device) to confirm base
within the test shaft and monitored by linear vibrating wire
cleaning procedures; and crosshole sonic logging (CSL)
displacement transducers (LVDTs) attached to the top of
and thermal integrity profiling (TIP) to assess concrete
the shaft. The Osterberg cell expansion was measured
integrity.
using LVDTs. All instrumentation was monitored using a
The axial bi-directional static load test was conducted
datalogger collecting data at 30 second intervals.
in general accordance with ASTM 8169-18 (Standard Test
Method for Deep Foundations Under Bi-Directional Static
5.2 BDSLT Results
Axial Compressive Loading) using Procedure A: Quick
Test loading schedule, until termination of the test upon
The instrumentation installed in the test shaft allowed for
reaching the ultimate loading condition. The test
the determination of the based strength and sidewall shear
progressed with the Osterberg cell assembly pressurized
strength (or mobilized stresses) if the failure load is not
in ten nominally equal increments, resulting in a maximum
reached). The stress strain relationship (stiffness) of the
bi-directional load of 13.37 MN applied to the shaft above
base of the base and sidewall was also obtained. These
and below the Osterberg cell. Each successive load
parameters are important for the assessment of the design
increment was held constant for eight minutes by manually
capacity of the production drilled shafts, the settlement,
adjusting the Osterberg cell pressure, followed by an
and the load sharing between the sidewall and the base.
approximately one-minute period to increase to the next
The results of the BDSLT on the test shaft are shown in
loading increment. The loading was halted after increment
Figure 6.
 zones above and below showed a consistent response and
were used to obtain the sidewall shear strength for the
moderately weathered to fresh bedrock. Similarly, the
sidewall shear strength for the weathered shale zone was
obtained from instruments bounding the weathered zone of
the rock.

Figure 6. O-Cell load-displacement – gross load-

displacement graph for the test shaft base and sidewall

5.2.1 Sidewall Resistance

The ultimate sidewall capacity of the test shaft was

calculated using Equation 9.

𝑃
[9]
𝜋𝐷 × 𝐿

where P is the maximum applied load before significant

upward movement, D is the shaft diameter and L is the rock Figure 7. Shear strength profile - socket shear strength
socket length. The diameter of the test shaft was 1.5 m and profile and net load vs displacement during O-cell test
the length of the test socket consisted of approximately loading
0.7 m and 3.15 m of weathered and fresh rock,
respectively, for a total length of 3.85 m.
The test shaft was loaded to 13,400 kN before The following ultimate (unfactored) sidewall capacities
significant upward movement of the test shaft socket was were obtained from the bi-directional static load test:
observed without increase in the load. The resulting shear • Moderately weathered to fresh shale: 1,200 kPa
strength profile is shown in Figure 7. • Weathered shale: 600 kPa
Based on this result, an average ultimate sidewall
capacity of 740 kPa was obtained. However, for this test 5.2.2 Base Resistance
pile the average ultimate sidewall capacity is not
considered appropriate for assessment of the ultimate When significant upward movement of the test shaft socket
capacity of the test shaft and production rock sockets due occurred (P = 13,400 kN), the axial bi-directional static load
to defects observed in the lower section of the test shaft test had not reached the ultimate resistance at the base.
(above the O-Cell). The lower average ultimate side friction While the ultimate base resistance was not fully mobilized,
capacity of the test shaft socket in the Osterberg cell test is the test proved a base resistance of at least 19,300 kPa
attributed primarily to an anomaly indicative of a reduced over the loaded area (base plate of 940 mm diameter)
concrete cover in the lowest zone of the test shaft socket using the Chicago method. The resulting mobilized base
which was observed in both the CSL and TIP testing, strength versus displacement is shown in Figure 8.
discussed in further detail below.
For design purposes based on the results of the bi- 5.2.3 Inspection and Quality Assurance
directional static load test, the instrumented zone of the
moderately weathered to fresh rock socket where the
anomaly was not observed and all the strain gauges in the
 concrete cover in this zone which was deemed acceptable
for the O-Cell test. The anomaly from 8.8 m to 9.4 m
corresponded with lower sidewall shear strength
measurements as discussed previously.

6 DRILLED SHAFT CAPACITIES BASED ON STATIC

TEST RESULTS

Based on the results of the BDSLT, geotechnical

resistance factors of 0.6 and 0.9 were applied for ULS and
SLS, respectively, and a consequence factor of Ψ = 0.9
was applied.
The results derived from the axial bi-directional static
load test were directly applicable to the design of the
production drilled shafts. The sidewall shear strength
obtained provided appropriate design values for the
moderately weathered to fresh shale in the rock socket,
and for the weathered shale above the top of the rock
socket and below the permanent liner. Moreover, based on
the borehole results and geotechnical laboratory testing
from boreholes drilled within the pier footprints, the average
unconfined compressive strength and modulus was
considered to be consistent at and below the production
drilled shaft founding level.
Figure 8. Mobilized base strength vs displacement for The results obtained from the bi-directional static load
0.94 m diameter base plate in Chicago method, and test results were used to create load deformation plots and
adjusted for 1.5 m diameter base plate loading profiles for a single drilled shaft using RSPile (by
Rocscience Inc.). For the analysis the caisson was divided
into 50 elements. For estimating the serviceability
The results of the inspection or quality assurance testing geotechnical resistance for 10 mm of settlement the
aided in the interpretation of the results as indicated in analysis included the compression of the caisson as well
earlier sections. as the geotechnical settlement.
The SONICaliper results indicate that the calipered The estimated unfactored and factored ULS and SLS
diameter of the rock socket above the Osterberg cell for the 1.5 m diameter caisson with 2.0 m and 2.5 m long
assembly was 1500 mm, in accordance with the design for rock sockets are summarized in Table 1.
the test and production shafts.
Based on observations from the downhole camera
video, the shale bedrock in the rock socket sidewalls was Table 1. Estimated Geotechnical Resistance Values from
observed to be of good to excellent quality from the bottom Static Load Test
of the casing to the water level (which was drawn down to
as low as 1.4 m above the base of the shaft). The SID Minimum Rock Socket Length: 2.0 m 2.5 m
camera inspection of the socket base showed up to 10 mm ULS – Unfactored 50,000 kN 53,000 kN
to 20 mm of sediment in the west quadrant of the socket, ULS – Factored 27,000 kN 28,600 kN
and additional cleaning of the base was completed prior to
SLS – Unfactored 26,800 kN 28,060 kN
SQUID inspection. Based on the SQUID test results,
average thickness of sediments of less than 8 mm, at least SLS – Factored 21,700 kN 22,700 kN
50% of the base having less than 8 mm of sediment was
achieved, and the maximum thickness of sediment at any
place on the base did not exceeding 15 mm. These For 2.0 m and 2.5 m long rock sockets the results of the
measurements met the criteria specified for the production analysis indicated that at the early stages of loading, the
drilled shafts on this project. sidewall and the base share the load approximately
The CSL and TIP results were consistent with each equally; however, as the load increases beyond
other and demonstrated no major anomalies in the test approximately 20,000 kN, the sidewalls behave in a more
shaft, although minor anomalies were noted from top of plastic manner and the contribution of the base to the
concrete to a depth of 0.3 m interpreted by to be due to the capacity becomes the major component.
concrete not being fully cured at the time of CSL testing
which was considered acceptable for the O-cell test as the
test shaft will not be subject to top-down loading. A minor 7 COMPARISON OF ANALYSIS AND STATIC TEST
anomaly was also noted from a depth of 8.8 m to 9.4 m DESIGN
which is just above the top plate of the O-cell assembly
which the result of the CSL and TIP indicative a reduced
The factored geotechnical resistances obtained from the Bridge Foundation Report), and 30M12-513 (Bi-Directional
axial bi-directional static load test were slightly higher than Static Load Test – Credit River Twin Bridge).
those obtained from the static analyses completed in the
bid and execution stage design, which were in turn ACKNOWLEDGEMENTS
significantly higher than those used in the Reference
Concept Design for the QEW/Credit River Twin Bridge. All team members involved in the design, installation,
The resulting unfactored geotechnical resistances are oversight, and reporting of the BDSLT are acknowledged.
summarized in Table 2 and show that the design values Special thanks are extended to Infrastructure Ontario, MTO
from analysis and static load test methods produced similar Major Projects Office, lead designer AECOM, EDCO
results; the corresponding factored ULS and SLS values (EllisDon-Coco Group Joint Venture), LoadTest, and
are also summarized below, noting the increased Subsurface Geotech Inc.
resistance factors employed for BDSLT.
REFERENCES

Table 2. Summary of the Geotechnical Resistances ASTM. 2018. ASTM 8169-18 Standard Test Method for
Estimated from the Analysis and BDSLT Methods Deep Foundations Under Bi-Directional Static Axial
Compressive Loading
Maximum Rock Bieniawski, Z.T. 1976. Rock mass classification in rock
Analysis BDLST Analysis BDSLT engineering. Exploration for rock engineering, proc. of
Socket Length
the symp., (ed. Z.T. Bieniawski), Cape Town: Balkema
Unfactored ULS (kN) Factored ULS (kN) 1: 97-106.
2.0 m 48,610 50,000 17,500 27,000 Canadian Geotechnical Society (CGS). 2006. Canadian
2.5 m 52,780 53,000 19,000 28,600 Foundation Engineering Manual, fourth edition (4th Ed.)
Canadian Standards Association, 2019. Canadian
Unfactored SLS (kN) Factored SLS (kN) Highway Bridge Design Code (CHBDC) and
2.0 m 24,305 26,800 17,500 21,700 Commentary on CAN/CSA-S6-19. CSA Group.
FHWA. 2010. FHWA Publication No. FHWA-NHI-10-016
2.5 m 25,695 28,060 18,500 22,700
Drilled Shafts: Construction Procedures and LRFD
Design Methods
Golder Associates Ltd. 2021. Summary Report: Bi-
The unit factored ultimate axial resistances (per square directional static load test – Credit River Twin Bridge,
metre) for base and sidewall are summarized in Table 3. QEW/Credit River Improvement Project, Mississauga,
Ontario, Infrastructure Ontario RFP No. 19-110.
Report No. QEWCR-SUB-GEN-GEO-QEWCR-SUB-
Table 3. Factored Axial Geotechnical Resistances at ULS GEN-GEO-RPT-002-R1 O-Cell Test (19122889D)
Goodman, R. E. 1989. Introduction to Rock Mechanics
Calculation Method Base (kPa) Sidewall (kPa) (2nd Ed.). John Wiley & Sons.
Analysis 10,880 630 Hoek, E., Carranza-Torres, C.T., and Corkum, B., 2002.
Hoek-Brown failure criterion – 2002 Edition. 5th North
BDLST 10,695 650
American Rock Mechanics Symp., Toronto, Canada. 1:
267-273.
Horvath, R.G. and Kenney, T.C. 1979. Shaft Resistance of
8 CONCLUSIONS Rock Socketed Drilled Piers. Symposium on Deep
Foundations, ASCE, New York, N.Y. 1: 182–214.
The results of the BDSLT validated the use of higher International Society of Rock Mechanics (ISRM). 1985.
geotechnical resistances for shorter rock sockets in the Suggested method for determining point load strength.
Georgian Bay shale formation, and offers information for Elsevier BV.In Int. J. of Rock Mech. Min. Sci. &
the design of more efficient rock sockets on other sites and Geomech. Abstr. 22, (2): 51–60.
contracts in this shale formation. Marinos, P., and Hoek, E. 2000. GSI: a geologically friendly
SQUID testing allows assessment of the base cleaning tool for rock mass strength estimation. ISRM
procedures and confirmation that the target sediment international symposium. OnePetro.
thicknesses have been achieved, giving greater Pells, P.J.N., Rowe, R.K. and Turner, R.M. 1980. An
confidence to design with high load transfer to the drilled experimental investigation into side shear for socketed
shaft base. Quality assurance testing (CSL and/or TIP) piles in sandstone. Int. Conf. Structural Foundations on
allows for both confirmation of the integrity of the drilled Rock, Sydney, Balkema.
shafts as well as greater insight into interpretation of the Rowe, R.K. and H.H. Armitage. 1987. A Design Method for
O-cell test results if anomalies are observed in the drilled Drilled Piers in Soft Rock, Canadian Geotechnical
shaft, such as those seen in the test shaft. Journal. 24: 126–142.
Sowers, G. F. 1970. Introductory Soil Mechanics and
For additional information the foundation reports related to Foundations. Macmillan New York, pp. 395–6
the QEW/Credit River Twin Bridge piers and BDSLT are Wyllie, D. C. 1999. Foundations on rock: Engineering
available at https://foundation.mto.gov.on.ca/ under practice, second edition (2nd ed.). Spon Press.
GEOCRES numbers 30M12-500 (QEW/Credit River Twin
An Improved Approach for Frost Depth
Evaluation Considering Unfrozen Water in
Frozen Soil
Greg Qu and Nam Pham
WSP, Oakville, Ontario, Canada

ABSTRACT
The frost depth is known to depend on the unfrozen water content in frozen soil. However, current engineering practice of
frost depth calculation often ignores the unfrozen water and assumes that all water in the soil completely freezes below
0°C. Fine-grained soil can contain significant amounts of unfrozen water at freezing temperature, particularly for fine-
grained soil in the range of temperatures of practical importance for frost depth. The assumption of fully frozen water in
soil often leads to a less conservative underestimation of the frost depth by about 15% to 30% for fine-grained soils. There
is a need to develop a practical and industry-friendly approach for frost depth assessment, which takes account of the
unfrozen water content in frozen soil. The work by Tice et al. (1976), Anderson and Ladanyi (2003) recommended a
correlation between unfrozen water in soil and temperature with a good match with the test data. But this approach requires
a non-standard laboratory test (liquid limit test with N=100), which limits its application in engineering practice for
commercial projects. This paper proposed an improved approach, which simplifies the correlation and does not need any
non-standard tests for the input parameters. It requires only two basic input soil parameters, total initial water content and
liquid limit from routine Atterberg limit tests, both of which are usually available in most engineering projects. A sensitivity
study was carried out to evaluate the proposed approach using laboratory tests of about 20 soils from the literature. The
results shows that the proposed correlation matches well with the lab data and conventional approach and is considered
suitable for practical engineering assessment of frost depth.

RÉSUMÉ
La profondeur de gel est connue pour dépendre de la teneur en eau non gelée dans le sol gelé. Cependant, la pratique
technique actuelle du calcul de la profondeur du gel ignore souvent l'eau non gelée et suppose que toute l'eau du sol gèle
complètement en dessous de 0 °C. Le sol à grains fins peut contenir des quantités importantes d'eau non gelée à la
température de congélation, en particulier pour les sols à grains fins dans la plage de températures d'importance pratique
pour la profondeur de gel. L'hypothèse d'eau entièrement gelée dans le sol conduit souvent à une sous-estimation moins
prudente de la profondeur de gel d'environ 15% à 30% pour les sols à grains fins. Il est nécessaire de développer une
approche pratique et adaptée à l'industrie pour l'évaluation de la profondeur du gel, qui tienne compte de la teneur en eau
non gelée dans le sol gelé. Les travaux de Tice et al. (1976), Anderson et Ladanyi (2003) ont recommandé une corrélation
entre l'eau non gelée dans le sol et la température avec une bonne correspondance avec les données d'essai. Mais cette
approche nécessite un test de laboratoire non standard (test de limite de liquidité avec N = 100), ce qui limite son
application dans la pratique de l'ingénierie pour les projets commerciaux. Cet article propose une approche améliorée, qui
simplifie la corrélation et ne nécessite aucun test non standard pour les paramètres d'entrée. Il ne nécessite que deux
paramètres de sol d'entrée de base, la teneur en eau initiale totale et la limite de liquidité des tests de limite d'Atterberg de
routine, qui sont généralement disponibles dans la plupart des projets d'ingénierie. Une étude de sensibilité a été réalisée
pour évaluer l'approche proposée à l'aide d'essais en laboratoire d'environ 20 sols de la littérature. Les résultats montrent
que la corrélation proposée correspond bien aux données de laboratoire et à l'approche conventionnelle et est considérée
comme appropriée pour l'évaluation technique pratique de la profondeur du gel.

1 INTRODUCITON soil freezes over a wide range of temperatures from -20°C

to 0°C. Frost action in the fine-grained soil can cause
The current engineering practice to calculate frost depth significant damage to the structure due to the form of ice
often ignores the unfrozen water and assumes that all lenses and consequent frost jacking.
water in the soil completely freezes below 0°C (see There is a need to develop a practical and industry-
Canadian Foundation Engineering Manual, 2006 and the friendly approach for frost depth assessment, which takes
guidelines by U.S. Army, 2012). The volumetric latent heat account of the unfrozen water content in frozen soil, using
of fusion, as a key parameter for the frost depth (see the the routine standard laboratory tests available for most
modified Berggren equation in CFEM 2006) is governed by engineering projects. The currently available approaches
the content of the water in soil, which can be frozen into (Anderson and Tice 1972, Anderson and Morgenstern
ice. This assumption leads to underestimating the frost 1973, Tice et al. 1976, Anderson and Ladanyi 2003 and Hu
depth, particularly for fine-grained soils, where the water in et al. 2020) require inputs from non-standard laboratory
tests (i.e., the specific surface areas of soil particles, liquid 𝛼 = 𝑤𝑢1 represents the unfrozen water content in
limit test with N=100, or nuclear magnetic resonance tests), frozen soil at 𝑇 = −1℃.
which limits their application in practice. 𝛽 is a material parameter.
This paper presents an improved and practical Anderson and Tice (1972) proposed a correlation
approach to take into account the impact of unfrozen water between the soil specific-area and the parameters of 𝑤𝑢1
in frost depth evaluation. This approach requires two input and 𝛽. However, the lab test to determine specific area of
parameters, initial water content and liquid limit of soil, both soil is complicated and is often not available for most
of which are available from standard laboratory tests in engineering projects.
most engineering projects. This approach intends to With further studies, the research team (Tice, Anderson
support a preliminary assessment of frost depth in and Banin 1976) later proposed a correlation using a liquid
engineering practice. For critical projects, design engineers limit test and a non-standard liquid limit test to obtain the
should consider a field test (thermistor) or advanced water content corresponding N=100 (where N is the
laboratory tests, for example, nuclear magnetic resonance number of blows required to close the standard groove in
(NMR) test. the liquid limit test). Although the correlation worked very
well to match the lab test data, this correlation has not been
2 BACKGROUND widely used in practice, likely due to the following two
factors:
The theoretical methodology for frost depth has been well 1. It requires the input parameters of 𝐿𝐿𝑁=25 (Liquid
established and the engineering practice usually adopts Limit) and 𝐿𝐿𝑁=100 from non-standard lab tests
the modified Berggren equation, as recommended by (i.e., non-standard liquid limit test of N=100, N is
CFEM (2006) and U.S. Army (2012). the number of blows required to close the
standard groove in the liquid-limit test). The non-
standard lab tests are usually not available in
𝑋 = 𝜆√2𝑘𝑓 𝐼𝑠 ⁄𝐿𝑠
commercial engineering projects to support the
where assessment of frost depth using this correlation.
𝑋 = frost depth 2. The 𝑤𝑢 approaches to an infinitely high value, as
𝐼𝑠 = surface freezing index the temperature approaches 0℃. This may cause
𝑘𝑓 = thermal conductivity of the frozen soil confusion and numerical issues in the analyses.
𝐿𝑠 = volumetric latent heat of the soil There are some recent research studies to assess
𝜆 = a dimensionless coefficient unfrozen water content in frozen soil, as summarized by Hu
One key input for the frost depth calculation is the et al. (2020) and Anderson and Ladanyi (2003). However,
volumetric latent heat of fusion (𝐿𝑠 ) for soil, which is the these studies focused more on accurate estimation of
amount of energy required to freeze the unfrozen water in unfrozen water content, which inevitably introduces more
soil to the frozen state. Therefore, the unfrozen water in fitting parameters or required further advanced lab tests.
frozen soil has a significant impact to the latent heat and The objective of this paper is to develop a practical
consequently the frost depth. As shown in the formula correlation of unfrozen water content for engineers to
below, the input of 𝑤𝑐 is usually assumed to be the total calculate the frost depth using only standard lab tests
water content (CFEM 2006), ignoring the unfrozen water in (Atterberg limit test).
frozen soil.
𝐿𝑠 = 𝛾𝑑 𝑤𝑐 𝐿
where 40%
𝛾𝑑 = dry unit weight of soil Correlation proposed in this paper
Unfrozen Water Content, %

35%
𝑤𝑐 = water content of soil which is assumed to be Correlation by Tice et al. 1976
completely frozen into ice 30% Reference Points
𝐿= latent heat of fusion of water to ice, 334 kJ/kg 25%
20%
The studies have been carried out to quantify the
unfrozen water content in frozen soil by researchers 15%
(Anderson and Tice 1972, Anderson and Morgenstern 10%
1973, Aderson et al. 1973, Tice et al. 1976, and Anderson 5%
and Ladanyi 2003).
The formula below was proposed by these researchers 0%
0

-2

-5
-1

-3

-4

to represent the correlation between unfrozen water

content with temperature, as shown in Figure 1a. One Temperature, degree C
limitation of this correlation is that the 𝑤𝑢 approach to an Figure 1a. Comparison of Correlation of 𝑤𝑢 versus T
infinitely high value which may even exceed the total water proposed in this paper and that by Tice et al. 1976.
content as the temperature approaches to 0℃. (𝑤𝑐 =22%, LL=40%, 𝛼= 𝑤𝑢1 =10.6%, 𝛽= - 0.25)

𝑤𝑢 (𝑇) = 𝛼 × (−𝑇)𝛽
where
𝑇 is the temperature in Celsius degree (℃).
 50% considered to have negligible impact on the frost
Lab Test Data (Tice et al. 1976) depth calculation, see the assessment in Section
Unfrozen Water Content at -1 C deg,
Correlation from Eq. [2] 4 for details.
40%
3. The proposed approach agrees well with the
Linear (Lab Test Data (Tice et al. original one proposed by Tice et al. (1976)
30% 1976)) regarding the correlation between liquid limit and
the unfrozen water content in frozen soil at 𝑇 =
Wu1

−1℃. Figure 1b shows a comparison with the test

20% data of eight soils indicates, indicating both
correlations match well with the data. Table 1a
10% listed the parameters for the eight soils in the
assessment.

0% Correlation using Test Data in Literature

20%

40%

60%

80%
0%

100%
Liquid Limit (LL), % Table 3 summarizes the calibrated 𝑤𝑢1 and 𝛽 for about 20
different soils from the literature data. Figure 3a shows the
Figure 1b. Comparison of Correlations of Liquid Limit and calibrated fitting curves using Eq. [1] match reasonably well
𝑤𝑢1 proposed in this paper and that by Tice et al. 1976. with the test data from Anderson and Morgenstern (1973).
In Figure 3b, the unfrozen water content is normalized with
the value at -1°C. It shows the test data falls in a relatively
3 PROPOSED CORRELATION FOR UNFROZEN
narrow range. The values of 𝛽 for these four soils vary from
WATER CONTENT IN FROZEN SOIL
– 0.24 to - 0.6 correspondingly.
The following presents the correlation formula proposed to
4 PARAMETER OF 𝛽
estimate the unfrozen water content in frozen soil.
For the 20 soils listed in Table 1b, the soil parameter 𝛽
𝑤𝑐 − (𝑤𝑐 − 𝑤𝑢1 ) × (−𝑇) 𝑓𝑜𝑟 𝑇 ∈ (0, −1℃) [1𝑎]
𝑤𝑢 (𝑇) = { varies typically from - 0.15 to - 0.40, with an average of
𝑤𝑢1 × (−𝑇)𝛽 𝑓𝑜𝑟 𝑇 < −1℃ [1𝑏] about - 0.25. Figure 2 shows the upper and lower bounds
of the curve with 𝛽= - 0.15 and - 0.4, respectively.
[1] Given the narrow range of the soil parameter 𝛽, this
paper proposed to use the default value of 𝛽 = -0.25.
𝑤𝑢1 = 𝑤𝑢,𝑇=−1℃ = 0.35 × 𝐿𝐿 − 3% [2] Section 6 presents the results of sensitivity analyses of 𝛽
for the impact to frost depth within the range from -0.15 to
where - 0.4. It is shown that the influence of 𝛽 is insignificant for
𝑤𝑢1 is the unfrozen water content in frozen soil at 𝑇 = this range, less than 5% of frost depth.
−1℃, which can be estimated using the equation
[2] for preliminary evaluation or specific tests.
𝛽 is a material parameter with a default value of -0.25.
Alternatively, this parameter can be obtained by
fitting the data from laboratory tests. Section 4 30%
provides further discussion for this parameter.
Unfrozen Water Content, Wu

Eq. for T > -1°C

25%
Eq. for T < -1°C
Figure 2 illustrates the correlation from Eq. [1]. In 𝑤𝑐
general, the unfrozen water content (𝑤𝑢 ) decreases with 20% Reference Points
the lower temperature. For temperature from 0 to -1°C, 𝑤𝑢 Upper Bound
15%
reduces linearly from the total water content, 𝑤𝑐 to 𝑤𝑢1 , Lower Bound
see Eq. [1a]. It is noted that literature data suggests a 10%
𝑤𝑢1
correlation between 𝛼 and the limit liquid of soil. Below -
1°C, the 𝑤𝑢 reduction is in a linear log-log correlation which 5%
is governed by the slope of the log-log plot, 𝛽, see Eq. [1b].
0%
The following describes the main features of the
-10

-15

-20
0

-5

proposed approach. Temperature, degree C

1. For the temperature range from 0 to −1℃, the
Figure 2. Illustration of Correlation (Equation [1]) Between
proposed approach adopts a linear correlation of
Unfrozen Water Content in Frozen Soil and Temperature.
unfrozen water content from the total water
content (𝑤𝑐 ) to 𝑤𝑢1 , as shown in Figure 1a. Eq.
(Notes: 𝑤𝑐 =22%, LL=40%, 𝛼 = 𝑤𝑢1 =10.6%, 𝛽 = -0.25 for
[1a] presents the mathematic formula.
base-case, 𝛽= - 0.4 for lower bound and 𝛽= -0.15 for upper
2. This approach uses a default value of 𝛽 and
bound)
eliminates the requirement for non-standard liquid
limit tests, see Eq. [1b]. This simplification was
 100
90
0.0
Assuming fully frozen water
Unfrozen Water Content,%

80
70 β= -0.25 (Default)
60 β= -0.15
0.5
50
40 β= -0.4
30
20 1.0

Frost Depth, m
10
0
0

10
Temperature, -T 1.5
No.1 Umiat Bentonite No.2 Hawaiian Clay
No. 3 Suffield Silty Clay No.4 Dow Field Silty Clay

Figure 3a. Correlation for the test data from Anderson and 2.0
Morgenstern (1973).

2.5

2.0 Site 1 Site 2 Site 3

Unfrozen Water Content, Normalized

1.8
3.0
1.6
1.4
1.2
Figure 4a. Sensitivity of Frost Depth at Three Sites for Four
1.0
Cases: (1) Assuming fully frozen water (ignoring Wu), (2)
0.8
Considering unfrozen water in frozen soil using Eq. [1] with
0.6
β = - 0.25, (3) Eq.[1] with β = - 0.4, (4) Eq.[1] with β = - 0.15.
0.4
0.2
0.0
3
0

10

100%
Frost Depth Ratio of F.f / F.u

Temperature, -T
No.1 Umiat Bentonite No.2 Hawaiian Clay
No. 3 Suffield Silty Clay No.4 Dow Field Silty Clay 90%

80%
Figure 3b. Correlation for the test data from Anderson and
Morgenstern (1973).
70%
Site 1 Site 3
60%
10% 20% 30% 40%
Initial Water Content, Wc
Figure 4b. Sensitivity of Frost Depth with Initial Water
Content.

(Frost Depth Ratio: Frost Depth for the Case Ignoring Wu

divided by that for the Case Considering unfrozen water in
frozen soil using Eq. [1] with β = - 0.25)
5 INCOPERATION IN FROST DEPTH CALCUATION The following summarizes the main findings:
1. The assumption of fully frozen water in frozen soil
This section presents the proposed procedure in the frost (i.e., ignoring the unfrozen water, 𝑤𝑢 ) would lead
depth calculation (i.e., modified Berggren approach) to to an unconservative estimate of frost depth, by
consider the unfrozen soil water content. about 15% to 30%, as shown in Figure 4b. The
The average unfrozen water content within the frost underestimation of frost depth appears more
depth can be estimated using the equation below: significant for the soil with lower initial water
contents, as shown in Figure 4b.
0
2. The frost depth is not sensitive for the range of 𝛽
𝑤𝑢_𝑎𝑣𝑔 = ∫ 𝑤𝑢 (𝑇)𝑑𝑇 from -0.15 to -0.4, which yields a difference of less
𝑇𝑠
than 5% (about 2% to 4%) in comparison with the
where 𝑇𝑠 represents the equivalent surface temperature in base-case value (𝛽 = - 0.25), see Table 3. The
℃. variation of frost depth within the range of 𝛽 is
The ice content in frozen soil can be obtained using the less than 4%. As such, using the default value of
formula below to account for the unfrozen water content in 𝛽 (-0.25) is considered acceptable for most
frozen soil. engineering evaluations.

𝑤𝑖 = 𝑤𝑐 − 𝑤𝑢_𝑎𝑣𝑔 7 SUMMARY AND CONCLUSIONS

This paper presented a practical approach for engineers to

The volumetric latent heat of the soil can be estimated
calculate the frost depth considering the unfrozen water in
from the equation below:
frozen soil. In comparison with the conventional approach
which assumes all water is frozen in soil, this approach
𝐿𝑠 = 𝛾𝑑 × 𝑤𝑖 × 𝐿
provides a more conservative frost depth (typically about
15% to 30%) for the fine-grain soil. The frost depth
where 𝛾𝑑 is dry soil unit weight. L is the latent heat of fusion calculation approach was coded in Mathcad and Excel
of water to ice, with the typical value of 334 kJ/kg. formats and is available for download at the website listed
The impact of unfrozen water is negligible for the in the reference (Qu, 2023). The following summarizes the
volumetric heat capacity and thermal conductivity (Nixon main findings.
and McRoberts 1973). The following provides the
procedure to take these into account for reference only. 1. The proposed approach is practical as it requires only
The volumetric heat capacity for frozen soil can be two basic input parameters, i.e., total initial water
estimated using the relationship to account for 𝑤𝑢 : content and liquid limit, which can be obtained from
𝐶𝑓 = 𝛾𝑑 /𝑟𝑤 (0.17 + (𝑤𝑐 − 𝑤𝑢 ) + 0.5 × 𝑤𝑢 ) × 4.187 × 106 𝐽/ routine standard tests and are usually available in
(𝑚3 ℃) most projects.
The thermal conductivity for frozen soil can be obtained 2. The proposed approach would lead to a more
by using 𝑤𝑖 instead of 𝑤𝑐 for Kersten approach or other conservative frost depth (typically about 15% to 30%)
acceptable equivalent, as shown below for the fine-grained for fine-grained soil, in comparison to the
soil type. conventional approach ignoring the unfrozen water in
frozen soil. For coarse-grained soil, the proposed
6 SENSITIVITY ANALYSES approach yields a similar frost depth estimate with the
conventional approach as the unfrozen water content
The sensitivity analyses were conducted to investigation in frozen coarse-grained soil is relatively low.
the following two questions: 3. The following summarizes the limitations of the
1. What is the impact of unfrozen water content in proposed approach.
frozen soil to the calculated frost depth. o If salts are present in soil, a correction should be
2. How sensitive is the value of 𝛽 to the calculated applied to the proposed approach for frost depth
frost depth? The three cases were evaluated with calculation, depending on the salinity
β = - 0.25 (base case, default value), -0.15 (upper measurement from the site.
bound) and -0.4 (lower bound). o Caution should be applied for soils with liquid
The frost depth calculation was conducted at three sites limits over 100, as the correlation from the
with different climate conditions. Site 1 is in south Ontario. proposed approach is based on the soil test data
Site 2 is in Alberta. Site 3 is in Northern Ontario. Table 2 with liquid limits less than 100.
presents the climate conditions for the three sites. Table 3 4. Future studies using laboratory or field tests will be
presents the calculated frost depths for the approaches helpful to verify the findings presented in the paper
accounting for the unfrozen water in frozen soil at the three regarding the impact of unfrozen water in frozen soil
sites with different climate conditions. Figure 4a shows the on frost depth.
results.
Figure 4b presents the impact of initial water content to
the frost depths for the approaches accounting for the
unfrozen water in frozen soil the at the three sites with
different climate conditions.
Table 1a. Soil properties with LL and 𝑤𝑢1 from literature Table 2. Summary of Site Conditions in Sensitivity Study

Soil 𝐿𝐿, % 𝑤𝑢1 ,% Notes Parameters Site 1 Site 2 Site 3

West Lebanon gravel Low LL value 𝐼𝑎 ( C -days)
o
467 855 1558
17.2 2.1
(<149 um) nf 0.9 0.9 0.9
Basalt (powdered) 17.4 3.5 𝑤𝑐 (%) 22% 22% 22%
Fairbanks Silt 26.7 4.8
LL(%) 40% 40% 40%
Limonite 40 8.0
𝛾𝑠 (kN/m ) 3
18 18 18
Dow Field Silty Clay 34.3 11.0
MAAT (oC) 8.5 6.4 2.1
Rust 48.7 14.0
t (days) 81 120 157
Suffield silty clay 40.1 14.0
Soil Type (F/C)1 F F F
Kaolinite 78.9 22.0
Note: 1. “F” represents Fine-grained soil and “C”
Hawailian clay 98.6 32.4 High LL value represents Coarse-grained soil.

Notes:
1. LL represents for Liquid Limit.
2. Data inferred from Tice et al. 1976.

Table 1b. Summary of soil properties from literature Table 3. Sensitivity of Frost Depths to 𝛽

Soil 𝛽 𝑤𝑢1 Reference Assumptions Site 1 Site 2 Site 3

Anderson et al Ignoring 𝑤𝑢 0.94 1.34 1.95
Fairbanks silt -0.33 4.8
(1973) (80%) (80%) (80%)
Hawaiian clay -0.24 32.4 Same as above
Umiat bentonite -0.34 67.6 Same as above 𝛽 = - 0.25 1.18 1.68 2.44
Wyoming bentonite -0.29 56.0 Same as above (Default,base-case) (100%) (100%) (100%)
Oliphant et al.
Morin clay -0.51 13.1
(1983) 𝛽 = - 0.4 (LB) 1.15 1.63 2.34
Caen silt -0.23 9.5 Smith (1984) (97%) (97%) (96%)
Calgary silt -0.36 9.6 Patterson and 𝛽 = - 0.15 (UB) 1.21 1.72 2.52
Allendale clay -0.19 15.7 Smith (1981) (102%) (102%) (103%)
Inuvik clay -0.25 14.5 Smith (1985)
Tomokomai clay -0.31 19.5 Kay et al. (1981)
Sandy Loam* -0.20 4.3 Tsytovich (1975)
Loam* -0.18 8.5 Same as above
Clay* -0.20 13.5 Same as above
Clay with Same as above
-0.19 23
montmorillonite
Kinelian Clay* -0.25 26.0 Tsytovich (1960)
Pleistocene Clay* -0.18 14.0 Same as above
Pleistocene Clayey Silt Same as above
-0.18 9.5
with Some Sand*
Silty Sand with Some Same as above
-0.15 5.0
Clay*
Suffield Silty Clay* -0.4 14 Anderson and
Morgenstern
Dow Field Silty Clay* -0.6 11 (1973)

Notes:
1. The data with “*” represents the interpretation of soil
parameters made in this paper.
2. The data without “*” indicate the interpretation of soil
parameters made by the authors of the paper or
summarized by Anderson and Ladanyi (2002).
 Patterson, D. E., and M. W. Smith. 1981. The
8 LIST OF PARAMETERS measurement of unfrozen water content by time-
domain reflectometry: Results from laboratory tests.
N: the number of blows required to close the Can. Geotech. J. 18(1): 131–44.
standard groove in the liquid-limit test.
𝑤𝑐 : total water content in frozen soil, including both Hu, G., Zhao, L., Zhu, X., Wu, X., Wu, T., Li, R., & Hao, J.
unfrozen and frozen water. (2020). Review of algorithms and parameterizations to
𝑤𝑢 : unfrozen water content in frozen soil. determine unfrozen water content in frozen
𝑤𝑢_𝑎𝑣𝑔 : average unfrozen water content in frozen soil. soil. Geoderma, 368, 114277.
𝑤𝑖 : ice content in frozen soil.
𝛾𝑑 : dry soil unit weight. Nixon, J. F., & McRoberts, E. C. (1973). A study of some
𝛾𝑠 : bulk soil unit weight. factors affecting the thawing of frozen soils. Canadian
𝛾𝑤 : unit weight of water. Geotechnical Journal, 10(3), 439-452.
𝐼𝑎 : air freezing index
𝐼𝑠 : surface freezing index. US Army Corps, 2012, Pavement-transportation Computer
L: latent heat of fusion of water to ice, typical value Assisted Structural Engineering (PCASE)
of 334 kJ/kg. Implementation of the Modified Berggren (Modberg)
𝐿𝑠 : volumetric latent heat of the soil. Equation for Computing the Frost Penetration Depth
𝑘𝑓 thermal conductivity of the frozen soil. within Pavement Structures.
𝑤𝑢1 : unfrozen water content in frozen soil at 𝑇 = −1℃.
Tsytovich, N.A., 1975. The Mechanics of Frozen Ground.
It can be estimated using the equation [2].
Edited by George Swinzow, U.S. Army Cold Region
𝛽: soil parameter for correlation between unfrozen
Research and Engineering Laboratory.
water content in frozen soil and temperature,
depending on the soil types, see Table 1b.
Tsytovich, N.A., 1960. Bases and Foundations on Frozen
LL: Liquid limit of soil.
Soil. National Academy of Sciences, National Research
𝑇: Temperature in Celsius, degree.
Council.
𝜃: Absolute value of temperature in Celsius, degree
Tice, A. R., D. M. Anderson, and A. Banin. 1976. The
(always a positive value).
Prediction of Unfrozen Water Contents in Frozen Soils
from Liquid Limit Determinations.U.S.Army Cold
Regions Research and Engineering Laboratory Report
CRREL 76-8.
REFERENCES
Qu, G, 2023, MathCAD Calculation Sheet for Frost Depth
Andersland, O.B. and Ladanyi, B., 2003. Frozen ground
Calculation - Considering Unfrozen Water Content in
engineering. John Wiley & Sons.
Frozen Soil, website address:
https://www.researchgate.net/publication/370604091_Fro
Anderson, D. M., and N. R. Morgenstern. 1973. Physics,
st_Depth_Calculation_-
chemistry, and mechanics of frozen ground: A review.
_Considering_Unfrozen_Water_Content_in_Frozen_Soil
In North Am. Contrib.2nd Int. Conf. on Permafrost,
Yakutsk, USSR. Washington, D.C.: National Academy
of Sciences, pp. 257–88.

Anderson, D.M. Tice, A.R. and McKim, H.I. 1973,The

unfrozen water and the apparent specific heat capacity
of frozen soils. In North Am. Contrib. , 2nd Int. Conf. on
Permafrost , Yakutsk, USSR. National Academy of
Sciences, pp289-95.

Anderson, D.M. and Tice, A.R. 1972, Predicting unfrozen

water contents in frozen soils from surface area
measurement. In Frost Action in Soil. National
Academy of Sciences, pp12-18.

Canadian Geotechnical Society, 2006, Canadian

Foundation Engineering Manual (CFEM), the 4th
edition.

Technical Manual by U.S. Department of the Army and the

Air Force. (1983). TM-5-852-4: Arctic and subarctic
constructions for structures. Department of the Army
and the Air Force
Settlement Control of The Embankment
Along Commissioners Street Using Ground
Improvement
Mark Tigchelaar, P.Eng., Yared Sholla, P.Eng., PMP, Wayne
Sunghyun Park, EIT & Omar Fayez, EIT
GeoSolv Design-Build Inc., Woodbridge, Ontario, Canada
Allen Bowers, PhD, PE.
Geopier, a division of CMC, Davidson, North Carolina, USA

ABSTRACT
The Port Lands Flood Protection and Enabling Infrastructure (PLFPEI) Project is a $1.25 billion program that will revitalize
800-acres of flood prone land. In connection with the Port Lands Flood Protection Enabling Infrastructure project, a grade
raise is proposed along Commissioners Street and Don Roadway to protect local neighbours from floodwaters from the
Don River in the event of extreme weather. A new fill of up to about 4m in height will be placed to raise the existing grade
within the right of way of the roadways. Due to the extensive thickness of compressible peat and organic soils, significant
settlement of the foundation soils is anticipated. Following a comprehensive review of different design alternatives to control
the post-construction settlement of the foundation soils supporting the embankment to an acceptable limit, ground
improvement utilizing a rigid inclusion method was chosen as the preferred option. This paper presents the results of the
foundation investigation program, design methodology, performance specifications, load transfer mechanism, and the
results of settlement monitoring of the embankment during and following construction.

RÉSUMÉ
Le projet de protection contre les inondations « Port Lands Flood Protection and Enabling Inrastructure » est un
programme de 1,25 milliard de dollars qui revitalisera 800 acres de terres sujettes aux inondations. Dans le cadre du
projet, une surélévation du niveau de terrain est proposée le long de la rue Commissioners et de la Don Roadway afin de
protéger les habitants locaux contre les eaux de crue de la rivière Don en cas de conditions météorologiques extrêmes.
Un remblai d'une hauteur pouvant atteindre environ 4 mètres sera utilisé pour rehausser le niveau de terrain existant dans
l'emprise des routes. Étant donné l'épaisseur considérable des tourbes compressibles et des sols organiques, un
tassement significatif des sols de fondation est prévu. Suite à un examen approfondi des différentes alternatives de
conception pour contrôler le tassement post-construction des sols de fondation soutenant le remblai dans une limite
acceptable, la méthode d'amélioration du sol utilisant l’inclusion rigide a été choisie comme option privilégiée. Cet article
présente les résultats du programme d'étude des fondations, la méthodologie de conception, les spécifications de
performance, le mécanisme de transfert de charge et les résultats de surveillance des tassements du remblai pendant et
après la construction.

1 INTRODUCTION subsequent sections are not able to support without

significant settlements over time. The approach
The Port Lands Flood Protection and Enabling embankments to the bridges, which are supported on piles,
Infrastructure (PLFPEI) Project introduces a newly installed must adhere to strict long-term total and differential
naturalized river extension to the Don River through old settlement criteria to ensure proper performance of the
hydraulic fill zones that were both structurally poor and system.
environmentally impacted. This lined riverbed creates a Various approaches were considered including pile-
much-needed relief for large rain events and its supported platforms, ground improvement using preload
development is one of the most ambitious and revitalizing and surcharging, and ground improvement using rigid
projects to come to Toronto in decades. The new river cuts inclusions.
through or is immediately adjacent to current at-grade road Based on soil conditions and loading, which is a
routes including Commissioners Street and the Don function of embankment height, some sections of the
Roadway. These roadways must be raised to meet the roadway were improved using preloading/surcharging and
bridge elevations to allow the routes to cross the higher other sections directly abutting the bridge were improved
river level while being able to maintain the necessary river with rigid inclusion ground improvement techniques.
traffic proposed. In addition, a flood protection landform, This paper presents the subsurface conditions,
coincident with the Don Roadway is to be constructed to settlement criteria for the project, settlement analysis, finite
ensure that future flooding will not inundate the Port Lands element modelling, and settlement monitoring results. This
development area. paper considers the calculated and measured settlements
As a result of the necessary overpasses, up to 4m of from zones of roadway that were preloaded and the
new grade is to be placed for the new roadway surface calculated and measured settlements of zones of roadway
creating an embankment. This grade will result in new load, embankment that were supported on ground improvement.
which the existing poor hydraulic fills described in The detailed design analysis and results obtained using
rigid inclusions will be discussed in more detail and the the commissioners bridge (station 0+320 to 0+500) contain
preloaded sections will only be discussed as a means of pockets and layers of organic silt at varying depths within
adjacent comparison for both calculations and actual the sand layer. The depth of bedrock varies along
measurements in the field. These two different methods of Commissioner Street, deeper at the west ends and
ground improvement used for different sections of the new becomes shallower on the east side.
embankment to control settlement will then be compared. The various pockets and layers of organic silt were
encountered between depths of 14m and 14.5m, 20m and
1.1. Subsurface Conditions 21m and 26m and 33m. Where these deep pockets of
organic silts present, the depth of bedrock varies from
The soil profile along Commissioners Street generally about el. 33m to 38 m. However, east of station 0+560, the
consists of up to 3m thick fill consisting of sand and gravel, bedrock is located at depth 12 to 15m (elevation 64 to61m).
silty sand, and sandy silt. The fill materials overlay a Groundwater was encountered about 1.5 m below
discontinuous layer of organic rich alluvial soils (peat or existing grade (about el. 75 m). The soil stratigraphy and
organic silts) ranging in thickness from 2m up to about 8m, parameters as summarized by the geotechnical engineer
which in turn overlies loose to compact sands of up to 2m of record is presented in Table 1. A generalized soil profile
thickness, becoming dense to very dense at depth. The complete with stationing is contained in Figure 1.
dense sand is followed by bedrock for the majority of the
road alignment (station 0+560 to 0+820). Sections west of

Table 1. Soil Stratigraphy and parameters

Soil Layers Avg Unit Initial E OCR POP 𝐶𝑐 𝐶𝑟 𝐶𝑣 k, 𝐶𝛼 /𝐶𝑐

Moisture, Weight γ Void MPa 𝑘𝑃𝐴 𝑚2 /𝑦𝑟 𝑚/𝑑
𝑤𝑎 𝑘𝑁/𝑚3 Ratio,
𝑒0
Fill 1 - 20.0 - 64 - - - - - - -
Fill 2 - 19.0 - 18-34 - - - - - - -
Organic Silt 1 38 16.6 1.0 13-16 1.5 28 0.46 0.09 72 1.6E-03 0.04
Organic Silt 2 51 15.4 1.3 7 1.5 25 0.61 0.12 75 1.8E-03 0.05
Peat 166 11.5 4.2 4 1.0 12 1.99 0.40 100 3.1E-03 0.06
Loose Sand - 19.0 - 15-48 - - - - - - -
Compact Sand - 20.0 - 64-91 - - - - - - -

Figure 1. Generalized Soil Profile for Commissioners street (East of STA 0+420)
 1.2. Settlement Criteria engineer recommended that ground improvement
measures be taken to meet the project settlement
Post-construction settlement criteria at 100 years is performance criteria. Ultimately the owner selected
presented in Table 2. ground improvement for settlement mitigation beyond
Station 0+330m based on technical and financial
Table 2. Post-Construction Settlement Criteria evaluations. The technical aspect will now be presented.

Maximum Total Settlement Limits 1.4. Rationale for Rigid Inclusion Ground Improvement
Distance from Bridge 0-20 20-50 >50
Abutment (m) Due to the organic and very loose nature of the
Total Settlement (mm) 25 50 100 upper hydraulic fill soils, ground improvement using
Differential Settlement (mm/m) 1.5 2.0 2.5 stone (such as Rammed Aggregate Piers or Stone
Columns) was not deemed an efficient method to control
settlements adequately. The relatively high loads and
the potential for further future degradation of the
1.3. Settlement Estimate without Ground Improvement organics in the upper zone soils would lead to bulging of
the aggregate piers and long-term settlement.
Based on the soil parameters summarised in Table 1, Thus, the geotechnical designer selected rigid
the project geotechnical engineer calculated settlements inclusion ground improvement approaches, proposing a
for an unimproved condition, the results of which are design that penetrated all organic zones in the profile.
summarized in Table 3. The settlement estimates show The rigid inclusion was chosen over the conventional
settlement magnitudes that do not meet the performance deep foundations, such as driven or drilled piles, for its
criteria of Table 2 and demonstrate that ground economic value, as it does not require pile caps for load
improvement measures are required. The first transfer and the production rates are typically faster. The
alternative considered to mitigate the settlement was to design-build tender for ground improvement required the
surcharge the site and monitor settlement for up to six design-build contractor to complete a detailed design to
months. control long term settlements to the criteria summarized
Analysis using the surcharge method showed in Table 2.
acceptable results for the section from 0+000m to The Geopier® GeoConcrete® Column (GCC) System
0+330m, where the embankment height is 1.6m or less. was ultimately chosen to support the embankments
As seen from estimates in Table 3, surcharge and immediately adjacent to the bridge structure, where the
preload could not achieve the settlement requirement grade raise fill for the embankment was the highest (up
beyond chainage 0+330m. Therefore, the geotechnical to 4m).

Table 3. Settlement Estimate Without Ground Improvement

Long-Term Post Construction Settlement at the Roadway Grade (100 year)

Station 0+080 0+270 0+410 0+460 0+580 0+600
Borehole Utilized BHG18-14 BHG18-16 BHG18-99 BHG18-98 BHG18-80 BHG18-80
Settlement Criteria 50 50 50 50 50 20
Unimproved 125 100 800 1,200 1,000 1,000
Post-construction
Settlement (mm)

Preload & Surcharge 50 40 180 250 250 250

Settlement
Surcharge/Duration 1.5m/6mths 1.5m/4mths 1.5m/4mths 1.5m/4mths 1.5m/4mths 1.5m/4mths
Rigid Inclusions NA NA 80 50 50 20
Settlement

using a strong static force augmented by dynamic

2.0 GEOPIER® GEOCONCRETE® COLUMN (GCC) vertical impact energy. The GCC elements for this
DESIGN site were typically installed to depths ranging
between 8 to 12 m, depending on the location of the
The design approach of a GCC element is to transfer the medium-dense to dense sand. The proprietary
load from the embankment, roadway surface, and/or mandrel includes an internal flow restrictor to
utilities through the upper poor soils, to a stiffer and/or prevent soil from entering the tamper foot and
denser layer below. This stiffer/denser bearing layer is mandrel during driving. The process fully displaces
further improved through the creation of the GCC soils laterally, and the flow restrictor and vertical
element. The installation process is as follows and is impact energy and crowd force results in
further described in Figure 2: densification and reinforcement of the soils below
• A specially designed proprietary closed-system
mandrel and tamper foot is driven into the ground
 the tip of the mandrel and along the sides of the monitoring of the pressure inside the mandrel, to
element where feasible. ensure that the correct concrete volume is extruded
• As the apparatus is driven to the design depth, at the right elevation to form a high-quality element.
concrete is pumped through the closed-system
mandrel into the driving head. While pumping As previously described, the vertical ramming,
concrete under pressure, the mandrel is raised proprietary internal flow restrictor and crowd force with
thereby extruding concrete, and then driven back repeated re-driving create a large bulb on an improved
down the full design depth. soil base. The embankment load is transferred through
the concrete shaft to the enlarged bottom bulb, which is
• The proprietary driving head restrictor engages the
in turn supported by a competent bearing layer, whether
extruded concrete, forcing the concrete downward it be sand or bedrock. Because additional layers and
and outward to form a bottom bulb. The bulb is pockets of organic soil are present beneath this bearing
enlarged by using a repeated up-down driving cycle layer, a careful evaluation of the lower zone
until the bottom bulb meets the design compressibility and the ability of the improved lower
requirements. This process results in the creation dense sand layer to redistribute loads was performed.
of an enlarged pier base, improvement at the base, The GCC system was designed to adequately support
and a bearing area on competent/improved soil that the embankment loads while avoiding having to
affords significant load transfer to the base penetrate to depths of up to 25m or more. The capacity
materials. The mandrel is then raised while of the GCC elements was checked by performing
several full-scale modulus test on an installed sacrificial
continuously pumping concrete under pressure to
GCC element. Each element is designed to carry a
form a column shaft to the top of pier elevation. design load of 540kN (120kips), confirmed through
• The entire process utilizes a proprietary quality testing to 200% using testing as per ASTM D1143.
control method, which involves continuous

Figure 2. Illustration of GCC installation process

3.0 GCC LOAD TRANSFER MECHANISM The majority of the settlements were estimated to be
substantially completed during the 90 day wait period
Because of their often very high stiffness relative to that due to the elastic settlement within the sands. This
of the soil, GCC elements attract high stresses that must settlement was confirmed with the monitoring program.
be transferred from the supported structure to the GCC The post construction settlements, which occur after the
elements. This load transfer is typically accomplished elastic settlement, are conceived to be driven by the
through an engineered load transfer platform (LTP) or a consolidation of the clay and organic layers below the
bearing layer. The amount of settlement at 100 years
granular zone of a certain thickness between the tops of
after construction was calculated to be less than 50 mm
the piers and the underside of the embankment or (2 inches).
structures to be supported. The load arches from the
GCC through the LTP material to provide a more uniform 5.0 FINITE ELEMENT ANALYSIS
support condition beneath the structure. If the spacing of
the GCC elements and/or thickness of the granular LTP Finite Element Analysis (FEA) was performed to verify
is such that uniform support cannot be provided, a stiffer the design spacing of the GCCs, verify it provides
LTP created out of concrete may be required. uniform support to the new road, and to estimate the
The LTP was constructed using granular material normal and shear stresses on the GCCs that develop
with 30% maximum fines content and a minimum of 30- from embankment fill, pavement, and traffic loading. The
settlement results from the FEA was also compared with
degree friction angle. Finite Element Analysis (FEA)
those from the Settle 3D model.
using Plaxis software showed that the load arching will
occur within the new fill and the existing fill to ensure that Plaxis 2D was used to conduct the numerical studies.
the loads are fully transferred to the rigid inclusions The work was performed by:
within the zone above the compressible peat or organics. • Selecting critical sections for analysis
In certain areas where utilities were present either • Creating a generalized soil model for each section
below the cone of arching (no support) or were too close • Discretizing the generalized model into finite
to the tops of the GCC element (nonuniform support), a elements
LTP consisting of either a concrete mud mat or layers of • Assigning material constitutive models and
geogrid and stone were designed to transfer the load to parameter values to each layer
the GCCs. • Establishing initial stresses
• Apply the embankment fill, pavement and 12 kPa
4.0 SETTLEMENT ANALYSIS traffic loads at representative stages
• Solve for the resulting element stresses and nodal
Because the GCCs are designed to transfer the entire displacements.
embankment load to the sands below, lower zone
settlements below the bulbs of the GCCs can be 5.1 Model Geometry for Critical Section
calculated using Settle 3D software with load application
at the same elevation as the tip of the GCC elements. Figure 3 shows the geometry and layering of the critical
Please see Table 4 for a summary of calculated cross-section located at Station 0+330. This is the critical
settlements at various times for this project. dross section because it represents the maximum GCC
spacing (S) to embankment fill height (Hf) ratio (S/Hf) for
the project. Other less-critical sections had a greater
Table 4. Summary of Expected time and Settlement amount of embankment fill but a close GCC spacing.
These cross sections were also studied. The FEA model
Station Time Expected Total Post-Construction was created from the generalized profile shown in Figure
(yr) Settlement Settlement 1. Axisymmetric formulation with a radius of 0.9 m is
(mm) (mm) used for analysis, which represents the response of a
0+430 0.25 72 40 unit GCC cell located under the embankment installed at
100 119 47 a spacing of 1.6 m on-center.
0+460 0.25 37 13
5.2 Constitutive Models and Parameter Values
100 49 <25
0+600 0.25 <25 <25 Table 5 provides the constitutive models and parameter
100 <25 <25 values used for each layer in the model. Because the
initial stresses are large compared to the applied change
Following the installation of the rigid inclusions, in stress, linear elastic – perfectly plastic (Mohr-
Coulomb) models were used for the asphaltic concrete
settlement monitoring was performed during and after
and soil layers. Drained parameter values for both shear
the embankment construction for a minimum of 90 days. strength and compressibility were used to appropriately
The results of settlement monitoring are presented in represent both initial stress levels and long-term
(section 5.7) of this paper. deformations.
 Figure 3. Geometry and layering

Table 5. Plaxis Parameters

Unit Young’s
Cohesion, su Poisson’s
Materials Elevation (m) Weight Φ’ (deg) Modulus E Model
c’, (kPa) (kPa) Ratio (v)
(kN/m3) (MPa)
Roadway 78.1 to 77.5 21 20 32 24 (500 ksf) 0.3 MC
Embankment Fill 77.5 to 76.6 20 0 32 24 (500 ksf) 0.3 MC
Fill 1 76.6 to 75.6 20 0 34 38 0.3 MC
Fill 2 75.6 to 72.8 19 0 29 15 0.3 MC
0.22 (Es)
Peat 72.8 to 70.5 11.5 0 22 13 0.48 (10 ksf, 0.4 MC
Eoed)
Compact Sand 70.5 to 62.5 20 0 34 38 0.3 MC
Sandy Silt 62.5 to 59.0 19 0 29 38 0.3 MC
Compact Sand 59.0 to 57.0 20 0 34 38 0.3 MC
0.92 (Es)
Organic Silt 2 57.0 to 55.0 15.4 0 22 13 1.96 (41 ksf, 0.4 MC
Eoed)
Compact Sand 55.0 to 50.5 20 0 34 38 0.3 MC
0.92 (Es)
Organic Silt 2 50.5 to 44.5 15.4 0 22 13 1.96 (41 ksf, 0.4 MC
Eoed)
Compact Sand 44.5 to 38.5 20 0 34 38 0.3 MC
Bedrock <38.5
Unit Young’s
Cohesion, Φ’ Poisson’s
Materials Elevation (m) f’c (MPa) Weight Modulus Model
c’, (kPa) (deg) Ratio (v)
(kN/m3) E (MPa)
14’ GCC 76.6 to 68.5 31 22 - - 26,345 0.2 LE
5.3 Computed Stresses at the Completion of Traffic The relative shear stress computed and presented in
Loading Representing Long Term Conditions Figure 6 shows that the embankment fill load arches to
the GCC elements. The red zones signify that the soil
After the initialization of stresses in the model, the shear capacity is fully developed within these soil
vertical, horizontal, and relative shear stresses were elements. At the bottom of the GCC, the red zones show
computed at the completion of embankment filling, the way in which end bearing is developing in the sandy
pavement placement and at the completion of traffic soil.
loading stages. The vertical stresses presented in Figure The relative shear stress levels in the upper organic
4 show that the GCC elements attract most of the soils remain at the original relative shear stress values
embankment load with a maximum GCC stress of indicating that little to no shear stress is being transferred
approximately 1180 kPa computed near the bottom of to the organic soils during loading. This ultimately results
the element. The stress contours in the soil material in little deformation in this layer.
between the GCC elements show the attraction of the
load to the GCC elements through skin friction.

Figure 6. Relative Shear Stress – Traffic load

Figure 4. Vertical Stress – Traffic load 5.4 Computed Settlements

The horizontal stresses shown in Figure 5 are less Figure 7 presents vertical displacements at the end of
than 10% of the compressive strength (f’c) of the pavement construction. Please note that these
compressive strength of concrete that the GCC elements computed displacements are total long-term
are made of. Empirically, the flexural strength ranging deformations. In all cases, the computed settlements are
between 10 and 15 percent of the compressive strength uniform throughout the mesh. This is because the total
of a particular concrete is acceptable. settlements are dominated by lower zone organic soil
settlements whereby the entire upper zone is settling
uniformly downward due to lower zone compression. As
a result, the computed long-term top-of-mesh differential
settlement is negligible. Long-term settlements of 70mm
for embankment filling and 128mm for filling plus
pavement construction were computed.

5.5 Computed Stresses within the Upper Organic Soils

Figure 8 presents the computed top-of-organic layer

vertical stress conditions for each loading stage.
Comparing conditions at a mesh radius of 0.5 m, the
initial stress of 51.78 kPa increases to 51.85 kPa during
embankment filling, to 51.925 kPa during paving, and
52.08 kPa during traffic loading. The change of pressure
for the entire load sequence is 0.3 kPa. Confirming the
stress conditions presented above, the loading
sequence applied by filling, paving, and traffic loading
Figure 5. Horizontal Stress – Traffic load results in virtually no applied stress to the organic soils.
 Fill Height

Settlement

Figure 9. Settlement monitoring result 90 days after

embankment placement

Figure 7. Vertical Displacement – End of Pavement 5.7 Discussion

Based on the analyses performed and settlement

monitoring results, the following are noted:
• The result of the FEA showed a total system
settlement of 70mm (2.8 inches) for embankment
filling and 128mm (5.0 inches) for filling plus
pavement construction.
• The computed settlements are uniform in the upper
zone and are dominated by the long-term lower
zone consolidation.
• Because of the high stiffness of the GCC elements
relative to the soil, they transfer the embankment
load to the granular lower zone soil resulting in
negligible stress induced in the upper organic
materials. Thus, the settlements that occurs during
filling and paving are a result of immediate elastic
settlement in the granular embankment material
and upper soil layers.
Figure 8. Vertical Change in Stress on Peat Layer • The effectiveness of this approach (bypassing the
organic layer) is demonstrated by comparing the
5.6 Settlement Monitoring settlements of the GCC supported section with
those of the pre-loaded section. The GCC
As part of the project specifications, settlement supported section experienced significantly less
monitoring was performed, and weekly readings were settlement than the pre-loaded section and over a
taken for a period of 90 days after the completion of shorter time frame.
embankment filling and before the construction of • As shown in Figure 8, the settlement monitoring
pavement and other settlement sensitive utilities. result shows the analysis and the assumptions
The section supported by the Geopier GCC system made are appropriate and reasonably conservative.
was represented by monitoring points SP-CS-09C, 10A, • Although the remaining long-term settlements from
10B, 11,12, 13, 14, 15BR2 and 16R. The settlement paving (128 mm – 70 mm = 58 mm or 2.3 inches)
result of the monitoring point showing the most vertical are slightly greater than the project settlement
movement within the ground improvement section is requirements of 50 mm, the settlement monitoring
shown in Figure 9, which shows settlements of less than result shows that this value likely results from an
50mm before pavement placement, which is the pre- over-conservative representation of the lower zone
construction settlement. This value is lower than those compressibility in the FEA and that actual long-term
computed using Settle 3D and Plaxis, thus providing settlements will be less.
confidence that the overall settlement criteria of the
project meet design specifications.
6.0 CONCLUSION AND RECOMMENDATIONS 7.0 References

6.1 Conclusion
Fox, N.S. and Cowell, M.J. (1998). “Geopier Soil
After three (3) months of monitoring and a general trend Reinforcement Manual.” Geopier Foundation
to the asymptote to indicate the settlement has Company, Blacksburg, Virginia.
ostensibly concluded, the project settlement criteria are
adequately met by the GCC ground improvement Lawton, E.C., Fox, N. S., and Handy, R. L. (1994).
system for the section of roadway embankment that it is “Control of settlement and uplift of structures using
supporting. The GCC supported section, when short aggregate piers.” In-situ Deep Soil
compared to an adjacent area that was preloaded, Improvement, Geotechnical Special Publication No.
displayed reduced settlement magnitude, and reduced 45, American Society of Civil Engineers, 121-132.
settlement duration.
ASTM. (1994). ASTM D 1143: Standard Method for Piles
6.2 Recommendations under Static Axial Compression Load (Reapproved
1994).
It is recommended to continue monitoring the preloaded
section to determine when the settlement has WSP, (2020). “Draft Assessment of Ground
concluded, and further studies should be conducted on Improvement Requirements along Commissioners
additional sections of this roadway including major Street Revision 1, Port Lands Flood Protection
infrastructure items such as trunk sewers supported by Enabling Infrastructure Project”
ground improvement.

6.3 Acknowledgements

The authors gratefully acknowledge Dr. Jianhua Li for his

valuable contribution to performing the Finite Element
Analysis and reviewing this paper.
Wednesday, October 4, 2023

CASE STUDIES I
The case for true probabilistic internal
stability design of geosynthetic MSE walls
R.J. Bathurst
GeoEngineering Centre at Queen’s-RMC, Kingston, Canada

ABSTRACT
The paper makes the argument for a fully probabilistic approach to quantify the margins of safety for the internal stability
limit states for geosynthetic mechanically stabilized earth (MSE) walls. The three limit states considered are tensile
strength, pullout and the soil failure limit state that appears in the stiffness method. The methodology is based on statistical
quantification of the accuracy (bias) of load and resistance terms that have been collected from large databases of
instrumented MSE walls for load side models, and laboratory measurements compiled in databases for resistance side
models. A closed-form solution is presented to compute the reliability index for internal stability limit states with lognormal
distributions of bias values and nominal values. The concept of level of understanding that appears in Canadian load and
resistance factor design (LRFD) practice appears in the formulation and thus allows for the injection of experience into the
probabilistic approach. The general approach provides a quantitative link between the classical factor of safety used in
working stress design past practice, and margins of safety expressed within a probabilistic framework. The closed-form
solution is used to compute the reliability index for the three limit states for as-built polyester (PET) strap MSE walls.

RÉSUMÉ
Le document plaide en faveur d’une approche entièrement probabiliste pour quantifier les marges de sécurité pour les
états limites de stabilité interne des murs géosynthétiques en terre stabilisée mécaniquement (MSE). Les trois états limites
considérés sont la résistance à la traction, l’arrachage et l’état limite de rupture du sol qui apparaît dans la méthode de
rigidité. La méthodologie est basée sur la quantification statistique de la précision (biais) des termes de charge et de
résistance qui ont été recueillis à partir de grandes bases de données de parois MSE instrumentées pour les modèles côté
charge, et des mesures de laboratoire compilées dans des bases de données pour les modèles côté résistance. Une
solution fermée est présentée pour calculer l’indice de fiabilité pour les états limites de stabilité interne avec des
distributions log-normales des valeurs de biais et des valeurs nominales. Le concept de niveau de compréhension qui
apparaît dans la pratique canadienne de calcul des facteurs de charge et de résistance (LRFD) apparaît dans la formulation
et permet ainsi l’injection d’expérience dans l’approche probabiliste. L’approche générale établit un lien quantitatif entre le
facteur classique de sécurité utilisé dans la pratique antérieure de la conception des contraintes de travail et les marges
de sécurité exprimées dans un cadre probabiliste. La solution de forme fermée est utilisée pour calculer l’indice de fiabilité
pour les trois états limites des parois MSE en polyester (PET) conformes à l’exécution.

1 INTRODUCTION databases. These data are a key component of

probabilistic design because they capture the uncertainty
Current practice in Canada for the internal stability design in load and resistance model predictions (i.e., model error).
of mechanically stabilized earth (MSE) walls follows the The paper then goes on to show how these model errors
load and resistance factor design (LRFD) approach are combined with uncertainty in the choice of nominal load
adopted in the Canadian Highway Bridge Design Code and resistance model estimates consistent with the
(CHBDC) (CSA 2019). If a limit state is just satisfied at time concept of “level of understanding” found in the CHBDC.
of design then the structure is safe, however, how safe in The paper concludes with examples of the calculation
probabilistic terms is unknown. Probabilistic reliability- of probabilistic margins of safety for tensile strength, pullout
based design for internal stability limit states for MSE walls and the soil failure limit states for PET strap MSE walls
constructed with geosynthetic and steel reinforcement using different load and resistance models.
materials provides a more nuanced appreciation of the
margin of safety for these systems compared to
conventional factor of safety, partial factor, and LRFD 2 BACKGROUND
approaches.
The paper introduces the basics of probabilistic The three limit states that are the focus of this paper are
analysis and design for internal stability limit states for shown in Figure 1.
geosynthetic MSE walls. The general approach uses a In this paper, each limit state equation contains a single
closed-form solution for reliability index that is easily resistance term and a single load term due to soil self-
implemented in a spreadsheet. The paper demonstrates weight as follows:
how bias statistics for tensile load and pullout model
accuracy have been gathered from instrumented wall load
measurements and from laboratory pullout box test
 The parameters Rm and Qm are the actual (measured)
resistance and actual (measured) load, respectively. Of
course these values are not available at time of design,
rather the designer computes nominal values for resistance
(Rn) and load (Qn) using closed-form (analytical) solutions
found in design codes and design manuals, and sometimes
from the research literature. Example equations are shown
in Figure 1. The ratio of nominal resistance and nominal
load terms can be recognized as the nominal factor of
safety Fn = Rn/Qn that is familiar from allowable stress
design past practice.
Nominal values can be transformed to measured
values by multiplication with bias values (R and Q) if these
values are available. In simple terms, a bias value
quantifies the accuracy of a model prediction. A perfect
model has a bias value of 1. This is an unlikely occurrence
Figure 1. Nominal tensile load (Qn) and nominal for simple soil-structure interaction problems of the type
resistance (Rn) for internal limit states for geosynthetic encountered for the internal stability design of MSE walls.
MSE walls. H = height of wall, z = depth of reinforcement Fortunately, bias data are available in the literature as
layer below crest of wall,  = unit weight of reinforced demonstrated later in the paper. If nominal values and bias
values are treated as random values with known
soil,  = soil friction angle, Le = length of reinforcement
distributions, Monte Carlo simulation can be used to
layer beyond the internal failure surface, Sv =
compute the probability of failure Pf. The probability that a
contributory area (vertical spacing) of each
limit state is not satisfied (failure) is expressed as follows:
reinforcement layer, Tmax = maximum tensile load (e.g.,
kN/m), K = equivalent lateral earth pressure coefficient,
Rc = coverage ratio, Tal = maximum allowable tensile Pf = P(g  0) = 1 − () [2]
load (e.g., kN/m), Pc = pullout capacity (e.g., kN/m), F*
= dimensionless interface shear coefficient, Tj = tensile where, () is the standard normal cumulative distribution
load corresponding to soil failure. function (NORMSDIST() in Excel). The link between
probability of failure (Pf) and reliability index () is shown in
Figure 2.
Rm R R n Fortunately, if all random variables are usefully
g= −1 = −1 [1] approximated by lognormal distributions, the following
Qm Q Qn
closed-form solution is available to compute  and follows
from basic probability theory (Bathurst and Javankhoshdel
2017):

 μ μ (1+COVQn )(1+COVλQ ) 
2 2

ln  (λR Rn
) 2 2

 μ λQ μ Qn (1+COVRn )(1+COVλR ) 

β=
2 2 2 2
(1+COVQn )(1+COVλQ )(1+COVRn )(1+COVλR ) [3]
2 2
(1+ρ R COVRn COVλR ) (1+ρ Q COVQn COVλQ )
ln 2
(1+ρ n COVRn COVQn )

This equation can be used in an Excel spreadsheet to solve

for  and to carry out parametric analyses. Parameters Rn,
Qn, R and Q are mean values of nominal resistance
and load values (Rn and Qn), and mean values of
resistance and load bias values (R and Q), respectively.
Their corresponding coefficients of variation (COV) are
denoted as COVRn, COVQn, COVR and COVQ. The
nominal resistance value (Rn) and nominal load value (Qn)
computed at time of design for a limit state design equation
are equivalent to Rn and Qn in the above equation.
Lognormal distributions for load and resistance bias values
have been shown to be sufficiently accurate for practical
purposes for the internal limit states of MSE walls
Figure 2. Probability of failure (Pf) and link to reliability constructed with extensible (geosynthetics) and
index (). Here  and  are the mean and standard inextensible (steel) reinforcement materials (ignoring
deviation of the distribution of g. corrosion) (e.g., Bathurst et al. 2019b,c, 2020; Miyata et al.
2019a,b). The remaining terms in Equation 3 are bias Simple linear regression between computed load bias
correlation coefficients Q and R between bias values and values and predicted load can be carried out. The R in R 2
corresponding nominal load and resistance values, that is an output from simple linear regression is the
respectively. It is often the case that the accuracy of a absolute value of Pearson’s correlation coefficient. The
model (bias) varies with the magnitude of predicted stiffness load model gives Q = +0.09 which can be taken
nominal value. The (Pearson) correlation coefficient as 0 at a level of significance of 5%. From a bias
captures this dependency. The remaining parameter in this dependency point of view, the stiffness method is a better
formulation is n which is called the nominal correlation model than the simplified method because the latter shows
coefficient between nominal load and nominal resistance. a strong negative bias dependency with predicted load at
This parameter has a non-zero value when there are a 5% level of significance (Q = −0.41).
common random variables in the equations for Rn and Qn
(e.g., soil friction angle and unit weight in the equations for 3.2 Resistance
Tmax and Pc that appear in Figure 1) (see Lin and Bathurst
2018). Equation 3 can be written as: CDF plots for two different geogrid pullout test data sets
are presented in Figure 3b. The measured pullout
capacities were taken from the results of pullout tests
β = A  ln(Fn ) + B [4] carried out on specimens of different geogrids embedded
in cohesionless soil materials. Comparison of the bias
where, A and B are equations containing the statistical statistics for the two pullout capacity models shows that
quantities in Equation 3. If all statistical quantities in
Equation 3 remain constant, then this equation shows that
reliability index varies as the logarithm of nominal factor of
safety; thus, Equations 3 and 4 provide a quantitative link
between the classical factor of safety used in past practice,
and margins of safety expressed probabilistically.

3 BIAS VALUES

3.1 Load

Tensile loads under wall operational conditions have been

collected from full-scale instrumented structures by the
author and co-workers over many years (e.g., Allen and
Bathurst 2015). For example, measured loads were
deduced from 96 measurements recorded from 21 walls
constructed with planar polymeric geogrid reinforcement
layers (Bathurst et al. 2019a). The matching predicted
a)
loads were calculated using two different load models;
simplified method and stiffness method. The details of the
analytical models (shown in Figure 1) are not explained
here for brevity, and can be found elsewhere (e.g., Allen
and Bathurst 2015, 2018). Regardless, these details are
not necessary for the objectives of the paper and the main
narrative. Each pair of measured and predicted data was
used to compute the bias of the predicted value. The
computed bias values are plotted using cumulative
distribution function (CDF) plots as shown in Figure 3a. The
mean and coefficient of variation (COV) of both bias data
sets were used to compute the approximations shown in
the figure. The bias data for both models are reasonably
well represented by a straight line in the CDF plots with a
logarithmic bias axis. The visually poorer fit at the lower
tails is not a concern because it is the load bias data at the
upper tail that most influences the probability that the limit
state function (g) is not satisfied. The stiffness method load
model can be judged to be a better model than the
b)
simplified load model because the average bias value is
Figure 3. Cumulative distribution function (CDF) plots for a)
close to 1, and the spread in bias values is less. reinforcement tensile load bias values for geogrid MSE
As noted earlier, bias dependency with nominal load walls using AASHTO simplified and stiffness load models,
values is a potentially important contributor to the estimate b) geogrid pullout bias values using two pullout models
of the reliability index for MSE wall limit state functions. Its (PM1 from AASHTO (2020) and PM2 from Huang and
contribution (Q) is considered explicitly in Equation 3. Bathurst 2009 ) (data from Bathurst et al. 2019a,c).
from the perspective of model accuracy, PM2 is better than
PM1. The corresponding correlation coefficients for pullout
models PM1 and PM2 are R = −0.46 and 0 at a level of
significance of 5%. Again, pullout model PM2 can be
judged to be superior to model PM1.
The mean of the bias values for the strength limit state
using geosynthetic reinforcement is taken as the nominal
allowable tensile strength (Tal) and COVR > 0 is computed
from the variability in ultimate tensile strength from multiple
tensile tests on the same material taken to rupture (i.e., R
= measured strength/nominal ultimate strength). The same
approach is taken for steel reinforcement with adjustments
made for the influence of corrosion (Bozorgzadeh et al.
2020a,b; Bathurst et al. 2021).

4 NOMINAL VALUES

As explained earlier, nominal values are computed at time

Figure 4. Sao Paulo wall (data from Bathurst et al. 2020).
of design using closed-form analytical equations for load
and resistance terms in each limit state function. The
estimate of Qn and Rn will have some uncertainty because common input parameters and are thus uncorrelated.
soil properties such as the soil friction angle () and unit However, for the pullout limit state there are common
weight () that appear in Figure 1 have variability. Using parameters for MSE walls (i.e.,  and ). Hence, the non-
estimates of mean and COV of these soil parameters, the zero value of n must be evaluated on a project-specific
mean and COV of the nominal values are easily calculated basis. To simplify calculations in this paper, n = 0.
(i.e., Rn, Qn and COVRn, COVQn). However, there are
other sources of uncertainty that influence the best
estimate of nominal values; these are the quantity and 5 EXAMPLE PET STRAP MSE WALLS
quality of materials testing, knowledge of project-specific
foundation conditions and geometry, confidence in the The examples here are taken from Bathurst et al. (2020).
models used and their applicability to project-specific The structures are full-scale instrumented test walls
conditions, and so on. A more holistic approach to account constructed in Japan, Brazil and the US. The Sao Paulo
for all sources of uncertainty is to lump all sources of wall in Brazil is shown in Figure 4.
uncertainty into a “level of understanding”. This approach Figure 5a shows the distribution of  for the soil failure
is used in load and resistance factor design (LRFD) for limit state with depth below the wall crest. The margins of
foundations in Canada (Fenton et al. 2016; CSA 2019). safety expressed as  can be seen to vary between walls
The approach was adopted to reward designers with higher and between reinforcement layers in each structure. The
resistance factors in LRFD limit state design equations target  value for the soil failure limit state for geosynthetic
when the level of understanding is higher than typical. It is reinforced soil walls is  = 1.0 (Pf = 1/6). All walls satisfy
admittedly subjective but has the advantage of injecting this limit state even though the walls were not originally
human judgement into the design process. Bathurst et al. designed using the stiffness method. The target  value
(2019b,c) recommended COV values of 0.1, 0.2 and 0.3 may appear low, however this is a serviceability limit state
for the COVRn and COVQn for “high”, “typical” and “low” as explained by Bathurst and Allen (2023). If this
levels of understanding for nominal load and resistance serviceability limit state is not satisfied, deformations may
quantities in Equations 1 and 3 when applied to internal become large, but collapse of the wall is not expected. A
limit states for MSE walls. This approach and these values value of  = 1 (Pf = 1/6) was used for LRFD calibration of
are used with the exception of the resistance terms for the soil failure limit state that applies only to the stiffness
tensile strength and soil failure limit states. For example, method (Bathurst and Allen 2023).
the resistance Rn = Tal is taken as deterministic for Figure 5b shows the linear relationship between the
geosynthetic walls and thus COVRn = 0. The uncertainty in logarithm of nominal factor of safety and reliability index 
the tensile strength of the reinforcement is captured by the which is expected from Equation 4. The minimum
variability in strength from multiple tensile tests on the deterministic margin of safety for these walls corresponds
same material quantified by COVR > 0 as explained to a factor of safety Fn = 1.9 for the Delaware wall.
earlier. The same treatment applies for the soil failure limit Tensile strength and pullout limit states demonstrated
state for geosynthetic MSE walls when the stiffness next are ultimate limit states and a larger value of  = 2.33
method is used for design (Bathurst and Allen 2023). For (Pf = 1/100) is recommended for reliability theory-based
the same tensile strength limit state, the nominal cross- design and LRFD calibration. This value may also appear
correlation between reinforcement load (Qn) using the low as well, but it is important to recognize that MSE walls
simplified and stiffness load models, and reinforcement are highly strength redundant structures. Should one
strength (Rn = Tal) is ρn = 0. This is because the models for element fail, there are neighbouring elements that can
reinforcement strength and reinforcement load do not have compensate.
 a)
a)

b) b)
Figure 5. a)  distribution for the soil failure limit state Figure 6. a)  distribution for tensile strength limit state
using AASHTO (2020) stiffness load model, and b) using different load models, and b)  versus nominal
 versus nominal factor of safety (Fn) for the soil failure factor of safety for tensile strength limit state (data from
limit state (data from Bathurst et al. 2020). Note: margin Bathurst et al. 2020). Note: margin of safety against Tmax
of safety against Tmax ≥ TJ at maximum allowable ≥ Tal and COVQn = COVRn = 0 unless noted otherwise.
reinforcement strain εmax = 2% and COVRn = COVQn = 0.
reinforcement loads with depth for simple reinforcement
arrangements that follow notions of classical earth
Figure 6 shows the results of similar analyses for the pressure theory and from which these methods are
tensile strength limit state for the Sao Paulo wall using adapted. A distinctive feature of the trend in β values
three different load models found in AASHTO (2020). In computed using the stiffness method is a limiting constant
fact, the coherent gravity method is specified in AASHTO β value for layers located closest to the bottom of the wall.
for steel (inextensible) MSE walls only. Nevertheless, This outcome is expected because the stiffness method
Miyata et al. (2018) showed that some designers have predicts constant tensile load under operational conditions
used the coherent gravity method for PET strap walls once a critical depth below the crest of the wall is reached
because of their relatively high stiffness compared to (Allen and Bathurst, 2015).
polymeric sheet reinforcement products. The data plots for each wall in this figure were
The plots in Figure 6a show that the margin of safety is computed with COVQn = 0, which gave the lowest β value
above the recommended target β = 2.33, and in most cases for all load calculation methods with the exception of the
above β = 3.54. β values are very similar for loads stiffness method. Figure 6a also illustrates the influence of
calculated using the AASHTO coherent gravity method and level of understanding using the AASHTO stiffness
the AASHTO simplified method. The safest outcomes method. The influence of level of understanding using all
correspond to loads computed using the stiffness method. three load methods is not shown in order to avoid visual
The first two load methods in each figure show a general clutter. For low level of understanding (COVQn = 0.3), the
trend of decreasing margins of safety with depth. This may probabilistic margin of safety for all layers is lower than the
be expected since these methods predict increasing case when the concept of level of understanding is ignored
 the topmost reinforcement layer. Experience with design of
MSE walls has shown that the layers closest to the top of
the wall are most likely to have the lowest margins of safety
in both deterministic and probabilistic design. A practical
solution is to locally increase the length of these layers.
This increases the pullout capacity which in turn increases
the nominal factor of safety as seen in Figure 7b.
Figure 7b shows that using pullout models 2 and 3 will
satisfy Fn > 2 and  > 2.33 without having to increase the
length of any layer. Another example using an as-built full-
scale geogrid MSE wall can be found in the paper by
Bathurst and Allen (2021).

6 CONCLUSIONS

The paper shows how load measurements from large

databases of instrumented MSE walls for the load models
a) in a limit state function, and laboratory measurements
compiled in databases for the resistance side models, are
used to quantify model accuracy (bias). Bias statistics are
used to compute probabilistic margins of safety for tensile
strength, pullout and soil failure for geosynthetic MSE
walls. This paper and prior similar work by the author and
co-workers has shown that model accuracy can vary widely
between different load models and resistance models used
for a particular class of MSE wall type and between
different wall types. Consequently, true margins of safety
in probabilistic terms are a function of wall type and the
accuracy (bias) of the models for resistance and load sides
in limit state design equations for tensile strength, pullout
and soil failure that appear in current North American
design codes.
The general approach holds promise to move current
design practice for MSE walls in North America to a
b) rigorous probability of failure framework and, within that
Figure 7. Sao Paulo wall: a)  distribution for pullout limit framework, to make informed choices between competing
state using AASHTO coherent gravity load model and solutions for MSE walls performing the same retaining wall
three different pullout models, b)  versus nominal function.
factor of safety for pullout failure limit state (data from The general approach provides a quantitative link
Bathurst et al. 2020). Note: margin of safety against between deterministic working stress design margins of
Tmax ≥ Pc and COVQn = COVRn = 0. safety and probabilistic margins of safety while at the same
time allowing for the injection of experience through the
concept of level of understanding. At the very least the
(as in US LRFD practice). However, computed  values probabilistic approach advanced in this paper can be
are well beyond practical levels and are shown here to viewed as a complimentary approach to current working
show trends only. stress design and LRFD practice for MSE wall designers.
Reliability  versus logarithm of nominal factor of safety
plots appear in Figure 6b. For the coherent gravity method
load case, the minimum nominal factor of safety for the ACKNOWLEDGMENT
reinforcement layers closest to the bottom of the wall is less
than 2.5, but the margin of safety in probabilistic terms is Financial support for this study was provided by the Natural
above the minimum target reliability index of  = 2.33. Sciences and Engineering Research Council (NSERC) of
Figure 7a shows the distribution of  for the pullout limit Canada through Grant Number RGPIN-2018-04076.
state for the Sao Paulo wall. For visual clarity, results are
shown for a single load model (AASHTO (2020) coherent
gravity method) and COVQn = 0 to highlight the influence of
choice of pullout model on calculation outcomes. The
details of the pullout models can be found in Bathurst et al.
(2020). For each layer, the margin of safety expressed by
 increases in the order of pullout model 1, 2 and 3. Pullout
model 1 falls slightly below the target value of  = 2.33 for
REFERENCES CSA 2019. Canadian Highway Bridge Design Code.
CAN/CSA-S6-19. Canadian Standards Association
Allen, T.M. and Bathurst, R.J. 2015. An improved simplified (CSA), Mississauga, Ontario, Canada.
method for prediction of loads in reinforced soil walls. Fenton, G.A., Naghibi, F., Dundas, D., Bathurst, R.J. and
Journal of Geotechnical and Geoenvironmental Griffiths, D.V. 2016. Reliability-based geotechnical
Engineering 141(11): 04015049. design in the 2014 Canadian Highway Bridge Design
Allen, T.M. and Bathurst, R.J. 2018. Application of the Code. Canadian Geotechnical Journal 53(2): 236-251.
Simplified Stiffness Method to design of reinforced soil Huang, B. and Bathurst, R.J. 2009. Evaluation of soil-
walls. Journal of Geotechnical and Geoenvironmental geogrid pullout models using a statistical approach.
Engineering 144(5): 04018024. ASTM Geotechnical Testing Journal 32(6): 489-504.
Allen, T.M., Nowak, A.S. and Bathurst, R.J. 2005. Lin, P. and Bathurst, R.J. 2018. Influence of cross-
Calibration to determine load and resistance factors for correlation between nominal load and resistance on
geotechnical and structural design. Transportation reliability-based design for simple linear soil-structure
Research Board Circular E-C079. Transportation limit states. Canadian Geotechnical Journal 55(2): 279-
Research Board, Washington, DC. 295.
AASHTO 2020. LRFD Bridge Design Specifications, 9th Miyata, Y., Bathurst, R.J., and Allen, T. M. 2018. Evaluation
Ed., American Association of State Highway and of tensile load model accuracy for PET strap MSE walls.
Transportation Officials (AASHTO), Washington, DC, Geosynthetics International, 25(6), 656-671.
USA. Miyata, Y., Bathurst, R.J. and Allen, T.M. 2019a.
Bathurst, R.J. and Allen, T.M. 2021. Reliability assessment Calibration of PET strap pullout models using a
of internal stability limit states for two as-built statistical approach. Geosynthetics
geosynthetic MSE walls. International Journal of International 26(4): 413–427.
Geoengineering Case Histories 6(4): 67-84. Miyata, Y. and Bathurst, R.J. 2019b. Statistical assessment
Bathurst, R.J. and Allen, T.M. 2023. LRFD calibration for of load model accuracy for steel grid reinforced soil
soil failure limit state using the Stiffness Method. walls. Acta Geotechnica 14: 57-70.
Canadian Geotechnical Journal
(https://doi.org/10.1139/cgj-2022-0499) (online).
Bathurst, R.J., Allen, T.M., Lin, P. and Bozorgzadeh, N.
2019a. LRFD calibration of internal limit states for
geogrid MSE walls. Journal of Geotechnical and
Geoenvironmental Engineering 145(11): 04019087.
Bathurst, R.J., Allen, T.M., Miyata, Y., Javankhoshdel, S.
and Bozorgzadeh, N. 2019b. Performance-based
analysis and design for internal stability of MSE walls.
Georisk 13(3): 214-225.
Bathurst, R.J., Bozorgzadeh, N., Miyata, Y. and Allen, T.M.
2021. Reliability-based design and analysis for internal
limit states of steel grid-reinforced mechanically
stabilized earth walls. Canadian Geotechnical Journal
58(5): 695-710.
Bathurst, R.J. and Javankhoshdel, S. 2017. Influence of
model type, bias and input parameter variability on
reliability analysis for simple limit states in soil-structure
interaction problems. Georisk 11(1): 42-54.
Bathurst, R.J., Lin, P. and Allen, T.M. 2019c. Reliability-
based design of internal limit states for mechanically
stabilized earth walls using geosynthetic reinforcement.
Canadian Geotechnical Journal 56(6): 774-788.
Bathurst, R.J., Miyata, Y. and Allen, T.M. 2020.
Deterministic and probabilistic assessment of margins
of safety for internal stability of as-built PET strap
reinforced soil walls. Geotextiles and Geomembranes
48: 780-792.
Bozorgzadeh, N., Bathurst, R.J. and Allen, T.M. 2020a.
Influence of corrosion on reliability-based design of steel
grid MSE walls. Structural Safety 84: 101914.
Bozorgzadeh, N., Bathurst, R.J., Allen, T.M. and Miyata, Y.
2020b. Reliability-based analysis of internal limit states
for MSE walls using steel strip reinforcement. Journal of
Geotechnical and Geoenvironmental Engineering
146(1): 0401911.
Design And Construction Of A Circular
Secant Pile Shaft For Connection To A Deep
Collector Sewer In Ottawa
Michael Cunningham & Troy Skinner
Marathon Underground Constructors Corporation, Greely, Ontario,
Canada

ABSTRACT
Construction of Stage 2 of Ottawa’s Confederation Line LRT system required re-routing of a shallow sewer near Cleary
Avenue and making a connection for it to the existing 1.5 m I.D. West Nepean Collector (WNC) sewer at about 13 m depth.
A water-tight shaft needed to be constructed to make that connection due to ground conditions consisting of glacial till with
intermittent thick water-bearing sand layers. An 8 m I.D. circular secant pile shaft was constructed to make that connection.
Challenges for the design and construction of the shaft included having to span over and around the WNC without
damaging it as well as sealing the ‘windows’ around the WNC against groundwater inflow and potential ground loss around
the shaft. The WNC was constructed in the 1950’s and the condition of the pipe was of concern. The method of construction
of the WNC was also unknown. Therefore, the pipe required monitoring for strains and vibration and was continuously
inspected by internal video camera feed for the duration of secant pile drilling above the pipe and during excavation of the
shaft. A pre-excavation grouting program of the windows around the WNC was also implemented through sleeves in the
secants to address the groundwater inflow concerns. This case study provides details on the design, instrumentation,
monitoring, construction, and performance of the shaft.

RÉSUMÉ
La construction de l'étape 2 du réseau de LRT de la Ligne de la Confédération d'Ottawa a nécessité le détournement d'un
égout peu profond près de l'avenue Cleary et son raccordement à l'égout West Nepean (WNC), de diametre 1.5 m, existant
à environ 13 m de profondeur. Un puits étanche à l'eau devait être construit pour établir cette connexion en raison des
conditions du sol constituées de till glaciaire avec des couches de sable épaisses intermittentes contenant de l'eau. Un
puits circulaire de 8 m de sécante a été construit pour établir cette connexion. Défis particuliers pour la conception et la
construction du puits, y compris la nécessité de s'étendre au-dessus et autour du WNC sans l'endommager, et comment
sceller les « fenêtres » autour du WNC contre l'afflux d'eau souterraine et la perte de sol potentielle autour du puits. Le
WNC a été construit dans les années 1950 et l'état du tuyau était une préoccupation. La méthode de construction du WNC
était également inconnue. Le tuyau devait donc être surveillé pour détecter les déformations et les vibrations, et être
inspecté en permanence par une caméra vidéo interne, pendant la durée du forage de pieux sécants au-dessus du tuyau
et pendant l'excavation du puits. Un programme de jointoiement des fenêtres autour de la CNB avant l'excavation a
également été mis en œuvre, à l'aide de manchons dans les sécantes, pour répondre aux préoccupations relatives à
l'afflux d'eau souterraine. Cette étude de cas fournit des détails sur la conception, l'instrumentation, la surveillance, la
construction et les performances de l'arbre.

1 INTRODUCTION AND PROJECT DESCRIPTION Designers, which is a joint venture that included WSP and
Hatch.
Construction of Stage 2 of Ottawa’s Confederation Line The LRT system was to be constructed on an alignment
LRT system required re-routing of a shallow sewer near through the area of Richmond Road and Cleary Avenue in
Cleary Avenue and making a connection from it to the an approximately 10 m deep cut-and-cover tunnel, and
existing West Nepean Collector (WNC) sewer at about 13 construction of the system would sever existing utilities.
m depth. This paper describes the design and construction EWC Designers had therefore identified that an existing
of the shaft needed for that connection, including notable shallow sanitary sewer in this area needed to be re-routed
details about the instrumentation program that was used to and a new connection made to the deeper collector sewer
monitor the condition of the WNC pipe while the shaft was (the WNC). Figure 1 shows the profile view of the
constructed. connection.
The owner of the LRT project is the City of Ottawa
(City). The overall design-build contractor for the Phase 2
Confederation Line construction is a joint venture of Kiewit,
Eurovia, and Vinci Construction (KEV). Marathon
Underground Constructors Corporation (Marathon) was
retained as a subcontractor to KEV. The design of the
permanent works was completed for KEV by EWC
 Figure 3. Plan view of new manhole (EWC Designers
2021)

Figure 1 Profile view of connection to WNC (EWC

Designers 2020)

Details on the original construction of the WNC were

limited, but it was apparently built in the late 1950’s and
likely by tunnelling. As-built drawings and a previous
condition survey carried out on behalf of the City indicated
the sewer to consist of an arch/horseshoe-shaped concrete
pipe, and to generally be in relatively good condition, but to
also have some areas of significant infiltration (see Figure
2).

Figure 4. Cross section of new manhole (EWC Designers

2021)

2 SUBSURFACE CONDITIONS

The subsurface conditions had been previously

investigated at the site by an initial investigation by the City
and then by some additional investigation on behalf of
KEV. A general stratigraphic profile for the shaft location
is shown in Figure 5.
Figure 2. WNC pipe cross section

EWC Designers developed a preliminary design for the

connection which would involve construction of a shaft
overtop of the WNC, removal of the top of the WNC pipe
down to about the bottom of the curved/arched upper-
portion of the pipe (‘springline’), construction of new
footings alongside the pipe wall, and then installation of a
new precast manhole structure (with drop-shaft pipe)
above that level, followed by backfilling between the
manhole and shaft (see Figures 3 and 4).
 The shaft design (see Figures 6 and 7) needed to consider
a few constraints and considerations, as follows:
1. Due to the risk of encountering permeable sandy layers
below the groundwater level, an essentially water-tight
shaft was required. Pre-excavation dewatering was not
considered feasible due to the variable presence,
depth, and thickness of the sand layers, as well as
project specifications which precluded groundwater
level lowering in this part of the overall project.
2. The WNC needed to remain operational during all
phases of this work, including during the actual
connection. The shaft therefore needed to ‘doghouse’
over and around the pipe and could not cut off flow in
the pipe. The shaft construction would therefore leave
unsealed ‘windows’ above and around the pipe, with
the risk of there being flowing ground conditions
through those windows.
3. There was concern by the owner and project team over
possible impacts to the WNC pipe. The construction
plan therefore needed to provide confidence that the
pipe would not be damaged and could be monitored for
the duration of construction.

4 SHAFT DESIGN

A circular secant pile shaft was selected as the preferred

shoring system for the shaft. The shaft had an inside
diameter of 8 m. At-rest earth pressures were used for the
design, in view of the stiff behaviour of this shoring system.
A total of 34 secants were used, with a diameter of 1,000
Figure 5. Stratigraphic profile. mm, resulting in an overlap of approximately 168 mm.
Some additional details about the design are as follows:
The geotechnical data showed the shaft to be located in an 1. The shaft was designed for the surcharge from an 80-
area where the bedrock surface dips steeply down to the
ton excavator working directly beside the shaft, albeit
east (to an apparent infilled valley). To the depth of the
supported on a steel ‘road plate’ to better distribute
shaft construction (13 m), the ground conditions consisted
the loads. For the analysis, the incremental but
of thin surficial layers of sandy fill and stiff weathered brown unbalanced lateral earth pressure from this surcharge
clay, over a thick deposit of glacial till. The bedrock surface
was distributed onto the shaft walls and the effect of
was interpreted to be at about 20 m depth.
this surcharge evaluated using a simplified structural
The glacial till was the important stratum from the
analysis of the wall. From this analysis it was
perspective of the shaft design, and generally consisted of determined that a ring beam was not required to
a gravelly silty sand, which is fairly typical of the glacial till
support the upper portion of the shaft walls. The
in the Ottawa area. However, the till deposit had also been
bending resistance of the secant shaft wall, which is
shown to have intermittent layers of ‘cleaner’ sand, with a
composed of overlapping drilled shafts and is
typical thickness of about 0.6 m. There was therefore a therefore not a monolithic concrete wall, was
concern about water inflow and ground loss if these layers evaluated based on shear-friction between secants,
were exposed during the shaft construction.
in consideration of the ring-compression in the shaft
The glacial till was also indicated to be fairly dense and
wall. The upper portion of the shaft, above the WNC,
bouldery, with bedrock coring methods having been was therefore entirely un-reinforced.
required for the boreholes to penetrate the deposit. The 2. The secants directly over and immediately adjacent
groundwater level was indicated to be at about 8 m depth,
to the WNC needed to be terminated slightly above
which is untypically deep for the Ottawa area and was
the crown of the WNC pipe. To that depth, the shaft
considered to possibly indicate the effects of localized was designed for compression-ring conditions.
subsurface drainage due to leakage into the WNC pipe. Conversely the remaining secants were extended 3
Accordingly, it was considered that this groundwater level
m below the bottom-of-excavation level (see Figure
might only reflect the piezometric level in the immediate
7). For this section of the shaft walls, every second
area of the WNC pipe and/or at the level of the pipe invert,
secant pile was reinforced with steel beams (W610)
with there being a potential for higher/perched conditions and spanned vertically between the embedded pile
and a gradient towards the pipe.
toes, and a reinforced concrete ring beam was
constructed slightly above the WNC.
3 CONSTRAINTS AND DESIGN CONSIDERATIONS
 Both strain gauges were installed perpendicular
to the pipe axis.
Figure 6. Plan view of shaft design.
3. A vibration monitor was similarly installed on the
crown of the pipe, beneath the centre of the shaft.
The vibration monitor was connected to a flashing
3. In the ‘short’ secants, directly over and beside the
light at ground surface, which would illuminate if
WNC, 4-inch nominal diameter steel pipes were
vibrations hit a ‘review’ level.
installed to allow subsequent grouting of the
4. Strain gauges were also installed on the crown of
‘windows’ around the WNC (see Figure 12).
the WNC pipe at locations under the secant pile
shaft walls, to monitor the pipe during secant pile
drilling (e.g., in the case of impact by the drill
5 INSTRUMENTATION PROGRAM
tooling). Two gauges were also installed in both
locations, to provide redundancy.
The instrumentation program implemented for this work
5. As excavation progressed downwards, survey
included the following components:
targets were installed at three levels on the
1. An illuminated camera was installed within the
interior faces of the shaft walls, and at eight
WNC pipe, immediately upstream of the shaft
positions around the circumference, so that the
position and looking downstream, so that any
deformations of the shaft could be monitored.
impacts to the pipe could be monitored in real-
time. After a ‘fixed’ camera was damaged during
The instrument positions are shown on Figure 9.
high-flow conditions in the collector sewer, the
camera was ultimately installed on a floating
platform, so that it could be removed during high
flow conditions (see Figure 8).
2. Strain gauges were installed on the crown of the
WNC pipe, under the centre of the shaft, to
monitor pipe strains during excavation. For
redundancy, two strain gauges were installed,
which ultimately proved to be a good decision.
 Figure 7. Cross section of shaft design.

Figure 8. Image from live camera feed.

Figure 9. Instrument positions.

6 SHAFT CONSTRUCTION

The steps involved in the shaft construction were as

follows:
1. Four additional boreholes were drilled by
Marathon for the sole purpose of confirming the
position of the WNC, to provide greater
confidence that subsequent drilling of the secant
 piles would not damage the WNC, whose position
was known only from a survey of the widely
spaced manhole locations, and whose outside
diameter was uncertain. Those boreholes were
drilled slightly closer to the WNC than the nearest
‘deep’ secants (100 mm inwards) – see Figure 6.
For the borehole drilling, rotary drilling methods
were used to advance a casing. Once each
borehole had advanced below the theoretical top-
of-pipe level, and if/where refusal was
encountered on boulders, coring was only
allowed in maximum 300 mm long runs, to confirm
that the borehole was advancing through a
boulder and not concrete. Van Ruth plugs were
ready on-site, to allow sealing of any holes made
in the WNC pipe walls, should the boreholes end
up penetrating the pipe. The two boreholes on the
north side of the pipe encountered a thin layer of
concrete at about the pipe crown level and it was
decided to shift the shaft 200 mm north to be more
certain of the secants not impacting the WNC
pipe. And an additional borehole was drilled on
the south side of the pipe to be more certain of the
south limit of the WNC pipe with the northward
shift of the shaft.
2. An EPS template was then used to construct a
concrete guide wall at ground surface for the
subsequent secant pile drilling (see Figure 10).
Figure 11. Photograph of secant pile drilling.

4. For the secants over the WNC, 4-inch nominal

diameter schedule 40 plain steel pipe with plastic
caps was set into the concrete (water-filled to
avoid floatation) at the locations shown on Figure
6, to allow for grouting.

5. Once the secants had cured, the grout pipes were

drilled out and pre-excavation grouting of the
‘windows’ carried out. For each grout-hole, this
procedure involved first advancing the drill hole to
the depths indicated on Figure 12. Two hoses
were then lowered down each grout hole. One
hose was equipped with a spinning nozzle, and
water was pumped down the hose at about 1,000
to 2,000 PSI to erode a larger cavity above and
Figure 10. Photograph of guide wall construction around the pipe. A second pipe was lowered
down the hole to just above the nozzle, connected
3. Secant pile drilling started once the guide wall had
to a hydro-vac truck, and was used to suction
cured. A Liebherr LB 36 foundation drill was used
excavate the material eroded by the spinning
for the secant pile drilling. Drill tooling consisted
water jet. A neat cement grout was then pumped
of augers, drilling buckets, and 1,000 mm
down (using a tremie line) at about 300 to 600 PSI
diameter segmental casing. Tremie concreting
methods were required for all secant pile
construction due to groundwater inflows into the
casings during drilling. The steel beams were
wet-set in the casings after concreting. The
design had these beams only positioned within
the bottom portions of the secants, extending
below the ring beam, where a complete
compression ring could not be provided at the
WNC pipe level. It is notable that, with the wet-
setting method, there was some uncertainly about
the orientation of the beams once installed.
 to fill the cavity.

Figure 14. Analysis of bending moments within WNC pipe.

Figure 12. Grout hole positions and depths.

6. Excavation then proceeded downward within the

shaft.
7. Once the excavation reached just above the
WNC, the cast-in-place reinforced-concrete ring-
beam was installed.
8. Once the ring-beam had cured, excavation
proceeded down around the pipe.
9. The excavation needed to extend to about the
invert level of the WNC to thereby reach the
manhole footing level. A key challenge was how
to set equipment on top of the WNC, without
damaging it, to excavate down around the pipe
(see Figures 13, 14, and 15).

Figure 15. Photograph of excavator working on WNC pipe.

10. The footing and bottom sections of the pre-cast

manhole were then set into place and the interior
backfilled up to spring line level of the WNC pipe
(see Figures 16 and 17).

Figure 13. Excavator positioning on top of WNC

Figure 16. Photograph of manhole installation. Figure 17. Photograph of backfilling.

11. The top of the sewer was then removed, the Some particular challenges that were encountered
connection completed, the remaining sections of during this process were as follows:
manhole installed, and the annulus between 1. During secant pile construction, additional
manhole and shaft was backfilled. concrete volumes were required versus the
theoretical volume of each drill-hole. As well,
some upward water flow was observed
around and through the concrete when the
casing was pulled (see Figure 18). The
additional concrete volume was attributed to
possible heaving and over-drilling of sandy
soil at the pile tip level, due to a sandy zone at
bottom-of-shaft level as inferred by drill
cuttings. The concrete quality was therefore a
concern for subsequent shaft excavation.
However, no defects in the concrete were
observed during excavation at depth.
Figure 18. Photograph of water flow around casing.

2. The WNC was found to be significantly thicker

than expected (per the theoretical cross-
section of the pipe) and appears to have been
constructed by hand-mining, possibly while
working in compressed air (see lagging
boards around pipe in Figure 19). Some of the
secants over the pipe were found to have
come into contact with the pipe (although the Figure 19. Photograph of groundwater at bottom of shaft.
instrumentation never registered any worrying
levels of strain).
3. The grouting did not appear to have been fully 7 RESULTS OF INSTRUMENTATION PROGRAM
effective at sealing the ‘windows’ around the
WNC; however, there was no trouble with The most significant challenge with the instrumentation
inflow or ground loss around the pipe itself, program was survival of the instruments. The initial
likely due to the groundwater being at about camera installation was damaged by high flows when the
pipe invert level (probably having been sewer surcharged during a storm event. A modified system
lowered due to long-term leakage into the was then developed, with the camera being located on a
pipe). floating platform, which could be removed when the pipe
4. The groundwater level was, however, was expected to surcharge.
encountered at almost exactly the founding Challenges were also experienced with failure of
level of the new manhole and, with the sandy several of the strain gauges. The choice to have two strain
ground conditions encountered at that same gauges at each location, for redundancy, was ultimately a
level, deeper excavation would not have been good decision.
possible without additional measures such as The key outcomes of the instrumentation program were
active dewatering (see Figure 19). as follows:
1. The strains in the pipe roof/crown were never
significant or in excess of the ‘review’ level.
Interestingly, the strains measured for the
underside of the pipe roof were all
compressive, but also quite small.
2. No vibration events occurred in excess of the
vibration limit.
3. The secant shaft wall deflection, as measured
by survey targets on the pile faces, never
exceeded 2 mm.
ACKNOWLEDGEMENT

The instrumentation for this project was supplied and

installed by Subsurface Geotech Inc., whose involvement
was key to the success of this work.

REFERENCES

EWC Designers 2020. Drawing EJC-S2GUID-UTL-DWG-

3139.
EWC Designers 2021. Drawing EJC-S2GUID-UTL-DWG-
5005.
Lightweight Cellular Concrete to Mitigate
Highway Settlement Over Organic Deposits
- Highway 63:08 South of Fort McMurray,
Alberta
Ken Froese & Renato Clementino
Thurber Engineering Ltd., Edmonton, Alberta, Canada
Brad Dolton
CEMATRIX, Calgary, Alberta, Canada
Roger Skirrow
Alberta Transportation and Economic Corridors, Edmonton, Alberta, Canada

ABSTRACT
A section of Highway 63 in northern Alberta constructed in 1975 experienced excessive long-term settlement due to
consolidation of peat and organic layers which had not been removed. The combined thickness of these layers was up to
2.6 m. The area was remediated in 2021 by sub-excavating to place a 500 mm layer of lightweight cellular concrete (LCC)
with a 400 mm clay cap below the pavement structure to reduce embankment loading and bridge the soft subgrade.
Extensometers installed in the remediated zone show a significant reduction in the rate of consolidation of only 15 mm
over two years. Thermistor strings determined that the LCC reduced frost penetration by 1 m and that the thickness of
cover (clay cap and GBC) over the LCC was sufficient to reduce the risk of differential icing.

RÉSUMÉ
Un tronçon de la route 63 construit en 1975 dans le nord de l'Alberta a subi un tassement excessif à long terme en raison
de la consolidation de la tourbe et des couches organiques qui n'avaient pas été retirées. L’épaisseur combinée de ces
couches atteignait jusqu'à 2,6 m. La zone a été corrigée en 2021 par une sous-excavation pour placer une couche de 500
mm de béton cellulaire léger (BLC) ainsi qu’une couche d'argile de 400 mm sous la structure de la chaussée afin de réduire
la charge du remblai et de combler le sol de fondation sensible. Des extensomètres ont été installés dans la zone corrigée
et ont mesuré une réduction significative du taux de consolidation à seulement 15 mm sur deux ans. Des thermocouples
ont aussi démontré que le BLC réduisait de 1 m la profondeur du givre et que l'épaisseur de la couverture (couches d'argile
et de base granulaire) sur le BLC était suffisante pour réduire le risque de givrage différentiel.

1 INTRODUCTION 2.1 History

This paper presents a case study on the use of lightweight There was limited formal documentation of the distress at
cellular concrete (LCC) to bridge soft, organic subgrade this site. However, anecdotal evidence provided by
along a portion of Highway 63, which had experienced highway maintenance staff for this portion of the highway
persistent settlement issues. Based on the stratigraphy and indicated that settlement had been ongoing for many years.
observed distress, a remedial measure was selected and Periodic patching was undertaken to maintain the highway
implemented in conjunction with a pavement overlay grade. The most recent patching was completed in 2015
project. The repaired section and an adjacent untreated (Figure 2), and evidence of fresh settlement was reportedly
section of the highway were instrumented to monitor the observed within 24 hours. Maintenance personnel also
effectiveness of the solution. The thermal profiles below the noted that there were springs in the area. There were
treated and untreated sections were monitored to provide significant difficulties encountered during the excavation of
design information for future projects. wet, organic soils from the subgrade while constructing the
The site is located about 75 km south of Fort McMurray, adjacent north-bound lanes (NBL), resulting in heavy
Alberta, on the south-bound lanes (SBL) of Highway 63, construction equipment becoming mired.
Control Section 8 at km 27.9, as shown in Figure 1. The NBL were constructed in 2014 and at a lower,
Highway 63 is the main highway connecting to Fort grade than the SBL. This can be seen on the highway
McMurray and important industrial facilities in the north and profile shown in Figure 3. Table 1 compares the 2018 SBL
carried about 4,000 vehicles per day in 2018 and 2019 centreline survey with the SBL survey taken in 2014 (during
when the engineering study and design were completed. NBL construction) and indicates that a significant thickness
The SBL were constructed in 1975 and served as the of asphalt was placed between those two surveys.
original Highway 63 prior to twinning. Design began in 2018 for a scheduled overlay of the
SBL and this site was noted as requiring additional
investigation and consideration. Thurber supported the
2 BACKGROUND Prime Consultant, ARA Consulting (now McIntosh Perry)
with investigation, design, construction input, and A site inspection was carried out in May 2018 to review the
instrumentation monitoring. general site conditions and finalize the drilling program.
The highway embankment was about 12.4 m wide, 2 m to
3 m high, and with sideslopes inclined between 3.5H and
4H:1V. Large gravel pits were located on both sides of the
highway. There were access roads to the pits on both sides
of the highway. The ditches appeared relatively well
drained, with culverts flowing from east to west. There was
a subdrain from the NBL full-depth excavation that
daylighted in the median ditch and drained north toward a
culvert. The vegetation in the vicinity was predominantly
mature trees with areas of peat (muskeg) and low willow.
The peat deposits did not appear to be deep and were
relatively dry at the time of the site reconnaissance.
Pavement distress appears to be related to vertical
deformation of the surface, although only a few cracks
were noted. A recent patch was observed over this zone,
as shown on the site plan and in Figure 2. Observations of
south-bound vehicles driving through the area confirmed
that there was some unevenness at this location and at a
small area further south of the patch (in the vicinity of
TH18-8).

Figure 1: Site Location

Table 1: Difference between SBL 2018 and 2014 Surveys

Highway Surface Difference (m)

Relative Location
Kilometer (2018-2014)
28.1 South of patch 0.07
28.0 In patch 0.05
27.9 In patch 0.23
27.8 In patch 0.11 Figure 2: Looking north at patch over distressed area
27.7 North of patch 0.01

Later in May 2018, Thurber undertook a drilling

2.2 Physiological and Geographic Setting program consisting of 15 test holes advanced using solid
stem augers to a maximum depth of 17.5 m. The test holes
The general topography slopes downward from the were drilled through the highway embankment in the
southeast toward the northwest. The bedrock depth is distressed area and adjacent ditches on either side of the
estimated to be greater than 50 m below the ground highway lanes. These locations are shown on the site plan
surface and dips downward toward the southeast in Figure 4. Open PVC standpipe piezometers were
(Pawlowicz and Fenton 1995). Based on surficial geology installed at six locations adjacent to the highway.
mapping, the site is situated predominantly on the Horse
River Till ground moraine (composed of clay till) with a 2.4 Stratigraphy
glacio-fluvial kame deposit toward the south and east on
the higher ground (composed of poorly-sorted outwash The generalized soil profile, in descending order, consisted
sands and gravels) as well as an outwash sand and gravel of the pavement structure (asphalt over granular base
deposit toward the northeast (Bayrock and Reimchen course), sand and clay fill, organic soils, and layers of
1974). native sand and clay till. Table 2 summarizes the thickness
Hydrogeological mapping in this area indicates that of these main layers below the highway. (The thick asphalt
regional flow is downward toward the McMurray Formation, layer at some locations made it difficult to measure the
which discharges into the Clearwater and Athabasca River thickness of the granular base course below.) Test holes
valleys (Ozaray 1974). There were no wells reported in the TH18-1 through TH18-8 were drilled through the pavement
local area, so there is limited information on local flow from north to south, with TH18-2 through -6 located in the
trends, such as the potential for springs and artesian patched portion of the highway. TH18-1 and -7 were
conditions. located within 20 m of the patch and TH18-8 was 155 m
south of the patch.
2.3 Site Investigation
Figure 3: SBL and NBL profiles at the site

Figure 4: Plan of test holes with encountered organic thickness and layout of instrumentation

Alberta Transportation and Economic Corridors (TEC) of cement-stabilized soil, which was encountered in the
records indicate that the nominal asphalt pavement three attempts made to unsuccessfully complete TH18-7
thickness was between 180 mm and 244 mm, which (premature refusal encountered in fill at only 2.3 m deep).
agrees with 170 mm average thickness at the test holes Organic materials (peat and soils with high organic
drilled outside the patched area. The asphalt in the distress content) were encountered at all the test holes except
area was between 180 mm and 630 mm thicker than the TH18-6 and -7, which were in the distressed area, and
average. The historical records from grade widening in TH18-15 located along the east ditch. The thickness of
1995 showed that the existing embankment had 225 mm peat below the highway embankment was between 0.3 m
and 2.1 m with water contents between 55 and Mitigation options consisted of: do-nothing; full-depth peat
375 percent, whereas the test holes in the ditch removal; partial peat removal and reconstruction with
encountered peat up to 4.4 m thick with water contents up geogrid reinforcement; and partial depth excavation and
to 403 percent. At TH18-5 and -8, there was a layer of black reconstruction with lightweight fill.
to grey organic soil containing shells and wood chips with The do-nothing option was to proceed with the overlay
water contents up to 98 percent. In Figure 3, it can be seen with the understanding that ongoing patching would still be
that the dips in the May 2018 SBL centreline profile required. Full sub-excavation would provide the best long-
correspond with the thicker portions of the organic layer. term performance but would be difficult to construct given
The water levels measured in the standpipes (located the depth of soft soil to remove, the high groundwater table,
in the ditch) approximately one month after installation and the soft subgrade below the peat requiring bridging
were between 1.2 m and 2.3 m below ground surface. with geogrid and/or gravel. A partial sub-excavation to
Seepage during drilling was encountered in all test holes at place geogrid reinforcement would even out the areas of
similar depths. settlement but still be subject to long-term consolidation
issues.
Ultimately, LCC was selected as it required less sub-
Table 2: Thickness of Stratigraphic Units excavation, would reduce the overall embankment loading
on the organic soils by up to 35 percent, and could be
Layer Thickness (m), TH18- placed quickly, minimizing traffic delays. Furthermore, the
strength of the LCC was greater than that provided by
Material 1 2 3 4 5 6 7 8
geogrid-reinforced gravel, increasing the layer coefficient
Asphalt 0.16 0.35 0.65 0.55 0.80 0.36 0.15 0.19 from 0.14 to 0.2 as well as providing better bridging over
Granular the soft layer. Expanded Polystyrene (EPS) foam blocks
0.21 0 0.26 0.25 0 0.40 0 0
Base1 were considered as an alternative lightweight fill but are
Fill 1.43 1.94 1.49 1.5 3.31 2.29 2.21 3.61 more difficult to place and have limited flexural strength
Peat 0.3 0.31 2.1 0.3 0.99 0 N/A2 0.7 being composed of distinct blocks. Polyurethane foam
Organic injection was also considered, as this method had been
- - - - 1.60 - N/A 2.3 used successfully on other projects to create columns that
Soil
Total carry the embankment loading to the underlying inorganic
0.30 0.31 2.10 0.30 2.59 - N/A 3.00 soils (Yu et al, 2013 and Smith et al, 2011). However, at
Organics
this site, the peat zone was not consistent in depth or
1
Where distinct from the underlying embankment fill
thickness, making it difficult to target for reinforcement and
2
Shallow refusal in cement-stabilized material the estimated cost was higher than for the LCC approach.
All the excavation options also considered the potential
to lower the vertical profile by up to 1.0 m to better match
3 DESIGN the NBL alignment and to further reduce the embankment
loading. The final vertical profile is shown in Figure 2.
3.1 Assessment Thurber recommended that the partial sub-excavation
and LCC placement option be pursued as it was slightly
At the commencement of the study, there were concerns less expensive than the other remediation options and
that the distress might be related to slope instability; resulted in the greatest reduction in embankment load.
however, given the relatively flat terrain at the site, this was
deemed unlikely. The distress appeared to be limited to a 3.3 Cellular Concrete Defined
portion of the SBL coincident with a reported peat zone not
excavated during original construction. This layer was Cellular concrete, sometimes referred to as foam concrete
removed during construction of the NBL where similar (CROW 2003), is a lightweight construction material
distress has not been observed. Thus, the roadway consisting of Portland Cement or Portland Limestone
settlement observed at the site is likely related to long-term Cement, water, foaming agent, and air. Fly ash and blast
creep (likely due to secondary consolidation) of the organic furnace slag are often added to the mix to customize
deposits exacerbated by the continued placement of material properties. Lightweight cellular concrete (LCC)
asphalt to patch the area over the last 40 years. It is usually contains no sand or aggregate. Fresh cellular
reasonable to assume that primary consolidation was concrete is highly flowable and can be pumped into place
completed within a few years from after completion of over long distances through flexible hoses. In most cases,
original construction. The extent of settlement along the cellular concrete is cast-in-place.
SBL was assumed to correspond with the recent patch, The air bubbles within the concrete create a lightweight,
which was verified by the locations where peat was insulating material. LCC can have wet densities from
observed in the test holes. The thin layer of peat at TH18- 250 to 1600 kg/m3; however, most lightweight fill
1 outside of the patch appears to be fully consolidated as applications are placed at wet densities of 400 to
there was no distress observed at, or further north, of that 600 kg/m3.
test hole.
3.4 Differential Icing Concerns
3.2 Options Considered
Based on previous TEC experience, the use of insulative Construction commenced on April 22, 2021, and the overall
products immediately below a pavement structure can project was completed on June 21, 2021
cause differential icing which is a significant hazard to
traffic. This is due to the different thermal regime in the 4.2 LCC Placement
insulated section. This concern can be partly mitigated by
using a soil cap over the insulation or deeper burial to The LCC production unit is shown in Figure 4. Cement
provide a heat reservoir (“heat sink”). was delivered using bulker trailers to the silos located at
According to Côté and Konrad (2006), the risk of the rear of the cellular concrete pump unit. Water was
highway surface icing occurs when the temperature of the delivered to the site and stored in a flexible water storage
pavement surface is colder than the frost point (which is the bladder.
sub-zero equivalent of the dew point so the temperature at During placement, LCC is highly flowable and is nearly
which moisture will condense as frost on an exposed self-leveling. The maximum slope that may be held with
surface). Once the frost point temperature drops below the LCC at a density of 475 kg/m3 is approximately 1%,
surface temperature of the pavement (which warms provided the lift thickness is 250 mm or less. Since the
somewhat during the day from radiant energy), icing is no vertical profile of the roadway was an average of
longer a concern. They define the phenomenon of approximately 3.7%, wood formwork lined with
differential icing as “the movement of moisture from the air polyethylene sheets were placed transversely across the
to the colder surfaces of these insulated sections, creating excavation at approximately 16 m spacing. Formwork was
either hoarfrost (or white frost) or black ice.” Differential also used to contain the LCC along the roadway shoulder.
icing can occur when the surface temperature of the A view of the formwork and freshly-placed LCC after
insulated pavement is lower than both the surface pouring the first half is shown in Figure 5. Once the first lift
temperature of the conventional pavement and the frost was in place, the transverse formwork was removed, and
point. the areas between the “steps” of LCC were filled in. Filling
Once the soil overlying the insulation has completely in the “steps” required two pours over two consecutive days
frozen, the insulated pavement surface will cool faster than to complete.
the uninsulated pavement, which still has upward flow of The first and second halves of the roadway were
latent heat. Thus, they concluded that a thicker layer of completed May 6 to 8 and May 27 to 29, 2021, respectively.
cover (that is, deeper burial of the insulation) will provide The dry mix unit can produce up to 100 m3/hour; daily pour
longer protection against differential icing. When modeling volumes were limited due to the LCC geometry. Daily
Quebec climate conditions they determined that between production ranged from 73 to 689 m3/day, with an average
October and November (typically) both conventional and of 300 m3/day. A total of 1,800 m3 of LCC were placed over
insulated pavements were at risk of icing; however, once six pour days.
the freezing front reaches the insulation, there also arose The project specified an LCC wet (as-cast) density of
a risk of differential icing which continued later into the 475 kg/m3 (+/- 10%) with a minimum compressive strength
winter. This occurs because the insulated pavement of 0.4 MPa at 28 days. Quality Control (QC) was conducted
structure is colder than the conventional pavement, thus continuously throughout the operation to measure LCC
remaining below the frost point temperature later into the properties. Wet densities ranged from 440 to 490 kg/m3,
season. (Côté and Konrad 2006). with an average of 460 kg/m3. Cylinders were cast to verify
To mitigate the potential for differential icing on this compressive strength as per ASTM C495, Standard Test
project, a nominal 300 mm clay cap between the pavement Method for Compressive Strength of Lightweight Insulating
structure and the LCC was used. This represented a Concrete. Based on QC data, the average LCC 28-day
compromise between the 0.3 m of gravel recommended by compressive strength was 1.5 MPa. Minimum and
the LCC supplier and the 0.6 m of clay recommended by maximum compressive strengths were 1.1 and 1.9 MPa,
TEC. Clay was selected in preference to gravel due to its respectively.
higher thermal capacity, which should provide better heat
retention and thus reduce the potential for differential icing.
The final embankment design consisted of 240 mm
asphalt, 400 mm of GBC, the nominal 300 mm clay cap,
and 500 mm of LCC. The clay cap thickness tapered from
400 mm at centreline to 200 mm at the shoulder to provide
cross-drainage below the GBC.

4 CONSTRUCTION

4.1 Overall Project Schedule

The LCC installation was part of the larger pavement

overlay project which was completed during the summer of
2021. One-lane highway traffic was maintained on the SBL
by reconstructing one half of the embankment at a time.
Figure 4: Lightweight Cellular Concrete Production Unit
(taken by CEMATRIX)
 The instrumentation installation was a combined effort
by the Contractor, their electrical subcontractor, and
Thurber. Holes were drilled near the centre of the SBLs
from the top of the GBC layer using a conventional
geotechnical auger drilling rig. The holes were grouted
after the instruments were installed. The instrumentation
cables were routed in conduit (a minimum of 75 mm deep
to protect from high temperatures during asphalt
placement) to the edge of the highway embankment and
then in combined larger-diameter PVC conduits (with a
minimum cover of 300 mm to protect against equipment
damage) to the datalogger enclosure on the far side of the
west ditch. The instrumentation, conduits, and enclosure
were completed on May 17, 2021, and Thurber finished the
wiring of the multiplexers and datalogger on June 6, 2021.
The completed enclosure is shown in Figure 6.

Figure 5: Finished surface of the first half of LCC where

colour bands show the pour stages (photo taken May 8,
2021, taken by CEMATRIX)

Typically, backfill material may be placed and

compacted on top of the LCC within 1 or 2 days after
placement, depending on weather conditions. Construction
traffic is not allowed to drive directly on the LCC without the
first lift of backfill in place. A small bulldozer or skid steer
loader is commonly used to spread out the backfill material.
A smooth drum roller with static (no vibration) compaction
is required to compact the first 150 mm lift of backfill.
Vibratory compaction is permitted during placement of the
subsequent lifts.
For this project, the material placed immediately over
the LCC was low plastic clay (Standard Proctor Maximum
Dry Density, SPMDD, of 1927 kg/m3 at 12.6% optimum Figure 6: Datalogger enclosure and weather station in the
water content). The clay was placed according to TEC west ditch of Highway 63.
specifications (95% of SPMDD) without vibration. The
water content was close to the optimum and density
requirements were confirmed by nuclear densometer Local climatic data is collected by a ClimaVUE50
testing. Approximately 1,270 m3 of clay was placed to build weather station installed at the datalogger enclosure, which
the LCC cap. measures temperature, humidity, barometric pressure,
wind speed and direction, rainfall, and solar radiation.
4.3 Instrumentation Installation Snow, ice, and hail are not measured by this type of
weather station. The weather station, instrumentation,
Instrumentation was installed to assess both settlement Campbell Scientific CR6 datalogger, and Campbell
and thermal regime to better understand the performance Scientific AM16/32B multiplexers are powered by two 12V
of LCC for future projects. 100Ahr batteries and a vertically-inclined100W solar panel.
Two Geokon Model 1150 (A-3) vibrating wire After the instrumentation was installed, the Contractor
extensometers were installed in the treated section. The final-graded the surface of the GBC and placed the asphalt
extensometers were nominally 6 m in length with pavement. The final surface was chip sealed to increase
monitoring points installed targeting the base of the LCC, pavement roughness to counter the potential differential
the base of the peat, and the underlying native clay till. icing hazard.
Four Geokon Model 3810 thermistor strings were The datalogger is downloaded twice per year as part of
installed: two in the treatment zone (THERM21-2 and -3), TEC’s Geohazard Risk Management Program.
one as a Pavement Control south of the treatment zone
(THERM21-1), and one as a Grassed Control in the west
ditch (THERM21-4) adjacent to the datalogger enclosure 5 ANALYSIS
pedestal. The thermistor strings were 6 m in length with six
monitoring points (one metre spacing). The cable lengths 5.1 Settlement
were adjusted in the field such that installation depths were
between 3.3 m and 5.3 m. Tighter spacing was used in two The depths of the extensometer monitoring points are
of the installations to refine the measurement of the thermal summarized in Table 3. The south extensometer, EXT21-
gradient in the upper 3 m.
1, is about 2.4 m higher in elevation than the north still well within tolerances for roadway safety and
extensometer, EXT21-2. Given the constraints of the rod maintenance. These rates since remediation compare
lengths, EXT21-1 could not be installed into the underlying favourably with the maximum settlement rate of almost
inorganic soils, whereas EXT21-2 had two points within 60 mm per year measured between 2014 and 2018
that unit. The extensometers were installed on May 18, (Table 1).
2021, and the datalogger became operational on June 6, The settlement points near the base of the LCC
2021, so the plots start with small unrecorded measure the change that occurs over the GBC, clay cap,
deformations. The instruments were installed near the and LCC. It can be assumed that the LCC is volumetrically-
centreline of the SBL where the clay cap thickness was stable and GBC is relatively insensitive (being well
approximately 400 mm. compacted and free-draining); thus, the heaving and
For both extensometers, the lower probe is considered settling observed is likely occurring within the clay cap.
as the stable reference point and the settlement measured There appears to have been some initial consolidation of
at that depth is, therefore, the total settlement at that the clay between 1.5 mm and 2.5 mm, which is about
location and is comprised of the settlements that occurred 0.5 percent of the 400 mm-thick lift. The heave amplitude
in the pavement structure (including the LCC and clay cap), of 3 mm is not of concern from a highway operation
which occurred in the underlying fill and organic soils, and perspective. The low plastic clay was placed at optimum
the minor amount that occurred within the till itself. The moisture content, so volumetric change is expected to
data are summarized in Table 4 and shown in Figure 7 remain within this minor change measured so far.
(only south extensometer data is shown). This heave occurs over a consistent time span. At
EXT21-1, heave was observed between October 21, 2021,
and April 28, 2022, in the first winter and started on
Table 3: Settlement Measurement Depths November 7, 2022, in the second winter and had not
dissipated when the readings were taken April 21, 2023
EXT21-1 (SOUTH) EXT21-2 (NORTH) (less than 0.5 mm of heave remained). The other
Depth Depth
extensometer, EXT21-2, showed the same pattern, with
Soil Unit Soil Unit the heave occurring between November 24, 2021, and
(m) (m)
April 24, 2022, in the first winter and between December 1,
0 Top of GBC 0 Top of GBC
2022, and April 12, 2023, in the second winter. It is likely
1.33 Near base of LCC 1.26 Near base of LCC that this heave is the result of ice lens formation during the
4.83
Organic Silt (just
4.76
Clay Till (just below winter and then dissipation during the spring.
below peat) peat)
Organic Soil (just
5.83 above Native Clay 5.76 Clay Till Table 4: Measured Settlement (June 2021 to April 2023)
Till)
EXT21-1 (SOUTH) EXT21-2 (NORTH)
Thickness of 2.59 m 2.10 m
Organic Soils (Peat and OG Silt) (Peat)
Maximum Total
3.4 mm 15.2 mm
Settlement
Current Total
2.1 mm 13.9 mm
Settlement
Current Peat
1.8 mm 11.6 mm
Settlement
Pavement 1.8 mm to 3.5 mm to
Structure
Variability1 -3.4 mm -3.04 mm
1
Incorporates GBC, clay cap, and LCC; negative values indicated
heave

5.2 Climatic
Figure 7: South Extensometer Measurements
The weather station measures several parameters that
could be considered in assessing the thermal performance
of the pavement. For the purposes of this paper, air
Trends have developed at both locations and indicate
temperature (Table 5) was used as the key parameter for
that settlement has been reduced compared to the
understanding local climatic conditions.
presumed pre-construction rates. At the south end, the
overall settlement is relatively inconsequential (about
3 mm) and generally trends downward but is dominated by
heaving in the upper layers. At the north end, the maximum
settlement was 15 mm which is greater than the south but
Table 5: Late Fall, Winter, and Early Spring Air per Côté and Konrad 2006). Against the right-hand axis is
Temperature Comparison the differential between the insulated and uninsulated
pavement temperatures. From Figure 8, it can be seen that
Winter 2021/2022 Winter 2022/2023 there are a few short periods in October and November
where the near-surface temperature measured in the
Max Min Avg. Max Min Avg.
insulated pavement is colder than the uninsulated
Sep. 22.0 1.1 10.7 32.4 2.3 13.0 (negative differential), which is when Côté and Konrad
Oct. 14.3 -7.1 3.9 21.1 -3.4 6.8 (2006) observed the highest risk in their modeling for
Nov. 12.5 -22.4 -3.9 8.9 -28.7 -8.0 Quebec climates; however, the largest and sustained
Dec. 0.8 -34.0 -19.6 -2.3 -34.8 -18.3 difference is seen in December and January when the
Jan. 2.8 -35.5 -15.8 3.3 -27.4 -8.3
insulated pavement can be 4°C to 5°C colder than the
uninsulated. However, that period is also when the risk is
Feb. 6.9 -30.3 -15.1 5.1 -32.3 -12.2 typically low for icing as the frost point will be lower than
Mar. 15.0 -25.0 -5.3 8.4 -27.1 -8.5 the pavement surface temperature.
Apr.1 12.0 -19.7 -0.4 14.6 -10.0 2.1
Freezing
1943 17051 6 CONCLUSION
Index
Note: All values are in Celsius except Freezing Index which is in
°C-Days as in the sum of the daily mean degree-days where the The purpose of incorporating LCC into this settlement
air temperature was below 0°C. repair was three-fold. First, to simplify construction by
1 – Data was truncated on April 20. placing a flowable product over the soft subgrade rather
than attempting to bridge over it with fill and geotextiles.
Secondly, to contribute to the overall reduction of the
5.3 Thermal embankment load. Third, to provide bridging of the
overlying pavement structure and live loads from traffic
The data from the thermistors showed the expected over the soft subgrade. The bridging success is difficult to
pattern: temperature moderated with depth becoming assess; however, construction was straightforward and the
near-constant at 5.3 m below GBC, shallow temperatures rate of settlement has been significantly reduced. Falling
measured in the paved areas were significantly higher in weight deflectometer testing may be used in the future to
the summer and colder in the winter compared to the provide an indication of structural improvements.
grassed area, and insulative benefits of the LCC were A clay cap was introduced to reduce potential for
clear. One of the two thermistors in the LCC-insulated differential icing as the greater thermal capacity of clay
section (THERM21-3) was not used due to malfunction. would be the equivalent of a thicker layer of gravel, thus
The maximum temperatures measured at each node of reducing both the amount of material required and the
the remaining three thermistor strings are shown in Table 6 depth of subexcavation. Based on limited assessment of
and grouped by season. Table 6 also includes the the thermistor data, it appears that the near-surface
estimated maximum depth of frost penetration at each temperatures measured in the insulated section were
location. somewhat colder in the winter than at the uninsulated
The depth of frost penetration was shallower for the section. The largest difference occurred in mid- to late-
LCC section by about 1.0 m. The LCC is located at 1.04 m winter when the risk of differential icing is already reduced
below the roadway surface and Table 6 shows that the at the entire pavement surface. It appears that the risk of
estimated date on which the frost penetrated to this depth differential icing is relatively low; therefore, the thickness of
was essentially the same (only two days different) for both the clay cap was sufficient.
insulated and conventional pavements (within the likely As far as the LCC supplier is aware, this is the first
error of using linear interpolation between thermistor nodes project utilizing a clay cap instead of gravel. There were
at the conventional pavement). Côté and Konrad (2006) difficulties encountered with placing the clay. During
suggest that the risk of differential icing period starts when placement, the clay tended to slide on the smooth LCC
the frost penetrates to the insulation depth. At this site, the surface making compaction difficult. It made more difficult
frost penetrated to that depth below both pavement types as the Contractor could not use a sheepsfoot roller due to
at the same time, which may indicate a lower risk of the potential higher contact pressures damaging the LCC
differential icing. surface. For future projects using a clay cap, consideration
The other factor to consider is when the surface may be given to placing a thin lift of gravel (150 mm
temperature of the insulated section is colder than suggested) between the clay and the LCC to reduce the
uninsulated and colder than the frost point. The difference potential for the clay to slip during compaction.
between the pavement surface temperatures was The moisture conditioning required to achieve
assumed to be represented by the difference in the 0.1 m compaction meant the surface of the clay cap was
below GBC thermistor nodes. somewhat soft when proof-rolled and the team elected to
Figure 8 plots the air temperature and frost point use a woven geotextile to provide additional reinforcement
temperature (calculated using Côté and Konrad 2002) at the base of the GBC.
against the 0.1 m node temperatures for both winter
periods (all temperatures are the 6:00 AM daily reading as
Table 6: Seasonal Maximum Temperatures Measured at Each Thermistor Node

Summer 2021 Summer 2022

Depth (m) (T1, T2, T4)1 T1 Pvt Control2 T2 LCC T4 Grass2 T1 Pvt Control2 T2 LCC T4 Grass2
0.1, 0.1, 0.3 28.7 29.8 17.3 25.3 26.4 15.6
1.1, 1.3, 1.3 20.5 10.3 9.9 20.2 11.3 9.4
2.1, 2.3, 2.3 14.1 8.9 7.0 14.2 9.6 7.3
3.1, 3.3, 3.3 11.1 7.7 5.9 11.4 8.2 5.9
4.1, --, 4.3 9.6 -- 5.3 9.5 -- 5.5
5.1, --, 5.3 8.5 -- 5.0 8.3 -- 5.2
Winter 2021/2022 Winter 2022/2023
Depth (m) (T1, T2, T4) T1 Pvt Control2 T2 LCC T4 Grass2 T1 Pvt Control2 T2 LCC T4 Grass2
0.1, 0.1, 0.3 -25.1 -28 -0.1 -16.7 -17.3 0.1
1.1, 1.3, 1.3 -17.8 -0.8 2.0 -9.9 -0.4 2.6
2.1, 2.3, 2.3 -1.5 0.8 3.5 -0.7 1.1 4.6
3.1, 3.3, 3.3 0.5 1.7 4.3 0.8 2.0 5.0
4.1, --, 4.3 1.3 -- 4.7 1.3 -- 5.4
5.1, --, 5.3 2.1 -- 4.6 2.4 -- 5.0
Est. Frost Depth 2.85 1.8 0.35 2.58 1.58 N/A
1 – T1 refers to thermistor THERM21-1 and so forth. Thermistor depths are listed in order of T1, T2, T4. Thermistor T2 was installed
shallower to place more nodes in the upper portion of the pavement structure.
2 – T1 was installed south of the treatment as a pavement control and T4 was installed in the adjacent ditch as a grassed control.

Figure 8: Winter 2022/2023 Air, Frost Point, and Near-Surface Thermistor Temperatures

7 FUTURE STUDY station at the site and can be used to calculate the frost
point temperature (methods provided by Côté and Konrad
The key parameters to measure in the field to determine 2002). The depth of frost penetration can be interpolated
the potential for differential icing (according to Cote and from the thermistor node data. These parameters need to
Konrad 2006) are the air temperature, humidity, surface be combined with the surface temperature of the asphalt,
temperature, and depth of frost penetration. Air which is difficult to measure on an active highway. The next
temperature and humidity are measured by the weather step would be to model the thermal regime of both
insulated and uninsulated pavement sections to back- Smith, J. Gräpel, C., Skirrow, R., Oad, S., Reynolds, N.,
calculate the surface temperatures, which can then be and Moroschan, C. 2011. Foam column injection to a
compared with the frost point and depth of frost penetration highway embankment foundation – results of a full
to determine when icing could have occurred. There are scale pilot study: The 2011 Annual Conference of the
additional parameters measured by the weather station Transportation Association of Canada, Edmonton,
that could be included in the modeling, such as wind speed Alberta.
and direction and solar radiation.
The continuing settlement trends will be directly
assessed via the extensometers and generally assessed
over the next several years through TEC Construction and
Maintenance operations. If significant distortions are noted,
direct survey measurements will be undertaken in support
of overlay designs. Overlays would be the simplest method
of addressing continued settlement if it occurs. The first two
years of data indicate the frequency of such patching
should be significantly reduced because of the remedial
measures implemented at this site.

8 ACKNOWLEDGEMENTS

The authors wish to thank and acknowledge ARA

Engineering Consulting Engineers (now McIntosh Perry),
Prime Consultant for this project, for the opportunity to
assist in the engineering design and construction as well
as for the provision of survey, construction information, and
drawings incorporated into this paper. We also thank and
acknowledge Alberta Transportation and Economic
Corridors for providing technical input during design,
historical information utilized in preparing this paper, and
for their permission to publish. We specifically
acknowledge Roger Skirrow, Director of Geotechnical
Services, for his review of this paper. We are grateful to
CEMATRIX for their commitment to the research aspects
of this project, including the generous supply of the
instrumentation installed at the site.

9 REFERENCES

Côté J. and Konrad, J-M. 2006 Granular Protection Design

to Minimize Differential Icing on Insulated Pavements,
Canadian Geotechnical Journal, 43: 260-272.
Côté J. and Konrad, J-M. 2002. A Field Study of Hoarfrost
Formation on Insulated Pavements, Canadian
Geotechnical Journal, 39: 547-560.
CROW 2003. Roads and car parks on foam concrete.
CROW, Ede, Gelderland, The Netherlands.
Bayrock, L.A., and Reimchen, T.H.F. 1974. Surficial
Geology – Waterways (NTS 74D). Alberta Research
Council Map 148 (scale 1:250,000).
Ozaray, G.F. 1974. Hydrogeology of the Waterways -
Winefred Lake Area, Alberta. Alberta Research Council
Report 74-2 and Map 106 (NTS 74D and 73M,
scale 1:500,000).
Pawlowicz, J.G. and Fenton, M.M. 1995. Drift Thickness of
Alberta. Alberta Geological Survey May 227 (scale
1:2,000,000).
Yu, L., Wang, R., Skirrow, R. 2013. The application of
polyurethane grout in roadway settlement issues. Geo
Montreal 2013, Canadian Geotechnical Society,
Montreal, Quebec, Canada.
Drilled Shaft Installation Under Polymer
Slurry: A Case Study in Barrie, Ontario
Mark Henderson, Lisa Coyne, WSP Canada Inc.
Rui Arco, Luis Pardo, Aecon Foundations
Tony Sangiuliano, Andrew DeSira, Ministry of Transportation of
Ontario

ABSTRACT
Drilled shaft foundations for Ministry of Transportation Ontario (MTO) projects have traditionally been installed using casing
to ensure sidewall stability during construction. However, site access, subsurface conditions, and caisson sizes can result
in construction challenges related to laydown area, equipment size demand and installation efficiency. At the Highway
400/Essa Road Overpass Replacement Project in Barrie, Ontario, 20 m long, 1.5 m diameter drilled shafts have been
installed in silty sands with a high groundwater table. The use of full-depth temporary casing was considered but was
found to be impractical due to the extremely limited working footprint in the roadway median adjacent to live traffic lanes
and existing infrastructure, and the risk of basal instability. Hence, a polymer slurry drilling system with short surficial
casing was selected by the Contractor as the preferred method for stabilizing the shafts in these challenging geotechnical
and site conditions. Polymer slurry testing, base cleaning, concrete mix design and testing, and pile integrity testing were
completed to support the construction and quality control.

RÉSUMÉ
Les fondations des puits forés pour les projets du Ministère des Transports de l'Ontario (MTO) ont traditionnellement été
installées à l'aide d'un tubage pour assurer la stabilité des parois latérales pendant la construction. Cependant, l'accès
au site, les conditions du sous-sol et la taille des caissons peuvent entraîner des difficultés de construction liées à la zone
de dépôt, à la taille de l'équipement requis et à l'efficacité de l'installation. Dans le cadre du projet de remplacement du
passage supérieur de l'autoroute 400/Essa Road à Barrie, en Ontario, des puits forés de 20 m de long et de 1,5 m de
diamètre ont été installés dans des sables limoneux avec une nappe phréatique élevée. L'utilisation d'un tubage
temporaire à pleine profondeur a été envisagée mais s'est avérée peu pratique en raison de l'empreinte de travail
extrêmement limitée dans le terre-plein central de la route, à proximité des voies de circulation et de l'infrastructure
existante, et du risque d'instabilité basale. L'entrepreneur a donc choisi un système de forage à boue polymère avec un
court tubage superficiel comme méthode privilégiée pour stabiliser les puits dans ces conditions géotechniques et de site
difficiles. Les essais de boue polymère, le nettoyage de la base, la conception et les essais du mélange de béton et les
essais d'intégrité des pieux ont été réalisés pour soutenir la construction et le contrôle de la qualité.

1 INTRODUCTION High traffic volumes on Essa Road required that two

lanes (one in each direction) be maintained during
The Highway 400/Essa Road Overpass Replacement construction. With the limited cross-sectional width on
Project in Barrie, Ontario involves the replacement of the Essa Road, the contractor’s work area needed to be
existing 50 m wide, six-lane, single-span overpass minimized. To decrease the contractor’s work area and
structure with an 80 m wide, ten-lane, two-span structure. reduce excavation work by avoiding roadway protection,
The new abutments were designed to be supported on caissons were selected as the preferred foundation option
driven steel H-piles, whereas the centre pier was designed to support the centre pier. The pier foundations consisted
to be supported on caisson foundations. This paper of a single row of 22 caissons (two piers comprised of 11
presents the design and construction of the caissons for caissons each for the northbound and southbound
the centre pier based on the foundation investigation, with structures).
a particular focus on polymer slurry drilling to facilitate
caisson installation in challenging subsurface conditions
and the quality control measures used to confirm the
effectiveness of this method.

2 SITE CONDITIONS AND CONSTRAINTS

The existing Essa Road Overpass is located on Highway

400 about 3 kilometres from the Mapleview Drive and
Dunlop Street interchanges in Barrie, Ontario. The
replacement structure accommodates six lanes of Hwy 400
traffic for the interim configuration, with future widening to
ten lanes for the ultimate configuration.
 effective stress and SPT methods for assessing axial
resistance were considered:
▪ Effective Stress (σ‘v ) Method using combined shaft
resistance coefficient, β (as outlined in Canadian
Foundation Engineering Manual, CFEM)
▪ Effective Stress Method (σ‘v) Method using
coefficient of lateral earth pressure, Ks (as partially
outlined in CFEM)
▪ Effective Stress Method (‘σ‘v) Method by Poulos
and Davis (1980)
▪ SPT Method by Meyerhof (1976)
▪ SPT Method by Decourt (1995)

Limiting values of ultimate shaft and base resistance

within the non-cohesive deposits were applied to the
results of the static analysis methods according to
American Petroleum Institute (API), 1984. Engineering
Photograph 1. Constrained work area for drilled shaft judgment based on previous experience in similar
construction at centre pier conditions was also applied to the range of results to select
an appropriate factored ultimate geotechnical resistance
value for design. The soil strata thicknesses and material
3 SITE INVESTIGATION AND SUBSURFACE parameters applied within these analyses are summarized
CONDITIONS in Table 1.
Two boreholes and one Dynamic Cone Penetration Test
(DCPT), advanced to depths of about 20 to 30 metres, Table 1. Soil strata thicknesses and material parameters
were carried out in 2016 as part of preliminary design. The applied for caisson design analyses
foundations field investigation included two boreholes
advanced during preliminary design and twelve boreholes SPT N Model Parameters1
advanced near the foundation elements during detailed Soil Strata
Range γ φ β Kb Nt,q,b
design. Two of the twelve boreholes were advanced to
Loose to very
approximately 40 metres depth to support foundation
dense silty
design recommendations for the piers. Standard sand to sand 8-64 20 32-33 0.4 - -
Penetration Testing (SPT) was carried out in all boreholes (14.5 m to
and five standpipe piezometers were installed to monitor 17 m thick)
the stabilized groundwater level at the site.
In general, the subsurface conditions at the Essa Road Dense to very
Overpass consist of fill and surficial deposits underlain by dense silty 50-75 (Nt)
sand to silt 37-89 21 35 0.4 140 76 (Nq)
an extensive non-cohesive deposit of sand to silty sand (1 m to 3.5 m 64 (Nb)
approximately 30 to 35 metres in thickness, which has a thick)
variable state of compactness ranging from loose to very 1
where γ is the unit weight (kN/m3), φ is the friction angle (o), β is the
dense. This deposit is interlayered with various fine- combined shaft resistance coefficient, Kb is the base factor from
Decourt’s, method, and Nt, Nq, and Nb are bearing capacity factors
grained deposits and underlain by very dense or hard
deposits of silts, sands, and clays, which extend to at least
40 to 45 metres below ground surface.
The groundwater level is generally at about 3 m below Using the methods and model parameters described
the natural ground surface. Pressurized groundwater above, in conjunction with the results of the foundation
conditions (confined by discontinuous clayey sand investigation, the factored ultimate geotechnical resistance
interlayers) are present within the lower portion of the sand (f-ULS) values for various caisson diameters shown in
deposit. Table 2 were developed for design.

4 CAISSON DESIGN Table 2. Factored ultimate geotechnical resistance values

for caisson design
The typical approach for predicting deep foundation
capacity on MTO projects is to use static formulae; in situ Caisson Anticipated Caisson
static load tests may be completed, particularly where f-ULS1 (kN)
Length Founding Stratum Diameter
significant numbers and lengths of deep foundation
Dense silty sand 0.9 m 1,200
elements are required. The design factored ultimate
geotechnical resistance for the centre pier caissons was 18 m and gravel to silty 1.2 m 2,000
sand 1.5 m 2,700
assessed using static methods of analyses based on
1
fundamental soil strength parameters. The following five The factored serviceability geotechnical resistance for 25 mm of
settlement was greater than the factored ultimate geotechnical
resistance.
 6 DRILLED SHAFT INSTALLATION
The factored geotechnical resistance values were
calculated on both shaft and tip resistances; appropriate 6.1 Equipment Selection, Logistics, Size and Capacity
quality assurance procedures were required during
construction to minimize both side wall and base Based on the subsurface conditions at the site,
disturbance. conventional drilling and installation techniques for drilled
Figure 1 shows the various caisson components and shafts would likely utilize a full-length temporary casing
geometry that were designed for the project. with the shaft advanced under a positive head of pressure
using water or drilling fluid. However, constraints
associated with the site working footprint made these
conventional methods impractical. Therefore, to optimize
the equipment footprint and provide adequate shaft stability
measures, a polymer slurry drilling method was selected by
the contractor. Adoption of this method reduced the drill
equipment size and had the advantage of increasing
productivity, as it eliminated the need to extract long casing
strings. Moreover, polymer slurry could be re-used
throughout the project and easily disposed upon
completion of the works.

6.2 Overall Construction Procedure and Sequence

The overall construction procedure and sequence for each

drilled shaft under polymer slurry was as follows:

1) A pilot hole was augered into the ground using a 6-

metre-long, 1,700 mm diameter single-wall temporary
casing, as captured in Photograph 2; this temporary casing
was left protruding approximately 1.5 metres above grade.

Figure 1. Typical pier and caisson foundation details

5 CONTRACT SPECIFICATIONS

A special provision was incorporated into the contract

tender documents to specify caisson requirements,
construction materials, and construction methods including
means of controlling the shaft excavation, continuity of
operations, reinforcement steel placement, placement of
concrete, and construction tolerances.
Key specifications for the caisson installations included
the selection of equipment and installation method to
ensure adequate advancement into compact to very dense
native soils, and to ensure stability of the side walls and
base of the shaft excavation, given the high groundwater
level and presence of pressurized groundwater conditions
at the site. Further, monitoring of drilling fluid properties
were specified to ensure functionality during shaft
advancement, as well as shaft base cleaning given that the
caisson capacities had a significant end-bearing
component. Non-destructive quality-control testing
consisting of Pile Integrity Testing (PIT) and Crosshole Photograph 2. Advancement of temporary casing prior
Sonic Logging (CSL) were specified to assess the to drilling with polymer slurry
structural integrity of the caissons post-construction.

2) The polymer slurry drilling fluid was mixed in a slurry

storage and fabrication plant and tested by an experienced
polymer slurry technician prior to the commencement of the
drilling operation. The slurry plant used for the project is
captured in Photograph 3. 4) A drilling bucket with a cleaning edge was used once the
excavation approached the design tip elevation. At the
design tip elevation, the shaft base was inspected using a
Shaft Quantitative Inspection Device (SQUID), and a
supplemental slurry test was carried out, as captured in
Photograph 5, to confirm that the polymer slurry properties
stayed within acceptable parameters.

Photograph 3. Polymer slurry storage and fabrication

plant

3) The polymer slurry was pumped into the pilot hole and
drilling commenced using a 1,500 mm diameter double-cut
drilling bucket equipped with a fluid bypass aperture. The
excavation was stabilized using the polymer slurry and
standard fluctuations of the slurry level were controlled
using the temporary casing, which also served as a
protection measure for crew members working near the
shaft. An 83,000-litre tank was used for storage of the
polymer slurry. Stoppages did not exceed 48 hours; during
stoppages, slurry properties and level were monitored and
maintained as required.
Photograph 5. Slurry testing after advancement of shaft
to the design tip elevation.

5) Upon approval of the shaft base by the Foundation

Engineering Specialist as part of the Construction Contract
Administration tam, the installation of the steel reinforcing
cage proceeded, followed by concrete placement using
tremie techniques in accordance with the Special Provision
for the contract. During concrete placement, the polymer
slurry was recovered and reused or disposed as per the
project specifications.

A caisson installation report was prepared for each

drilled shaft to document verticality checks, caisson toe and
cut-off elevations, the top of steel reinforcing cage
elevation, details of obstructions or unusual geotechnical
conditions encountered, concrete testing results, and slurry
testing results.

6.3 Polymer Slurry Mix Design, Testing and

Management

Table 3, below, provides the polymer slurry specifications

for the project.
Photograph 4. Excavation of drilled shaft with a slurry
level near the ground surface.
 Freshly 6.4 Steel Reinforcing Cage Installation and Concrete
Slurry Prior to Placement
Made/ Test Method
Property Concrete
Regenerated
≤ 1.01 The caisson reinforcing cage was installed after approval
Mud Weight
(freshly Density
of the caisson base by the Foundation Engineering
Specific made) Balance Specialist. Steel reinforcing cages were preassembled and
Gravity ≤ 1.04 stored in the laydown area located across the Essa Road
≤ 1.01 (API 13B-1,
ASTM D4380 - westbound traffic lane. As the limited working footprint did
(freshly not allow for the staging of the cages next to the caisson
regenerated) 84)
construction activities, a support crane positioned in the
laydown area was used to hoist and lower the cage into the
Marsh Funnel excavation in single lengths.
and Cup The fresh concrete properties and workability retention
Viscosity (API 13B-1, periods that were established with the concrete supplier
75 - 145 65 - 150
(sec/quart) ASTM were as follows:
D6910/D6910M
-09) ▪ Slump Flow (ASTM C1611) greater than 400
millimetres at the end of tremie application (3
pH, if 6.0 – hours);
7.0 – 11.5
contaminated 11.5 PH paper Strips ▪ Inverted Cone Outflow Test (EFFC/DFI Guide
pH, in (ASTM D4972- to Tremie Concrete for Deep Foundations)
standard 7.0 – 9.5 7.0 – 9.5 01)
less than 7 seconds at the end of tremie
conditions application (3 hours);
▪ Slump (ASTM C143) greater than 180 mm at
Sand Content
the final extraction of the temporary casing (4
Sand Content1 (API 13B-1, hours); and
≤ 0.5% ≤ 1.0%
(% by volume) ASTM D4381-
06) ▪ Bleed Rate (ASTM C232) less than 0.1
1
to be measured immediately prior to concrete placement. millilitres per minute during final extraction of
the temporary casing (4 hours)
Table 3. Polymer slurry requirements The concrete mix design specifications were as follows:
▪ Minimum compressive strength: 30 MPa at 28
days;
Slurry samples were obtained after mixing freshly
hydrated slurry, after recycling of used slurry, at the ▪ 19 millimetre maximum aggregate size;
beginning of each work shift, every four hours during the ▪ 3.5% to 6.5% air entrainment; and
drilling phase (or every two hours during the drilling phase
if the previous sample did not have consistent viscosity and ▪ Maximum 25% slag.
pH properties), prior to cleaning the bottom of the shaft, The concrete was placed through the slurry using a
and prior to concrete placement. 250-millimetre-diameter tremie pipe and hopper, as
Samples tested after mixing freshly hydrated slurry captured in Photograph 6. A separating inflatable plug was
were retrieved directly from the middle portion of the utilized to prevent mixing of concrete with any fluid in the
storage tank. Samples tested after recycling of used slurry tremie pipe. A minimum of 5 metres of the shaft length was
or at the beginning of each work shift were retrieved from placed prior to the first split of the tremie. During
the middle of the shaft. Samples tested prior to cleaning subsequent placement of concrete, the tremie pipe was
the bottom of the shaft and prior to concrete placement progressively withdrawn as the concrete level rose in the
were retrieved from within 0.6 metres from the bottom of excavation. The tremie pipe was immersed between 3
the shaft. metres and 8 metres into the concrete at all times.
The polymer slurry contained an additive that promoted
the agglomeration of the soil particles held in suspension
throughout the slurry column. These particles were
continuously removed during advancement of the shaft and
recycling of the slurry, keeping the slurry properties
consistent and eliminating the risk of excessive sediment
to accumulate at the base of the shaft. Adjustments to the
polymer slurry dosage were made based on changes in soil
conditions (particularly while drilling through fine-grained
soils that increased the specific weight of the slurry), and
fluid loss (while drilling through coarse-grained sands and
gravels); however, such adjustments were rarely required
throughout the duration of the project.
 tip elevation, the shaft base was inspected using a SQUID
unit, as required in the contract specifications, to measure
the sediment thickness at the bottom of the shaft. A SQUID
consists of three penetrometer probes that can record
accurate displacement and force measurements to provide
an objective, quantitative assessment of sediment
thickness at the base of a drilled shaft.
The SQUID unit was pin-connected to the caisson rig’s
Kelly bar, as captured in Photograph 7. Each SQUID test
proceeded by slowly lowering the Kelly bar without rotation
until the entire weight of the Kelly bar was transferred to,
and resting on, the SQUID unit. Penetrometer force and
displacement measurements were continuously acquired
and reported to the Foundation Engineering Specialist. A
test run was terminated once two of the three
penetrometers registered a force greater than 10
kilonewtons, or the maximum penetrometer travel of 152
millimetres was reached for any one of the penetrometers.

Photograph 6. Concrete placement set-up showing

tremie hopper and splice cage connected to temporary
casing.

The placement of concrete upward from the base of the

caisson displaced the slurry, which was pumped back from
the pile head to the storage tanks. A pump with an output
equal to 60 cubic metres per hour (i.e., an output equivalent
to the output of the concrete pump) was utilized. The
concrete was overpoured near the top of the caisson to
ensure stability of the excavation side walls during
extraction of the temporary casing. A concrete report
documenting the pertinent information related to the tremie
method was generated for each caisson.

6.5 Disposal

Once the polymer slurry was no longer required, the

fluid was treated on site and discharged in accordance with
the project requirements. To break down the polymer Photograph 7. SQUID unit preparing to be lowered to
chains, sodium hypochlorite was added to the polymer and the base of a drilled shaft
a buffer consisting of citric acid was used to correct the pH.
Once neutralized, the polymer slurry met the same Five SQUID runs were performed for each drilled shaft:
thresholds set out for stormwater or groundwater and could one in the centre of the shaft as well as in the four
be safely disposed. quadrants surrounding the shaft centre, as shown in Figure
2.
In the contract tender documents, the thickness of
7 INSPECTIONS AND TESTING sediment, loose material, or debris at the base of the drilled
shaft was defined as the difference in the displacement
7.1 Shaft Quantitative Inspection Device (SQUID) plate measurements that occurred between a
Testing penetrometer force of 0.089 kilonewtons to 0.71
kilonewtons. Since a drilled shaft base often contains
Full-time geotechnical inspection and monitoring was irregularities from a level surface due to pilot holes or
carried out by the Foundation Engineering Specialist to grooves from cutting teeth or drilling tools, a SQUID run
oversee the soil and groundwater conditions encountered was considered complete provided the debris thickness
during caisson advancement. Upon reaching the design could be determined from a minimum of two force versus
displacement plots. A maximum sediment thickness of 25
millimetres, with at least 50% of the base of each shaft
having less than 25 millimetres of sediment, was specified
for the contract.

Figure 2. Typical test locations for a SQUID test. One test

comprises five measurements; each measurement is
comprised of two or three displacement readings
(depending on the presence of irregularities at the base of
the shaft), with the average displacement taken as the
debris thickness in millimetres.

All caisson base subgrades were approved based on Photograph 7. PVC tubing (top middle of photograph)
the results of the SQUID testing, confirming the and analyzer software for CSLs
effectiveness of both the polymer slurry and of the base
cleaning methods.
Each caisson requiring CSL contained six access
7.2 Pile Integrity Testing (PIT) tubes; in total, nine paths (tube pairings) were tested (six
perimeter paths and three cross sectional paths). Upon
PITs were carried out at least seven days after shaft completion of logging, each path was assigned a rating
concrete placement, or after the concrete had achieved 75 based on a percentage of the delay in first arrival time of
percent of the design strength (whichever occurred first). the signal transmission or energy reduction in decibels.
PIT is a low strain integrity test that uses pulse echo or All four tests showed clear CSL data of high quality and
transient response methods to detect potential defects demonstrated general homogeneity of the concrete along
such as major cracks, necking, soil inclusion, or voids. The the length of each caisson, again confirming the
PITs were carried out on all caissons and in accordance effectiveness of the polymer slurry and tremie placement.
with ASTM D6760. No major anomalies were detected in
the data collected from any of the tests, demonstrating the 8 PROJECT COSTS
effectiveness of the polymer slurry and tremie methods.
Velocity wave traces were clear and of high quality, and the Adoption of polymer slurry drilling techniques provided
profiles displayed a consistent pattern from caisson to significant cost savings for the project. The use of
caisson. conventional drilling methods utilizing segmental casing
require the mobilization of additional equipment,
7.3 Crosshole Sonic Logging (CSLs) specifically larger caisson equipment combined with an
oscillator and tanks to store and treat the water during pile
CSLs were carried out on four caissons in accordance with installation. Moreover, a minimum of six sections of
ASTM D6760. During drilled shaft construction, 50 mm segmental casing would be required and mobilized from
steel tubes were fastened to the steel reinforcement cage the adjacent laydown area, which would significantly
approximately equidistant around the inner perimeter of the decrease productivity and cause more frequent disruptions
cage. The tubes were filled with water following concrete to traffic flow.
placement. Data was recorded for the full length of the Polymer slurry drilling allowed for significantly faster
caisson in 50 mm intervals as the probes were raised in the tremie concrete placement than conventional methods, as
tubes, from the bottom of the shaft to concrete cut-off there was no need to remove the temporary casing until the
elevation. Photograph 7 shows the PVC tubes and end of concrete placement. This decreased the placement
analyzer for a CSL on one of the caissons. time by approximately 50 percent and promoted the
required concrete workability retention, in turn reducing the
risk of concrete anomalies from irregular concrete flow.
9 CONCLUSION

The successful construction of caisson foundations for the

Highway 400/Essa Road Overpass Replacement Project in
Barrie, Ontario showed that the polymer slurry drilling
system can be an effective alternative to conventional
drilling methods in challenging subsurface conditions. The
polymer slurry drilling system addressed both construction
and logistical challenges for the project; the system
reduced the amount of drilling equipment, drill fluid
treatment, and trucking that otherwise would have been
required for conventional drilling methods while minimizing
disruptions to traffic. Moreover, the system reduced
concrete placement time while increasing overall drilling
productivity through the elimination of multiple casing
strings.
Lastly, the project showed that contract specifications
for drilled shafts under polymer slurry can be appropriately
established; slurry properties were monitored through a
robust quality control program that employed
straightforward testing methods, each caisson base was
inspected via a SQUID test that provided an objective,
quantitative assessment of base cleanliness, and the use
of PITs and CSLs provided a non-destructive means to
verify the structural integrity of the caissons.

10 REFERENCES

Canadian Geotechnical Society. 2006. Canadian

Foundation Engineering Manual, 4th Edition,
BiTech Publisher Ltd., British Columbia, Canada.
Decourt, L. 1995. Prediction of load-settlement
relationships for foundations on the basis of the
SPT-T, Ciclo de Conferencias Internationale,
Leonardo Zeevaert, UNAM, Mexico, 85-104.
Meyerhof, G.G. 1976. Bearing Capacity and Settlement
of Pile Foundations: The Eleventh Terzhagi
Lecture, J. of Geotech. Engrg. Div., ASCE, 102,
GT3: 195-228. Discussion in 103, GT3 and GT4,
Closure in Vol. 103, GT9.
Poulos, H.G. and Davis E.H. 1980. Pile Foundation
Analysis and Design, John Wiley and Sons, New
York, NY, USA.
Golder Associates Ltd. 2018. Foundation Investigation
and Design Report, Essa Road Overpass (Site No.
30X-0178/B1&2), Highway 400 / Essa Road
Interchange Reconstruction, Barrie, Ontario, MTO
G.W.P. 2337-16-00, Assignment No. 2017-E-0076
(MTO GEOCRES No. 31D-767)
Prediction and mitigation of construction
induced ground vibrations on sites in
Alberta

Bernie Mills and Kyle Noble

Stantec Consulting Ltd., Calgary, Alberta, Canada

ABSTRACT
With an increase in urban and industrial development, the need for better and more reliable vibration prediction tools for
the geotechnical engineering community has increased in recent years. Having a simple and practical means of predicting
construction induced ground vibrations is important as an initial pre-construction screening tool. The selection of site and
project specific vibration thresholds is crucial to form the basis of compliance and mitigation requirements. In this paper,
vibration monitoring results from three Alberta construction sites are described. Vibration sources included impact pile
drivers, vibratory pile drivers, and other construction equipment. Pre-construction predictions were completed with
empirical vibration attenuation models, then improved with site specific “vibration versus distance” relationships. Initial
models were calibrated with field measured data for improved predictions as construction progressed at the sites. The
case studies show examples of mitigations used to reduce the potential for public annoyance and infrastructure damage.

RÉSUMÉ
Avec une augmentation du développement urbain et industriel, le besoin d'outils de prédiction des vibrations meilleurs et
plus fiables pour la communauté de l'ingénierie géotechnique a augmenté ces dernières années. Disposer d'un moyen
simple et pratique de prédire les vibrations du sol induites par la construction est important en tant qu'outil de dépistage
initial avant la construction. La sélection de seuils de vibration spécifiques au site et au projet est cruciale pour former la
base des exigences de conformité et d'atténuation. Dans cet article, les résultats de la surveillance des vibrations de trois
chantiers de construction en Alberta sont décrits. Les sources de vibrations comprenaient des enfonceurs de pieux à
impact, des enfonceurs de pieux vibrants et d'autres équipements de construction. Les prédictions de pré-construction ont
été complétées avec des modèles empiriques d'atténuation des vibrations, puis améliorées avec des relations « vibrations
versus distance » spécifiques au site. Les modèles initiaux ont été calibrés avec des données mesurées sur le terrain pour
améliorer les prévisions à mesure que la construction progressait sur les sites. Les études de cas montrent des exemples
de mesures d'atténuation utilisées pour réduire le risque de gêne publique et de dommages aux infrastructures.

1 INTRODUCTION Having a good understanding of vibration thresholds

applicable for a specific site, project or important
Higher-density development is becoming more common infrastructure is crucial to form the basis of compliance and
place and has led to more construction within urban mitigation requirements. Likewise, having a simple and
residential and commercial areas and near important practical means of predicting construction induced
underground infrastructure such as pipelines in industrial vibrations is also important as an initial pre-construction
settings. Construction activities such as impact and screening tool and to establish construction and monitoring
vibratory pile driving and the operation of construction plans that reduce the risk of excessive vibrations from
equipment such as vibratory compactors, excavators, and annoying the public or damaging important infrastructure.
dump trucks generate ground vibrations. Demolition
activities, roadway maintenance operations, and In this paper, vibration monitoring results from three
highway/roadway traffic are also sources of ground Alberta construction sites are described. Vibration sources
vibrations. If the amplitudes are high enough, ground included impact pile drivers, vibratory pile drivers, and
vibrations have the potential to damage structures, cause other construction equipment such as vibratory drum
cosmetic damage, or disrupt the operation of sensitive rollers. Pre-construction predictions were completed with
equipment. Ground vibrations and ground borne noise can empirical vibration attenuation models, then improved with
also be a source of annoyance to the public who live site specific “vibration versus distance” relationships. Initial
nearby. As a result, the need for better and more reliable models were calibrated with measured data for improved
vibration prediction tools and applicable threshold criteria predictions as construction progressed at the sites. The
for the geotechnical engineer has increased in recent case studies also show examples of mitigations used to
years. reduce the potential for infrastructure damage.

Canada does not have a specific vibration standard or 2 VIBRATION STANDARDS AND CRITERIA
code, so the geotechnical engineering community typically
references vibration standards from other countries such Geotechnical engineers need to use proper vibration
as the US, Britain, Switzerland, and Germany as examples. standards or threshold criteria for their specific project
considering any applicable bylaw criteria, project type, and maximum allowable PPV for bridges be 50 mm/s.
potential nearby vibration receivers such as the public, CHBDC reports that a PPV limit of 100 mm/s is
buildings, and any other important nearby structures such not uncommon for bridges and should be
as buried pipelines. The following sections provide a addressed on a case-by-case basis.
general summary of vibration standards and criteria
considered more relevant for the case studies presented in 2.1.2 American Guidelines
this paper.
The more common and relevant guidelines for construction
2.1.1 Canadian Standards induced ground vibrations in Canada is Caltrans (2020)
and the FTA (2018) guidance manuals.
Construction induced ground vibrations are not specifically
addressed in the Canadian National Building Code. While 2.1.2.1 Caltrans (2020) Guidance Manual
several Canadian cities such as Toronto have established
vibration criteria in city bylaws, the main code reference is The California Department of Transportation,
the Canadian Standards Association (CSA) S6:19 Transportation and Construction Vibration Guidance
Canadian Highway Bridge Design Code (CHBDC 2019). Manual (Caltrans, 2020) is a comprehensive manual that
Section and commentary references in CHBDC include: provides practical guidance on the topic of vibrations
induced by construction equipment.
• Section 6.11.4.13 Pile Driving Induced Vibrations. Caltrans uses the peak particle velocity (PPV) to
The effects of vibration induced by pile installation describe vibrations in the ground. PPV is generally
shall be considered to mitigate adverse human accepted as the most appropriate descriptor for evaluating
response and potential structural damage or the potential for building damage. Caltrans most often uses
potential damage to the supporting ground. a vertical PPV description because vibration amplitude
• C6.11.1.3 Selection of Deep Foundation Units. along the ground surface is usually, but not always,
When deep foundation units are driven adjacent greatest in the vertical direction (Hendriks 2002). More
to existing structures or services, precautions importantly, the vertical component is usually
should be taken to avoid damage to the existing representative of the vibration in all three orthogonal
installations from heave, vibration effects, or directions and is the most easily measured.
ground displacements and compaction. This type
of damage can be minimized by using non- Caltrans uses the simplified propagation model
displacement piles and selective pre-augering proposed by Wiss (1981). The simplified model uses an “n”
during installation. value, which is a composite value that accounts for the
• C6.11.4.13 Pile Driving Induced Vibrations. The geometric and material damping. Woods and Jedele
(1985) developed values for “n” related to generic soil types
structures and foundations in the vicinity of the
from field construction data.
pile installation operations should be pre-
surveyed and a real-time vibration monitoring As described in Section 3, the majority of the empirical
program should be conducted when pile driving PPV predictions used in this paper were from Caltrans
induced vibrations are anticipated. The vibrations using the source reference PPVs values from FTA (2018).
exceeding certain levels has the potential to
damage structures, disrupt sensitive equipment 2.1.2.2 FTA (2018) Guidance Manual
and disturb individuals. These vibrations can also
have detrimental effects on fresh concrete. The Transit Noise and Vibration Impact Assessment by the
• For pile driving, the CHBDC references the Federal Transit Administration (FTA, 2018) is a
California Transportation and Construction comprehensive guidance manual for addressing noise and
Induced Vibration Guidance Manual (2004) and vibration impacts for transit construction projects in the US.
comments that it provides empirical relationships Table 1 provides a portion of the comprehensive list of
for a wide range of construction equipment. vibration source levels for construction equipment.
CHBDC also references work by Wiss (1981),
Bornitz (1931), and Woods and Jedele (1985). Table 1. Vibration Levels for Construction Equipment
• CHBDC states that the vibration thresholds for
damage and human discomfort are different for Equipment PPV at 25 feet (in/s)
transient versus steady state dynamic loads. It Pile Driver (Impact) Upper range 1.518
states that the thresholds for steady state are less Typical 0.644
than those of transient sources. The vibrations Pile Driver (Vibratory) Upper range 0.734
from impact pile driving are considered a steady Typical 0.170
state vibration source. Vibratory Roller 0.210
• To evaluate the potential for damage to structures Large bulldozer 0.089
and human annoyance, CHBDC references Loaded trucks 0.076
similar vibration criteria as reported in the current
Caltrans (2020) Tables 19 and 20. FTA states that the levels provided in the above table
• CHBDC also references damage threshold are reasonable estimates for a wide range of soil conditions
criteria from Dowding (1996) that suggests and cautions that there is a considerable variation in
reported ground vibration levels from construction W = nominal energy of the hammer (J)
activities. x = distance measured along ground surface (m)

2.1.2.3 British Report – TRL (2000) 3.2 Vibratory Pile Drivers

The report by the Transportation Research Laboratory is a Based on the literature review by Caltrans, the PPV for
comprehensive report titled “Groundborne vibration vibratory pile drivers can be estimated as follows:
caused by mechanized construction works.” It was
prepared for the Civil Engineering, Highways Agency and PPVVibratory Pile Driver = PPVRef (25/D)n [4]
includes a review of vibrations from piling works. The
empirical PPV predictions by Attewell (1995) for impact and Where:
vibratory pile drivers were used throughout this paper. PPVRef = 0.65 in/sec for a reference pile driver at 25 ft
Attewell (1995) describes a probabilistic approach by D = distance from pile driver to the receiver in ft
providing formulas that are either “more probable levels of n = 1.1 is the attenuation rate through the ground
vibration” or “high level of confidence that the predicted
values will not to be exceeded.” Vibratory pile drivers generate the maximum vibration
levels during the start-up and shut-down phases of the
operation because of the various resonances that occur
3 CONSTRUCTION VIBRATION PREDICTIONS during vibratory pile driving. The above equation can be
used to estimate the vibration amplitude during the
To assess the potential for vibrations to damage structures resonant start-up and shut-down phases of the pile driving
or annoy people, a simple and practical means for operation.
estimating the PPVs from various types of equipment and
soil types at various distances is provided by Caltrans As reported by Attewell (1995), the PPV from impact
(2020), FTA (2018), and Attewell (1995) as reported in TRL pile drivers can be estimated as follows:
(2000). The PPVs estimated using the methods described
by Caltrans are expected to be typical worst-case values For a high level of confidence not to be exceeded:
and should be used as guidelines only.
vres = 1.8 [Wc0.5/x] [5]
3.1 Impact Pile Drivers
For a more probable level of vibration:
As reported by Caltrans (2020), the PPV from impact pile
drivers can be estimated as follows: vres = 1.0 [Wc0.5/x]0.95 [6]

PPVImpact Pile Driver = PPVRef (25/D)n x (Eequip/ERef) [1] Where:

vres = resultant velocity (mm/s)
Where: Wc = nominal energy per cycle of the hammer (J)
PPVRef = 0.65 in/sec for a reference pile driver at 25 ft x = distance measured along ground surface (m)
D = distance from pile driver to the receiver in ft
n = 1.1 is the attenuation rate through ground 3.3 Other Construction Equipment
ERef = 36,000 ft-lb (rated energy of reference pile driver)
Eequip= rated energy of impact pile driver in ft-lbs Using the vibration source level amplitudes provided by
FTA (2018) as described in detail in Section 2.1.2.2, the
Literature indicates that the value of “n” in the above PPV from various equipment can be estimated using the
equation is generally between 1 to 1.5. Caltrans suggests relationship by Caltrans (2020) as follows:
the value to use is 1.1. Vibration estimates may be refined
further using values of “n” that are based on the soil type PPVEquipment = PPVRef (25/D)n [7]
classification as described in more detail in Caltrans.
Where:
As reported by Attewell (1995), the PPV from impact PPVRef = reference PPV at 25 ft
pile drivers can be estimated as follows: D = distance from equipment to the receiver in ft
n = 1.1 is the attenuation rate through ground
For a high level of confidence not to be exceeded:

vres = 1.5 [W0.5/x] [2] 4 EVALUATING POTENTIAL IMPACTS

For a more probable level of vibration: There is limited consistency in the literature between
categorization of effects of structural damage and human
vres = 0.76 [W0.5/x]0.87 [3] response. It is apparent that damage thresholds for
continuous sources are less than those for single-event or
Where: transient sources. Many types of construction activities fall
vres = resultant velocity (mm/s) between a single event and a continuous source. An
impact pile driver, for example continuously generates plate compactors (hoe-packs) can also generate high
single transient events. As reported by Caltrans (2020), it vibration levels. General methods for reducing vibrations
is reasonable to separate into two categories: continuous for impact pile drivers and vibratory compaction equipment
and transient. are provided below:

To assess structural damage and human response Reduction Measures for Impact Pile Driving:
potential from construction induced ground vibrations, the • Predrilling.
PPVs estimated using the empirical formulas in Section 3 • Jetting using a mixture of air and water.
can be compared to the threshold criteria provided in • Using non-displacement piles.
Tables 2 and 3. • Using pile cushioning.
• Scheduling specific times to minimize disturbance
Table 2 Vibration Damage Potential Threshold Criteria and impact to people.
(after Caltrans 2020) • Alternatively, impact pile driving can be avoided
using non-impact pile drivers such as vibratory
Maximum PPV (in/s) pile drivers or switching to cast-in-place piles.
Continuous
Structure and Condition Reduction Measures for Other Construction
Transient Frequent
Sources Intermittent Equipment:
Sources • Increasing the distance between the source and
Fragile buildings 0.2 0.1 receiver. An example of this measure is provided
Historic buildings 0.5 0.25 in Case Study 3.
Older residential structures 0.5 0.3 • Choose alternative construction methods by
(7.6 mm/s) selecting non-impact methods.
New residential structures 1.0 0.5 • Phase construction activity so as not to occur in
Modern 1.0 0.5 the same time-period.
industrial/commercial • Route heavy-loaded trucks away from residential
streets and select streets with the fewest homes.
Table 3 Vibration Annoyance Potential Criteria • Scheduling the work when most people are not in
(after Caltrans 2020) the area and avoid nighttime activities.

Maximum PPV (in/s)

6 ALBERTA CASE STUDIES
Continuous
Human Response
Transient Frequent The following three case studies are described to illustrate
Sources Intermittent the general procedure of:
Sources
Barely perceptible 0.04 0.01 • Developing an initial pre-construction screening
Distinctly perceptible 0.25 0.04 procedure using available empirical PPV models
Strongly perceptible 0.9 0.10 and relationships to predict the vibration levels,
Severe 2.0 0.4 • Comparing these initial predictions to the
threshold criteria established for the project to
In Tables 2 and 3, transient sources are single isolated determine if there are potential issues,
events such as blasting. Continuous/frequent intermittent • Carrying out field vibration monitoring during
sources include impact pile drivers, vibratory pile drivers construction activities to verify, calibrate and
and construction equipment such as vibratory compaction. refine the input parameters and models,
• Using the improved models to predict future
For human response, The City of Calgary Roads for vibrations across the site with more accuracy,
example uses a human response to traffic vibrations scale. • Establishing mitigations as required to keep the
Humans tend to be sensitive to particle velocities with vibration levels below the required thresholds.
frequencies in the range of 8 to 80 Hz. The typical
frequencies measured from construction equipment varies The field measured PPVs described in the following
between 10 and 40 Hz, within the range that is sensitive to case studies were the maximum PPVs measured in any of
humans. Based on the City of Calgary scale: PPV of 2.5 the three orthogonal directions. Typically, the maximum
mm/s to 5 mm/s “begins to annoy” people, and above 5 PPV was in the vertical direction.
mm/s is “annoying” to people.
6.1 Impact Pile Hammer - Industrial Site
(Case Study 1)
5 VIBRATION MITIGATION STRATEGIES
The first case study is an industrial construction site in
Impact pile driving can be the most significant source of northern Alberta that involved the installation of driven steel
vibration at construction sites. Other activities and pipe piles for support of a concrete overpass bridge
equipment such as vibratory drum rollers and vibratory
structure. The foundations consisted of 711 mm diameter
steel pipe piles with a wall thickness of 19 mm. The piles
were driven open ended to practical refusal in the
underlying bedrock using a hydraulic impact hammer. The
soil conditions consisted of loose to compact sand
overlying clay overlying bedrock. Pile embedment depths
varied from approximately 16 m to 18 m.

This project consisted of several hundred piles with

vibration monitoring on a total of 39 piles. The piles were
driven using a DFI HHB600-60 hydraulic hammer with a
rated energy of 105.6 kJ or 77,886 ft-lbs. A specified
percentage of the piles were also PDA tested. Out of the
39 piles that were tested for ground vibrations, a total of 12
piles were also PDA tested.
Figure 1 Predicted Maximum PPV versus Distance for
The main vibration receptors at this site were existing Impact Pile Driver
pipelines varying from approximately 8 m to more than 50
m away from the hydraulic pile driver. The pipelines were 6.1.2 Revised Predictions versus Field Measurements
buried in the order of 2 m to 3 m or more below ground
surface. A project specific maximum PPV threshold of 50 In general, the sequence of piling started with the piles that
mm/s was established for the pipelines. An alert level of 30 were the furthest away from the pipelines to obtain an
mm/s and reporting sequence was also established to understanding of the “PPV versus distance” relationship
provide notice for the case when this level was exceeded. before moving on to piles that were closer to the pipelines.

A vibration monitoring plan was developed by the Upon closer review and field survey, it was determined
contractor to ensure the vibration levels in the ground at the that the closest pile was 8.5 m away from the pipeline.
pipelines were below this 50 mm/s threshold. The plan Figure 2 shows the field measured PPV versus distance
called for several vibration seismographs to be placed on results for the initial 38 piles at distances in the range of 18
the ground surface above the pipelines to be monitored. m to 54 m away. A logarithmic best fit line was used for all
Contingency plans were developed by the contractor for 38 piles and included the higher capacity piles. It is
the case where the measured vibrations were approaching acknowledged that including the higher capacity piles to
or exceeding these threshold values. make the 8.5 m prediction was conservative. This best fit
line was projected to the closest pile distance of 8.5 m
6.1.1 Initial Predictions away.

Initial pre-construction PPV predictions were carried out

using the empirical relationships by Caltrans (2020) and
Attewell (1995), using source reference values for impact
pile drivers from FTA (2018). The soils were classified as
Class III soils with an attenuation rate of n=1.1. The results
are shown in Figure 1.

Due to the relatively large hammer energy and

requirement for piles to be driven to practical refusal in
bedrock using the hydraulic impact hammer, it was
considered appropriate to use the “upper bound” and
conservative source levels for impact pile driving from FTA.
Similarly, a second model by Attewell (1995) for a “high
confidence not to exceed” level was also used. Figure 1
shows the results of the initial vibration predictions of 53
mm/s to 62 mm/s for the closest distance of 8 m between Figure 2 Field Measured Maximum PPV versus Distance
the pile driver and the pipeline. for Impact Pile Driver

As shown in Figure 1, there was a relatively high As shown in Figure 2, the field measured PPV data
probability that the closest pile at approximately 8 m away matched the typical and upper bound Caltrans models very
would produce ground vibrations greater than the 50 mm/s well. It is important to note that there was a grouping of
threshold. For this reason, it was recommended for the much higher capacity piles as indicated in Figure 2 that
contractor to start the piling process with the furthest piles were driven to practical refusal with a much higher
first and to drive the closest piles near the end of the piling termination criteria as compared to the other piles.
process. Confirmation of the higher capacity piles were made using
PDA. This probably explains why these piles plotted on or
above the upper bound Caltrans and Attewell (1995) lines.
 Initial pre-construction vibration predictions were
At 8.5 m, the revised prediction was between 35 mm/s carried out for the piles using the PPV models by Caltrans
and 58 mm/s, as this pile was a lower capacity pile with a and Attewell (1995). The results of the predictions are
blow count termination criterion similar to the piles that plotted on Figure 3. It was predicted that at 10.9 m away,
plotted below the upper bound Caltrans line. As shown in the maximum PPV for Structure 1 piles would be between
Figure 2, the field measured PPV at 8.5 m away was 41 4 and 15 mm/s using the typical and upper bound Caltrans
mm/s, below the 50 mm/s threshold. relationships and as high as 30 mm/s using the “high
confidence not to exceed” model by Attewell (1995).
6.2 Vibratory Pile Hammer – Industrial Site
(Case Study 2) 6.2.2 Field Measurements

The second case study is an industrial construction site in The sequence of piling started with the piles that were the
northern Alberta that involved the installation of very large furthest away from the pipelines to obtain an understanding
diameter driven steel pipe piles for two structures. The piles of the “PPV versus distance” relationship before moving on
were driven open ended using a vibratory hammer. These to the piles that were closer to the pipelines.
piles were driven during the winter, and as such predrilling
of the frozen ground was carried out before the piles were For the closest Structure 1 pile, the maximum field
installed. measured PPVs at 10.9 m away were 9.3 mm/s, 8.0 mm/s
and 6.3 mm/s showing a decreasing PPV trend as the pile
The main vibration receptors at this site were existing embedment depth increased. For the closest Structure 2
pipelines varying from approximately 10.9 m to more than pile, at a distance 29 m away, the maximum field measured
100 m away from the vibratory pile drivers. The depth of PPVs were 5.9 mm/s and 2.6 mm/s showing a similar
the pipelines was approximately 1.5 m below ground decreasing PPV trend as the pile embedment depth
surface. A project specific maximum PPV threshold of 50 increased. These field measured results are plotted on
mm/s was established for the pipelines. An alert level of 35 Figure 3 and compared to the initial predictions. The field
mm/s and reporting sequence was also established to measurements matched the typical and upper bound
provide advance notice for the case when this level was Caltrans and the more probable Attewell (1995) models
exceeded. reasonably well. The “high confidence not to exceed” line
by Attewell (1995) was considerably higher by a factor of
A vibration monitoring plan was developed to ensure 2x or more as compared to the field measurements.
the piling activities were below this 50 mm/s threshold. The
plan called for the vibration seismographs to be placed on
the ground surface above the pipelines to be monitored.

The Structure 1 piles were 2590 mm diameter piles with

wall thickness of 16 mm and were driven to minimum
embedment of 15 m using an APE Model 400 vibratory
hammer. The closest Structure 1 piles were 10.9 m away
from the pipelines. The piles, embedment and hammer for
the Structure 2 piles were the same as Structure 1 with the
closest pile at 29 m away from the pipelines.

A series of trial vibration measurements on 1220 mm

diameter pipe piles with wall thickness of 16 mm were also
carried out using a smaller 75HD PTC vibratory hammer.
These measurements were carried out during the
predrilling activities and set up of the 2590 mm diameter Figure 3 Predicted versus Field Measured Maximum PPVs
piles described above for Structure 1. These for Vibratory Pile Driver
measurements on the 1220 mm diameter piles were
carried out at a location where the pipelines were greater For the 1220 mm diameter trial piles using the smaller
than 100 m away. vibratory hammer for pipelines that were more than 100 m
away, the field measurements were generally consistent
6.2.1 Initial Predictions with the Caltrans model using the typical FTA PPV
reference value.
Initial pre-construction PPV predictions were carried out
using the empirical relationships by Caltrans (2020) and 6.3 Construction Equipment – Residential Setting
Attewell (1995), using source reference values for vibratory (Case Study 3)
pile drivers from FTA (2018). The frozen soils were
modeled using an attenuation rate of n =1.1. The energy The third case study involved vibration monitoring services
per cycle of the APE Model 400 that was used in the during construction of a local drainage improvement project
Attewell (1995) formula was estimated to be 32,000 Joules. within an older residential neighborhood in Calgary,
Alberta. The purpose of the monitoring program was to
assess if the construction induced vibrations had the
potential to annoy the public and/or damage residential
structures.

The study included an initial vibration pre-construction

screening assessment followed by on-site vibration
monitoring using multiple seismographs on city property.
The vibration monitoring was not carried out on private
residential property and no pre-construction or post-
construction surveys were completed.

Construction activities and equipment included

excavations using tracked and rubber-tired backhoes,
delivery of fill materials by dump trucks, spreading of fill
materials using bulldozers and backhoes, compaction of
fills in lifts using a vibratory drum roller and hydraulic plate
compactor (hoe-pack). Figure 4 Predicted Maximum PPV versus Distance for
Construction Equipment (Log-Log)
The main vibration receptor for this case study was
older residential structures, residential occupants, and the 6.3.2 Revised Predictions versus Field Measurements
public. A review of the site layout and the proposed
construction activity locations showed that the residential A field vibration monitoring program was undertaken and
structures were approximately 10 m to more than 30 m included development of PPV versus distance
away, with several residences as close as 8 m away. measurements for three different construction activities
and equipment as construction progressed. This field
In the absence of a specific vibration standard for program allowed site and equipment specific attenuation
damage to residential structures, a conservative damage curves to be developed and comparison to the empirical
threshold PPV of 7.6 mm/s as reported by Caltrans (2020) relationships and PPVref values proposed by Caltrans and
for older residential structures was used. For human FTA. Especially important was to confirm the attenuation
response, we used the City of Calgary scale: PPV of 2.5 models for both the sheepsfoot vibratory roller and the hoe-
mm/s to 5 mm/s “begins to annoy” people, and above 5 pack as this equipment was considered to have the highest
mm/s is “annoying” to people. vibration potential. Figure 5 shows the results of the field
measurements and the predicted power law relationships
6.3.1 Initial Predictions that were developed for each activity and piece of
equipment.
Initial pre-construction PPV predictions were carried out
using the empirical relationships and PPVref values for As shown in Figure 5, the revised maximum vibration
various construction equipment as outlined by Caltrans level of 7 mm/s was predicted for the hoe pack and
(2020) and FTA (2018). The soils at the site were classified sheepsfoot vibratory drum roller at a distance associated
as either Class II or Class III soils with attenuation rates of with the closest residence at 8 m away. It is important to
n=1.3 or n=1.1, respectively. Figure 4 plotted in log-log note that this revised prediction of 7 mm/s is higher than
scale shows the results of the initial vibration predictions the initial prediction of 5 mm/s using the available published
considering both the damage threshold for older residential empirical relationships and reference PPV values. This
structures, and the vibration levels that are “annoying” to revised predicted vibration level of 7 mm/s was also lower
people. than the damage threshold used for older residential
structures of 7.6 mm/s, so the probability of causing
As shown in Figure 4, compaction activities using a damage to older residential structures was considered low.
vibratory roller has the potential to cause the highest
vibrations. The PPVref for a hydraulic plate compactor (hoe- Regarding public annoyance potential, a review of the
pack) was not available for the initial screening review. It field measured frequencies showed values typically in the
was assumed that the hoe-pack equipment would have a range of 10 to 40 Hz, which is in the range that is sensitive
similar vibration potential as a vibratory drum roller. to people according to the City of Calgary scale. Given the
Based on the results of the initial screening review, it revised predicted PPV and the measured frequencies, the
was predicted that the construction equipment and above analysis showed that there was a relatively high
activities at 8 m (closest residence) or more away would be probability that this construction activity would be annoying
outside of the damage threshold to older residential to the residences and the public and complaints could be a
structures. However, due to the lower human annoyance real possibility.
thresholds, it was predicted that the construction activities
would potentially be annoying to the public. A fourth construction activity was also recorded and
shown as “outliers” on Figure 5. This activity consisted of
an excavator bucket breaking concrete. The vibration
seismograph was located approximately 8 m to 10 m away
when this event occurred with measured PPVs of 32 mm/s
and 34 mm/s. The closest residence was approximately 16 amplification effects, potential resonance, and
m away when this event occurred. The contractor was combined effects from multiple equipment.
asked to move this activity to an approved location at an • If the amplitudes are high enough, ground
appropriate distance away from the residences. This fourth vibrations have the potential to damage structures
activity shows the importance of vibration monitoring of and annoy the public.
construction activities to ensure PPV’s are lower than • Although not discussed in this paper, settlement
target thresholds. of the surrounding soil as the result of pile driving
has the potential to damage surface and buried
infrastructure.
• There are inconsistencies in the literature on
damage thresholds to structures and public
perception. For this reason, it is crucial to assign
site and project specific threshold criteria that are
representative for a project, structure type, and
geographic location.
• Depending on project requirements, design and
construction specifications should include
vibrations and vibration damage thresholds for
nearby infrastructure including the requirement of
a vibration monitoring plan (VMP). VMPs are an
important component of the overall construction
quality management plan for a project. The VMP
should include details of the how and where the
Figure 5 Revised Predicted PPV versus Distance using field monitoring will be conducted and should
Field Measurements for Construction Equipment include mitigation strategies for the case of
thresholds being exceeded. Depending on the
project, the VMP is often developed by the
7 LESSONS LEARNED contractor and approved by the designers and
owner.
From the perspective of the geotechnical engineer and the
owner of potentially vibration sensitive infrastructure, the
following lessons learned are provided: 8 CONCLUSIONS

• The case studies showed three examples of how The three case studies presented were successfully
relatively simple empirical vibration prediction constructed and met the vibration damage threshold
models were used to estimate pre-construction criteria established. The importance of field vibration
vibration levels followed by verification and monitoring as part of the vibration monitoring plan cannot
refinement using field measurements as be overstated. By having a plan and properly executing the
construction progressed. plan, vibrations can be managed to acceptable levels.
• The empirical predictions described in this paper
need to be used with caution and are considered
reasonable estimates for a wide range of soil 9 REFERENCES
conditions. The researchers caution that there is
a considerable variation in reported ground Attewell P. B. 1995. Tunnelling contracts and site
vibration levels from construction activities. investigation. Chapman and Hall, London.
• The main factors influencing construction induced Bornitz G. 1931. Über die Ausbreitung der von
ground vibrations for the models and case studies Großkolbenmaschinen erzeugten
presented are as follows: equipment type, source Bodenschwingungen in die Tiefe. J Springer, Berlin.
energy level, distance along the ground surface Canadian Standards Association (CSA) S6:19. 2019.
between the source and receiver, and soil class Canadian Highway Bridge Design Code (CHBDC).
attenuation rate. California Department of Transportation (Caltrans). 2020.
• It is acknowledged that the distance factor is Transportation and Construction Vibration Guidance
described by other researchers as a sloping line, Manual.
r measured from the pile tip to the receiver. There Dowding, C. H. 1985. Blast vibration monitoring and
is evidence that using the sloping line is a more control. Prentice-Hall. Englewood Cliffs, NJ.
accurate way to predict vibrations from pile Hendriks, R. 2002. Transportation related earthborne
driving, since the energy source is often vibrations (Caltrans experience). California Department
transferred from the pile tip to the soil as reported of Transportation. Sacramento, CA.
by Massarach and Fellenius (2015). Hiller D. M., and Crabb G. I. 2000. Transportation
• The case studies presented considered single Research Laboratory, Report 429, Ground borne
source ground vibrations only. Many other factors vibration caused by mechanised construction works,
may be applicable for a project such as building
 Quality Services — Civil Engineering, Highways
Agency.
Transportation Research Laboratory (TRL). 2000.
Groundborne vibration caused by mechanized
construction works. Prepared for Quality Services –
Civil Engineering Highways Agency, D M Miller and G I
Crabb. TRL Report 429
US Department of Transportation. 2018. Transit Noise and
Vibration Impact Assessment Manual. Federal Transit
Administration, FTA Report No.0123.
Wiss, J. F. 1981. Construction vibrations: state-of-the-art.
Journal of the Geotechnical Division 107(GT2):167–
181.
Woods, R. S., and L. P. Jedele. 1985. Energy-attenuation
relationships from construction vibrations. Pages 187–
202 in G. Gazetas and E. T. Selig (eds.), Vibration
Problems in Geotechnical Engineering: Proceedings of
a Symposium. Sponsored by the Geotechnical
Engineering Division, in conjunction with the American
Society of Civil Engineers Convention, Detroit, MI,
October 22, 1985. New York: American Society of Civil
Engineers.
Wednesday, October 4, 2023

FOUNDATIONS III
Bedrock Stabilization Design for the World’s
Largest Rollercoaster, Qiddiya Six Flags
Project, Qiddiya, Saudi Arabia
Imad Alainachi, Faisal Abou-Seido, Maha Saleh & Fernanda
Carozzi
Paterson Group, Ottawa, Ontario, Canada

ABSTRACT

Construction of the Falcon’s Flight rollercoaster along the rugged terrain of Tuwaiq Mountains entails major geotechnical
challenges. The rollercoaster, located within the Qiddiya Six Flags theme park project is designed to be built on a 65 m
high cliff which dives into a 160 m long talus slope in an area that is prominent to sheer cliffs and active rockfall events.
These challenges are discussed in this paper along with the proposed remedial measures for rockfall stabilization,
including but not limited to blasting, slope stability and 3-D rockfall analysis, design of netting and rock anchors, and rock
catch fence design. The paper also provides an overview of the inspection and monitoring plans designed to ensure the
long-term durability of the bedrock face during the service life of the proposed rollercoaster.

RÉSUMÉ

La construction des montagnes russes Falcon's Flight sur le terrain montagneux de la chaîne de Tuwaiq dans le cadre
du projet Qiddiya Six Flags comporte des défis géotechniques majeurs car il est sur une falaise de 65 m de haut et
plongera dans un talus de 160 m dans une zone qui est caractérisée par des falaises abruptes et des chutes de pierres
actives. Ces défis sont abordés dans cet article, ainsi que les mesures correctives proposées pour la stabilisation des
chutes de pierres, y compris, mais sans s'y limiter, le dynamitage, la stabilité des pentes et l'analyse des chutes de pierres
en 3D, la conception de treillis et d'ancrages de roches, et la conception de barrières de retenue de roche. L'article donne
également un aperçu des plans d'inspection et de surveillance qui ont été proposés pour assurer la durabilité à long terme
de la paroi rocheuse pendant la durée de vie prévue des montagnes russes proposées.

1 INTRODUCTION The Qiddiya project includes multiple facilities such as

stadiums, an art complex, racetracks and a golf course. In
1.1 General addition, the city will incorporate the theme park named
Six Flags Qiddiya (MoMRAH 2022)
Qiddiya is a unique sport, entertainment, and cultural The development of the record-breaking theme park
project proposed to be constructed 40 km southwest of includes the construction of the Exa-coaster “Falcon’s
downtown Riyad, the capital city of the Kingdom of Saudi Flight rollercoaster”, which is proposed to be the longest,
Arabia (Figure 1). The proposed destination city, spanning tallest, and fastest rollercoaster in the world. Flacon’s
over 340 km2, is one of the giga projects presented in Flight will include a 4 km-long track that will climb a 65 m-
Vision 2030 Strategy of Saudi Arabia (Habibi 2019; Rivera high vertical rock cliff followed by a dive into a 160 m-long
2021; Saudi Gazette 2018). The construction of the project talus slope, and reaching a speed of up to 250 km/hour
began in mid-2018 and is scheduled to open in 2023 (Cripps 2021).
(Qiddiya 2023). Due to the current conditions of the bedrock cliff, the
area is prominent to sheer cliffs and active rockfall events.
Therefore, bedrock stabilization measures are required,
including blasting the unstable bedrock face, stabilizing
the bedrock face, providing rockfall measures and
monitoring plans to ensure the long-term durability of the
bedrock face during the design service life of the proposed
rollercoaster. The geotechnical challenges involved in the
construction of this project, the proposed remedial
measures, and monitoring plans are discussed in this
paper.

1.2 Geology of the Study Area

The construction of the Exa-coaster project is taking place

along the rugged terrain of Tuwaiq Mountain (Figure 2)
Figure 1: Location of Qiddiya Project (Paterson 2022a). The nearly parallel sequence of Tuwaiq
mountain are west-facing scarps developed in the Jurassic observed that these oblique joints control the overhangs
age and are generally crop out as a north-south belt of the on the western cliff face and some of the planar sliding
Central Arabian Arch (Al-Hinai et al. 1997; Powers et al. failures on the eastern cliff face. The western cliff face
1966). The middle Jurassic outcrops of the study area (Figure 4) was also found to have high degrees of erosion
show limited to non-existent overburden material, and its and a large open crack. Noticeable tension cracks and
rock units are described to be marly limestone to limestone potential hazards of rockfall and wedge failure (including
with persistent sub horizontal grading. The area of the overhangs, near vertical and persistence fractures) have
rollercoaster includes the vertical cliff, with its also been observed on the cliff face along with detached
corresponding skirt escarpment and Wadi Hanifah lower boulders on the slope surface. Planer sliding failures were
plateau. The steep cliff presents primarily two faces: an also observed on the eastern cliff face.
eastern and a western cliff face (Bamousa et al. 2020;
Paterson 2022b). The photogrammetric survey conducted
as part of the cliff inspection showed that the bedrock 3 ROCKFALL STABILIZATION ANALYSIS
outcrops primarily consist of Mesozoic limestone with
persistent sub horizontal bedding (Figure 3) (Paterson Based on the risk assessment analysis which was
2022c). completed for the study area, it was determined that the
existing and potential geotechnical hazards, including
rockfalls and wedge failures, would lead to considerable
damage to site, equipment, and temporary structures
during construction. These consequences would lead to
significant delays in the project timeline and result in
additional construction and redesign costs. It might also
cause injuries and/or fatality to site personnel and the
public before, during, and after construction in the project
area (Paterson 2022c). Therefore, to manage these risks,
bedrock stabilization measures were proposed, which
would include determining the safe location of the line of
drilling and blast excavation, completing a rock face
stabilization analysis, and implementing rockfall remedial
measures. Monitoring plans were also found to be
necessary to ensure the long-term sustainability of the
bedrock face during the service life span of the proposed
Figure 2: The Mountainous terrain of Tuwaiq mountain rollercoaster.
chain
3.1 Slope Stability Analysis

A series of 2-D slope stability analyses were conducted

mainly using Rocscience SLIDE 2 software, to assess the
stability of the cliff faces for slope failure under existing
conditions. At least 7 slope sections, spaced at 15 m, were
analyzed along each cliff face, under static and seismic
conditions. The 2-D analysis was conducted using the
General Limit Equilibrium (GLE) analysis method. The
generalized Hoek-Brown material model was used for
modeling bedrock and the Mohr Coulomb material model
was used for modeling soils (Abramson et al. 2001;
Rocscience 2023). The rock and soil parameters adopted
in the design were obtained from the geotechnical
investigation reports completed in the site area and are
shown in Table 1. The analyses were conducted for both
Figure 3: Panoramic view of the cliff showing the bedrock pre-blasting and post-blasting conditions.
outcrops
3.1.1 Pre-blasting Analysis

2 GEOTECHNICAL CHALLENGES The purpose of the pre-blasting analysis was to identify a

primary location of the line of drilling with respect to the
Based on geotechnical investigation reports prepared for existing rock conditions to ensure a safe working platform
the study area, it was noted that the bedrock within the for the construction team and to eliminate the weathered
upper 30 m of the cliff is significantly weathered. (Paterson bedrock face and overhangs. In order to verify the model
2022d; Powers et al. 1966). Sub-vertical and oblique joints input data and analysis procedure, the pre-blasting
were identified during the cliff inspection, and it was analysis was followed by a trial blast operation.
Figure 4: Westren cliff face and associated geotechnical hazards

Table 1. Material properties used in SLIDE 2 for slope blasting operations. The set back distance takes into
stability analysis account removing the majority of the weathered bedrock
face, and reducing the number of major fractures, fissures,
Characteristics Value and clay/sand filled seams within the influence zone of the
cliff face. Furthermore, the proposed setback will provide
Silty Sand with gravel, cobbles, and boulders:
a safe working platform for heavy equipment and
Unit Weight (γ) 21 kN/m3 personnel and to accommodate a 6 m working platform at
Drained cohesion (C) 1 kPa the lower blast line.
Friction Angle (∅) 35O
3.1.2 Post-blasting Analysis
Limestone:
The post-blasting conditions were reviewed to determine
Unit Weight (γ) 24.6 kN/m3 the final location of the main blast line, taking in
Uniaxial Compression Strength (UCS) 35 – 50 MPa considration other factors, such as tension cracks, bedrock
Elastic Modulus (Ei) 35.4 GPa cracks and fissures filled with clay/sand seams, as well as
Geological Strength Index (GSI) 38 – 65 the structure loads. This analysis also aimed to determine
the optimum rock supporting system to ensure the
Intact Rock Constant (mi) 9
stabilization of rock face during and after blasting
Disturbance Factor 0.7 operation. Furthermore, this analysis was conducted both
under static and seismic conditions using a horizontal
ground acceleration of 0.1g, which was obtained from the
The results of the pre-blasting analysis showed a factor seismic zonation map of Saudi Arabia (Al-Hadad et al.
of safety of 1.3-1.5 for the studies scenarios. The potential 1992).
slip surface obtained from this analysis are shown in The results of post-blasting analysis showed computed
Figure 5. factors of safety exceeding 1.5 and 1.1 when analyzed
Based on the analysis results for the pre-blasting under static and seismic conditions, respectively. A sample
conditions, and the observations resulting from the trial of the potential slip surfaces obtained from this analysis is
blasting of the cliff face, it was noted that a setback of 4 m shown in Figures 6-7.
from the uppermost edge of the cliff is required for the line
of drilling in order to ensure the rock stability during the final
 and stabilization measures, as described in section 4
below.

3.2 Rockfall Analysis

3-D statistical analysis was completed for the roller coaster

area, using Rocscience Rocfall 3 to assess the risk of
rockfall on both sides of the cliff. The primary goal of the
rockfall analysis was to verify the suitability of the proposed
line of blast by identifying the spatial extent of potential
rockfall hazards under existing conditions (short term). The
other objective was to design the rockfall catch fence
system to protect the footings of the proposed structure at
the ascent and descent sections that are likely to be
impacted by post-construction potential rockfall hazards
during operation of the rollercoaster.
Figure 5: Results of pre-blasting slope stability analysis The 3-D analysis was conducted using the Lumped
Mass analysis method (Rocscience) and the input
parameters were obtained from the geotechnical reports
and summarised in table 2. The analyses were also
conducted for pre-blasting and post-blasting conditions, as
stated earlier.

Table 2. Input parameters used in Rocfall 3 for rockfall

analysis

Parameter (1) Value Variability

Scree Slope:
Normal Coefficient of Restitution (Rn) 0.31 75
Tangential Coefficient of Restitution (Rt) 0.81 75
Friction Coefficient 0.50 50

Bedrock:
Figure 6: Results of post-blasting slope stability analysis Normal Coefficient of Restitution (Rn) 0.50 50
under static conditions
Tangential Coefficient of Restitution (Rt) 0.55 20
Friction Coefficient 0.90 20
(1)
All parameters are uniformly distributed

3.2.1 Pre-blasting Analysis

The analysis under pre-blasting conditions was conducted

based on the existing cliff conditions, taking into
consideration rock release points at the crest of the cliff,
and the shape and orientation of the slope of the talus.
Although identified as a potential rockfall hazard, existing
blocks on the talus were not considered in this analysis
because one of the main design requirements is that all
existing large rock boulders will be removed from the
rollercoaster area prior to construction (Paterson 2023).
Based on the pre-blasting analysis results (Figure 8), it
was determined that a spatial extent of blasting and cliff
Figure 7: Results of post-blasting slope stability analysis stabilization of around 140 m from each the nose of the cliff
under seismic conditions on both sides (east and west) would be sufficient to provide
a safe area for the rollercoaster during construction and
operation. Furthermore, it was observed that most of the
Results of the post-blasting analysis constituted a potential falling rocks were following a path which moves
strong base for the design of the other bedrock remediation away from the rollercoaster alignment.
 The rockfall trajectory, as predicted from the 3-D
analysis, is shown in Figure 9. The calculated total kinetic
energy and bounce height of the rock boulders for the
proposed fencing system are summarized in Table 3.
These results constituted a strong base for the final design
of the fencing system, as described in section 4 below.

Table 3. Energies and bounce height determined from

post-blasting rockfall anlaysis

Total Kinetic Maximum Kinetic Maximum

Cliff side
Energy (KJ) Energy (KJ) Height (m)
East Cliff 625 1875 6.7
West Cliff 313 939 1.1

4 ROCK PROTECTION DESIGN

Based on the results of the slope stability analysis and the

rockfall analysis discussed earlier, the following
stabilization and protection measures were recommended:
Figure 8: Rockfall path obtained from pre-blasting rockfall
4.1 Stabilization Netting System
analysis
For long-term stability, and to account for the weather and
deterioration of the bedrock at the cliff face, a stabilization
netting (rock anchoring and meshing) system was
proposed to be placed on the east and west face of the
cliff. The primary purpose of the netting system was to hold
potential failure blocks in place and limit the unlikely
rockfall occurrences from impacting the ride foundations.
The netting system should be placed immediately following
the completion of blasting operations and the subsequent
scaling of the blasted face of the cliff.
The netting system was designed following the
guidelines provided in CIRIA 775 (Koe et al. 2018). The
proposed netting system consisted of a mesh with a tensile
strength of up to 250 kN/m in addition to a series of rock
anchors to hold the mesh in place and stabilize/prevent
large block failure. Furthermore, the netting system
required border steel wire ropes along the boundary of the
system, mainly at the top and sides of the mesh. The length
of the rock anchors ranged between 5.0 m and 7.0 m for
the upper stabilized face and the lower stabilized face,
respectively. These assumptions were based on the
material properties of the cliff rocks (table 4).(Koe et al.
2018).

4.2 Rock Catch Fence

Figure 9: Rockfall path obtained from post-blasting rockfall
analysis The rock fall analysis also revealed the necessity of placing
a catch fencing system along the talus to protect the
structure from rockfalls and reduce the risks of damage
3.2.2 Post-blasting Analysis and injuries.
Based on the results of the rockfall analysis, it was
The analysis for post-blasting conditions aims to determine determined that the catch fencing system should be
the suitable rockfall remediation measures for the designed following the parameters presented in Table 5.
rollercoaster both during construction and operation (such Moreover, the location of the fences will be determined
as rockfall catch fencing system) based on the observed based on the final conditions of the terrain and the actual
rock trajectory and the impact energy on the fence. location of the rollercoaster footings.
Table 4. Input parameters used in Rocfall 3 for rockfall 5.2 Monitoring System
analysis
In addition to the field inspections, it was required to install
Parameter Value a vibration monitoring system. Such a system allows
surveillance of the vibrations arising from the blasting
RQD 80
works during construction and monitoring of the bedrock
Jn 9 stabilization and rockfall protection measures throughout
Jr 1.5 the design life of the structure. Prior to installation of the
Ja 0.75 system, a detailed survey of the existing slope should be
Jw 1 completed to identify weak or localized fractures that may
SRF 2.5 be affected during blasting operations. The vibration
monitoring system should include the installation of
Q 7
seismographs at the design depths to automatically report
Average Joint Spacing 6m any vibration that is higher than the specified thresholds.
Persistence 15 m Also, displacement monitors, such as ShapeArrays, were
Unit Weight 24.6 kN/m3 recommended to be installed along the descending and
UCS 20 MPa ascending portions of the rollercoaster. This system
monitors the displacement within the rock face and sends
Joint Friction Angle 35o
an automated report in a specified time range (i.e. monthly)
Design Block Size 4 m3 to ensure that the rock is not being pulled excessively away
Failure Mechanism Toppling and Wedge from the footing location during operations. A remote
monitoring system was also recommended to support the
detection of changes or incipient defects in the netting
Table 5. Energies and bounce height determined from system and the catch fence resulting from the rollercoaster
post-blasting rockfall analysis. operation.

Cliff side Height (m) Length (m) Post spacing (m) 5.3 Performance Testing
East Cliff 6 177 8-12
Performance testing was also required for all stabilization
West Cliff 3 132 8-12
measures to be completed during and after installation.
Among others, it was recommended that a minimum of 5%
of the installed rock anchors should undergo pull tests.
5 INSPECTION AND MONITORING PLANS Such a test ensures that the rock anchor design is
applicable for the expected loads. Suitability testing was
A comprehensive plan including periodic inspections in also required on three rock bolts prior to construction.
combination with real-time monitoring was proposed for Similarly, performance testing on the catch fencing system
this project Considering the geotechnical challenges of the according to the manufacturer’s specifications was also
proposed project in combination with the numerous visitors required.
expected at the subject site during the normal operation of
the rollercoaster, to have a detailed plan in place during
construction and operation of the rollercoaster, to include 6 CONCLUSIONS
periodic inspections, real time monitoring, and testing to be
confirm that construction is being done in accordance with 2-D limit equilibrium analysis and 3-D statistical analysis
the geotechnical requirements of the project and to revise are useful tools for assessing geotechnical challenges
the design if needed. The inspection and monitoring plan associated with large excavations and civil engineering
is intended to provide a safe working environment while design in areas of challenging terrain, when appropriate
ensuring that the bedrock is well protected during and after input properties are provided. The Qiddiya Six Flags
the construction operations, and over the service life of the Project engaged numerical modeling in conjunction with
proposed rollercoaster. geotechnical investigation, visual inspections, and
engineering judgement to assess geotechnical risks
5.1 Field Inspections associated with the design and construction of a record-
breaking rollercoaster in a high, vertical cliff, and to
Field inspections include, but are not limited to, site develop optimized safety measures to ensure the long-
verification of the design blasting layout and ensuring that term durability of the bedrock face during the design
the blast line is located within a safe area, free of excessive service life of the proposed rollercoaster.
bedrock cracking and well prepared for drilling equipment. Through 2-D slope stability analysis, it was possible to
Field inspection is also recommended to include pinpoint geotechnical hazards and simulate possible
inspection of rock protection and stabilization, such as rockfalls that may occur along the cliff during blasting and
pullout testing to verify the performance of the rock scaling operations during the construction phase of the
anchors, inspection of the netting system (rock anchors proposed structure.
and mesh) during and after installation, and inspection of
rockfall catch fence during and after construction.
 Rockfall hazards were simulated using 3-D rockfall Paterson. 2022c. Cliff Area Structures of Qiddiya Six
analysis, and related remediation measures were Flags Project: Existing Investigation, Cliff
designed accordingly. Inspection, and Slope Stability Analysis Report.
Results from 2-D and 3-D modeling were considered to Technical Report PG6503-5. Paterson Group
design the bedrock stabilization approach, and to design Inc.
inspection, monitoring, and testing plans to be conducted Paterson. 2022d. Cliff Area Structures of Qiddiya Six
prior, during, and after the construction phases, as well as Flags Project: Risk Assessment and Methods
during the operation of the proposed rollercoaster. Statement. Technical Report PG6503-3.
Paterson Group Inc.
Paterson. 2023. Cliff Area Structures of Qiddiya Six Flags
7 ACKNOWLEDGEMENT Project: Rockfall Analysis Report. Technical
Report PG6503-7. Paterson Group Inc.
The authors would like to thank Qiddiya Investment Powers, R., Ramirez, L., Redmond, C., and Elberg, E.
company, and Haif Contracting c/o Co-Mender for the 1966. Geology of the Arabian Peninsula. United
opportunity to present this paper. Also, the authors would States Department of the Interior, Geological
like to thank Mr. David Gilbert, Mr. Joey Villeneuve, and Survey.
Mrs. Stephanie Boisvenue from Paterson Group for their Qiddiya. 2023. Six Flags Qiddiya begins design of world’s
input and support. fastest roller coaster. Available from
https://qiddiya.com/en/media/press-office/six-
flags-qiddiya-begins-design-of-world-s-fastest-
8 REFERENCES roller-coaster/.
Rivera, M. 2021. Qiddiya: The Capital of Entertainment in
Abramson, L.W., Lee, T.S., Sharma, S., and Boyce, G.M. Saudi Arabia. Rosen Research Review, 2(3): 10.
2001. Slope stability and stabilization methods. Rocscience. Rocfall3 User Guide. Available from
John Wiley & Sons. https://www.rocscience.com/help/rocfall3/docum
Al-Hadad, M., Siddiqi, G., Al-Zaid, R., Arafah, A., entation.
Neciogh, A., and Turkelli, N. 1992. Seismic Rocscience. 2023. Slide2 User Guide. Available from
hazard and design criteria for Saudi Arabia. In https://www.rocscience.com/help/slide2/docume
Proceedings of the 10th World Conference on ntation.
Earthquake Engineeing Madrid, Spain. pp. 449- Saudi Gazette. 2018. King Salman launches Qiddiya
454. Project Near Riyadh. Available from
Al-Hinai, K.G., Dabbagh, A.E., Gardner, W.C., Khan, https://saudigazette.com.sa/article/533746.
M.A., and Saner, S. 1997. Shuttle imaging radar
views of some geological features in the Arabian
Peninsula. GeoArabia, 2(2): 165-178.
Bamousa, A.O., Daoudi, M., and Hachemi, K. 2020.
Tectono-geomorphic development of an active
transtensional depression: a case study of
Mountain Tuwaiq Plateau, south Riyadh City,
Saudi Arabia. Carbonates and Evaporites, 35(4):
127.
Cripps, K. 2021. Record-breaking roller coaster will travel
more than 155 miles per hour. Available from
https://www.cnn.com/travel/article/world-fastest-
roller-coaster-six-flags-qiddiya.
Habibi, N. 2019. Implementing saudi arabia’s vision 2030:
An interim balance sheet. Middle East Brief, 127:
1-9.
Koe, A., Murphy, W., and Nicholson, R. 2018. Rock
Netting Systems-Design, Installation and Whole-
life Management. CIRIA.
MoMRAH. 2022. Kingdom of Saudi Arabia’s Report on
the Implementation of the New Urban Agenda.
Ministry of Municipal, Rural Affairs & Housing.
Kingdom of Saudi Arabia.
Paterson. 2022a. Cliff Area Structures of Qiddiya Six
Flags Project: Basis of Design Report. Technical
Report PG6503-1. Paterson Group Inc.
Paterson. 2022b. Cliff Area Structures of Qiddiya Six
Flags Project: Geotechnical Feedback Report.
Technical Report PG6503-8. Paterson Group
Inc.
Design and Construction of Rock Anchors
in the Mining Industry
Hafeez Baba & Ljiljana Josic
SNC-Lavalin Inc., Toronto, Ontario, Canada
Jason Wang & Lingam Vaithilingam
SNC-Lavalin Inc., Toronto, Ontario, Canada

ABSTRACT
Rock anchors have been used successfully for a wide range of engineering applications. These have been used for
increasingly higher loads and prestressing conditions, temporary or permanent applications, as a single or a group of
anchors, as vertical or at an angle, on land and over water. The most notable application is for long span suspension bridge
foundations; examples of other applications are retaining walls for deep excavation and tie-down of wind turbines. Mining
structures, be it an onshore process facility or a nearshore marine facility, are typically heavy loaded. Many mines facilities'
design-life keeps on increasing and require upgrades or expansion of the existing structures. These structures or their
upgrades result in significant axial tensile and lateral loads; often requiring the use of grouted rock anchors. This paper
presents design basis and few construction examples related to some mining structures where grouted rock anchors were
used to provide an anchorage system to resist uplift and lateral loads.

RÉSUMÉ
Les ancres de roche ont été utilisées avec succès pour une large gamme d'applications d'ingénierie. Ceux-ci ont été
utilisés pour des charges et des conditions de précontrainte de plus en plus élevées, des applications temporaires ou
permanentes, individuelou en groupe d'ancrages, verticaux ou en angle, sur terre et sur l'eau. L'application la plus notable
est celle des fondations de ponts suspendus à longue portée. Des exemples d'autres applications sont les murs de
soutènement pour l'excavation profonde et l'arrimage des éoliennes. Les structures minières, qu'il s'agisse d'une
installation de traitement à terre ou d'une installation marine côtière, sont généralement très chargées. La durée de vie
théorique de nombreuses installations minières ne cesse d'augmenter et nécessite des mises à niveau ou l'expansion des
structures existantes. Ces structures ou leurs mises à niveau entraînent d'importantes charges de traction axiale et latérale,
nécessitant souvent l'utilisation d'ancrages au roc cimentés. Cet article présente une base de conception et quelques
exemples de construction liés à certaines structures minières où des ancrages au roc cimentés ont été utilisés pour fournir
un système d'ancrage pour résister aux charges de soulèvement et latérales.

1 INTRODUCTION requiring series of upgrades or expansion of the existing

structures. These structures or their upgrades result in
Rock anchors have been used successfully for a wide significant axial tensile and lateral loads, often requiring the
range of engineering applications on land and over water. use of grouted rock anchors as a part of deep foundation
These have been used for increasingly higher load system.
capacities and prestressing conditions. The most notable
application is for long span suspension bridges. Examples
of other applications are retaining walls for deep
excavation, slope stabilization, provide support for marine
structures, and tie down or provide uplift and lateral
resistance for structures such as wind turbines. These
have been used either as a single or as a group for
temporary and/or permanent applications. These can be
placed vertically or at an angle as the design may demand.
These are finding applications in more demanding and
unprecedented situations. Therefore, require knowledge
update, extra care and a high-level engineering judgment
when using these for unprecedented situations.
Mining structures are often heavy loaded, be it an
onshore process facility or a nearshore marine facility.
Marine facilities are usually located in deeper waters to
Figure 1. A Typical Deep Water Mining Related Operation
allow sufficient draft for heavy loaded bulk ore carriers.
Marine Terminal for Illustration Only
Figure 1 shows a typical marine quay wall. Figure 2 shows
a marine quay wall supported by grouted rock anchors.
Authors have experience with soil and rock anchors, on
Many mining facilities design-life keeps on increasing and
land and over water in various industries applications but
this paper presents only grouted rock anchor experiences coupler, centralizer, corrosion protection, bond breaker;
in the mining industry while soil anchor experiences will be grout and grout tube; bearing plate and anchor heads; and
presented at another opportunity. This paper presents few instrumentation, if any. The system can typically be
design, construction, and testing examples related to described as a bar grouted in a predrilled hole also referred
mining structures where high-strength grouted rock to as bond zone/ or length.
anchors were used to provide an anchorage system to
resist uplift and lateral forces. This paper also presents a
conceptual design for a related application where a
relatively large group of high-strength, prestressed grouted
rock anchors are used to provide a robust anchorage
system to resist uplift forces and by prestressing foundation
subgrade to provide enhanced lateral resistance.
Considering its permanent application and longer design
life with no direct access to observe prestress during its
service life; a unique design, construction, and monitoring
features were required.

Figure 3. A Typical Make-up of a Grouted Rock Anchor

(PTI 2014)

The pull out or tensile resistance is generated in this

section. The bar extends to the surface connected with the
anchor head or structure. The extension part may be
unbonded and is also referred to as free stressing length.
For greater lengths of bar, couplings can be used to join
factory bar segments, typically 13 to 15 m length.
Centralizers are used to position the bar inside the drill hole
for uniform grout cover. The anchor head is a device by
which the prestressing force is permanently transmitted
from the prestressing steel to the bearing plate. For some
applications the free stressing length can be significantly
longer than the bonded zone length. If center-to-center
spacing between bond lengths is less than 1.2 m, then
staggering the bond lengths may be adopted.
Rock anchors provide a versatile solution for mobilizing
axial resistance in the direction of its pull or shear
resistance in compression zones when prestressed.
Bonded zone can be placed in competent material at
greater depths and at varying angles. Anchors are equally
suitable for a small working space or low headroom and
can be installed as a part of new deep foundation system
or an existing foundation/structure in a refurbishment
situation. This versatile nature makes rock grouted anchor
a popular alternative for providing solution to mobilize
Figure 2. Shows a Marine Quay Wall Supported by tensile resistance or to provide compression.
Grouted Rock Anchors (Typical Illustration)
3 DESIGN OF ROCK ANCHOR
It is hoped that this paper will provide some better
understanding of the current process of design and A rock anchor may fail in one of the following modes (Refer
construction of the high strength prestressed grouted rock to Figure 4):
anchor application in the mining or other industry. a) Failure of the anchor bar (steel)
b) Failure at the grout/anchor (steel) interface (or bond)
2 MAKE-UP OF A GROUTED ROCK ANCHOR c) Failure at the rock/grout interface (or bond)
d) Failure within the rock mass or rock mass pullout
A typical make-up of a grouted rock anchor with enhanced failure
corrosion protection is shown in Figure 3. The main Each one failure mode is considered in the design. All
elements of a grouted rock anchor system are the bar, are important since the system is only as strong as its
weakest mode of failure. However, it is usually the • Strength and modulus of elasticity of the rock;
grout/rock interface which is the weakest and governs the • Discontinuities in the rock mass, including spacing,
design for widely spaced anchors. Rock anchor design is orientation, and width of bedding planes, joints, and
based on an assumed uniform bond stress along the fractures;
bonded zone of the anchors. Bond length dimensions shall • Minerals in the rock, which may lubricate the bond
be adequate to transfer the design load from the length or reduce the grout strength;
prestressing steel to the ground for the service life of the • Method of drilling and cleaning the drill hole;
anchor. The design of the bond length shall consider the • Drill hole wall roughness;
properties of the ground and the anchor installation • Timing between drilling and grouting in soft rocks;
techniques. The performance of all anchors shall be • In-place strength of grout;
verified by field testing.
• Grouting methods, pressures, and mixture designs;
and
• Bond length.

Typical average ultimate bond strengths (rock/grout)

range from about 0.15 MPa to 3.0 MPa for Marl to Granite.
depending on the rock type. Individual capacities shall be
checked with a rock mass pullout cone and group capacity
checks shall be performed.
Rock mass pullout is an important consideration in the
design of rock foundations. As shown in Figure 4d an
inverted rock cone failure is assumed and further details
elaborated in Figure 5. The position and angle of apex is
usually arbitrary, but the common practice is to assume
conservatively the position at center of the bond zone and
the apex angle of 60o for soft rock and 90o for hard rock.
Rock shear strength on the failure surface of the cone is
either ignore or some reasonable value is assumed. Where
large number of anchors are used at tight spacing and
pullout cones may overlap, the pullout failure, as opposed
to rock-grout bond failure, may govern the design.

Figure 4. Shows Modes of Failure of a Grouted Rock

Anchor (Modified after Brown 2015)

The required bond length is estimated from the

following equation: Figure 5. Detailing a Typical Cone Geometry for Uplift
Capacity Calculation for a Single Rock Anchor (Modified
after Brown 2015)

Steele and grout (concrete) aspects follow the state-of-

the-art analytical means while as the geotechnical aspect
of the current design basis are still at best empirical.

where: 4 INSTALLTION AND TESTING

Lb = bond length
P = design load for the anchor 4.1 Bar Anchors
d = diameter of the drill hole
𝞃u = average ultimate bond strength along interface Bar anchors are usually shop-fabricated in accordance with
between grout and ground project approved drawings and specifications.
FS = factor of safety on average ultimate bond strength
4.2 Drilling and Drill Hole Size
The average ultimate bond strength in rock depends on
the following:
The drill hole diameter shall be sized to allow for the installation challenges. It can be a simple rock/grout
placement of bar (and any instrumentation) and grout interface determination or a comprehensive arrangement
placement around the bar with reasonable grout cover and to determine the load transfer mechanism or any other
corrosion protection to it. Percussion or core drilling design challenges such as creep etc.
methods can be used. Open hole drilling techniques with
water flushing may be used for rock anchors if the drill hole
remains stable. Drilling observations shall include changes
in rock types, drilling rates, loss of flushing medium or more
advanced diagraphy drilling techniques where all these
observations are instrumented, and data profiles logged
and displayed on monitors during installation or for latter
records and analysis.

4.3 Water pressure testing

Water pressure testing of core holes shall be performed at

depths which lie within the planned anchor bond length.
The results are used to assess the probable need to pre-
grout the anchor drill holes. Water pressure typically 35
kPa more than the hydrostatic head and as measured at
the top of the hole. If the leakage from the hole over a 10-
minute period exceeds 10.3 L of water, then the hole shall Figure 5. Shows a Photo Collage Related to a Rock Anchor
be pre-grouted, redrilled, and retested. If the subsequent Test in Progress
water pressure test fails, the entire process shall be
repeated until acceptable results are attained. Sometimes a quick test method such as ASTM D3689:
Procedure A, is used on sacrificial anchors of shorter
4.4 Homing lengths (1 or 1.5 m) in sound rock outside of the footprint
of actual application to determine or verify ultimate
Careful inspection; centralisers; mechanical aid in grout/rock bond strength. Load is applied in increments to
lowering. Care shall be taken to avoid any damage to any failure.
parts of the anchor assembly. For example, for a structural upgrade of an existing
onshore mining facility in Sudbury region where grouted
4.5 Grouting rock anchors were used as a part of a deep foundation
system, a concept similar to the details shown in Figure 6,
The grout mixture used for rock anchors shall be to enhance the uplift resistance of the deep foundation
pumpable, stable and provide required strength. Mixers, system.
storage tanks, and pumps shall have adequate capacity
and shall be sized to allow continuous grouting of an
individual anchor within a period of less than 1 hour.

4.6 Anchorage System and Instrumentation

Bearing plate, nut, any wedges, etc., make the anchorage

system. Care shall be exercised to obtain the continuity of
corrosion protection in the vicinity of the anchorage.

4.7 Stressing

Stressing and testing are required to demonstrate that the

anchor meets the design requirements. Stressing and
testing are commonly applied through hydraulic jacks. A
typical arrangement for a rock anchor test in progress is
shown in Figure 2 and Figure 5. Typically, dial gauges are
used to measure movements and load cells, or
elastomagnetic sensors are used to measure load. Loads
may be sustained for a longer period to check for creep.
The sequence of stressing is applied after the alignment
load (AL) and is typically defined by the project
requirements. Stressing at any time shall not be greater
than 80% of Fpu (ultimate load capacity) of the
prestressing steel. Preproduction tests may be required to Figure 6. Schematic of a Grouted Rock Anchor as a
determine ultimate grout-ground bond strength and Component of a Deep Foundation System – FHWA 2000
 A large number of anchors were installed on this
Many grouted rock anchors, some details shown in project. Performance tests are conducted on selected
Figure 7, formed part of the deep foundation system. production anchors (typically 2% of installed anchors).
Preproduction test on 64 mm threadbar Dywidag anchor These are conducted by incrementally loading and
grouted 1.5 m in sound rock to 150 mm diameter hole unloading the anchor in accordance with the project
resulted in an ultimate bond strength of about 2100 kPa requirements. Proof tests are carried out on all production
(design assumption was 2000 kPa). This facility had very anchors not subjected to a performance test to 1.33 of the
tight headroom and space restrictions; for production design loads. Load is applied in increments and maintained
anchors, shorter length bars were coupled together to for certain durations.
produce the required bar lengths.
Canadian Foundation Engineering Manual (CFEM 5 QUALITY CONTROL
2006) refers to a preproduction test as a design test or as
a design verification test. It is a cyclic loading pattern and The quality control of anchor system starts with the quality
sequenced to twice the design load with sustained loading control of materials, during installation and testing. A mil
at certain loads. As an example, a typical test result from a certificate of bars, couplers, plates etc is required for the
nearshore mine application where a 64 mm diameter bar, anchors. A diagraphy drilling is preferred where one can
Grade 80 anchor with about 3 m in rock bond of about 200 record: drilling rates, thrust levels, flush pressure, and
mm diameter and about 40 m of free length are presented circulations, any losses, torque etc. Drill hole alignment
in Figures 8 and 9. The design was based on CFME 2006 tolerance, tightness, and quality control requirements for
guidelines, assuming an ultimate bond of 730 kN/m grout mix and during grout installation and are usually
(Granite or Basalt rock type). defined in the specifications.

Figure 8. Load – Displacement Curve for Anchor Test

Figure 7. Shows a Photo Collage of Grouted Rock Anchor

Installation as a Part of Micro-piling System for a
Foundation Refurbishment Project for a Mine Process
Plant in Northern Ontario

One of the preproduction tests presented above

indicated it did achieve the assumed bond capacity and
movement criteria up to 125% design load but marginal
increase in displacements beyond that load. While most of Figure 9. Load – Displacement Curve for Anchor Test
the other tests met both these criteria.
5.1 Acceptance Criteria

Typically, three groups of acceptance criteria are required

to satisfy: creep, movement, and lock off loads. Typical
creep sources are movements at rock/grout or steel/grout
interface; and stress relaxation in steel. Residual
movement increases with load and is typically larger in soil
than in competent rock. All these criteria are detailed in
PTI 2014.

6 LOAD TRANSFER MECHANISM

Several authors have questioned the assumption of

uniform bond stress along the bonded zone of the anchors.
They have tried to use nonlinear relationships along the
ground–grout interface to analyze the behavior of ground
anchors based on elastic-plastic analysis subjected to
tensile loads. For anchor bond zones soils operating in
tension, the initial load is resisted by the upper grout body
with negligible load diffusing to the bottom of the anchor.
As the load is incremented, strain along the grout-soil
interface may exceed the soil peak strain or ultimate Figure 11. Typical Axial Pile Load Transfer-Displacement
interface bond strain, which results in lowering of the bond (t-z) Curves – API 2000 (presented for Illustration only)
stress along this length and a shift in the peak bond stress
towards the end of the anchor as illustrated in Figure 10.
and fatigue is often a critical design factor. Tiikkainen
2017, attempted finite element analysis for static and
dynamic loading, including fatigue under cyclic loadings
for wind turbine foundations supported on rock anchors.
Rock anchor forces and rock compression profile were
presented but the strain gauge profile was not available.
This would need further work to conclude its application.

8 AGING AND AGING MANAGEMENT

The long-term stress sustainability of prestressed grouted

rock anchors constitutes an important aspect of the
Figure 10. Mobilization of Bond Stress for a Tension stability of the structure retained by these structural
Anchor after FHWA 1999 members. The loads supplied by the anchors must be
For example, Bryson and Giraldo 2020, attempted to always sufficient so that the stability of the structure is not
develop more realistic load-displacement models using an compromised.
approach of t-z, similar to well-known practice for pile load- Mbenmokrane and Gerardballiv 1991, reported five-
displacement assessment, see Figures 11 and 12. Strain year monitoring of tensile load losses on nine prestressed
gauges were installed in the anchor bonded length and cement-grouted rock anchors, with varying bond lengths in
from the load–displacement data collected; the load- sound rock and free lengths. These were experimental set
transfer mechanism was investigated numerically using ups; two such results are reproduced here as Figure 13. It
load-transfer models that include softening behavior. Also, was observed that the load variations versus time make up
strain gauge data were used to assess the long-term two phases: phase I, lasting about 6 months, throughout
variation of the loads along the bonded length of the which the losses of load are rapid; and phase II, which is
anchors. This exercise was limited to the shale stratum but marked by a slower and uniform rate of load loss. The
was not conclusive as the testing was not continued to overall losses are less than 10%.
ultimate failure.
Possible aging affects could be stress relaxation and
corrosion of the anchors. Both mechanisms are being
7 DYNAMIC LOADING AND FATIGUE addressed in the design to ensure successful long-term
performance of rock anchors.
Structures like wind turbine foundations are exposed to
hundreds of millions stress changes during their lifetime
 9 INSTRUMENTATION

Figure 14 presents some typical load measurement

instrumentation for anchors. Measurement of loads on
individual anchors may be performed using load cells,
elastomagnetic sensors, liftoff tests. Specially designed
anchor heads will be required to allow later lift-off readings
and load adjustments.

Figure 14. Typical Load Measurement Instrumentation for

Anchors
In experimental set ups or actual applications strain
gauges within the anchor profile and extensometers within
the rock profile have also been used.

10 PRESTRESSED APPLICATION

There are many situations where the free stressing zone

is compressed by prestressing the anchors to enhance the
shear resistance of compressed zone. It usually comprises
a large group of anchors, a conceptual formation is shown
Figure 12. Analysis of Preproduction Tests: (a) Load in Figure 15. Sometimes, these are permanent
Transfer Models; (b) Measured and Calculated Response applications with longer design life with no direct access to
(Bryson and Giraldo 2020) observe anchor heath or its prestress condition during its
service life. Therefore, requiring a robust corrosion
protection, improved construction, special design
allowances, and better instrumentation and monitoring
features.

Figure 15. A Conceptual Prestressed Grouted Rock

Figure 13. Changes in Anchor Prestress Load Anchor Application to Improve Shear Resistance
 4th Edition. Canadian Geotechnical Society.
3 CONCLUSIONS Bryson L.S. and Giraldo J. R. 2020. Analysis of case study
presenting ground anchor load-transfer response in shale
stratum. Can. Geotech. J. Vol. 57, 2020.
Rock grout anchors provide a versatile solution for Tiikkainen, H. 2017. Analysis of Experimental
mobilizing axial resistance in the direction of its pull or Measurements on the Rock Anchored Wind Turbine
shear resistance in compression zones formed by Foundation. University of Oulu, Degree Programme of
prestressing. Bonded zone can be placed in competent Mechanical Engineering Master’s thesis
material at greater depths, at varying angles, on land or in
water. Anchors are more adapt for a small working space API 2000. American Petroleum Institute, Recommended
or low headroom or over water and can be installed as a Practice for Planning, Designing and Constructing Fixed
part of new shallow or deep foundation system or through Offshore Platforms—Working Stress Design. API
an existing shallow or deep foundation systems in a Recommended Practice 2A-WSD (RP 2A-WSD) 21st
refurbishment situation. This versatile nature makes rock Edition, December 2000.
grout anchors a more popular alternative for providing a FHWA 1999. U.S. Department of Transportation, Federal
solution to mobilize resistance or to provide compression. Highway Administration. Geotechnical Engineering
Rock grouted anchors find their application in multiple Circular No. 4, Ground Anchors and Anchored System.
loading scenarios with increasing load and prestress Publication No. FHWA-IF-99-015, June 1999.
capacities. These are used in mining and other demanding
industrial applications with great confidence. There is
continuous improvement in materials, installation, testing,
monitoring, and performance. This paper demonstrated
some examples of successful onshore and nearshore
applications in mining industry. There has been a lot of
progress in establishing load-displacement model similar
to t-z model used for pile capacity (load-displacement)
assessment; however, more work is needed to cover more
soil and rock types for better understanding and wider
application.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the contribution of

several individuals to this paper from the SNC-Lavalin Inc.
offices, our clients, and general industry partners without
their assistance, this paper would not have been possible.

REFERENCES

PTI 2014. Recommendations for Prestressed Rock and

Soil Anchors. Post-Tensioning Institute, PTI DC35.1-14
Brown, E. T. 2015. Rock engineering design of post-
tensioned anchors for dams – A review. A Journal of Rock
Mechanics and Geotechnical Engineering
B. Brahim and G. Ballivy, “Five-year monitoring of load
losses on prestressed cement-grouted rod anchor,”
Canada Geotechnique, vol. 28, pp. 668–677, 1991.
ASTM D3689. Standard Test Methods for Deep
Foundations Under Static Axial Tensile Load. ASTM 2007.
FHWA 2005. U.S. Department of Transportation, Federal
Highway Administration. NHI Course No. 132078,
Micropile Design and Construction Reference Manual.
Publication No. FHWA NHI-05-039, December 2005.
CFEM 2006. Canadian Foundation Engineering Manual,
Design And Construction of A 9-Meter-
High Embankment Over Champlain Sea
Clay
Gohe A Metaferia 1, Louis D’Amours 2
Company Name, City, Province/State, Country
1 SNC Lavalin Inc, Ottawa, ON, Canada
2 SNC Lavalin Inc, Montreal, QC, Canada

ABSTRACT
In Eastern Canada, dealing with normally, or minimally over-consolidated clays from the Champlain Sea deposit often
demands innovative approaches. This study focuses on a case where a 9-meter-high embankment was needed for a fast-
paced P3 transit project in the Canadian Capital. The project faced tight scheduling, space constraints, and rather strict
settlement limit tolerances. Investigating the site's silty clay deposit overlying bedrock revealed a lower over-consolidation
ratio (OCR) than initially estimated, affecting its behavior. The selected solution involved rapid consolidation of the clay
using vertical wick-drains and surcharging. This enhanced the clay's strength, supporting a retaining structure. Additionally,
lightweight EPS embankment fill minimized settlements. Geotechnical monitoring aligned closely with predictions,
validating the approach. Despite challenges from the sensitive clays, the project successfully managed embankment
construction, meeting performance criteria through innovative geotechnical interventions and careful monitoring.

Dans l'Est du Canada, les argiles normalement ou peu sur-consolidées de la mer de Champlain exigent souvent des
approches novatrices. Cette étude porte sur un cas où un remblai de 9 mètres de haut était nécessaire pour un projet de
transport ferroviaire dans la capitale canadienne. Le projet était confronté à un calendrier serré, à des contraintes d'espace
et à des tolérances de limites de tassement plutôt strictes. L'étude du dépôt d'argile silteuse du site recouvrant le
substratum rocheux a révélé un taux de sur-consolidation (OCR) inférieur à celui initialement estimé, affectant son
comportement. La solution retenue consistait en une consolidation accélérée de l'argile à l'aide de mèches drainantes
verticales et d'une surcharge. Cela a renforcé la résistance de l'argile, supportant une structure de soutènement. De plus,
le remblai en EPS a davantage minimiser les tassements. Les données d’instrumentation géotechnique ont étroitement
aligner avec les prévisions, validant l'approche de conception. Malgré les défis posés par les argiles sensibles, le projet
fut un succès grâce à des interventions géotechniques innovantes et à un suivi attentif.

1 INTRODUCTION characteristics of the silty clay deposit, through a

comprehensive geotechnical investigation campaign. 2)To
In present day Eastern Ontario and Quebec region, outline the design process, including the application of
known as the Ottawa River and St Laurence Valleys, laid a accelerated consolidation techniques using vertical wick-
vast shallow sea over thousands of years ago Geologically drains and surcharging, coupled with the strategic use of
recognized as the Champlain Sea. This vast area has since lightweight EPS embankment fill to manage settlements
seen the sea parted and deposits of the Champlain Sea effectively. 3) To validate design assumptions and predict
Clay, or colloquially known as Leda clay occupy a ground response during construction using Finite Element
considerable portion of the overburden stratigraphy. modeling (Plaxis2D) and a comprehensive geotechnical
This deposit is infamous for its high sensitivity to instrumentation and monitoring program (GIMP).
remolding and its rather low pre-consolidation pressures.
Leda Clay has been deemed responsible for numerous 2 PROJECT BACKGROUND AND SETTING
infrastructure failures, and some of the most devastating
landslides in the region’s history. While it would be SNC-Lavalin was retained to carry out the geotechnical
preferable avoid such geology all together, it is however design for the Trillium Line Extension Project (the Project)
important to understand, master its defining geotechnical in Ottawa, ON.
characteristics to better inform and optimize the design and The Project is currently in construction to extend the
construction of infrastructure bearing overtop it. existing OC Transpo line to the proposed Limebank Station
The primary objective of this paper is to present a case (near the intersection of Limebank Road and Earl
study detailing the geotechnical design and construction Armstrong Road). The total length of the proposed
approach undertaken by SNC-Lavalin for the Trillium Line extension is approximately 15 km, which includes an
Extension Project in Ottawa, ON. The paper seeks to approximate 4 km Airport spur to the existing Ottawa
address the challenges posed by the presence of Macdonald-Cartier International Airport (OMCIA).
Champlain Sea Clay and highlights innovative and efficient An approximately 9-meter-high approach embankment –
solutions applied to optimize the project's schedule and The Embankment at Limebank Road – constitutes part of
accommodate right-of-way constraints. Specifically, the the East approach to the Limebank rail Bridge. This 9 m
objectives are as follows: 1) To provide a detailed analysis high embankment, at the structure, gradually tappers down
of the subsurface conditions, particularly the consolidation over about 300 meters stretch to reach the existing grade.
The embankment is retained by a Mechanically Stabilized
Earth (MSE) wall to the North and sloped to the South.

3 SUMMARY OF SUBSURFACE INVESTIGATION

An initial geotechnical baseline geotechnical data report

provided by the project indicated that in general, the
overburden encountered at the site consists of a silty clay
deposit underlain by a granular till overlaying Sandstone
bedrock of the March Formation.
In-situ testing was completed to obtain geotechnical
parameters such as undrained shear strength and N(SPT).
Similarly, 3 oedometer tests were carried out on samples
from three boreholes: RS-20 located 60 m south of the
station, RS-17 at 75 m north and RS- 15A at 300 m east of
the East Abutment). These boreholes are shown (red
marker) in Figures 1 and 2.

Figure 3 Effective and Pre-Consolidation Pressures Vs

Elevation _ Baseline Data

Figure 1 Investigation site plan Subsequent supplementary investigation completed by

SNC Lavalin, including sCPTu sounding, higher frequence
in-situ undrained shear strength measurements using the
Nilcon vane, further uni-axial consolation, and battery of
laboratory index testing campaign revealed considerably
tighter over-consolidation margins.
Figure 4, below, illustrated the interpreted Effective as well
as Pre-consolidation pressures versus elevation.

Figure 2 Stratigraphic Cross-section

The Pre-Consolidation Pressures (PCP) obtained from the

odometer tests are respectively 160, 285 and 190 kPa.
These odometer tests results showed that the clay layer is
in an over consolidation condition with ratios ranging from
OCR=3,4 to 4,5 - this translates to an over-consolidation
margin (OCM) of about 125kPa, as shown in Figure 3,
below where a summary of the information interpreted for
the site that shows the effective pressure at the site
alongside the estimated pre-consolidated pressures are
presented. Figure 4 Effective and Pre-Consolidation Pressures Vs
Elevation _ Supplementary Data

Pre-consolidation pressure of clay deposit is one of the key

properties of Champlain Sea Cla. The state of practice for
investigations for such a clay dictates that the minimum
value of the pre-consolidation pressure of the clay deposit
must be established. For this purpose, the usual state of
practice is the provide a clear shear strength profile with
values at every 1 m (or less) and then to take a
representative intact sample at the minimum shear
strength values level to get a sample from such a layer to
conduct an Odometer test (Standard Test Methods for One
Dimensional Consolidation Properties of Soils). The data The site at Limebank embankment is characterized by the
obtained in such a process will give us the minimum value presence of compressible silty clay layer with minimal
of clay’s pre-consolidation pressure. The data obtained margin between effective and pre-consolidation pressures
from this minimum value and odometer test are used to as illustrated earlier.
estimate the magnitude of total settlement and also the An evaluation of the consolidation parameters across the
time of consolidation. These 2 parameters are essential to full stretch of the embankment show that the clay at the site
design the bridges, buildings and mainly the approach is over-consolidated with governing margins ranging from
embankment fill to bridges. These parameters are of key about 50kPa to 65kPa at an approximate elevation of about
importance in the design of engineered structures and the 89 masl to 90 masl. The approach fill was expected to
evaluation of their performance. (ASTM D2435) induce a peak pressure of about 180 to 200 kPa on the
native soils.
The data gap identified in the baseline information is that
process explained earlier was NOT applied and instead,
the shear strength was measured seemingly sporadically,
and samples of intact clay submitted to consolidation test
were NOT targeted at the governing lowest values as state
of the practice dictates it in such a Champlain sea clay
deposit.

The design pre-consolidation pressure (PCP) envelope is

further optimized along the guideway covered under the
scope of the current design report by utilizing location
specific data where appropriate. It is evident that the silty
clay layer encountered presents the same governing pre-
consolidation pressure of about 60 kPa at an elevation of
about 89.25 m to 89.50 m. That said, the pre-consolidation
pressure envelop varied slightly through the stretch of the
approach embankment.

4 PROPOSED INFRUSTRUCTURE AND DESIGN

An approximately 9-meter-high approach embankment –

The Embankment at Limebank Road – is planned at the
East approach to the Limebank structure. The present
report covers the stretch of the guideway from STA 16+900
to 17+500, where the embankment height is at its
maximum of about 9 m at the East abutment (~17+210)
and decreases to about 0 m around STA 17+500. Due to
property limit restraints as well adjacent roads, a
combination of MSE walls and slopes are proposed to
retain the embankment fill on the north side. Figure 5 Additional loading as a result of embankment
construction

This pressure will stress the silty clay layer past its highest
historical stress – the pre-consolidation pressure – thus
triggering consolidation settlements, in addition to
recompression settlements. This pressure would produce
significant long-term settlement in the order of 850 mm to
950 mm primarily due to the consolidation of the clay layer
near the highest embankment height. A gross estimation of
total settlement under the centerline of the embankment is
shown below.
Figure 6 - Gross settlement estimation over time

Consequently, an optimized accelerated consolidation

surcharging program was designed implemented. This
entailed in the installation of vertical wick-drain in a
triangular pattern spaced at 1.2m and extending to base of
the near normally consolidated silty clay deposit. The
ground response in terms of settlement, pore water
pressure and strength gains overtime were modeled
through Finite Elements using Plaxis2D and later
confirmed with a of purpose-built geotechnical
instrumentation and monitoring program.

These models were reproduced at set intervals along the

embankment stretch to capture the ground response at
varying embankment heights, as well as we changing
ground conditions.

5 GEOTECHNICAL INSTRUMENTATION AND

MONITORING PROGRAM

As stated earlier, given that the clay thickness

under Limebank is of about 6 m to 10 m, pre-loading the
clay with wick drains (installed in
the silty clay layer) in order to have a reasonable
consolidation time was implemented. The primary purpose
of wick drains is to shorten drainage distance and
consolidation delays. It is important to plan the
embankment construction to avoid sudden over-
pressurizing of the area and triggering global instability. As
such, Bbar should be limited to 0.8 to ensure stability
(Leroueil et al., 1977). Fill placement should be placed
such that this threshold is not surpassed. SCs and VWPs
are specified to be installed at the following locations.
Consequently, the fill placement was monitored both
visually by means of regular survey reports as shown
below, and a network instruments.

Figure 10 GIMP monitoring data

Following the near perfect validation of the FE Plaxis 2D

modeling results through the GIMP monitoring data, the
design of the final embankment, including the retaining wall
on the norther face was confirmed, and constructed.

6 EMBANKMENT AND RETAINING WALL DESIGN

The right of way limitation North of the project alignment
dictated that a vertical retaining structure be incorporated
into the approach embankment as depicted below. that
Figure 7 Embankment construction survey monitoring being said, the less than favorable strength parameters of
the soft clay deposit posed challenges to the retaining wall
design and required advanced analyses to ensure the
requisite Factors of Safety for external stability in particular
where satisfied. The final embankment design leveraged
heavily the consolidation and settlement analyses
discussed earlier and had to further deepened the
approach by examining the enhanced strength parameters
of the clay deposit as a result of consolidations beyond the
PCP limits as detailed by Mesri (1989).

Figure 8 GIMP instruments layout

Figure 11 Original conceptual design

Additionally, a design change in the geometry of the

embankment as shown below allowed further optimization
to “distance” the bearing pressures from the surface of clay
deposit by elevating the retaining wall:

Figure 12 Optimized embankment geometry.

Moreover, an added design constraint at the east abutment

was the presence of a diamond rail crossing supported on
Figure 9 GIMP control unit the approach embankment. Such crossings are known for
their stringent requirements as it relates to both settlement, comprehensive geotechnical instrumentation and
and elastic deflections. As such, the retained design monitoring program (GIMP) provided a robust foundation
solution for the embankment directly east of the bridge was for decision-making throughout the project. The near-
conceived to virtually eliminate any concern of settlement perfect correlation between the monitoring data and the
by incorporating the use of EPS – a lightweight material modeling results instilled confidence in the final design and
weighing less than 1 kn/m3. Such innovative design embankment construction, ensuring a stable and safe
ultimately allowed for a near null net additional loading on structure.
the silty clay deposit. A deeper look at the EPS mechanical
properties was also completed to ensure the material The lessons learned from this case study can be invaluable
offered sufficient stiffness under dynamic loading to for future projects facing similar challenges involving
mitigate deflections during rail traffic. The use of EPS was sensitive clays and time-sensitive schedules.
extended to a comfortable offset from the location of the Understanding the defining geotechnical characteristics of
diamond cross where impacts from settlement of the the Champlain Sea clay and adopting innovative and
remainder of the embankment was eliminated, as shown in adaptive design solutions are crucial to the successful
the below figure. execution of infrastructure projects in the region.

Overall, this case study exemplifies how a collaborative

and multidisciplinary approach, coupled with cutting-edge
geotechnical investigation techniques and state-of-the-art
monitoring systems, can lead to the effective and safe
construction of complex infrastructure in challenging
geological settings. The successful completion of the 9-
meter-high embankment at Limebank Road serves as a
testament to the ingenuity and expertise of the engineering
team involved in this remarkable project.

Figure 13 EPS Embankment configuration

7 CONCLUSION
In conclusion, the case study presented in this paper
showcases a successful and comprehensive approach to
address the challenges posed by construction on sensitive
clays of the Champlain Sea deposit in Eastern Canada.
The project's location in the Canadian National Capital
Region and its association with a multi-billion-dollar transit
initiative demanded innovative and efficient solutions to
optimize schedule, accommodate right-of-way constraints, Figure 14 Embankment currently under active use
and meet stringent performance criteria. (dynamic rolling stock testing)

The geotechnical investigation campaign carried out by

SNC-Lavalin played a crucial role in providing accurate
data on the subsurface conditions, especially the over-
consolidation ratio (OCR) of the silty clay deposit. The
identification of a lower OCR value than initially estimated
required the team to rethink and tailor their design solutions
accordingly.

The design process involved a combination of accelerated

consolidation through vertical wick-drains and surcharging,
complemented along with the strategic use of lightweight
EPS embankment fill to manage settlements effectively.
The approach successfully mitigated potential long-term
consolidation settlements, meeting the specified
thresholds. The implementation of a Mechanically
Stabilized Earth (MSE) wall and slopes on the northern
face of the embankment to accommodate right-of-way
constraints demonstrated a balanced and versatile design
approach.

The validation of design assumptions and predictions

through Finite Element modeling (Plaxis2D) and a
Field Research on Soil-Structure Interaction
of Sustainable Basement Using Mass Timber
Panels
Fernanda B. T. Imamura, Yuxiang Chen, Lijun Deng, Ying Hei Chui
Department of Civil and Environmental Engineering – University of
Alberta, Edmonton, AB, Canada

ABSTRACT
Resilient and mechanically stable cross-laminated timber (CLT), popular in mid-rise and high-rise buildings, is promising
in basement construction for many advantages over concrete. The reinforced concrete basement is Canada's most popular
solution in low-rise construction. However, these basements are known to have performance issues with the propagation
of cracks and moisture migration problems. Wood-based foundations can provide a sustainable alternative solution for
basement construction once mass timber produces lower carbon emissions than concrete in the life cycle. Before
proposing this new basement material, the geotechnical aspect of mass timber basements should be examined. An
experimental building with a CLT panel basement was completed at a cohesive soil site in Edmonton. The basement was
3 m by 6 m in plan and 2 m deep. The site profile consists of glaciolacustrine clay, with a more clayey composition at
shallower depths. In-situ monitoring instrumentations were extensively used to measure the earth pressure, CLT deflection,
timber strain, and soil temperature and moisture content. This paper presents the details of a field experiment development
proposed to assess the soil-structure interaction of a CLT basement. Earth pressure cells show measurements of lateral
earth pressure values ranging between the at-rest lateral and active earth pressure. Strains measurements are affected
by the changes in the relative humidity and the wall volume change due to moisture content changes.

RÉSUMÉ
Le bois lamellé-croisé (CLT) résistant et mécaniquement stable, populaire dans les immeubles de moyenne et grande
hauteur, présente des perspectives prometteuses dans la construction de sous-sols en raison de nombreux avantages
par rapport au béton. Le sous-sol en béton armé est la solution la plus populaire au Canada dans la construction de faible
hauteur. Cependant, ces sous-sols sont connus pour avoir des problèmes de performance liés à la propagation de fissures
et aux problèmes de migration d'humidité. Les fondations à base de bois peuvent offrir une solution alternative durable
pour la construction de sous-sols, d'autant plus que le bois massif produit moins d'émissions de carbone que le béton au
cours de son cycle de vie. Avant de proposer ce nouveau matériau de sous-sol, l'aspect géotechnique des sous-sols en
bois massif devrait être examiné. Un bâtiment expérimental avec un sous-sol en panneaux CLT a été achevé sur un site
de sol cohésif à Edmonton. Le sous-sol mesurait 3 m sur 6 m en plan et avait une profondeur de 2 m. Le profil du site se
compose d'argile glaciolacustre, avec une composition plus argileuse à des profondeurs moins profondes. Des instruments
de surveillance in situ ont été largement utilisés pour mesurer la pression du sol, la déflexion du CLT, la contrainte du bois,
ainsi que la température et la teneur en humidité du sol. Cet article présente les détails d'un développement d'expérience
sur le terrain proposé pour évaluer l'interaction entre le sol et la structure d'un sous-sol en CLT. Les cellules de pression
du sol montrent des mesures de valeurs de pression latérale du sol variant entre la pression latérale au repos et la pression
latérale active. Les mesures de contraintes sont influencées par les variations d'humidité relative et les changements de
volume du mur dus aux modifications de la teneur en humidité.

1 INTRODUCTION conditions, and propensity to defective construction

(CMHC 2007a, CMHC 2007b).
Concrete has been the dominant material used in the Recent advances in mass timber products (MTP) have
basement construction of low-rise residential light wood opened the door to sustainable alternate solutions to
frame buildings in North America due to its durability and replace concrete to overcome deficiencies and reduce
strength. However, frequent performance problems are carbon footprints possibly. Mass timber refers to panelized
associated with concrete basements, such as dampness, wood construction products developed from smaller pieces
cracking, and high thermal loss. Alternative basement of wood, and the most noticeable product of this kind is
systems have been developed and used commercially. cross-laminated timber (CLT) (Kordziel 2018).
One example is the Permanent Wood Foundation (PWF) CLT may pose a solution to basement construction due
(ANSI/AWC 2015, CSA 2016, CMHC 2007b). PWF is a to its availability in large dimensions, high strength-to-
light wood foundation system, structurally made of lumber weight ratio and excellent dimensional stability compared
framing and sheathing, intended to support low-rise with traditional wood products. Cross-laminating layers
buildings. Despite the inclusion of the PWF basement in make CLT of dimensional lumber into a thick structural
building codes, these systems have not been widely panel using a durable adhesive, as shown in Figure 1, and
accepted as a replacement for the traditional concrete are fabricated commonly in three-ply (105mm), five-ply
basement due to various challenges, such as higher cost, (175mm), seven-ply (245mm) and nine-ply (315mm)
lack of resiliency to deal with a wide range of soil thickness.
 movement and the pattern of these movements
(translation, rotation, flexure) (Clayton et al., 2013).
For rigid retaining walls, earth pressure and water pressure
can be calculated using theoretical solutions from
Rankine's or Coulomb's earth pressure theories, which
predict a linear earth pressure distribution behind the
retaining wall. These methods provide limited information
on the distribution and magnitude of lateral earth pressures
induced by different wall displacements (Sadrekarimi and
Damavandinejad, 2013).
The earth pressure distribution becomes more
complicated for flexible walls since it depends on wall
deflection and soil deformation; in this case, an "apparent
earth pressure" distribution may be adopted for
design (CFEM, 2006). The notion of "apparent earth
pressure" was presented by Terzaghi and Peck (1947)
when observing flexible walls. Field measurements on full-
scale walls and model tests helped develop a back
calculation for the earth pressure distribution on these
Figure 1. Cross-Laminated Timber layering (from walls, demonstrating that loads are more concentrated
https://www.structuremag.org) near the support in a flexible wall, depending on the soil
type. The design of flexible retaining walls relies on these
CLT developed in central Europe about twenty years ago theoretical earth pressure distributions. However, these
as a substitute for concrete slabs and tilt-up walls in low- diagrams may be unrelated to displacement control needs
rise buildings. Still, it has been used in tall wood buildings (Boone and Westland, 2006).
worldwide for the last ten years, mainly as a load-bearing Long-term field experiments conducted to measure
wall and floor system. FPInnovations published a earth pressure are limited, especially when dealing with
comprehensive handbook on CLT's production process, structures that are not impactful, such as bridges. Among
applications, performance and design in 2011 and updated the long-term field studies, it is common to find publications
it in 2019 (FPInnovations 2019). related to bridge abutments or robust and tall retaining
CLT foundation walls can be categorized as PWF due walls. Huntley and Valsangkar (2013) show findings from
to similar intended usage and demands, offering benefits the first three years of monitoring the lateral earth pressure
over traditional basement systems, such as better material of an integral-abutment bridge in New Brunswick.
insulation property, ease of prefabrication, and possibly Instruments were installed to monitor the bridge and
lower environmental footprints. A preliminary assessment measure foundation movements, backfilling response and
conducted by Daneshvar et al. (2022) on the possible good temperature. The measured earth pressure was affected
impacts of using CLT to replace concrete structures by the ambient temperature changes with variations in the
revealed that the global warming potential, representing magnitude of the earth pressure. In summer, the lateral
carbon dioxide emission, can be reduced by almost 80% in earth pressures increase as the abutment wall moves
a 50-year lifespan. toward the backfilling, and in winter, it decreases as the
Designing earth-retaining structures, such as wall contracts.
basements, requires knowledge of the earth load and earth Analytical, numerical and experimental methods have
pressure distribution acting on these structures. Limitations revealed that wall displacement significantly affects the
in the existing methods to estimate the lateral earth magnitude and distribution of static earth pressure.
pressure may include an assumption of a lateral earth However, this area is still poorly understood because such
pressure distribution that may not correspond to the actual a problem is one of the most complicated soil-structure-
distribution and lateral earth pressure states, which can interaction problems (Bakr,2018).
change depending on the wall deformation. These A review of the literature and practice suggests the
methods also disregard the effects of soil-structure following gaps:
interaction and construction systems. Many factors can 1. There was no prior large-scale field investigation of
affect the earth pressure behaviour, such as the retaining CLT basement construction.
wall material and construction type, the backfill soil type 2. Earth pressure distribution on CLT walls has never
and the seasonal temperatures. been investigated.
This paper will present a field experiment developed to
2 LITERATURE REVIEW assess CLT basements on a large scale. The soil-structure
interaction will be observed in an actual boundary
The earth pressure acting on a soil-retaining structure condition. The broader objectives are to understand the
depends on the lateral deformation of the soil. Unless the lateral earth pressure distribution and magnitude to guide
deformation conditions can be estimated accurately, no the future design of CLT basements, understand how CLT
theoretical solution can accurately predict the earth retaining walls interact with cohesive soils and the impact
pressure distribution. Many factors can affect the of seasons on the basement wall deformation.
magnitude and pattern of the earth pressure, such as how
the wall is placed, the way this wall moves, the amount of
3 MATERIALS AND METHODS 3.1.1 Structural details

An evaluation of this variety of parameters and interactions Daneshvar et al. (2022) described the preliminary
would be difficult to conduct in a laboratory setup, so a field structural calculation to determine the thickness of the CLT
experiment was proposed. The field experiment allows the panels for field development. The thickness of these panels
investigation of soil-structure interactions under actual and the number of laminations were hand calculated
boundary conditions. The design and basement according to the backfilling height and soil pressure. The 3-
construction will be shown in the following topics. ply panel was selected in spandrel orientation, measuring
105 mm thick (3 laminations measuring 35 mm each).
3.1 Field testing design details Hand calculations estimated the self-tapping screws
(choice of connections) spacing, quantities and length.
A large-scale field test at the University of Alberta south
campus farm in Edmonton, Alberta, was designed to have 3.1.2 Wall envelope
dimensions compatible with a house basement. The
chosen dimensions were 6 m (20 ft) in length, 3 m (10 ft) The wall assembly was designed to protect the CLT
deep and 2.4 m (8 ft) in height, with 1.8 m (6 ft) of backfilling basement against moisture. The granular layer provides a
to simulate an actual scale basement construction. Six CLT preferable path for water to flow. The drain connects to a
panels measuring 3 m x 2.4 m were ordered. Figure 2 sump pump. Each wall was built with a different water
shows the CLT basement design cross-sections. barrier solution: Test Wall 1 was protected with a plastic
dimpled membrane, Test Wall 2 was covered with a self-
adhesive asphalt membrane, and the two narrow walls that
are not instrumented were protected with a fluid
membrane, that turns into a rubbery membrane when dry.
A 7.60 cm (3") thick XPS rigid insulation was placed outside
the CLT panels. Figure 4 shows each wall's water
protection, insulation and drainage details.

Figure 2. CLT basement design - Cross-sections.

Dimension unit: m.

The actual basement during construction is depicted in Figure 4. CLT Wall Envelope Design
Figure 3.
3.2 Geotechnical site investigation

The site profile consists of glaciolacustrine clay, with a

more clayey composition at shallower depths. The in-situ
test results indicate a 1.0 m thick layer of silty organic clay
soil overlying a uniform clay layer up to 5 m thick. Beyond
this layer is a 1.5 m thick silty clay layer, overlaying a 1.5
m thick silty sand layer. Underlying this material is silty
sand with interbedded clay layers until the end of the
borehole test, at 10 m deep.
From the CPTu results, illustrated in Figure 5,
parameters such as the qt (tip resistance), fs (sleeve
friction) and u2 (pore pressure), other soil parameters can
be estimated. Pore pressure measurements were obtained
by the dissipation test conducted in the sandy layers of the
soil profile. From the dissipation test data, the groundwater
table (GWT) is estimated to be 5.3 m below ground. The
Figure 3. CLT basement during construction (photo taken undrained shear strength (Su) was calculated by
in December 2021) performing unconfined compressive strength (UCS) tests,
vane shear tests on Shelby tube samples and using a Table 1. Sensors for the field investigation on soil-wall
semi-empirical relationship to correlate Su from CPTU interaction
tests. The undrained shear determined by the vane shear
test shows values between 50 kPa and 110 kPa. Sensor Quantity Type
On average, the specific gravity of solids (Gs) was 2.8
Soil Water Content 6 CS655
at a 2.7 m deep. The bulk unit weight varied between 17.25 Reflectometer Probe
kN/m3 and 19.2 kN/m3. In situ, moisture contents were
Thermocouples 32 Type T
determined at depths between 0.60 m and 2.75 m. The
moisture content generally increased linearly with depth, Relative Humidity 17 HM1500LF-ND
reaching an average value of approximately 34% at 2.75 Strain Gauges 16 PFLW-30-11-6LJCT-F
m. Estimated parameters will support the field experiment 120-Ohm Resistors 16 4WFBS120
results in the discussion. Soil Pressure 4 KDE-200 kPa
Transducers

(a) (b) (c) (d)

Figure 6. Sensor location in cross-section - Test Wall 1: a)

CS655 probes, b) Thermocouples and Relative humidity,
Figure 5. Soil's undrained shear strength profiles from
c) Earth pressure cells, d) Strain gauges
CPTu tests
4 RESULTS AND DISCUSSIONS
3.3 Field experiment instrumentation
The data presented in this section were collected between
In-situ monitoring instrumentations have been planned to
measure the earth pressure using soil pressure June 16, 2022, and August 2, 2023 (0 to 410 days). The
preliminary data was measured from April to June 2022,
transducers (earth pressure cells), wall temperature and
but signals were unstable. The data acquisition system was
relative humidity to interpret strain measurement induced
only finalized on April 2022, meaning that the strain values
by the CLT panel moisture content, wall strain using strain
do not reflect the loading "zero" condition once loading
gauges for use in wood materials, and soil temperature and
during construction could not be measured. No changes in
moisture content using soil water content reflectometer
the load were made after construction, but a large
probes. The data acquisition system comprises one
settlement was observed on May 2022, mainly because
CR3000 mini logger, a CR310 datalogger and six
backfilling was done during the winter and was not
AM16/32B, 16 or 32 channel multiplexers with remote data
compacted.
collection. Sensor reading was collected every 30 minutes.
Data from earth pressure cells, strain measurements
Table 1 shows the selected sensors and the total number
from strain gauges, temperature and relative humidity of
of sensors installed for both test walls.
CLT panels will be used to interpret the soil-structure
Test Walls 1 and 2 counts on the same number of
interaction. Data from soil water content volumetric probes
sensors and sensor locations. Figure 6 illustrates the
will provide input on soil temperature and in-situ moisture
sensor's positions in cross-section, highlighting the depth
soil moisture content.
of each sensor.
From January 7, 2022, 198 days after the data
collection started to March 16, 2022, 266 days after the
data collection began, heating was provided inside the
basement. The temperature and relative humidity changes
of CLT panels posed by the heating source are relevant to
the data interpretation.

4.1 Earth pressure cells

As part of the soil-structure interaction analysis, a pair of

Type KDE-200 kPa, a dual-diaphragm soil pressure
transducer, 50 mm in diameter, were installed at each test
wall to measure the soil lateral earth pressure. After the
CLT panel installation, insulation placement and roof
construction, still with an open excavation, the earth
Figure 8. Profile temperature for Test Wall 1, at 1.5 m
pressure cells were secured to the external rigid insulation
below ground level
layer. After the sensor installation, backfilling was started.
The presented data were collected after the settlement
The heating source inside the basement seems to affect
occurrence. Figure 7 illustrates readings from the earth
soil temperatures between 210 and 270 days.
pressure cells installed on Test Wall 1. The estimated
active lateral pressure from the theories at 0.5 m is -90 kPa,
4.3 Moisture content
and at 1.5 m is – 170 kPa. At rest, lateral earth pressure
was estimated at 6 kPa at 0.5 m and 18 kPa at 1.5 m.
Understanding the CLT panel's relative humidity and
However, on average, the measured lateral earth pressure
moisture content is vital to estimate the portion of strain
is -0.17 kPa at 0.5 m and -0.03 kPa at 1.5 m.
affected by the change in wood moisture. In the basement,
The earth pressure cells were installed in contact with
temperature measured at the panel surface and relative
a 7.6 cm thick rigid insulation that is part of the basement
humidity measured at 2.5 cm inside the CLT panel were
protection envelope. The effects of having this layer may
used to estimate the panel moisture content. The soil
be reflected in the earth pressure cell's results. It is
probes measure temperature and volumetric water content
expected a reduction in the lateral earth pressure acting on
in the soil. From that measurement, the soil gravimetric
the all, with that inclusion, according to Azzam &
moisture content was determined. Figure 9 shows the
AbdelSalam (2015). However, this effect was not evaluated
indoor air and inside the panel relative humidity. The
for this paper.
estimated CLT panel moisture content and soil moisture
content are illustrated in Figure 10.

Figure 7. Test Wall 1 lateral earth pressure measured at Figure 9. Test wall 1 relative humidity measured 2.5 cm
1.5m below ground level inside the CLT panel, at 1.5 m below ground level.

4.2 Temperature profile

Several temperature measurements were made in different

locations in the CLT basement. The most important are the
internal and external CLT panel surfaces measured by
thermocouples. The outdoor air temperature data was
collected from the South Campus weather station, and the
water content reflectometer probe measured the soil
temperature. Figure 8 shows the temperature profile
measure at Test Wall 1 at a depth of 1.5m below ground
level. Observe the indoor heating effects on the
measurements from 198 to 266 days. A power outage
occurred at approximately 240 days, but it quickly Figure 10. Test Wall 1 moisture content at 1.5 m below
recovered. ground level.
4.4 Strain gauges installation of the primary sensors was finished, and stable
data collection started on June 16, 2022.
Strain values were corrected to the wood thermal Based on the available field monitoring data, the
expansion and seem highly affected by the panel's relative readings from the Earth Pressure Cells:
humidity, moisture content and internal temperature. The • Measured lateral earth pressure (σh) values sit
effects of the relative humidity in wood strain must be between the calculated at-rest lateral earth pressure
carefully analyzed, which is why an additional laboratory (σh0) and the calculated active lateral earth pressure
test calibration was conducted to understand the strain (σha).
range caused by the changes in the relative humidity. A • The installation location may impact the
curve "strain versus relative humidity" was prepared from measurements once the presence of rigid insulation
the laboratory calibration, illustrated in Figure 11. A curve- could reduce the applied pressure to the CLT wall.
fitting method was used as in Equation 1: Results from strain gauges suggest that:
• The relative humidity effects on strain have caused
𝑆𝑚𝑐 = 0.05 (𝛥𝑅𝐻)3 + 1.45 (𝛥𝑅𝐻)2 + 15.6 𝛥𝑅𝐻 + 3.83 significant axial strain on the walls.
• The moist-induced strain follows the swelling and
where Smc is the estimated strain in με, and ΔRH is the shrink pattern of the wall and the relative humidity
incremental relative humidity in %. trend.
Equation 1 was used to estimate the portion of the • The mechanically induced strain may indicate that
measured strain related to the changes in the CLT panel when the wall swells, the wall volume change makes
moisture content reflected by the changes in relative the panel move toward the soil, showing negative
humidity. strain measurements on the internal surface of the
CLT wall. When the wall shrinks, the reduction in
volume moves the panel toward the basement's
interior, showing positive strain measurements.

6 ACKNOWLEDGEMENTS

The project was funded by the Alberta Innovates. We

appreciate the technical support of Dr. Haitao Yu,
Landmark Group of Companies. We also thank the
collaboration of other researchers who supported the field
investigation and lab work in various stages: Mohammad
Fathi, Mahboobeh Fakhrzarei, and Mohammad
Rezvanpour.

Figure 11. Strain gauge calibration to relative humidity 7 REFERENCES

The moist-induced strain was calculated using the data Azzam, S. A., and AbdelSalam, S. S. 2015. EPS Geofoam
corrected to the wood thermal expansion and the to reduce lateral earth pressure on rigid walls.
laboratory calibration equation. After that, the mechanically International Conference on Advances in Structural and
induced strain was estimated. Figure 12 shows the three Geotechnical Engineering, ICASGE'15, Hurghada,
strain measurements. Egypt.
ANSI/AWC. 2015a. Permanent Wood Foundation Design
Specification with Commentary. Leesburg, VA.
ANSI/AWC. 2015b. NDS National Design Specification
(NDS) for Wood Construction. Leesburg, VA.
Bakr, J. A. 2018. Displacement-based approach for
seismic stability of retaining structures. Ph.D. thesis,
School of Mechanical, Aerospace and Civil
Engineering, The University of Manchester, UK.
Boone, S. J., and Westland, J. 2006. Design of excavation
support using apparent earth pressure diagrams:
consistent design or consistent problem? Proceedings
of the 5th ISSMGE, Amsterdam.
CGS (Canadian Geotechnical Society). 2006. Canadian
Foundation Engineering Manual. Vancouver, BC.
Figure 12. Test Wall 1 estimated strain for the internal
Clayton, C. R. I., Woods, R. I., Bond, A. J., and Milititsky,
face of the CLT panel at 1.5 m below ground level
J. 2013. Earth pressure and Earth-retaining Structures.
CRC Press, Taylor and Francis Group, Boca Raton, FL.
5 CONCLUSIONS
CMHC. 2007a. Monitored thermal performance of ICF
walls in MURBs. Research Highlights 65863. Canada
A 6 m by 3 m large-scale CLT wall basement was designed
Mortgage and Housing Corporation, Ottawa, ON.
and constructed. The field experiment development of the
CMHC. 2007b. Preserved wood is a potential source of
arsenic and fungi in the foundation walls of homes.
Research Highlights 65891. Canada Mortgage and
Housing Corporation, Ottawa, ON.
CSA (Canadian Standards Association). CSA S406-16.
2016. Specification of permanent wood foundations for
housing and small buildings. Canadian Standards
Association. Toronto, ON.
Daneshvar, H., Fakhrzarei, M., Imamura, F., Chen, Y.,
Deng, L., and Chui, Y. H. 2022. Structural Analysis and
Design of Sustainable Cross-Laminated Timber
Foundation Walls. Buildings, 12 (7): 979.
FPInnovations. 2019. CLT Handbook. Special Publication
SP-53. Vancouver, BC.
Huntley, S. A., and Valsangkar, A. J. 2013. Field monitoring
of earth pressures on integral bridge abutments.
Canadian Geotechnical Journal. 50 (8): 841-857.
Huntley, S. A., and Valsangkar, A. J. 2018. Nine-year field-
monitoring data from an integral-abutment bridge.
Proceedings of the IFCEE conference.
Kordziel, S. 2018. Study of Moisture Conditions in A Multi-
Story Mass Timber Building Through the Use of
Sensors and WUFI Hygrothermal Modeling. Master of
Applied Sciences Thesis, Colorado School of Mines,
Golden, CO.
Sadrekarimi A., and Monfared S.D. 2013. Numerical
Investigation of The Mobilization of Active Earth
Pressure on Retaining Walls. International Conference
on Soil Mechanics and Geotechnical Engineering,
Paris.
Evaluation of Creep in a Normally
Consolidated Glacio-Lacustrine Clay
Mohammad Kermani, Michel Demers Bonin
WSP Canada Inc., Montréal, QC, Canada
Marielle Limoges Shaigetz, Edouard Masengo
Agnico Eagle Mines Ltd., Toronto, ON, Canada

ABSTRACT
It is recognized that glacio-lacustrine varved clay in Abitibi region of Quebec, Canada presents a moderate level of strain-
weakening that needs to be considered in the design of embankments over these clays. Moreover, failure under sustained
shear load, i.e., creep-induced failure, has been identified as a potential failure mechanism. The literature suggests that
for a material that is susceptible to strain weakening, any mechanism that causes deformations large enough to bring the
material to instability could lead to a failure. In this paper, it is suggested that advanced laboratory testing, field monitoring
data, and stress-deformation modeling could provide a framework to judge whether creep in the normally consolidated
clayey foundations of embankments could lead to a failure. In this study, the containing structures of a tailings storage
facility (TSF) of a mine site located in Abitibi region were studied and this framework was examined.
For the studied dikes and their upstream raises, undrained shear tests in the field and laboratory on the normally
consolidated clay showed some level of brittleness, which implied that the clay under the dikes and the tailings pond is
susceptible to strain weakening under undrained loading. The monitoring data along with the construction and deposition
stages suggests that creep and consolidation were taking place at the same time. The coupled stress-flow numerical
models using FLAC software also indicate that the clay behavior under construction was partially drained. Therefore, clay
is expected to also behave partially drained during creep. Thus, both field data and numerical models indicate that in the
stress path space, the clay should not approach the instability line while undergoing creep. Therefore, creep failure was
judged to be unlikely.

RÉSUMÉ
Il est reconnu que l'argile glacio-lacustre varvée dans la région de l'Abitibi au Québec, Canada, présente un niveau modéré
d’écrouissage par déformation, qui doit être pris en compte dans la conception des remblais sur ces argiles. De plus, la
rupture sous une charge de cisaillement constante, c'est-à-dire la rupture induite par le fluage, a été identifiée comme un
mécanisme de rupture potentiel. La littérature suggère que pour un matériau sensible à l'anti-écrouissage, tout mécanisme
qui provoque des déformations suffisantes pour amener le matériau à l'instabilité peut conduire à une rupture. Il est
suggéré que les indications provenant d'essais de laboratoire avancés, les données de surveillance sur le terrain et la
modélisation contrainte-déformation pourraient fournir un cadre pour évaluer si le fluage dans les fondations constitués
d’argile normalement consolidé pourrait conduire à une rupture. Dans cette étude, les structures de confinement d’un parc
à résidus miniers situé dans la région de l'Abitibi ont été étudiées.
Des essais de cisaillement non drainés sur l'argile normalement consolidée, sur le terrain et en laboratoire ont montré un
certain niveau de fragilité, ce qui suggère que l'argile sous les digues et sous le parc à résidus est susceptible à l’anti-
écrouissage sous l'effet d'une charge dans les conditions non-drainée. Les données de surveillance recueillies au cours
des phases de construction et de déposition suggèrent que le fluage et la consolidation ont lieu en même temps. Des
modèles numériques couplés contrainte-écoulement utilisant le logiciel FLAC calibrés sur les données de terrain ont
également indiqué que le comportement de l'argile en cours de construction est partiellement drainé. Par conséquent, on
s’attend à ce que l'argile se comporte de manière partiellement drainée pendant le fluage. Ainsi, les données de terrain et
les modèles numériques indiquent que dans l'espace de cheminement des contraintes, l'argile ne s'approcherait pas de
la ligne d'instabilité pendant le fluage. Ainsi, la rupture due au fluage a été jugée peu probable.

1 INTRODUCTION failure of clay should be identified as a possible failure

mechanism.
The Abitibi gold belt located in Québec, Canada is home In this study, a literature review was carried out on the
for several mine sites with tailings storage facilities (TSF), creep phenomena, and creep failure in fine grained plastic
as well as waste rock and overburden stockpiles, water soils. The ultimate objective was to evaluate the behavior
retaining dikes, and road embankments. At certain mine of the foundation clay at two TSF dikes of a mine site, that
sites, these structures include large embankments are constructed over up to 25 m of glaciolacustrine varved
constructed over foundations partly constituted of glacio- clay, and to assess the influence of creep on the
lacustrine varved clay. This material shows a moderate performance of these containment structures. In particular,
level of strain weakening. This implies that creep-induced the impact of creep on the stress-strain-strength
characteristics (strain weakening) of the foundation
comprised of cohesive soils was examined and it was negative pore water pressure along planes with greatest
evaluated whether this occurs under drained or undrained shear stress and strain. With time, water migrates into
condition. The assessment could finally shed some light on these zones and causes weakening and strength
whether creep deformations could lead to loss of strength decrease. This seems to have been the cause for failure
and failure in these particular cases. in several cuts in heavily overconsolidated brown London
This paper summarizes the literature review, provides clay 40 to 70 years after the excavation. However, this
a framework for evaluation of creep behavior of mechanism is unlikely to occur in the case of
embankments on soft clays, and applies it to TSF dikes of embankments on clays.
a mine site. As the focus is on large embankments on
cohesive material particular attention will be paid to 3.2 Drained vs. Undrained Creep
normally consolidated clays.
It is important to recognize the difference between drained
2 DESCRIPTION OF THE SITE and undrained creep.

The gold mine is located in the Abitibi green belt geological 3.2.1. Drained creep
region in the Province of Quebec, Canada. The studied site
is located within the footprint of the former glacial lake It is generally accepted that drained creep of a
Barlow-Ojibway. Tailings are discharged in a TSF that was compressive soil leads to further densification of the
partially built on a varved glacio-lacustrine clay. This material and gradually leads to an increase in its shear
stratigraphic unit is susceptible to generate shear-induced resistance. However, for materials with cementation and
pore water pressure when subjected to a deviator stress, bonding, creep strains could lead to breakage of the bonds
and possibility susceptible to strain-weakening, i.e. a and consequently loss of strength, and possibly a brittle
decrease in shear strength as the clay undergoes shear failure.
strain. On the other hand, as explained above, in heavily
Throughout the years, a series of geotechnical overconsolidated dilative clays (e.g., London Clay) an
investigations including borehole sampling, Field Vane increase in the water content during drained creep might
testing, (Seismic) Cone Penetration Tests with porewater lead to a loss of strength.
pressure measurements (CPTu and SCPTu), and
installation of monitoring instruments were conducted at 3.2.2. Undrained Creep
the site. Several clay samples were collected at the site
using thin wall tubes, Laval sampler, and advanced Figure 1 taken from Mitchell and Soga (2005) describes
laboratory testing including triaxial and direct simple shear how undrained creep may lead to a reduction in strength in
tests were carried out. The results of some of these tests a normally consolidated clay. The stress path from 0 to 1
are used for this study. shows the path that a normally consolidated clay might
undergo during undrained loading; e.g., construction of an
3 A BRIEF LITERATURE REVIEW ON CREEP embankment. The path from point 1 to 2 shows undrained
RUPTURE IN SOILS creep in a contractive material, which leads to an increase
in the pore water pressure. 1 to 2’ shows drained creep,
3.1 Creep Rupture in Soils which leads to a decrease in the pore water pressure until
the stress state reaches the drained loading stress path.
Creep is defined as continued deformation of materials Once sheared again to failure, both tests end up on the
under sustained load. In geotechnical engineering, what is same φ’ line, given that there is no mineralogical, chemical
usually meant by creep is continued straining under or biological change. However, from 2 to 3, compared to 2’
constant deviatoric stress. Therefore, for instance if the to 3’ a smaller change in deviatoric stress is needed to
pore water pressure changes without any changes in the reach the failure. In other words, undrained creep of a
total stress components (constant deviatoric stress) the contractive material leads to strength loss.
resulting strain is still called creep. The same phenomenon can happen when time-
Besides the fact that creep deformations may lead to dependent shear deformation in adjacent soil causes shear
displacements beyond the tolerable displacements of a deformation in clay.
structure, the strength and the stress-deformation behavior On the other hand, by evaluating different stress state
of a soil may change because of creep. conditions under an embankment, Leroueil et al. (2001)
Mitchell and Soga (2005) indicate that a soil may argue that there is no evidence that creep deformation in
undergo strength loss due to creep in the following undrained conditions under an embankment load leads to
circumstances: a reduction in strength. In other words, in their experimental
• If a significant portion of the strength is due to study, the undrained shear strength for two specimens with
cementation, creep deformations may cause failure of similar void ratio (the void ratio remains constant during
cemented bonds, and lead to strength loss. undrained creep) was the same, no matter the state of
• Shear causes changes in pore pressure during effective stress before shearing.
undrained deformation, and changes in water content These observations should be valid only for normally
during drained deformation. consolidated clays that do not show strain weakening and
• In heavily overconsolidated clays, undrained creep should not be generalized to all clays.
deformation may cause dilation and development of
 Also, soil strength in terms of effective stresses does where creep rate is high and decreases with time, called
not change unless there are chemical, biological, or primary creep; the subsequent period, in which the strain
mineralogical changes during the creep period. Thus, the rate is constant with time, called secondary creep or steady
critical state line remains the same before and after creep. state creep; and the third stage is tertiary creep, during
Similar conclusion can be made from studies by Leroueil et which strain rate increases until failure occurs (creep
al. (2001) and Mitchell and Soga (2005). rupture). When the applied deviatoric stress is considerably
smaller than the material’s shear resistance, only the
primary creep takes place. As the deviatoric stresses get
closer to the peak shear strength of the material, secondary
and tertiary creep stages can occur.

Figure 2. Definition of Creep Stages in a Triaxial

Apparatus: a) Strain versus Time, b) Log (strain rate)
versus Log (time). from Augustesen et al. (2004).

It could be concluded that, by monitoring the strain rate

Figure 1. Effects of Undrained Creep on the Strength of vs. time under sustained shear stress level in the field, one
Normally Consolidated Clay (from Mitchell and Soga, can determine if the clay is undergoing primary or
2005) secondary creep. Secondary creep should be taken as an
alert.
In summary, the studies on undrained creep in normally It is important to note that the time-dependent deformation
consolidated clay are mainly concentrated on ductile of clay under an embankment could be due to creep or
materials. There is no clear agreement on whether the consolidation. Therefore, monitoring of pore water
undrained creep leads to a reduction in the shear strength. pressure is an essential component to judge whether the
However, it is generally accepted that the critical state line ongoing deformations are due to creep, consolidation or a
would not change due to undrained creep. It is also combination of both.
generally accepted that undrained creep can eventually
lead to failure. Many efforts have been deployed during the 3.4 The Onset of Creep Failure in Overconsolidated
late 1960s until present to quantify the creep deformations Clays
and try to predict eventual failure. This is more elaborated
in the following sections. Lefebvre (1981) published the results of eight triaxial
isotropically consolidated drained (CID) creep tests under
3.3 Different Stages of Creep as Observed in Triaxial reconsolidation pressure of 5 to 25 kPa lasting up to 100
Tests days on an overconsolidated Nicolet marine clay. Figure 3
shows stress-strain curves both for monotonic loading and
To study the time dependent deformation of earth materials creep tests. He observed that samples that have been
through triaxial compression tests, the axial and confining loaded past the post-peak envelope eventually failed with
pressures are kept constant, and the axial and lateral time. By referring to this study, as well as the works of
strains are measured with time. In such experiments, Leroueil (1998) and Hunter and Khalili (2000), Fell et al.
usually, attention is paid to the shear stress level (σ1/σ3) (2000) argue that creep failure can occur at stresses
dependency of the strain rate. While in oedometer testing smaller than the peak strength, possibly at any stress
the deformation in constant stress is called secondary greater than the fully softened or critical state strength.
compression, in triaxial compression tests it is called creep Note that Nicolet Clay is a sensitive marine clay, and
(Augustesen et al. 2004). Contrary to secondary generalization of the creep behavior observed by Lefebvre
compression, the creep process can lead to failure (1981) to normally consolidated Glacio-lacustrine clay
because of the boundary conditions and the principal should be done with caution. Also, these tests were done
stress ratio. in drained conditions under small confining stresses, trying
Figure 2 shows a typical strain versus time curve (in to simulate slopes in clay.
linear scale), and axial strain rates versus time curve (in Mesri and Shahien (2003) analyzed 99 case histories
log-log scale) in a triaxial creep test. In triaxial tests, of slope failures in 36 soft clays to stiff clays and clay
depending on stress level and material properties, creep is shales. They back calculated mobilized shear strength in
considered to take place in three stages: The initial period, first time slope failures as well as reactivated slides. In most
cases the back-calculated mobilized strength was in fact at
or near the fully softened condition. As implied from consolidated clay shows a certain level of strain
Skempton (1970), the fully softened strength is equal to the weakening. Therefore, in this article it is assumed that the
peak strength of a normally consolidated clay. For the creep failure can occur possibly at any stress level greater
reactivated slides, the residual strength was mobilized. than the fully softened strength in normally consolidated
They cite from Lefebvre (1981) and Mesri and Abdel- glacio-lacustrine clay.
Ghaffar (1993) that for Eastern Canada clays, the For slopes in intact and fissured clay, it is generally
mobilized strength in first-time slides is near the large- accepted that the accumulation of creep strains with time
strain strength. It can be concluded from Mesri and results in a reduction of the strength with time (Tavenas
Shahien (2003) work that creep failure could happen at and Leroueil 1981). For the case of embankments on clays,
stresses greater than fully softened or residual strength, normally, the clay consolidates at the same time as creep
similar to what is implied from Lefebvre (1981). deformations occur. Therefore, two mechanisms act
against each other.

Figure 3. Stress Strain Curves for monotonic and Creep

Test Results for Nicolet Clay (Source: Lefebvre 1981.)

Skempton (1964), as reported by Tavenas and Leroueil

Figure 4. Stress Conditions for Undrained and Drained
(1980) also mentioned that the failure in natural slopes in Creep Tests (adapted from Tavenas et al. 1978)
fissured clays is governed by the residual shear strength of
the soil. Therefore, the same conclusion as Lefebvre
3.6 Concluding Remarks of the Technical Literature
(1981) could be drawn from Mesri and Shahien (2003) and
Review
Tavenas and Leroueil (1980) studies: for slopes in clay, if
the shear stress is greater than the residual strength, Figure 5 shows stress-strain curve for a soil that shows
failure could happen.
strain weakening and loss of strength after peak. Point A is
On the other hand, Morgenstern and Tchaenko (1967)
the beginning of shearing from the in-situ state of stress,
and James (1971) indicate that the mobilization of residual
and point D being the state at large strains (residual or
strength takes very large displacements and development critical state). Three mechanisms could lead to the loss of
of a continued failure surface. They conclude that the strength from point A to D: further shearing (A-D by
residual state cannot exist in first-time failure.
monotonic loading path), dynamic loading (A-B-D), and
Figure 4 shows the stress state for several triaxial creep
creep (A-C-D). Therefore, the possibility of failure due to
tests carried out by Tavenas et al. (1978). Also shown in creep becomes a question of brittleness, stress level,
this figure is the limit state curve for t = 1 day (i.e., the locus allowed deformation, and whether the deformations
of states of stresses for which failures happened sooner
exceed a certain level of strain, or whether a discrete shear
than 1 day past the application of the load). The red arrows,
surface is developed.
that are added to the graphic by the authors, show the
stress states at which creep rupture has occurred in creep
tests carried out on the same soil. It appears that all the
creep tests that lead to creep failure were done at stress
states very close to the zone defined for the limit state for t
= 1 day. Thus, these results on overconsolidated marine
clay appear to be different from other studies presented in
this section.

3.5 The Onset of Creep in Normally Consolidated

Clays
Although it is expected that creep failure should be less
probable in normally consolidated clays under the load of
an embankment, the conclusion of Lefebvre (1981) on
overconsolidated clays should be considered applicable to Figure 5. Mechanisms that Could Lead the Loss of
normally consolidated clays as well, in case the normally Strength in Soil (adapted from Olson and Stark 2003)
According to Fell et al. (2000), the threshold shear strain 5.1 Geotechnical Shear Tests Data
level for failure due to creep to occur is the shear strain
associated with the peak strength in a monotonic shear In the absence of creep tests, monotonic shear tests on
test. normally consolidated clay shed some light on the strain
weakening behavior. Data on direct simple shear (DSS)
4 STRAIN WEAKENING AND BRITTLENESS laboratory testing, triaxial laboratory testing, and field vane
tests were available for the clay in the studied site. Below
Strain weakening is defined as a material reaching a peak are the main outcomes from those tests.
in the stress-strain curve and then a reduction in strength • DSS tests: Data from DSS testing carried out on the
as it goes towards the steady state. clay shows that the normally consolidated clay shows an 𝐼𝐵
Strain weakening can be quantified through the up to 12% (𝑆𝑡 =1.14).
brittleness index: • Triaxial tests: Undrained triaxial compression tests on
𝑆𝑢𝑝𝑒𝑎𝑘 −𝑆𝑢𝑝𝑜𝑠𝑡−𝑝𝑒𝑎𝑘
𝐼𝐵 = [1] the studied clay in normally consolidated state show an 𝐼𝐵
𝑆𝑢𝑝𝑒𝑎𝑘
as high as 26% (𝑆𝑡 =1.35).
or sensitivity:
𝑆𝑢𝑝𝑒𝑎𝑘 Both in triaxial and DSS tests, the brittleness of
𝑆𝑡 = [2] normally consolidated clay tends to increase with the
𝑆𝑢𝑝𝑜𝑠𝑡−𝑝𝑒𝑎𝑘
increase in consolidation stress.
Depending on the application, the post-peak strength • Field vane tests (FVT): FVTs carried out on normally
could refer to the remolded strength. The higher 𝐼𝐵 or 𝑆𝑡 is, consolidated clays under the dikes shows an 𝐼𝐵 as high as
the higher is the post-peak strength loss. ICOLD (2022) 41% (𝑆𝑡 =1.7). In case of FVT, post-peak strengths could
mentions that although this ratio represents one significant be reached at very large deformations, and 𝐼𝐵 values from
aspect of brittle behavior (i.e., the magnitude of the post- field vane tests are not necessarily comparable to those in
peak strength loss), it does not address the strain aspect the laboratory. Also note that the concept of weakening due
(i.e., the rate of strength loss). to creep for stresses greater than the residual strength
ICOLD (2022) mentions that clays with plate-like (Lefebvre 1981) is developed around laboratory test data.
minerals (e.g., kaolinites, montmorillonites) can undergo In summary, laboratory and field data show a certain
strain weakening under both drained and undrained level of strain weakening, which implies that the material is
shearing conditions; but due mostly to other phenomena susceptible to strength loss due to creep.
than those occurring in cohesionless soils. Strain
weakening of such clays occurs as clay particles reorient 5.2 Field Data
during shearing, creating polished surfaces along the shear
surface. In drained shearing of overconsolidated clays, the The deformation trends observed in clay layer through the
diminishing shear resistance is caused by the development inclinometers helped determine zones of shearing. The
of continuous bands of strongly orientated particles to the relative horizontal displacement of the top and bottom of
residual strength. In undrained conditions, the shear- the identified layers divided by the thickness of the layers
induced pore water pressure plays the primary role in the provided an estimation of shear strain. The shear strains
shear resistance loss. evolved through time, were evaluated in parallel with the
Historically, it has been presumed that a normally data on the construction (rate of rise and geometry),
consolidated clay would not show strain weakening deposition of tailings, and pore water pressure and lateral
behavior (Mitchell and Soga 2005; Leroueil 2001). This is displacements measurements. Periods with no
generally accepted in the literature for ‘’textbook’’ clays. construction or deposition of tailings were chosen to
However, direct simple shear (DSS) and triaxial evaluate the creep deformations. As explained in section
compression laboratory tests, as well as field vane tests on 3.3, secondary creep is identified with a constant rate of
normally consolidated Barlow-Ojibway clays show a strain deformation. Therefore, signs of steady rate of strain,
weakening behavior, possibly due to structuration and which could indicate secondary creep, were sought.
bonding developed during consolidation in the laboratory
and in the field. Therefore, it is legitimate to assume that Figure 6 shows an example of monitoring data for 18
normally consolidated clay is also susceptible to the loss of months (6b), with a 6-month period with no construction or
strength due to creep. deposition (6a). The shear strains were calculated based
on inclinometer readings and considering the thickness of
5 PROPOSED APPROACH FOR ASSESSMENT OF the identified clay layer with large shear strain. Because of
THE SUSCEPTIBILITY TO CREEP FAILURE the noises in the deformation data, the trends are studied
rather than punctual data.
Considering the literature on creep failure, including the The shear strain curves from the field data could not be
three stages of creep, the concept of strain weakening and directly compared to the laboratory curves in Figure 2. The
the creep failure threshold, as well as drained vs. laboratory curves show the strain and strain rates after the
undrained creep, it is proposed that results from laboratory end of a shearing stage after an isotropic or anisotropic
and field shear testing, field monitoring data, as well as consolidation stage. In the field, however, consolidation
numerical modeling could allow assessing the potential for and loading take place in parallel continuously. Therefore,
creep failure. The following sub-sections summarize data engineering judgement is required to distinguish the effect
collected at the site studied and discusses whether these of creep.
clues indicate creep failure is judged probable. A dashed gray line is included in
 Figure 6b which represents a constant shear strain rate. The coupled stress-deformation elasto-plastic models
The clay shear strain continues to increase at a decreasing developed for the dikes and the TSF using FLAC software
rate, and the pore water pressure in clay constantly were calibrated based on the laboratory testing results and
decreases with time. Therefore, the deformations during field vane shear tests, i.e., that it includes the potential for
this period could be mainly attributed to consolidation that strain weakening of the clay but not its viscosity. The
followed the increase in shear and confining stresses models were capable of matching the pore water pressures
during loading stage. Although creep may also be taking and deformations observed in the field. This indicates that
place, a decrease in pore water pressure under constant partial drainage of clay was occurring during construction
shear stress implies that the material is moving towards a and deposition. As the clay behaves partially drained
more stable state of stresses. Also, clay would be during construction, it is expected that it should also
undergoing primary stage creep, as the rate of strain is behave partially drained during creep.
reducing with time. From another point of view, undrained creep tests in the
laboratory include an undrained shearing phase followed
by an undrained creep phase. On the other hand, at the
studied site, it is accepted that (from instruments and
stress-deformation model) the shearing phase before the
creep phase is partially drained. Therefore,
• as the loading is partially drained, it is expected that
creep cannot be fully undrained.
• Even if the creep happens undrained, it is not realistic to
expect the same behavior in the field as the undrained
creep tests in the laboratory.
Figure 7 shows the effective stress paths during
construction and deposition for different locations in the
clay foundation. The stress paths move along a line
showing that the stress ratio does not increase with
construction. Therefore, the clay does not approach, and
moves away from the failure line during the construction
and deposition. It is unlikely that creep deformations move
(a) the stress state towards the instability line. Note that the
models did not take into account the clay’s viscous
behavior. However, they can provide a reasonable
estimation of stress path and porewater pressure
generation and dissipation.

Figure 7. Effective Stress Paths Computed in the Clay Unit

under the TSF.
(b)
6 CONCLUSIONS AND RECOMMENDATIONS
Figure 6. Construction, deposition, and pore water
pressure history at a dike, along with the shear strain In this study, a review of the creep phenomenon in clay and
In general, the displacement data showed that after the its theoretical basis was presented. Then, clues from a
construction and deposition periods the strain in the clay mine site located in Abitibi region of Québec were
continues to increase, but at a smaller rate than during the presented and it was discussed if creep failure is likely to
loading. One important observation is that since the happen.
installation of the instruments, in general, the pore water It is important to recognize that one of the limitations of
pressure in the clay is constantly decreasing. This implies the literature review was that laboratory testing that
that the deformation of the clay during and after the investigates creep is mainly done on overconsolidated
application of the load is partially drained. undisturbed clay. Also, most of the studies discuss creep
behavior in a natural slope or a cut in overconsolidated
5.3 Numerical Models clays, which is different from the case of an embankment
constructed on clays at stresses that brings the clay in a
normally consolidated state.
 It could be implied from studies on creep in clays that Ladd, C. C. (1991). Stability evaluation during staged
creep failure could occur at any stress higher than the fully construction. Journal of geotechnical
softened or residual strength. In other words, strain engineering, 117(4), 540-615.
weakening due to creep could lead to failure. Additionally, Lefebvre, Guy. (1981) Fourth Canadian Geotechnical
clues from laboratory testing on the clay show that normally Colloquium: Strength and slope stability in Canadian
consolidated glaciolacustrine clay is susceptible to strain soft clay deposits. Can. Geotech. J., 18, 420-442
weakening (i.e., it shows strain weakening in stress- Leroueil, S. (1998) Elements of time-dependent
deformation behavior in shear tests). Therefore, it was mechanical behavior of overconsolidated clays. Proc.
concluded that, creep failure is a legitimate case to 51stCanadian Geotechnical Conf., Edmonton, Vol.2:
consider in the stability assessment of the dikes on clay. pp.671-677.
The field data showed that the clay is being Leroueil S., Demers, D., Saihi, F. (2001). Considerations
consolidated as well as being sheared under the dikes. on stability of embankments on clay. Soils and
Based on the pore water pressure measurements and the Foundations Journal Vol. 41, No. 5, 117-127.
stress-deformation models, shearing during construction Mesri, G., and Abdel-Ghaffar, M. E. M. (1993) ‘‘Cohesion
has been partially drained. Therefore, it suggests that intercept in effective stress-stability analysis.’’ J.
creep is also occurring in partially drained conditions. Geotech. Eng., 119~8, 1229–1249.
Drained creep is generally accepted to increase the shear Mesri, G. and Shahien, M. (2003). Residual Shear Strength
strength of a normally consolidated clay where the bonds Mobilized in First-Time Slope Failures. ASCE journal of
between clay minerals are already broken. This geotechnical and geoenvironmental engineering.
assessment would not have been possible without the use January 2003
of a sophisticated stress-deformation model that was Mitchell, James Kenneth, SOGA, Kenichi, et al. (2005)
calibrated on the in-situ instruments measurements, the Fundamentals of soil behavior. New York: John Wiley
laboratory testing results and the construction history of the & Sons.
dike and deposition of the tailings. Morgenstern, J, and Tchalenkjo. S. (1967). Microstructural
For embankments on clays, the loss of strength due to observations on shear zones from slips in natural clays.
large deformations in drained conditions would not need to Proceedings of the Geotechnical Conference, Oslo,
be considered as the undrained shear strength considered Vol. 1, pp. 147-152.
in the limit equilibrium analyses is expected to be smaller Tavenas, F., Leroueil, S., La Rochelle, P., Roy, M. (1978)
than the strength calculated based on large deformations Creep behavior of an undisturbed lightly
friction angle. overconsolidated clay. Can. Geotech. J., 15,402-423
It is worth noting that when compared to monotonic Tavenas, F., & Leroueil, S. (1981). Creep and failure of
loading, the difference between the actual state of stresses slopes in clays. Canadian Geotechnical Journal, 18(1),
and the residual stresses in a creep failure is less 106-120.
significant. Therefore, in case of failure, as the material Morsy, M. M., Morgenstern, N. R., D.H. Chan (1995).
moves from the yield to its critical state, it will experience a Simulation of creep deformation in the foundation of Tar
smaller loss of strength. In the case of normally Island Dyke. Can. Geotech. J. 32: 1002-1023
consolidated clay, as the bonds between the particles are Olson, S. M., & Stark, T. D. (2003). Yield strength ratio and
already broken, the behavior is generally expected to be liquefaction analysis of slopes and embankments.
even more ductile. In other words, even if the failure due to Journal of Geotechnical and Geoenvironmental
creep happens in a normally consolidated clay that shows Engineering, 129(8), 727-737.
strain softening behavior, it should not be a highly brittle Skempton, A. W. (1964). Long term stability of clay slopes.
failure. Géotechnique, 14(2), pp. 77-101
Skempton, A. W. (1970). First-time slides in over-
7 REFERENCES consolidated clays. Geotechnique, 20(3), 320-324.

Augustesen, A., Liingaard, M. and Lade, P.V., 2004.

Evaluation of time-dependent behavior of
soils. International Journal of Geomechanics, 4(3):
137-156.
Fell, R., Leroueil, S., Hungr, O., Riemer, W., (2000)
Keynoter lecture – Geotechnical engineering of the
stability of natural slopes and cuts and fills in soil –
January 2000
Hunter, G. and Khalili, N. (2000). “A simple criterion for
creep induced failure of over-consolidated clays”.
Pro.GeoEng 2000 Conference, Melbourne.
ICOLD. (2022). Tailings Dam Safety, Bulletin 194. Draft
version. Committee on Tailings Dams and Waste
Lagoons. 202 p.
James, P. M. (1971). The role of progressive failure in clay
slopes. Proceedings, 1st Australia-New Zealand
Conference on Geomechanics, Vol. 1, pp. 344-348.
Potential Issues on MSE external stability
design in light of the recent Canadian and
AASHTO Bridge Design Codes
Dan V. Dimitriu, Nazmur Rahman
WSP E&I Canada Limited, Windsor, Ontario, Canada

ABSTRACT
The paper examines differences between the Canadian and the American designs with respect to the external stability
(bearing-sliding-eccentricity) of the MSE walls. It is noted that in general, the Canadian design is more conservative. A
particular wall example founded on seismic Site Class C soils in an area in British Columbia at the border with USA is
examined to provide a quantitative assessment of the conservatism in the Canadian design. The differences between the
design results originate from some differences in the prescribed load factors, resistance factors, but more so from some
major differences in the methods of evaluation of the seismic hazard and wall inertia forces. Both codes are ambiguous
with respect to the effects of the load inclination on the foundation bearing resistances. In case of strong earthquakes, the
load inclination factors may completely break down.

RÉSUMÉ
L'article examine les différences entre les conceptions canadiennes et américaines en ce qui concerne la stabilité externe
(appui-glissement-excentricité) des murs TSE. Il est à noter que la conception canadienne est en général plus
conservatrice. Un exemple particulier de mur fondé sur des sols de fondation de classe sismiques C, qui se trouve en
Colombie-Britannique dans une zone située juste à la frontière avec les États-Unis, fournit une évaluation quantitative du
conservatisme de la conception canadienne. Les différences entre les résultats de conception proviennent de certaines
différences dans les facteurs de charge prescrits, les facteurs de résistance, mais plus encore de certaines différences
majeures dans les méthodes d'évaluation de hasard sismique et des forces d'inertie des murs. Les deux codes sont
ambigus en ce qui concerne les effets de l'inclinaison de la charge sur les capacités portantes des fondations. En effet, en
cas de fort tremblement de terre, les facteurs d'inclinaison de la charge peuvent conduire à un effondrement complet.

1 INTRODUCTION the consideration of the load inclination on the foundation

bearing resistances. Also, in both codes the prescribed
The provisions for the external stability design of the expressions for the load inclination factors may
Mechanically Stabilized Earth (MSE) wall according to breakdown in the case of large ratios of horizontal to
the 2019 Canadian Highway Bridge Design Code vertical loads.
(CHBDC) and the 2020 LRFD Bridge Design
Specifications (AASHTO) have been compared and
determined that the Canadian design is more 2 LOAD COMBINATIONS, LOAD FACTORS AND
conservative. The roots of the differences have been RESISTANCE FACTORS
traced from the load and resistance factors to the
prescribed levels for the seismic hazard and to A summary of the typical load combinations and
differences regarding some of the methods of associated load and resistance factors are summarized
calculations of the wall inertial forces and load inclination in Tables 1, 2 and 3 below.
factors. To get a quantitative feeling of the magnitude of While the serviceability (“service” in AASHTO) limit
the differences, an example wall (Figure 1), founded on state (SLS) combinations in both codes are intended to
fully drained cohesionless foundation soils located in an check the wall deformation performance under normal
assumed seismic Site Class C area (Latitude 49.0030, operation, it is customary to conduct checks of the
Longitude -122.750) at the Canadian-USA border in external stability conditions (bearing-sliding-eccentricity)
British Columbia (Figure 2) was considered. as well.
Major differences have been noted in terms of the The ultimate (“strength” in AASHTO) limit states
magnitude of the base design seismic events to be (ULS) combinations are intended to ensure stability /
considered in the design of MSE walls; in Canada it is survival of the structure under significant load
the approximate 2475-year return event while in USA it combinations that are anticipated to reasonably occur
is the approximate 1000-year return event. In addition, during the life of the structure. In practice the design at
there are tangible differences to the determination of the ULS operates with the stresses imposed onto the
wall inertial forces and their combinations with the design foundation soils ensuring that those stresses do not
seismic earth pressures. Another significant difference is exceed the foundation resistances against failures.
in the approach to the effects of the load inclination AASHTO 2020 does not include a load similar to
factors while both codes are rather ambiguous regarding Combination ID = 4, shown in Table 1, as mandated in
CHBDC 2019 and hence no comparisons could be As mentioned before, the percent values in brackets
made. Regardless, Combination ID = 4 was examined in indicate by how much the specific parameter is more
this study just for the case it might have been a more conservative than in the other code. Notably, virtually all
critical design criterion than other combinations. resistance specified by CHBDC are more conservative
than those specified by AASHTO.

Table 1. Typical Load Combinations for Retaining Walls

Table 2 Load Combination Factors
Load
CHBDC 2019 AASHTO 2020
Combination
ID

AASHTO
CHDBC
1A SLS-1 Bearing Service I Bearing

20191
Load Combination ID

2020
1B SLS-1 Sliding / Service I Sliding /
Eccentricity Eccentricity
2A ULS-1 Bearing Strength I Bearing
2B ULS-1 Sliding / Strength I Sliding /
Eccentricity Eccentricity
3A ULS-5 Bearing Extreme Event I

LL3
EQ

EQ
EP

EP
DL

DL
Bearing

LL
3B ULS-5 Sliding / Extreme Event I
Eccentricity Sliding / Eccentricity
4 ULS-9 Bearing Not applicable

1.0 (11%)
1A & 1B

1.0

1.0

0.9

1.0

1.0
0

0
1.35 (8%)

1.5 (20%)

(2.9%)
1.25

1.25

1.75
1.7
2A

0
1.5 (20%)

(2.9%)
1.25

1.75
1.01

1.7

1.0
2B

0
0.52 (∞%)
3A & 3B

1.01

1.01

1.0

1.0

1.0

1.0
0

Figure 1. Arrangement of Example MSE wall

The load factors associated with the load

combinations defined in Table 1 are provided in Table 2.
1.35

1.25

NA

NA

NA

NA
4

The abbreviations in Table 2 represent: DD = Dead

Loads, EP = Loads due to Earth Pressure, LL = Live
Loads, EQ = Earthquake Loads. 1
CHBDC Article 6.19.6 provisions specific to MSE design
The load factors in AASHTO 2020 are equal to, or 2
AASHTO Article C3.4.1 recommends LL factor of 0.5 during
larger (more conservative) compared to CHBDC 2019. extreme seismic events as “reasonable for a wide range of
The percent values in brackets indicate by how much the average daily traffic” (Turkstra’s rule)
specific parameter is more conservative than in the other
3
For all sliding/eccentricity focused checks LL factor is taken 0
code. Most notably, the earth pressure load factor at ULS for live loads directly on top of wall
design for static conditions is by a sizeable 20% larger in
ASSHTO. However, this initial AASHTO conservatism
will “dissipate” and will revert into a more aggressive final Table 3. Summary Resistance Factors
design compared to CHBDC design.
The relevant resistance factors are presented in Load
Table 3. In both codes the guidelines for resistance Combination CHBDC1,2 AASHTO
factors are well defined for the ultimate load ID
combinations but there is an ambiguity regarding the 1A & 1B3 0.4 (bearing) 0.4 (bearing)
resistance factors for SLS / service limit states. For the 0.7-0.8-0.9 (sliding) 1.0 (sliding)
(25%)
latter, interpretations were made as detailed in the
2A 0.45-0.5-0.6 0.65
footnotes to Table 3.
(30%)
2B 0.7-0.8-0.9 1.0
 (25%) = 1 and ic is less than 1.0 depending on the undrained
3A 0.65-0.7-0.84 0.9 shear strength, H, B, L, Nc and ϴ.
(28.5%) It is worthwhile to note that for both codes it is
3B 0.9-1.0-1.04 1.0 conceivable that for some combinations of load
4 0.45-0.5-0.6 N/A inclinations, soils strength parameters, and footing
1
CHBDC defines a range the resistance factors depending on dimensions, the inclination factors may become zero, or
the “degree of understanding”. For this study the only “typical”
values (in bold Italics) have been considered
even breakdown (negative, or imaginary values).
2
According to CHBDC Article 6.5 all resistance factors must be CHBDC is rather ambiguous on the use of the
adjusted by a failure “consequence" factor ranging from 0.9 to inclination factors for the calculation of the bearing
1, 1.5. The "typical" consequence factor of 1.0 has been resistance. The main Article 6.19.9.4 states: “For the
considered purpose of calculating the bearing resistance of the
3
For SLS/Service the resistance factors are interpreted to be gravity mass of the MSE wall system at ultimate limit
applied to the back-calculated driving forces/pressures causing states, an equivalent strip footing shall be assumed
acceptable settlements / displacements (e.g., 25 mm or 50 mm having width of dimension B at the foundation level”.
for some MSE walls). Such deformation study is beyond the
scope of this paper and for comparison purposes the ratios
From this it can be interpreted that all the provisions
nominal bearing resistance over the average bearing pressures stipulated for the determination of the bearing resistance
were used. For SLS/Service sliding verifications a resistance should be considered for the regular strip footings.
factor of 0.8 was applied for CHBDC method (Table 6.2) and However, Article C6.19.9.4 in the Commentaries reads:
1.0 for AASHTO method (Article 11.5.7) “The effect of eccentricity and load inclination is
4
CHBDC Table 6.3 specifies resistance factors equal for static accommodated by the introduction of an effective width,
loads +0.2 to a maximum of 1.0. This applies to force-based B' = L-2e, instead of the actual width”. Therefore, this
seismic design statement may be interpreted that the inclination factors
could be ignored because the eccentricity, e, is not
influenced by the ratio H/V.
3 LOAD INCLINATION AND ECCENTRICITY AASHTO at the commentaries to Article 10.6.3.1.2a
EFFECTS addressed the theoretical determination of the bearing
resistance of shallow footings includes a statement
It is largely accepted in the geotechnical practice that the
reading: “Most geotechnical engineers nationwide have
bearing capacity of the foundation soil under a footing
not used the load inclination factors”. AASHTO further
could be significantly impacted by the load inclination
added: “In practice, therefore, for footings with modest
and eccentricity. It has been long adopted in the industry
embedment, consideration may be given to omission of
to express the bearing capacity of the foundation soils as
the load inclination factors”. These statements seem like
a sum of three fundamental functions, qu =
an inducement to ignore the effects of the inclination that
F(c)+F(d)+F(B,ϒ) attributed each to the soil cohesion, c,
have quite a strong contribution in the reduction of the
to the footing embedment, d, and to the footing width, B, bearing resistances especially at structures subjected to
and unit weight of the foundation soil, ϒ, respectively. substantial and permanent lateral loads.
Each of the above functions are presented as a product
of several factors: bearing capacity factors Nc, Nq, Nɣ
(functions of the soil friction angle, ϕf), load inclination 4 SEISMIC HAZARD AND DESIGN LOADS
factors ic, iq, iϒ (in general functions of applied vertical, V,
and horizontal, H, loads, footing length, L, and width, B, The basic design seismic hazard level in the Canadian
the “azimuth” angle, ϴ, measured from the long axis to codes (CHBDC 2019, NBC 2020) applicable to the force-
the projected direction of the load in the plane of the based design is defined for a 2% probability of
footing, the soil cohesion, c, and friction angle, ϕf), and exceedance in 50 years (about 2475-year return period)
the shape factors accounting for the L/B ratios. while in the USA code (AASHTO 2020) the design
In case of long footings with L>>B loaded almost earthquake is defined for a 7% probability of exceedance
uniformly along the footing length (i.e., plane-strain /2- in 75 years (about 1035-year return). Hence, right from
dimensional cases typical for retaining structures), the here it should be expected that significantly higher
shape factors have no effect being essentially equal to design seismic loads are considered in Canada. It is true
1.0. The inclination factors are always ≤ 1.0 and in some that the Canadian codes also include provisions for
instances could be significantly less than 1.0. performance based-design methods that are allowed to
According to CHBDC, the inclination factors ic, iq and consider lower levels of hazards such as 475-year and
iɣ are described in Article 6.10.2 and depend exclusively 975-year return periods.
on the load inclination angle from the vertical and the As an example, consider a particular geographic
effective friction angle of the foundation soil. According zone at the Canadian-USA border as shown in Figures
to AASHTO, the inclination factors are described in 2a and 2b. Due to the geographic proximity it is
Article 10.6.3.1.2a based on the classical theoretical conceivable that the actual intensity of the seismic
considerations (Vesic, 1975) and recommended also in hazard should be the same. The reference Peak Ground
the Canadian Foundation Engineering Manual (CFEM, Accelerations (PGAref) for the two sites have been
2006). They incorporate the effects of the loads (H, V), determined from the Earthquake Canada website
effective soil friction angle and cohesion of the Canada Seismic Hazard Tool (CSHT) according to the
foundation soils, the azimuth angle and the ratio between 2020 National Building Code of Canada (NBC), and from
the footing dimensions. In case of frictionless soils, iɣ =iq
the PGA contour maps provided in AASHTO 2020 under correct comparison between the two codes, a site
Article 3.10.2.1. correction factor, F(PGA), interpolated between 1.1 and
For Canada the three values of PGA shown in Figure 1.0 must be applied to the US PGAref, corresponding to
2a are for the 2475-year, 975-year and 475-year return PGAref=0.3 and PGAref>0.4, respectively (AASHTO
periods, respectively. Interestingly, the 2020 CHST now Table 3.10.3.2-1). Hence, the site adjusted design PGA
provides directly the seismic design parameters for all in AASTHO becomes 0.33 for the example site and Site
site classes from A to E and accordingly the site Class C.
coefficients F(PGA) required in CHBDC, Table 4.8, to From Figures 2a and 2b it is obvious that for the Site
determine the site adjusted PGA are no longer Class C the intensity of the design seismic hazard in
necessary. However, it appears that the 2020 NBC Canada expressed by the PGA at the selected wall
seismic hazard levels may have been increased location could be by about 48% larger than in the US
compared to 2015 NBC levels. As a quick comparison, design. Even assuming similar return period around
according to CHBDC 2019 the site adjusted PGA for 1000 years, the Canadian uses PGA by 5.1 to 15.7%
stronger earthquakes (PGAref ≥ 0.3) would decrease for larger (more conservative) than in US.
site classes D and E since F(PGA) < 1.0. However, the
updated 2020 CSHT indicates a tangible increase of the 4.1 Design Seismic Coefficient
site-adjusted PGA for the same lower site categories
compared to what is obtained using the CHBDC The intensity of the design seismic loads for retaining
methodology but with the latest values for the reference structures is calculated using the seismic coefficient, kh.
PGA (i.e., PGA for Site Class C). For example, at the Both CHBDC and AASHTO indicated that the seismic
location in Figure 2 and the 2475-year event, CSHT horizontal acceleration coefficient (k h) should be taken
indicates PGA=0.488 for Site Class C and the CHBDC as half the site-adjusted PGA where the walls and the
site factor becomes F(PGA)=0.884 for a Site Class D. backfill can move 25 to 50 mm. For walls that are
Hence, according to CHBDC approach, the adjusted restrained against lateral movement, the seismic lateral
PGA for Site Class D becomes 0.431. The latter is by earth pressures should be obtained using the
more than 22% less conservative than the PGA=0.528 Mononobe-Okabe (M-O) formulation, or the General
obtained from NBC 2020 CSHT for Site Class D. Limit Equilibrium (GLE) methods using a seismic
horizontal acceleration coefficient (kh) equal to the site-
adjusted PGA.
AASHTO 2020 allows for additional reduction of the
kh by a factor α = 1+0.01h(0.5FvS1/kho-1) on the
accounts of the wave scattering, if the walls / slopes are
taller than 6.0 m and the Site Class are C, D, or E. The
expression for α is empirical and as such the wall height,
h, should be expressed in feet. Fv is the Site Factor for
long-period range of acceleration spectrum S1 (T=1.0 s),
and kh0 is the site adjusted peak acceleration for zero
displacement.
a. Canada Based on the above, for the example MSE wall
PGA = 0.488 / 0.347 / 0.257 (Site Class C) located within a Site Class C environment kh and the
seismic earth pressure coefficient, Kae, using the M-O
expression are summarized in Table 4 below.
The results in Table 4 illustrate significant
conservatism by more than 60% of the design seismic
coefficient, kh, and by 23% at the seismic earth pressure
coefficient, Kae.

4.2 Wall Inertial Loads

b. USA According to the Canadian practice the wall inertia forces

PGAref = 0.3 (Site Class B) are calculated as khW, where W is the weight of the
PGA = 0.33 (Site Class C) entire wall. However, the weight and thickness of the
facing are typically ignored in the external stability
analyses.
Figure 2. Location of MSE Wall Example (Figure 2a from
According to AASHTO method, since the MSE wall
CSHT and Figure 2b from AASHTO)
is not a rigid block, only a portion of the reinforced soil
mass, called “effective mass”, corresponding to 50% of
the “effective” wall height would shake in phase with the
By convention, the seismic hazard reference
maximum peak acceleration. The effective height is
parameters in CHBDC were provided for Site Class C
equal to the total height of the wall if the top of the wall is
(Very Dense Soil and Soft Rock) while in AASHTO the
horizontal. The inertial effects of the mass of the facing
reference parameters are provided for Site Class B
are added to the effective mass inertia. In the case of the
(Rock with Vs=[2500 to 5000 ft/s]). Therefore, for a
wall example with a typical cruciform facing of 150 mm 5 RESULTS EXAMPLE MSE WALL
width (linear weight of 24kN/m3x0.15mx7.3m
=26.28kN/m), the resulting wall inertia loads, Pir, are A comparison of the inclination factors obtained for the
summarized in Table 5 below. example MSE wall using the CHBDC and AASHTO
methods is made in Table 6. Because only cohesionless
foundation soil was considered, the ic factors are not
Table 4. Design Seismic Coefficients included in Table 6.
The results of the external design are shown in Table
7 and are compared thru the synthetic Capacity /
Adjusted PGA

Kae-Canada /
Site Class C

Demand (C/D) ratios for the relevant load combinations.

Kae-USA
Design
Event

AASHTO stipulates that Combination ID= 3 (Extreme I)

Kae
F v1
S1

α1

kh
consider two subcategories of combination of the inertial
and seismic earth pressures, namely:

• Case 1 consisting of 50% wall inertia + 100%

Canada
2475yr

seismic earth pressure; and

0.488

0.521

0.244

0.469

1.23
1

• Case 2 consisting of 100% wall inertia + 50%

seismic earth pressure but not less than the static
earth pressure).
Canada
975yr

0.347

0.352

0.173

0.428

1.12
1

Table 6. Comparisons Inclination Factors

Load CHBDC AASHTO

Combination iɣ iq iɣ iq
Canada
475yr

0.257

0.248

0.129

0.39

1.02

ID
1

1 0.47 0.79 0.65 0.79

2 0.45 0.78 0.62 0.77
3 0.04 0.51 0.51 0.69
1035yr

0.225

0.152

0.382
0.33-
USA

1.95

0.92

In Table 7, Combination ID = 3 compares the regular

CHBDC ULS-5 with the AASHTO most conservative
Parameters not used in CHBDC
1
between Case 1 and Case 2 under Extreme-I
combination. In CHBDC such type of subcategories of
combinations of seismic earth pressures and wall inertial
Table 5. Example Wall 100% Inertia Loads, Pir (kN) forces are not distinctly specified. The Commentaries to
CHBDC (Article C6.14.9) mentions that various
Site Class C Canada USA Pir - Canada / international agencies adopt subcategories of load
Adjusted PGA Pir - USA combinations including different proportion of the inertial
0.488-(Canada 207.7 NA1 3.03 forces from 25% to 100%, but these are in conjunction
2475yr) with 100% of the kinematic effects in the case of
0.347-(Canada- 147.3 NA1 2.15 liquefiable soils.
975yr)
Table 7 illustrates that for the studied example wall at
0.257-(Canada- 109.8 NA1 1.60
475yr) SLS/Service combinations, CHBDC design is more
0.33-(USA- - 68.42 - conservative by more than 14% at bearing mostly due to
1035yr) the difference in the definition of the inclination factor, iϒ.
1
NA = Not Applicable At sliding CHBDC is more conservative by more than
2
Without wave scattering effect 24% at sliding while AASHTO is marginally (by 3.8%)
more conservative at eccentricity.
At static ULS-1/Strength-I combinations CHBDC is
The example suggest that the inertial load is by more by 49.7% more conservative at bearing. At sliding the
than 200% more conservative in CHBDC than in level of conservatism in CHBDC narrows to 9.9%. At
AASHTO. The guidelines regarding the evaluation of the eccentricity the differences are less than 1.0%.
inertial loads and their combination with the seismic At seismic ULS-5/Extreme-I the CHBDC design
earth pressures is not detailed at all in CHBDC for the indicates catastrophic failures at bearing and sliding for
non-liquefiable soils. As such there is no basis for the example wall arrangement while AASHTO suggests
scaling down for the combination of 100% of inertia with tangible overdesign with C/D > 3 at bearing and C/D >
100% of seismic earth pressure which likely is largely 1.3 for sliding. According to CHBDC method, the
overconservative. example wall to meet C/D = 1.0 at sliding under ULS-5
Table 7. Summary Results - Example MSE wall

Relevant Load Capacity / Demand (C/D) Control Ratios

Combination
ID BEARING SLIDING ECCENTRICITY
CHBDC AASHTO CHBDC AASHTO CHBDC AASHTO

1 4.9 5.59 1.57 1.96 2.83 2.75

(14.1%) (24.8%) (-2.8%)

2 1.61 2.41 1.21 1.33 3.68 3.65

(49.7%) (9.9%) (-0.8%)

3 0.39 3.26 0.80 1.32 1.81 3.01

(736%) (65%) (66.3%)

combination will require an increased width to 9.3 m, Limited management and to Taner Aktas, P.Eng., for the
hence a B/H ratio of 1.27 compared to B/H = 0.73 for the logistic support in the preparation of this paper.
stable wall design according to AASHTO method. In
other words, for the selected example the Canadian
design leads to a wall by 75% wider (more massive) than 8 REFERENCES
in USA.
At eccentricity all checks pass comfortably with C/D AASHTO, LRFD Bridge Design Specifications, 9th
> 1.8 for CHBDC and C/D > 2.75 for AASHTO. Edition, 2020.
Canadian Research Council Canada, National Building
Code of Canada, 2020, Volume 1.
6 CLOSING REMARKS Canadian Research Council Canada, National Building
Code of Canada, 2015, Volume 1.
The study was limited to the external stability and fully Canadian Geotechnical Society, Canadian Foundation
drained conditions. Engineering Manual, 4th Edition, 2006.
The Canadian design of the external stability of MSE CSHT, 2020 National Building Code of Canada Seismic
wall is more conservative than the USA although the load Hazard Tool (earthquakescanada.nrcan.gc.ca).
factors in the USA code trend higher than in the CSA Group, Canadian Highway Bridge Design Code,
Canadian code. The sources of the conservatism CSA S6:19.
originate from lower resistance factors by an average of Vesic A. S. 1975. Bearing Capacity of Shallow
28% in CHBDC combined with a significantly and Foundations, Foundation Engineering Handbook,
inherently lower inclination factor iɣ recommended by the Editors Winterkorn, H.F. and Fang H-Y., Van
Canadian code. A particularly significant impact has the Nostrand Reinhold Co.
inclination factor at ULS-5 where, according to CHBDC,
iɣ rapidly approaches zero, and hence, virtually
annihilates the bearing resistance in the case of drained
cohesionless foundation soils.
Another major source of the difference between
CHBDC and AASHTO originates from the significantly
higher levels of the design earthquake in Canada
(approximate 2475-year return) versus in USA
(approximate 1000-year return). This difference is
further amplified by the differences in the methods of
considering the wall inertial loads and their combination
with the seismic earth pressures.
The seismic hazard levels according to NBC 2020
seemed to have trended upwards, deepening the gap
between the CHBDC and AASHTO regarding the
specified levels for the design seismic hazard in Canada
and USA, respectively.

7 ACKNOWLEDGEMENTS

The authors are deeply grateful to the WSP E&I Canada

Conception et construction d’une passerelle
à motoneiges sur la rivière Aisley
Pierre Vannobel, ing., M.Sc.A. &
Jose Marcel Bustamante Bedoya, ing., M.Sc.
Hydro-Québec, Montréal, Québec, Canada
Benoît Turgeon, ing.
WSP, Ville de Saguenay, Québec, Canada

ABSTRACT
Three snowmobile auxiliary bridges were built in 2022 and 2023 to provide a safe crossing of the Romaine Salmon River
and of the Aisley River. The auxiliary bridges, on the edge of Route 138, are connected to the Trans-Québec no. 3
snowmobile trail that links Tadoussac to Baie-Johan-Beetz on the north shore of the St. Lawrence River. This article
specifically addresses the design and construction of the abutment’s foundations of the Aisley River auxiliary bridge. They
consist of deep steel piles driven to refusal in overburden and to rock. These foundations, constructed in the shelter of
riverbank sheet piles, are located in the limits of the Goldthwait Clay Sea. Vibrations during pile driving to refusal were
measured to ensure compliance with MTQ vibration criteria at the nearby Route 138 bridge. Dynamic monitoring was done
on piles to confirm their capacity.

RÉSUMÉ
Trois passerelles à motoneiges ont été construites en 2022 et 2023 afin de pouvoir traverser sécuritairement la rivière à
saumon Romaine ainsi que la rivière Aisley. Les passerelles, situées en marge de la route 138, sont agencées au sentier
de motoneiges Trans-Québec no. 3 qui relie Tadoussac à Baie-Johan-Beetz en rive nord du fleuve St-Laurent. Cet article
traite particulièrement de la conception et de la construction des fondations des culées de la passerelle sur la rivière Aisley.
Elles sont constituées de pieux profonds en acier foncés au refus dans le mort-terrain et au roc. Ces fondations, construites
à l’abri de palplanches, sont situées dans les limites de la mer d’argile Goldthwait. Les vibrations lors du battage des pieux
au refus ont été mesurées afin d’assurer le respect des critères de vibration du MTQ au pont de la route 138 à proximité.
Des analyses dynamiques ont été effectuées sur les pieux afin de confirmer la capacité des pieux.

1. INTRODUCTION

Hydro-Québec a réalisé les nouveaux aménagements

hydroélectriques du complexe La Romaine sur la rivière
Romaine. Or, Hydro-Québec et les citoyens du milieu
anticipaient que l’exploitation des nouveaux
aménagements hydroélectriques du complexe La
Romaine allait affecter la formation du couvert de glace sur
les rivières Romaine et Aisley. Conséquemment, les
motoneigistes ne pourraient plus passer de façon
sécuritaire sur ces rivières (Figure 1).
Pour remédier à cette problématique, dans un premier
temps, un service d’escorte automobile 24 heures sur 24 a
été mis en place afin que les motoneigistes puissent
circuler sécuritairement sur les ponts de la route 138, au
droit des rivières Romaine et Aisley. Puis en 2022 et 2023,
trois passerelles à motoneiges ont été construites par
Hydro-Québec pour permettre aux motoneigistes de Figure 1. Couvert de glace affecté par les nouveaux
traverser de façon sécuritaire la rivière à saumon Romaine aménagements sur la rivière Aisley
et la rivière Aisley adjacente, tributaire de la rivière
Romaine. Ces travaux ont été réalisés sous la supervision
d’Hydro-Québec, également maître d’œuvre. Ces passerelles en acier sont agencées au sentier de
Ces trois passerelles ont été conçues par la firme WSP. motoneiges Trans-Québec no 3, qui relie les villes de
Elles ont été construites par l’entreprise SEIE avec l'aide Tadoussac et Baie-Johan-Beetz, qui se trouvent
de ses deux sous-traitants: SBP et Les Constructions respectivement sur la Côte-Nord et la moyenne Côte-Nord
Mackenzie Inc. (rive nord du fleuve St-Laurent). Ces passerelles à
motoneiges sont situées en marge de la route 138 vers le
PK 3 de la rivière Romaine, à une cinquantaine de
kilomètres en aval de l’aménagement hydroélectrique
Romaine-1 le long de la rivière (Figure 2).
La passerelle de la rivière Aisley a été construite à
proximité d’un pont existant, édifié en 1964 et présent le
long de la route 138 sur la rivière Aisley, en aval de la
nouvelle passerelle. Les culées en béton armé de ce pont
existant reposent sur des pieux en bois battus au refus au
till ou au roc. Or, la distance la plus courte entre les culées
de ce pont et celles de la nouvelle passerelle à motoneiges
est, en son point le plus rapproché, de seulement 26 m, ci
qui implique des précautions à prendre lors des travaux.
Cette passerelle en acier galvanisé, d’une longueur de
34,5 m, a la particularité d’être située dans les limites de la
mer d’argile Goldthwait (Figure 3). En 1980, un très grand
glissement par étalement dans l’argile avait emporté la
route 138 à quelques kilomètres de cette passerelle, ce qui
dénote la grande sensibilité de ce dépôt argileux marin
dans ce secteur.

Figure 3. Limites de la mer Goldthwait (Figure modifiée de

Leroueil et coll., 1985)
Route 138
Sentier de motoneiges
Passerelles à motoneiges 2.1 Dépôt d’argile de la mer Goldthwait

Figure 2. Localisation des passerelles à motoneiges L’argile marine du type CL de la mer Goldthwait, au site de
la passerelle Aisley, montre une faible plasticité et un
indice de liquidité de l’ordre de 2,1 à 3,3, ce qui lui confère
Cet article traite principalement, du point de vue une grande sensibilité au remaniement (Tableau 1). Sa
géotechnique, de la conception et de la construction des résistance au cisaillement non drainé Su est de 35 kPa min
fondations des culées de cette nouvelle passerelle de la alors que sa résistance remaniée Sur est de l‘ordre de 0,1
rivière Aisley, située dans les limites de la mer d’argile kPa. La grande sensibilité de l’argile a été considérée par
Goldthwait. le concepteur, ce qui a permis le bon déroulement des
travaux d’excavation et de construction de la passerelle.

2. CONCEPTION ET CONSTRUCTION DE LA
PASSERELLE ET AGENCEMENT AU SITE

Les deux culées en béton armé de la passerelle à

motoneiges, munies de murs en retour en console,
reposent sur des pieux foncés au refus au roc ou au mort-
terrain en rives de la rivière (Figure 4). Les fondations
profondes des culées de la passerelle ont été ceinturées
de palplanches en rives de la rivière au niveau du dépôt
d’argile sensible à très sensible de surface. Chaque culée
est appuyée sur six pieux verticaux et six pieux inclinés.
Lors de la phase des travaux, la passerelle a été
modélisée en 3D par Hydro-Québec à l’aide du logiciel
Catia, afin de permettre de bien ajuster l’agencement de la Figure 4. Concept de la passerelle Aisley
passerelle avec les conditions au site. Cette vue 3D selon
l’axe ne montre que la moitié de la structure (Figure 4).
Tableau 1. Caractéristiques de l’argile de type CL testée analyse par équation d’ondes a été réalisée à l’aide du
logiciel GRLWeap par la firme spécialisée en fondation
Forage F-3 Forage F-4 profonde CBF. Cette analyse était nécessaire afin de
Teneur en eau (%) 48,1 44,7 à 51,4
confirmer la capacité géotechnique des pieux au refus et
d’évaluer le critère de refus requis, dans le but de fournir la
Limite liquide (%) 34 29
bonne capacité portante escomptée. Selon cette analyse,
Limite plastique (%) 21 20 un critère de refus de 25,4 mm d’enfoncement au
Sable (%) 2,4 1,8 maximum pour 10 chutes d’un marteau d’un poids de 29,9
Silt 46,5 54,3 kN avec une chute de 0,91 m permettrait d’atteindre la
Argile 51,1 43,9 capacité portante géotechnique pondérée requise (ÉLUL)
de 600 kN. La contrainte de compression dans l’acier était
estimée à environ 210 kPa max, soit inférieure à la limite
élastique de l’acier de 310 kPa.
2.2 Excavations du mort-terrain
2.5 Vibrations engendrées par le battage des pieux au
Les excavations de mort-terrain aux deux appuis de la
refus
passerelle ont été requises afin de permettre la réalisation
sécuritaire des travaux et la mise en place de plateformes
La vitesse des particules au pic qui se produit à un endroit
temporaires stables. Ceci a ainsi permis de mettre en place
donné, en fonction de la distance du fonçage d’un pieu par
des palplanches et des pieux puis de manipuler et de
battage, ainsi que la sévérité et la perceptibilité des
mettre en place la structure en acier de la passerelle. Les
vibrations transmises peuvent être évaluées à partir des
pentes d’excavation étaient de 2H :1V avec protection en
travaux de Bay (2003) en fonction de la compacité des sols
enrochement afin d’assurer la stabilité des pentes.
en place.
L’évaluation de la transmission des vibrations est
2.3 Palplanches
influencée par plusieurs autres paramètres comme la
présence d’une croûte de sols raides en surface, d’une
Des palplanches ont été foncées au site sur une
excavation profonde, de la proximité du socle rocheux ou
profondeur de plus de 5 m, en guise de batardeaux
de la présence d’affleurements rocheux. Une telle
temporaires. Ainsi, il a été possible d’atteindre le niveau
évaluation ne peut en aucun cas être substituée à la
requis d’arasement des pieux foncés aux deux appuis sous
mesure des vibrations réelles au site lors des travaux de
le lit de la rivière et réaliser à sec la structure de béton armé
mise en place des pieux.
de la culée au-dessus. La grande sensibilité de l’argile au
Les vibrations transmises au pont existant du Ministère
site a fait en sorte qu’elle se remanie et perd donc de sa
des Transports du Québec (MTQ) lors du battage au refus
résistance sur le périmètre des palplanches sur quelques
des pieux de la nouvelle passerelle à motoneige ont été
décimètres d’épaisseur lors de leur fonçage par vibrations
évaluées en fonction de nos conditions au site et en se
(Figure 5). Ce phénomène a dû être considéré lors de la
fondant sur les travaux de Bay (2003). Les vibrations,
vérification de la stabilité des caissons en palplanches.
estimées empiriquement, sont d’environ 1 mm/s. Elles sont
donc fortement perceptibles, mais ne génèrent pas de
dommages au pont existant du MTQ et respectent le critère
du Ministère des Transports du Québec, qui est de 3 à 8
mm/s max.
Deux sismographes munis d’un géophone triaxial ont
été installés sur le béton et sur les remblais d’approche du
pont existant sur la route 138. Ces instruments, qui
permettent de mesurer l’amplitude et la fréquence des
ondes vibratoires selon les trois composantes
orthogonales indépendantes (longitudinale, verticale et
transversale), ont été installés afin de vérifier et de
confirmer le respect des critères de vibration du MTQ lors
du battage des pieux au refus à la nouvelle passerelle. Les
vibrations mesurées au pont existant, lors du battage des
pieux à la passerelle au refus au till ou au roc, étaient de
Figure 5. Remaniement de l’argile lors du fonçage des l’ordre de 0,1 à 1,5 mm/s, soit sensiblement égales à celles
palplanches préalablement estimées empiriquement (Figure 6). Il est à
noter que le simple déplacement de l’instrument par un
technicien a engendré une valeur enregistrée de l’ordre de
2.4 Conception des pieux 2 mm/s. La fréquence verticale mesurée était de l’ordre de
8 à 70 Hz. La faible valeur des vibrations mesurées
Avec le peu de capacité portante des sols en place, le choix confirme que le battage des pieux à la passerelle n’était
d’utiliser des pieux comme support de la structure s’est vite pas dommageable pour l’intégrité du pont existant et de
imposé. Les pieux sont du type circulaire en acier, à bout ses fondations.
fermé à l’aide d’une pointe de fonçage, d’un diamètre de
0,323 m et d’une épaisseur d’un peu plus de 8 mm. Une
Figure 6. Vibrations mesurées au pont existant sur la 138 lors du battage des pieux (Source : Englobe)

2.6 Contrôle qualitatif sur les pieux battus au refus Le principal point contrôlé pour les pieux battus au refus
franc au roc a consisté à vérifier précisément que les pieux
Une analyse dynamique a été effectuée par une firme ne relèvent pas suite à leur enfoncement. Un des 12 pieux
spécialisée sur deux des 12 pieux foncés au refus à la au roc à la culée Est a justement relevé sous l’action du
culée ouest afin de confirmer, à l’aide du logiciel Capwap, principe d’Archimède, avant que du béton ne soit appliqué
leur capacité portante pondérée adéquate au refus dans le dans ces 12 pieux.
mort-terrain constitué en profondeur de till glaciaire. Suite Les 12 pieux foncés à la culée Ouest ne pouvaient pas
à la première analyse dynamique effectuée sur le premier relever étant donné que la forte friction latérale sur les fûts
pieu battu au refus, le critère de refus a été sécuritairement des pieux était supérieure à la poussée d’Archimède vers
ajusté à 19 mm d’enfoncement au maximum pour 10 le haut sur ces pieux.
chutes de 0,91 m d’un marteau d’un poids de 29,9 kN. Les
résultats ont démontré une capacité portante pondérée
supérieure à celle estimée; elle était de 900 kPa avec un
coefficient de tenue géotechnique de 0,5 au lieu de 600
kPa en tête du pieu estimé sécuritairement initialement en
conception avec un coefficient de tenue géotechnique de
0,4 (Figure 7). Aucun endommagement aux pieux n’a été
noté (Figure 8). Quant à la capacité portante pondérée des
12 pieux au refus franc au roc à la culée est, elle n’a pas
été testée par un essai dynamique.

Figure 8. Sommet des pieux en acier

2.7 Culées en béton armé et structure en acier

Les deux culées en béton armé ont nécessité la mise en

place de 125 m3 de béton armé contenant 8 300 kg d’acier.
Des coffrages temporaires en bois ont été préalablement
installés afin de permettre le bétonnage des culées (Figure
Figure 7. Exemple de résultat d’analyse dynamique à l’aide
9).
du logiciel Capwap (Source : PDA Consultants Inc.)
Des remblais de protection des rives et des culées,
constitués de blocs de 0,5 à 0,9 m de dimensions, ont été
placés en rive.
 2.8 Assemblage et mise en place de la passerelle à
motoneiges

La passerelle a été assemblée sur place à l’appui gauche

(Figure 10). Elle a été poussée en place longitudinalement
le 24 mai 2023 (Figure 11).

3. CONCLUSION

La passerelle à motoneiges Aisley, ainsi que les deux

autres sur la rivière Romaine à proximité, seront en
opération à l’hiver 2023-2024. Ce type de passerelle à
motoneiges, qui doit être conçue en fonction des conditions
Figure 9. Culées en béton armé propres à chaque site et construite en respect de
l’environnement, de la sécurité de chacun et de la qualité,
constitue une mesure de compensation efficace pour
assurer à long terme la sécurité et l’agrément des usagers
du milieu.

4. REMERCIEMENTS

Les auteurs remercient leurs collègues des départements

suivants qui se sont assurés de la conformité du présent
article: Intégration et ingénierie – Géotechnique
Unité Expertise et conception – Hydraulique, hydrologie et
barrages en remblai – Direction Sécurité des barrages et
infrastructures; Direction Expertise et soutien technique –
Sécurité des barrages et infrastructures; Ingénierie du
Chantier Romaine-4; Administration de contrats Chantier
Romaine-4; Communications entreprise II.
Figure 10. Assemblage sur place de la passerelle en acier
à l’appui gauche (Source : LP Drone)
5. RÉFÉRENCES

Bay, J.A. 2003. A summary of the Research on Pile Driving

Vibrations, Proceedings of the Pile Driving Contractor’s
Association 7th Annual Winter Roundtable, Atlanta,
Georgia, United States.
Dionne, J.-C. 1977. La mer de Goldthwait au
Québec, Géographie physique et Quaternaire, 31 :61-
80.
Englobe, 2023. Rapport de suivi de vibrations du 02 au 08
février 2023, No de référence 35-02207319.000-0100-
HS-R-0123-00, Varennes, Québec, Canada.
Forage CBF. 2022. Pieux – Calculs structuraux –
Passerelle de motoneige – Au-dessus de la rivière
Aisley, Havre-Saint-Pierre, Québec, Canada.
GHD, 2018. Étude géotechnique - Nouvelles passerelles –
Rivières Romaine et Aishley – Route 138, Havre-Saint-
Pierre, Québec. No de référence : 11184188-
A1, Rimouski, Québec, Canada.
Figure 11. Passerelle en acier en place (Source : LP GHD, 2022. Avis géotechnique – Construction d’une
Drone) passerelle au-dessus de la rivière Aisley – Route 138,
Havre-Saint-Pierre, Québec. No de référence :
12592108-A2, Québec, Québec, Canada.
Hydro-Québec. 2020. Bilan des activités
environnementales 2019. Complexe de la Romaine.
Hydro-Québec. 2022. Fourniture et installation de
passerelles pour motoneigistes près de la route 138 sur
les rivières Aisley et Romaine – Contrat R3-10-
06. Aménagement hydroélectrique de la Romaine-3.
Clauses techniques particulières – Pour construction.
No de référence : WSP 181-13003-01.
Hydro-Québec. 2022. Note interne – Visite chantier et
tranchées d’exploration pour installation de passerelles
à motoneige – amont du pont Benoît-Vigneault (P-
02495W), amont du pont Alexandre –Tanguay (P-
02498E) et amont du pont P-02497. Code de
classement : CI-2022-0027-01.
Hydro-Québec. 2023. Maquette Catia 3D, Montréal,
Québec, Canada.
Leroueil S. et coll. 1985. Remblais sur argile molle.
Technique et documentation Lavoisier, Paris, France,
342 pages.
Ministère des Transports du Québec, 2022. Courriel du
MTQ à Hydro-Québec. R3-10-06 – Demande de
permis de voirie – contrôle des ondes vibratoires. 22
juillet 2022.
PDA consultants inc. 2023. Rapport Final – culée 1 –
Analyse de battage de pieux. Passerelle de motoneige
– Au-dessus de la rivière Aisley, Havre-Saint-Pierre
(Québec). Longueuil, Québec, Canada.
TR3E Experts-Conseils inc. 2022. Conception du
batardeau de l’axe 1. Batardeau – Passerelle Aisley.
Notes de calcul. Rimouski, Québec, Canada.
TR3E Experts-Conseils inc. 2022. Conception du
batardeau de l’axe 2. Batardeau – Passerelle Aisley.
Notes de calcul. Rimouski, Québec, Canada.
 Wednesday, October 4, 2023

NUMERICAL MODELS II
Probabilistic analysis of an embankment
under different rainfall events considering
spatial variability of soil strength
parameters
Leila Baninajarian1, Sina Javankhoshdel1 & Rashid Bashir2
1Rocscience, Inc., 54 St. Patrick St., Toronto, ON, M5T 1V1

Canada
2York University, 11 Arboretum Lane, Toronto, ON, M3J 1P3

Canada

ABSTRACT
Studies have shown that embankments built with fine materials such as silt are more susceptible to extreme rainfall events
than those built with coarse materials. In most of these studies, it was assumed that soil strength and hydraulic parameters
such as unit weight, friction angle, cohesion, and hydraulic conductivity are random variables. However, soil properties
and hydraulic properties can vary spatially within a soil profile. This has been shown by several researchers by carrying
out stochastic analyses. Therefore, as a preliminary study, this research investigates the influence of spatial variability of
soil strength parameters including cohesion, friction angle and unit weight on the probability of failure and a simplistic
assumption of constant hydraulic conductivity throughout the soil profile is made. The probabilistic analysis is carried out
for an extreme rainfall event. The results are compared with analyses in which assumption of random variable soil
parameters is made. The results show a reduction in the value of the total probability of failure for the case with spatially
variable soil strength parameters compared to the case with random soil parameters. The spatially variable case better
describes the practical site condition.
RÉSUMÉ
Des études ont montré que les remblais construits avec des matériaux nobles tels que le limon sont plus sensibles aux
précipitations extrêmes que ceux construits avec des matériaux grossiers. Dans la plupart des études mentionnées ci-
dessus, on a supposé que certains des paramètres hydrauliques tels que la conductivité hydraulique sont des variables
aléatoires, tout en tenant compte de la variabilité aléatoire de l’angle de frottement du sol. Cependant, les propriétés du
sol et les propriétés hydrauliques peuvent varier spatialement à l’intérieur d’un profil de sol. Cela a été démontré par
plusieurs chercheurs en effectuant des analyses stochastiques. Par conséquent, en tant qu’étude préliminaire, cette
recherche étudie l’influence de la variabilité spatiale des paramètres de résistance du sol, y compris la cohésion, l’angle
de frottement et le poids unitaire sur la probabilité de défaillance. Une hypothèse simpliste de conductivité hydraulique
constante dans tout le profil du sol est faite. L’analyse probabiliste est réalisée pour un événement pluvieux extrême. Les
résultats sont comparés à des analyses dans lesquelles l’hypothèse de paramètres pédologiques variables aléatoires est
faite. Les résultats montrent une réduction de la valeur de la probabilité totale de rupture pour le cas avec des paramètres
de résistance du sol spatialement variables par rapport au cas avec des paramètres de sol aléatoires. Le cas spatialement
variable décrit mieux l’état pratique du site.

1 INTRODUCTION distribution of pore water pressure, and ultimately, the

stability of the soil embankments (Pk et al., 2019).
Studies have shown that water infiltration, commonly This research aims to examine how the spatial
occurring during rainfall, can have a significant impact on variations in soil strength parameters, such as friction angle
the stability of soil embankments. Additionally, the moisture cohesion, and unit weight, influence the likelihood of failure
levels in the slope before rainfall can also influence the in the soil embankments. The study employs a probabilistic
vulnerability of soil embankments to instability. Soil analysis that considers the spatial variability of soil
geotechnical characteristics, which govern the stability of parameters for an embankment subjected to an extreme
earth embankments, can vary spatially due to the inherent rainfall event with varying initial saturation conditions. The
heterogeneity of soils. Moreover, uncertainties arising from results obtained are then compared to those from
measurement errors and inadequate testing further deterministic analysis and probabilistic analyses ignoring
contribute to the uncertainty surrounding the soil properties spatial variability of some soil parameters Mentioned in the
used in analyses. Therefore, it is crucial to consider all following
these uncertainties when evaluating the stability of
embankment slope (Baninajarian et al., 2020). 1.1 Background
The surface of soil embankments interacts with the
atmosphere, and the exchange of water at the soil-
atmosphere boundary affects soil water storage, The stability of both natural and constructed soil slopes can
be influenced by changes in extreme rainfall patterns
caused by climate change. For instance, embankments failure (PF) in the soil embankment under rainfall events
which are constantly exposed to atmospheric conditions with variably saturated initial conditions is estimated. The
are susceptible to slope failures triggered by rainfall. Metaheuristic approach known as Non-circular Particle
Baninajarian et al. in (2020), conducted a research Swarm Search is used as the search method to identify the
study focusing on the impact of extreme rainfall on a soil critical slip surface, while the General Limit
embankment slope located in Niagara Falls, Ontario. The Equilibrium/Morgenstern-Price method (Price &
primary objective of their study was to develop fragility Morgenstern, 2015) is applied in this study (Rocscience,
curves for the slope, considering two distinct types of 2022).
failures: general failures and shallow failures. To assess In this research, the shear strength of soil is only based
the slope's stability throughout extreme events over time, on cohesion and frictional strength. In order to be more
the researchers employed a coupled hydro-geotechnical conservative, the contribution of suction to shear strength
model as a means of evaluation. is neglected.
The fragility curves were developed using the First Order Simulations are conducted considering transient flow
Second Moment (FOSM) method, and the associated analysis, and the factor of safety of the embankment is
reliability analyses assumed that various soil parameters evaluated at various time steps. Deterministic analyses are
were random variables. The findings from their research performed to establish a benchmark for comparing the
revealed that embankments constructed using fine results of probabilistic analyses, demonstrating the impact
materials, specifically silts, exhibit heightened susceptibility of input parameter variability on the probability of failure.
to an elevated likelihood of failures during extended rainfall The probabilistic analyses aim to evaluate the
events characterized by longer return periods. probability of slope failure, by considering the inherent
In 2017, Javankhoshdel et al. investigated the effect of uncertainties associated with certain soil parameters such
random and spatial variability of soil strength parameters, as friction angle, cohesion, and unit weight. Spatial
and cross-correlation between them on probability of failure variability analyses are carried out to quantify the effect of
of simple slopes with cohesive and cohesive frictional soil. spatial variation of soil parameters within a soil mass on the
They performed several analyses using 2D random finite probability of failure. When employing the RLEM approach
element method (RFEM) and 1D and 2D random limit for probabilistic analyses, it is necessary to define random
equilibrium method (RLEM) to identify to compare RLEM variables and their corresponding statistical parameters.
and RFEM approaches. This includes determining the probability distribution
function (PDF) of the variables, as well as establishing the
Cami et al., (2020) provided a comprehensive database standard deviation and scale of fluctuation in both
of horizontal and vertical scale of fluctuation values horizontal and vertical directions.
obtained from various locations and materials, based on
published case studies which represents the spatial 1.2 Geometry and Materials
correlation length essential for accurately characterizing
and simulating a spatially variable field. Chash, The embankment profile analyzed in this study is depicted
in Figure 1. The embankment has a symmetrical geometry,
Javankhoshdel et al. (2022) conducted a comparative and only half of the domain is simulated. It has a height of
study to assess the accuracy and time efficiency of the 8 meters with side slopes of 2H:1V. Additionally, a 3-meter
calculation process in probabilistic and stochastic slope width unpaved shoulder is assumed at the top of the
stability analyses, specifically focusing on the RFEM and embankment. To minimize the influence of side boundary
RLEM approaches. They concluded that RLEM conditions, the distance between the slope toe and the right
demonstrates superior performance compared to RFEM in side of the model is set to more than three times the height
slope stability analyses. Additionally, they noted that of the slope (Rahardjo et al., 2010).
RFEM's performance is highly influenced by the mesh size The embankment fill material studied in this research is
and field discretization required to create a spatially sandy silt, hereafter referred to as "silt." The material
variable field. On the other hand, RLEM is less sensitive to properties considered for the silt include a saturated unit
these factors as it utilizes the method of slices, where weight (𝛾) of 19 kN/m3, a friction angle (ϕ) of 27°, and a
adding more slices does not significantly impact small value of 2 kPa for the effective cohesion.
computation time. For simplicity, a constant hydraulic conductivity function is
Griffiths et al., (2009) investigated the probability of assumed throughout the soil profile in this study. The
failure of slopes using the first-order reliability method and hydraulic behavior of unsaturated soils is characterized by
the random finite-element method. The results showed that the soil water characteristic curve (SWCC) and the
simplified probabilistic analyses which do not consider unsaturated hydraulic conductivity function (HCF), which
spatial variability of soil properties can lead to are determined using the van Genuchten-Mualem
unconservative estimates of slope failure probability and approach (Mualem, 1976). Figure 2 shows the Soil-water
more advanced probabilistic methods are warranted. Characteristic Curves for silt based on (Baninajarian,
In this study, the RLEM approach is utilized to perform 2020).
a probabilistic analysis using the Slide2 software, which is
a 2D limit equilibrium slope stability program (Rocscience,
2023). This software is employed to conduct finite element
seepage analysis for groundwater and assess the factor of
safety (FOS) using LEM. Moreover, the probability of
 The initial moisture condition influences the pore water
pressure within the embankment, which significantly
impacts slope stability. Baninajarian, (2020) evaluated
slope moisture conditions by calculating the average
degree of saturation within the slope area and performed
statistical analyses to identify the average and maximum
moisture conditions within the slope and determine critical
times when the distribution of pore water pressure (PWP)
poses a risk to slope stability. In the current study, long-
term variations of the spatial distribution of pore water
pressure (PWP) within the embankments corresponding to
the 90th percentile of saturation (P90%) is used.
Figure 1. Embankment considered in this study (modified
The water table was conservatively assumed to be at
from Baninajarian, 2020)
the natural ground surface, located 4m below the slope toe.
At the soil-atmosphere interface, a flux boundary
representing an actual rainstorm pattern, known as the
Chicago design storm, based on Keifer & Chu, (1957), was
applied. This approach has been recommended by the
Ministry of Transportation Ontario for assessing the impact
of storms on drainage systems (MTO, 1997). Figure 3
illustrates the design storms used in this study. The
Chicago curve, developed by (Baninajarian, 2020) is based
on future Intensity-Duration-Frequency (IDF) curves for a
48-hour rainfall event with a 100-year return period in the
city of Niagara Falls. However, the simulation was
extended beyond the duration of the rainfall event to 60
hours due to high water retaining capacity of silt.

Figure 2. Soil water characteristic curves for silt material,

used in this study (modified from Baninajarian, 2020)

In this study, the variability of three crucial soil

parameters: the angle of internal friction, cohesion, and unit
weight are examined. Various researchers have suggested
a coefficient of variation (COV) ranging from 2% to 20%.
for the angle of internal friction, 10% to 50% for the
cohesion, and 3% to 7% for the for the unit weight (Harr,
1984, Kulhawy, 1992, Phoon & Kulhawy, 1999,
Javankhoshdel & Bathurst, 2016). In this study, a COV of
15%, 50%, and 5% are chosen for the angle of internal
friction, cohesion, and unit weight, respectively. The
probability distribution function chosen for ϕ, c, and 𝛾 is the
lognormal distribution. The soil parameters considered in
the probabilistic analyses are summarized in Tables 1, and
for the probabilistic slope stability analyses, the Latin-
Hypercube method is employed with 10000 samples.
Figure 3. Chicago curve for 48-hour extreme rainfall – City
of Niagara Falls (modified from Baninajarian, 2020)
Table 1. Mean values and standard deviations of soil 2 ANALYSIS AND RESULTS
parameters considered in this study
Parameters Mean value Standard Deviation 2.1 Seepage and Deterministic Analysis
Friction Angle 27 4 PWP distribution at different times after the extreme
Unit Weight 19 1 precipitation event are illustrated in Figure 4. The change
Cohesion 2 1 in the phreatic surface and pore water pressure due to the
infiltration of the water into the slope are shown in the
following figures. Figure 4(a) indicates the initial PWP
conditions in the slope. At this time, rainfall starts and the
pore water pressure in the embankment starts to increase
1.3 Initial and Boundary Conditions
by the infiltration as it can be seen from figure 4(b) to 4(d),
which show the PWP distribution at the peak of the rainfall,
end of it, and after 60 hours respectively. Due to the low
permeability of material, the PWP increases even 12 hours
after the end of the rainfall.
(a)

(b)

Figure 5. Temporal variation of FOS obtained from the

deterministic analysis

2.2 Single Random Variability Analysis (SRV)

For the probabilistic analyses, soil unit weight, friction

angle, and cohesion are considered random variables
which their values are as presented earlier in Table 1. The
(c) probability of slope failure and values of mean FOS at
different stages obtained from SRV analysis are shown in
Figure 6. This figure shows an insignificant change in the
mean FOS over time. The mean FOS changes from around
1.275 at initial hours of rainfall to almost 1.26 after 60 hours
thereafter. These values are relatively close to values
obtained from deterministic analyses.
In the current study, 1.10 is considered as the target
FOS for the estimation of PF. The variation of probability of
failure over time is shown in Figure 7. It shows the
probability of failure obtained from SRV analyses varies
(d) from around 20% to 22%. It is noticeable that these values
are relatively high considering an acceptable mean FOS of
about 1.3 and based on target PF = 0.01% for permanent
well-engineered and constructed unreinforced soil slopes
and embankments (Silva et al., 2008). This indicates that
ignoring the spatial variability of soil properties results in an
unrealistically high probability of failure.
A comparison between the results from current study
based on the critical FOS of 1.0 and previous study
conducted by Baninajarian et al., (2023) shows that
Figure 4. Spatial distribution of PWP in the slope at the (a) although the value of mean factor of safety obtained from
initial stage, after (b) 18.5 hours, (c) 48 hours, and (d) 60 probabilistic analyses in both studies are the same, the PF
hours increases to around 30%. This shows even a low value of
soil cohesion plays a critical role on the probability of slope
The temporal variation of FOS of slope at various times failure.
during the rainfall event is presented in Figure 5. The
saturated conditions in the slope and increase of pore
water pressure due to the rainfall results in reducing the 2.3 Stochastic Analysis
effective soil stress and soil shear strength and
consequently, the factor of safety decreases. However, the 2.3.1 Spatial Variability Analysis
variation of FOS is rather insignificant, under the rainfall, at
different stages due to the higher retention and lower To consider spatial variability of soil parameters,
conduction capacity of the silt. To investigate the influence horizontal correlation of 33.2, and vertical correlation of
of the variation of soil input parameters on the probability 2.08 as recommended by Cami et al. (2017) are
of failure, probabilistic analyses are carried out and considered in the analyses. The variation of mean FOS and
discussed in the following sections.
PF over time obtained from stochastic analysis is
presented in Figure 6 and Figure 7.
The stochastic analyses yield a probability factor (PF)
ranging from 6% to 8% for a factor of safety (FOS)
approximately equal to 1.23, which is significantly lower
than the PF obtained from SRV analyses. Nevertheless,
when comparing these findings to the results of a prior
study conducted by Baninajarian et al. (2023), it is evident
that considering cohesion in the analyses leads to a more
pronounced increase in the PF.

2.3.2 Spatial Variability Analysis Considering Cross-

Correlation

Slope stability analyses using cross-correlated soil

properties can predict reasonable probabilities of failure. In
this study, a negative cross-correlation of -0.5 between
cohesion and friction angle and a positive cross correlation Figure 7. Variation of PF over time - SRV Analysis and
of 0.5 between friction angle and unit weight are Stochastic Analysis
considered (Javankhoshdel & Bathurst, 2016). The
negative cross correlation between friction angle (Phi) and
cohesion can be observed from Figure 8 which illustrates
the friction angle and cohesion contours for a specific slip
surface. It can be seen that in the area with greater Phi
value, there exists a lower cohesion value.
As shown in Figure 6 and Figure 7, mean FOS resulted
from stochastic analyses considering above mentioned
cross correlation coefficients is around 1.23 with a PF of
about 2%. These figures indicate that considering the
mean factor of safety of around 1.2, stochastic analyses
without cross-correlated variables lead to a relatively high
probability of failure compared to the analyses with cross-
correlated variables.

Figure 8. Friction angle and cohesion contours for a

specific slip surface

3 CONCLUDING REMARKS

This research aims to examine how the spatial

variations in soil strength parameters, such as friction angle
cohesion, and unit weight, influence the likelihood of failure
in soil embankments. The study employs a probabilistic
Figure 6. Temporal variation of FOS over time - SRV analysis that takes into account the spatial variability of soil
Analysis and Stochastic Analysis parameters for an embankment subjected to an extreme
rainfall event with different initial saturation conditions. The
RLEM approach has been for the probabilistic analysis.
Results showed that considering spatial variability of soil
properties and the negative cross-correlation between the
cohesion and friction angle will reduce the probability of
failure significantly.
4 REFERENCES Input Motion in Stochastic Slope Stability.
Proceedings of the TMIC 2022 Slope Stability
Baninajarian, L. (2020). Effect of Future Extreme Conference (TMIC 2022), 179–198.
Precipitation Events on the Stability of Soil https://doi.org/10.2991/978-94-6463-104-
Embankments Across Ontario. 3_17
https://yorkspace.library.yorku.ca/xmlui/handl MTO. (1997). Drainage Management Manual. Ronin
e/10315/37975 House Publishing, under contract from Ministry
Baninajarian, L., Bashir, R., Ghassemi, A., DeSira, of Transportation of Ontario, Ottawa, Ontario,
A., & Sangiuliano, T. (2020). Developement of Canada.
fragility curves for soil embankment slopes due Mualem, Y. (1976). A new model for predicting the
to future extreme rainfall events. hydraulic conductivity of unsaturated porous
Cami, B., Javankhoshdel, S., Lam, J., & Bathurst, R. media. Water Resources Research, 12(3),
J. (2017). Probabilistic Analysis of a Tailings 513–522.
Dam using 2D Composite Circular and Non- https://doi.org/10.1029/WR012i003p00513
Circular Deterministic Analysis, SRV Phoon, K.-K., & Kulhawy, F. H. (1999). Evaluation of
Approach, and RLEM. Geo Ottawa. geotechnical property variability.
Cami, B., Javankhoshdel, S., Phoon, K.-K., & Ching, Pk, S., Bashir, R., & Beddoe, R. (2019). Effect of
J. (2020). Scale of Fluctuation for Spatially climate change on earthen embankments in
Varying Soils: Estimation Methods and Values. Southern Ontario, Canada. Environmental
ASCE-ASME Journal of Risk and Uncertainty Geotechnics, 8(2), 148–169.
in Engineering Systems, Part A: Civil https://doi.org/10.1680/JENGE.18.00068
Engineering, 6(4). Price, V. E., & Morgenstern, N. R. (2015). The
https://doi.org/10.1061/AJRUA6.0001083 Analysis of The Stability of General Slip
Griffiths, D. V., Huang, J., & Fenton, G. A. (2009). Surfaces.
Influence of Spatial Variability on Slope Https://Doi.Org/10.1680/Geot.1968.18.3.393,
Reliability Using 2-D Random Fields. Journal 18(3), 393–394.
of Geotechnical and Geoenvironmental https://doi.org/10.1680/GEOT.1968.18.3.393
Engineering, 135(10), 1367–1378. Rahardjo, H., Nio, A. S., Leong, E. C., & Song, N. Y.
https://doi.org/10.1061/(ASCE)GT.1943- (2010). Effects of Groundwater Table Position
5606.0000099 and Soil Properties on Stability of Slope during
Harr, M. E. (1984). Reliability-based design in civil Rainfall. Journal of Geotechnical and
engineering. Henry M. Shaw Lecture, Dept. of Geoenvironmental Engineering, 136(11),
Civil Engineering, North Carolina State 1555–1564.
University, Raleigh, NC. https://doi.org/10.1061/(ASCE)GT.1943-
Javankhoshdel, S., & Bathurst, R. J. (2016). 5606.0000385
Influence of cross correlation between soil Rocscience, S. 2D limit equilibrium slope stability
parameters on probability of failure of simple analysis. (2022). Slide2 Overview.
cohesive and c-φ slopes. Canadian https://www.rocscience.com/help/slide2/overvi
Geotechnical Journal, 53(5), 839–853. ew
https://doi.org/10.1139/CGJ-2015- Silva, F., Asce, M., Lambe, ; T William, Asce, H. M.,
0109/ASSET/IMAGES/CGJ-2015- Marr, W. A., & Asce, F. (2008). Probability and
0109IEQ45.GIF Risk of Slope Failure. Journal of Geotechnical
Javankhoshdel, S., Luo, N., & Bathurst, R. J. (2017). and Geoenvironmental Engineering, 134(12),
Probabilistic analysis of simple slopes with 1691–1699.
cohesive soil strength using RLEM and RFEM. https://doi.org/10.1061/(ASCE)1090-
Http://Dx.Doi.Org/10.1080/17499518.2016.12 0241(2008)134:12(1691)
35712, 11(3), 231–246. Vanapalli, S. K., Fredlund, D. G., Pufahl, D. E., &
https://doi.org/10.1080/17499518.2016.12357 Clifton, A. W. (1996). Model for the prediction
12 of shear strength with respect to soil suction
Javankhoshdel, S., Rezvani, M., Fatehi, M., & Characterization of the Mechanistic-Empirical
Jamshidi Chenari, R. (2022). RLEM versus Pavement Design Guide Input Parameters for
RFEM in Stochastic Slope Stability Analyses in the Resilient Modulus of Ontario Subgrade
Geomechanics. 241–250. Soils View project Model for the prediction of
https://doi.org/10.1061/9780784484036.025 shear. Article in Canadian Geotechnical
Keifer, C. J., & Chu, H. H. (1957). Synthetic Storm Journal. https://doi.org/10.1139/t96-060
Pattern for Drainage Design. Journal of the
Hydraulics Division, 83(4), 1–25.
Kulhawy, F. H. (1992). On the Evaluation of Soil
Properties. Geotechnical Special Publication,
31, 95–115.
Malekpoor, P. S., Lopez-Querol, S., &
Javankhoshdel, S. (2023). Spatial Variability of
Seepage and stability analyses for flood
embankment design
Huang, Ellen & Liu, Xiteng
Geotechnical Services Team, Stantec Consulting Ltd., Edmonton,
Alberta, Canada
Wood, Matt & Ofield, David
Water Services Team, Stantec Consulting Ltd., Edmonton, Alberta,
Canada

ABSTRACT
This paper provides a case study on the geotechnical design of a flood barrier system to reduce the risk of flooding for
EPCOR’s Water Treatment Plants (WTP), located on the lower terrace of the North Saskatchewan River (NSR) in
Edmonton. Earthen embankments were adopted as the main flood barrier in the site. The flood barrier was designed
following the US Army Corps of Engineers (USACE) design guidelines for various loading scenarios and conditions.
Coupled seepage and stability analyses were performed to evaluate the potential seepage and stability of the proposed
flood embankments using GeoStudio SEEP/W and SLOPE/W programs. Based on the analysis results, the thickness and
uniformity of the “impervious” blanket layer (native clay or clay fill) beneath the flood embankments is critical to seepage
flow and heaving during the flood event. While the amount of seepage flow is insignificant, the uplift pressure at the
downstream toe of the flood embankment could cause concerns of internal erosion and/or heaving. The USACE method
treats flood as a steady state condition. In reality, due to the transient nature of the flood event, the majority of the
embankment fill and foundation soil will remain unsaturated during the 1:500-year flood event. Such results and
conclusions bring some discussions on the design criteria for similar flood embankments design. The case study also
provides more value for future applications with similar conditions for economical design.

RÉSUMÉ
Cet article présente une étude de cas sur la conception géotechnique d'un système de barrière anti-inondation visant à
réduire le risque d'inondation pour les usines de traitement des eaux (WTP) d'EPCOR, situées sur la terrasse inférieure
de la rivière Saskatchewan Nord (NSR) à Edmonton. Des remblais en terre ont été adoptés comme principale barrière
antiinondations sur le site. La barrière antiinondation a été conçue conformément aux directives de conception de l'US
Army Corps of Engineers (USACE) pour différents scénarios et conditions de charge. Des analyses couplées d'infiltration
et de stabilité ont été effectuées pour évaluer l'infiltration potentielle et la stabilité des digues proposées en utilisant les
programmes GeoStudio SEEP/W et SLOPE/W. D'après les résultats de l'analyse, l'épaisseur et l'uniformité de la couche
de couverture "imperméable" (argile naturelle ou remblai d'argile) sous les digues d'inondation sont critiques pour le flux
d'infiltration et le soulèvement pendant l'événement d'inondation. Bien que la quantité de flux d'infiltration soit insignifiante,
la pression de soulèvement au niveau du pied aval de la digue d'inondation pourrait causer des problèmes d'érosion
interne et/ou de soulèvement. La méthode de l'USACE traite l'inondation comme une condition de régime permanent. En
réalité, en raison de la nature transitoire de l'événement d'inondation, la majorité du remblai de la digue et du sol de
fondation restera non saturée pendant l'événement d'inondation de 1:500 ans. Ces résultats et conclusions amènent des
discussions sur les critères de conception pour les digues d'inondation similaires. L'étude de cas fournit également plus
de valeur pour les applications futures avec des conditions similaires pour une conception économique.

This paper takes one of the WTPs – Rossdale WTP as

1 INTRODUCTION a case study, presents the methodology used for the
EPCOR Water Services Inc. operates two water treatments geotechnical design of a flood barrier system to reduce the
plants (WTPs) in the City of Edmonton – Rossdale WTP risk of flooding, summarizes the seepage and stability
and E.L. Smith WTP. Both plants are located on the lower analysis results, and discusses the design criteria for
terrace of the North Saskatchewan River (NSR) valley and similar flood embankment design.
can therefore be vulnerable to flooding. Due to the major 2 SITE DESCRIPTION AND LOCAL GEOLOGY
flooding event across Alberta in 2013, EPCOR’s insurance
provider FM Global was prompted to assess the Edmonton The Rossdale WTP site is located on the north side of the
WTPs’ vulnerability to flood from the NSR. NSR terrace immediately east of Walterdale bridge in
Stantec was retained by EPCOR to provide preliminary downtown Edmonton, Alberta. The Rossdale WTP has
engineering design for earthen embankments at the WTPs been in operation for over 100 years. The plant supplies
that could protect the facilities from a 1 in 500-year flooding Edmonton and its surrounding areas with drinking water.
event. Geotechnical analyses were performed to evaluate The site is relatively level with a gentle slope towards
the potential seepage and stability of the proposed earthen south to the NSR. Majority of the site is at elevations
embankments under various conditions. between 623 m and 625 m. Published local geology of
Edmonton (Bayrock and Hughes 1962, Kathol and
McPherson 1975) indicates that the NSR flood plain is 3.1 Seepage analyses
underlain by alluvial deposits and followed by interbedded
clay shale and sandstone bedrock of the Edmonton Seepage analyses were conducted to estimate location of
Formation. The alluvial deposits consist predominately of the phreatic surface, potential seepage to landside (plant
fine to medium grained sand with some silt and clay. side), and uplift pressure on the base of top stratum at the
Coarse sand and gravel is present in many places, landside toe of the embankments.
especially in or close to the river channel. A 2-Dimensional (2-D) finite element seepage program
Based on the historical boreholes data, the general soil Seep/W 2019 was used to assess event-based seepage
stratigraphy at the site consists of various fills overlying 4 potential on the preliminary embankment design. The
m to 7 m thick alluvial deposits overlying interbedded analyses focused on the local (fluvial) aquifer and the
bentonitic shale bedrock. The alluvial deposits typically potential for seepage immediately under or on the plant
include silty clays and/or clayey silts, sands, and gravel in side of the embankment.
descending order. Bedrock is expected to be encountered Seepage analyses were performed on three selected
approximately 9 m to 14 m below existing ground surface representative cross sections (namely A-A’, B-B’, C-C’) at
(an approximate elevation of 612 m to 615 m). the site (Figure 1). The results of Section B-B’ were
Based on the historical borehole data, the groundwater presented and discussed in the paper. The thickness of top
level was at approximately 7 m to 10 m below ground stratum (blanket layer) and underlying alluvial deposits
surface (bgs) (corresponding elevation of 616 m to 614 m) were defined by reviewing the historical borehole logs in
within the sand/gravel layer. the surrounding area of. The 2-D finite element model was
analyzed for both steady state seepage and transient
3 GEOTECHNICAL DESIGN OF FLOOD seepage conditions.
EMBANKMENT For steady state seepage analyses, the peak flood level
based on the 1:500-year flood was applied to the upstream
Two types of flood protection system - earth embankment side (river side) of the embankment as a total head
and pile supported concrete wall are proposed. Earthen boundary. Free seepage was allowed on the plant side.
embankments are proposed as the main flood barrier at the For transient seepage analyses, the 1:500-year flood
site. Flood walls are proposed in areas where spatial design hydrographs were used as boundary condition
constraint exists. The site plan and proposed flood (Tetra Tech 2020). The initial groundwater conditions were
embankment locations of the Rossdale WTP site are based on regional groundwater levels obtained from
shown in Figure 1. The top of the flood embankment is set previous field investigations.
to protect a 1:500-year flood plus a 1.0 m freeboard, which In the seepage analyses, both saturated and
equals to approximate elevation of 625.6 m at this site. In unsaturated hydraulic conductivities were used to properly
general, the proposed earthen embankments will be account for water migration in unsaturated zone above
constructed on the existing ground with a crown width of phreatic surface and to better predict the phreatic surface.
3.0 m and 3H:1V side slopes. For transient analyses, water storage functions were also
The flood barrier was designed following the USACE assigned to each soil unit above the bedrock. The
design guidelines for various loading scenarios and saturated hydraulic conductivities used in the analyses are
conditions. Coupled seepage and stability analyses were summarized in Table 1. They generally represent upper
performed to evaluate the potential seepage and stability bound values of previous field test results (Stantec 2021)
of the proposed flood embankments using GeoStudio for conservatism. The unsaturated hydraulic conductivities
SEEP/W and SLOPE/W programs. and water storage functions were estimated based on the
soil gradation curves (Seep/W 2012, Benson 2014).
Table 1. Saturated Hydraulic Conductivities

Material Type Horizontal Hydraulic Vertical Hydraulic

Conductivity kx, m/s Conductivity kz, m/s
Embankment Fill1 5.0 x 10-9 1.0 x 10-9
Clay Fill 1.0 x 10-8
2.0 x 10-9
Silt 1.0 x 10-6
2.0 x 10-7
Silty Sand 2.0 x 10-5
2.0 x 10-5
Sand and Gravel 1.0 x 10-4
1.0 x 10-4
Bedrock 1.0 x 10-10
1.0 x 10-11
1
medium to high plastic clay with plastic index between 20 and 40
Seepage analysis results of Section B-B’ are presented in
Figure 2 and Figure 3 and summarized in Table 2 below.

Figure 1. Rossdale WTP Site Plan and Flood

Embankments Location
Figure 2. Total Head Contour Peak Flood - Transient Flow

Figure 3. Total Head Contour Peak Flood - Steady State Seepage

Table 2. Seepage Analysis Results during Peak Flood For the proposed flood mitigation earthen
embankments, various loading conditions to which
Scenario Landside Seepage Uplift Pressure embankments and foundations may be subjected were
(m3/d/m) (kPa) considered in the slope stability analyses. According to
Steady State Seepage 2.0 x10-4 19.9 USACE EM-1110-2-1913 the loading conditions to be
considered are as follows:
Transient Flow 0 0 Case I - End of construction. This case represents
undrained conditions for impervious embankment and
From the seepage analysis results, under steady state foundation soils, i.e., excess pore water pressure is present
flood condition, the estimated seepage to plant side was under additional loading. Analysis for the End of
very small, up to 2.0 x10-4 m3/day per linear meter of the construction case was considered by using both total
embankment. Furthermore, the weight of the surficial stress approach (i.e., undrained shear strength) and
“impervious” soil stratum was greater than 1.5 times the effective stress approach with excess pore pressure. For
calculated uplift pressures underneath the surficial soil. each loading situation, only the case with lowest resulting
Therefore, no drainage or pressure relief is required on the Factor of Safety (FOS) are presented.
plant side. Under transient condition, no potential seepage Case II – Peak flood. This condition represents the
or uplift issues occurred on the plant side under the 1:500- water level reaching peak flood. Plant side embankment
year flood event. slope stability was examined using the following two pore
3.2 Slope stability analyses water pressure regimes:
• Case IIA: steady state seepage during peak flood
Slope stability analyses were carried out using a 2-D limit which means the peak flood stays long enough so
equilibrium computer program Slope/W 2019. The selected that steady state seepage occurs. This is the case
sections and corresponding finite element models required by USACE EM-1110-2-1913.
developed in the seepage analyses were used for slope
stability analyses. The pore-water pressure conditions for
stability analyses were imported from the seepage analysis
results (i.e., coupled stability and seepage analysis).
 • Case IIB: transient flow under 1:500-year return Case IV - Earthquake. Earthquake loading was
period flood event. Under transient flow, the modelled as a constant force in the pseudo-static (seismic)
embankment and foundation soil may not be analysis. The effects of earthquake shaking in the
saturated during flooding and is considered more horizontal and vertical directions were modelled using
representative of likely conditions. seismic coefficients kh and kv, respectively. Based on the
Case III - Rapid drawdown. Upstream slope failures NBC-AB (2019), the Peak Ground Acceleration (PGA) for
can result from a rapid drawdown of the water level, this site was estimated to 0.083. The coefficient kh was
whereby the river level falls faster than the soil can drain. taken as 50% of the PGA at the ground surface. A vertical
This causes the development of excess pore water seismic coefficient kv equal to 2/3 of kh was used. The kv
pressure in the embankment. The pore pressure that might acting in both upward and downward directions were
remain or develop in the embankment during the rapid analyzed, and the lowest factor of safety was selected.
drawdown of the reservoir was estimated through seepage The minimum required safety factors for the slope
analysis, and the result was applied to analyze the stability stability analyses under the aforementioned loading
of the upstream slope. Similar to peak flood, two pore water conditions are defined according to the USACE EM-1110-
pressure regimes were used in the rapid drawdown 2-1913 and Canada Dam Safety Guidelines (2013 Edition).
analyses: Soil parameters used for slope stability analysis are
• Case IIIA: assumed water retreat in 0.5 day from presented in Table 3. The values were estimated based on
steady state seepage after peak flood. This is the the field and laboratory test results from previous
case required by USACE EM-1110-2-1913. investigations (Stantec 2021) and published data of similar
• Case IIIB: assumed water retreat in 0.5 day from materials.
transient flow after 1:500-year return period flood Slope stability analyses were performed on the
event. Again, this case has low pore pressure in representative cross sections. River side and plant side
general and is considered more representative of slope stabilities were examined.
likely conditions. Results of the slope stability analysis of Section B-B’
are presented in Figures 4 to 9 and summarized in Table 4
below.
Table 3. Soil Parameters for Slope Stability Analysis

Material Type Soil Model Unit Weight, Cohesion, Friction Angle, Excess Pore Cu2 (kPa)
(kN/m3) c’ (kPa) φ’ (°) Pressure, Ru
Embankment Fill M-C 19 5 24 0 -
Clay Fill M-C 18 2 24 0.25 30
Silt M-C 18 0 25 0 30
Silty Sand M-C 18 0 30 0 -
Sand and Gravel M-C 20 0 32 0 -
Bedrock 1
High Strength 21 - - - -
1
assumed to be impenetrable
2
undrained shear strength
Table 4. Summary of Slope Stability Analysis Results

Sub-Case Factor of Safety

Target Calculated
End of Construction 1.3 2.43
Peak Flood - Steady State Seepage 1.4 2.11
Peak Flood - Transient Flow 1.4 2.77
Rapid Drawdown - Steady State Seepage 1.0 1.26
Rapid Drawdown - Transient Flow 1.0 2.21
Seismic Loading 1.0 2.46
Figure 4. Slope Stability Analysis Results - End of Construction

Figure 5. Slope Stability Analysis Results - Peak Flood (Steady State Seepage)
Figure 6. Slope Stability Analysis Results - Peak Flood (Transient Flow)

Figure 7. Slope Stability Analysis Results - Rapid Drawdown (Steady State Seepage)
Figure 8. Slope Stability Analysis Results - Rapid Drawdown (Transient Flow)

Figure 9. Slope Stability Analysis Results – Seismic

Based on the slope stability analyses, the calculated FOS adequate factor of safety. Therefore, no drainage
for all cases met the targeted FOS. The designed 3H:1V and/or pressure relief measures plant side of the
side slope is considered adequate for both river side and embankment is required.
plant side slopes from a stability perspective. • Due to the transient nature of the flood event,
majority of the embankment fill and foundation soil
3.3 Summary of Analyses Results
will remain unsaturated during the 1:500-year flood.
Based on the geotechnical engineering analyses results, This further reinforces the unnecessity of any
the proposed earthen embankments in both WTP sites can seepage control measures stated above.
satisfy the seepage and slope stability requirement from a • The designed 3H:1V embankment side slopes can
geotechnical perspective. The findings are summarized as satisfy the stability requirements for various loading
follows: scenarios specified in USACE EM-1110-2-1913.
• Under steady state 1:500-year peak flood,
negligible seepage will be encountered on the plant
side of the embankment due to the presence of
relatively thick (typically greater than 2.0 m)
“impervious” blanket consisting of native clay, clay
fill, and/or silt fill. The uplift pressure under the
“impervious” blanket was counterbalanced by its
self-weight and shear strength of the blanket with
4 CONCLUSIONS AND DISCUSSIONS
The flood embankment was analyzed and designed in
accordance with the USACE Method in terms of both
seepage and stability. Coupled seepage and stability
analyses were conducted for the embankment design.
The USACE method treats flood as a steady state
condition. In reality, due to the transient nature of the flood
event, the majority of the embankment fill and foundation
soil will remain unsaturated during the 1:500-year flood
event. From seepage and slope stability analysis results
of the proposed flood embankments, it can be seen that the
USACE method gives very conservative results for the
proposed flood embankment design. Such results and
conclusions bring some discussions on the design criteria
for similar future applications with similar conditions for
more economical design.
5 ACKNOWLEDGEMENTS
The authors wish to thank the EPCOR for sponsoring the
Rossdale WTP Flood Mitigation Embankments Design
project and their support.
6 REFERENCES
Bayrock, L.A. and Hughes, G.M. 1962, Surficial Geology of
the Edmonton District. Alberta, Alberta Research
Council.
Benson, C.H., Chiang, I., Chalermyanont, T. et al. 2014.
Estimating van Genuchten Parameters α and n for
Clean Sands from Particle Size Distribution Data,
Geotechnical Special Publication. February 2014.
Canadian Dam Association, 2013. Dam Safety Guidelines
2007 (2013 Edition).
GEO-SLOPE International Ltd. 2012. Seepage Modeling
with SEEP/W - An Engineering Methodology, July 2012
Edition.
Kathol, C. and McPherson, R. 1975, Urban Geology of
Edmonton. Bulletin 32, Alberta Research Council.
National Research Council of Canada. 2019. National
Building Code of Canada Alberta Edition (NBC-AB,
2019), Volume 1.
Stantec Consulting Ltd. 2021. EPCOR WTP Flood
Mitigation Embankments –Geotechnical Desktop Study
Technical Memorandum. File No. 110146440, March
2021
Tetra Tech Canada Inc. 2020. Technical Memo: North
Saskatchewan River Flood Hydrographs and
Bathymetry for Groundwater Modelling Study of
Rossdale and E.L. Smith Water Treatment Plants,
Edmonton Alberta, September 28,2020. File:
ENG.EGEO03539-01
US Army Corps of Engineers (USACE). 2020. Engineering
and Design - Design and Construction of Levees,
Manual No. EM 1110-2-1913. April 2020
Analysis of Structure Damage from an
Adjacent Excavation
S.J. Boone, Ph.D., P.Eng.
Ground Rules Engineering Inc., London, ON
J.L. Carvalho, Ph.D., P.Eng.
WSP, Mississauga, ON
M. Kanungo, M.E.Sc., P.Eng.
GHD, Mississauga, ON

ABSTRACT
During construction of a water treatment plant addition, the existing structure suffered significant damage, had to be
removed from service and a forensic investigation ensued. Ground and structure responses to the adjacent excavation
were evaluated using analytical methods and finite difference numerical modelling. Input parameters for the soil constitutive
model were calibrated to laboratory data and preloading responses of the site predating construction of the original
treatment plant. The numerical model simulated the entire construction history of the existing building and neighbouring
excavation. The numerical modelling, coupled with analytical evaluation of other contributing ground displacement
mechanisms, corresponded to displacement measurements and remedial grout volumes, and provided valuable insight
into issues that are vital in selection, design, and construction of excavation support systems.

RÉSUMÉ
Lors de la construction d'une usine de traitement des eaux, la structure existante a subi des dommages importants, a dû
être retirée du service et une enquête médico-légale s'en est suivie. Les réponses du sol et de la structure à l'excavation
adjacente ont été évaluées à l'aide de méthodes analytiques et d'une modélisation numérique aux différences finies. Les
paramètres d'entrée pour le modèle constitutif du sol ont été calibrés sur les données de laboratoire et les réponses de
préchargement du site avant la construction de la station de traitement d'origine. Le modèle numérique a simulé tout
l'historique de construction du bâtiment existant et des fouilles avoisinantes. La modélisation numérique, associée à
l'évaluation analytique d'autres mécanismes de déplacement du sol contributifs, correspondait aux mesures de
déplacement et aux volumes de coulis correctifs, et a fourni des informations précieuses sur les problèmes vitaux dans la
sélection, la conception et la construction des systèmes de support d'excavation.

1 INTRODUCTION loose, laminated silt, clay and fine sand at the site
prompted a preloading program to control settlement and
During construction of a potable water treatment plant allow use of a relatively shallow and rigid raft foundation.
addition, the existing facility suffered significant damage, Preload settlement was measured for about 12 weeks;
had to be removed from service and a forensic however, monitoring was started about five days after fill
investigation ensued. Ground and structure responses to placement started to allow for protection of monitoring
the adjacent excavation were evaluated using analytical instrument cables, thus early ground responses were
methods and finite difference numerical modelling. Input missed. Measured settlement remained unchanged after
parameters for the soil constitutive model were calibrated the first 45 days. After removing the preload, the treatment
to laboratory data and preloading responses of the site facility was constructed where the southern two thirds of
predating construction of the original treatment plant. The the building housed treatment tanks and the northern third
numerical model simulated the entire construction history included process equipment rooms and a three-story open
of the existing building and neighbouring excavation. The equipment bay. The building and tanks were constructed
numerical modelling, coupled with analytical evaluation of as an integral structure of reinforced concrete walls, floors,
other contributing ground displacement mechanisms, and roofs (Figure 1). Until the planned plant expansion, an
corresponded to displacement measurements and earth berm surrounded three sides of the chorine contact
remedial grout volumes, and provided valuable insight into chamber structure. Six key events in building the existing
issues that are vital in selection, design, and construction plant are summarized in Table 1 (Stages 1 through 6).
of excavation support systems. In the late-2000s construction began on a major facility
expansion immediately beside the existing plant. The
2 SITE & CONSTRUCTION HISTORY facility addition required a 5.5 m deep excavation adjacent
to the chlorine contact chambers (water storage tank) and
Development of the water treatment plant site evolved over the new structure and internal equipment was to be
more than a century, initially using a gallery of intake pipes supported on driven H piles. Interlocking 10 m long Z60
buried in the natural sand and gravel of a former river steel sheets were driven about 1 m away from and parallel
meander. A new treatment facility was built on the site in to the exterior western wall of the chlorine contact
the late 1990s (see Figure 1). A deep layer of saturated, chamber. Soil and rock anchors provided resistance to
Figure 1. Profile through existing treatment plant, area of addition and ground conditions. Inset illustrates treatment plant
structure in northern third of building.

horizontal earth and surcharge pressures. Steel HP250x85

piles were driven at 2.4 m centers, inboard of and abutting
the sheets, intended to resist the vertical load component Table 1. Key stages of site development & construction
from the anchors. Interlocking sheet piles were also used
to support a smaller excavation about 25 m west of the Stage Description Stage Description
existing facility. 1 Open site 10 Partially excavate
Score marks on the north and west concrete walls of 2 Preloading 11 Dewater
the chlorine contact chamber were established for 3 Remove preload 12 Deepen sump
elevation monitoring points once the earth berm was 4 Construct plant 13 Top anchors
5 Fill tanks 14 Install H piles
removed. The temporary surveying benchmark was 6 Construct berm 15 Excavate notch
located on an exterior concrete pad near the north side 7 Remove berm 16 Lower anchors
equipment bay door frame. 8 Excavate support 17 Drain contact
Shortly after the lower level of anchors was installed sump bench chamber tanks
from within a narrow “notch” excavation (Stage 15), plant 9 Drive sheet piles
operators heard “two loud bangs” emanating from
somewhere within the existing structure. The following
morning, personnel found leaking cracks and a void 3 SUBSURFACE CONDITIONS
beneath the western edge wrapping around the north and
south sides of the chlorine contact chambers. About two In the mid-1990s, ground conditions were investigated with
weeks later, similar cracking or “bangs” were heard, leaks 12 boreholes (two of which were in the structure area) and
worsened, and new cracks opened. The plant owner conventional standard penetration testing (SPT), three test
initiated a building structural inspection and forensic pits, dynamic cone penetration tests, one flat-plate
investigation. dilatometer (DMT) sounding, 10 grain size distribution
A supplementary instrumentation and monitoring tests, one oedometer test, and two direct shear tests.
program was initiated immediately by the contractor These explorations identified the stratigraphy of sand and
following the damage event. Instruments included gravel fill (reworked local materials), natural sand and
precision survey targets on the sheet piles, the gravel, fine-grained deposits (silt, clay and fine sand) and
supplementary H-piles, the chlorine contact chamber, and bedrock as illustrated in Figure 1. In the late 2000s, two
other areas of the structure, vertical and horizontal different geotechnical consultants completed
electrolevels, tilt meters, extensometers, and investigations at this site for the treatment plant expansion.
inclinometers. The first investigation included five boreholes, none of
Key stages during the plant expansion are also which penetrated the bedrock and the second consisted of
summarized in Table 1 (Stages 7 through 17). The last two boreholes in the plant expansion area, one of which
stage considered as part of the forensic analysis was when included rock coring. Laboratory testing was limited to
the chlorine contact chamber was unloaded to help water content determinations. Figure 1 illustrates the
mitigate further damage to the overall structure. subsurface conditions at the treatment plant and addition.
 Immediately following the damage, the owner initiated Table 2. Summary of geotechnical parameter values
a subsurface investigation consisting of eight boreholes
and eight grain size distribution tests, with the boreholes Parameter Sand & Gravel Silt, Clay & Sand
scattered around the existing facility and new addition SPT 'N' Value 12-52 7-12
footprint. Another investigation by the contractor included Gravel (%) 46-71 0
pre-bored Ménard pressuremeter tests (PMT) in the fine- Sand (%) 24-42 0-9
grained deposit within four boreholes, two borehole shear Silt (%) 69-90
5-12
Clay (%) 6-25
tests (BST) in the fine-grained soils, three boreholes with
Water content (%) 2-25 13-42
conventional standard penetration testing, and flat-plat Liquid Limit 18-39
dilatometer testing at five locations. Laboratory testing for Plastic Limit 13-18
both investigations consisted of water content Plasticity Index np-23
determinations. Su (kPa) 70-240
Later, during the forensic investigation, two additional ' (degrees) 41-42.5 30-40
boreholes were completed, one of which included large Dry density, d (kN/m3) 19-21.5
volume sampling of the native and fill sand and gravel and Relative density, Dr (%) 35-65
a porous tip piezometer installed in the fine-grained Wet density, w (kN/m3) 21-22 16-20
deposits. Piezocone penetration testing (CPTu) was also ch x10-1 cm/s 2 to 7
conducted through the saturated fine-grained soils within cv x10-4 cm/s 6-23
which nine pore water pressure dissipation tests were OCR 1.7-4
completed. Laboratory testing included natural water Cr 0.005-0.05
content determinations, mechanical and hydrometer grain Cc 0.05-0.37
size distribution analyses, two oedometer tests and four Su = undrained shear strength; = effective stress angle of
isotropically consolidated undrained triaxial compression internal friction; cv = horizontal coefficient of consolidation; cv =
vertical coefficient of consolidation; OCR = overconsolidation
tests on thin-wall tube samples, three large box direct
ratio; Cr = recompression index; Cc = virgin compression index;
shear tests and a maximum-minimum density test of np = non-plastic
reconstituted samples of the sand and gravel. Table 2
summarizes basic geotechnical properties of the strata as
derived from the field and laboratory testing. 4 NUMERICAL MODELING

Figure 2. Examples of calibrating numerical model input parameters to site-specific laboratory and field testing.
 4.3 Selection & calibration of geotechnical input
parameters

To overcome biases in forensic numerical analyses when

the end-performance is known (e.g., Lambe 1973; Boone
2005), input values to the numerical model were calibrated
to laboratory oedometer and triaxial test results, with
interpretation of field variability based on site-specific
correlations to indicator parameters of water content and
CPTu tip resistance (see examples in Figure 2). The
calibrated FLAC input data was then used to simulate site
preloading and compared to available monitoring data
(Figures 2 and 3).

4.4 Structural Components

All structural members, except the tiebacks, were

modelled as two-dimensional beam elements. The
Figure 3. Simulation of preload and comparison to
tiebacks were modelled using one-dimensional axial
measured settlement.
elements attached to the grid along the bond zone length
so that cable element forces could be tracked. The
4.1 Model
reinforced concrete structure was modelled by simplifying
the beam and column construction along two east west
The FLAC (Itasca 2008) finite difference software was used
planes, representative of the northern third and southern
for this analysis for its ability to model large strains and
two thirds of the facility. The simplified beams were all
coupled mechanical and fluid flow problems. The model
considered to be rigidly connected and capable of moment
geometry was based directly on drawings of the existing
transfer. To avoid unnecessary complications associated
pre-treatment structure, preload construction, and shoring
with self-weight beam bending the beams were considered
using a uniform square mesh of 0.25 m, adjusted to
weightless. Between the beams and columns, walls were
triangular elements at any angled zones. The boundaries
modelled as materials with strength and deformation
were set at more than 3 times the total soil thickness to the
properties consistent with reinforced concrete with
nearest structural element. Vertical boundaries permitted
deformation modulus values reduced proportional to the
vertical displacement but prevent horizontal displacement.
cross-sectional area perpendicular to the model plane. The
The bottom boundary prevented vertical and horizontal
simulated density of the walls was chosen to ensure
displacement to simulate a rough soil-rock contact so that
appropriate contact stresses between the building
failure paths would be forced through the soil mass. The
structure and soil. Based on an evaluation of building
pore water pressure profile at the vertical and bottom
structure dimensions and materials the raft-soil contact
boundaries was fixed to represent the pre-construction
pressure was expected to be 55 to 60 kPa. In those areas
ground water conditions and cross-boundary flow was
with water tanks, the contact pressure was estimated to be
permitted.
about 135 kPa, with an overall average contact pressure
(north and south building sections) estimated to be about
4.2 Soil Constitutive Behaviour
105 kPa. Interfaces between the beam elements and
ground were modelled with stiffness of at least one order
The cap-yield model was used to simulate soil behaviour
of magnitude greater than the soil modulus values to avoid
since it considered non-linear stress-strain soil behaviour,
inappropriate influence on overall behaviour. The
dilation or contraction of the soil under loads and coupled
interfaces were also modeled to be incapable of resisting
fluid-flow and mechanical stress-strain (pore water
tensile forces but capable of transmitting shear forces so
pressure) responses to loading. The model included strain-
that separations between the structure and ground could
hardening/softening behaviour, an elliptical volumetric
occur depending on relative displacement and stiffness
yield surface and deformation modulus values defined as
characteristics. In one of the modelling scenarios, the
functions of confining stress and simulation of pore water
chlorine contact chamber was permitted to rotate at a joint
pressure responses to contraction or dilation under shear
in the raft foundation at the junction with the remaining
or compression stresses. Key geotechnical input
treatment plant building to understand the influence of the
parameters included the tangent shear modulus, 𝐺𝑟𝑒𝑓 𝑒
,
multi-story open equipment bay that was not as heavily
slope of isotropic laboratory mean normal stress-strain interconnected with the chlorine contact tank as the
response, 𝐾𝑟𝑒𝑓𝑖𝑠𝑜
, oedometer modulus, 𝐸𝑟𝑒𝑓 𝑜𝑒𝑑
, all at the southern 2/3 of the structure.
reference effective pressure of 100kPa, and an exponent
controlling the confining stress-dependent change in 4.5 Modelling Results
modulus values, m. Site stratigraphy was modelled by a
system of 0.5 m thick layers to account for geotechnical Results of the numerical modelling are illustrated in Figures
parameter variability. 4 and 5. The open equipment bay within the northern part
Figure 4. Simulation of chlorine contact chamber exterior
wall settlement compared to survey monitoring
measurements. Coloured circles and numbers refer to the
corresponding line colours and the stages of construction
in Table 1.

of the structure likely acted as a structural fuse such that

when stresses exceeded the tensile strength of the few
beams and upper part of the raft, the structure no longer
redistributed stresses eastward and the chlorine contact
chamber section became free to rotate. Prohibiting
formation of a structural joint for the southern part of the
structure and allowing joint formation on the northern part
clearly matched well with the available survey data.

5 VOIDS
Figure 5. Simulation of east-west chlorine contact
While the numerical modelling, calibrated to past site chamber settlement profile (top) and lateral displacement
performance, provided a reliable method for simulating and of the sheet pile support system (bottom). Coloured circles
tracking performance of the excavation and building at and numbers refer to the corresponding line colours and
many stages, it did not address other observations at the the stages of construction in Table 1.
site.
Voids were identified beneath the exterior perimeter of and 25 mm/s with many instances well over this range to
the chlorine contact chamber west wall and the western as much as 112 mm/s. Using the methods of Whyley and
half of the north and south walls. Once the chlorine contact Sarsby (1992) and Hope and Hiller (2000) and equipment
chamber was emptied (Stage 17), the floor was cored in data for the ICE 812 vibratory and B-4505 impact
multiple locations to define the extent of the voids and allow hammers, ground ppv values were estimated to be as
filling with low-mobility grout. Figure 6 illustrates the much as 100 mm/s at the piles to about 50 mm/s at the
locations that accepted grout and showed communication closest point of the chlorine contact chamber. Vulnerability
between holes and those that did not. In some instances, of the soils to vibration-induced liquefaction and
void thicknesses on the order of 25 mm were identified in densification was evaluated using CPTu and SPT data and
holes that eventually did not accept grout. It was concluded methods of Massarsch (2002), PDI (2010) and Taylor
that the inferred void height was associated with (2011). The evaluation determined that pile-driving-
disturbances from the core hole drilling and that the void induced liquefaction and settlement (hazard extent) was
did not propagate further from the hole. The grouted area probable for about 50% of the soil profile depth within a
shown in Figure 6 was based on assuming that grout radial zone of 4 to 8 m from each pile.
travelled half the distance between holes that accepted Based on the work of Clough and Chameau (1980) and
grout and those that did not. Grout communication was Drabkin et al. (1996), settlement at the boundary of the
confirmed among all holes shown as accepting grout. One chlorine contact chamber from driving individual sheet piles
additional large volume void was identified during grouting was estimated to be on the order of 35 to 70 mm. Driving
over a location where a tieback was drilled through one of the supplementary H piles and building piles exacerbated
the buried and abandoned water intake pipes. Ground settlement. Figure 6 illustrates contours of likely cumulative
losses were not evident during the tieback drilling other pile-driving-induced ground settlement based on these
than loss of drilling fluid circulation and excessive grout analyses. As a comparison, the temporary benchmark
injection. Given the patterns of void thickness and grout used by the contractor for monitoring surveying, located on
volumes, the cause of the voids was suspected to be a concrete pad outside the north side equipment door,
vibration-induced densification of the granular soils. settled 9 mm during the time that the sheet piles were
Limited vibration monitoring of the concrete structure installed and prior to driving the structural piles.
indicated that sheet pile driving likely resulted in typical
structural peak particle velocities (ppv) between at least 5
Figure 6. Plan view of physical features of chlorine contact chamber, excavation support, exploratory core holes,
grout injection results, and estimated settlement induced by pile driving. Contours of settlement shown in 25 mm
increments from zero (green) to >150 mm. Numbers next to hole markers indicate measured thickness of voids (mm).

While the analysis of vibration-induced densification of • there was no monitoring of the shoring system or
the ground was inexact, the results were rational and adjacent ground prior to the damage event
consistent with the injected grout volumes, measurements • DMT data were not reliable, and settlements estimated
of void thickness (except one area), grout communication using Schmertmann’s (1987) methods would have
between injection holes, and the temporary benchmark resulted in values 10 times those measured
settlement in both pattern and magnitude. This mechanism • BST tests could not discern pore-water pressure
was thus considered the likely cause of the perimeter void effects on the results and the tests were not conducted
around the otherwise heavily loaded chlorine contact slowly to reduce pore water pressure influences on the
chamber. results
• PMT holes were not controlled well, pore water
pressure effects and interpretation methods were
6 COMPLICATING FACTORS inappropriate for the highly layered and silty ground
conditions
During the forensic investigation, multiple other problems • hammer systems (e.g., automatic, rope and cathead)
related to the shoring design and initial post-failure used for obtaining SPT data were not consistently
investigations were identified: identified
• unidentified “corrections” were applied to SPT data on
• the first design of the shoring system did not account the borehole records from one consultant, obfuscating
for the surcharge pressures of the existing treatment field data and prohibiting rational comparison to other
facility SPT data
• the wale-to-sheet pile distance was smaller than the H • other numerical models used to explain the site
pile depth so the wales had to be cut and reinforced to behaviour used traditional finite-element software,
permit supplementary H pile driving linear-elastic, perfectly plastic Mohr-Coulomb
• the supplementary H piles were not connected to the constitutive models (ignoring unloading and small
wales or wall at the time the upper level of anchors strain responses), and modelled the building as a
were stressed or when damage occurred simple, continuous rigid or flexible beam – and none
• the buried and abandoned pipes were not identified as matched the observed conditions
part of the contract information
• the temporary benchmark used for surveying was too These issues led to a variety of alleged causes of the
close to the ground and building that were subject to problems experienced at this site, such as:
displacement and was ultimately within the zone of
influence and moved
• the buried and abandoned water intake pipes were the Massarsch, K.R. 2002. Effects of Vibratory Compaction.
primary source of settlement Vibratory Pile Driving and Deep Soil Compaction –
• dewatering, necessitated by the design, was the TRANSVIB2002, Proceedings of the International
source of settlement Conference on Vibratory Pile Driving and Deep Soil
• ground conditions were different (softer) than Compaction, Belgium, Holeyman et al. eds., Balkema,
expected 33 – 42.
• construction of the berm surrounding the north, west Pile Dynamics Inc. (PDI). 2010. GRLWEAP™ - Wave
and south sides of the chlorine contact chambers Equation Analysis of Pile Foundations. Pile Dynamics
created the observed voids Inc.
• removal of the berm, necessitated by design, resulted Schmertmann, J. 1987. Dilatometer to Compute
in rebound of the structure and the observed cracking Foundation Settlement. Use of In Situ Tests in
Geotechnical Engineering, Geotechnical Special
None of these other alleged causes matched the patterns Publication No. 6, ASCE, 303-321.
and sequences of displacement and damage; however, the Taylor, O-D., S. 2011. Use of an Energy-Based
alleged causes prompted the need for detailed evaluations Liquefaction Approach to Predict Deformation in Silts
to discern and weigh the importance of each in the overall Due to Pile Driving. Ph.D. Dissertation, University of
forensic investigation. Rhode Island, 303 pp.
Whyley, P.J., and Sarsby, R.W. 1992. Groundborne
7 CONCLUSIONS vibration from piling. Ground Engineering, 32–37.
Woods, R.D. and Sharma, V.M. 2004. Dynamic Effects of
The detailed numerical modelling, calibrated to site- Pile Installations on Adjacent Structures, International
specific laboratory and field testing and a prior site preload Edition. A.A. Balkema Publishers, New York, 163 pp.
enabled the full sequence of site development to be reliably
simulated. Key to this work was developing a rigorous
model of the building structure and adapting it to recognize
the different stiffness and weak point in the northern third
of the structure. The results of both modelling scenarios
(south and north building models) explained the different
building settlement responses identified by the surveying
as well as the rationale for the loud “bangs” heard when
concrete fractured in tension. Coupled with analytical
evaluation of vibration-induced settlement, the work
described and summarized in this paper fully explained the
complex responses of the building, shoring and ground,
and was supported by limited monitoring data. With results
available for each construction stage, many of the alleged
causes of building damage could be evaluated and
ultimately dismissed.

Boone, S.J. 2005. General Report: Deep Excavations. Fifth

International Symposium on Geotechnical Aspects of
Underground Construction in Soft Ground, International
Conference on Soil Mechanics and Geotechnical
Engineering, Balkema, 81 – 92.
Clough, G.W. and Chameau, J.-L. 1980. Measured Effects
of Vibratory Sheetpile Driving. Journal of the
Geotechnical Engineering Division, ASCE, 106(10),
1081 – 1099.
Drabkin, S., Lacy, H., and Kim, D.S. 1996. Estimating
Settlement of Sand Caused by Construction Vibration.
Journal of Geotechnical Engineering, ASCE, 122(11),
920 – 928.
Hope, V.S. and Hiller, D.M. 2000. The prediction of
groundborne vibration from percussive piling. Canadian
Geotechnical Journal, Vol. 37, 700 – 711.
Itasca 2008. FLAC2D user manual. Itasca Consulting
Group Inc., Minneapolis, Minn.
Lambe, W.T. 1973. Prediction in soil engineering.
Geotechnique, 23(2), 149 – 202
Understanding the influence of boundary
conditions and thermophysical soil
parameters on thermal modelling in
permafrost regions
Konstantin Ozeritskiy
Department of Civil Engineering – University of Calgary, Calgary,
AB, Canada
Jocelyn L. Hayley
Department of Civil Engineering – University of Calgary, Calgary, AB, Canada
Alexey Gunar
Department of Geocryology – Lomonosov Moscow State University, Moscow, Russia

ABSTRACT
This article uses an approach which was developed at the Department of Geocryology of Lomonosov Moscow State
University. This approach provides a method to calibrate the thermal boundary condition on the ground surface using
borehole soil temperature measurements in combination with thermophysical soil parameters. This article presents thermal
calculations using one-dimensional heat conduction. Some errors may occur during surveys when determining the
thermophysical properties of soils and surface boundary conditions. The paper examines the influence of these errors on
simulated thermal fields. Suggestions for minimizing the influence of errors in the thermal modelling arising from the
thermophysical soil properties are suggested. The performed calculations identify that the quality of boundary conditions
affects the result of thermal modelling more than the quality of thermophysical properties. In addition, to conduct more
accurate thermal modelling, it is necessary to take a more detailed approach to the characterization of snow and vegetation
covers.

RÉSUMÉ
Cet article utilise une approche qui a été développée au Département de Géocryologie de l'Université d'État de Moscou
Lomonossov. Cette approche fournit une méthode pour étalonner la condition limite thermique à la surface du sol en
utilisant des mesures de température du sol dans les trous de forage en combinaison avec des paramètres
thermophysiques du sol. Cet article présente des calculs thermiques en utilisant la conduction de chaleur
unidimensionnelle. Des erreurs peuvent survenir lors des enquêtes pour déterminer les propriétés thermophysiques des
sols et les conditions limites de surface. Le document examine l'influence de ces erreurs sur les champs thermiques
simulés. Des suggestions sont proposées pour minimiser l'influence des erreurs dans la modélisation thermique découlant
des propriétés thermophysiques du sol. Les calculs effectués montrent que la qualité des conditions limites affecte
davantage le résultat de la modélisation thermique que la qualité des propriétés thermophysiques. De plus, pour réaliser
une modélisation thermique plus précise, il est nécessaire d'adopter une approche plus détaillée de la caractérisation des
couvertures de la neige et de la végétation.

1 INTRODUCTION remains a significant gap in the methodology for

determining and verifying the initial boundary conditions.
The development of territories with permafrost is inevitably The lack of a generally accepted approach to the
accompanied by significant environmental disturbances, implementation of thermal modelling is a serious scientific
changes in the thermal regime of soils, and geocryological challenge since predictive thermal modelling is one of the
conditions, leading to negative cryogenic processes that crucial parts in the design of engineering structures on
have the potential to cause damage to infrastructure. An permafrost soils and, thus, in the development of the
accurate geotechnical forecast makes it possible to northern regions.
develop preventive measures to mitigate negative This paper considers the influence of errors in the initial
consequences and improve the quality of design solutions and boundary conditions of the model on the predicted
necessary for the development of territories and the temperature regime of soils.
preservation of a safe ecological state of the environment.
One of the main components of the geotechnical
forecast is thermal modelling. Setting the appropriate initial 2 MODERN ISSUES OF GEOTECHNICAL
and boundary conditions is critical to accurately model the FORECAST
thermal behaviour of permafrost. This requires careful
consideration of the local environment and the physical One of the main tasks in solving the problems of
processes occurring in the soil. developing territories with permafrost soils is predicting
While the mathematical component of the thermal changes in geocryological conditions. The development of
solution does not require further development, there such forecasts should be based on a comprehensive study
of possible changes in climatic, natural, and environmental 3 INITIAL CONDITIONS
conditions, which is a complex problem (Ershov et al.,
2001). Having carried out geocryological studies and The initial temperature distribution in the soil is a critical
knowing the nature of anthropogenic/technogenic impacts, input for the thermal modelling of permafrost. This is
it is possible to make a geotechnical forecast. A typically obtained from measurements of the soil
geotechnical forecast is a way to predict the qualitative or temperature at depth.
quantitative characteristics of changes in the permafrost The other important factor to consider is the
under the influence of certain types of impacts in different thermophysical properties of the soils. The thermophysical
areas. properties of soils during the conductive transfer of thermal
Ground-surface conditions heavily influence the energy are evaluated by three main characteristics: heat
thermal regime of permafrost soils. They vary based on capacity, thermal conductivity, and thermal diffusivity. The
local air temperature, wind speed, precipitation, vegetation thermophysical properties of soils are characterized by a
type and cover, snow cover, slope and aspect, and solar significant dependence on the composition, structure, and
radiation. Anthropogenic impacts can change one or more state of the soils.
components of the energy balance, leading to a change in
the temperature regime of permafrost. 3.1 Heat capacity
Different landscapes, as well as relief affect snow cover
distribution. Various studies show that the direction of the The volumetric heat capacity of soils (kJ/(m3°C)) is the
prevailing winds and the positioning of the slopes of natural amount of heat required to change the temperature of a
irregularities affect snow depth (Gascoin et al., 2013, unit volume of soil by one degree. Due to the heat capacity
Draebing et al., 2017). In the Arctic, snow tends to have of a substance being equivalent to the sum of its
ecosystem-specific properties (Sturm et al., 1995), components’ heat capacity, the volumetric heat capacity of
generally associated with landscapes and climates. For soil can be obtained by adding the volumetric heat capacity
example, increased accumulation of snow at the bottom of of its different constituent parts.
the embankment occurs in the tundra due to wind
redistribution and has a warming effect on the temperature 3.2 Thermal conductivity
regime of the soil (O'Neill et al., 2017).
Simulation results for Barrow, Alaska, for the period The thermal conductivity of soils (W/(m°C)) can be
1977–1998 showed that about half of the warming of expressed as the amount of heat that flows across a
permafrost at a depth of 20 m was associated with an surface of a unit area of soil in a unit of time under a unit of
increase in the thickness of the snow cover and the rest the temperature gradient of the soil surface. There is an
with an increase in air temperature (Stieglitz et al., 2003). intrinsic relation between thermal conductivity and mineral
The results of another study indicate that snow cover can content, porosity, degree of water saturation, and
lead to higher soil temperatures in northern regions over a temperature.
long-term period, especially in areas covered by There are laboratory and calculated methods for
continuous permafrost (Park et al., 2015). On the other determining the thermal conductivity of soils. In the
hand, the absence of seasonal snow cover in laboratory, researchers have used transient line heat
discontinuous and sporadic permafrost regions may be a source methods to measure thermal conductivity.
key factor for permafrost development (Zhang, 2005). Typically, a probe for this measurement consists of a
According to a study by Grünberg et al. (2020), to needle with a heater and temperature sensor inside
determine the relationship between vegetation and (Decagon Devices, Inc., 2016). The accuracy of
permafrost, a strong influence of vegetation on snow determining thermophysical parameters is ±10% in the
accumulation was found. The ability of vegetation to trap interval from 0.2 to 4.0 W/(m°C).
snow has been described previously (Frost et al., 2018). In more recent times, three studies by Balland and Arp
Furthermore, snow in shrub areas tends to be thicker and (2005), Coté and Konrad (2005), and Lu et al. (2007) have
more effective as a heat insulator per unit volume than the tried to improve on Johansen’s model for calculating
snow outside of shrub patches (Sturm et al., 2001). thermal conductivity of soils. The comparison results by
The development of territories in permafrost regions Barry-Macaulay (2015) demonstrate that, in general, all
has been a significant factor in forming modern four models show good agreement between experimental
landscapes. In addition, the development of territories has thermal conductivity and modelled thermal conductivity.
also led to the creation of new infrastructure and The root mean square error value can reach 0.15 for fine-
transportation networks, such as highways and railways. grained soils and 0.20 for coarse-grained soils.
Human activities such as mining, oil and gas development,
and infrastructure construction can alter the landscape by 3.3 Thermal diffusivity
removing vegetation cover and altering the surface
topography. Therefore, geotechnical forecasting in The rate of change of the temperature field over time in a
developed areas is complicated, and many factors cause medium without internal heat sources is determined by the
uncertainties in surface conditions. The task of thermal diffusivity (m2/s). In heat transfer analysis, thermal
geotechnical forecasting is precisely the elimination of diffusivity is the thermal conductivity divided by volumetric
these uncertainties. capacity at constant pressure.
4 BOUNDARY CONDITIONS where Rsnow is the thermal insulance of snow cover,
(m2°C)/W, and Rvegetation is the thermal insulance of
The assignment of boundary conditions on the ground vegetation cover, (m2°C)/W.
surface is complicated, which is explained by the presence In turn, thermal insulance is determined in terms of the
of three types of heat transfer there simultaneously: thermal conductivity of the material:
conductive, convective, and radiant. Consequently, many
parameters determine this heat transfer. R = l/λ [9]
The surface energy balance equation can be written for
a heat transfer analysis as follows: where l is the length of the heat circuit section, m, and
λ is the thermal conductivity of the material, W/(m°C).
(qns-qnl ) = qsens+qlat+qg [1] The equation obtained by Proskuryakov B.V has been
widely used in engineering practice:
where qns is net solar radiation, qnl is net terrestrial
radiation, qsens is sensible heat flux, qlat is latent heat flux, λ = 0.021+1.01·10-3·ρ [10]
and qg is ground heat flux.
Convective heat transfer involves the combined According to a comparative study conducted by Osokin
processes of conduction and heat transfer by bulk fluid et al. (2017), values calculated using Equation 10 correlate
flow, referred to as heat advection. The rate equation well with the data obtained from field measurements.
describing convective heat transfer (Newton’s Law of The values of the thermal conductivity of vegetation for
Cooling) is: frozen and non-frozen conditions depend on humidity
(Gavriliev, 2004; Aleksyutina et al., 2011). The value of
q = h(Tg-Ta) [2] thermal conductivity of vegetation can vary in a wide range
of values, from 0.05 to 0.46 W/(m°C) for the thawed state
where q is the surface heat flux due to convection, and from 0.07 to 1.14 W/(m°C) for the frozen state.
W/m2, h is the convection heat transfer coefficient, However, empirical relationships have not yet been
W/(m2°C), Tg temperature of the surface, °C, and Ta the developed.
temperature of the fluid outside the thermal boundary layer, The article by Gorelik and Pazderin (2017) proposes to
°C. select the upper boundary conditions so that under these
Based on Equations 1 and 2, the surface temperature conditions, the soil temperature at the depth of zero annual
can be written as: amplitude and the depth of seasonal thawing should
coincide with their measured values during surveys. Thus,
Tg = Ta+(R-P)/h [3] the problem of determining the initial boundary conditions
can be reduced to the selection of the value of the thermal
where R is the radiation balance of the soil surface, resistance of the vegetation cover.
W/m2, and P is heating loss from the day surface due to Scientists have long identified the surface cover’s
evaporation and heating of underlying rocks and phase poorly constrained thermal properties as a critical source
transitions in them, W/m2. of uncertainty for ground temperature modelling
Based on the works of Khrustalev L.N., semi-imperial (Westermann et al., 2013; Decharme et al., 2016). Careful
formulas for calculating these parameters were derived selection and accurate specification of boundary conditions
(Ershov et al., 1999): are essential for obtaining reliable predictions of
temperature distribution.
R = 0.61TSI-20 [4]
P = 0.49TSI-60 [5]
5 BASIS OF SIMULATION
where TSI is the total solar irradiance, W/m2.
Ershov (1999) proposes the following formula for A borehole located on the Yamal Peninsula was selected
determining the heat transfer coefficient due to wind speed: to perform thermal modelling. The landscape of this area is
a well-drained patchy polygonal tundra (including slope
h = 2.4v+4.3, if v<4.6 m/s or [6] edges), mainly with herbaceous-moss-lichen vegetation.
h = 3.7(v-1), if v>4.6 m/s [7] The profile along the borehole is an interbedding of fine
and silty non-saline sands to a depth of 13.5 meters. Then,
where v is wind speed, m/s. to a depth of 21.5 m, there are light loams of medium
Based on the above, the surface temperature can be salinity. Heavy, moderately saline loams were found from
used considering the so-called radiative correction (Eq. 4 a depth of 21.5 m to 25 m. Table 1 shows the values of the
and 5), and the heat transfer coefficient due to wind can be physical characteristics of soils. Wtot – in situ volumetric
used based on Eq. 6 or 7. water content; λf and λth - thermal conductivity in frozen and
Since there are usually natural covers in the form of unfrozen state, W/(m°C); Cf and Cth heat capacity in frozen
snow or vegetation on the soil surface, the heat transfer and unfrozen state, kJ/(m3°C); and freezing point, °C.
coefficient between the soil surface and air can be written The nearest meteorological station is located 120 km
in the following form: from the borehole and located near Seyaha. Climatic
characteristics are given in Table 2.
htotal = 1/(1/h+Rsnow+Rvegetation) [8]
Table 1. Physical properties of soils active layer thickness differs from 0.9 to 1.1m, the
temperature at the depth of zero annual amplitude is -5.2°C
Soils Wtot λf λth Cf Cth FP and the depth of zero annual amplitude is 16.0m.
Silty sand 0.271 2.48 1.91 1818 2614 -0.12

Depth, m
Fine sand 0.231 2.56 2.09 1854 2590 -0.15 Temperature, °C
Silt 0.161 1.44 1.43 2509 2905 -1.07 -25 -20 -15 -10 -5 0 5
Silty clay 0.152 1.45 1.44 2541 2686 -2.04 0

5
25.08.18
01.01.19
10
01.04.19
01.01.20
01.04.20 15
01.10.21
01.01.22 20
01.04.22
01.10.22
25
Figure 2. Initial temperature curve

6 THERMAL MODELLING

Figure 1. Space image of the borehole location A transient analysis in TEMP/W was used to perform the
initial simulation. Measured ground temperatures on 01
Table 2. Average monthly climatic characteristics of the October 2021 were implemented as the initial ground
Seyakha meteorological station temperature condition for the original scenario.
A Convective Surface boundary condition was used to
Month Temperature Snow depth Snow Wind speed
estimate the ground’s thermal response to simulate the
(°C) (m) density (m/s) complex soil-climate interaction. The Convective Surface
(t/m3) boundary condition applies Newton’s Law of Cooling (Eq.
2) to calculate a heat flux. The temperature was set
Jan -23.8 0.31 0.27 6.0
considering the radiative correction according to Equations
Feb -23.7 0.33 0.29 5.6 4 and 5.
Mar -19 0.33 0.31 5.8 As a result of the calibration modelling, such heat
Apr -13.9 0.33 0.32 5.7 transfer coefficient values were selected at which the
May -5.3 0.35 0.23 6.0 temperature values at the depth of zero annual amplitudes
Jun 3.1 - - 5.3 and the thickness of the active layer corresponded to those
measured during surveys.
Jul 8.9 - - 4.4
The selected values of the temperatures and heat
Aug 8.9 - - 5.2 transfer coefficient, given in Table 3, gave satisfactory
Sep 4.2 - - 5.9 results relative to those measured in the field. The
Oct -4.3 0.13 0.03 6.2 simulated depth of the active layer and temperature at the
Nov -15.1 0.22 0.15 6.1 depth of zero annual amplitudes were found to be in close
agreement with field measurements, with errors not
Dec -19.8 0.28 0.21 6.3
exceeding 6%.
Jan -23.8 0.31 0.27 6.0 Figure 3 shows the temperature curves of the simulated
and field data for April and October. The deviation in the
simulated data under 1.0m in April may occur because the
Temperature measurements were regularly taken in model uses monthly mean snow depths. For more accurate
the borehole, which is shown in Figure 2. According to the modelling, it is necessary to use values that change over
measurements, the depth of zero annual amplitudes, the 10 days. At the same time, the difference in the upper 1.0
temperature at the depth of zero annual amplitudes and m is because under natural conditions the atmospheric
also the thickness of the active layer were determined. The temperature can vary greatly during the day, and the
sensor located close to the surface is very sensitive to this 35%, while the wind speed can fluctuate by 50% in both
change. However, the fit obtained here is considered sides. However, for a more detailed assessment of the
sufficient for the purposes of this paper. effect of snow cover over time, an intermediate value of
10% in snow depth was taken. Table 6 shows the heat
Table 3. Values used for the surface boundary condition in transfer coefficient values used for sensitivity analysis.
the original simulation scenario Additionally, the values of the thermal conductivity of
soils were changed by the amount of possible error in their
Month Temperature Heat determination. According to available data, the value of
(°C) transfer such an error can be ±10%. Table 4 lists the thermal
coefficient conductivity values used for sensitive analysis.
(W/(m2°С))
Jan -23.8 1.120 7.1 Influence of thermal conductivity
Feb -23.7 1.101
According to Figure 4, an error in determining the
Mar -16.3 1.039
thermal conductivity gives a slight deviation of the
Apr -10.5 1.008 temperature curve from the original scenario. The
May -1.5 1.431 temperature at a depth of zero annual amplitudes differs by
Jun 7.5 3.097 0.1°C, comparable to the accuracy of temperature
Jul 14.0 2.946 determination in the borehole. The deviation of the
Aug 12.5 3.083
temperature curves from the original scenario in the range
of 2–15 m does not exceed 0.3°C. In a specific model, no
Sep 6.7 3.173
influence on the depth of the active layer can be traced.
Oct -2.1 3.581
Nov -15.1 1.428 Table 4. Thermal conductivity values (W/(m°C)) used for
Dec -19.8 1.294 sensitivity analysis
Jan -23.8 1.120
Soils +10% λf +10% λth -10% λf -10% λth
Silty sand 2.73 2.10 2.23 1.72
Depth, m

Temperature, °C
Fine sand 2.81 2.30 2.30 1.88
-16.0 -12.0 -8.0 -4.0 0.0 4.0
Silt 1.59 1.57 1.30 1.28
0
Silty clay 1.59 1.58 1.30 1.29
2
Depth, m

Temperature,°C
4
-6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0
6 0

8 2

10 4
01.04.20
12 6
01.10.21 8
14
Simulated 01.04 10
16
Simulated 01.10 12
18
Figure 3. Comparison of field data with simulated Original scenario
14
temperatures
+10% Thermal conductivity
16
7 SENSITIVITY ANALYSIS -10% Thermal conductivity
18
The main objective of this study is to compare the Figure 4. Influence of thermal conductivity on the
influence of the initial boundary conditions on the model. temperature curve
To fulfill the set goal, components of surface conditions
were changed, such as the thickness of the snow cover By correcting the surface conditions in the winter period
and wind speed. According to data analysis from the by about 1.5 - 2.5%, it is possible to achieve a decrease in
Seyakha weather station over the past 20 years, the the influence due to thermal conductivity on the
deviation of snow cover from the average value can reach temperature curve and thus minimize the effect of error in
determining the thermal conductivity. It should be noted 7.2 Influence of boundary conditions
that these values may differ depending on the geotechnical
profile. According to Figure 6, an error in determining the depth
The values of heat transfer on the surface used in the of snow gives a significant deviation of the temperature
calibration are given in Table 6, and the results of the curve from the original scenario, even with a deviation of
calibration of the models are shown in Figure 5. In this 10%. The temperature at a depth of zero annual
case, the deviation of the temperature curves from the amplitudes differs by 0.3°C for a 10% snow deviation and
original scenario in the range of 2 - 15 m does not exceed by 1.2°C for a 35% snow deviation. In a specific model, no
0.1°C. influence on the depth of the active layer can be traced.
Nevertheless, according to Figure 7, an error in
Table 5. Values of convective heat transfer coefficient determining the wind speed gives insignificant deviation.
(W/(m2°C)) used for calibration of models with different
thermal conductivity profiles Table 6. Values of convective heat transfer coefficient
(W/(m2°C)) used for sensitivity analysis
Month +10% λ -10% λ
Jan 1.137 1.092 Month +10% -10% +35% -35% +50% -50%
Snow Snow Snow Snow Wind Wind
Feb 1.118 1.074 depth depth depth depth speed speed
Mar 1.055 1.013 Jan 1.231 1.027 1.640 0.850 1.146 1.080
Apr 1.024 0.983 Feb 1.210 1.010 1.609 0.837 1.129 1.064
May 1.452 1.395 Mar 1.143 0.953 1.525 0.788 1.063 1.005
Jun 3.097 3.097 Apr 1.110 0.924 1.481 0.764 1.031 0.977
Jul 2.946 2.946 May 1.569 1.315 2.067 1.094 1.474 1.367
Jun 3.097 3.097 3.097 3.097 3.346 2.826
Aug 3.083 3.083
Jul 2.946 2.946 2.946 2.946 3.244 2.744
Sep 3.173 3.173
Aug 3.083 3.083 3.083 3.083 3.336 2.818
Oct 3.635 3.492
Sep 3.173 3.173 3.173 3.173 3.396 2.875
Nov 1.449 1.392 Oct 3.853 3.345 4.757 2.871 3.849 3.201
Dec 1.313 1.261 Nov 1.566 1.312 2.064 1.091 1.469 1.363
Jan 1.137 1.092 Dec 1.421 1.187 1.884 0.985 1.326 1.240
Jan 1.231 1.027 1.640 0.850 1.146 1.080
Temperature, °C
Depth, m

-6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 Temperature, °C

Depth, m

0
-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0
2 0

4 2

6 4

6
8
8
10
10
12
Original scenario 12
14 Original scenario
+10% Thermal conductivity 14 +10% Snow depth
16
-10% Snow depth
-10% Thermal conductivity 16
18 +35% Snow depth
Figure 5. Calibration of temperature curves for different 18 -35% Snow depth
thermal conductivity profiles Figure 6. Influence of snow cover on the temperature curve
 Temperature, °C
Depth, m Temperature, °C Time, Years
-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 0 20 40 60 80 100
0 -4.0

2
-4.5
4
-5.0
6

8 -5.5
10
-6.0
12

14 -6.5
Original scenario
16 +50% Wind speed
18 -50% Wind speed
Figure 7. Influence of wind speed on the temperature curve

Figure 8. Comparison of the influences of different factors

8 RESULTS AND DISCUSSION
on the temperature on the depth of zero annual amplitude
over time
This study uses model verification methods developed at
the Moscow State University at the Department of
Geocryology. According to the results obtained, the
9 CONCLUSION AND NEXT STEPS
method used has good accuracy in determining the initial
boundary conditions on the surface.
This study summarizes the results of thermal behaviour
The results of the sensitivity analysis can be divided
under the change in different factors, such as depth of
into 5 points:
snow cover, wind speed and thermal conductivity. Based
• Errors in determining the thermal conductivity do on the results, error in determining the boundary conditions
not affect the ground thermal regime process significantly. has a much more noticeable effect in thermal modelling.
• A slight deviation caused by an error in Under the influence of various types of development,
determining the thermal resistance can be reduced by many components of the natural environment can change.
calibrating the surface boundary conditions. This study shows that it is essential to correctly determine
• Snow coverage significantly impacts the thermal these changes due to anthropogenic impacts for accurate
regime of permafrost. An error in determining snow depth thermal modelling.
can lead to poorly constrained essential components in There is a need to improve the quality and accuracy of
geotechnical design. the geocryological forecast in order to understand and
• The wind component appears to have a negligible mitigate anthropogenic impacts on permafrost, leading to
impact on the temperature regime of soils in natural more sustainable and economic northern development.
conditions. Moreover, in natural conditions, the The long-term goal is to develop a formalized approach to
temperature regime of soils is mainly determined by the accurately and quantitatively determine anthropogenic
properties of the surface covers. influences in geocryological studies.
• As demonstrated in Figure 8, even a relatively
small error of 10% can have a significant impact on the
early stages of building and structure operation. 10 ACKNOWLEDGEMENTS
Furthermore, an error of 35% in determining the height of
snow cover can lead to a substantial influence on the We would like to express our gratitude to Sergey Pravov
temperature regime, subsequently affecting the bearing and Alexandra Popova for generously sharing the data
capacity of frozen soils. In contrast, the other components used in this study. We would also like to thank Teddi
investigated in this study were found to have a negligible Herring for insightful comments and suggestions, which
impact on geotechnical projects, both in the short and long greatly improved the quality of the paper. Financial support
term. for this research was provided by the Natural Sciences and
Engineering Research Council of Canada (NSERC).
11 REFERENCES O’Neill, H., & Burn, C. R. 2017. Impacts of variations in
snow cover on permafrost stability, including simulated
Aleksyutina, D. & Motenko, R. 2011. Thermal properties for snow management, Dempster Highway, Peel Plateau,
vegetation cover for AGU Fall Meeting ’11, San Northwest Territories. Arctic Science, 3(2), 150–178
Fransico, CA, USA. 0408-. Osokin, N.I., Sosnovskiy, A.V. & Chernov, R.A. 2017.
Balland, V., & Arp, P. A. 2005. Modeling soil thermal Effective thermal conductivity of snow and its
conductivities over a wide range of conditions. Journal variations. Earth's Cryosphere. 21. 60-68.
of Environmental Engineering and Science, 4(6), 549– Park, H., Fedorov, A. N., Zheleznyak, M. N., Konstantinov,
558. P. Y., & Walsh, J. E. 2015. Effect of snow cover on pan-
Barry-Macaulay, D, Bouazza, A., Wang, B., & Singh, R. M. Arctic permafrost thermal regimes. Climate Dynamics,
2015. Evaluation of soil thermal conductivity models. 44(9-10), 2873–2895.
Canadian Geotechnical Journal, 52(11), 1892–1900. Stieglitz, M., Dery, S. J., Romanovsky, V. E., & Osterkamp,
Côté, J. & Konrad, J.-M. 2005. A generalized thermal T. E. 2003. The role of snow cover in the warming of
conductivity model for soils and construction materials. arctic permafrost. Geophysical Research Letters,
Canadian Geotechnical Journal, 42(2), 443–458. 30(13), 1721–n/a.
Decagon Devices, Inc. 2016. KD2 Pro Thermal Properties Sturm, M., Holmgren, J., & Liston, G. E. 1995. A Seasonal
Analyzer Operator’s Manual. Snow Cover Classification System for Local to Global
Decharme, B., Brun, E., Boone, A., Delire, C., Le Moigne, Applications. Journal of Climate, 8(5), 1261–1283.
P., & Morin, S. 2016. Impacts of snow and organic soils Sturm, M., McFadden, J. P., Liston, G. E., Chapin, F. S.,
parameterization on northern Eurasian soil temperature Racine, C. H., & Holmgren, J. 2001. Snow–Shrub
profiles simulated by the ISBA land surface model. The Interactions in Arctic Tundra: A Hypothesis with
Cryosphere, 10(2), 853–877. Climatic Implications. Journal of Climate, 14(3), 336–
Draebing, D., Haberkorn, A., Krautblatter, M., Kenner, R., 344.
& Phillips, M. 2017. Thermal and mechanical Westermann, S., Schuler, T. V., Gisnås, K., & Etzelmüller,
responses resulting from spatial and temporal snow B. 2013. Transient thermal modelling of permafrost
cover variability in permafrost rock slopes, steintaelli, conditions in Southern Norway. The Cryosphere, 7(2),
swiss alps: thermal and mechanical responses to snow 719–739.
in permafrost rock slopes. Permafrost and Periglacial Zhang, T. 2005. Influence of the seasonal snow cover on
Processes, 28(1), 140–157. the ground thermal regime: An overview: snow cover
Ershov, E.D. et al. 1999. Fundamentals of geocryology. and ground thermal regime. Reviews of Geophysics
Volume 5. Permafrost Engineering. Moscow University (1985), 43(4).
Press, Moscow, Russia (In Russ.).
Ershov, E.D. et al. 2001. Fundamentals of geocryology.
Volume 4. Dynamic geocryology. Moscow University
Press, Moscow, Russia (In Russ.)
Frost, G. V., Epstein, H. E., Walker, D. A., Matyshak, G., &
Ermokhina, K. 2018. Seasonal and Long-Term
Changes to Active-Layer Temperatures after Tall
Shrubland Expansion and Succession in Arctic Tundra.
Ecosystems (New York), 21(3), 507–520.Gascoin, S.,
Lhermitte, S., Kinnard, C., Bortels, K., & Liston, G. E.
2013. Wind effects on snow cover in Pascua-Lama, Dry
Andes of Chile. Advances in Water Resources, 55, 25–
39.
Gavriliev, I. 2004. Thermophysical properties of natural
environment components in permafrost. Novosibirsk:
Russian Academy of Science. 146 p. (In Russ.).
Gorelik, J.B.; Pazderin, D.S. 2017. Correctness of
formulation and solution of thermotechnical problems in
forecasting temperature field dynamics in the
foundations of constructions on permafrost. Kriosfera
Zemli, 21, 49–59. (In Russ.)
Grünberg, I., Wilcox, E. J., Zwieback, S., Marsh, P., &
Boike, J. 2020. Linking tundra vegetation, snow, soil
temperature, and permafrost. Biogeosciences, 17(16),
4261–4279.
Lu, S., Ren, T., Gong, Y., and Horton, R. 2007. An
improved model for predicting soil thermal conductivity
from water content at room temperature. Soil Science
Society of America Journal, 71(1): 8–14.
 Preliminary results of thermo-hydro-
mechanical modelling of large-strain
deformation of ground in Canadian cold
regions subject to climate change
Anna Pekinasova, Jocelyn L. Hayley
Department of Civil Engineering, University of Calgary
Brandon Karchewski
Department of Geoscience, University of Calgary,
Calgary, Alberta, Canada

ABSTRACT
Understanding the dynamic nature of frozen ground and thawing permafrost is vital for engineering adaptation of
transportation infrastructure in cold regions to climate change. A large-strain thaw consolidation model is more appropriate
than a small-strain model for ice-rich permafrost terrain, fine-grained soils, and frozen ground with excess ice. To forecast
the impacts of climate change, it is important that models account for the complex and dynamic heat energy boundary
conditions which may be influenced by short-wave and long-wave solar radiation, humidity, precipitation, and wind speed
in addition to air temperature. This study builds on recent modelling work on nonlinear thermo-hydro-mechanical coupled
large-strain consolidation (Dumais and Konrad 2018, Yu et al. 2020a, 2020b) and presents preliminary results towards
developing an open-source tool for forecasting the impacts of climate change on infrastructure in Canadian cold regions.

RÉSUMÉ
Comprendre la nature dynamique du sol gelé et du dégel du pergélisol est essentiel pour l'ingénierie de l'adaptation des
infrastructures de transport dans les régions froides au changement climatique. Un modèle de consolidation de dégel à
grande déformation est plus approprié qu'un modèle à petite déformation pour les terrains de pergélisol riches en glace,
les sols à grains fins et les sols gelés avec un excès de glace. Pour prévoir les impacts du changement climatique, il est
important que les modèles tiennent compte des conditions aux limites complexes et dynamiques de l'énergie thermique
qui peuvent être influencées par le rayonnement solaire à ondes courtes et à ondes longues, l'humidité, les précipitations
et la vitesse du vent en plus de la température de l'air. . Cette étude s'appuie sur des travaux de modélisation récents sur
la consolidation non linéaire couplée thermo-hydro-mécanique à grandes déformations (Dumais et Konrad 2018, Yu et al.
2020a, 2020b) et présente des résultats préliminaires en vue de développer un outil open source pour prévoir les impacts
du changement climatique. changements sur les infrastructures dans les régions froides du Canada.

1 INTRODUCTION There is therefore a need for practical and openly available

tools to evaluate and analyze existing and future
Understanding the settlement and heaving of ground infrastructure resilience to climate change. Such tools will aid
experiencing freeze-thaw cycling is a vital component of in the identification of potential risks and budget allocation for
understanding and analyzing infrastructure built in northern operations and maintenance to reduce and offset the
and permafrost regions subject to climate change. Due to the negative impacts of climate change.
increasing impacts of climate change in cold regions, there is Climate change forcing in northern regions (e.g.
an urgent need to advance the understanding of freezing and increasing annual mean temperature, increasing
thawing ground behaviour in order to develop resilient precipitation, increasing relative humidity, surface ponding,
infrastructure and promote sustainable socioeconomic streamflow, soil moisture, and/or droughts) causes regional
climate adaptation in these important Canadian communities. and local changes by affecting the ground temperature
Heaving and thawing of the ground are two major causes distribution, the thickness of the active layer, and the depth to
of infrastructure destruction in cold regions (Yu et al. 2020a). stable ground temperature (cf. Smith and Riseborough 1996,
The freezing process can trigger migration of water from Riseborough et al. 2008, An et al. 2017, Harris et al. 2017).
unfrozen to frozen zones (cryosuction) and formation of ice- There have been significant advances in formulation and
lenses (Yu et al. 2020a). When ice-rich material thaws, the numerical analysis of freeze-thaw processes in cold regions
ground loses strength due to high pore water pressure, and permafrost terrain and in modelling of surface energy
develops substantial thaw settlement, leads to instability, and boundary conditions, but these have not been fully integrated
increased runoff and flooding (cf. Hjort et al. 2018). These into a practical engineering analysis tool. Our research
effects result in increases in repairs, cost of maintenance, focusses on the impact of climate change on how the ground
and unreliable operations, all of which have undesirable moves – thaw consolidation and frost heave – using a novel
impacts on the socioeconomic aspects of northern combination of coupled equations to accurately quantify the
communities and the environment (cf. Hjort et al. 2018). freeze-thaw processes with state-of-the-art models for the
surface energy balance. The goal of our work is to develop a A useful application of the Lagrangian coordinate
practical open-source thermo-hydro-mechanical (THM) transformation is for the calculation of total settlement of a
modelling tool to bridge the gap between disciplines of layer experiencing simultaneous consolidation and heave
geotechnical engineering, hydrogeology, and climate processes in terms of the void ratio distribution and the
science. This modelling tool will assist in the design and Lagrangian coordinate system
analysis of existing and future infrastructure for engineers,
geoscientists, climate scientists, owners, and policymakers,
and other stakeholders. [2]

where H [m] is the original thickness of the layer, Z = z0 is the

2 MATHEMATICAL FORMULATION
coordinate of the original ground surface, and Z = z0 + H is a
point below which negligible deformation occurs (e.g. below
The theory of ground freeze-thaw and thaw consolidation
the bottom of a permafrost layer). We can see that this
involves thermo-hydro-mechanical coupled processes
calculation is simply the original layer thickness minus the
accounting for heat transfer, pore fluid migration, and
volume change ratio integrated over the entire layer. Note
deformation resulting from a nonlinear void ratio-
that points with (1+e)/(1+e0) > 1 contribute to heave and
permeability-effective stress relationship. It is important to
points with (1+e)/(1+e0) < 1 contribute to a settlement which
recognize that in a layer experiencing seasonal variations in
may both be occurring simultaneously at different depths
surface boundary conditions, the ground temperature
within the soil column and a positive (negative) value for s(t)
distribution may oscillate with depth, and different parts of the
would represent a net settlement (heave) of the layer.
surface layer may experience consolidation or heave
processes simultaneously. This is more complex than a
2.2 Thermal equations
monotonic freezing or thawing process for which analytical
solutions are available (Stefan 1889, Sumgin et al. 1940,
We write the heat energy balance in the unfrozen or frozen
Carlson 1952, Carslaw and Jaeger 1959, Kudryavtsev et al.
soil by accounting for conduction, advection, sensible heat
1977, Lunardini 1978, Wright et al. 2003, Andersland and
storage due to temperature change, and latent heat storage
Ladanyi 2004), since there are interannual variations in the
due to phase change as follows (Dumais and Konrad 2018)
thermal regime of the subsurface (Romanovsky and
Osterkamp 1995, Riseborough et al. 2008), and necessitates
numerical solution of a coupled system of partial differential
equations. We follow the formulations presented by Dumais
and Konrad (2018) and Yu et al. (2020a, 2020b), which we [3]
summarize here briefly for context. For more detail, we refer
the reader to the original references and prior work.
where λ [J·s-1·K-1] is the bulk thermal conductivity, T [°C or K]
is the temperature, qw [m·s-1] is the pore water flux rate per
2.1 Lagrangian coordinate transformation
total unit area, Cw [J·K-1·m-3] is the volumetric heat capacity
of water, C [J·K-1·m-3] is the bulk volumetric heat capacity, L
To capture the coordinate change due to thaw consolidation
[J·kg-1] is the specific latent heat of fusion of water, ρi [kg·m-
and heaving we can use a coordinate transformation 3] is the density of ice, S [-] is the degree of saturation of
w
technique that captures the moving boundary conditions
unfrozen pore water, t [s] is time, and a superposed dot
which can be mapped onto non-moving Lagrangian
indicates a time derivative. Assuming fully saturated
coordinates to simplify the calculation of the THM formulation.
conditions (Sw + Si = 1), we compute the bulk thermal
In the case of large strain consolidation and heave, the
conductivity and bulk volumetric heat capacity using
change of coordinates is entirely described by the change in
geometric and arithmetic averages, respectively, as follows
volume between any two reference points. As such, we can
(Côté and Konrad 2005)
relate derivatives (and infinitesimals) with respect to the
global (moving) coordinate z [m] to the fixed (Lagrangian)
coordinate Z [m] via the following relationship (cf. Dumais and [4]
Konrad 2018)
[5]
[1]
where subscripts s, w, and i refer to solids, water, and ice,
where e = e(z, t) is the void ratio at some position z and time respectively. To compute ∂Sw/∂T, we follow Yu et al. (2020b)
t [s], and e0 = e(z, 0) is the initial void ratio at the same in using the Sw – T relation by Nishimura et al. (2009) as
reference position. There is a 1-to-1 mapping between any follows
moving reference point z and a corresponding reference point
Z in the Lagrangian coordinates. For brevity, we present the
remaining equations in this document with respect to the [6]
Lagrangian coordinate system, though the physical
relationships originate in the global (moving) coordinate
system.
where α and β are empirical values that depend on the state with no overburden pressure, σ1 [MPa] is the local
material and temperature T must be in [°C] here. Eq. [6] only stress including overburden pressure and additional pressure
applies for T ≤ Tf where Tf = 0°C = 273.15 K is the freezing associated with the void ratio, and b1 [-], b2 [°C-1], and b3
temperature of water and we assume that Sw = 1 and [MPa-1] are soil constants that should be determined
∂Sw/∂T = 0 for T > Tf. experimentally (Yu et al., 2020a). In Eq. [9], the ± sign is
positive when Ṫ < 0 and negative when Ṫ > 0.
2.3 Hydraulic equations
2.4 Effective stress hydromechanical equations
The pore fluid flux qw behaves differently depending on
whether the soil is unfrozen or frozen. For the unfrozen soil, we follow Yu et al. (2020b) in explicitly
In the unfrozen soil (T > Tf), we assume that fluid flux is coupling the hydraulic mass balance and the momentum
governed by the 1D large-strain consolidation process balance equations assuming all deformation is due to the
involving the transfer of excess pore pressure to effective consolidation process to obtain the following balance
stress, since at hydrostatic conditions the flux rate will be equation in terms of effective stress σ′ and void ratio e
zero. We use Darcy’s law to relate the pore fluid flux qw to the
gradient of effective stress σ′ as follows
[10]

[7]
where all variables are as introduced previously. The
effective stress – void ratio relationship used to calculate the
where k [m·s-1] is the permeability (hydraulic conductivity) of stress-strain coefficient ∂σ′/∂e is the well known
the unfrozen soil, γw [N·m-3] is the unit weight of water, Gs [-] semilogarithmic curve as follows (Yu et al. 2020b)
is the specific gravity of the solids, and we evaluate ∂σ′/∂Z
using chain rule as (∂σ′/∂e)(∂e/∂Z) where ∂σ′/∂e is the slope
of the effective stress – void ratio curve, which we discuss [11]
further in Section 2.4. The bracketed term in Eq. [4] is the
difference between the effective stress gradient (∂σ′/∂Z)′ at
hydrostatic conditions (i.e. buoyant unit weight γ′) and the where Ccu and Cru are the compression and rebound indices
current effective stress gradient ∂σ′/∂Z. This difference goes of unfrozen soil, pc′ is the preconsolidation pressure, and the
to zero as excess pore pressure dissipates approaching points {σ′cu0, ecu0} and {σ′ru0, eru0} are reference locations for
hydrostatic conditions. We also note that permeability the normal consolidation line (NCL) and the unloading-
depends on void ratio, and we follow Yu et al. (2020b) in reloading line (URL), respectively. Yu et al. (2020b) explain
writing in more detail about upper and lower limits of validity of this
model with respect to grain separation (above some void ratio
[8] esep) and particle breakage (below some void ratio emin), but
we assume for the analysis herein that emin < e < esep.

where Cku [-] is a permeability change index for unfrozen soil, 2.5 Total stress mechanical equations
{k0, eu0} are a reference permeability and corresponding void
ratio, m ≥ 1.0 is a multiplier accounting for a permeability For the frozen soil, there is deformation due to fluid migration,
increase in thawed soil, and elim, emin, and etr are reference phase change, and total stress variation, which introduces
void ratios, which we do not explain in detail here. We present some complication owing to the stress-strain relationships, as
Eq. [8] here mainly to emphasize that we obtain the gradient suggested by Konrad and Samson (2000) and Yu et al.
∂k/∂Z in the advection term of Eq. [3] via (∂k/∂e)(∂e/∂Z) similar (2020b). We can write the volume change due to fluid
to the effective stress gradient in Eq. [4]. migration in the frozen soil as
For frozen soil (T ≤ Tf) we use a water flux function, which
has a detailed history of investigation, although a complete [12]
review of the literature is outside of the scope of this paper.
The fluid flux in frozen soil depends on several factors
including temperature gradient, segregation potential, soil where we use Eq. [9] to compute the water flux qw. There is
type, void ratio, and stress state (cf. Konrad and Morgenstern also volume change due to phase change between water and
1982, Michalowski 1993, Michalowski and Zhu 2006, Yu et ice which follows
al. 2020a). For the model herein, we follow Yu et al. (2020a)
and write [13]

where Gi [-] is the specific gravity of ice and the value (1 – Gi)
[9] ≈ 9% represents the expansion (contraction) of pore volume
during a freezing (thawing) process. The deformation due to
stress variation follows the momentum balance in terms of
where Ṫref [°C⋅s-1] is a small positive reference temperature
total stress σ
rate, SP0 [m2⋅°C-1⋅s-1] is the segregation potential of the soil
corresponding to void ratio e0 on the frozen front at steady
 At the surface boundary, we follow Yu et al. (2020b) in
[14]
applying prescribed temperature boundary conditions with a
harmonic seasonal variation, which we explain in more detail
where the second term represents the total unit weight of a in Section 4.1. To accurately model the impact of complex
partially frozen soil in which the pores are fully saturated by climate forcing, in future work we will incorporate a surface
ice and water. We note that Yu et al. (2020b) present a energy balance (SEB) boundary condition (cf. Foken 2008,
simplified form of this relationship An et al. 2017, van Huissteden 2020, Westermann et al.
2023) incorporating incoming and reflected short-wave
radiation, incoming and outgoing thermal radiation of Earth's
[15] surface, sensible heat flux exchange by convection of air,
latent heat flux by evaporation or condensation of water, and
where the second term implies an assumption that the initial ground heat flux by migration of heat into and out of the
reference state with void ratio e0 corresponds to unfrozen ground. Each term involves specific formulae and parameters
conditions. We cannot make that assumption for cases where that characterize the energy balance at the surface.
we examine permafrost ground that is initially frozen at some Westermann et al. (2023) provide a detailed overview of
locations. We write a semilogarithmic total stress – void ratio these components within a framework of process-based tile
relationship for the frozen soil as modelling useful for medium to large scale regional climate-
ground interaction modelling. We refer the reader to the
original work and references therein for more detail. As with
[16] the lower boundary, we assume full drainage at the upper
boundary so we apply a prescribed void ratio similar to
where the point {ef0, σf0} is a reference point and Cf is a Eq. [20], but assume that we can calculate the void ratio
based on the applied stress conditions using Eq. [11] or [16]
temperature-dependent compression index (Li et al. 2014, Yu
depending on whether the surface soil is frozen or unfrozen.
et al. 2020b)

[17] 3 NUMERICAL IMPLEMENTATION

where a1, a2, and a3 are empirical material parameters. In this section we present a brief overview of the weak form
equations for solving the system of thermo-hydro-
2.6 Boundary conditions mechanically coupled equations using the finite element
(Galerkin) method, modified after Yu et al. (2020b), and
For the boundary condition at large depth Z = z = z0 + H, we outline the implicit time stepping scheme with iterative
assume a fixed geothermal gradient G = dT/dz [°C·m-1] so we correction. To invite transdisciplinary collaboration building
write the ground heat flux rate qg as bridges between the disciplines of geotechnical engineering,
hydrogeology, and climate science we implement the
[18] numerical code as an open-source Python package frozen-
ground-fem available at:
https://github.com/annapekinasova/frozen-ground-fem/
where the thermal conductivity λ is evaluated at the
In the code repository we provide examples of usage, and we
integration point nearest to the boundary where the void ratio
encourage those interested in collaboration to reach out for
e and degree of saturation with water Sw is approximately
collaborator access and/or to fork the repository to modify
constant in time. Note that at the lower boundary we assume
implementation details to their own needs.
the contribution of advection to the ground heat flux to be
negligible and only account for conduction in the geothermal
3.1 Thermal update equations
temperature gradient boundary condition. We assume full
drainage at the lower boundary so we apply a boundary
We form the integrated residual heat balance equation from
condition for water flux qw as
Eq. [3] and apply integration by parts to obtain
[19]

where q̂w is a prescribed flux that we compute using Eq. [7]

or Eq. [9] at the integration point nearest to the boundary if T
> Tf or T ≤ Tf, respectively, at that point. This is equivalent to
a fully drained boundary condition since we allow the [21]
prescribed fluid flux to be driven by other processes. We
assume negligible deformation below the lower boundary so
where δT represents residuals or errors in the approximate
[20] solution for T, a superposed dot represents a time derivative,
and a subscript comma represents differentiation with respect
to the adjacent variable. Subdividing evaluation of Eq. [21]
where ê is a prescribed void ratio fixing the void ratio at the
over a discretized system of finite elements, interpolating
lower boundary to the initial void ratio.
solution variables and residuals with the same weighting
functions (i.e. Galerkin method), and summing element making use of Eq. [8] to calculate ∂k/∂e and k. Discretization
vectors and matrices respecting connectivity of nodes and integration of Eq. [25] leads to the global residual
between elements leads to the following discretized global equation
residual equation
[27]
[22]
where e is the global vector of nodal void ratios, Q is the
where T is the global vector of nodal temperatures, Φ is the global fluid flux vector (from the boundary term and first
global heat flux vector (from the boundary term in Eq. [21]), integral in Eq. [25]), K is the global stiffness matrix (from the
H is the global conductive-advective heat flow matrix (from second integral in Eq. [25]), and M is the global mass matrix
the first integral in Eq. [21]), C is the global heat storage (from the third integral in Eq. [25]). Again, for arbitrary
matrix (from the second integral in Eq. [21]), and superscript residuals δe, the expression in parentheses in Eq. [27] must
T is the transpose operator. Since the residuals δT are be zero, which leads to the following update equation
arbitrary, the expression in parentheses in Eq. [22] must be
zero, and applying an implicit time stepping approach leads
to the following update equation

[28]

[23] where Δej is the void ratio correction vector at iteration j, ΨHMj
is a residual hydromechanical vector, and we compute
weighted vectors and matrices as in Eq. [24].
where ΔTj is the temperature correction vector at iteration j,
ΨUj is a residual heat flux vector that goes to zero at 3.3 Implicit time stepping with iterative correction
convergence of a time step, and we compute any weighted
vector or matrix Aαj as We use an implicit time stepping scheme with iterative
correction of material parameters and global matrices at each
[24] time step. We improve on the iterative correction algorithm of
Yu et al. (2020b) by incorporating the update of material
Note that the global matrices in Eq. [23] depend on qw and e parameters, global coefficient matrices, and the void ratio
so we achieve coupling by updating them as part of the same change due to ice-water phase change within the iterative
iteration loop with the hydromechanical equations. correction scheme. This can be computationally expensive
for large models with thousands of elements, but we find that
3.2 Hydromechanical update equations for the 1D large-strain freeze-thaw problem at field scale, the
number of elements required is on the order of hundreds, so
We combine Eqs. [10], [12], and [13] forming the integrated the added expense does not substantially increase the
residual hydromechanical equation and apply integration by simulation time in absolute terms. The implicit time stepping
parts to obtain scheme with iterative correction at each time step proceeds
as follows:

0. Set initial conditions for t, T, e, e0, and boundary

conditions and initialize global vectors for solution
variables Tt and et.
1. Initialize λ (Eq. [4]); C (Eq. [5]); Sw and ∂Sw/∂T
(Eq. [6]); k and ∂k/∂e (Eq. [8]); qw (Eq. [7] or [9]); σ′,
∂σ′/∂e, and pc′ (Eq. [11]); and σ and ∂σ/∂e (Eq. [16])
based on Tt and et.
2. Integrate to initialize global vectors and matrices Φt,
Ht, and Ct (Eq. [21]); and Qt, Kt, and Mt (Eq. [25]).
[25] 3. Initialize iteration counter j = 0, update boundary
conditions at t + Δt, and integrate to initialize Φt+Δtj
and Qt+Δtj.
where δe represents residuals or errors in the approximate 4. Initialize Tt+Δtj = Tt, et+Δtj = et, Ht+Δtj = Ht, Ct+Δtj = Ct,
solution for e, H(T – Tf) is the Heaviside function that is 1 for Kt+Δtj = Kt, and Mt+Δtj = Mt.
T > Tf and 0 for T ≤ Tf, and we evaluate the following 5. Initialize weighted vectors and matrices Φαj, Qαj, Hαj,
derivative as Cαj, Kαj, and Mαj (Eq. [24]).
6. Compute Ψuj and solve for ΔTj (Eq. [23]), then
[26] update Tt+Δtj+1 = Tt+Δtj + ΔTj.
7. Compute ΨHMj and solve for Δej (Eq. [25]), then
update et+Δtj+1 = et+Δtj + Δej.
 8. Update λ, C, Sw, ∂Sw/∂T, k, ∂k/∂e, qw, σ′, ∂σ′/∂e, pc′, (a)
σ, and ∂σ/∂e based on Tt+Δtj+1 and et+Δtj+1, integrate
to update Φt+Δtj, Qt+Δtj, Ht+Δtj, Ct+Δtj, Kt+Δtj, and Mt+Δtj,
and update Hαj, Cαj, Kαj, and Mαj.
9. Check for convergence of relative error norms
||ΔTj|| / ||Tt+Δtj+1|| and ||Δej|| / ||et+Δtj+1||. If not
converged, set j = j + 1 and go to to Step 6.
10. Compute additional void ratio change Δet+Δt due to
total stress variation (Eq. [14]).
11. Set Φt = Φt+Δt, Qt = Qt+Δt, Ht = Ht+Δt, Ct = Ct+Δt,
Kt = Kt+Δt, Mt = Mt+Δt, and t = t + Δt.
12. Check if final time tf reached. If not, go to Step 3.

4 PRELIMINARY RESULTS

Current work on fully coupled THM modelling of thaw (b)

consolidation (Dumais and Konrad 2018) and freeze-thaw
heave and consolidation (Yu et al. 2020a,b) demonstrate
results and performance of models for relatively thin surface
layers <50 cm thick. In particular, Yu et al. (2020b)
demonstrate model performance for freeze-thaw on a layer
of 40 cm thickness, indicating the ability of the model
framework to capture pore water accumulation and talik
(unfrozen soil layer with overlying frozen soil, often with
confined excess pore pressure) formation. Both prior works
use proprietary software packages for solution of partial
differential equations using the finite element method such as
COMSOL (Dumais and Konrad 2018) or MATLAB (Yu et al.
2020b). The present authors implement the coupled THM
framework as an open-source Python package to promote
collaborative development and open access for researchers,
practitioners, and students. However, we highlight the
importance of scaling solutions to more realistic dimensions Figure 1. Stabilization of thermal profile under harmonic
for simulating permafrost thaw, active layer thickening, and surface temperature conditions. (a) Seasonal temperature
thermokarst and talik formation where the full problem profiles (dashed lines at 5-year intervals and thick solid line
domain may be ~80 m thick with active layer thickness up to converged results after 30 years of simulated time). (b)
~25 m (cf. Andersland and Ladanyi 2004). Inherently, there Converged temperature envelopes (thick lines) and
are issues with scaling the numerical solutions from ~0.5 m temperature profiles (dashed and dotted lines).
thick up to full field scale, and with our preliminary results we
highlight some of the challenges. We compare and contrast
the numerical solution parameters necessary for stability of Cs = 1.964×106 J/m3/°C, Cw = 4.204×106 J/m3/°C, Ci =
the thermal component in Section 4.1 and of the 1.881×106 J/m3/°C; densities ρs = 2650 kg/m3, ρw = 1000
hydromechanical coupled consolidation component in kg/m3, ρi = 910 kg/m3; and an initial void ratio (frozen) e0 =
Section 4.2. 0.3. We simulate 30 years of harmonic surface temperature
variation until the temperature profiles and envelopes
4.1 Permafrost, active layer, and temperature profiles under converge using a time step of Δt = 6.065×105 s ≈ 1 week,
harmonic annual temperature variation implicit time stepping factor α = 0.5 (Crank-Nicolson scheme),
and iterative correction error tolerance of 1.0×10 -3. Figure 1
We demonstrate “spin-up” or “burn-in” of ground temperature shows the results of the simulation with a depth to zero
profiles under harmonic seasonal temperature variation using annual amplitude of ~25 m and temperature gradient
the thermal equations and numerical solution scheme. We approaching the geothermal gradient below this depth with
simulate a 50 m thick soil column sufficient to capture the permafrost persisting below 50 m depth which is a typical
active layer, depth to zero annual amplitude of ground temperature profile in a permafrost region (cf. Andersland
temperature, and permafrost layer. We use 24 elements with and Ladanyi 2004). We emphasize here (for comparison with
2 nodes each, an upper boundary with a harmonic prescribed the consolidation simulation) that the result was stable and
temperature with annual mean of -10 °C and amplitude of accurate with elements ~2 m thick and a time step of
25 °C (Tmin = -35 °C in mid-Jan, Tmax = +15°C in mid-Jul), and ~1 week; the simulation took ~1 min on a system with an Intel
a lower boundary with a geothermal gradient G = dT/dz = Core i7-8700K CPU (3.70 GHz x 12 cores) and 32 GB of
0.03 °C/m = 30 °C/km. We assume thermal conductivity RAM. No special effort was made toward code optimization
values λs = 7.0 W/m/°C, λw = 0.563 W/m/°C, or parallelization apart from using vectorized operations via
λi = 2.22 W/m/°C; volumetric heat capacity values NumPy arrays in Python.
4.2 Large strain hydromechanical coupled simulation of a
consolidating layer with double drainage

To demonstrate a substantial challenge in scaling the fully

coupled THM framework to practical dimensions, we simulate
consolidation of a 20 cm thick layer with double drainage
under a finite total stress increment and self weight. We
assume volumetric, hydraulic, and effective stress material
properties the same as those provided in Table 1 of Yu et al.
(2020b), which represent a clayey soil of relatively low
hydraulic conductivity. We apply an initial void ratio profile as
shown in Figure 3 (e = 1.374~1.525), consistent with a small
surface load (1 kPa), effective self-weight, and the NCL,
which we allow to stabilize for 20 min of simulated time, after
which we observe negligible changes in the void ratio and
effective stress profiles and negligible settlement. After
stabilization, we apply a vertical stress increment of 10 kPa
(applied via fixed void ratio boundary conditions at top and
bottom) and simulate 1120 min of consolidation which
stabilizes at ~30 mm of settlement (~15% of original layer
thickness) as shown in Figure 2. Using 200 elements Figure 2. Simulation of consolidation settlement of a 20 cm
(element thickness 10 mm), we found that a numerical time thick layer of fine-grained material using the current large-
step of Δt = 120.0 s was necessary for a stable and accurate strain framework with semi-logarithmic k – e – σ′
solution, which is slightly larger than the time step of 50 s relationships (solid line) and small-strain Terzaghi
used by Yu et al. (2020b) for thaw consolidation a layer with consolidation based on averaged k and ∂σ′/∂e values
the same thickness. For comparison, in Figures 2 and 3 we (dashed line).
also show predictions of Terzaghi small-strain consolidation
based on the average initial permeability k and average
secant ∂σ′/∂e values. Contrary to statements in early large- and implementation on high-performance computing (HPC)
strain consolidation models (Davis and Raymond 1965, systems to bridge the gap between simulations of lab-scale
Gibson et al. 1967) the excess pore pressure dissipates and realistic spatial and temporal field-scale. Also the
slower in the current large-strain model compared to Terzaghi concept of residual stress is critical for properly capturing the
small-strain and the coefficient of consolidation behaviour of thawing ice-rich soils, and future work is needed
k(∂σ′/∂e)/[(1+e0)γw] is not constant. We can explain the slower to validate this model after incorporating the residual stress
rate of excess pore pressure dissipation by observing in component. The aim of this research is to provide a fully
Figure 3(c) that the initial decrease in void ratio near the validated THM code that captures both the thaw
boundary drainage points leads to a rapid decrease in consolidation and freeze-thaw processes in soils to predict
permeability k which governs the effective permeability of the ground response under complex climate conditions. We
entire layer through the consolidation process, even though demonstrated performance of an open-source Python
the permeability near the centre of the layer may be an order package under development for this purpose for thermal and
of magnitude higher. We emphasize that stable and accurate consolidation components separately. We encourage
results for the 20 cm layer (compare to 50 m layer in the interested readers to follow development via the GitHub link
thermal simulation) required elements of ~10 mm thickness provided and to connect regarding interest in collaboration.
(compare to ~2 m thick elements) and a time step of 120 s
(compare to ~1 week). The simulation took ~20 min (compare
to ~1 min) on the same system as in the previous example. 6 ACKNOWLEDGMENTS
In our opinion, the numerical efficiency of the large-strain
consolidation component poses the most substantial Financial support for this research was provided by the
challenge to scaling the THM framework to field-scale for Natural Sciences and Engineering Research Council of
simulation of climate impacts on permafrost layers and Canada (NSERC).
overlying infrastructure. Bridging this gap will be the focus of
our work in this area in the near future.
7 REFERENCES

5 CONCLUSIONS AND FUTURE WORK An, N. et al. (2017). Assessment of the methods for
determining net radiation at different time-scales of
Yu et al. (2020b) demonstrate model performance with meteorological variables. Journal of Rock Mechanics and
oscillating temperature boundary conditions over a simulated Geotechnical Engineering, 9(2), 239–246.
5 day period, which is a worthwhile proof-of-concept for Andersland, O.B. and Ladanyi, B. (2004). Frozen ground
modelling climate driven changes to permafrost. Based on engineering (2nd edition). John Wiley & Sons.
our experience with replicating similar results, substantial Carlson, H. (1952). Calculation of the depth of thaw in frozen
effort will be necessary on code optimization, parallelization, ground. In Frost Action in Soils (Vol. 2, pp. 192–223).
Figure 3. Simulation of consolidating layer using large-strain hydromechanical framework. (a) Void ratio profiles at initial (solid
black), intermediate (dotted lines, arrows indicating time), and final (dashed line) conditions. (b) Corresponding effective stress
profiles. (c) Corresponding permeability (hydraulic conductivity) profiles.

Carslaw, H.S. and Jaeger, J.C. (1959). Conduction of Heat in Michalowski, R.L. and Zhu, M. (2006). Frost heave modelling
Solids (2nd ed.). Oxford University Press. using porosity rate function. International Journal for
Côté, J. and Konrad, J.-M. (2005). A generalized thermal Numerical and Analytical Methods in Geomechanics,
conductivity model for soils and construction materials. 30(8), 703–722.
Canadian Geotechnical Journal, 42(2), 443–458. Nishimura, S. et al. (2009). THM-coupled finite element
Davis, E.H. and Raymond, G.P. (1965). A Non-Linear Theory analysis of frozen soil: formulation and application.
of Consolidation, Géotechnique, 15(2), 161-173. Géotechnique, 59(3), 159–171.
Dumais, S. and Konrad, J.-M. (2018). One-dimensional large- Riseborough, D. et al. (2008). Recent advances in permafrost
strain thaw consolidation using nonlinear effective stress– modelling. Permafrost and Periglacial Processes, 19(2),
void ratio – hydraulic conductivity relationships. Canadian 137–156.
Geotechnical Journal, 55(3), 414–426. Romanovsky, V.E. and Osterkamp, T.E. (1995). Interannual
Gibson, R.E. et al. (1967). The Theory of One-Dimensional variations of the thermal regime of the active layer and
Consolidation of Saturated Clays, Géotechnique, 17(3), near-surface permafrost in northern Alaska. Permafrost
261-273. and Periglacial Processes, 6(4), 313–335.
Harris, S.A., Brouchkov, A. and Guodong, C. (2017). Smith, M.W. and Riseborough, D.W. (1996). Permafrost
Geocryology. CRC Press. monitoring and detection of climate change. Permafrost
Hjort, J. et al. (2018). Degrading permafrost puts Arctic and Periglacial Processes, 7(4), 301–309.
infrastructure at risk by mid-century. Nature Communications, Stefan J. (1889). Über einige Probleme der Theorie der
9(1), 5147. Wärmeleitung. Sitzungsber. Akad. Wiss. Berlin Math. Kl.,
Konrad, J.-M. and Morgenstern, N.R. (1982). Prediction of 98 pp. 473–484.
frost heave in the laboratory during transient freezing. Sumgin, M.I. et al. (1940). General frost studies. Izd. Ak.
Canadian Geotechnical Journal, 19(3), 250–259. Nauk SSSR.
Konrad, J.-M. and Samson, M. (2000). Hydraulic conductivity Westermann, S. et al. (2023). The CryoGrid community
of kaolinite-silt mixtures subjected to closed-system model (version 1.0) – a multi-physics toolbox for climate-
freezing and thaw consolidation. Canadian Geotechnical driven simulations in the terrestrial cryosphere.
Journal, 37(4), 857–869. Geoscientific Model Development, 16(9), 2607–2647.
Kudryavtsev, V.A. et al. (1977). Fundamentals of frost forecasting Wright, J.F., Duchesne, C. and Côté, M.M. (2003). Regional-
in geological engineering investigations, Cold Regions scale permafrost mapping using the TTOP ground
Research and Engineering Lab. temperature model. 8th International Conference on
Li, S., Lai, Y., Pei, W., Zhang, S. and Zhong, H. (2014). Permafrost, 1241–1246.
Moisture–temperature changes and freeze–thaw hazards Yu, F., Guo, P., Lai, Y. and Stolle, D. (2020a). Frost heave
on a canal in seasonally frozen regions. Natural Hazards, and thaw consolidation modelling. Part 1: A water flux
72(2), 287–308. function for frost heaving. Canadian Geotechnical
Lunardini, V. J. (1978). Theory of n-factors and correlation of Journal, 57(10), 1581–1594.
data. 3rd International Conference on Permafrost, 40–46. Yu, F., Guo, P., Lai, Y. and Stolle, D. (2020b). Frost heave
Michalowski, R.L. (1993). A constitutive model of saturated and thaw consolidation modelling. Part 2: One-
soils for frost heave simulations. Cold Regions Science dimensional thermohydromechanical (THM) framework.
and Technology, 22(1), 47–63. Canadian Geotechnical Journal, 57(10), 1595–1610.
Finite Element Modelling of Frozen Soil- Pile
Interaction
Faranak Sahragard, Mehdi Pouragha & Mohammad Rayhani
Department of Civil and Environmental Engineering –Carleton
University, Ottawa, ON, Canada

ABSTRACT
Numerical simulation of soil-pile interaction poses challenges pertaining to the accurate modelling of the interface. The
problem is more relevant to piles in frozen soils where the bearing capacity of the pile relies mostly on the cohesive strength
of the adfreeze bonds between frozen soil and pile surface. This paper describes an interface model that used concurrent
damage frictional mechanisms to capture the brittle response of the adfreeze bonds. The model is implemented into an
FEM solver (ABAQUS). Comparisons with direct shear test results on such interfaces indicate the accuracy of the model
in capturing the temperature and pressure dependency of the peak stresses at the interface. This interface model has also
been adopted to carry out finite element simulations of a pull-out experiment carried out in lab. The simulation is shown to
successfully predict the stiffness and the maximum tensile force with high accuracy without additional calibrations, which
further verifies the performance of the developed interface model.

RÉSUMÉ
La simulation numérique de l'interaction sol-pieu pose des défis concernant la modélisation précise de l'interface. Le
problème est plus pertinent pour les pieux dans les sols gelés où la capacité portante du pieu repose principalement sur
la force de cohésion des liaisons adgelées entre le sol gelé et la surface du pieu. Cet article décrit un modèle d'interface
qui utilise des mécanismes de frottement de dommages simultanés pour capturer la réponse fragile des liaisons adfreeze.
Le modèle est implémenté dans un solveur FEM (ABAQUS). Les comparaisons avec les résultats d'essais de cisaillement
direct sur de telles interfaces indiquent la précision du modèle dans la capture de la dépendance à la température et à la
pression des contraintes maximales à l'interface. Ce modèle d'interface a également été adopté pour effectuer des
simulations par éléments finis d'une expérience d'arrachement réalisée en laboratoire. Il est démontré que la simulation
prédit avec succès la rigidité et la force de traction maximale avec une grande précision sans étalonnages
supplémentaires, ce qui vérifie davantage les performances du modèle d'interface développé.

1 INTRODUCTION adfreeze strength of model PVC pile-frozen soil

interface. Liu et al. (2014) conducted large-scale direct
Pile foundations are widely used to support structures on shear tests on the frozen silty clay-concrete interface,
frozen soils due to their relatively high bearing capacity, employing a constant displacement rate. Their findings
resistance to frost jacking and minimum disturbance to revealed a significant correlation between temperature
frozen ground thermal regime during installation. In such and both peak and residual adfreeze strength. The study
cold regions, pile foundations derive their bearing established a linear relationship between temperature
capacity mainly from the adfreeze bond that forms variations and the strength characteristics of the
between frozen soil and pile surface. The strength of the interface. In their study, Abdulrahman (2019) conducted
adfreeze bond, in turn, depends on multiple parameters direct shear tests and pile pull-out tests to investigate the
such as temperature, soil type, pile surface roughness, interface behavior of ice-poor and ice-rich soils, as well
normal pressure at the interface, loading rate, degree of as steel piles. The research aimed to explore the impact
saturation, and unfrozen water content, the effects of of temperature on the interface characteristics. The pile
which have been previously explored to various extents, tests were conducted under varying conditions, including
through experimental studies (Crory, 1963; Ladanyi, laboratory and field settings, considering different types
1995; Morgenstern et al., 1980). of piles. Another study by Sun et al. (2021) focused on
The adfreeze strength of frozen soil-pile interface is investigating the frozen silt-concrete adfreeze strength.
commonly explored through two types of experiments: They carried out orthogonal direct shear tests and tensile
pile pull-out test and direct shear test. Parameswaran tests, considering factors such as moisture content,
(1978) performed model pile tests, where piles made of temperature, and dry density. The researchers
different materials (wood, steel, or concrete) embedded suggested an exponential relationship between shear
in frozen fine sand were subjected to different loading strength and tensile strength at the interface, highlighting
rates. The results suggested a power-law expression to the influence of these factors on adfreeze strength.
relate adfreeze strength to the loading rate. Biggar and Furthermore, Gao et al. (2022) conducted a series of
Sego (1993) analyzed the adverse effect of salinity on multistage creep tests to examine the creep behavior of
the small-scale concrete model pile-frozen clay where 𝒕 is the nominal traction vector containing three
interface. Their research specifically considered the components: 𝑡𝑛 , 𝑡𝑠 , and 𝑡𝑡 which represent the tractions
effect of ice content on the behavior of the pile-frozen along normal and the two tangential directions, related
soil interface. Based on their findings, the authors to the deformation along these directions given by 𝛿𝑛 , 𝛿𝑠 ,
proposed a creep model that accounts for the influence and 𝛿𝑡 . The elastic moduli are given by 𝐾𝑛𝑛 , 𝐾𝑠𝑠 , and 𝐾𝑡𝑡 .
of ice content on the interface behavior. The elasticity along normal and shear components is
On the numerical simulation front, despite the recent assumed to be uncoupled, resulting in zero off-diagonal
advances, the issue of realistically modelling the terms in the stiffness matrix.
adfreeze interface behaviour requires more careful The onset of damage, where the degradation of
attention. Previous studies commonly resort to simple cohesion begins, is determined by defining a damage
pre-packaged models for representing the interface initiation criterion. In this research quadratic nominal
characteristics, neglecting the complexity of its behavior. stress criterion is selected which can be represented as:
For example, Coulomb friction model is among the most
commonly used models which has been adopted with
frictional and/or cohesive terms either of which are ⟨𝑡 ⟩ 2 𝑡 2 𝑡
2
sometimes made temperature-dependent, see for { 𝑡𝑛𝑜 } + {𝑡 𝑜𝑠 } + {𝑡 𝑜𝑡 } = 1 [2]
instance Zhu et al. (2012), Fei et al. (2019), and Wang 𝑛 𝑠 𝑡
et al. (2017). However, Coulomb-based models alone
can be deemed too simple in that they overlook the brittle
failure and damage characteristics of adfreeze bonds. where tno , tso and tto are peak nominal stresses when
In this study, we adopt an interface model based on deformation is purely along 𝑛, 𝑡, and 𝑠 directions. The
damage and friction mechanisms for simulating adfreeze symbol ⟨ ⟩ represents the Macaulay brackets which is
interface strength in FEM simulations of frozen soil-pile defined as:
interface. The model parameters are made temperature-
dependent with the interface maximum strength being
also pressure-dependent. The model has been 𝑥, 𝑥 ≥ 0
⟨𝑥⟩ = { [3]
implemented in a FE solver (ABAQUS) to simulate direct 0, 𝑥 < 0
shear tests carried out by Aldaeef and Rayhani (2021).
The comparison with the experimental data shows the
accuracy of the interface model in capturing temperature Once initiated, the damage level evolves with shear
and pressure dependency of the shear strength. The and normal deformation resulting in degradation of
calibrated model is then used to simulate the pull-out test stiffness. According to the traction-separation behavior
by Aldaeef and Rayhani (2021). The results indicate that observed in direct shear test results, a linear damage
the simulations with the developed interface model evolution is adopted. Tractions are calculated by a
accurately capture the pull-out experiments conducted in damage factor, 𝐷, which represents the level of
the lab conditions. degradation and evolves from 0 at the beginning of
damage to an ultimate value of 1. Figure 1 illustrates a
linear damage law. The traction components affected by
2 INTERFACE MODEL damage are determined by:
The interface of frozen soils and piles often undergo a
brittle failure signified by a sudden and catastrophic loss (1 − 𝐷)𝐾𝑛𝑛 𝛿𝑛 , 𝑡̄𝑛 ≥ 0
of load-carrying capacity without significant warning or 𝑡𝑛 = { [4]
𝑡𝑛̄ , 𝑡𝑛̄ < 0
deformation. The abrupt failure may lead to excessive
settlement, structural damage, or even collapse of the
supported structure. Therefore, reliably modelling such 𝑡𝑠 = (1 − 𝐷)𝐾𝑠𝑠 𝛿𝑠 [5]
brittle responses becomes crucial for assessing the
safety and stability of structures.
The developed interface model includes frictional 𝑡𝑡 = (1 − 𝐷)𝐾𝑡𝑡 𝛿𝑡 [6]
and damage components that are concurrently activated
to capture the pressure dependency of the maximum
adfreeze strength. Equation 4 ensures that there is no reduction to
A damage mechanism is used here to model the normal stiffness in case of compression. The evolution
degradation of adfreeze strength due to brittle failure.
of the damage factor (𝐷) is described as:
Before the damage threshold is reached, a linear elastic
modulus describes the relation between the traction and
the deformation at the interface: 𝑓 𝑚𝑎𝑥 −𝛿 ∘ )
𝛿𝑚 (𝛿𝑚
𝐷= 𝑚
𝑚𝑎𝑥 (𝛿 𝑓 −𝛿 ∘ )
[7]
𝛿𝑚 𝑚 𝑚
𝑡𝑛 𝐾𝑛𝑛 0 0 𝛿𝑛
𝒕 = [𝑡𝑠 ] = [ 0 𝐾𝑠𝑠 0 ] [𝛿𝑠 ] [1]
𝑡𝑡 0 0 𝐾𝑡𝑡 𝛿𝑡
where 𝛿𝑚 ∘ 𝑓
and 𝛿𝑚 are effective displacements at depth of frozen soil and thickness of steel plates are 20
damage initiation and complete damage, and 𝛿𝑚 𝑚𝑎𝑥
is mm and 25.4 mm respectively.
maximum effective displacement in loading history. The
effective displacement is defined as:

𝛿𝑚 = √⟨𝛿𝑛 ⟩2 + 𝛿𝑠 2 + 𝛿𝑡 2 [8]

Figure 1. Bilinear damage law ((Abaqus

Documentation, 2014)

To include the effect of friction on interface shear

stress after damage initiation, the contribution of
cohesion and friction to shear stress is defined as
follows: Figure 2. Linear dependency of (a) cohesion and (b)
friction of interface on temperature ((Aldaeef and
Rayhani, 2021)
𝜏𝑓 = 𝐶(𝑇, 𝐷) + 𝜎𝑛 tan(𝜑(𝑇)) [9]

The soil used in the tests was classified as poorly

The temperature dependency of cohesion and graded sand, representing ice-poor soil conditions. Steel
friction is described based on the experimental results plates were employed to mimic the surface
from Aldaeef and Rayhani (2021), as illustrated in Figure characteristics of a typical steel pile. Strain-controlled
2. conditions were maintained during the tests, with various
temperatures and normal stresses considered. The
experiments also examine the effect of loading rate
3 FINITE ELEMENT MODELLING which is not considered herein. We limit our simulations
to experiments with displacement rates of
ABAQUS software is used here for implementing the 0.011 mm.min-1.
interface models and simulating direct shear and pull-out The steel is modeled as a linear elastic material with
experiments. The interface model is implemented via a elastic modulus of 210 GPa and Poisson’s ratio of 0.3.
UMAT subroutine. The model uses interface elements The tests were performed in four sub-zero temperatures
and assigns a fictitious thickness to convert forces and which are considered in the simulation: -1o C, -4o C, -7o C
displacements into stresses and strains. A unit thickness and -10o C. Temperature dependent Mohr-Coulomb
is chosen in our case. A single layer of cohesive model is used to describe the stress-strain response of
elements is used at the interface and tie constraints are the frozen soil. The corresponding magnitudes of friction
employed to ensure that the displacement between the angle and cohesion obtained from the test results are
cohesive elements and the surrounding elements listed in Table 1. The displacement of steel plate is
remains continuous. restricted in horizontal directions and frozen soil is
subjected to a shear displacement of 3 mm. Three
3.1 Simulation of direct shear tests normal pressures of 0.025 MPa, 0.1 MPa and 0.2 MPa
are applied and maintained during the shearing stage.
The direct shear experiments on the adfreeze strength The geometry of the model is shown in Figure 3. C3D8
of frozen sand reported by Aldaeef and Rayhani (2021) elements are assigned to soil and steel plate and
is used here for calibration and primary evaluation of the cohesive layer is modeled using COH3D8 element type.
models’ accuracy. The interface area in the modified Based on the experimental results, an elastic shear
shear box had dimensions of 60 mm by 60 mm. The stiffness of 1.15 GPa. m−1 is considered along all
directions for the interface model. This is a simplifying
assumption since the interface is usually stiffer along the
normal direction.
The temperature dependent cohesion and friction at
the interface follow the curves given in Figure 2. The only
𝑓
remaining parameter for damage models is 𝛿𝑚 which is
kept at 1mm.

Table 1. Temperature dependent soil properties

(Aldaeef and Rayhani2021)

Temperature (oC) Cohesion (MPa) Tan(Φ)

-1 0.44 1.76
-4 0.94 1.73
-7 1.74 3.10
-10 2.26 3.40

Figure 3. FE model geometry and boundary conditions

The traction-separation responses of interface are

illustrated in Figure 4. The results show that the
simulations can successfully capture both temperature
and pressure dependency of the peak shear stress, as Figure 4. Comparison between the results obtained
well as the post-peak brittle failure. Importantly, the latter from FE simulation and experiments (Aldaeef and
falls beyond the capability of simple interface models Rayhani, 2021)). Solid lines and dashed lines represent
based only on Mohr-Coulomb criterion. the experiment and FEM results, respectively.
The deviation between the peak traction observed in
the test results and the FE simulation can be attributed
to errors in estimation of cohesion and friction angle from 4 FE MODELLING OF PILE PULL-OUT TESTS
direct shear tests results, as well as the simplified linear USING CONCURRENT DAMAGE-FRICTIONAL
relationships used to describe the temperature MODEL
dependency of cohesion and friction properties. It is
worth noting that, as another simplifying assumption, the As a further verification, the calibrated interface model is
interface model uses the same values of friction angle at employed to simulate a pull-out test on steel piles in
the initial and residual states. However, the test results frozen soil as reported by Aldaeef and Rayhani (2021).
indicate that the actual residual friction is smaller than The details of the test and FE simulation procedure are
the friction observed at the peak which results in defined in the following section.
overestimation of residual shear stress. The issue is, The pull-out test modelled in ABAQUS is conducted
nevertheless, not pursued here since the accuracy of the on a steel model pile with an outer diameter of 50 mm
experimental data in measuring the residual strength is
and length of 200 mm embedded in a frozen soil sample
lower.
with length of 200 mm, width of 100 mm and height of frozen soils at different temperatures and normal
200 mm. The test is carried out under controlled pressures. Comparisons with the experimental data
laboratory conditions at a constant temperature of -7o C. indicate the accuracy of the model in capturing the
The loading rate and frozen soil type are similar to the pressure and temperature dependency of the adfreeze
interface direct shear tests described in earlier sections. interface strength. For further verifications, the same
The same material properties of frozen sand, steel pile interface model is utilized to simulate a pull-out test
and interface elements defined in the FE modeling of carried out under the lab conditions. The results
direct shear tests are applied in this simulation. The demonstrate a good agreement between the lab
gravity force is applied to the whole model before the experiments and the finite element simulations when the
pull-out loading step. To establish boundary conditions, pile pull-out test conditions, such as soil type and strain
the base of the model is constrained from movement in rate, closely resemble the original test conditions based
the x, y, and z directions, while the walls are restricted on which the interface model was calibrated. The verified
only in the direction normal to each wall surface. A interface model can be used in future for a more reliable
displacement of 3 mm is gradually imposed on the pile simulation of frozen soil-pile interaction with more
head. complex thermo-mechanical loading conditions.

5 RESULTS AND DISCUSSION 7 REFERENCES

The force displacement curves obtained from the FE pile Abaqus, V. 2014. 6.14 Documentation, ‖ Dassault Syst.
pull-out model is compared to the test results in Figure Simulia Corp, 651: 2–6.
5. The prediction of FE model demonstrates a good Abdulrahman, A.A.H. 2019. Load Transfer and Creep
agreement with laboratory pile test in terms of peak pull- Behavior of Pile Foundations in Frozen Soils.
out force and the corresponding pile displacement. This Carleton University.
accuracy is significant considering that the interface Aldaeef, A.A., and Rayhani, M.T. 2019a. Load transfer
parameters are calibrated based on the direct shear test and creep behavior of open-ended pipe piles in
and the model parameters are not changed frozen and unfrozen ground. Innovative
retrospectively. The simulation predicts a sharper post- Infrastructure Solutions, 4: 1–14. Springer.
peak stress drop. Such an error is, however, expected Aldaeef, A.A., and Rayhani, M.T. 2019b. Interface shear
since, in the lack of reliable experimental data, the post strength characteristics of steel piles in frozen clay
damage behaviour of the model was not accurately under varying exposure temperature. Soils and
calibrated. Moreover, the evolution of damage can also Foundations, 59(6): 2110–2124. Elsevier.
depend on normal pressure, which is ignored in the Aldaeef, A.A., and Rayhani, M.T. 2021. Pile-soil
model. interface characteristics in ice-poor frozen ground
under varying exposure temperature. Cold
Regions Science and Technology, 191: 103377.
Elsevier.
Biggar, K.W., and Sego, D.C. 1993. Field pile load tests
in saline permafrost. I. Test procedures and
results. Canadian Geotechnical Journal, 30(1):
34–45. NRC Research Press Ottawa, Canada.
Crory, F.E. 1963. Pile foundations in permafrost. In
Proceedings of Permafrost International
Conference.
Fei, W., Yang, Z.J., and Sun, T. 2019. Ground freezing
impact on laterally loaded pile foundations
considering strain rate effect. Cold Regions
Science and Technology, 157: 53–63. Elsevier.
Gao, Q., Wen, Z., Zhou, Z., Brouchkov, A., Wang, D.,
and Shi, R. 2022. A creep model of pile-frozen soil
interface considering damage effect and ice effect.
Figure 5. Comparison between FE model results and International Journal of Damage Mechanics,
experimental results (Aldaeef and Rayhani, 2021) 31(1): 3–21. Sage Publications Sage UK: London,
England.
Ladanyi, B. 1995. Frozen soil — structure interfaces. In
6 CONCLUSION In Studies in Applied Mechanics. Elsevier. pp. 3–
33.
This study outlines a model for the adfreeze strength of Liu, J., Lv, P., Cui, Y., and Liu, J. 2014. Experimental
pile-frozen soil interface, that includes damage and study on direct shear behavior of frozen soil–
friction mechanisms. The model has been implemented concrete interface. Cold regions science and
into ABAQUS and used to reproduce several previously technology, 104: 1–6. Elsevier.
conducted direct shear tests on the interface of steel and
Morgenstern, N.R., Roggensack, W.D., and Weaver,
J.S. 1980. The behaviour of friction piles in ice and
ice-rich soils. Canadian Geotechnical Journal,
17(3): 405–415. NRC Research Press Ottawa,
Canada.
Parameswaran, V.R. 1978. Adfreeze strength of frozen
sand to model piles. Canadian geotechnical
journal, 15(4): 494–500. NRC Research Press
Ottawa, Canada.
Sun, T., Gao, X., Liao, Y., and Feng, W. 2021.
Experimental study on adfreezing strength at the
interface between silt and concrete. Cold Regions
Science and Technology, 190: 103346. Elsevier.
Wang, T., Liu, J., Tai, B., Zang, C., and Zhang, Z. 2017.
Frost jacking characteristics of screw piles in
seasonally frozen regions based on thermo-
mechanical simulations. Computers and
Geotechnics, 91: 27–38. Elsevier.
Zhu, J., Zou, L.Q., and Fu, H.L. 2012. Numerical
Simulation Analysis on Effects of Freezing and
Thawing on Workover Rig Anchor Pile Bearing
Performance. Applied Mechanics and Materials,
204: 714–717. Trans Tech Publ.
Modelling of Consolidation Settlements of
Organic Soils
Ahmed Moslem, Sumi Siddiqua, Ramy Saadeldin & Ahmed Mahgoub
School of Engineering, University of British Columbia, Okanagan Campus,
Kelowna, BC Canada

ABSTRACT
Organic soils are characterized by their high organic contents, high initial water contents, high compressibility, and low
shear strength. Organic soils exhibit high volume changes under loading conditions and undergo significant primary
consolidation and continuous long-term secondary consolidation settlements. The mechanical behaviour of organic soils
is complex and requires extensive soil testing to obtain representative geotechnical parameters to model its behaviour.

This paper includes a review of the key constitutive models that could be utilized to model soft soils. Afterward, finite
element analyses were developed using PLAXIS 2D for a well-documented case study of an embankment on soft clay
deposit. The finite element model is validated against field measurements using different constitutive soil models to
investigate their adequacy in modelling organic soil behaviour. Then, the recommended soil model was used to simulate
available laboratory tests (i.e., consolidation tests) of organic soils with different organic contents to verify the limitations
and drawbacks of the recommended constitutive model. As a result of this review and the numerical modelling analyses,
an approach to simulate the consolidation behaviour of organic soils has been developed.

RÉSUMÉ
Les sols organiques sont caractérisés par leur contenu organique élevé, leur teneur initiale élevée en eau, leur
compressibilité élevée et leur faible résistance au cisaillement. Les sols organiques présentent des changements de
volume élevés dans des conditions de chargement et subissent une consolidation primaire importante et un tassement de
consolidation secondaire continu à long terme. Le comportement mécanique des sols organiques est complexe et
nécessite des essais de sol approfondis pour obtenir des paramètres géotechniques représentatifs afin de modéliser son
comportement.

Cet article comprend un examen des principaux modèles constitutifs qui pourraient être utilisés pour modéliser les sols
mous. Par la suite, une analyse par éléments finis a été développée à l'aide de PLAXIS 2D pour une étude de cas bien
documentée d'un remblai sur dépôt d'argile molle. Le modèle d'éléments finis est calibré par rapport à des mesures sur le
terrain à l'aide de différents modèles de sol constitutifs pour étudier leur adéquation à la modélisation du comportement
organique du sol. Ensuite, le modèle de sol recommandé a été utilisé pour simuler les tests de laboratoire disponibles
(c'est-à-dire les tests de consolidation) de matières organiques avec différents contenus organiques afin de vérifier les
limites et les inconvénients du modèle constitutif recommandé. À la suite de cet examen et de la modélisation numérique,
une approche pour simuler le comportement de consolidation des sols organiques a été développée.

1 INTRODUCTION 2017), (Tschuchnigg & Schweiger, 2018), (Nithurshan et

al., 2019), (Feng et al., 2021), and (Long et al., 2022). The
The Canadian landscape contains over 1.5M km2 of majority of these models assume the Soft Soil Creep (SSC)
organic soils (such as Muskeg soils). Organic soils model for simulating organic soils. Wong & Somanathan
generally require special consideration for infrastructure (2019) performed analytical analysis on high organic peat
design as they exhibit high volume changes under loading that produced comparable results to the Mohr-Coulomb
conditions and undergo significant primary consolidation model and oedometer test settlement results.
and continuous secondary consolidation settlements. The Constitutive soil models available for modeling the
engineering problems associated with organic soils have consolidation behavior of soils have generally been
been researched and supported by enormous field and formulated based on experimental data and theoretical
laboratory tests in Canada since 1945 (Pihlainen, 1963). principles (Lade, 2005). The capabilities and deficiencies
Limited research has, however, been completed to of these models are not easy to determine, and the
model and validate the complex consolidation behavior of requirements for the determination of input parameters are
the Canadian Muskeg soils. In addition, the behavior of not uniform (Lade, 2005). The finite element program
Muskeg soils has not been well documented for (PLAXIS) is one of the most common finite element
engineering design purposes. Available constitutive soil programs that employ several constitutive models,
models typically used to simulate soft /organic soils have including Mohr-Coulomb model (MC), Hardening Soil
to be supported by related field and laboratory testing model (HS), Hardening Small Strain model (HSsmall), Soft)

before they are used for these Muskeg soils. Soil model (SS), Soft Soil Creep (SSC) model, and
Advanced constitutive soil models are available to Modified Cam-Clay model (MCC) (Bently, 2021). Each of
predict the short-and long-term compressibility responses these models has some limitations in modelling organic
of soils due to embankment loads (Tyurin & Nevzorov, soils.
 This paper has examined the adequacy of MC, HS, The Hardening Soil (HS) model is considered as an
HSsmall, SS, SSC, and MCC constitutive models for the extension of the Mohr–Coulomb Model (Voznesensky &
consolidation behavior of highly organic soils based on a Nordal, 1999). The HS model is, however, proven to
parametric study for a well-documented case study. A address the limitation of the Soft Soil (SS) model with
numerical modelling approach was developed and respect to the over-consolidated soil behaviour by using
validated from the modeling results generated using the the theory of plasticity rather than the theory of elasticity
different constitutive models. Finally, consolidation tests (Müthing et al., 2018) and (Schanz, 1999). In the HS
conducted on various organic soil samples have been model, the total strains are developed based on the
modelled using the selected constitutive model to plasticity theory using the stress dependency of the soil. A
determine its limitations. volumetric cap yield surface was incorporated with the
ability to expand due to the plastic strain with consideration
2 METHODOLOGY of the pre-consolidation pressure.
The Hardening Soil model with small-strain stiffness
The following summarizes the numerical analysis (HSsmall) incorporates a stiffness-strain relation based on
conducted using the finite element program PLAXIS 2D the hyperbolic model developed by Hardin & Drnevich,
V20. (1972). The small-strain stiffness refers to the stiffness of
• A detailed review of six constitutive soil models the soil when subjected to very small strains or
generally used for soft soils were first completed. This deformations. When the applied stress exceeds the yield
included a parametric study to determine the various stress, the model incorporates hardening behaviour to
parameters required for each model. capture the non-linear behavior of the soil. The HSsmall is
• The constitutive soil models were used to predict the considered as an extension of the HS model.
settlements over time for the case study embankment The Soft Soil (SS) model utilizes the failure criterion of
and were compared with field measurements. The the Mohr-Coulomb, which is the friction angle (φ') and the
results were used to understand the limitations of each cohesion (c') components. The SS model is mainly
soil model and to validate the selected constitutive dependent on a linear stress-dependency stiffness. The SS
model. model is based on the MCC model on primary compression
• A parametric study was finally conducted to further for normally consolidated clay soil (Roscoe & Burland,
examine the constitutive model in simulating the effect 1968). The yield surface cap of the SS model represents
of the organic content of soft soils. an ellipsoidal cap of perfectly plastic Mohr-Coulomb at the
failure state in the (p'-q') plane. In the SS model, stress
3 SOIL MODELS states are not allowed over the MC failure condition. In
addition, softening strain is not allowed.
Various soil constitutive models are available in PLAXIS to The Soft Soil Creep (SSC) model is a rate dependent
model soft clayey soils. The suitability of each of these that combines the isotache principle of viscous
models depends on the soil type and its geological compression with the Cam Clay model of shear strength
composition. The determination of representative soil with further development of the SS model (Grimstad &
parameters depends significantly on the availability of Nordal, 2018), and (Vermeer et al., 1998). The SSC model
laboratory and field tests of soil stiffness and strength is designed to consider creep relaxation, creep hardening,
parameters. and critical state shear deformations.
In order to model the complex behavior of highly Although, the HS and SS models are typically used to
organic soils that extend to great depths throughout predict the behaviour of compressible soft soils. The SSC
Canada posing many challenges to various engineering model has increasingly been referred to as the most
applications, this research first examined the well- appropriate constitutive model for accounting for the
developed soil models that are commonly used for very soft secondary compression in soft soil.
to soft soils. The investigation of the soil models was aimed The Modified Cam Clay (MCC) model is an elastic-
at selecting the soil model that can depict the organic soil plastic strain hardening model (Roscoe & Burland, 1968).
behaviour. The features and capabilities of each of these The derivation of the MCC model is based mainly on the
models were well reviewed. critical state theory, including the logarithmic relationship
Most soft soil types exhibit primary compression that is between void ratio (e) and the mean effective stress (p') in
followed by creep settlements, especially organic clays and virgin isotropic compression.
peat. Organic soils under certain conditions of imperfect The MCC yield surface represents an ellipse in (p'-q')
aeration are decomposed and, with pressure, introduces plane. The yield surface defines the boundary between the
certain creep behavior. The models considered are Mohr- elastic and plastic strain increments for the stress path. The
Coulomb model (MC), Hardening Soil model (HS), stress paths within this boundary give elastic strains,
Hardening Small Strain model (HSsmall), Soft Soil model whereas stress paths tend to cross the yield surface give
(SS), Soft Soil Creep (SSC) model, and Modified Cam-Clay elastic and plastic strains. The top of the ellipse intersects
model (MCC). a line called the critical state line (CSL). The Tangent (M)
The Mohr-Coulomb (MC) soil model is a linearly elastic, is the slope of the CSL, which determines the height of the
perfectly plastic soil model. The model input parameter ellipse, and 𝜆 and 𝜅 are two compression indices. The pre-
requires a constant soil stiffness, and the model consolidated pressure regulates the hardening through the
computations are relatively fast due to a linear increasing volumetric strain.
soil stiffness with respect to depth.
4 MODEL DEVELOPMENT from the boundary conditions. The model consisted of
around 11,096 elements and 17,096 nodes.
4.1 Case Study
4.2 One-Dimensional Consolidation
The case study used to validate the soil models consisted
of an earth embankment built on soft clay deposit as To examine the Soft Soil Creep (SSC) model, the time-
documented in previous research studies by the Australian dependent one-dimensional (1-D) consolidation was
Research Council Centre. Figure 1 presents the soil and modelled for organic soils with a range of organic contents
embankment profiles. The soil profile consisted of 2.7 m from 15% to 95%. The modelling loading conditions, and
thick soft alluvium deposit that is underlain by a 12 m thick geometry were in accordance with the ASTM D 2435. Each
very soft clay. The earth embankment was 3 m in height load increment was applied for a constant time of 24 hrs in
with a 15 m wide crest and side slopes of 3H: 2V. The each stage. The consolidation testing results were
construction of the embankment was completed in three collected from various sources (Hameedi, Al Omari, et al.,
stages within approximately 60 days. Pre-installed vertical 2020), (De Guzman & Alfaro, 2018), (Amuda et al., 2019)
drains (PVDs) were used to expedite the soil consolidation and (Mohamad et al., 2021).
with a surface drainage layer of 0.6 m thick. The geometry Testing performed by Hameedi, Al Omari, et al., (2020),
of the soil strata under the embankment load has been Johari et al., (2016) consisted of an organic soil with an
described by Buddhima Indraratna et al., (2018) , as shown organic content of 15%. The loading scheme included a
in Figure 1. stress range of 25 kPa - 1600 kPa. Testing performed by
The PVDs were installed to a depth of 15 m in a square De Guzman & Alfaro, (2018) consisted of a peat sample
pattern at 1.2 m spacing to cover the 24 m width of the main with an organic content of 57% with a loading range of 3.5
embankment area. The water table was considered at the kPa - 122 kPa. Testing performed by Johari et al., (2016)
ground surface. The soil layers underneath the consisted of peat samples with an organic content of 83%
embankment were determined to be fully saturated. with a loading range of 5 kPa - 320 kPa. Testing performed
The model captured the analysis by Buddhima by (Amuda et al., 2019) consisted of amorphous peat with
Indraratna et al., (2018). The developed mesh is shown in an organic content of 92% with a load range of 10 kPa -
Figure 2. A 6-noded 2D triangular element was used in the 160 kPa.
plane strain analysis. A fully fixed boundary condition was
assumed at the base of the model. A free boundary 5 SOIL PROPERTIES AND PARAMETERS
condition was assigned in the vertical direction and roller
boundary condition assigned in the horizontal direction. 5.1 Case Study
The lateral extent of the model was assumed at 50 m
from the centerline of the embankment that was considered Table 1 summarizes the parameters selected for the
sufficient to model the settlements associated with the embankment and soil layers for the six constitutive models.
embankment loading and to avoid any significant effects The parameters were determined based on field and

Figure 1. Embankment and Soil Profiles

 Figure 2. Developed Mesh

laboratory testing as included in (Pineda et al., 2016) and 6 RESULTS AND DISCUSSION
(Buddhima Indraratna et al., 2018). The field testing
included seismic dilatometer tests (SDMTs) and cone Figure 3 shows the predicted settlements near the ground
penetration testing with pore pressure measurements surface versus time for the case study embankment using
(CPTu tests). The laboratory testing included oedometer the six constitutive soil models (MC, HS, HSsmall, SS,
tests (Constant Rate of Strain (CRS) tests, Incremental SSC, and MCC) as well as the field measurements. Point
Loading (IL) tests, and undrained triaxial tests. (a) as shown on Figure 2 was the location of the settlement
For the simulation of the PVDs, the horizontal monitoring from the field program (Buddhima Indraratna et
permeability in the undisturbed zone for the plane strain al., 2018).
model was based on the in-situ permeability. The ratio of Figure 3 shows the deficiency of the Mohr-Coulomb
the permeability in the undisturbed zone to that in the model in predicting consolidation settlements for soft soils.
smear zone ((kh/ks), and the normalised size of smear As can be seen from Figure 3, the SSC model and HSsmall
zone (ds/dw) were evaluated to be 2.3 and 7.8, predicted settlements close to the field measurements for
respectively (B Indraratna & Redana, 2014). primary and secondary consolidation. The HS model
predicted with higher settlements within the first year from
5.2 One-Dimensional Consolidation construction. In contrast, the SS model predicted close
primary consolidations settlements, but underestimated
Table 2 summarizes the parameters selected to model the the secondary settlements. This behaviour was previously
time-dependent one-dimensional (1-D) consolidation, observed by (Hameedi, Fattah, et al., 2020). It can be also
using the SSC model, for organic soils with a range of noted that the SS model predicted much lower settlements
organic contents from 15% to 95%. compared to the SSC model.
Figure 4 shows the excess pore pressure (EPWP)
versus time for the case study embankment using the six
Table 2. The parameters for consolidation test model constitutive soil models (MC, HS, HSsmall, SS, SSC, and
OC% λ* κ*
γ(KN/
Source
MCC) as well as the field measurements. Point (b) at a
m3) depth of 6.0 m as shown on Figure 2, was the location of
15 0.057 0.0193 17.6 Hameedi et al., (2020) the monitoring point from the field program (Buddhima et
57 0.20 0.0167 12 De Guzman & Alfaro, (2018) al., 2018). As presented in Figure 4, the excess pore
83 0.125 0.0158 10.75 Johari et al., (2016)
pressure increased during the embankment construction
92 0.179 0.0719 11 Amuda et al., (2019)
95 0.224 0.0245 12 De Guzman & Alfaro, (2018) period then the dissipation was well-developed at the end
of the consolidation process. All soil models except MC
predicted excess porewater dissipation in general
 Figure 3. Predicted settlements using soil models Figure 4. Predicted excess water pressures using the
and field measurements at point a. soil models and field measurements at point b.

Figure 5 One-Dimensional Consolidation Test Model Results for Different Organic Contents.

agreement with the field measurements for the primary measurements. The HS and HSsmall have slightly
consolidation period. overestimated the excess porewater pressures during the
The SSC and MCC models have predicted excess primary consolidation. The MC model under-estimated the
porewater pressures in very close agreement with the field excess pore pressure.
 In general, the compressibility behavior of organic soils Tagung–in Memoriam of Professor Tom Schanz.
is functional of its organic content. Figures 5 A to 5 E, show
the void ratio versus the effective vertical from the one- Hameedi, M. K., Al Omari, R. R., & Fattah, M. Y. (2020).
dimensional consolidation test model for different organic Compression and Creep Indices of Organic Clayey
contents between 15% to 95%. This has been completed Soil. IOP Conference Series: Materials Science and
to validate the SSC model developed in this research to Engineering, 671(1).
use in the simulation of the Muskeg soils in Western
Canada. Hameedi, M. K., Fattah, M. Y., & Al-Omari, R. R. (2020).
The SCC model captured well the soil consolidation for Creep characteristics and pore water pressure
organic soils. However, it is noted that the model may changes during loading of water storage tank on soft
slightly misestimate the compressibility of the soils based organic soil. International Journal of Geotechnical
on its organic content. This can generally be overcome by Engineering,14(5),527–537.
validating the model to field or laboratory testing results as
well as by further incorporating the creep behaviour of Hardin, B. O., & Drnevich, V. P. (1972). Shear modulus and
organic soils. damping in soils: design equations and curves.
Journal of the Soil Mechanics and Foundations
7 SUMMARY AND CONCLUSIONS Division, 98(7), 667–692.

This paper presented a numerical modeling approach to Indraratna, B, & Redana, I. W. (2014). Numerical modeling
examine the available soil constitutive models for the of vertical drains with smear and well resistance
simulation of soft soils. Six constitutive soil models (MC, installed in soft clay. May.
HS, HSsmall, SS, SSC, and MCC) were used in the finite
element program PLAXIS 2D for the simulation of a case Indraratna, Buddhima, Baral, P., Rujikiatkamjorn, C., &
study embankment built on multi-layer soft soil deposit as Perera, D. (2018). Class A and C predictions for
well as for one-dimensional consolidation tests conducted Ballina trial embankment with vertical drains using
on organics soils with various organic contents. standard test data from industry and large diameter
In general, the SSC model captured the consolidation test specimens. Computers and Geotechnics, 93,
of the soft clay deposit. The model was found to be capable 232–246.
of estimating the primary consolidation settlements and
excess porewater pressures with time. The settlements Johari, N. N., Bakar, I., Razali, S. N. M., & Wahab, N.
estimated by the model started to deviate slightly from the (2016). Fiber Effects on Compressibility of Peat. IOP
field measurements after the primary compression. Conference Series: Materials Science and
It has been noted that the SSC soil model may Engineering, 136(1).
misestimate the soil consolidation settlements based on its
organic content. This highlights the importance of carefully Lade, P. V. (2005). Overview of Constitutive Models for
assuming the soil parameters to represent the behavior of Soils. 40771(June), 1–34.
organic soils as well as to capture its creep behaviour.
Long, M., Paniagua, P., Grimstad, G., Trafford, A., Degago,
8 REFERENCES S., & L’Heureux, J. S. (2022). Engineering properties
of Norwegian peat for calculation of settlements.
Amuda, A. G., Hasan, A., Unoi, D. N. D., & Linda, S. N. Engineering Geology, 308(June), 106799.
(2019). Strength and compressibility characteristics
of amorphous tropical peat. Journal of Mohamad, H. M., Zainorabidin, A., Musta, B., Mustafa, M.
GeoEngineering, 14(2), 85–96. N., Amaludin, A. E., & Abdurahman, M. N. (2021).
Bently. (2021). Connect edition V21.01 PLAXIS 3D - Compressibility behaviour and engineering
Material Models Manual. Plaxis 3D Connect Edition properties of north borneo peat soil. Eurasian Journal
V21, 1–274. of Soil Science, 10(3), 259–268.

De Guzman, E. M. B., & Alfaro, M. C. (2018). Geotechnical Müthing, N., Zhao, C., Hölter, R., & Schanz, T. (2018).
Properties of Fibrous and Amorphous Peats for the Settlement prediction for an embankment on soft
Construction of Road Embankments. Journal of clay. Computers and Geotechnics, 93, 87–103.
Materials in Civil Engineering, 30(7), 1–12.
Nithurshan, M., Nitharshan, R., & Nasvi, M. C. M. (2019).
Feng, R., Wang, L., Wei, K., & Zhao, J. (2021). Compressibility Behaviour of Peat Stabilized with Fly
Consolidation settlement of soil foundations Ash Using DMM: An Experimental and Numerical
containing organic matters subjected to Study. International Journal of Trend in Scientific
embankment load. Geomechanics and Engineering, Research and Development, Volume-3(Issue-3),
24(1), 43–55. 432–438.

Grimstad, G., & Nordal, S. (2018). On the modelling of soft Pihlainen, J. A. (1963). A review of muskeg and its
clay. Aktuelle forschung in der bodenmechanik 2018: associated engineering problems.
tagungsband zur 3. Deutschen Bodenmechanik
Pineda, J. A., Suwal, L. P., Kelly, R. B., Bates, L., & Sloan,
S. W. (2016). Characterisation of Ballina clay.
Geotechnique, 66(7), 556–577.

Roscoe, K., & Burland, J. B. (1968). On the generalized

stress-strain behaviour of wet clay.

Schanz, T. (1999). Formulation and verification of the

Hardening-Soil Model. RBJ Brinkgreve, Beyond
2000 in Computational Geotechnics, 281–290.

Tschuchnigg, F., & Schweiger, H. F. (2018). Embankment

prediction and back analysis by means of 2D and 3D
finite element analyses.

Tyurin, D. A., & Nevzorov, A. L. (2017). Numerical

simulation of long-term peat settlement under the
sand embankment. Procedia Engineering, 175, 51–
56.

Vermeer, P. A., Stolle, D. F. E., & Bonnier, P. G. (1998).

From the classical theory of secondary compression
to modern creep analysis. Proc. Computer Methods
and Advances in Geomechanics, 2469–2478.

Voznesensky, E. A., & Nordal, S. (1999). Dynamic

instability of clays: an energy approach. Soil
Dynamics and Earthquake Engineering, 18(2), 125–
133.

Wong, L. S., & Somanathan, S. (2019). Analytical and

Numerical Modelling of One-Dimensional
Consolidation of Stabilized Peat. Civil Engineering
Journal, 5(2), 398.
Wednesday, October 4, 2023

TAILINGS AND ROCK

Decommissioning the RM3 Drain at the
Selbaie Mines Reclaimed Tailings Facility: A
Field Trial Study
Anne-Marie Dagenais
AMD Solutions, Île Bizard, Quebec, Canada (Formely with WSP
Canada Inc. & EOR at Selbaie Mines)
Michaël Demers-Bonin
WSP Canada Inc., Montreal, Quebec, Canada
Ashley Gusikoski
BHP – Legacy Assets, Saskatoon, Saskatchewan, Canada
Vincent Martin
Newmont – Mine Éléonore (Formely with WSP Canada Inc.)

ABSTRACT
The former Selbaie mines produced 45 million tonnes of acid-generating tailings stored in the tailings management area
(TMA). Reclamation started in 2000 and was mostly finished by 2006. The site is now in active care and maintenance.
Reclamation involved placing a 1 m thick till cover over most of the tailings, promoting clean runoff and maintaining the
phreatic surface within the TMA. The cover slopes towards the northern end towards the Retention Pond and the spillway
in Dam RM3.
During Dam RM3 spillway construction, a rock drain and sumps called RM3 drain were built. The system collects water,
which flows to a manhole and discharges through an outlet pipe in the spillway wall to control water levels in the acid-
generating waste rock shell, preventing contamination of the Retention Pond.
A field trial evaluated decommissioning the drain to simplify water management, improve dam safety, and manage risks
associated with the pipe in the spillway. The trial assessed dam stability, flow regime in the Retention Pond area and TMA
dams, and water quality.

RÉSUMÉ
Les anciennes mines Selbaie ont produit 45 millions de tonnes de résidus générateurs d'acide, entreposés dans le parc à
résidus (PÀR). Les travaux de remise en état ont commencé en 2000 et se sont terminés en 2006. Le site est aujourd'hui
sous surveillance active. La remise en état a consisté à placer une couverture de till de 1 m d'épaisseur sur la plupart des
résidus, afin de favoriser un écoulement de surface et de maintenir la nappe phréatique à l'intérieur de ces derniers. Le
recouvrement favorise un écoulement vers le bassin de rétention à l'extrémité nord puis vers un déversoir dans la digue
RM3.
Lors de la construction du déversoir de la digue RM3, un drain rocheux et des puisards appelés Drain RM3 ont été
construits. Le système recueille l'eau, qui s'écoule vers un trou d'homme et se déverse par un tuyau de sortie dans le mur
du déversoir pour contrôler les niveaux d'eau dans l’enrochement amont de stériles générateurs d'acide, empêchant ainsi
la contamination du bassin de rétention.
Un essai sur le terrain a permis d'évaluer l’impact de la mise hors service du drain ce qui permettrait de simplifier la gestion
de l'eau, d'améliorer la sécurité de la digue et de gérer les risques associés à la conduite sous le déversoir. L'essai a
permis d'évaluer la stabilité de la digue, le régime d'écoulement dans la zone du bassin de rétention et dans les digues du
PÀR, ainsi que la qualité de l'eau.

1 INTRODUCTION proportion of tailings was used as underground backfill.

Ponds and channels were built to collect and manage
The Selbaie mine site (Selbaie) was in operation between contact water from the stockpiles and the TMA. There are
1981 and 2004. It extracted approximately 56.5 million 19 dams and 8 ponds at Selbaie and the site is in active
tonnes of copper, zinc, gold and silver ore, from care and maintenance (Figure 1).
underground and open pit mines. The TMA is delimited by eight dams built and raised
Operations generated wastes including: waste rock, over the years according to the deposition and water
pyritic waste rock, mine tailings, and overburden. Waste management plan. Tailings deposition started in the south
rock (34 million tonnes) was accumulated in the waste, on higher ground and progressed towards the west and
pyrite, and soil stockpiles, and was also used as dam north, using internal dykes made most likely of waste rock.
construction material. As for tailings, it is estimated that 45 The TMA is divided into two sectors by internal dams, with
million tonnes were stored in the tailings management area the sector to the north lower by several meters. The tailings
(TMA), occupying an approximate area of 180 ha. A small
water pond was in the northern sector towards the end of safety context to aid in making an informed decision
operation. regarding decommissioning of the drain.
Reclamation of the site began in 2000 and was largely
complete by 2006, with the reclamation of the TMA 2 BACKGROUND INFORMATION
completed by 2005 (SNC-Lavalin, 2006), at which time
Selbaie entered the active closure care and maintenance A first step of the trial process was to develop a field
phase. Current operations at the site involve monitoring, procedure based on current monitoring and field
inspections, surveillance, maintenance, water observations, as well as data on drain configuration, and
management, and water treatment. anticipated response of key parameters in relation to dam
Reclamation of the TMA involved placement of an safety and water quality during the trial. A review of
approximately 1 m thick cover of till over most of the tailings construction reports and monitoring data was carried out to
area. The objective of the till cover was to promote clean prepare a detailed procedure, which included the trial
runoff and maintain an elevated phreatic surface within the methodology and equipment, a monitoring and
tailings. The cover surface is inclined towards the northern surveillance program and an emergency response plan.
end of the facility where a pond, referred to as the Thresholds were also derived based on the review of
Retention Pond, is located. The pond acts as a water cover monitoring data available and supporting analyses. The
over the tailings in that sector of the TMA. Surface runoff is sections below give an overview of the data reviewed and
conveyed from the southern sector to the northern sector analyses completed to develop the field trial procedure.
via a broad-crested spillway built in the internal dam. Water
in the northern sector can be discharged to the 2.1 RM3 Drain Construction
environment or circulated through the water management
and treatment system, depending on its quality. According to the as-built report on reclamation works at
Dam RM3 is located north of the TMA. It contains the Selbaie (SNC-Lavalin, 2006), the dam upstream shell,
tailings and the Retention Pond. During upgrades to the made of rockfill, appeared to be conveying water from the
RM3 spillway in 2004, a rock drain and sumps, referred to Retention Pond and the tailings area during the
as the RM3 Drain, was built. It was upgraded in 2005 to its construction of the spillway in 2004 and at the time the
current configuration, The water collected by the RM3 reclamation work was undertaken. The water level in the
Drain flows to a manhole, from where it discharges into an upstream shell of Dam RM3 had to be lowered, using
outlet pipe, which passes through the concrete wall of the temporary pumping wells excavated upstream of the
RM3 spillway. The intent at the time of the RM3 Drain spillway, to prevent water inflow in the excavation made for
construction was to control the water level in the acid- the construction of the spillway. Following completion of the
generating waste rock shell of the dam and thus prevent spillway, a rock drain was built directly upstream to control
potential contamination of the Retention Pond directly the water level in the acid-generating waste rock shell and
upstream of the dam. thus prevent potential contamination of the Retention Pond
In 2016, BHP initiated a review of the Selbaie closure (SNC-Lavalin, 2006). The RM3 drain, composed of waste
strategy, including assessing the RM3 drain to understand rock, is approximatively 40 m long by 2 m wide and 1 m
its role and influence on the TMA hydrogeological behavior. high (Figure 2).
Opportunities to decommission the drain and potential dam The drain is connected to two sumps. Water
safety implications were also discussed at the time. collected in the sumps flows via two 200 mm diameter
In 2020, a field trial procedure was prepared to gain perforated HDPE pipes to a manhole in which a drop-inlet
insight on the feasibility of decommissioning the drain, mechanism is installed. It then leaves the manhole via a
which could simplify water management and improve dam 300 mm diameter riser pipe reducing to 200 mm (Figure 3).
safety by eliminating risks associated with the presence of The manhole outlet pipe passes through the RM3
a pipe through a dam. spillway wall. The drain elevation is 3.6 m below the
The objective of the field trial was to evaluate the impact concrete spillway invert. On the downstream side of the
of closing the drain on dam stability, flow regime in the RM3 spillway, the outlet pipe connects to a Y-valve that
Retention Pond area and TMA dams, and water quality in allows discharging to the Recirculation Pond at the toe of
the Retention Pond. This paper presents the field trial Dam RM3 or the Floodplain at the toe of Dam RM5CW.
procedure as well as the conclusions and perspectives The estimated discharge rate from the drain is around
gained from the trial. It also details the proposed path 540,000 m3/year (Wood, 2020).
forward to improve the understanding of the role of the
drain within the closure strategy as well as within a dam
Figure 1. Selbaie Mines Site Layout (Golder, 2021)

Figure 2. RM3 drain Plan View (adapted from SNC-Lavalin, 2006)

 From a stability point of view, the latter would represent
a worst-case scenario during the field trial. Considering the
uncertainty on the state of the tailings (loose and
contractive, or dense and dilative) and the cohesive soil
units (normally consolidated or over-consolidated), both
effective stress analyses (ESA) and undrained strength
analyses (USA) methods were used to evaluate the static
stability, considering prudent assumptions for geotechnical
parameters, based on historical field tests. The authors
recognize that the field trial is essentially an analogue of
rising pore pressures at constant deviator stress. Should
the tailings be identified as contractive, their stress states
relative to the instability line would govern their
susceptibility to static liquefaction. Then, USA stability
analyses were carried out using both static peak and post-
liquefaction shear strength parameters for tailings
combined with an elevated pond water level.
The cross-sections selected for the stability analyses
are located on Dam RM3 along its highest elevation and
through the spillway, and on Dam RM5CW. These cross-
sections were chosen since it is believed that they
Figure 3. Sketch of the manhole with the riser pipe during represent the most sensitive areas during the RM3 drain
construction (adapted from SNC-Lavalin, 2006) test.
When considering the results for the ESA method, the
2.2 Water management and operational water levels factor of safety (FoS) evaluated is above 1.3 for both pond
water level conditions. It is noteworthy that the FoS criteria
The Retention Pond collects surface runoff from the TMA used at the time of design was 1.3 and dams at the site are
cover over a catchment area of 173 ha. It provides flow currently under assessment to verify compliance with the
attenuation of surface runoff from the TMA and short-term Global Industry Standard on Tailings Management
storage prior to discharge in the spring. The Retention (GISTM; ICMM, 2020). Analyses carried out according to
Pond also generally acts as a local water recharge into the the USA method results in FoS above 1.3, except for the
tailings. scenarios with a high water table at Dam RM3 and Dam
The water level in the Retention Pond can be controlled RM5CW, in which case the FoS was 1.2. Similar results
by siphons or gravity flow through the spillway where it were obtained for post-liquefaction conditions. The
discharges to the environment. Water from the Retention similarity between FoS for different conditions is due to the
Pond can be discharged to the environment through the critical sliding surface located mostly within downstream
RM3 spillway for a short period of time during spring freshet dilative structural components and the native cohesive unit
when it is compliant with regulatory requirements (Rio foundation. The sliding surface does not involve the
Algom, 2017). Pond operating water levels vary between tailings. Considering the duration of the trial, the potential
elevations 277.9 m and 279.0 m, which is the maximum rise of the phreatic surface and pond water level was
elevation of the RM3 spillway stop-log structure above the deemed to have little effect on the overall stability for Dams
concrete spillway invert. RM3 and RM5CW. Thresholds were defined for monitoring
The elevation of the riser pipe intake is assumed to be of water levels during the trial.
at 276.3 m. Hence, a hydraulic gradient is created from the
Retention Pond towards the RM3 drain. Shutting off the 2.4 Water Quality
RM3 drain could potentially translate into a rise of the
phreatic surface in the dam core and shell as well as water Water quality data were reviewed to compare water quality
inflow into the Retention Pond and TMA surface which at the different sampling points, identify parameters to
could potentially affect dam stability and water quality in the monitor during the field trial and define thresholds for these
Retention Pond. parameters. The locations chosen were the Retention
Pond, the RM3 drain, the tailings area, Dam RM3 and Dam
2.3 Physical Stability RM4. More specifically, sampling points reviewed were
located at:
The effect of a potential rise of the phreatic surface and • RM3 drain at pipe: Chemistry is expected to
pond water level was evaluated on both overall dam represent the water signature associated to pore
stability and pore water pressure in tailings. Analyses were water from waste rock in the upstream dam shell
carried out considering two scenarios: and tailings.
• Retention Pond water level at normal operating • Retention Pond: Chemistry is expected to
water level. represent water signature associated to
• Retention Pond water level at maximum operating precipitation and surface runoff.
water level controlled by the invert of the spillway • Tailings management area: Groundwater wells
stop-log structure. installed in the tailings at different locations in the
 TMA (Figure 1). Chemistry represents water potential other water sources to the flow measured in the
signature associated to tailings pore water. RM3 drain. The following scenarios represent the two end-
• Dam core and upstream shell: Groundwater wells points, with regards to the contribution of the Retention
installed in the core of Dam RM4 and the lower Pond to the flow in the drain and describe potential
portion of Dam RM3 upstream shell. Chemistry is response in water quality and piezometric surface:
expected to be influenced by tailings and waste • If the Retention Pond is the major contributor to
rock pore water signature. the flow measured in the RM3 drain, the phreatic
Water samples at these locations were analyzed for surface in the dam would eventually reach a level
general chemistry parameters, major anions and dissolved comparable the water level in the pond and there
metals. A preliminary evaluation of water chemistry data would be little to no gradient between the two.
from the observation wells suggests that water sampled in Thus, little to no change in water quality in the
the RM3 drain is chemically closer to contact water quality Retention Pond would be expected.
from within Dam RM3 shell than tailings pore water. • If drainage from the TMA is the major contributor
This trend is observed in the relationships between to the flow measured in the RM3 drain, a rise of
parameters such as conductivity, sulfates, alkalinity, the phreatic surface in the dam could eventually
copper, calcium, magnesium and zinc, where RM3 drain cause an inversion of the gradient between the
water quality has the low alkalinity, elevated conductivity Retention Pond and the RM3 drain.
and sulfate concentration of waste rock, as opposed to the Considering these potential scenarios, the duration of
higher alkalinity and lower sulphate concentration of the trial has been limited to four weeks to minimize
tailings. These preliminary observations suggest that water potential impact to the water quality in the Retention Pond.
entering the RM3 drain is a mixture of Retention Pond (low The trial procedure included water level monitoring to
conductivity and sulfate concentration) acquiring the observe the extent of the effect of closing the RM3 drain on
chemical load (sulfates and metals) of waste rock the phreatic surface in the TMA and surrounding dams, and
surrounding the drain and/or of inflow of water flowing water level in the Retention Pond. Conductivity
through the dam waste rock shell. measurements was also included at strategic locations
The trends observed between sulfates and electrical throughout the trial based on the relationship between
conductivity suggested that electrical conductivity could be sulfates and conductivity with the objective of limiting the
used as a monitoring parameter during the field trial to effect on the water quality of the Retention Pond Trigger
evaluate the effect of RM3 drain closure on water quality in levels are defined in Section 4.0 The test methodology is
the Retention Pond. presented in Section 4.0. Monitoring frequency was
modulated by field trial phase and anticipated effect at the
monitoring points.
3 BASIS FOR FIELD TRIAL PROCEDURE

Water chemistry data reviewed suggests that water in the 4 TEST METHODOLOGY
RM3 drain appears to originate from waste rock pore water
flowing through the upstream dam shell. Upstream shells The field trial closure of the Dam RM3 drain was carried out
of Dams RM3, RM5CW, RM5CE, and RM5B, delimiting the between 13 June 2021 and 25 August 2021. The drain
northern sector of the TMA, are assumed to be connected outlet valves were closed on June 29th. The RM3 drain
allowing water to flow through to the drain. This connection was reopened a month later on July 28th, and monitoring
is assumed to be truncated at Dam RM4 and Dam RM5A was continued until August 25th. The following steps
by the cores of Dams RM7 and RM8 connecting to the summarize the methodology according to the procedure
cores of the perimeter dams (Figure 1). described in Golder (2021):
Also worth mentioning, the proximity of the Retention
Pond and configuration of the area suggests a possible 4.1 Phase I – Baseline period
hydraulic connection between the upstream shell of Dam
RM3 and the Retention Pond. Even if there is a potential Phase I occurred between 13 and 29 June 2021. At the
hydraulic connection, the water quality in the Retention start of the field trial, pressure transducers (TD Diver, Van
Pond appears to be unaffected by the water in the Essen Instruments) were installed in selected wells and in
upstream dam shell. This suggests that, under current the manhole of RM3 drain. These wells were selected for
conditions, the Retention Pond generally maintains a their proximity to the manhole, anticipated reaction time
hydraulic gradient towards the upstream shell throughout and sensitive location with regards to dam stability. Manual
the year. measurements with a water-level meter were also taken
Upon closure of the RM3 drain, a rise of the phreatic periodically in wells further away from the drain outlet, in
surface in the upstream shell and dam core of Dam RM3 is the Retention Pond and Recirculation Pond.
expected. The extent of this rise will depend in part on the Water quality samples were collected once during the
duration of the trial, the hydraulic properties of the tailings baseline period at selected wells, in the Manhole, the
and the materials composing the dam, and on the potential Recirculation Pond and the Retention Pond.
sources of water flowing to the drain area. The response in A cursory visual inspection of Dams RM3 and RM5CW
terms of water quality and position of the piezometric and northern sector of the TMA was completed to get an
surface at the different monitoring locations will also overview of the initial site conditions and to identify existing
depend on the contribution of the Retention Pond and
features (e.g., boils, exfiltrations, surface erosion) prior to [DBM1]in suspension were observed along the Retention Pond
closing the drain. beach, near Dam RM3 on its upstream side (Figure 4). The
The Retention Pond was lowered below normal boils were presumed to be caused by an increase in the
operation water levels by pumping to better observe the hydraulic gradient from Dam RM3 toward the Retention
effect of closing the drain on water-level readings and Pond caused by a rise in the water pressure head in the
water quality, over a short period of time. dam core and upstream shell combined to the low water
level in the pond.
4.2 Phase II – Closing the drain

On 29 June 2021, the drain outlet valves were closed. An

inspection of the valves, couplings and exposed sections
of the pipeline was then completed.

4.3 Phase III – Field trial

Phase III occurred between June 29th and July 28th.

Monitoring and surveillance was done daily according to
the monitoring and surveillance program. The daily routine
of the field personnel was:
• Visual inspection of the pipeline between the
manhole and the valves for signs of leakage,
turbid exfiltration or surface erosion.
• Visual inspection of Dams RM3 and RM5CW for
signs of new seepage points or changes in
existing seepage points, boils, presence of Figure 4. Example of boils observed upstream of RM3 Dam
particles in seepage or boils, discoloration in the in the Retention Pond.
pond water and signs of instability.
• Water quality sampling, monitoring of physico-
chemical parameters and/or water level of 6 RESULTS
selected wells, depending on monitoring
schedule, and verification against trigger levels. The following subsections present results of the water-level
• Reporting, including daily inspection forms, daily measurements, the physico-chemical parameters
reports including general view photos of the monitoring and the water quality sampling carried out
upstream and downstream slope and specific between June 14th and August 25th. The results are
observations. presented by measured parameters:
• Meetings were held twice a week between DMI, • Section 6.1 – Water level;
the site operator, BHP and Golder (now WSP) to • Section 6.2 – Water physico-chemical
discuss status of the trial, address any safety parameters;
concerns raised during the test, review the • Section 6.3 – Water quality.
preliminary results and to discuss action plans in Total precipitation data measured at the Selbaie weather
the event that a warning level was reached. station was reviewed for comparison purposes to the water
level and pH data. The 2021 Summer at Selbaie Mines was
4.4 Phase IV – Reopening the drain relatively dry, with no significant storm events or long
duration precipitation events. This created favorable
On July 28th, the valve was reopened to discharge water conditions for the RM3 drain field trial, since changes in
in the Recirculation Pond. water levels in wells or in the Retention Pond could more
easily be attributed to the closure or reopening of the outlet
4.5 Phase V – Return to baseline valves of the RM3 drain.
Phase V occurred between July 28th and August 25th. 6.1 Water levels
During this phase, field personnel continued to monitor
water level and physico-chemical parameters until water Water levels were monitored to observe the extent of the
levels were close to pre-trial levels. Water quality sampling effect of closing the RM3 drain on the TMA and TMA dams
and inspection of Dam RM3 and RM5CW were carried out phreatic surface and water level in the Retention Pond and
according to monitoring and surveillance schedule. The the Recirculation Pond. The main observations which
field trial was completed on 25 August 2021. arose from the trial are:
• The change in the water levels in the wells in the
dams upstream rock shell was observed within
5 FIELD OBSERVATIONS
hours of closing and reopening the drain. The
change in water levels at wells located in the dam
Dams RM3 and RM5CW and the RM3 spillway remained
foundations and the tailings within the TMA
in good condition during the test. Boils with fine particles
appeared attenuated and may have been
 influenced by external elements such as dam upstream shell and core, RM3 drain, as well as the
precipitation or the surrounding ponds. Retention Pond and Recirculation Pond.
• The field trial highlighted the extent of the zone of A Piper diagram presented in Figure 6 represents
influence of the RM3 drain on the water level in graphically the water chemistry of the samples taken in the
the TMA dams rock shells (Figure 5). It reaches selected wells and ponds. The following observations can
the southern sector of the TMA, with water-level be made on the chemistry measurements:
changes less than 24 hours after closing or • Water chemistry data from samples taken in the
reopening the drain in wells located roughly 1 km manhole and the wells located in the dam rock
from the drain (e.g., Dam RM5A – PU-20; Dam shell and dam core is similar, with low alkalinity
RM4 – PO-06-04). This indicates that there are and elevated sulphate concentration.
some hydraulic connections between the TMA • Water chemistry data from samples taken in the
and dams RM3, RM4 and RM5A. Whether the wells located in the tailings show higher alkalinity
connections occur within the upstream rock shell and lower sulphate concentration than the
or some rock structure passing below the core of samples collected from the manhole.
dams RM7 and RM8 is unknown. • Water chemistry data from samples taken in the
• During the test, the water level in the Retention Retention Pond show lower alkalinity and lower
Pond was lower than in dams RM3, RM5CW, sulphate concentration than the samples
RM5CE and RM5B (both rock shells and cores). collected from the manhole.
The hydraulic gradient direction in the TMA for the • Sample water chemistry from the Retention Pond
duration of the test was from Dams RM3 and show a decrease in relative alkalinity after the test
RM5CW toward the Retention Pond; this is (Phase V).
supported by the water levels monitored in that
area and the appearance of boils along the
northern beaches of the Retention Pond.

Figure 6. Water chemistry data for all samples taken

7 CLOSING REMARKS

A field trial to inform on the feasibility of decommissioning

Figure 5. Evolution of the water level in the dam rock shells the RM3 drain permanently was conducted between 13
and daily precipitation data. June 2021 and 25 August 2021. The objective of the trial
was to evaluate the impact of closing the RM3 drain on dam
6.2 Water physico-chemical parameters stability, flow regime in the Retention Pond area including
the dams, and water quality within the Retention Pond by
No discernible trend was noted with regards to pH and way of monitoring water levels and key parameters within
electrical conductivity measurements during trial. The pH the TMA dams and in the tailings.
values measured in the manhole and all the wells remained The test procedure included five phases, each with its
lower than the pH in the Retention Pond, while the own set of monitoring, surveillance, and sampling
conductivity values remained higher for the duration of the requirements. A baseline for the trial was established
field trial. during a 2-week period, from June 13th to June 29th, before
closing the valves at the outlet of the drain to provide a
6.3 Water quality reference point. Following the baseline phase, the RM3
drain was closed for four weeks, from 29 June to 28 July
Water chemistry was analysed during the field trial mainly 2021. The last phase took place after reopening the drain
to investigate the origin of the water entering the drain. from July 28th to August 25th, and was essentially a period
Evolution of water chemistry during the trial was also of monitoring the return to baseline conditions.
looked at to evaluate the effect of closing the RM3 drain. Water levels, pH, electrical conductivity, and water
Samples were collected before, during and after the trial at quality were monitored at selected wells in the TMA and
selected wells representing tailings pore water chemistry, along the dams during the field trial; measurements were
modulated by trial phase and anticipated effect at follow to better understand the impact of closing the drain
monitoring point. from a dam safety, environmental and operational
Based on the observations above, the following standpoint.
conclusions and outcome stem from the trial: A longer field trial (over five-month period at a
• Closing the drain will affect the position of the minimum, and maybe up to one year to cover all 4 seasons)
phreatic surface in dams and TMA, and by extent would benefit the understanding of TMA hydrogeological
dam stability through a potential rise in pore water conditions and help validate the site water balance, and the
pressure and decrease in effective stress in dam hydrogeological and contaminant transport models for
materials and foundations. Stability should be current versus potential future conditions. Monitoring
reviewed to consider potential change in phreatic changes over a longer period would inform on conditions
conditions and assess the need for mitigation at hydraulic equilibrium, time to reach equilibrium,
measures and their extent. sensitivity to precipitation events and effect on water
• Drainage from the TMA through the upstream balance and cover performance.
rock shells of the dams appears to be the major The RM3 drain field trial was a valuable learning
contributor to the flow in the RM3 drain. The till exercise for the teams involved in the management of
layer onto the dam upstream rock shell limits flow Selbaie mine. The trial highlighted the challenges in
between the Retention Pond and the rock shell. managing complex reclaimed facilities with passive closure
• Although not directly observed during the trial, as a long-term objective. Thus, long-term passive closure
closing the drain may affect water quality in the must be examined from a global site standpoint since the
Retention Pond based on field observation of individual components of a TMA have an impact on the
boils and hydraulic gradient between the rock overall performance of a site.
shell and Retention Pond during trial.
Decommissioning of the Retention Pond by
backfilling should be envisioned in the future to 8 ACKNOWLEDGEMENTS
prevent further degradation of pond water quality
and water treatment if drain is decommissioned. The authors want to thank the staff at the Selbaie Mine from
• With regards to site operations and reclamation, BHP and Denison Environmental Services that made this
closing the RM3 drain will impact the position of work possible.
the phreatic surface within the TMA. This may be
beneficial to the current closure strategy, which is
based on maintaining a high degree of saturation 9 REFERENCES
in the tailings. This may cause an increase in
exfiltration rates and potentially impact Golder Associés Ltée (Golder). 2016. Instrumentation
groundwater quality and water quality in the program – Installation Report. N/Ref.
peripheral ditches. 002-1538775-RA-Rev0. April 2016.
• Overall water volumes to treat may also change, Golder. 2021. Field Trial Detailed Action Plan for
which might impact water chemistry. This could Decommissioning the RM3 drain – Selbaie Mines.
influence the quantity of lime consumption and/or N/Ref. 031-1897974-19002-Rev1. June 2021.
the water treatment process during the transition Golder. 2022. RM3 drain Decommissioning Field Trial –
period. Selbaie Mines. N/Ref. 002-21452482-RA-Rev0. June
2022.
A clear understanding of TMA hydrological and International Council on Mining and Metals (ICMM). 2020.
hydrogeological conditions is needed before a plan for Global Industry Standard on Tailings Management.
decommissioning the drain can be implemented given August 2020.
potential impacts to dam stability, site operations and the Rio Algom. 2017. Selbaie Mine Water Management
long-term closure strategy. Therefore, decommissioning of Operating Strategy and Plan. Internal report.
the RM3 drain should be part of a broader reflection on the SNC-Lavalin. 2006. Supervision des travaux de
closure strategy for Selbaie mine to enable informed restauration pour l’année 2004. Volume 1. N/Ref.
decisions for the path forward. M-6826 (603692). Volume 2 and 3. N/Ref. M-6742
In follow up to the field trial, a staged approach was (603692).
proposed to provide necessary information to support an Wood. 2020. Selbaie Site Characterization Report. N/Ref.
informed decision with respect to decommissioning the TE183027. 17 January 2020
RM3 drain. The first stage, data acquisition, would feed the
second stage which is the preparation of the main studies
needed to understand how the drain closure may affect
overall long term site conditions and operations, and
assess potential methods for decommissioning. The
outcome of the study stage would feed into a risk-benefit
analysis that would enable BHP to decide on the course of
action with a deeper understanding of the drain and its
effect on dam stability and how it is integrated into the long-
term closure strategy. A long-term field trial could then
Effect of tailings migration on hydraulic
conductivity of mine waste rock
Samy García-Torres, Florian Behlke & Carlos Ovalle
Department of Civil, Geological and Mining Engineering,
Polytechnique Montréal Research Institute of Mining and
Environment (RIME) UQAT-Polytechnique, Québec, Canada

Serge Ouellet
Canadian Malartic Mine, Agnico Eagle Mines, Québec, Canada

ABSTRACT
Significant amounts of waste rock and tailings (WR&T) are generated by the mining industry, representing a
challenge in terms of waste management. Co-placement of WR&T is considered an alternative with positive effects.
Nevertheless, countereffects related to the interaction between both materials could arise, particularly due to
migration of tailings fine particles inside WR pores. The objective of this work is to improve the understanding of
mine WR&T interaction in terms of hydro-geotechnical properties. A series of slurry tailings migration tests within
a large column of saturated WR were carried-out. The effect of WR grading was studied testing three different
materials with constant dmax=50 mm and different Cu. The results indicate that the retained mass of tailings along
the column height reached a maximum value of 20%, showing a homogeneous behaviour of tailings retention
through the entire column. The saturated hydraulic conductivity of the WR presents a maximum reduction factor of
13.

Keywords: Tailings, mine waste rock, waste co-disposal, hydro-geotechnical properties.

RÉSUME
Des quantités importantes de roches stériles et de résidus (RS&R) sont générés par l'industrie minière, ce qui
représente un défi en termes de gestion des déchets. Le co-placement de RS&R est considéré comme une
alternative avec des effets positifs. Néanmoins, des contre-effets liés à l'interaction entre les deux matériaux
pourraient survenir, notamment en raison de la migration de particules fines des résidus à l'intérieur des pores des
RS. L'objectif de ce travail est d'améliorer la compréhension de l'interaction des RS&R en termes de propriétés
hydro-géotechniques. Une série d'essais de migration de boues de résidus dans une grande colonne de RS
saturée a été effectuée. L'effet de la granulométrie des RS a été étudié en testant trois matériaux différents avec
un dmax=50 mm constant, et différents Cu. Les résultats indiquent que la masse retenue des résidus le long de la
hauteur de la colonne atteint une valeur maximal de 20%, montrant un comportement homogène de rétention des
résidus à travers la colonne. La conductivité hydraulique saturée des RS présente un facteur de réduction
maximale de 13.

Mots clés: Résidus miniers, stériles miniers, co-placement des déchets, propriétés hydro-géotechniques.

1 INTRODUCTION Managing WR&T together could be an alternative to

enhance the materials behaviour, taking advantage of (1)
Significant volume of waste rock (WR) and tailings the mechanical properties of WR to mitigate tailings
generated by mining operations are frequently placed in liquefaction, and (2) the hydraulic properties of tailings to
piles and poured in impoundments, respectively. The avoid ARD of WR (Jehring & Bareither, 2016; Wilson et
optimal placement and safe deposition of large waste al., 2003; Zhang et al., 2021). Co-disposal can be
amounts results challenging as the properties of both managed by simply mixing WR&T before deposition
materials could generate long-term pollution in the field. (also called co-mingling), or by placing the materials in
Saturated tailings exhibit low shear strength, are prone batches where WR and tailings are directly in contact and
to liquefaction, and consolidation may take several interact (also called co-placement).
years. On the other hand, WR remains unsaturated with An interesting case of WR&T co-placement is the use
relatively high void ratio and therefore great exposure to of waste rock inclusions (WRI) in tailings impoundments.
undergo acid rock drainage (ARD) (Olson et al.,2003; This method is used at Canadian Malartic Mine in
Bussière ,2007; Wilson, 2008). Quebec and consists in placing linear inclusions of WR
Due to the disadvantages of traditional waste rock inside the tailings impoundment, which is thus
treatments, improved techniques with high benefits in divided in several zones. WRIs act as drainage ways and
cost, safety and sustainability aspects are evaluated and
accelerate consolidation of tailings, contributing to
recommended nowadays (Gowan et al., 2010; Ulrich &
improve the physical stability (James et al., 2013;
Coffin, 2015; Wickland et al., 2010). For instance, co-
disposal of waste rock and tailings (WR&T) in the same Aubertin et al., 2019). However, WR&T interactions can
facility presents significant advantages and could result in significant tailings fine particles migration into
promote a better social acceptance (Wilson et al., 2003). the pores of the WR material. This could have an effect
on the hydraulic conductivity of WR, greatly affecting its

1
drainage properties. Essayad (2021) studied this topic Table 1. WR properties.
for relatively low hydraulic gradients and well graded WR
materials and showed that migration does not have a T1 T2 T3
significant effect on the hydraulic conductivity of WR. dmax (mm) 50 50 50
However, other co-placement solutions could be dmin (mm) 1.4 0.425 0.15
exposed to more extreme conditions. For instance, open
d50 (mm) 22 20 19
pit backfilling where WR&T are placed separately but
remain in contact. In this case, significant particle d60 (mm) 27 25.1 25
segregation of WR material upon in pit deposition could d10 (mm) 4.7 3 2.1
take place, and coarser WR could be exposed to eo 0.51 0.50 0.43
significant tailings migration. Moreover, due to steep Cu 6 9 12
slopes of open pits in hard rock mines, when backfilling
with slurry tailings the rate of height raise can be
extremely high (about 100 m per year according to 2.2 Physical modelling
Ouellet et al., 2021), generating very high hydraulic
gradient during drainage through WR&T interaction. A physical model was developed to reproduce the WR
Further understanding is required on the interactions behaviour influenced by tailings migration. The
and mixing of co-placed WR&T materials. The objective experimental set-up and non-standard procedure
of this work is to study the effects of tailings migration proposed by Essayad (2021) were considered (Figure
through WR material on mixing ratio and hydraulic 2a). A column of 300 mm in diameter and 900 mm in
conductivity. A series of large-scale laboratory column height was employed. The set-up includes perforated
tests were performed for a proper characterization of the aluminium plates placed at the top and bottom of the
material, considering different WR uniformity coefficient column, and a funnel located at the base for an efficient
(Cu) to take into account the effects of WR segregation material reception. Inlet and outlet valves allow slurry
during in pit deposition. injection from the top and slurry collection at the bottom
of the column, respectively. Pore pressure transducers
2 METHODOLOGY
(PPT) were placed along the column height, as shown in
Figure 2b. WR was placed inside the column in layers of
Three large-scale column tests were performed to
evaluate the hydraulic behaviour of WR after tailings 5.5 cm thick approximately, and compacted with 20 hits
migration, considering different particle size distributions per layer using a modified Proctor hammer. In order to
(PSD) of WR shown in Figure 1: T1, T2 and T3. Each accelerate the saturation process, CO2 was injected in
test has three essential stages on the same column of the WR sample before saturation with a flowrate of 2.0
WR material: (1) constant-head hydraulic conductivity L/min (Bouin et al., 2021). WR saturation was carried-
test on WR before tailings deposition, (2) tailing migration out over a period of two days and the saturation degree
test, and (3) constant-head hydraulic conductivity test was verified considering the control method of Chapuis
after tailings deposition and migration.
et al. (2007), which takes into account volume and mass
2.1 Material characterization variation of the specimen to calculate the degree of
saturation during the test. As an initial step, a first
WR and tailings materials obtained from Canadian validation of the column set-up was developed by
Malartic Mine (Quebec, Canada) were used for the performing a constant head permeability test using clean
experimental work. Tailings PSD is presented in Figure sand. A similar value of saturated hydraulic conductivity
1. Tailings samples were prepared considering a slurry (ksat) to that obtained from analytical methods (Chapuis
density of 63% of solids, which is the value used in the and Aubertin, 2003) was attained in this test.
field. WR relative density (Gs) is 2.78, and its main
characteristics for tests T1, T2 and T3 are summarized 2.3 Experimental procedure
in Table 1.
2.3.1 Stage 1: Hydraulic conductivity of WR

A constant-head permeability test considering a low

hydraulic gradient i=0.04 (ASTM D2434, 2006) was
performed as initial step to obtain the hydraulic
conductivity of the WR material. For each sample, 2
permeability tests were conducted, with a duration
between 2 to 6 minutes.

2.3.2 Stage 2: Tailings migration test

First, tailings followed a saturation procedure in a

separated deaeration column under applied vacuum,
before they were injected inside the column (Figure 2a).
The tailings column was connected to the top inlet valve
Figure 1. Particle size distribution of WR and tailings. of the WR column. For each test, 3 depositions of tailings

2
(D1, D2 and D3) were performed from the top of the 2.3.3 Stage 3: Hydraulic conductivity of WR after
saturated WR column, using a pump which supplies a tailings migration
pressure of 14 kPa and a flow rate 3 l/min. At the WR
column outlet valve, the outflow material was taken in A constant head hydraulic conductivity test was
intervals between 10 to 20 seconds per deposition during performed on the WR&T mixture after the migration test.
the entire time of the migration test. Tailings volume of The aim was to evaluate the variation of the WR
6.5 l were injected per deposition. Each deposition had a hydraulic conductivity due to tailings particles migration.
duration of 3 minutes approximately. Time and mass of First, the test was conducted considering a low hydraulic
the outflow were measured for a later analysis of the gradient i=0.04 (similar to Stage 1). Finally, four
tailings material. additional ascendant gradients of 0.05, 0.07, 0.11 and
0.20 were applied in order to replicate diverse deposition
a) scenarios.

3. RESULTS AND ANALYSIS

3.1 Migration of tailings fine particles in WR

During Stage 2 (migration tests), outflow samples were

taken from the column bottom outlet. The solid content
varied from the initial value of 63% to the minimum solid
content registered in the outflow of 3%, suggesting
significant material retention of fine material in the
column. During the first tailings injection D1, water within
WR pores was pushed-out of the column by the tailings
slurry. Therefore, low solid contents were registered in
the outflow. This similar trend was observed in tests T1,
T2 and T3 (see Figure 3). For the second and third
depositions of each test (D2 and D3), the initial solid
contents of the outflow present a peak value during the
first seconds of material collection, which is due to
settlement of previously deposited material. Higher solid
contents were registered over time for D3 outflows
compared to D2, as fine material accumulation inside the
column increased when consecutive depositions were
conducted.
b) The effect of tailings particle retention inside the WR
was investigated after dismantling the column in 10
layers. PSDs per layer along the column height were
analysed. The percentage of trapped tailings in the WR
pores was assessed through the ratio Vt/Vv %; where Vt
is the volume of solid tailings retained, and Vv is the
original void volume of WR. Solid tailings were identified
in each layer as the material finer than 315 µm. Figure 4
presents tailings entrapment registered per layer for the
different tests. The results indicate that for T2 and T3 the
first top layers of the column contain more trapped
material compared to the lower layers. In T2, at height of
49 cm, 17.5% of tailings volume was retained while 12%
was registered at height of 10 cm. Similar pattern was
observed in T3, in which the top layer contained 16% of
entrapment. In T1, the minimum and maximum
percentages of trapped tailings along the column height
were 8% and 12%, respectively, obtained at heights of
72cm and 32 cm, indicating a more uniform migration
compared to the better-graded samples. As expected,
the results indicate that lower entrapment along the
column is obtained in more uniformly graded WR due to
the greater volume of pores. This tendency of more
trapped tailings for higher Cu was also observed from
PSD curves of the dismantled column.
Figure 2. a) Experimental set-up of tailings migration
tests and b) WR column.

3
3.1 Influence of tailings migration on the hydraulic
conductivity of WR&T

As shown in Figure 5, ksat values obtained for WR at

low hydraulic gradient of i=0.04 in Stage 1 suggest an
increment of permeability at lower uniformity coefficient.
Results obtained during Stage 3 indicate a
considerable effect of the trapped tailings in the WR ksat,
particularly when subjected to lower hydraulic gradients.
Figure 6 presents ksat reduction factors of WR calculated
through the ratio of ksat-0 / ksat-f; where ksat-0 is the initial
WR hydraulic conductivity and ksat-f is the one obtained
after tailings migration. A higher reduction factor of 13
was obtained for T1, while reductions of 4.5 and 1.6 were
observed for T2 and T3, respectively.

Figure 4. Tailings entrapment along the column height.

Figure 5. WR initial hydraulic conductivity ksat at low

hydraulic gradient.
Figure 3. Tailings solid content in outflow samples
during Stage 3 (migration test).

4
 when better-graded samples were analysed. Moreover,
tailings migration considerably influenced the hydraulic
conductivity of WR depending on the Cu coefficient, and
ksat reduction factor was higher in WR having lower Cu.
Additional experimental work is essential to set
maximum ksat reduction factors as expected in the field,
where larger WR particles and significant segregation
could occur.

5 REFERENCES

ASTM (2006). Standard Test Method for Permeability of

Granular Soils (Constant Head) (D2434), ASTM
International, West Conshohocken, PA.
Aubertin, M., Jahanbakhshzadeh, A. and Yniesta, S.
(2019). The effect of waste rock inclusions on the
seismic stability of a tailings impoundment. 7th
Figure 6. Ksat reduction factor.
International Conference on Earthquake
Figure 7 shows a rising trend for ksat with an increase in Geotechnical Engineering ICEGE2019, Roma, Italy.
Bouin, C., Weber, S., Ethier, Y. A., Dubé, J. S., and
hydraulic gradients for the three WR samples. ksat shows
Duhaime, F. (2021). On preferred saturation methods
higher values for the different gradients evaluated for
for geotechnical flow tests. Geotechnical Testing
sample T1, due to its lower Cu=6. Values of ksat =0.054 Journal, 44(4), 1135-1152.
cm/s and ksat =0.040 cm/s were obtained for i=0.04 in T2 Bussiere, B. (2007). Hydrogeotechnical properties of
and T3, respectively. Moreover, a lower variation of ksat hard rock tailings from metal mines and emerging
per hydraulic gradient increment was observed in T1 geoenvironmental disposal approaches. Canadian
before it reached ksat=0.33 cm/s for i=0.2 (Figure 7). A Geotechnical Journal, 44(9), 1019-1052.
potential movement of tailings particles inside the column Chapuis, R.P. and Aubertin, M. (2003). On the use of the
eventually takes place at high gradients, resulting in Kozeny–Carman equation to predict the hydraulic
similar ksat for the 3 tests for i=0.2. conductivity of soils. Canadian Geotechnical Journal,
40: 616-628.
Chapuis, R.P., Masse, I., Madinier, B. and Aubertin, M.
(2007). A drainage column test for determining
unsaturated properties of coarse materials.
Geotechnical testing journal, 30(2).
Essayad, K. (2021). Évaluation multi-échelle de
l’instabilité interne et de la migration des résidus à
travers les inclusions de roches stériles. Ph.D. thesis,
Polytechnique Montréal.
Gowan, M., Lee, M., and Williams, D. J. (2010). Co-
disposal techniques that may mitigate risks
associated with storage and management of
potentially acid generating wastes. In Mine Waste
2010: Proceedings of the First International Seminar
on the Reduction of Risk in the Management of
Tailings and Mine Waste (pp. 389-404). Australian
Centre for Geomechanics.
James, M., Aubertin, M., and Bussière, B. (2013). On the
use of waste rock inclusions to improve the
performance of tailings impoundments. Proceedings
Figure 7. Hydraulic conductivity of the WR&T mixture at of the 18th international conference soil mechanics
different gradients. and geotechnical engineering, Paris, France.
Jehring, M. M. and Bareither, C. A. (2016). Tailings
4 FINAL REMARKS composition effects on shear strength behavior of co-
mixed mine waste rock and tailings. Acta
A series of laboratory tests were performed to analyse Geotechnica, 11(5), 1147-1166.
the effect of tailings particles migration in WR samples Olson, G. J., Clark, T. R., Mudder, T. I. and Logsdon, M.
with different Cu. The analyses were carried-out in 3 (2003). A novel approach for control and prevention
stages for more precise understanding: (1) constant- of acid rock drainage. Proceedings of the 6th
head hydraulic conductivity test on saturated WR, (2) International Conference on Acid Rock Drainage,
tailing migration test, and (3) constant-head hydraulic Cairns, Queensland, Australia.
conductivity test after tailings deposition and migration. Ouellet, S., Chapuis, S., Ovalle, C. (2021). Le projet de
The results show that the lower the WR Cu, the lower codéposition dans la fosse Canadian Malartic.
Symposium Rouyn-Noranda 2021 sur
the tailings entrapment and the higher the density of the
l’environnement et les mines. Université du Québec
outflow slurry. A greater retention of tailings inside the
en Abitibi-Témiscamingue (UQAT).
WR was more evident at higher layers of the column

5
Ulrich, B. and Coffin, J. (2015). Combined tailings and
mine waste. In Proceedings of the Tailings and Mine
Waste, Vancouver, BC, Canada.
Wickland, B. E., Wilson, G. W. and Wijewickreme, D.
(2010). Hydraulic conductivity and consolidation
response of mixtures of mine waste rock and tailings.
Canadian Geotechnical Journal, 47(4), 472-485.
Wilson, G. W., Plewes, H. D., Williams, D. and
Robertson, J. (2003). Concepts for co-mixing of
tailings and waste rock. Proceedings of the 6th
International Conference on Acid Rock Drainage,
Cairns, Queensland, Australia.
Wilson, G. W. (2008). Why are we still struggling with
acid rock drainage. Geotechnical News, 26(2), 51-56.
Zhang, C., Ma, C., Chen, Q., Liu, H., Wu, S., Pan, Z. and
Zhang, L. (2021). Influence of rock percentage on
strength and permeability of tailing-waste rock
mixtures. Bulletin of Engineering Geology and the
Environment, 80, 399-411.

6
A Conceptual Approach for Preliminary
Rock Slope Design in Remote Mountainous
Terrain
Jack Yu Guo, Douglas McLachlin
AECOM Canada Ltd., Mississauga, Ontario, Canada
Taesang Ahn
AECOM Canada Ltd., London, Ontario, Canada
Panda Bibhuti,
AECOM, Phoenix, Arizona, United States
Earl Underwood,
AECOM, Denver, Colorado, United States
Roberto de Moraes
AECOM, Oakland, California, United States

ABSTRACT

In recent years, researchers and geotechnical engineers have combined different methods to improve the accuracy and
reliability of rock slope designs. Rock slope stability designs are crucial for both mining and civil engineering projects. The
Q-slope and Slope Mass Rating (SMR) systems are widely used empirical methods that are based on the rock mass
classification system for assessing stable slope design angles. The representative stereonet plot is initially used when an
exposure of fractured rock is available to perform a kinematic analysis during preliminary design. Geophysical methods
are commonly used to determine the bedrock depth and strength of rock masses. The Q-slope method evaluates the
stability of rock slopes based on the geological conditions and engineering properties of the rock mass, while the SMR
method considers the slope geometry and the influence of discontinuities orientation on the slope stability and recommend
suitable support systems. Both the Q-slope and SMR methods require rigorous field investigations for the collection of rock
mass data. The limit equilibrium method (LEM) is widely used to determine the factor of safety (FOS) of a slope, which
normally requires comprehensive geotechnical investigations. However, sometimes it is challenging or impossible to obtain
comprehensive intrusive investigation data in remote areas due to the difficulty in site access (e.g., no access road in
mountainous terrain). Limited geological mapping and surface geophysical surveys are more feasible methods for
gathering geotechnical design parameters for remote sites. Therefore, it is important to adopt alternative design methods
for rock slope stability assessment during the preliminary design phase.. A conceptual approach is suggested in this paper
that combines Q-slope and SMR with LEM for preliminary rock slope designs under the commonly faced engineering
challenge when there is limited geological mapping, surficial geophysical surveys, and borehole investigation data. The
kinematic analysis and geophysical data are usually used to develop the design basis for the conceptual layers in the
stability section. A design flow chart is developed that may be used in the stability assessment. It is proposed that this
approach is more suitable during the pre-feasibility or feasibility design, and that it provides more useful guidance as input
to the detailed engineering design and construction stages.

RÉSUMÉ
Ces dernières années, les chercheurs et les praticiens ont combiné différentes méthodes pour améliorer la précision et la
fiabilité de la conception de pentes rocheuses. Les conceptions de stabilité des pentes rocheuses sont cruciales pour les
projets miniers et de génie civil. Les méthodes empiriques Q-slope et SMR sont utilisées pour évaluer les angles de
conception de pentes stables. Le plot stéréonet est utilisé initialement lorsque l'exposition de roches fracturées est
disponible pour effectuer une analyse cinématique. Les méthodes géophysiques sont couramment utilisées pour
déterminer la profondeur du socle rocheux et la résistance des masses rocheuses. La méthode Q-slope évalue la stabilité
des pentes rocheuses en fonction des conditions géologiques et des propriétés de l'ingénierie de la masse rocheuse,
tandis que la méthode SMR considère la géométrie de la pente et l'influence de l'orientation des discontinuités sur la
stabilité de la pente. La méthode d'équilibre limite (LEM) est largement utilisée pour déterminer le coefficient de sécurité
des pentes, ce qui nécessite normalement des enquêtes géotechniques approfondies. Cependant, il est parfois difficile ou
impossible d'obtenir des données d'investigation intrusives complètes dans des zones éloignées en raison de la difficulté
d'accès au site (par exemple, pas de route d'accès en terrain montagneux). La cartographie géologique spécialisée et les
enquêtes géophysiques de surface sont plus faisables dans de telles situations. Il est donc important d'adopter des
méthodes de conception alternatives pour l'évaluation de la stabilité des pentes rocheuses. Tout d'abord, une revue
complète des méthodes Q-slope et SMR est réalisée. Une approche conceptuelle est proposée dans cet article qui
combine Q-slope et SMR avec LEM pour la conception préliminaire de pentes rocheuses dans le cadre des défis
d'ingénierie couramment rencontrés, par exemple, la conception avec cartographie géologique, enquêtes géophysiques
de surface et données d'enquête de forage limitées. L'analyse cinématique et les données géophysiques sont
généralement utilisées pour développer les couches conceptuelles de la section de stabilité. Un organigramme de
conception est également fourni. Il est estimé que cette approche est plus adaptée à la conception de pré-faisabilité ou de
faisabilité et pourrait fournir des orientations d'ingénierie plus précises à la conception détaillée et à l'étape de construction.

1 INTRODUCTION quantitative definition of the correction factors. Later other

classification systems, such as Chinese Slope Mass Rating
Driven by an increasing global demand for minerals and System (CSMR, Chen 1995), Continuous Slope Mass
metals, the mining industry has been and will continuously Rating (Tomás et al. 2007), and the Graphical Slope Mass
be growing. New mines and expansion of existing mines Rating (Tomás et al. 2012) were developed by adopting the
require new infrastructure such as pits, tailings ponds, SMR system. The use of the SMR system has been
dams, pipelines, mills etc. Some of the existing pits also reported globally for use in more than 50 countries on all
require expansion. The related construction often includes five continents. In Asia where its use is very common it has
extensive slope excavations with significant rock removal. been incorporated to some technical guidelines and
Rock slope design is a critical aspect of mining operations. regulations for slope design (Romana et al. 2015).
To ensure both the safety and economic viability of new SMR is calculated based on the unadjusted RMR89 and
mine development, proper design of rock excavation four additional correction factors that depend on the
slopes is essential. relative orientation between the discontinuities and the rock
Over the past half-century, rock classification systems slope angle:
such as RMR and Q have been developed for assessing
rock conditions, and these were initially applied in tunneling SMR = RMR + (𝐹1 × 𝐹2 × 𝐹3 ) + 𝐹4 [1]
design projects (Bieniawski 1993, Barton 1988, Hudson
and Harrison 2000). Subsequently they were modified and Where RMR is the unadjusted RMR after Bieniawski
adapted for use in the design of rock slopes, i.e., SMR and (1989); and F1 to F4 are the correction factors that are
Q-Slope (Romana 1985, Barton and Bar 2015). However, calculated based on Table 1.
where sufficient detailed information is available, the limit
equilibrium method (LEM) (Romana et al. 2015) is widely
used for rock slope design. Typically the design process Table 1 Correction parameters for SMR modified from
using LEM involves a thorough geotechnical and Romana (1985) by Romana et al. (2003)
hydrogeology investigation through the intrusive method
using borehole or corehole drilling. However, in areas of Failure Type VF F N UF VUF
remote mountainous terrain, intrusive investigations may P |αj-αs|
be very expensive due to limited site access. Therefore, an |αj-αs- >30° 30°-20° 20°-10° 10°-5° <5°
T
alternative design approach that considers less intrusive 180°|
F1 0.15 0.40 0.70 0.85 1.00
investigation methods may be beneficial for remote areas, Relationship F1=(1-sin|αj-αs|)2
especially during in the early design stage. |βj| <20° 20°-30° 30°-35° 35°-45° >45°
In this paper, a conceptual approach for preliminary F2
P 0.15 0.40 0.70 0.85 1.00
rock slope design is presented for sites located in remote T 1.0
Relationship F2=tan2 βj
mountainous terrain.
P βj-βs >10° 10°-0° 0° 0°-(-10°) <(-10°)
T βj+βs <110° 110°-120° >120° - -
F3 0 -6 -25 -50 -60
2 SMR AND Q-SLOPE Relationship
F3 (BIENIAWSKI ADJUSTMENT RATINGS FOR JOINTS
ORIENTATION, 1976)
F4 Adjusting Natural
Presplitting
Smooth Basting or Deficient
The RMR system was developed by Professor Z.T. Factor for Slope Blasting Mechanical Blasting
Bieniawski in the early 1970s (Bieniawski, 1989). Six Excavation +15 +10 +8 0 -8
parameters are needed in the RMR system including the Method
Note:
following: VF-very favourable; F-favourable; N-normal; UF-unfavourable; VUF-Very
1. uniaxial compressive strength of rock material, unfavourable
2. rock quality designation (RQD), Failure: P-planar; T-toppling
3. spacing of discontinuities, Dip direction: αj-discontinuity; αs-slope
Dip: βj-discontinuity; βs-slope
4. condition of discontinuities,
5. groundwater conditions, and
The SMR classes and probability of failure are shown
6. orientation of discontinuities.
together with some recommended support measures in
Table 2.
The RMR system was designed to be used for tunneling,
slopes and foundations. The differences between the
Table 2 Description of SMR classes and recommended
applications typically use a rating adjustment for
slope support guidelines based on SMR and modified from
discontinuity orientations. The original RMR is not
Romana (1985) by Romana et al. (2003).
applicable to rock slopes as there is a lack of guidance for
the correction factors. The SMR method was developed
from the RMR by Romana (1985) by introducing a
 The Q system was later modified and adopted for use
in the design of rock cuts and open pit excavations (Barton
and Bar 2015; Bar and Barton 2017). Changes were made
to two of the Q parameters Jw and SRF. Jw was replaced
by Jwice in Q-slope, and it considers long-term exposure to
various climatic and environmental conditions. The
strength reduction factor, SRFslope, was enhanced to take
into account slope physical conditions, stress-strength
ratios and the effects of major discontinuities. The other
parameters (RQD, Jn, Ja, and Jr) remained unchanged.
However, orientation adjustments were introduced to the
term Jr/Ja to consider both sides of potential wedges. The
Qslope function is expressed as:

RQD Jr Jwice
×( ) ×
Q slope = Jn Ja o SRF [3]
Block size × Shear Strength × External factor
Continuous functions for the correction factors (F1 to
F3) were proposed by (Tomás et al. 2007) and are shown The Qslope parameters may be selected based on
in Table 3. The continuous functions are valuable tools for Tables 4, 5, 6 and 7. For the shear strength term, (Jr/Ja),
engineers and geologists to use in the slope design the least favorable value should be selected, or an average
process, as they may be programmed and integrated into shear strength when wedge failure is anticipated
software applications to aid in the analysis and optimization (Jr/Ja)A×(Jr/Ja)B. Table 5 provides the correction factors for
of rock slope designs. set A and set B. If the measured RQD is less than 10, then
RQD=10 shall be used in the Qslope function.
Table 3 The continuous functions proposed by (Tomás et
al. 2007)
Table 4 The parameter table for Jn, Jr, Ja based on Bar and
Factor Planar Toppling
Barton (2017)
16 3 1
𝐹1 = − tan−1 ( (|𝐴| − 17)) Joint set number, Jn
F1 25 500 10
𝐴 = 𝛼𝑗 − 𝛼𝑠 A= αj-αs-180° Description Value
9 1 17 A Massive, no or few joints 0.5-1
𝐹2 = + tan−1 ( 𝐵 − 5) B One joint set 2
F2 16 195 100 1
𝐵 = |𝛽𝑗 | C One joint set plus random joints 3
1 1 D Two joint sets 4
𝐹3 = −30 + tan−1 𝐶 𝐹3 = −13 − tan−1 (𝐶 − 120) E Two joint sets plus random joints 6
F3 3 7
𝐶 = βj − βs 𝐶 = βj + βs F Three joint sets 9
G Three joint sets plus random joints 12
Dip direction: αj-discontinuity; αs-slope H Four or more joint sets, random, heavily jointed 15
Dip: βj-discontinuity; βs-slope J Crushed rock, soil-like 20
Joint roughness number, Jr[note]
The Q-system was developed for tunnelling or rock (a) Rock-wall contact, (b) contact after shearing
slope design applications in Norway in 1974 (Barton et al. A Discontinuous joints 4
B Rough or irregular, undulating 3
1974). The Q-system is based on a numerical assessment C Smooth, undulating 2
of rock exposures, drill cores or tunnel faces, using the D Slickensided, undulating 1.5
following six parameters: E Rough or irregular, planar 1.5
F Smooth, planar 1.0
G Slickensided, planar 0.5
1. RQD © No rock-wall contact when sheared
2. Number of Joint Sets, Jn H Zone containing clay minerals thick enough to 1.0
3. Roughness of the most unfavorable joint or prevent rock-wall contact.
J Sandy, gravely or crushed zone thick enough to 1.0
discontinuity, Jr prevent rock-wall contact.
4. Degree of alteration or filling along the weakest Joint alteration number, Ja
joint, Ja (a) Rock-wall contact (no clay fillings, only coatings)
A Tightly healed, hard non-softening, impermeable 0.75
5. Water inflow, Jw filling, i.e., quartz or epidote
6. Stress condition, SRF B Unaltered joint walls, surface staining only 1.0
C Slightly altered joint walls. Non-softening mineral 2.0
The overall rock mass quality, or Q, is calculated as coatings, sandy particles, clay-free disintegrated
rock, etc.
follows: D Silty- or sandy-clay coatings, small clay 3.0
disintegrated rock, etc.
RQD Jr Jw E Softening or low friction clay mineral coatings, 4.0
Q= ∗ ∗ [2] i.e., kaolinite or mica. Also chlorite, talc, gypsum,
Jn Ja SRF graphite, etc., and
small quantities of swelling clays
(b) Rock-wall contact after some shearing (thin clay fillings, probable thickness & 1–
5 mm)
 F Sandy particles, clay-free disintegrated rock, etc. 4.0 E Residual slope with significant transport of material
20
G Strongly over-consolidated non-softening clay 6.0 downslope
mineral fillings SRFb stress and strength
H Medium or low over-consolidation, softening, 8.0 Description σc/σ1a Value
clay mineral fillings F Moderate stress-strength range 50-200 2.5-1
J Swelling-clay fillings, i.e., montmorillonite. Value 8-12 G High stress-strength range 10-50 5-2.5
of Ja depends on percent of swelling clay-size H Localized intact rock failure 5-10 10-5
particles and access to J Crushing or plastic yield 2.5-5 15-10
water K Plastic flow of strain softened material 1-2.5 20-15
(c) No rock-wall contact when sheared (thick clay/crushed rock fillings)
SRFc major discontinuity
M Zones or bands of disintegrated or crushed rock 6,8, or 8-
Description I II III IV
and clay (see G, H, J for description of clay 12
condition) L Major discontinuity with little or no clay 1 2 4 8
N Zones or bands of silty- or sandy-clay, small clay 5.0 M Major discontinuity with RQD100 = 0b due 2 4 8 16
fraction (non-softening) to clay and crushed
OPR Thick, continuous zones or bands of clay (see G, 10, 13, or rock
H, J for description of clay condition) 13-20 N Major discontinuity with RQD300 = 0c due 4 8 12 24
to clay and crushed
Note: rock
Descriptions refer to small-scale features and intermediate scale I-Favorable; II-Unfavorable; III-Very unfavorable; IV-Causing failure if unsupported
features, in that order
Note:
1. add 1.0 if mean spacing of the relevant joint set is greater than 3 m
a. σc = unconfined compressive strength (UCS), σ1 = maximum
2. Jr = 0.5 can be used for planar, slickensided joints having lineations,
principal stress
provided the lineations are oriented for minimum strength
b. RQD at 1m perpendicular sample of discontinuity
3. Jr and Ja classification are applied to the discontinuity set or sets that
c. RQD at 3m perpendicular sample of discontinuity
are least favorable for stability both from the point of view of
orientation and shear resistance s, where τ ≈ σn tan−1(Jr ⁄Ja )
Barton and Bar (2015) propose a simple equation to
find the steepest slope angle for a slope less than 30 m
Table 5 The parameter table for O-factor based on Bar and high that is safe without reinforcement.
Barton (2017)
β = 20log10 Qslope + 65° [4]
Discontinuity orientation factor, O-factor
Description Set A Set B The equation is based on the back analysis of 412
Very favorably oriented 2.0 1.5 cases. The equation is based on the probability of the
Quite favorable 1.0 1.0 failure (PoF) of 1%. One of the benefits from the Q-slope
Unfavorable 0.75 0.9 system is that it allows field engineers or geologists to
Very unfavorable 0.5 0.8 quickly assess a probable safe slope angle based on solely
Causing failure if unsupported 0.25 0.5
the field geological mapping of a natural slope or upon the
daylight of the newly constructed slopes. Equations for
Table 6 The parameter table for Jwice based on Bar and higher PoFs were also provided (Bar and Barton 2017) that
Barton (2017) may be used either for the design of temporary slopes or
for a cost-benefit analysis of long-term safe slopes.
Jwice-Value It can be noticed that both SMR and Q-slope are
Description Desert Wet Tropical Ice
Env. Env. Stm. Wdg. additionally providing the rock mass characterizations to
Stable structure; competent rock 1.0 0.7 0.5 0.9 aid in overall design. The SMR method does not provide a
Stable structure; incompetent rock 0.7 0.6 0.3 0.5 safe slope angle as does Q-slope. Rather, SMR provides
Unstable structure; competent rock 0.8 0.5 0.1 0.3
Unstable structure; incompetent 0.5 0.3 0.05 0.2
a range of supporting recommendations based on the
rock calculated SMR values. In the context of rock slope design
Desert Env.- Desert environment; Wet Env.- Wet environment; Tropical STM- it is important to note that rock mass classification systems
Tropical Storm; Ice Wdg.- Ice wedging
Note:
“…were not intended to replace analytical considerations,
a. Drainage measures installed apply: Jwice×1.5 field observations, or engineer judgment; and they were
b. Slope reinforcement installed apply: Jwice×1.3 simply to be design aids, forming only a part of the rock
c. Drainage measures and slope reinforcement installed Apply: engineer’s ‘bag of tools’” (Bieniawski 1993).
Jwice×1.5×1.3

Table 7 The parameter table for SRF based on Bar and 3 SEISMIC REFRACTION SURVEY
Barton (2017)
Seismic refraction survey is a common method used to
SRFa physical condition
Description Value
evaluate the thickness of soil and weathered rock ,
A Slight loosening due to surface location, disturbance estimate the rippability of the underlying bedrock materials,
2.5 and identify anomalous zones with low compressional
from blasting or excavation
B Loose blocks, signs of tension cracks and joint wave velocity. The survey is conducted by transmitting
shearing, susceptibility to weathering, severe 5
disturbance from blasting
seismic energy into the ground and recording the arrival
C As B, but strong susceptibility to weathering 10 time of the direct and refracted compressional (P-waves)
D Slope is in advanced stage of erosion and loosening at a known distance along the ground surface using a
due to periodic erosion by water and/or ice-wedging 15 series of geophones.
effects
Seismic refraction survey is a very useful geophysical
method that is less restricted by access limitations,
especially in remote areas. But it is normally calibrated with and analyzing data on the rock formations, major faults,
intrusive investigation results, typically boreholes and and other geological discontinuity features of the study
coreholes, for a more accurate interpretation of the area. A kinematic analysis also known as the Stereo Plot
subsurface ground profiles. Method should be performed if surficial exposure of joints
In addition, the P-wave velocities obtained from a and discontinuities are available to determine if joint
seismic survey may also be used to corrleate some rock controlled failure modes are present. For slope-related
properties, such as density, porosity, and uniaxial design assessments, the geological mapping should be
compressive strength. Research has been carried out to completed based on the Q-slope system. Typically,
develop empirical correlations between rock properties and detailed surficial geological mapping will provide both the
P-wave velocities. Especially at the feasibility stage study, RMR and Q-slope values for the rock outcrops or the
these correlations can provide valuable input into the exposed slope faces. In addition, geological mapping is
stability analysis using empirical correlations based on a also a valuable tool for slope construction inspection, as it
limited amount of field data. Table 7 provides a summary allows for a rapid assessment of the stability of newly
of selected equations used for the correlations between excavated slopes.
rock properties and P-wave velocities. Seismic refraction surveys are typically conducted
during the PFS or FS stages of mining or civil engineering
Table 7 Selected correlations between rock properties and projects and they are often complemented by subsurface
P-wave velocities geotechnical investigations. The seismic survey results are
Parameters Equations Reference used together with the borehole records to determine the
Barton (2006) soil and weathered rock thickness and rock properties that
Vp=4.75ρ-7.3 modified from are used in the stability analysis. While boreholes are
(Vp in km/s; γ in g/cm3) Kujundzíc and typically planned during this stage, the data may not be
Density Grujíc (1966) sufficient for the design. Therefore, geological mapping
Vp=4.3183ρ-7.5071 Yasar and and geophysical surveys are valuable sources of
(Vp in km/s; γ in g/cm3) Erdogan (2004)
supplementary information for the design, particularly
Vp=0.0317σc+2.0195 Yasar and when borehole data is limited during the preliminary design
(Vp in km/s; σc in MPa) Erdogan (2004)
phase.
Cobanglu and The limit equilibrium method (LEM) is a widely used
σc = 56.71 Vp - 192.93 Celik (2008) for
sandstone,
method for slope design. However, it often necessitates the
(Vp in km/s; σc in MPa)
limestone development of precise rock parameters to ensure a robust
Tuğrul and Zarif design. Given the constraints outlined above, there is often
σc = 35.54Vp – 55 insufficient information to use the limit equilibrium analysis
(1999) for granitic
(Vp in km/s; σc in MPa) rocks method during the PFS stage.. Therefore, it is more
UCS σc = 0.0642Vp – 117.99 Sharma and efficient to use empirical methods such as SMR and Q to
(Vp in m/s; σc in MPa) Singh (2008) establish an assumed stable slope geometry prior to
σc =9.95 Vp^1.21 conducting the LEM. Accurate design parameters can be
Kahraman (2001)
(Vp in km/s; σc in MPa) obtained through a comprehensive geological mapping
σc =2.304 Vp^2.4315 Kilic and Teymen program, geophysical surveys, and a desktop review of the
(Vp in km/s; σc in MPa) (2008) available subsurface investigations.
Barton (2006) The flow chart in Figure 1 provides a conceptual design
σc ≈0.25 Vp^3 to Vp^3 modified from approach for use in remote mountainous areas with the
Ohkubo and constraints discussed above.
Teresaki (1977)

4 CONCEPTUAL DESIGN APPROACH FOR

REMOTE MOUNTAINOUS TERRAIN

The normal design sequence for mining expansion typically

include Pre-Feasibility Study (PFS), Feasibility Study (FS),
and Detailed Design stages. During the PFS stage, a
preliminary analysis of the potential mining project and
conceptual design is carried out to determine its overall
feasibility. Then in the FS stage a more detailed study is
carried out that builds on the findings of the PFS and it
includes comprehensive geotechnical investigations.
However, site accessibility is a common issue that often
restricts geotechnical investigations, especially in remote
mountainous terrain. In such conditions, specialized limited
geological mapping and surface geophysical surveys (such
as seismic refraction surveys) are more feasible.
Geological mapping is a routine task for any rock
Figure 1. Conceptual Design Approach Flow Chart
engineering investigation work. This involves collecting
 A kinematic analysis is first carried out using a 5 CONCLUSION
stereonet to identify the modes of the slope failure based
on the joint set orientations and determining their daylight When carrying out rock slope designs for projects
envelopes. The stability conditions may also be examined located in remote mountainous terrains a common
by developing the friction cone where the joints intersect. challenge is limited site accessibility and a lack of
The residual joint friction angle should be obtained from a subsurface investigation information. In this paper, the
direct shear test on a smooth, planar surface of a core authors present a conceptual design approach using a
sample. However, the shear test may also be conducted combination of empirical methods, i.e., SMR and Q-Slope,
on a saw cut sample, which will result a lower value followed by LEM analysis once more detailed subsurface
compared to the natural surface. The friction angle of a information is available. The design parameters may be
natural rock surface may be estimated by measuring the obtained through geological mapping based on Q-slope
joint roughness and using the following equation: and surficial geophysical surveys with limited confirmatory
boreholes. The paper also provides a review of the SMR
σc and Q-slope empirical methods, outlining the procedures
Friction angle = Φr + JRClog10 ( ′ ) [5]
σ to obtain the SMR and Q values. Several empirical
correlations between P-wave velocity and rock parameters
Where Φr is the residual friction angle, which could be are presented. A design flowchart is provided that
either measured directly or estimated based on published illustrates the overall concept of parameter study and slope
empirical ranges (Barton 1973 , Jaeger et al. 2009); JRC is stability analysis through the various design stages of a
the joint roughness that may be estimated based during the project. This approach is provides better engineering
geological mapping; σc is the rock strength, that can be guidance in the early stages of the design and helps to
obtained from lab test or Schmidt hammer during the focus the subsurface investigations required during the
geological mapping. Some of the empirical correlations detailed design and construction stages.
regarding the rebound number (Rn) and the strength of the
rock are listed in Table 8
REFERENCE
Table 8 Some correlations between rock strength and
Schmidt hammer rebound number Bieniawski, Z.T., 1993. Classification of rock masses for
Equations Reference engineering: the RMR system and future trends.
σc=6.9×10^[0.16+0.0087(Rn* ρ)] Deere and Miller (1966) In Rock Testing and Site Characterization (pp. 553-
(σc in MPa; ρ is density in g/cm3) for varied rock 573). Pergamon.
σc =0.477EXP(0.045Rn+ρ) Kidybinski (1980) for coal, Bieniawski, Z.T., 1989. Engineering rock mass
(σc in MPa; ρ is density in g/cm3) shale, mudstone classifications: a complete manual for engineers and
σc = 0.308Rn^1.327 Selby (1993) weak rock geologists in mining, civil, and petroleum engineering.
(σc in MPa) on coal bearing strata John Wiley & Sons.
σc = 12.74 EXP(0.0185(Rn*ρ)) Beverly et al. (1979) for Barton, N., 2006. Rock quality, seismic velocity,
(σc in MPa; ρ is density in g/cm3) varied rock attenuation and anisotropy. CRC press.
σc = 2.208 EXP(0.067Rn) Katz et al. (2000) for Barton, N., Lien, R. and Lunde, J.J.R.M., 1974.
(σc in MPa) Limestone, sandstone Engineering classification of rock masses for the design
of tunnel support. Rock mechanics, 6, pp.189-236.
Barton, N., 1988. Rock mass classification and tunnel
reinforcement selection using the Q-system.
The kinematic analysis typically provides the direction
In Symposium on Rock Classification Systems for
for the further numerical analysis using specific failure
Engineering Purposes, 1987, Cincinnati, Ohio, USA.
modes. However, before conducting a detailed numerical
Barton, N. and Bar, N., 2015, October. Introducing the Q-
analysis (usually LEM), it is better to first calculate the Q-
slope method and its intended use within civil and
slope and SMR values to determine the design slope angle
mining engineering projects. In ISRM Regional
and identify any potential support methods that may be
Symposium-EUROCK 2015. OnePetro.
required. Analysis using LEM requires a detailed
Bar, N. and Barton, N., 2017. The Q-slope method for rock
parameter study from the available geotechnical and
slope engineering. Rock Mechanics and Rock
geological information. As mentioned above, rock
Engineering, 50, pp.3307-3322.
parameters can be obtained through intrusive methods
Chen, Z., 1995, September. Recent developments in slope
such as boreholes or coreholes or along the excavated
stability analysis. In 8th ISRM Congress. OnePetro.
slope. Performing a cross-check between intrusive and
Çobanoğlu, İ. and Çelik, S.B., 2008. Estimation of uniaxial
non-intrusive data is imperative to confirm the selection of
compressive strength from point load strength, Schmidt
the rock parameters obtained from both methods. The LEM
hardness and P-wave velocity. Bulletin of Engineering
analysis may be performed to check the slope stability and
Geology and the Environment, 67, pp.491-498.
assess the associated FOS. Finally, the recommendation
Hudson, J.A. and Harrison, J.P., 2000. Engineering rock
of the slope design should be provided based on the LEM
mechanics: an introduction to the principles. Elsevier.
analysis with more precise engineering guidance during
Kahraman, S.A.İ.R., 2001. Evaluation of simple methods
the detailed design and construction stages.
for assessing the uniaxial compressive strength of
 rock. International Journal of Rock Mechanics and
Mining Sciences, 38(7), pp.981-994.
Kılıç, A. and Teymen, A., 2008. Determination of
mechanical properties of rocks using simple
methods. Bulletin of Engineering Geology and the
Environment, 67, pp.237-244.
Kujundzic, B. and Grujic, N., 1966, September. Correlation
between static and dynamic investigations of rock mass
in situ. In 1st ISRM Congress. OnePetro.
Ohkubo, T. and Teresaki, A., 1971. Oyo
Corporation. Technical Note RP-479.
Romana, M., 1985, September. New adjustment ratings for
application of Bieniawski classification to slopes.
In Proceedings of the international symposium on role
of rock mechanics, Zacatecas, Mexico (pp. 49-53).
Romana, M., Tomás, R. and Serón, J.B., 2015, May. Slope
Mass Rating (SMR) geomechanics classification: thirty
years review. In 13th ISRM international congress of
rock mechanics. OnePetro.
Romana, M., Serón, J.B. and Montalar, E., 2003,
September. SMR geomechanics classification:
application, experience and validation. In 10th ISRM
Congress. OnePetro.
Sharma, P.K. and Singh, T.N., 2008. A correlation between
P-wave velocity, impact strength index, slake durability
index and uniaxial compressive strength. Bulletin of
Engineering Geology and the Environment, 67, pp.17-
22.
Tomás, R., Delgado, J. and Serón, J.B., 2007. Modification
of slope mass rating (SMR) by continuous
functions. International journal of rock mechanics and
mining sciences, 44(7), pp.1062-1069.
Tomas, R., Cuenca, A., Cano, M. and García-Barba, J.,
2012. A graphical approach for slope mass rating
(SMR). Engineering Geology, 124, pp.67-76.
Tuğrul, A. and Zarif, I.H., 1999. Correlation of mineralogical
and textural characteristics with engineering properties
of selected granitic rocks from Turkey. Engineering
geology, 51(4), pp.303-317.
Jaeger, J.C., Cook, N.G. and Zimmerman, R.,
2009. Fundamentals of rock mechanics. John Wiley &
Sons.
Yasar, E. and Erdogan, Y., 2004. Correlating sound
velocity with the density, compressive strength and
Young's modulus of carbonate rocks. International
Journal of Rock Mechanics and Mining
Sciences, 41(5), pp.871-875.
Numerical analysis of the time-dependent
response of backfilled mine openings
in soft rocks

Abtin Jahanbakhshzadeh1, Jonathan D. Aubertin2 & Michel

Aubertin1
1Polytechnique de Montréal,

P.O. Box 6079, Station Centre-ville, Montréal, QC, Canada

2École de technologie supérieure (ETS),

Université du Québec, Montréal, QC, Canada

ABSTRACT
Backfilling is commonly used in underground mines to improve ground stability. Mine backfill can also serve to control wall
convergence and maintain the integrity of pillars by adding local support. The design of backfilled openings requires an
evaluation of the stress state in the filling material and surrounding rock mass. To date, most analyses of the stress
distribution in and around backfilled openings have been performed for hard rock mines with little deformation of the walls
after fill placement. The analysis however becomes much more complex when the rock is subjected to significant time-
dependent (creep) behavior as is often the cases in many soft rocks (rocksalt, potash). This article presents some results
from recently conducted simulations to assess the interaction between creeping rock and the backfill placed in mine
openings. The simulations are performed with FLAC with the time-dependant ISV-SH model. Numerical modelling results
are used to evaluate the evolution of the strain and stress distributions due to transient and steady-state creep in
underground backfilled openings, considering different geometries and properties. A brief discussion follow.
RÉSUMÉ
Le remblayage est couramment utilisé dans les mines souterraines afin d’améliorer la stabilité. Le remblai minier peut
également servir à contrôler la convergence des murs et à maintenir l'intégrité des piliers en ajoutant un support local. La
conception des ouvertures remblayées nécessite une évaluation de l'état de contrainte dans le remblai et le massif rocheux
environnant. À ce jour, la plupart des analyses de la distribution des contraintes autour de ces ouvertures ont été effectuées
pour des mines en roche dure avec peu de déformation des parois après la mise en place du remblai. Les analyses
deviennent toutefois beaucoup plus complexes lorsque la roche montre un comportement différé (avec fluage) comme
c'est le cas des nombreuses roches tendres (sel gemme, potasse). Cet article présente des résultats de simulations
récemment menées pour évaluer l'interaction entre la roche et le remblai placé dans les ouvertures minières. Les
simulations sont réalisées avec FLAC et le modèle ISV-SH. Les résultats de la modélisation numérique sont utilisés pour
évaluer la distribution des déformations et des contraintes en fonction du temps due au comportement différé transitoire
et stationnaire de la roche dans les ouvertures souterraines remblayées, pour différentes géométries et propriétés. Une
brève discussion suit.

1 INTRODUCTION this limitation, there has been significant progress made to

account for a variety of influence factors such as opening
Modern mining methods often rely on backfilling of shape and inclination, and the impact of porewater
underground openings to reduce the amount of wastes pressures and backfilling sequence (e.g. Aubertin et al.
disposed on the surface and related environmental impacts 2003; Li et al. 2003, 2005; Pirapakaran and Sivakugan,
(Aubertin et al. 2002; Benzaazoua et al. 2008). Mine 2006; Li and Aubertin 2008, 2009a; Helinski et al. 2007,
backfill also provides ground support to improve stability of 2010; Fahey et al. 2009; Jahanbakhshzadeh et al. 2017a).
rock mass, while contributing to increased ore recovery There is however a paucity of results for backfill placed
and reduced dilution (Hassani and Archibald 1998; Darling in openings in low porosity soft rocks, such as rocksalt,
2011). which require that the time-dependent deformation of the
A realistic evaluation of the stress state in and around rock and its effect on the backfill response be considered
backfilled openings is essential for their design. This explicitly.
requires an understanding of the geomechanical behavior In this article, the authors present an investigation on
of the backfill and surrounding rock. In some cases, the the behavior of backfilled openings in salt mines based on
stress distribution is intrinsically time-dependent, and this numerical simulations conducted with FLAC and the ISH-
effect must be taken into account. This is particularly the SH time-dependant model (Aubertin and Aubertin, 2022;
case for mine openings in low porosity soft rocks such as Aubertin et al. 2023). In these simulations, the backfill
rocksalt and potash. behavior follows the Mohr-Coulomb Elastoplastic model or
Time effects are commonly neglected in the vast the Duncan-Chang stress-strain equation. The result
majority of investigations conducted to evaluate the demonstrates how to assess the impact of time-dependent
behavior of backfilled stopes with two- and three- rock deformation on the stresses with backfilled openings
dimensional analytical and numerical solutions. Despite for salt mining conditions.
 model can represent the time-dependent inelastic behavior
2 NUMERICAL SIMULATIONS OF THE TIME- of rocksalt, considering its strong history- and time-
DEPENDENT BEHAVIOR OF SOFT ROCKS dependent response (Aubertin et al. 2023).

Numerical analyses are commonly used to assess the Unsupported circular opening
behavior of underground backfilled openings in rock (e.g.
Li et al., 2003, 2007; Li and Aubertin 2009a; Pirapakaran Simulations are first conducted on a circular horizontal
and Sivakugan, 2006; El Mkadmi et al., 2014; Falaknaz et opening to illustrate the nature of the time-dependent
al., 2015a, 2015b; Jahanbakhshzadeh et al., 2017a, b). As behavior of rocksalt. Figure 1 shows the equivalent (von
indicated above, most of these analyses neglect the effect Mises) deviatoric stress 𝜎̅ (MPa) distribution calculated
of the time-dependent response of the surrounding rock. with the ISV-SH model immediately after the application of
New simulation results are conducted here to evaluate the an external radial pressure (i.e. elastic response) and the
stress state in excavations created in low porosity soft evolution over time after 30 days, 90 days, and one year.
rocks, considering explicitly the time-dependent This plane-strain simulation with the ISV-SH model
deformation of the walls based on a model recently implemented in FLAC evaluates the time-dependent
proposed to simulate the response of openings in salt response of a cylinder with an internal diameter of 1 m and
mines. an external diameter of 16 m, subjected to an external
The simulations are conducted with the internal state stress of 8 MPa (based on Aubertin et al. 2023).
variable, strain hardening (ISV-SH) model implemented in
FLAC (Itasca, 2005, version 8), (Aubertin and Aubertin, Table 1: Parameters used for the simulations with FLAC.
2021; Aubertin et al. 2023). The simulations are first Calibrated values of the ISV-SH model parameters for
conducted to evaluate the response of an externally rocksalt (taken form Aubertin et al. 2023)
pressurized cylinder resembling a circular mine shaft. Parameters Value
In these calculations, total strains are obtained by adding
the elastic strains, given by Hooke’s law, and inelastic 𝜖0̇ (s-1) 4.0E-09
strains obtained from the ISV-SH model. The latter 𝜎0 (Pa) 2.0E+07
constitutive model can be expressed from the following
differential equations for the inelastic strain rate (Aubertin 𝑛 4.0
and Aubertin 2022): 𝐴 (s-1) 5.0E-12
̅ 𝑁
𝜎 𝐴5 5.0E+08
𝜀̇ 𝑖 = 𝐴 ( ) [1]
𝐾
𝑁 5
𝐾
𝐾̇ = 𝐴5 (1 − ) 𝜖̇𝑖 [2] 𝐾0 (Pa) 2.0E+06
𝐾′
𝐸 (GPa) 17.06
̅′
𝜎
𝐾′ = ≥ 1 𝑀𝑃𝑎 [3]
0.33
1
𝜀̇ 𝑖 𝑁 𝜈
( )
𝐴
m (PL exponent) 4.0
1
𝜖̇ 𝑖 𝑛 A (𝑃𝑎−𝑚 ∙ 𝑠 −1 ) 5.0E-14
𝜎̅′ = 𝜎0 ∙ sinh−1 ( ̇ ) [4]
𝜖0
(PL coefficient)
where 𝐴, 𝐴5 , 𝑁, 𝑛, 𝜎0 , and 𝜖0̇ are the 6 material parameters
determined from experimental results on rocksalt; 𝜎̅′ and
𝐾′ are the saturated (steady-state) values of the deviatoric
stress 𝜎̅ and the isotropic internal state variable 𝐾 which
controls strain hardening. Eq. 2 gives the evolution of 𝐾 (
1 MPa) which progressively converges towards the value
𝐾′ reached upon the inelastic steady state flow. As
illustrated in recent publications, the ISV-SH model
typically behaves as a classical strain-hardening law at
small inelastic strain in a creep test, but diverges from the
latter as it naturally evolves towards a stationary creep
condition at larger strain (Aubertin and Aubertin, 2022;
Aubertin et al. 2023).
Table 1 presents the essential parameter values used
in the simulations with FLAC, following the numerical
procedures described by Aubertin et al. (2023); the table
also gives the two parameters of the Norton Power Law
(PL) equation for steady-state creep, which is also
considered for comparative assessment. The recent
analyses with the ISV-SH model have shown how this
 a) Initial state b) 30 days support (backfill) is added immediately after the excavation
is created.
A monitoring point was set at P1 located at about 0.7 m
from the boundary on both sides (bottom and left side, i.e.
45° from the vertical axis). This point is used to compare
the results with the two scenarios (with and without filling).
The other input parameters of the numerical models are
given in Table 1.

c) 90 days d) 1 year
8 MPa

Rock mass

Figure 1. Equivalent deviatoric stress (MPa) around a

circular opening obtained with FLAC and the ISV-SH model
(see text for details); (a) initial stress state (elastic solution)
and stresses after (b) 30 days (c) 90 days and (d) 1 year
(adapted from Aubertin et al. 2023). •P1
Supporting material
3 SIMULATIONS OF BACKFILLED OPENINGS Figure 3. Geometric configuration of the conceptual and
numerical model used for the numerical simulations of a
The simulation procedure for backfilled openings can be circular opening with backfill (not to scale). Results shown
summarized as follows. The excavation is created in the below are taken at point P1. The applied external stress
rock in one step; The calculations are performed with FLAC equals 8 MPa.
by combining the linear elastic and the time-dependent
ISV-SH models. After creation of the excavation, the Figure 4 presents the iso-contour of the equivalent
backfill is added, with a behavior based on two different deviatoric stress (MPa) obtained with the ISV-SH model,
constitutive laws (with parameters calibrated on laboratory considering the presence of a stiff ground support within
experiment data). The Mohr-Coulomb criterion is first the opening. These results, which can be compared with
applied as part of the elasto-plastic model for the backfill. the case without support (Fig. 1), indicate that the stiff
This well-known model considers that the shear strength of support modifies substantially the stress state around the
the backfill can be expressed from the elastic parameters, opening.
cohesion and internal friction angle.
Additional simulations are also conducted with the a) 30 days b) 90 days
Duncan-Chang stress-strain model, also available in
FLAC, for the backfill material. This non-linear model
accounts for elastic and plastic behavior in a somewhat
more realistic manner.
All numerical simulations were conducted with the time-
dependent rock properties given in Table 1.

3.1 Circular opening

Figure 3 shows the conceptual model for the case of a stiff

(backfill) support added inside the circular opening. The
dimensions of the model are similar to the previous model Figure 4. Simulated iso-contour of the deviatoric stresses
described in Section 2; An external stress of 8 MPa is (MPa) around the backfilled cylindrical opening after (a) 30
applied, and the Mohr-Coulomb EP model is used for the days and (b) 90 days.
(stiff) backfill. The properties of this backfill take the
following values (based on Yin et al. 2020; Yan et al. 2021) Figure 5 presents the simulation results for the radial
and tangential stresses obtained with the ISV-SH model for
: unit weight of γ = 18.00 (kN/m3), elastic Young’s modulus the cylindrical opening filled with the stiff support material
E = 26 GPa, Poisson’s ratio 𝜐 = 0.2, internal (constant inside. These results are compared with those obtained for
volume) friction angle 𝜙′= 30°, cohesion c′ = 4 kPa, a steady-state (long term) stress condition given by the
dilatation angle ψ′ = 0. There is no pore water pressure so commonly used Norton power law (PL) model, for the case
effective and total stresses are the same. In this case, the without support. The figures provide a visual representation
of how the stresses change over time within the rock, under
a constant applied (external) stress state.

Figure 6. Simulation results for the vertical displacement at

point located about 0.7 m from the boundary on both sides
(P1) as a function of time; cases with and without backfill.
a)
The simulation results demonstrate that the use of a
stiff filling material can lead to a substantial reduction in
displacement within the opening; this also modifies the
stress distribution by increasing the radial stresses at and
near the wall. Filling may thus play a major role in
controlling the magnitude of time-dependent
displacements, which may contribute to enhancing the
stability of the excavation by increasing the confining
pressure on the rock.

b)
Figure 5. Simulation results obtained with the ISV-SH
constitutive model for the circular opening (Fig. 3)
submitted to an external constant pressure of 8 MPa after
30, 90 days and 365 days (1 year) a) tangential stress and
b) radial stress.; the analytical solution given by the power
law (PL) equation is also shown (stationary state, without
ground support).

Figures 6 and 7 show of the vertical displacements and

vertical stresses at point P1 for the circular opening with
and without (stiff) backfilling.
Figure 7. Simulated vertical stresses (MPa) at P1 as a
function of time; cases with and without backfill.

Figure 8 shows that the elastic modulus of the fill

material has a significant influence on the stress
distribution in the surrounding rock mass. The stiffer backfill
(26 Gpa vs 260 Mpa) reduces the displacements and
increases the normal stresses at the internal boundary.
 a) Initial

b) 1 year

Figure 8. Simulated vertical stresses at P1 as a function of

time; for cases without backfill and with backfills having two
different elastic moduli.

These simulations demonstrate, as expected, that

increasing the stiffness of the backfill material leads to a c) 5 years
significant decrease in displacement within the rock mass
near the excavation. This is accompanied by a substantial
increase in vertical stresses near the excavation, which
progressively evolves toward a stationary value, as
illustrated in Figure 8.

3.2 Rectangular opening

The ISV-SH model implemented in FLAC is also used to

assess the time-dependant behavior of rocksalt Figure 10. Iso-contour of deviatoric stresses around the
surrounding a rectangular backfilled mine opening. The unsupported mine opening: a) Initial condition (elastic
two-dimensional (plane strain) model in Figure 9 (not to response); b) after 1 year; c) after 5 years.
scale), identifies the geometric parameters (including the
stope dimensions) and the imposed boundary conditions. A simulation is also conducted for the case of backfill
added immediately (a few seconds) after creation of the
opening. Figure 11 shows the variations of vertical (Figure
11a) and horizontal (Figure 11b) stresses along the central
line of the backfilled opening, at different elevations and
time intervals. The Mohr-Coulomb EP model is used here
for the backfill. The results in Figure 11 indicate that the
time-dependent strains (shown in Figure 13) in the
surrounding rock mass have an impact on the normal
stresses of the backfill.
Figure 9. Geometric configuration of the
conceptual/numerical model used for the simulations of a
rectangular backfilled mine opening in rocksalt (not to
scale). The external pressure is 10 MPa.

Figure 10 shows the iso-contours of the deviatoric

stresses immediately after the creation of the opening, and
one year and 5 years later (without backfilling).
 can be significantly reduced by adding backfill. This in turn
influences the stress distribution in and around the
backfilled opening.

a)

Figure 13. Vertical displacements at top of the opening for

cases with and without backfill.

The choice of a suitable constitutive model for each

material is crucial in finite element analyses of geotechnical
problems.
One commonly used nonlinear stress-strain model for
geomaterials is the Duncan and Chang's equation (Duncan
and Chang 1970; Duncan et al. 1980). This model, which
is deemed more representative for salt backfill, is applied
here to represent the fill material in the numerical analyses.
The hyperbolic stress-strain model of Duncan and
Chang implemented in FLAC can be expressed as:
b)
Figure 11. Numerical simulation results of the time- 𝜎 𝑛 𝑅𝑓 (1−sin 𝜑)(𝜎1 −𝜎1 ) 2
𝐸𝑡 = 𝐾𝑃𝑎 ( 3) [1 − ] [5]
dependent interaction between the backfill and 𝑃𝑎 2𝑐 cos 𝜑+2𝜎3 sin 𝜑
surrounding rock mass: (a) vertical stress along the central 𝑚
line at different elevations; (b) Horizontal stresses along the 𝜎
𝐵𝑡 = 𝐾𝑏 𝑃𝑎 ( 3 ) [6]
central line. 𝑃𝑎

where Et is the tangent modulus of elasticity at any loading

Figure 12 shows the iso-contour of vertical stresses
stage, Rf is the failure ratio (defined as the ratio between
within the backfilled opening (represented by the Mohr-
the deviatoric stress at failure and the ultimate deviator
Coulomb EP model) at two different times, after 1 year (Fig.
stress Rf = 1 for backfill), c and φ are Mohr–Coulomb
12a) and after 5 years backfilling (Fig.12b).
strength parameters, σ1 is major principal stress, Bt is the
stress-dependent bulk modulus, Kb is the bulk modulus
a) 1 year b) 5 years number, and m is the bulk modulus exponent.
The Duncan and Chang model parameters are given in
Table 2. These were adapted here for salt backfill, based
on parameters proposed for sand (Hamrouni et al. 2022).
Parameters for the simulations conducted with the ISV-SH
model for rocksalt were taken in Table 1.

Table 2: Salt backfill parameters for the Duncan and

Chang’s equation.
Unite Backfill
Unit weight (kN/m3) 18.00
𝜙′ ° 30
Figure 12. Iso-contour of vertical stresses (MPa) inside the
Gref (kPa) 5E6
backfill (Mohr-Coulomb EP Model) and around the opening
at different times: a) after 1 years b) after 5 years. Rf 1
β 2.35
Figure 13 shows the vertical displacement at the center
of the opening roof for cases with and without backfilling.
This comparison confirms that the vertical displacement
 Figure 14 shown the iso-contour of the vertical stresses
in the backfilled opening when the Duncan and Chang’s 4 DISCUSSION AND FINAL REMARKS
equation is used for the filing material.
This article presents the results of a numerical investigation
a) Initial b) 1year of the time-dependent response of mine openings in
rocksalt, with and without ground support provided by
adding backfill. The ISV-SH time-dependent model,
implemented in FLAC, is used to analyze the behavior of
the rock around the mine openings. The investigation looks
into the effect of different opening geometries and backfill
properties.
The results illustrate how filling the mine opening
rapidly (and completely) with backfill can significantly
reduce the displacement of the surrounding wall. The
simulations also indicate that reducing the modulus of the
backfill (from 26 GPa to 260 MPa) results in a notable
reduction in vertical stresses around the excavation.
c) 5 years The results presented in this study focus on the time-
dependent behavior of rocksalt rock around backfilled
mining excavations. The simulations are conducted in
plane strain; the influence of the third dimension is thus
neglected. Further work is being conducted to investigate
the effect of opening geometry, and of the time delay
before placing the backfill in one or more excavations
(created in sequence). The influence of a gap space
between the fill and the opening wall (at the top) is also
being assessed. These additional aspects and other
Figure 14. Iso-contour of vertical stresses (MPa) around related factors will be addressed in future publications.
the opening at different times: a) initial b) after 1 year; c)
after 5 years.
ACKNOWLEDGEMENTS
Figure 15 shows the vertical stresses within the
backfilled opening for the two constitutive models used to The authors acknowledge the financial support from
simulate the backfill response. NSERC

REFERENCES

Aubertin, M., Bussière, B., and Bernier, L. 2002.

Environnement et gestion des rejets miniers [CD-ROM
Manual]. Montréal: Presses Internationales
Polytechnique.
Aubertin, J. D. Aubertin, M. and Jahanbakhshzadeh.A.
2023. Numerical implementation and application of an
internal state variable model to analyze the time-
dependent behavior of mining excavations in rock
salt. Canadian Geotechnical Journal. 60(6): 917-
935. https://doi.org/10.1139/cgj-2021-0593
Aubertin, J.D., and Aubertin, M. 2022. Comparative
assessment of transient creep models to analyze
Figure 15. Vertical stresses in the backfilled mine opening openings in low porosity soft rocks,
in rocksalt obtained for the two constitutive models applied Conference:GeoCalgary2022,CGS.pp.11.
to the backfill (Elasto-plastic Mohr-Coulomb and Duncan Aubertin, J. D., and Aubertin, M. 2021. Simulating the
and Chang). influence of rigid support on the time-dependent
response (creep) of soft rock around underground
It can be observed that the vertical (and horizontal) openings. Proc. GeoNiagara 2021, CGS, 8pages.
stresses along the central line obtained with both Aubertin, M., Li, L., Arnold, S., Belem, T., Bussiere, B.,
constitutive models for the backfill are in close to each Benzaazoua, M. and Simon, R. 2003. Interaction
other; there are nonetheless differences, which may between backfill and rock mass in narrow Stopes. In:
become significant when assessing the effect of backfill Culligan PJ, Einstein HH, Whittle AJ (eds) Proceedings
pressure on the damage zone developing around the mine of Soil and Rock America 2003, Verlag Glu¨ckauf
opening. Essen VGE, Essen, Germany., 1: 1157–1164.
Benzaazoua, M., Bussiere B., Demers I., Aubertin M., Fried Geotechnical and Geological Engineering 36(2), 1011-
E. and Blier A. 2008. Integrated mine tailings 1036. doi: 10.1007/s10706-017-0371-0.
management by combining environmental Li, L., Aubertin, M., Simon, R., Bussière, B. and Belem, T.
desulphurization and cemented paste backfill: 2003. Modeling arching effects in narrow backfilled
Application to mine Doyon, Quebec, Canada. Minerals stopes with FLAC, FLAC and Numerical Modeling in
Engineering, 21(4): 330-340. Geomechanics. 211–219.
Darling, P. 2011. SME mining engineering handbook (3rd Li, L., Aubertin, M., and Belem, T. 2005. Formulation of a
ed.), Society for Mining, Metallurgy and Exploration three dimensional analytical solution to evaluate
(SME), Denver. stresses in backfilled vertical narrow openings.
Duncan, J. M. and Chang, C.-Y. 1970. “Nonlinear analysis Canadian Geotechnical Journal, 42(6): 1705-1717.
of stress and strain in soils,” Journal of the Soil Li, L., Aubertin, M., Shirazi, A., Belem, T. and Simon, R.
Mechanics and Foundations Division, vol. 96, no. 5, pp. 2007. Stress distribution in inclined backfilled stopes.
1629–1653. MINEFILL 2007, April 29- May 2, 2007, Montreal,
Duncan, J. M., Byrne, P. M., Wong, K. S., and Mabry, P. Quebec, Canadian Institute of Mining, Metallurgy and
1980. “Strength, stress-strain and bulk modulus Petroleum (CIM), #2510.
parameters for finite element analyses of stresses and Li, L. and Aubertin, M. 2008. An improved analytical
movements in soil masses,” Tech. Rep. UCB/GT/80- solution to estimate the stress state in subvertical
01, University of California, Berkeley, Calif, USA. backfilled stopes. Canadian Geotechnical Journal,
El Mkadmi, N., Aubertin, M. and Li, L. 2014. Effect of 45(10): 1487-1496.
drainage and sequential filling on the behavior of Li, L. and Aubertin, M. 2009a. Numerical investigation of
backfill in mine stopes. Canadian Geotechnical Journal, the stress state in inclined backfilled stopes.
51(1): 1-15. International Journal of Geomechanics, 9(2): 52-62.
Fahey, M., Helinski, M., and Fourie, A. 2009. Some Pirapakaran K. and Sivakugan N. 2006 Numerical and
aspects of the mechanics of arching in backfilled experimental studies of arching effects within mine fill
stopes. Canadian Geotechnical Journal., 46(11), 1322- slopes. In 6th International Conference on Physical
1336. Modelling in Geotechnics - Physical Modelling in
Falaknaz, N., Aubertin, M. and Li, L. 2015a. Evaluation of Geotechnics - 6th ICPMG '06, August 4, 2006 - August
the Stress state in two adjacent backfilled stopes within 6, 2006. Taylor and Francis/Balkema, Hong Kong,
an elasto-plastic rock mass. Geotechnical and China, pp. 1519-1525.
Geological Engineering (available online, February Yan, B., Xingping L., Hanwen J., Erol Y., and Chen H.
2015). DOI: 10.1007/s10706-015-9868-6. 2021. "A Solution to the Time-Dependent Stress
Falaknaz, N., Aubertin, M. and Li, L. 2015b. Numerical Distribution in Suborbicular Backfilled Stope Interaction
investigation of the geomechanical response of with Creeping Rock." Advances in Civil
adjacent backfilled stopes. Canadian Geotechnical Engineering 2021 (March 23, 2021): 1–18.
Journal, 52(10): 1507-1525. http://dx.doi.org/10.1155/2021/5533980.
Hassani, F. and Archibald. J. 1998. Mine backfill. CD-Rom, Yin, S. Shao, Y. Wu, A. Wang, H. Liu, X. and Wang, Y.
Canadian Institute of Mine, Metallurgy and Petroleum. 2020. “A systematic review of paste technology in metal
Montreal, QC, Canada mines for cleaner production in China,” Journal of
Hamrouni A, Dias D, Guo X. 2022. Behavior of Shallow Cleaner Production, vol. 247, Article ID 119590, 2020.
Circular Tunnels—Impact of the Soil Spatial
Variability. Geosciences; 12(2):97.
https://doi.org/10.3390/geosciences12020097
Helinski, M., Fahey, M., and Fourie, A. 2007. Numerical
modeling of cemented mine backfill deposition. Journal
of Geotechnical and Geoenviromental Engineering,
133 (10): 1308-1319.
Helinski, M., Fahey, M., and Fourie, A. 2010. Coupled two-
dimensional finite element modelling of mine backfilling
with cemented tailings. Canadian Geotechnical
Journal, 47(11): 1187-1200.
Itasca Consulting Group, Inc. (Itasca). 2005. FLAC – Fast
Lagrangian Analysis of Continua. Version 5.00.
[Computer software and user manual.] Itasca
Consulting Group, Inc., Minneapolis, Minn.
Jahanbakhshzadeh, A., Aubertin, M., Li, L. 2017a. A new
analytical solution for the stress state in inclined
backfilled mine stopes. Geotechnical and Geological
Engineering 35: 1151-1167. doi:10.1007/s10706-017-
0171-6.
Jahanbakhshzadeh, A., Aubertin, M., Li, L. 2017b. Analysis
of the stress distribution in inclined backfilled stopes
using closed-form solutions and numerical simulations.
GEOLOGICAL AND GEOMECHANICAL
DIFFERENTIATION OF ROCK INTERFACES
BASED ON RECURSIVE MULTI-SCALE ANALYSIS
WITH TOPOGRAPHICAL DATA SETS
Niloufarsadat Sadeghi1, Jonathan D. Aubertin1 and Akram Deminiat1
1Construction Engineering Department, École de technologie supérieure (ETS), Montréal, QC, Canada

ABSTRACT

Proper geomechanical description of natural or engineered excavations promote reliable description of instability potential
and informative notions related to geological units’ structures and geometry. Rock mass features present different
geomechanical properties which highlight crucial information such as rock or slope instability indicators, failure
mechanisms, evaluation of geohazard phenomenon, and other engineering purposes. Most of the existing methods for
differentiating geological units are either geometry driven (e.g., differentiation based on joint set orientations) or are
empirical to human visual discrimination. This article presents a remote-sensing assisted method for differentiating,
describing, and quantifying geomechanical conditions and geological units along exposed rock outcrops.
The presented methodology is based on topographical characterization of rock surfaces from 3-dimensional digital
surface models captured with remote sensing tools. A fractal-based approach is utilized as a framework for geomechanical
differentiation of geological features. An initial recursive roughness-based analysis is performed to quantify scale-
dependency along the inspected rock outcrop. A second layer of recursive analysis is performed for small moving spherical
neighborhoods generated on the scan. Data sets from an underground tunnel wall are identified and subdivided with
respect to known geological units. Back analysis of the known topographical characteristics for different geological regions
is performed to evaluate the differences in terms of fractal components.

RÉSUMÉ

La description géomécanique appropriée des excavations naturelles ou artificielles favorise une description fiable du
potentiel d'instabilité et des notions informatives liées aux structures et à la géométrie des unités géologiques. Les
caractéristiques de la masse rocheuse présentent différentes propriétés géomécaniques qui mettent en évidence des
informations cruciales telles que les indicateurs d'instabilité des roches ou des pentes, les mécanismes de défaillance,
l'évaluation des phénomènes géohazards et d'autres objectifs d'ingénierie. La plupart des méthodes existantes pour
différencier les unités géologiques sont soit basées sur la géométrie (par exemple, différenciation basée sur les
orientations des ensembles de discontinuités), soit empiriques et reposent sur la discrimination visuelle humaine. Cet
article présente une méthode assistée par télédétection pour différencier, décrire et quantifier les conditions
géomécaniques et les unités géologiques le long des affleurements rocheux exposés.
La méthodologie présentée est basée sur la caractérisation topographique des surfaces rocheuses à partir de modèles
numériques de surface tridimensionnels capturés avec des outils de télédétection. Une approche basée sur les fractales
est utilisée comme cadre pour la différenciation géomécanique des caractéristiques géologiques. Une analyse initiale
basée sur la rugosité récursive est réalisée pour quantifier la dépendance à l'échelle le long de l'affleurement rocheux
inspecté. Une deuxième couche d'analyse récursive est réalisée pour de petits quartiers sphériques mobiles générés sur
le balayage. Des ensembles de données provenant d'un mur de tunnel souterrain sont identifiés et subdivisés par rapport
aux unités géologiques connues. Une analyse rétrospective des caractéristiques topographiques connues pour différentes
régions géologiques est réalisée pour évaluer les différences en termes de composantes fractales.

1 INTRODUCTION methods can be enhanced and rigorized through

systematic and semi-automated use of remote sensing
In-situ characterization of geomaterials is a significant step technologies. The present article focuses on derived
during the geotechnical investigation process. applications for geomechanical topographical
Conventional infield characterization methods are characterization of rock surfaces.
topographical or texture based, and indexed with empirical Remote sensing assisted rock mass characterization is
systems. Rock interfaces, for instance, are visually a popular alternative to visual surveys of rock masses since
assessed in terms of roughness and weathering, and it avoids numerous disadvantages through the means of
indexed with the well acknowledged systems of joint mathematical descriptions of 3-dimensional digital surface
roughness coefficient, JRC (Barton and Choubey, 1977) models that are captured with remote sensing tools such
and the Geological Strength Index, GSI (Hoek and Brown, as digital cameras (i.e., photogrammetry) (Travelletti et al.,
1980). Similarly, granular materials are evaluated in terms 2008; Nilsson, Edelbro and Sharrock, 2012; Kim, Gratchev
of texture, particle size, and visible perspective of water and Balasubramaniam, 2013; Bizjak and Geršak, 2018;
content (Committee, 1985). Such empirical evaluation Francioni et al., 2019; Uotinen et al., 2021) and laser
scanners (Glenn et al., 2006; Lato et al., 2009; Jaboyedoff
et al., 2012; Aubertin et al., 2017; Aubertin, Hutchinson and where the best fitted plane normal vector ⃗⃗⃗⃗⃗⃗
𝑛𝑟,𝑖 =
Diederichs, 2019; Aubertin and Hutchinson, 2022). Three- [𝑎𝑟,𝑖 , 𝑏𝑟,𝑖 , 𝑐𝑟,𝑖 ].
dimensional laser scanners rely on Light/Laser Ranging (Aubertin and Hutchinson, 2022) evaluated the overall
and Detection (LiDAR) technology to measure distances roughness of a given rock surface as the root mean
from a source and generate dense and precise point clouds squared (RMS) of all the obtained roughness values for
representative of surveyed topographies. LiDAR tools can points (𝑅𝑟,𝑖 ) along the considered surface:
be used in fixed positions mounted on tripods (i.e.,
terrestrial laser scanner, TLS) or attached to mobile
𝑁
equipment such as unmanned aerial vehicles UAV (i.e., 1 2
𝜉𝑅 = √ ∑(𝑅𝑟,𝑖 − 𝑅̅𝑟 ) [2]
drones). 𝑁−1
The mechanical behavior of rock masses at the 1

excavation scale is inherently dependent on the

characteristic superficial geometry of distinct geological where 𝜉𝑅 (m) and 𝑅̅𝑟 (m) are the root mean square (RMS)
units and rock interfaces (i.e., joints and discontinuities). and mean value of roughness measurements of individual
These features present varying geomechanical conditions points within the radius of measurement of r, respectively.
pertaining to structural geometry, weathering, excavation A number of alternative criterion and geometrical
method, level of damage, etc. Differentiating and modelling formulations for superficial roughness exist in the literature
natural and artificial features is an essential component of (Shepard et al., 2001; Berti, Corsini and Daehne, 2013).
the geotechnical investigation and characterization Other forms of statistical roughness representation may
process. consider spatial variation between points to quantify the
Topographical features of rock surfaces effectively topographical gradient and derive angle of inclination
differentiate geological units in terms of geotechnical and (McKean and Roering, 2004; Glenn et al., 2006; Poropat,
geomorphological conditions. The superficial roughness of 2008; Tatone and Grasselli, 2013; Lai, Samson and Bose,
rock interfaces represents a characteristic parameter that 2014). Directionality may also be incorporated into the
can be seen as a physical “signature” bearing the influence formulation via a weighted distribution, or through recursive
of formation, weathering, and excavation history. Such vectorized analysis of the surface considered (Grasselli
features can in fact be used to discriminate geological units and Egger, 2003; Haneberg, 2007). The ensemble of
along rock outcrops. Multi-scale evaluation of topographic geometrical and statistical formulations for surface
features can thus offer a meaningful spatial description of roughness has been correlated with various geotechnical
geological units with direct correlation of geomechanical parameters. The relationships between JRC and
properties. correlated roughness were reported in literature(El-
The present article investigates the concept of multi- Soudani, 1978; Tse and Cruden, 1979; Maerz, Franklin
scale roughness measurements and derived fractal and Bennett, 1990; Yu and Vayssade, 1991; Tatone and
parameters for the purpose of geomechanical Grasselli, 2013) . For instance, (Tse and Cruden, 1979)
characterization and differentiation. Recursive multi-scale proposed a correlation between JRC and the root mean
analysis (i.e., moving window fractal measurements) is square of the first deviation of the profile (Z2) (Myers, 1962).
implemented for topographical characterization to It can be intuited from Equation [2] and other
differentiate textures and surface conditions. The comparable formulations that the results are severely scale
methodology was applied to a segmented part of an dependent. Figure 1 represents visualizations of obtained
advancing tunnel wall comprising of contrasting point roughness values for two different radiuses of
topographical conditions. Results showed that the measurement on the original scan. It can be seen that
proposed methodology was able to differentiate between radius of 0.1 m leads to smaller roughness values which
various parts of the surveyed area. Proper detection of the results in smoother interpretation of the superficial
concrete layer along the studied span, validates the characteristics, while the radius of 1 m yields higher range
applicability of the implemented scale-independent of roughness values.
approach.

2 BACKGROUND

Roughness geometrically refers to the local deviation from

planarity (or linearity in 2D) at a certain scale. In this work,
roughness of a of point (Pi) along a surface is computed as
its distance from a plane fitted for all points within a
distance 𝑟. The roughness value 𝑅𝑟,𝑖 (𝑚) for point 𝑃𝑖 =
[𝑥𝑖 , 𝑦𝑖 , 𝑧𝑖 ] at radius 𝑟 (m) is computed as follow: Roughness (0.1) Roughness (1)

0.000000 0.340449 0.000001

0.064100
(a) (b)
𝑎𝑟,𝑖 𝑥𝑖 + 𝑏𝑟,𝑖 𝑦𝑖 + 𝑐𝑟,𝑖 𝑧𝑖 + 𝑑𝑟,𝑖
𝑅𝑟,𝑖 =
2 2 2 [1] Figure 1: Point roughness visualization on the original scan
√𝑎𝑟,𝑖 + 𝑏𝑟,𝑖 + 𝑐𝑟,𝑖
at (a): r = 0.1 m and (b): r = 1m
Numerous studies were conducted based on fractal theory log -log plane is associated with the scale at which the
introduced by Mandelbrot (Mandelbrot, 1967; Mandelbrot considered processes is influenced by different features
and Wheeler, 1983) for quantification of topographical affecting the overall surface roughness
features at multiple scales (Belem, Homand-Etienne and These observations highlight those topographical
Souley, 1997; Fardin, Stephansson and Jing, 2001; Fardin, measurements of superficial conditions describe
Feng and Stephansson, 2004; Fardin, 2008; Lai, Samson characteristic features and trends. Such conditions are
and Bose, 2014). The basis of this approach lies in hereafter used to explicitly differentiate geological units and
identification of the recuring trends of surface geometry at represent in three-dimensional space the geomechanical
different scales (Mandelbrot and Wheeler, 1983; Shepard, and physical conditions of active workings in rock
Brackett and Arvidson, 1995). The related parameters can excavations.
then be implemented to define a relationship between
observed surface roughness characteristics at different 3 EXPERIMENTAL SETTINGS
scales.
The main trend of plotted statistical roughness values The investigated point cloud data set in this study was
(here RMS of point roughness, 𝜉𝑅 ) against scale (i.e., collected from an advancing underground tunnel with
radius of measurement, r) in log – log plane can be FARO X-130 laser terrestrial laser scanner (FARO, 2015).
interpreted by a regression analysis on a power law This device has measuring range of 130 m with highest
equation (Shepard et al., 2001): resolution of 1.5 mm point spacing per 10 m from the
scanner.
A section of the generated point cloud including a
𝜉𝑅 = 𝐴𝑟 𝐻 [3] concrete layer was selected to compare and contrast the
different superficial conditions. Table 1 presents the
where A represents the amplitude parameter that defines general characteristic parameters of the studied point
the differences or surface slope at a specified scale which cloud. The selected example contrasts very different
is one the common terms for describing a self-affine fractal interfaces in the form of an advancing rock tunnel and
object (Fardin, Stephansson and Jing, 2001). The other shotcrete surface. The application of the method provides
descriptor of the fractal objects is fractal dimension, D a first proof of concept towards interpreting characteristic
characterizing the scaling behaviour of a surface geometry topographical data sets to differentiate spatially different
which illustrates how roughness values varies with scale. interfaces along rock excavations.
This parameter is obtained from the exponent in Equation Figure 3 (a) and (b) show 3-dimensional surface model
[3], H calling Hurst exponent (Hurst, 1951)by the relation H of the studied advancing tunnel in form of point cloud data
= E – D, where E is the Euclidean dimension (E = 3 for a set and the segmented studied section, respectively.
surface and E = 2 for a profile) (Fardin, Stephansson and
Jing, 2001). Described fractal parameters are used as
representative parameters of the surface roughness
conditions.
Figure 2 shows a schematic relationship between 𝜉𝑅
and r in a log – log plane. If the plot gives an acceptable
straight line, the parameter H is directly obtained from the
slope of fitted line on the plot. The amplitude parameter, A
is also directly extracted from the constant coefficient of the
power law equation.

(a)

H
𝜉𝑅 (m)

r (m)

Figure 2: log – log plot of RMS of point roughness values,

𝜉𝑅 against radius of measurement, r

(Aubertin and Hutchinson, 2022) used multi-scale analysis

to derive the parameters from Equation [3] and quantify
topographical conditions of underground excavations. The
parameters were used to contrast excavation methods, (b)
rock damage, and ground control conditions. It was found
that the value of exponent from Equation [3] is related to Figure 3: (a) 3-dimensional surface model of the advancing
dominant geometrical features. In addition, it was tunnel, (b) segmented studied section
concluded that a change in the main trend of 𝜉𝑅 – r plot in
 Table 1: General characteristics of the studied point cloud Table 2: Fractal parameters for rock and concrete section

Parameter Value Fractal dimension Amplitude

Average point spacing 0.01 m Rock 2.2212 0.0369
Scan width 6.03 m concrete 1.8253 0.0332
Scan height 5.67 m
Number of points 178192 4.2 Recursive moving window analysis
Average surface density 4886.46 / m 2
Figure 5 shows a schematic representation of the
Average volume density 85519.33 / m3 implemented overlapping moving window approach where
spherical neighborhoods with radius l =0.5 m were
considered around each point (Pex)k, (𝑘 𝜖 [0, 𝑛], 𝑛 is the
4 COMPUTATIONAL METHODS number of data points) in the point cloud. Subsequently,
RMS of the measured roughness values at each radius ri
4.1 Multi-scale analysis (𝑖 𝜖 [0, 𝑚], 𝑚 is the number of radiuses of measurement)
for the points falling inside the spherical neighborhoods
The explicit physical differentiation of surface conditions were calculated and assigned to the investigated point
along the considered tunnel was realized through recursive ((RMS r=ri) k).
multi-scale analysis. A subsample of the surface of interest Eventually, all the points are assigned a number of
was extracted and analyzed in details to exemplify some of RMS values for all the considered radiuses of roughness
the procedural steps of this work. A span of 1.5 x 1.5 m of measurement. A computation layout of the explained
the concrete layer and rock section was segmentate methodology is presented by Table 3.
manually to highlight the computation method involved in
full face characterization. Recursive roughness
measurement was carried over a large range of radii using
the open-source software CloudCompare (Girardeau-
Montaut, 2016).
Figure 4 plots 𝜉𝑅 with respect to radius of measurement
r for the two subsampled sections of the tunnel. The rock
and concrete sections present different scale dependent
topographical characteristics described by the fractal
parameters derived from Equation [3] and [4]. The obtained Spherical
fractal parameters for these two sections are presented in neighborhoods with
Table 2. Multi-scale analysis results from these two radius l =0.5 m
subsamples of the considered tunnel surfaces
demonstrate that the obtained fractal dimension for the
concrete layer is less than that of rock section. The reason
is that the concrete layer is obviously smoother than the Figure 5: Schematic representation of the implemented
rock face and as indicated by (Mandelbrot, 1967; moving window approach for obtaining RMS of roughness
Mandelbrot and Wheeler, 1983), the fractal dimension has values of the points within neighborhood with radius of l =
a minimum value of 1 for a perfectly smooth profile and a 0.5 m around each point in point cloud
maximum value of less than 2 for an extremely rough
undulating profile.
Table 3: Computation layout for the proposed
methodology

𝑦 = 0.0369 𝑥 0.7788 Points r1 r2 … rm

1 (RMS r=r1)1 (RMS r=r2)1 … (RMS r=rm)1

2 (RMS r=r1)2 (RMS r=r2)2 … (RMS r=rm)2

… … … … …
𝑦 = 0.0332 𝑥 1.1747 n (RMS r=r1) n (RMS r=r2) n … (RMS r=rm) n

Figure 6 represents the obtained RMS values for each

point against radius of measurement in a log-log plane.
Figure 4: 𝜉𝑅 – r plot for rock section and concrete section Fractal parameters for each spherical neighborhood were
obtained from the power law regression analysis.
 A

0.03664

𝑦 = 0.0834𝑥 + 0.0033
𝜉𝑅 (m)

r (m)

Figure 6: An example RMS (r) – r plot for a random

spherical neighborhood in the point cloud (b) 0.00006

Resulting fractal dimension D, amplitude parameter A

and the ratio of fractal dimension to the amplitude D/A
parameter D/A for each neighborhood were assigned to
their associated reference points and used as scalar field 13783.33
color codes to be visualized on the scan. These results are
illustrated by Figure 7.

5 RESULTS AND DISCUSSION

As Figure 7 represents, the proposed methodology was

able to differentiate the concrete layer on the scan.
However, the ratio of fractal dimension to the amplitude
parameter shows the most differentiable results in
(c) 78.057
comparison with visualization of other parameters
individually. Observed results indicates that both the fractal
parameters contribute to evaluation of the surface
roughness and each of them alone are not representative Figure 7: Visualization of (a) fractal dimension D, (b) the
enough for the differentiation purpose. amplitude parameter A, and (c) the ratio of fractal
dimension to the amplitude parameter D/A on the original
scan
D
6 CONCLUSION
2.860

A scale-dependent topographical characterization method

using fractal-based analysis was presented to develop an
automated and effective means of differentiation of
topographical conditions along the3-dimentional surface
models of rock outcrops generated by LiDAR scanners.
The proposed methodology was able to provide
meaningful and scale-independent characteristic
parameters for discrimination between the natural and
artificial (e.g., ground control material) features along the
(a)
0.827 studied rock surface in an underground tunnel.
For the validation of the proposed approach, back
analysis of known topographical condition of a part of 3 -
dimensional surface model of the advancing tunnel
consisting of a concrete layer was performed to evaluate
differences in form of scale dependent fractal parameters.
The major framework of this study was subdivided into
two main sections:

I. Performing a recursive roughness measurement

analysis for a number of radiuses to the
segmented sections of the studied area with
 known topographical conditions and plotting the Bizjak, K.F. and Geršak, A. (2018) ‘Quantified joint
RMS of obtained point roughness values against surface description and joint shear strength of small rock
the associated radiuses of measurements in log – samples’, Geologija, 61(1), pp. 25–32.
log plane and getting fractal parameters for each Committee, C.G.Society.F. (1985) Canadian
segmented spans to see the differences in terms Foundation Engineering Manual. Canadian Geotechnical
of these parameters. Society. Available at:
II. implementing the proposed approach to small https://books.google.ca/books?id=vFmJQgAACAAJ.
overlapping moving neighborhood sections on the El-Soudani, S.M. (1978) ‘Profilometric analysis of
scan and obtaining fractal parameters for all these fractures’, Metallography, 11(3), pp. 247–336. Available at:
sections and visualizing on the original scan. https://doi.org/https://doi.org/10.1016/0026-
0800(78)90045-9.
It was observed that the obtained fractal parameters for the Fardin, N. (2008) ‘Influence of structural non-
concrete layer are different from those obtained for the rock stationarity of surface roughness on morphological
section. The concrete layer's smaller fractal dimension characterization and mechanical deformation of rock
validates the practicality of the presented method for joints’, Rock Mechanics and Rock Engineering, 41(2), pp.
determining the smoother surface. 267–297. Available at: https://doi.org/10.1007/s00603-
Results from the second part of the study shows that 007-0144-9.
visualizing obtained fractal parameters for the spherical Fardin, N., Feng, Q. and Stephansson, O. (2004)
neighborhoods around the points on the original scan in ‘Application of a new in situ 3D laser scanner to study the
form of scalar field color codes is a promising way for scale effect on the rock joint surface roughness’,
differentiation of different topographical conditions of International Journal of Rock Mechanics and Mining
surveyed areas from the LiDAR scans in form of point cloud Sciences, 41(2), pp. 329–335. Available at:
data sets. However, visualization of the ratio of the https://doi.org/10.1016/S1365-1609(03)00111-4.
obtained fractal parameters on the scan provided better Fardin, N., Stephansson, O. and Jing, L. (2001) The
differentiation results between the concrete layer and the scale dependence of rock joint surface roughness,
rock section than either of these parameters separately International Journal of Rock Mechanics & Mining
indicating that both of the fractal parameters are needed to Sciences.
describe and differentiate scale-dependent topographical FARO (2015) ‘FARO, 2015. FARO Laser Scanner
behavior of rock surfaces. Focus 3D x 130. DallasNo Title’.
To increase the reliability of proposed methodology, it Francioni, M. et al. (2019) ‘A new fast and low-cost
is required to find the best descriptive relationship between photogrammetry method for the engineering
fractal parameters which could provide the most precise characterization of rock slopes’, Remote Sensing, 11(11),
results. Moreover, depending on resolution and density of p. 1267.
the scans, an automated detection of the optimized radius Girardeau-Montaut (2016) CloudCompare V2.11.
of spherical neighborhoods are required which will be Glenn, N.F. et al. (2006) ‘Analysis of LiDAR-derived
investigated in the future works of the authors. topographic information for characterizing and
differentiating landslide morphology and activity’,
7 REFERENCES Geomorphology, 73(1–2), pp. 131–148.
Grasselli, G. and Egger, P. (2003) ‘Constitutive law for
Aubertin, J.D. et al. (2017) ‘Application of LiDAR the shear strength of rock joints based on three-
technology for the investigation of rock mass response to dimensional surface parameters’, International Journal of
blasting’. Rock Mechanics and Mining Sciences, 40(1), pp. 25–40.
Aubertin, J.D. and Hutchinson, D.J. (2022) ‘Scale- Haneberg, W.C. (2007) ‘Directional roughness profiles
dependent rock surface characterization using LiDAR from three-dimensional photogrammetric or laser scanner
surveys’, Engineering Geology, p. 106614. point clouds’, in 1st Canada-US rock mechanics
Aubertin, J.D., Hutchinson, D.J. and Diederichs, M.S. symposium. OnePetro.
(2019) ‘Application of LiDAR technology for the Hoek, E. and Brown, E.T. (1980) ‘Empirical strength
assessment of mechanical scaling conditions in criterion for rock masses’, Journal of the geotechnical
underground mines and tunnels’, Geo St-John’s [Preprint]. engineering division, 106(9), pp. 1013–1035.
Barton, N. and Choubey, V. (1977) ‘The shear strength Hurst, H.E. (1951) ‘Long-term storage capacity of
of rock joints in theory and practice’, Rock mechanics, reservoirs’, Transactions of the American society of civil
10(1), pp. 1–54. engineers, 116(1), pp. 770–799.
Belem, T., Homand-Etienne, F. and Souley, M. (1997) Jaboyedoff, M. et al. (2012) ‘Use of LIDAR in landslide
‘Fractal analysis of shear joint roughness’, International investigations: a review’, Natural Hazards, 61(1), pp. 5–28.
Journal of Rock Mechanics and Mining Sciences, 34(3–4), Available at: https://doi.org/10.1007/s11069-010-9634-2.
pp. 130-e1. Kim, D.H., Gratchev, I. and Balasubramaniam, A.
Berti, M., Corsini, A. and Daehne, A. (2013) (2013) ‘Determination of joint roughness coefficient (JRC)
‘Comparative analysis of surface roughness algorithms for for slope stability analysis: a case study from the Gold
the identification of active landslides’, Geomorphology, Coast area, Australia’, Landslides, 10(5), pp. 657–664.
182, pp. 1–18. Lai, P., Samson, C. and Bose, P. (2014) ‘Surface
roughness of rock faces through the curvature of
triangulated meshes’, Computers & Geosciences, 70, pp. Uotinen, L. et al. (2021) ‘Photogrammetric Prediction of
229–237. Rock Fracture Properties and Validation with Metric Shear
Lato, M. et al. (2009) ‘Engineering monitoring of rockfall Tests’.
hazards along transportation corridors: using mobile Yu, X. and Vayssade, B. (1991) ‘Joint profiles and their
terrestrial LiDAR’, Natural Hazards and Earth System roughness parameters’, in International journal of rock
Sciences, 9(3), pp. 935–946. mechanics and mining sciences & geomechanics
Maerz, N.H., Franklin, J.A. and Bennett, C.P. (1990) abstracts. Elsevier, pp. 333–336.
‘Joint roughness measurement using shadow profilometry’,
International Journal of Rock Mechanics and Mining
Sciences & Geomechanics Abstracts, 27(5), pp. 329–343.
Available at: https://doi.org/https://doi.org/10.1016/0148-
9062(90)92708-M.
Mandelbrot, B. (1967) ‘How long is the coast of Britain?
Statistical self-similarity and fractional dimension’, science,
156(3775), pp. 636–638.
Mandelbrot, B.B. and Wheeler, J.A. (1983) ‘The Fractal
Geometry of Nature’, American Journal of Physics, 51(3),
pp. 286–287. Available at: https://doi.org/10.1119/1.13295.
McKean, J. and Roering, J. (2004) ‘Objective landslide
detection and surface morphology mapping using high-
resolution airborne laser altimetry’, Geomorphology, 57(3–
4), pp. 331–351.
Myers, I.B. (1962) ‘The Myers-Briggs Type Indicator:
Manual (1962).’
Nilsson, M., Edelbro, C. and Sharrock, G. (2012) ‘Small
scale joint surface roughness evaluation using digital
photogrammetry’, in ISRM International Symposium-
EUROCK 2012. OnePetro.
Poropat, G. V (2008) ‘Remote characterisation of
surface roughness of rock discontinuities’, in SHIRMS
2008: Proceedings of the First Southern Hemisphere
International Rock Mechanics Symposium. Australian
Centre for Geomechanics, pp. 447–458.
Shepard, M.K. et al. (2001) ‘The roughness of natural
terrain: A planetary and remote sensing perspective’,
Journal of Geophysical Research E: Planets, 106(E12), pp.
32777–32795. Available at:
https://doi.org/10.1029/2000JE001429.
Shepard, M.K., Brackett, R.A. and Arvidson, R.E.
(1995) ‘Self-affine (fractal) topography: Surface
parameterization and radar scattering’, Journal of
Geophysical Research: Planets, 100(E6), pp. 11709–
11718. Available at:
https://doi.org/https://doi.org/10.1029/95JE00664.
Tatone, B.S.A. and Grasselli, G. (2013) ‘An
investigation of discontinuity roughness scale dependency
using high-resolution surface measurements’, Rock
Mechanics and Rock Engineering, 46(4), pp. 657–681.
Available at: https://doi.org/10.1007/s00603-012-0294-2.
Travelletti, J. et al. (2008) ‘Monitoring landslide
displacements during a controlled rain experiment using a
long-range terrestrial laser scanning (TLS)’, International
Archives of Photogrammetry and Remote Sensing, 37, pp.
485–490.
Tse, R. and Cruden, D.M. (1979) ‘Estimating joint
roughness coefficients’, International Journal of Rock
Mechanics and Mining Sciences & Geomechanics
Abstracts, 16(5), pp. 303–307. Available at:
https://doi.org/https://doi.org/10.1016/0148-
9062(79)90241-9.
Wednesday, October 4, 2023

UNSATURATED SOILS
Numerical study of the influence of suction
increment duration upon the water retention
curve in a Tempe cell

Arij Krifa, Robert P. Chapuis

Département CGM, Polytechnique Montréal, Montréal (Québec),
Canada

ABSTRACT
Understanding unsaturated soils is crucial in geotechnical and agricultural engineering. The water retention curve (WRC),
which links soil water content and matric suction, is a key parameter in modeling soil behavior. The WRC is typically
measured with long laboratory tests using devices like the Tempe cell, which applies suction increments over time.
However, the 24-hour suction increment recommended by the ASTM D6836-16 standard may not be sufficient for soils
like silty sands, especially when approaching their water entry value (WEV). This paper presents a numerical study on the
WRC of silty sands using the Tempe cell to show that longer drainage periods are necessary to obtain an accurate WRC,
particularly for suctions approaching or exceeding the WEV.

RÉSUMÉ
Comprendre les sols non saturés est crucial en génie géotechnique et agricole. La courbe de rétention d'eau (CRE), qui
relie la teneur en eau du sol et la succion matricielle, est un paramètre clé dans la modélisation du comportement du sol.
La CRE est généralement mesurée au cours de longs tests de laboratoire à l'aide de dispositifs tels que la cellule Tempe,
qui applique des incréments de succion au fil du temps. Cependant, l'incrément de succion de 24 heures recommandé
par la norme ASTM D6836-16 peut ne pas être suffisant pour, les sols tels que les sables silteux, en particulier lorsqu'ils
approchent leur valeur d'entrée d'eau (WEV). Cet article présente une étude numérique sur la CRE des sables limoneux
à l'aide de la cellule Tempe pour montrer que des périodes de drainage plus longues sont nécessaires pour obtenir une
CRE présice, en particulier pour les succions approchant ou dépassant la WEV.

1 INTRODUCTION deux incréments successifs dans le but de donner

l’occasion à l’échantillon d’atteindre son équilibre. Dans ce
La courbe de rétention d'eau est une représentation contexte, Vogel et Cislerova (2007) ont trouvé qu’un
graphique de la relation entre la teneur en eau volumique incrément de succion plus court mène à des mesures plus
et la succion matricielle d'un échantillon de sol. précises de la courbe de rétention et réduit l’influence des
La CRE est un outil crucial dans l'étude des sols non effets dépendants du temps. Également, Ramezani et al.
saturés car elle fournit des informations sur la conductivité (2018) ont montré que les mesures de la courbe de
hydraulique du sol, sa résistance au cisaillement et ses rétention d'eau du sol sont plus précises lorsque les durées
caractéristiques de déformation. d'incréments de succion sont plus courtes, et déclarent
Cette mesure peut être réalisée à l'aide de différentes qu’une durée d'une heure donnait de meilleurs résultats.
méthodes, telles que la méthode de la centrifugation, la En comparaison, les durées d'incréments de succion plus
méthode du cylindre en pression, la méthode du papier longues, comme celles de 4 heures, peuvent conduire,
filtre, et la méthode de la cellule Tempe. Cette dernière selon eux, à des mesures moins précises.
consiste à faire chasser l’eau du spécimen en lui appliquant On peut remettre en question la durée raccourcie pour
des incréments de succion. La durée de ces incréments les mesures de la courbe de rétention d'eau du sol, en
présente une influence majeure sur la courbe de rétention particulier pour les sols tels que les sables silteux, qui
d’eau obtenue. En effet, Schindler et Müller (1997) notent deviennent très imperméables lorsqu'ils sont soumis à des
que la durée du test peut varier en fonction des propriétés succions approchant leur WEV. L’évacuation de l'eau
du sol et du niveau de précision souhaité, mais qu'il faut devient extrêmement lente et le volume d'eau libéré est si
généralement plusieurs jours pour effectuer un seul test en petit qu’il est difficile à mesurer.
utilisant une cellule Tempe. Selon Fredlund et Rahardjo Dans cet article, une étude numérique est menée sur la
(2012) la mesure de la CRE du sol est une tâche courbe de rétention d'eau obtenue par la cellule de Tempe
chronophage qui implique de tester un échantillon de sol à sur des sables silteux, dans le but de prouver qu'une
diverses teneurs en eau et valeurs de succion. Le temps période de succion de 24 heures n'est en fait pas suffisante
d'équilibration nécessaire pour que l'échantillon de sol pour compléter le drainage. Pour le sable silteux, l'erreur
atteigne une teneur en eau ou une valeur de succion sur la courbe de rétention d'eau expérimentale est
spécifique peut prendre plusieurs jours, en fonction du type quantifiée en fonction de la durée des étapes de drainage,
de sol et des conditions de test. pour la partie de la courbe de rétention d'eau entre sa AEV
D’autre part, certaines études proposent de raccourcir (air entry value) et sa WEV, puis pour la partie des succions
l’essai en adoptant des durées d’incréments de succion supérieures à la WEV.
plus courtes et de maintenir des périodes de repos entre
2 MATÉRIEL ET MÉTHODES 2,3.10-9 m/s qui correspond à la conductivitéé hydraulique
obtenue pour la WEV (Figure 2)
2.1 Logiciel utilisé À noter que la valeur de la conductivité hydraulique
saturée a été calculée en prenant la moyenne des valeurs
Dans le but de modéliser une vraie cellule Tempe, le obtenues à partir des méthodes de prédiction de Ksat de
logiciel d'éléments finis Seep/W (Geo-slope 2007) a été Hazen (1892) couplée avec Taylor (1948), Navfac (1974)
utilisé pour modéliser l'écoulement saturé et non saturé en et Chapuis (2004, 2008, 2012).
utilisant les équations complètes que lui sont associées.

2.2 Méthodologie

Le modèle numérique présenté durant cette étude est basé

sur un modèle physique simple qui simule l’écoulement
non saturé en régime transitoire à travers une cellule
Tempe mise en succion par paliers.
On illustre par le Tableau 1 suivant les propriétés du
sable silteux utilisé dans les simulations numériques.

Tableau 1. Caractéristiques du sable silteux

d10 d60 AEV WEV Ksat
e n θr
(mm) (mm) (kPa) (kPa) (m/s)
0.018 0.21 0.289 0.224 20 60 0.169 2.10-6

Pour le spécimen du sable silteux testé dans cette

étude, on a choisi de prédire la courbe de rétention du sol
en question selon le modèle prédictif de Kovács modifié
(MK) proposé par Aubertin et al. (2003) qui décrit la CRE
pour les sols non cohésifs dans lequel le degré de
Figure 1. La teneur en eau volumique θ (m3/m3) en fonction
saturation Sr est composé de deux composantes : une
de la succion matérielle (kPa)
composante capillaire Sc, et une composante adhésive Sa,
qui sont reliées par : 1.E-05

𝜃 = 𝑛[1 − 〈1 − 𝑆𝑎 〉(1 − 𝑆𝑐 ) [1]

Hydraulic Conductivity (m/s)

Dans lequel 〈〉 représente les crochets de Macaulay, ce 1.E-06

que signifie que 〈𝑥〉 = 0.5(𝑥 + |𝑥|). Les composantes Sa
and Sc sont définies comme suit :

ℎcoMK 2
𝑚MK
ℎcoMK 2 1.E-07
𝑆𝑐 = 1 − [( ) + 1] exp⁡ [−𝑚MK ( ) ] [2]
𝜓 𝜓

𝜓 ℎcoMK 2/3
ln⁡(1+ ) ( )
𝑆𝑎 = 𝑎𝑐MK [1 −
𝜓𝑟MK
] 𝜓𝑛
[3] 1.E-08
𝜓0 𝜓 1/6
ln⁡(1+ ) 1/3
𝑒 ( )
𝜓𝑟MK 𝜓𝑛

Dans les équations précédentes, ℎcoMK and 𝜓𝑟MK sont

1.E-09
associés à 𝑒𝑑10 et au coefficient d’uniformité 𝐶𝑈 , 𝑚MK est
1 10 100
relié à 𝐶𝑈 .Pour 𝑎𝑐MK on a associé la valeur de 0.01 et
Suction (kPa)
1E107 pour 𝜓0 .
Ce modèle a été choisi car il représente le mieux la Figure 2. Conductivité hydraulique K(m/s) en fonction de la
CRE pour les sables silteux surtout dans la zone résiduelle succion matricielle (kPa)
de la courbe où la succion appliquée atteint des valeurs
élevées. La courbe de rétention d’eau du sable silteux
étudié prédite selon le modèle décrit précédemment est La valeur de la conductivité hydraulique saturée de la
illustrée dans la Figure. 1. pierre poreuse KPP doit être celle d'une pierre poreuse qui
Après avoir obtenu la courbe de rétention du sable restera saturée pour la succion maximale visée dans
silteux étudié, on a pu estimer la courbe de la conductivité l'essai Tempe. Une pierre trop peu perméable va gêner ou
non saturée K(u) en utilisant le logiciel Seep/W. La CRE retarder le drainage du sol qui est testé. Cependant, la
ainsi que la valeur de la conductivité saturée Ksat ont été pierre poreuse ne peut pas être trop grossière. Cette valeur
utilisées pour générer la courbe K(u) en se basant sur le a été choisie en respectant le catalogue des pierres
modèle de van Genuchten et en la limitant à la valeur de poreuses en céramique de soilmoisture.com et ceci dans
le but de pouvoir réaliser les mêmes essais au laboratoire est constante et ceci fait le logiciel trouve de la difficulté à
et de bien valider les résultats numériques obtenus. bien converger.
La succion maximale visée dans cette étude est de
l’ordre de 1000 kPa. Pour cela on a choisi la céramique 2.3.3 Choix du maillage
(B15M1) qui peut rester saturée pour une telle pression vue
que sa valeur d’entrée d’air est de l’ordre de 220 psi qui est Dans le but d’avoir les résultats les plus optimisés possible
équivalent à une succion de 1500 kPa. Cette pierre on a procédé par une étude sur la taille du maillage le plus
poreuse possède aussi une conductivité hydraulique convenable. Pour cela on a fait tourner, en régime
saturée de 2.59.10-9 cm/s permanent, le même modèle avec différentes tailles de
grille et on a comparé les courbes qu’on obtient. On a noté
2.3 Création du modèle numérique qu’à partir d’une taille de 1 cm les courbes obtenues
suivent bien la courbe de rétention d’eau du spécimen
2.3.1 Géométrie de la cellule Tempe prédite par le modèle MK. On a adopté une grille
rectangulaire de taille de 1 cm tout au long de l’étude.
La cellule Tempe est de diamètre de 6 cm et d’une hauteur
totale de 5 cm (Figure 3). Dans la cellule on dispose d’une 2.3.4 Essais de la mesure de la CRE
chambre dans laquelle on introduit le spécimen de sol
qu’on veut mettre en succion qui est posé sur une pierre Une fois qu’on a bien configuré le modèle et que tous les
poreuse qui selon la norme ASTM-D6836 doit rester paramètres nécessaires pour le faire tourner ont été
saturée pour la succion maximale visée durant l’essai. choisis, on a fait tourner le modèle en faisant changer à
Le spécimen est séparé de la pierre poreuse par chaque fois l’incrément de succion. On a opté pour trois
l’introduction d’un élément interface d’une épaisseur de 2 durées d’incréments : 24 heures, 48 heures et 72 heures.
mm. Cet élément est important car il va permettre la Une fois que la simulation numérique de l’essai est
distribution égale de la succion du spécimen vers la pierre terminée on a procédé par le traçage des courbes de
poreuse dans le but de modéliser une cellule Tempe qui rétention d’eau résultantes. Et ce en calculant pour chaque
s’approche de ces caractéristiques le plus possible à la étape de succion la valeur de la teneur en eau volumique
réalité. en appliquant une moyenne de Riemann sur l’ensemble
des valeurs de la teneur en eau calculées par le logiciel
dans chaque nœud du maillage sur une ligne verticale.

2.3.5 Calcul des volumes sortants

Pour des raisons de validation des résultats obtenus des

courbes de rétention d’eau, on a aussi calculé les volumes
d’eau cumulés dans chaque étape de succion et également
pour chaque incrément de temps. En utilisant les vitesses
de Darcy de la sortie d’eau calculées par le logiciel dans
l’élément interface qui sépare le spécimen de la pierre
poreuse. Selon l’expression suivante :

𝑉1 +𝑉2
𝑑𝑣𝑜𝑢𝑡 = ( ) . (𝑡2 − 𝑡1 ). 𝐴 [4]
2

Dans l’équation précédente V1 et V2 désignent les

vitesses de Darcy pour les temps t1 et t2, A est la section
Figure 3. Géométrie de la cellule Tempe et position de
de la cellule Tempe.
l’élément interface

2.3.2 Conditions aux frontières 3 RÉSULTATS

3.1 Courbes de rétention d’eau

Le spécimen sera mis en succion par incréments. En effet,
on a introduit à la base de la cellule des charges
hydrauliques négatives qui simulent des succions. On a fait Les résultats obtenus confirment tous qu’un incrément de
24h n’est pas suffisant pour le drainage complet du
que chaque mise en succion dure un temps bien spécifique
spécimen testé du sable silteux après chaque étape de
qu’on peut facilement contrôler par le logiciel. Les charges
mise en succion. Les courbes obtenues sont toutes tracées
ont été choisies de façon à bien décrire la courbe de
rétention d'eau du sol étudié et qui progressent sur le même graphique (Figure 4) pour illustrer l’influence
de l’incrément du temps. Celles-ci montrent bel et bien que
géométriquement.
la durée de la mise en succion est d’une importance
On a assuré aussi que les premières succions
majeure dans la mesure de la courbe de rétention.
appliquées présentent des valeurs proches pour assurer
une bonne convergence initiale du logiciel vu que dans la Les courbes montrent que dans la zone de saturation
la durée de la mise en succion ne présente pas une
zone de saturation la valeur de la teneur en eau volumique
influence sur la mesure précise de la courbe de rétention
d’eau. Les courbes se superposent parfaitement et l’AEV Dans cette analyse, plusieurs observations ont été
est bel et bien respectée. Une fois que la succion dépasse faites sur l'erreur expérimentale dans la mesure de la
l’AEV on commence à voir une différence dans les valeurs caractéristique de rétention d'eau (CRE) en fonction de la
de la teneur en eau volumique. En effet la différence est succion appliquée au spécimen. Tout d'abord, il a été
visiblement remarquable à partir des succions qui remarqué que l'erreur pour le palier de 24 heures est la plus
dépassent la valeur d’entrée d’air. De même, la CRE de importante parmi les trois paliers adoptés. Cela pourrait
l’incrément de succion de 24h est la plus éloignée de la s'expliquer par le fait que la durée de cet incrément de
courbe réelle du matériau. Plus la durée de la mise en succion est la plus courte, ce qui laisse moins de temps
succion est importante, plus on s’approche de la courbe pour la répartition uniforme de l’eau dans l’échantillon.
réelle du matériau. Le décalage le plus flagrant réside dans De plus, il a été constaté que plus la durée des paliers
la zone résiduelle de la CRE. En effet, les courbes est importante, moins l'erreur est significative. Cela pourrait
simulées par le logiciel présentent toutes les trois (24 h, 48 être attribué au fait que des durées plus longues
h et 72 h) un palier et elles sont presque horizontales, ceci permettent une meilleure stabilisation des conditions du
est indication majeure que même avec des incréments de sol, réduisant ainsi les sources d'erreurs expérimentales.
temps très importants, voire dépassant les 72 h, le Il est également intéressant de noter que pour les
drainage de l’échantillon demeure incomplet pour des succions qui ne dépassent pas les 10 kPa, qui
succions importantes et que les essais raccourcis de la correspondent à l’AEV du sol, les erreurs expérimentales
mesure de la courbe de rétention d’eau via la cellule sont négligeables. En effet, dans la zone de saturation, les
Tempe sont en effet inachevés et les résultats qui en erreurs sont difficiles à identifier et à quantifier en raison de
découlent ne sont pas fiables. la stabilité des conditions du sol.
Dans le but de concrétiser les résultats obtenus on a En examinant la CRE pour une succion entre 20 et 100
procédé par un calcul de l’erreur selon la méthode de kPa, on a observé que l'erreur présente un maximum pour
l’erreur quadratique moyenne (EQM) entre les courbes les trois incréments. Cette zone correspond à la zone
simulées et la courbe input (Figure 5). transitoire de la CRE, où les propriétés du sol changent
rapidement en réponse à de petites variations de la teneur
0.25 en eau. Cette observation souligne l'importance de
mesurer avec précision la CRE dans cette plage de
Volumetric Water Content

0.2 succion.
Enfin, il est intéressant de noter que pour une succion
supérieure à 200 kPa, la courbe de l'erreur quadratique
0.15 moyenne (EQM) prend l'allure d'une droite linéaire de
(m3/m3)

pente positive (Figure 5). Cette observation pourrait être

interprétée comme un signe de l'inachèvement de l'essai
0.1
et que la courbe de rétention d'eau est fortement influencée
Input values par la durée de l'incrément de succion. Cela souligne
0.05 fem 24 h steps l'importance de bien définir les conditions expérimentales
fem 48 h steps pour obtenir des mesures précises de la CRE
fem 72 h steps
0 3.2 Résultats des volumes sortants cumulés
10 1 100 1000
Suction (kPa)
En raison du grand nombre de résultats similaires, il n'est
Figure 4. Les CRE obtenues par simulations numériques pas possible de les présenter tous de manière détaillée
pour les incréments 24 h, 48 h et 72 h en comparaison avec dans ce papier. La préoccupation majeure était portée sur
la courbe prédite la zone résiduelle où on a trouvé non seulement que
l’erreur augmente d’une façon régulière mais que les CRE
7.E-03
fem 24 h steps présentent un grand écart avec la courbe de rétention
6.E-03 fem 48 h steps prédite.
Pour des raisons de comparaison on va aussi illustrer
Mean Squared Error

fem 72 h steps
5.E-03 les résultats obtenus pour la succion appliquée de 10 kPa
qui est une succion qui réside dans la zone ou les erreurs
4.E-03 sur les courbes sont pratiquement nulles (Figure 6).
3.E-03 La courbe du volume sortant cumulé en fonction du
temps pour une succion de 10 kPa présente un palier
2.E-03 horizontal, ce qui indique que le drainage du spécimen est
achevé pour cette valeur de succion. Cette observation est
1.E-03 importante car elle permet de déterminer le point à partir
0.E+00 duquel l'échantillon a été entièrement drainé pour une
succion appliquée. Ainsi, pour cette valeur de succion, on
10 100 1 1000
Suction (kPa) peut affirmer en toute sécurité que le palier de l'essai est
complet. Cependant, pour les succions qui dépassent la
Figure 5. L’erreur quadratique moyenne en fonction de la
valeur d'entrée d'eau, qui est de l'ordre de 60 kPa, les
succion (kPa) pour les incréments de 24 h ,48 h et 72 h
courbes du volume sortant en fonction du temps présentent
à la fin de la durée de l'incrément de succion une pente s'écoule de l’échantillon, même si elle est présente en
presque linéaire positive. Cette observation suggère que quantités infimes.
pour des succions importantes, qui dépassent Cette constatation peut avoir des implications
essentiellement la WEV, et en adoptant un incrément de importantes dans la compréhension des processus de
succion de 24 heures, le drainage demeure incomplet et transport de l'eau dans les sols et les roches, ainsi que
l'eau excédentaire avance si lentement qu’elle reste dans dans la modélisation de ces processus.
le spécimen du sable silteux testé (Figure 6).
1.E+00
24h suction increment
1.E+00

Cumulative Outflow Volume (cm3)

AP = 10 kPa
1.E-01 48h suction increment
Cumulative Outflow Volume (cm3)

1.E-01 AP = 200 kPa 1.E-02 72h suction increment

1.E-02 AP = 500 kPa

1.E-03
AP = 1000 kPa
1.E-03 1.E-04
1.E-04 1.E-05
1.E-05 1.E-06
1.E-06 1.E-07
1.E-07 1.E-08
1.E-08 1 100 10000 1000000
Time (s)
1 100 10000
Time (s) Figure 7. Le volume sortant cumulé (cm3) en fonction du
temps (s) pour la succion appliquée de 1000 kPa et pour
Figure 6. Les volumes sortants cumulés (cm3) pour les incréments de succion de 24 h,48 h et 72 h
l’incrément de succion de 24 h pour les succions
appliquées de 10, 200, 500 et 1000 kPa en fonction du
temps (s) 4 DISCUSSION

Les cellules Tempe sont connues pour leur capacité à

Cette constatation est significative car elle montre que maintenir une teneur en eau à peu près constante dans
pour des valeurs de succion importantes, le drainage peut l'échantillon de sol pendant toute la durée de la mesure, ce
ne pas être complet même après une durée d'incrément de qui n'est pas toujours le cas avec d'autres méthodes. Cela
succion de 24 heures (durée recommandée par la norme permet d'obtenir des mesures plus précises de la teneur en
ASTM). Cela peut avoir des implications importantes dans eau à différentes succions matricielles.
la compréhension des processus de transport de l'eau Cependant, la mesure des courbes de rétention d'eau
dans les sols et les roches, ainsi que dans la modélisation à l'aide de cellules Tempe peut être un processus long, et
de ces processus. En effet, il est essentiel de comprendre la précision des résultats peut être affectée par des
comment l'eau se déplace dans les sols et les roches pour facteurs tels que la taille de l'échantillon et le type de sol
évaluer leur comportement hydromécanique et leur étudié et plus particulièrement par la durée de la mise en
réponse aux sollicitations extérieures. succion. Les simulations numériques sont alors proposées
Les courbes des volumes sortants cumulés en fonction pour des essais de mesure de rétention d’eau avec des
du temps pour les trois paliers (24 h, 48 h et 72 h) pour la incréments de succions assez longs dans le but de mettre
succion appliquée maximale visée dans cet essai qui est en évidence leur importance quant au comportement de la
d’une valeur de 1000 kPa sont illustrés ci-dessous pour voir CRE.
en détail le comportement de l’eau à la fin de la mise en On a pu observer dans cette étude que pour avoir une
succion (Figure 7) . mesure crédible de la courbe de rétention d’eau il faut
Les courbes présentées dans la Figure 7 montrent un envisager des incréments de succion qui sont assez lourds
comportement intéressant en ce qui concerne l'évacuation en temps. Par conséquent il est primordial de ne pas
de l'eau de l'échantillon. En effet, la pente de ces courbes arrêter prématurément les essais de succion sur un sol
reste positive tout au long de la durée de la mise en silteux, car cela peut entraîner des erreurs expérimentales
succion. Cela indique que l'échantillon continue à évacuer importantes. Par exemple, si l'on effectue un essai de
de l'eau, même si le volume sortant devient très faible et courbe de rétention sur un spécimen de sable silteux à
difficile à mesurer. l'aide d'une cellule Tempe, et que l'on mesure sa masse
Cette observation est importante car elle suggère que avec une balance de précision bien déterminée à la fin de
l'évacuation de l'eau de l'échantillon ne s'arrête pas chaque journée, il est crucial de ne pas arrêter l'essai avant
complètement, même lorsque le volume sortant devient que le drainage soit complet. Si l'on arrête l'essai trop tôt,
très faible. En d'autres termes, il y a toujours de l'eau qui on risque de sous-estimer le poids du spécimen, ce qui
faussera les résultats. De plus, il est important d’exécuter
une étape d'étalonnage de la balance pour s'assurer de sa Navfac (1974). Design Manual DM7- Soil Mechanics,
précision, car même de légères variations de poids Foundations, and Earth Structures. U.S. Government
peuvent avoir un impact significatif sur les résultats de Printing Office, Washington, D.C.
l'essai. En somme, il est essentiel d'être rigoureux dans la Ramezani, H., Parsinejad, M., & Taghizadeh-Mehrjardi, R.
réalisation des essais de succion sur sol silteux afin (2018). Effect of suction increment duration on soil
d'obtenir des résultats fiables et précis. water retention curve using Tempe cell. Journal of
Agricultural Science and Technology, 20(7), 1361-1371
Schindler, U., & Müller, L. (1997). Advantages and
5 CONCLUSION limitations of Tempe cells for determining soil water
retention curves. Soil Science Society of America
Dans ce papier, on a pu montrer à l’aide des simulations Journal, 61(4), 1024-1030.
numériques d’un essai de mesure de la courbe de rétention Tuller, M., & Or, D. (2002). Retention curve hysteresis and
d’eau d’un sable silteux par une cellule Tempe qu’un its effect on soil hydraulic properties. Water Resources
incrément de succion de 24 heures, qui est généralement Research, 38(12), 1294.
adopté dans les essais de laboratoire est insuffisant pour Vogel, T., & Cislerova, M. (2007). Time-dependent effects
pouvoir chasser toute goutte d’eau du spécimen testé. Les in the Tempe cell method for soil water retention
incréments de succion de 48 heures et de 72 heures ont measurement. Soil Science Society of America Journal,
été testés également et on a aussi constaté que même 71(3), 735-740.
avec des durées de mise en succion aussi importantes les
courbes de rétention demeurent non fiables. Il est
recommandé de voir les résultats qui peuvent découler
d’une étude qui tient en compte des incréments de succion
plus chronophages voir même une semaine ou plus.
En se basant sur les résultats illustrés dans ce
document, les auteurs trouvent que, pour des sols qui
deviennent très imperméables lorsqu’ils sont soumis à des
succions qui approchent et dépassent leur WEV, le
standard de l’application d’un incrément de succion de 24
h doit être révisé.
On avise que les simulations numériques ont des
limites en raison de leur dépendance aux paramètres
d'entrée et aux hypothèses faites lors de la modélisation.
Pour garantir l'exactitude des prévisions, la validation
expérimentale est nécessaire pour identifier les écarts
entre les résultats de simulation et le comportement réel du
système. Cette combinaison peut fournir une meilleure
compréhension des systèmes complexes.

6 RÉFÉRENCES

ASTM (2016). D. Standard Test Methods for Determination

of the Soil Water Characteristic Curve for Desorption
Using Hanging Column, Pressure Extractor, Chilled
Mirror Hygrometer, or Centrifuge, 1-22.
Aubertin, M., Mbonimpa, M., Bussiere, B. & Chapuis, R.P.
(2003). A model to predict the water retention curve
from basic geotechnical properties. Canadian
Geotechnical Journal, 40(6), 1104-1122.
Chapuis, R.P. (2004). Predicting the Saturated Hydraulic
Conductivity of Sand and Gravel Using Effective
Diameter and Void Ratio. Canadian Geotechnical
Journal, 41(5), 787-795.
Chapuis, R.P. (2008). Predicting the saturated hydraulic
conductivity of natural soils. Geotechnical News,
26(2):47–50.
Chapuis, R.P. (2012) Predicting the saturated hydraulic
conductivity of soils: a review. Bulletin of Engineering
Geology and the Environment, 71(3): 401–434.
Fredlund, D.G. and Rahardjo, H. (2012). Soil mechanics for
unsaturated soils. John Wiley & Sons.
Impact of specimen preparation techniques
on the Interpretation of the unsaturated
shear strength of granular soils
Emad Maleksaeedi, Sogol Sarlati & Mathieu Nuth
Department of Civil Engineering – University of Sherbrooke,
Sherbrooke, Quebec, Canada

ABSTRACT
Natural or man-made granular deposits are often unsaturated in practice where the presence of suction may contribute to
their shear strength. The unsaturated shear strength of granular materials can be evaluated in the laboratory in conjunction
with soil-water-retention (SWR) tests. However, before performing a test, specimen preparation techniques often alter the
particle configuration which may consequently have an impact on the interpretation of the achieved results. The objective
of this paper is to examine the impact of moist tamping and hydraulic deposition techniques on the interpretation of
unsaturated shear strength for granular materials. To do so, a series of unsaturated drained triaxial tests are carried out
on glacial till specimens from Quebec prepared with the mentioned techniques at different confining pressure. To
complement the shear tests, drying soil-water retention curves of the glacial till are also obtained using the same specimen
preparation techniques. The results showed that specimen preparation techniques could affect the shearing behaviour of
granular soils. Ignoring a uniform test protocol for laboratory measurements corresponding to the granular deposit
formation might affect the interpretation of the unsaturated shear strength since their behaviour is dependent on the
physical properties of particles such as the particle shape and fabric isotropy.

RÉSUMÉ
Les dépôts granulaires naturels ou artificiels sont souvent non saturés dans la pratique où la présence de la succion peut
contribuer à leur résistance au cisaillement. La résistance au cisaillement non saturée des matériaux granulaires peut être
évaluée en laboratoire en conjonction avec des essais de rétention d'eau dans le sol. Cependant, avant d'effectuer un
essai, les techniques de préparation des échantillons modifient souvent la structure des particules, ce qui peut avoir un
impact sur l'interprétation des résultats obtenus. L'objectif de cet article est d'examiner l'impact des techniques de bourrage
humide et de déposition hydraulique sur l'interprétation de la résistance au cisaillement non saturé pour les sols
granulaires. Pour ce faire, une série d'essais triaxiaux drainés non saturés est effectuée sur des échantillons d'un till
glaciaire du Québec préparés à l'aide des techniques mentionnées, à différentes pressions de confinement. Pour
compléter ces essais, les courbes de rétention d'eau du sol à séchage du till glaciaire sont également obtenues en utilisant
les mêmes techniques de préparation des spécimens. Les résultats ont montré que les techniques de préparation des
échantillons pouvaient affecter le comportement au cisaillement des sols granulaires. En négligeant un protocole d'essai
uniforme pour les mesures de laboratoire correspondant à la formation de dépôts granulaires, la résistance au cisaillement
non saturé pourrait être mal interprétée, car son comportement dépend des propriétés physiques des particules, telles que
la forme des particules et l'isotropie du sol.

1 INTRODUCTION • Depending on the anticipated stress path, the

specimen must have a uniform degree of saturation
Testing a homogeneous and uniform specimen before before testing.
laboratory testing is required to characterize the behaviour • The specimen should be well-mixed to avoid particle
of unsaturated granular soils. In an ideal case, it is vital to segregation.
replicate the natural particle configuration to obtain valid • Specimen preparation technique must simulate the
results. However, the reconstitution of granular soils to natural mode of soil deposition as much as possible.
replicate their natural configuration is considerably difficult. Even though several specimen preparation techniques are
Kuerbis and Vaid (1988) presented certain criteria for available in the literature, numerous studies have shown
specimen preparation techniques for reconstituted that granular soil behaviour is highly dependent on
granular soil. These criteria with some modifications to laboratory preparation techniques (Mulilis et al. 1977,
account for the unsaturated state of soil are presented Miura et al. 1982).
herein: Unfortunately, the common specimen preparation
• The method must be able to produce loose to dense techniques of unsaturated soils often lead to a wide
specimens within the practical density range difference in the strength of soil and therefore fail to
in-situ conditions. represent the strength properties of natural soil. Moreover,
• The specimen must have a uniform void ratio the impact of preparation techniques on the hydraulic
throughout. retention behaviour of granular soil is not fully understood.
This paper aims to present some insights into the
impact of specimen preparation techniques on the water
retention capacity and shearing behaviour of granular soils technique, and the initial state of soil on the hydraulic
through a series of suction-controlled triaxial and water retention behaviour of the soils. For these tests, only drying
retention tests on a glacial till sampled in Quebec province. SWRC was obtained. However, in this article only four tests
namely SGT1, SGT6, SGT7 and SGT8 are presented that
2 TESTING PROCEDURE were prepared using moist tamping or hydraulic deposition
techniques.
To achieve the experimental objectives of this study, a For specimens SGT1 and SGT6, initially, oven-dried
poorly graded glacial till containing a large proportion of batches of glacial till were moistened with distilled water to
sandy-sized particles, noted as 𝑆𝐺𝑇𝑖 (for soil water achieve a target gravimetric water content of 2.5%. The
retention test) and 𝑆𝑆𝐺𝑇𝑖 (for triaxial tests) is chosen. samples were stored in plastic bags for a day to evenly
In Quebec, based on the definition of Ministère du distribute the water in the soil. Afterwards, the samples
transport du Québec, this granular soil conforms to the were added to the modified oedometer cell and then
granulometric requirements of MG112-type soils compacted to reach the constant void ratio of 0.66 by
commonly used in practice as subgrade fills. It is a well- controlling the achieved height of the specimens by several
known fact that granular soils, such as the tested glacial till layers. The constant void ratio means that different
have relatively low hydraulic retention capacity compared non-standard compaction energies created different soil
to fine materials such as clay. structures. During SWR tests, SGT1 and SGT6 underwent
The basic soil properties of the tested soil are given in constant net stress of 4 kPa and 256 kPa.
table 1. According to the USCS soil classification system, The constant void ratio of 0.66 was deliberately chosen
the glacial till is categorized as a poorly graded sand for experimental tests on specimens of SGT1 and SGT6
(SP-SM). since the specimens prepared by the hydraulic deposition
technique had a void ratio of nearly 0.66 which makes it
Table 1. Characteristics of tested glacial till possible to compare both preparation techniques.
Characteristics (%) Glacial till In parallel, two specimens of SGT7 and SGT8 were
assembled by the hydraulic deposition technique.
emax 0.69
Samples from dry batches of glacial till were rained to
emin 0.30 the water-filled Tempe cell. Subsequently, the surface of
Cu 8.3 the specimens was smoothly flattened by a spatula and
Cc 0.7 then soil particles were allowed to deposit under their
Gs 2.73 weight overnight. After 24 hours, a clear film of water on
top of each specimen was carefully removed. The
Fine content (%) 8%
specimens were vacuumed as shown in Figure 1 and
Peak drained friction angle 37.0o suggested by Lade (2016) to ensure that there would not
Drained cohesion (kPa) 0.0 be any air bubbles within the pores. The weight and height
of each specimen were then measured for the calculation
2.1 Soil Water Retention Curve (SWRC) phase.
Once all specimens were ready, the test setup was
To obtain the soil water retention curves (SWRC), a completed, and the axis translation technique was
modified Tempe cell fabricated at the University of employed to impose the suction. During SWR tests, both
Sherbrooke along with a modified oedometer apparatus volume changes and water exchanges during the
was used (Maleksaeedi et al. 2016). The modified setup experiments were recorded.
operates in a way that it could measure both mechanical
and hydraulic loadings and their superimposed response
simultaneously.
Using the proposed setup, one can perform a variety of
stress-strain paths, including mechanical paths that
correspond to a variation of net stress under constant
suction or suction loading paths where a variation of
suction happens at constant net stress. This could be of
importance during the assessment of the
hydro-mechanical behaviour of soils, particularly for
expansive and collapsible materials. However, studies
suggest that hydraulic loading can induce stiffness for Figure 1. Bubble chamber with a vacuum applied to the
granular soils even though the suction-induced volume specimen to expedite the saturation process and to ensure
changes might be about 1.0% compared to the initial state full saturation (Lade, 2016)
of specimens (Lins and Schanz 2003; Maleksaeedi et al.
2016). Hence, as the main advantage of the proposed 2.2 Unsaturated triaxial test
setup, even volume changes as low as 1.0 % caused by
suction could be recorded under constant mechanical Using the suction-controlled setup developed by Ng et
loading. al. (2002), known as HKUST cell, several unsaturated
In total, eight tests were performed where the focus was drained tests were performed on the glacial till specimens.
to evaluate the soil structure, the specimen preparation In total, thirty tests were performed on specimens
prepared by moist tamping and hydraulic deposition
techniques. However, in this article only six tests namely suction of 5 kPa. Specimens SSGT25, SSGT26 and
SSGT7, SSGT8, SSGT9, SSGT25, SSGT26 and SSGT27 SSGT27 were loaded by suction of 10 kPa respectively.
are presented.
The main distinction between the saturated and 3 RESULTS AND DISCUSSIONS
unsaturated triaxial tests is the saturation phase. In the
saturated state, the assumption is that the volume of 3.1 SWRC of specimens with different preparation
displaced water during both the consolidation and shearing techniques
phase is equal to the overall volume changes of the
specimen if B-value (Skempton Coefficient B) is close to To compare the effect of preparation techniques on the
1.0. Studies by Yoshimi et al. (1989), Baldi and Nova outcome of hydraulic retention tests, specimen SGT6 was
(1984), Lipinski and Wdowska (2010), and Sugiyama et al. prepared by moist tamping technique and underwent a
(2016) suggested that the B-value depends on the applied mechanical loading up to 256 kPa while for SGT1, the net
back pressure, the compressibility of the soil skeleton, the stress of 4 kPa was applied. The specimens SGT7 and
soil properties, and the stress history. Measuring B-value SGT8 prepared by the hydraulic deposition technique
in the unsaturated state is not straightforward due to the underwent similar mechanical loading of 4 kPa and
presence of air dissolution and air compressibility. 256 kPa respectively.
Recently, Vernay et al. (2019) presented a relationship Figure 2 depicts the measured SWRCs of tested
between the B-value and the degree of saturation 𝑆𝑟 as specimens. After performing the tests, the specimens were
shown in Equation 1 where B-value is dependent on the removed to measure the final water contents. Overall, there
initial void ratio of the specimen, 𝑒𝑜 , the isotropic was a satisfactory agreement between the measured and
compressibility of the soil skeleton in elastic conditions, κ predicted values.
and the initial mean effective stress, 𝑃𝑜′ and the initial
degree of saturation:

1
𝐵= 𝑝′ 𝑒 (1−𝑆𝑟𝑜 )
(1)
1+ 𝑜 𝑜
𝜅𝑝𝑎𝑡𝑚

As can be seen in Equation 1, the B-value is highly related

to the SWRC, and the elasto-plastic behaviour of the soils
and it cannot be regarded as a standalone parameter for
the investigation of water quantity in the triaxial specimen.
Consequently, the saturation phase for unsaturated
soils is considered a challenge when specimens are
prepared by the moist tamping technique.
A similar procedure as mentioned in the previous
section was used to prepare specimens in the triaxial mold
while setting a target void ratio of 0.66.
The hydraulic deposition technique eliminated the need
for controlling the B-value since the specimens were
already saturated before the testing. For the saturation
phase of specimens prepared by the moist tamping
technique, back pressure was applied. To expedite the
saturation phase, a vacuum pump as suggested by Lade
(2016) was periodically used to impose a relatively small
negative pressure at the top of the specimen. Monitoring
the water volume/pressure controller, the amount of
exchanged water showed the achieved degree of
saturation in the specimens. At the degree of saturation, 𝑆𝑟
near 90%, the vacuum pump was continuously open to
extract the trapped air from the specimens. Once, no air
could be seen to escape from the top of the specimen, it
was assumed that specimens were saturated.
Once saturated, for each test, the specimen was
consolidated at three different confinements. Three
confinement pressure of 50 kPa, 100 kPa and 300 kPa
were arbitrarily chosen in this regard. These confinements
were applied to the specimens in a stepwise fashion. After
achieving mechanical equilibrium, axis translation
technique was used to apply suction to the specimens.
While adjusting the applied confining pressure, specimens Figure 2. The SWRC of the glacial till prepared by different
SSGT10 to SSGT12 (moist tamped) were loaded by the preparation techniques in terms of (a) degree of saturation
(b) gravimetric water content
 in unsaturated granular soils induced by changes in the
applied suction.
From Figure 2 one can deduce that at low mechanical Furthermore, from Figure 3.b. it can be deduced that
loading, specimens SGT1 and SGT7 had a similar trend. regardless of the preparation technique, the specimens of
However, the difference between the SWRCs for SGT6 the glacial till experienced an irreversible deformation
and SGT8 emerged at relatively higher confinement. The under the applied suction. It seems that the tested material
Air-Entry-Values (AEV) of specimens at low mechanical had a memory of suction meaning that it has never
loading were unaffected by the preparation method. Yet, at experienced a previous non-null level of suction.
256 kPa of mechanical loading, SGT8 had slightly higher This over-compressed behaviour is soil-structure
AEV compared to SGT6. The main distinction between the dependent on unsaturated glacial till which occurred well
SWRCs of specimens prepared with moist tamping and beyond the AEV of the specimens. Yet, compared to
hydraulic deposition techniques at high mechanical loading cohesive materials as they can undergo suction-induced
is that there is a shift to a higher level of suction. Meaning deformation as high as 20% - 40% depending on the clay
that specimens prepared by the hydraulic deposition content, for cohesionless material, the volume changes
technique had higher retention capacity at the same level due to suction are essentially limited to much lower values
of suction compared to the specimens prepared by the as in the glacial till that reached as high as 2.0%.
moist tamping technique.
Furthermore, it seems that SGT8 reached the residual
state at a higher saturation level compared to SGT6. The
dissimilar fabrics can be the main explanation in this
regard. As explored by Kumar Kodicherla et al. (2018) on
ISO standard sand with fairly poor-graded grain-size
distribution, depositional techniques such as the pluviation
method induce more pronounced inherent fabric
anisotropy compared to the moist tamping technique.
Assessment of the fabric anisotropy of an unsaturated
granular assembly is of utmost importance, yet it is a
challenging task and the available experimental methods
for quantifying soil fabric anisotropy are very limited. Being
out of the scope of this research, it is recommended for
future studies evaluate how the fabric anisotropy can affect
hydraulic retention and correspondingly the
hydro-mechanical behaviour of unsaturated granular soils.
As confinement increases due to the closure of pores
between the particles and having a more restrained particle
arrangement, the glacial till has a higher retention capacity
for specimens prepared by the hydraulic deposition
technique. The obtained results are consistent with
numerous findings of researchers in the literature.
Figure 3.a shows the measured volume changes of
tested specimens during the suction imposition. The results
to some extent confirmed the observations in Figure 2
where the looser specimens, under 4 kPa of mechanical
loading, deformed more than specimens loaded at higher
mechanical stress. For SGT1 and SGT7, the preparation
technique did not influence the overall suction-induced
volume changes but at higher mechanical loading the gap
between SGT6 and SGT8 emerged. Specimen SGT8 had
lower deformation compared to SGT6 indicating that the
specimen structure affects the deformation due to suction,
probably owing to the uniformity of particle arrangement
and the closure of loose gaps between the particles.
Although the applied mechanical stresses and
measured axial deformation are very small, for all
specimens, the observed behaviour indicates some
intergranular phenomena occurring due to changes in the
degree of saturation and suction. Under gravity and
constant net normal stress, coarse-particle skeletons
develop a higher state of density which is more stable
Figure 3. Suction-induced volume changes of the glacial
compared to looser initial structures. The results confirm
till in terms of (a) void ratio (b) deformation
the findings of Milatz et al. (2016) regarding the settlements
3.1.1 Shearing behaviour of unsaturated specimens technique are deformable due to their structure. Jang and
prepared by different preparation techniques Frost (1998) stated that moist tamping technique produces
very loose specimens with “bulked, honeycomb structure”
As discussed earlier, the specimen preparation in a moist state, particularly in the dry-of-optimum state.
technique can significantly affect the hydraulic retention For specimens SSGT 10 to SSGT12, the non-linear
behaviour of unsaturated granular soils. Particle elastic behavior was followed by a simple increase toward
arrangements and grain-size distribution can attribute to a plateau at large strain. The specimens were compressed
the water storage of granular soils and in return, the progressively and became denser until reaching the critical
inter-particle stress can change accordingly. The induced state. Looking at figure 3.b., it seems that the specimens
inter-particle stress due to suction influences the underwent a contractive deformation which as limited to
unsaturated shear strength of granular soils. about 6.0% deformation of the total height at the critical
state.
Figure 5.a demonstrates the obtained shearing results
for the specimens prepared by the hydraulic deposition
technique loaded at the suction of 10 kPa. Compared to the
figure 4.a, the specimen prepared by the hydraulic
deposition technique showed higher elastic rigidity with a
peak behavior at a relatively lower strain. In figure 5.a, the
higher confining pressure resulted in higher shear
resistance. This is probably due to the capillary bonding
effect on the particles where particles were pulled together
more firmly.
Figure 5.b illustrates the variation of volumetric strain in
relation to the axial strain for the unsaturated glacial till
specimens prepared by the hydraulic deposition technique.
As anticipated, the general trend of relatively rigid
specimens encompassed a small expansion zone followed
by an extended contraction zone.
The volume change behaviour of the glacial till
prepared by the hydraulic deposition technique is
dependent on the confining pressure. Generally,
specimens confined at higher pressure generated more
restricted volume change regardless of the level of suction.
The overall volume changes of suction-controlled tests
were about 6.0% to 8.0% of the initial height of the
specimens.
The presented results in this section confirm the fact
that the specimen preparation techniques can affect the
hydraulic retention behaviour and the shearing behaviour
of unsaturated granular soils. As stated by Kuerbis and
Vaid [1988], specimens prepared by the hydraulic
deposition technique are exceptionally homogenous to the
void ratio and the particle size gradation, regardless of
gradation and fine content. The hydraulic deposition
technique simulates well the soil fabric found within a
natural fluvial or fill deposits.
The results also highlight the importance of defining a
uniform test protocol for reconstituted unsaturated granular
soils before executing shearing and hydraulic retention
tests. This in part affects the interpretation of shear
characteristics. The specimens prepared by the hydraulic
deposition technique have in essence higher resistance
compared to the specimens prepared by the moist tamping
technique.
Figure 4. Variation of deviator stress with axial strain (b)
volumetric strain with axial strain for the glacial till
specimen prepared at 2.5% of water content and 5.0 kPa
of suction.
Figure 4.a illustrates the obtained stress-strain results
of specimens prepared by the moist tamping technique at
2.5% at the suction of 5 kPa as an example. The results
suggest that specimens prepared by moist tamping
 specimens prepared by hydraulic deposition technique
essentially have higher water retention capacity, higher
shear resistance and lower deformability compared to the
specimens prepared by moist tamping technique.
The results also highlight the importance of defining a
uniform test protocol for reconstituted unsaturated granular
soils before executing shearing and hydraulic retention
tests. The results herein can be used as a benchmark for
establishing uniform testing protocols for future codes and
standards.

2 ACKNOLEDGEMENT
The authors would like to thank Hydro-Quebec and Natural
Sciences and Engineering Council of Canada (NSERC) for
their financial support.

REFERENCES

Kuerbis, R., and Vaid, Y. 1988. Sand Sample Preparation-

The Slurry Deposition Method. Soils and Foundations,
28(4), 107-188.
Mulilis J.P., Seed H.B., Chan C.K., Mitchell J.K., and
Arulanandan K. 1977. Effects of sample preparation on
sand liquefaction, Journal of the Geotechnical
Engineering Division, ASCE, 103: 91-108.
Miura S., Toki S., and Tanizawa F. 1984. Cone penetration
characteristics and its correlation to static and cyclic
deformation—strength behaviors of anisotropic sand
Soils and Foundations, 24(2): 58-74.
Lins, Y., and Schanz, T. 2003. Determination of hydro-
mechanical properties of sand. Proceedings of the
international conference "From experimental evidence
towards numerical modeling of unsaturated soils “,
Weimar, Germany, Springer.15-32.
Maleksaeedi, E., Nuth, M., and Chekired, M. 2016. A
modified oedometer apparatus for experimentally
obtaining the soil-water retention curve. Geovancouver.
Vancouver, Canada.
Lade, P. 2016. Triaxial Testing of Soils. John Wiley and
Sons.
Ng, C., Zhan, L., and Cui, Y. 2002. A New Simple System
for Measuring Volume Changes in Unsaturated Soils.
Canadian Geotechnical Journal, 39(3), 757-764.
Figure 5. Variation of deviator stress with axial strain (b) Yoshimi, Y., Tanaka, K., and Tokimatsu, K. 1989.
volumetric strain with axial strain for the glacial till Liquefaction Resistance of Partially Saturated Sand.
specimen at the suction of 10 kPa Soils and Foundations, 29(3), 157-162.
Baldi, G., and Nova, R. 1984. Membrane Penetration
4 CONCLUSION Effects in Triaxial Testing. Journal of Geotechnical
Engineering, 110(1), 403-420.
This paper presents some insights regarding how Lipinski, M., and Wdowska, M. 2010. Saturation Criteria for
specimen preparation techniques can affect the Heavy Over-consolidated Cohesive Soils. Annuals of
interpretation of unsaturated shear test results of granular Warsaw University of Life Sciences- SGGW, Land
soils. Reclam, 295-302.
A series of suction-controlled drained triaxial tests Sugiyama, Y., Kawai, K., and Lizuka, A. 2016. Effects of
along with suction-controlled water retention tests were Stress Conditions on B-value Measurement. Soils and
performed at University of Sherbrooke on a glacial till from Foundations, 56(5), 848-860.
Quebec. The tested specimens were prepared by moist Vernay, M., Morvan, M., and Breul, P. 2019. Evaluation of
tamping and hydraulic deposition techniques. the Degree of Saturation using Skempton Coefficient B.
Based on the results, it was observed that the specimen Geomechanics and Geoengineering.
preparation techniques can have a significant effect on
both water storage capacity, volume changes induced by
suction, and the evolution of stress-strain during the
shearing. The results suggested that unsaturated
Kumar Kodicherla, S., Gong, G., Fan, L., Moy, C., and He,
J. 2018. Effects of Preparation Methods on Inherent
Fabric Anisotropy and Packing Density of
Reconstituted Sand. Cogent Engineering, 5, 1533363.
Milatz, M., Torzs, T., and Grabe, J. 2016. Settlements in
Unsaturated Granular Soils Induced by Changes in
Saturation and Suction. E-UNSAT 2016.
Jang, D. J., and Frost, J. D. 1998. Sand structure
differences resulting from specimen preparation
procedures. Proceedings of the Specialty Conference
on Geotechnical Earthquake Engineering and Soil
dynamics, Seattle, WA, USA., 234-245.
Influence of hydraulic conductivity of
tension crack on the stand-up time of
unsupported vertical trenches in
unsaturated soils
Bhagya B. Mayadunna & Won Taek Oh
Department of Civil Engineering – University of New Brunswick,
Fredericton, NB, Canada

ABSTRACT
It is well known that tension cracks have an adverse impact on the stability of a slope since it reduces the resistance to
slope failure due to the decrease in slip surface. In addition, when rainfall infiltrates into a tension crack, the strength of a
soil in the vicinity of a tension crack significantly decreases, which leads to sudden failure in a slope. However, limited
studies have been undertaken to investigate the influence of rainfall infiltration into tension cracks on the stability of
unsupported trenches in unsaturated soils. In the present study, a series of numerical analyses are conducted to estimate
the stand-up time of unsupported vertical trenches excavated in an unsaturated glacial till (i.e., Indian Head till). Tension
cracks were simulated by adopting two different approaches: i) material with zero strength and its own hydraulic
conductivity function and ii) void space. Stand-up times obtained with these two approaches are compared and the results
are discussed.

RÉSUMÉ
Il est bien connu que les fissures de tension ont un impact défavorable sur la stabilité d'une pente car elles réduisent la
résistance à la défaillance de la pente en raison de la diminution de la surface de glissement. De plus, lorsque les
précipitations s'infiltrent dans une fissure de tension, la résistance d'un sol à proximité de la fissure de tension diminue
considérablement, ce qui entraîne une défaillance soudaine d'une pente. Cependant, peu d'études ont été menées pour
étudier l'influence de l'infiltration des précipitations dans les fissures de tension sur la stabilité de tranchées non soutenues
dans des sols non saturés. Dans la présente étude, une série d'analyses numériques est réalisée pour estimer le temps
de maintien debout de tranchées verticales non soutenues excavées dans un till non saturé (c'est-à-dire un till glaciaire
d'Indian Head). Les fissures de tension ont été simulées en adoptant deux approches différentes : i) matériau sans
résistance et sa propre fonction de conductivité hydraulique, et ii) espace vide. Les temps de maintien debout obtenus
avec ces deux approches sont comparés et les résultats sont discutés.

1 INTRODUCTION (Baker 1981, Lee et al. 1988, Chowdhury and Zhang

1991). Existence of tension cracks accelerates failure
Unsupported trenching is the fundamental stage in mechanism, which leads to sudden failure in unsupported
infrastructure construction projects, and it is typically trenches (Hu 2000, Ping et al. 2005, Wang et al. 2010).
carried out at shallow depths where soils are in unsaturated Tension cracks are initiated when tensile stress exceeds
conditions (Yanamandra and Oh 2019). Hence, among the tensile strength of the soil (Bagge 1985). Temperature
various factors that affect the stability of unsupported variations, differential settlement and associated
trenches, matric suction profile is the governing factor that construction activities may be the primary causes for
should be considered in analyzing the stability of tension crack propagation (Li and Zhang, 2010). The
unsupported trenches in unsaturated soils. Matric suction presence of a tension crack not only adversely affects the
profile depends on the field conditions, including level of stability of an unsupported trench but also creates a
groundwater table, type of soil, and regional climate preferential path for infiltration of water. This phenomenon
variables such as rainfall and evaporation. Failures in makes an unsupported trench more susceptible to failure
unsupported trenches are commonly caused by the loss of due to a significant decrease in the matric suction in the
matric suction associated with rainfall infiltration. vicinity of a tension crack. Hence, the hydraulic conductivity
Several studies have been undertaken to investigate function of a tension crack is a key factor that should be
the stability of unsupported trenches considering considered in analyzing the stand-up time of an
aforementioned factors, which provided guidelines on the unsupported trench under rainfall event (Li et al. 2011).
design of unsupported trenches in unsaturated soils. In the present study, a series of numerical analyses
However, although unsupported trenches are designed were conducted to estimate the stand-up times of
properly considering the variation of matric suction profile, unsupported vertical trenches (hereafter referred to as
unexpected failure can still take place due to the UVT) excavated in a single unsaturated Indian Head till
development of tension crack on the crest. This is layer under different rainfall intensities. Tension cracks
attributed to the reason that a tension crack shortens the were simulated by adopting two different approaches:
slip surface, which in turn reduces the resistance to failure namely, i) a material with zero strength and its own
hydraulic conductivity function and ii) void space. Two uncracked soil. In the present study, a single tension crack
different geotechnical modelling software, SEEP/W and was considered in the numerical analysis. Hence, it was
SLOPE/W (GeoStudio 2020) were used jointly for this assumed that the permeability of tension crack is one
research. magnitude higher than uncracked soil.

2 BACKGROUND 3 SOIL PROPERTIES

2.1 Geometry of Tension Cracks In this study, it was assumed that UVTs were excavated
into a single glacial till layer. Glacial till is widely spread
Several research have been undertaken to analyze the across Canada due to the extensive glaciation that
stability of unsaturated soil slopes considering the occurred during the last Ice Age. Table 2 summarizes the
influence of tension cracks. Different widths of tension basic properties of glacial till samples that were obtained
cracks were used by researchers as summarized in Table from Indian Head, Saskatchewan, Canada (i.e., Indian
1. The width of tension crack is one of crucial factors that Head till, IHT).
governs the stability of slopes since it controls the
infiltration quantity. Based on the initial numerical analyses, Table 2. Basic soil properties of Indian Head till (Oh and
it was observed that a tension crack width of 5 mm did not Vanapalli 2010)
result in significant differences in the stand-up time
compared to the outcomes obtained with a 10 mm width.
Soil property Value
Therefore, for ease of modeling, numerical analyses were
carried out for tension cracks with a width of 10 mm. Plasticity index, Ip (%) 15.5
Specific gravity, Gs 2.7
Table 1. Widths of tension cracks used by other Void ratio, e 0.69
researchers in the stability analyses of unsaturated soil Saturated volumetric water content, sat 0.47
slopes
Saturated unit weight, sat (kN/m3) 20.7
Effective cohesion, c’ (kN/m2) 5
References tension crack width
Effective internal friction angle, ’ (°) 23.1
Ping et al. (2005) 0.29, 0.63 and 1.35 mm
Wang et al. (2011) 5 mm Figure 1 shows the soil-water characteristic curve (SWCC)
Yanamandra (2020) 0.1 m of uncracked IHT established based on the data available
in Oh and Vanapalli (2013). The best-fitting was obtained
using the model proposed by van Genuchten (1980; Eq. [1]
2.2 Hydraulic conductivity function of tension cracks ). The hydraulic conductivity function for uncracked IHT
was estimated using the model proposed by van
For unfractured soils, the maximum depth of rainfall Genuchten (1980; Eq. [2]), as shown in Figure 2, along with
infiltration is typically less than 3 m due to the low the hydraulic conductivity function of tension crack. As
permeability (Zhang et al. 2017). However, rainwater may mentioned earlier, it was assumed that the hydraulic
seep into deeper depth in cases where tension cracks are conductivity of a tension crack was one magnitude higher
developed in the vicinity of slopes or unsupported trenches. than unfractured soil. The volume change of the crack upon
Tension cracks provide preferential routes for rainfall drying and wetting are not considered in the numerical
infiltration into soil mass. Rainfall that infiltrated through the analyses.
tension cracks increases the pore-water pressure, which in mvG
turn reduces matric suction and consequently the shear ( θ − θr ) =  1 
strength of soil (Sasekaran 2011, Hu 2000, Zhang et al. Se =  (mvG = 1 − 1 nvG ) [1]
( θs − θr ) 1 + ( αψ )n vG

2020). In addition, rainfall that fills tension cracks not only
exerts an additional driving force on the slope or
unsupported trench but also further increases the
infiltration rate due to the water head. The permeability of  (nvG −1)  − mvG
 
2

 1 − ( αψ) 1 + ( αψ) vG 
n

   
a tension crack developed in an unsaturated soil is affected
k(ψ) = k sat     [2]
by the matric suction value (Zhang et al. 2010). This  1 + ( αψ)  nvG (
1 2 −nvG 2 )

indicates that the hydraulic conductivity function of a    
 
tension crack should be considered in estimating the stand-
up time of unsupported trenches in unsaturated soils under
where Se = effective degree of saturation, k(), ksat =
rainfall events.
hydraulic conductivity for unsaturated and saturated
Hu (2000) conducted stability analyses for a soil slope
condition, respectively,  = volumetric water content, s, r
that consists of a cracked layer underlain by intact soil. The = saturated and residual volumetric water content,
cracked layer was modeled as a type of soil with relatively respectively , nvG, mvG = fitting parameters for van
high permeability and low strength. The permeability of Genuchten (1980) model (obtained with the nonlinear
crack was assumed to be two magnitudes higher than
fitting program developed by Seki (2007)), and  = matric 1960; Kwan, 1971), which can subsequently increase the
suction shear strength of the soil. Hence, in the present study, for
conservative analyses, the variation of pore-water
The variation of shear strength with respect to matric pressure in association with excavation was neglected by
suction was estimated using the model proposed by analyzing the stability of unsupported vertical trenches
Vanapalli et al. (1996; Eq. [3]). This is the sole shear using SLOPE/W (GeoStudio 2020), assuming that the
strength model adopted in GeoStudio (2020). groundwater table remains unchanged during excavation
(Ileme and Oh 2019). SEEP/W (GeoStudio 2020) was used
to simulate rainfall infiltration into tension cracks.
  − r 
unsat = c  + (n − ua )tan  + (ua − uw )   tan  [3]
 s − r  The following steps were adopted in the numerical
analyses.
where unsat = shear strength of unsaturated soil, c’ =
Step 1:
effective cohesion, ’ = effective internal friction angle, n =
normal stress, ua = pore-air pressure, (n – ua) = net normal
Stability analysis was first conducted in SLOPE/W using
stress, uw = pore-water pressure, and (ua – uw) = matric
Bishop (1960)’s simplified method to determine the safe
suction
height of UVT for GWT = 3m. The safe height was defined
as the depth of UVT that shows factor of safety (FOS) =
0.50
1.2. The ‘Entry and Exit’ slip surface option was used to
define the potential slip surfaces. The exit was specified as
Volumetric water content, q

0.45
a point at the toe of UVT, and the entry was defined as a
0.40 range along the ground surface (Figure 3).
0.35

0.30 1.2
van Genuchten (1980)
Measured
0.25 6

0.20
5
0.15
Elevation (m)

4
0.10
0.1 1 10 100 1000
3
Matric Suction (kPa)
Figure 1. Soil-water characteristic curve of Indian Head till 2

10-6
1
Hydraulic conductivity, k (m/sec)

10-7
0
10-8 0 1 2 3 4 5 6 7 8 9 10
10-9 Distance (m)
10-10 Figure 3. Determination of safe height (FOS = 1.2) in
10-11 SLOPE/W
HCF (intact soil)
10-12 HCF (tension crack)
Step 2:
10-13
Once the safe height is achieved, additional stability
10-14
analyses were conducted to determine the distance and
10-15 depth of a tension crack, which results in FOS = 1.1. For
0.01 0.1 1 10 100 1000
this, ‘tension crack line’ was assigned to a certain depth in
Matric suction (kPa) SLOPE/W to determine the distance of a tension crack that
Figure 2. Hydraulic conductivity function (HCF) of intact soil leads to the minimum FOS for the targeted depth of tension
and tension crack crack. This procedure was repeated for various depths of
tension crack. Figure 5 shows the combinations of distance
4 METHODOLOGY and depth ratio of tension cracks that lead to the minimum
factor of safety. Distance and depth ratio used in Figure 5
A series of numerical analyses were conducted with a are defined in Figure 6. Based on the results in Figure 5,
groundwater table at 3 m from the ground surface under the tension crack whose distance and depth are 1.37 m
four different rainfall intensities; namely, 0.11, 0.36, 1.08 and 0.3 m, respectively is chosen in this study.
and 10.8 mm/hr, which are 0.3, 1, 3, and 30 times of ksat of
unfractured IHT, respectively. Excavation in a soil causes
a decrease in pore-water pressure (Bishop and Bjerrum,
 tension crack
1.1 Step 3:

depth
tension crack

6
distance After the distance and depth of a tension crack is decided,
seepage and stability analyses were conducted to
5 determine the stand-up time of UVT. The hydraulic
tension crack line conductivity of a tension crack was assigned by adopting
Elevation (m)

4 the following two approaches in numerical analysis.

3 i) Tension crack can simulate as a material with zero

strength and the hydraulic conductivity that is 10 folds
2
higher than that of unfractured soil for each matric
1
suction value (referred to as khcf)
ii) Tension crack can be simulated as a void such that
0 the infiltration rate into a tension crack is the same as
0 1 2 3 4 5 6 7 8 9 10 rainfall intensity (Yanamandra and Oh 2021) (referred
Distance (m) to as kvoid)
Figure 4. Determination of distance of tension crack that
leads to the minimum FOS for an assigned ‘tension crack To simulate rainfall infiltration, unit flux boundary conditions
line’ in SLOPE/W (i.e., 0.11, 0.36, 1.08 and 10.8 mm/hr) were applied on the
soil surface and at the bottom of UVT. In case khcf is
adopted, unit flux boundary conditions were assigned
directly on the soil surface, while for the case of kvoid unit
flux boundary conditions were applied directly at the bottom
of tension crack (Figure 7).
flux boundary condition
1.15
FOS = 1.1
1.10
FOS

1.05

1.00
0.70
0.95 0.65
tio

khcf: material with zero

0.60
ra

strength and hydraulic

0.90
ce

0.55 conductivity function

an

0.3
0.50
st

0.2
Di

Dept 0.1 0.45

h rat
io
Figure 5. Combinations of distance and depth ratio of (a)
tension cracks that lead to the minimum factor of safety

TDist

TDepth kvoid: void space

TDist: Distance of a tension crack from the

Hsafe top of the vertical face of the cut
Tdepth : Depth of tension crack
Hsafe : Safe height (FOS = 1.2)
TDist
Distance ratio =
Hsafe
TDepth
Depth ratio =
Hsafe (b)

Figure 7. Application of unit flux boundary conditions; (a)

Figure 6. Definition of distance and depth ratio used in khcf, (b) kvoid
Figure 5
Table 3. Variation of pore-water pressure distribution with 5 ANALYSES RESULTS
time in the vicinity of tension crack for khcf and kvoid
Table 3 shows the variation of pore-water pressure
khcf kvoid distribution with time in the vicinity of tension crack for khcf
and kvoid. Obvious difference in the pore-water pressure
distribution is observed for the first 2 days; however, the
difference becomes not significant thereafter. The variation
of FOS with time under different rainfall intensities for khcf
and kvoid are summarized in Table 4. The results in Table 4
are plotted in Figure 8. As can be seen, negligible
difference in FOS is observed between khcf and kvoid for
0.5 days each time step. This can be attributed to the reason that
the volume of water that infiltrates through the bottom of
the tension crack is not significant to induce a sudden
increase in pore-water pressure.

Table 4. Summary of variation of FOS with time under

different rainfall intensities for khcf and kvoid

1.04 days Rainfall Intensity (mm/hr)

Time
0.11 0.36 1.08 10.8
(Days)
khcf kvoid khcf kvoid khcf kvoid khcf kvoid

0 1.10 1.10 1.10 1.10 1.10 1.10 1.10 1.10

1 1.10 1.10 1.09 1.09 1.08 1.08 1.08 1.07

2 1.10 1.10 1.09 1.09 1.07 1.07 1.06 1.05

1.96 days
3 1.10 1.09 1.08 1.08 1.05 1.05 1.04 1.03

4 1.09 1.09 1.07 1.07 1.03 1.03 1.03 1.02

6 1.09 1.09 1.04 1.04 1.01 1.02 1.01 1.00

7 1.09 1.09 1.03 1.03

9 1.08 1.08 0.99 0.99

3.08 days 12 1.07 1.07

17 1.04 1.04

21 1.02 1.02

25 1.00 1.00

5.29 days

8.04 days

Figure 8. Variation of FOS with time under different rainfall

intensities for khcf and kvoid
Figure 9 shows the variation of stand-up time with respect ii) Rainfall infiltration into tension crack decreases the
to rainfall intensity for khcf and kvoid. For the purpose of stand-up time of UVTs irrespective of the rainfall
comparison, stand-up time obtained without tension crack intensity. The analyses results showed that the stand-
is presented as well. It is interesting to note that, for both up time decreases by 9 and 2 days for low (i.e., 0.11
with and without tension crack, critical decrease in stand- mm/hr) and high (i.e., 0.36, 1.08 and 10.8 mm/hr)
up takes place under the rainfall intensity equals to ksat (i.e., rainfall intensities, respectively.
0.36 mm/hr). Due to the reason aforementioned, no iii) Significant decrease in stand-up time took place under
difference was observed for the stand-up times between the rainfall intensity of 0.36 mm/hr both with and without
khcf and kvoid. However, the tension crack caused the tension crack, which equals ksat of unfractured soil. The
decrease in stand-up time by 9 days and 2 days for low decreases in stand-up time became negligible for the
(i.e., 0.11 mm/hr) and high (i.e., 0.36 1.08 and 10.8 mm/hr) rainfall intensities greater than ksat.
rainfall intensities, respectively when compared to those for
intact soil. REFERENCES

Bagge, G. 1985. Tension cracks in saturated clay cuttings,

Proceedings of the 11th International Conference, San
Francisco, pp. 393–395.
Baker, R. 1981. Tensile strength, tension cracks, and
stability of slopes, Soils and Foundations, 21(2): 1–17.
Bishop, A. W. and Bjerrum, L. 1960. The relevance of
the triaxial test to the solution of stability problems,
Proc., ASCE Res. Conf. on Strength of Cohesive Soils,
ASCE, New York, pp. 437-501.

Bishop, A.W. and Morgenstern, N. 1960. Stability

coefficients for earth slopes, Geotechnique, 10(4):
164–169.
Chowdhury, R.N. and Zhang, S. 1991. Tension crack and
slope failure, Slope Stability Engineering:
Developments and Applications, The Institution of Civil
Figure 9. Variation of stand-up time with respect to rainfall Engineers, London, pp. 27–32.
intensity with (khcf and kvoid) and without tension crack van Genuchten, M.T. 1980. A closed- form equation for
predicting the hydraulic conductivity of unsaturated
6 SUMMARY AND CONCLUSIONS soils, Soil Science of America Journal, 44: 892–898.
Hu, S. 2000. Reliability of slope stability considering
In the present study, the influence of hydraulic conductivity infiltration through surface cracks, Master’s Thesis,
of a tension crack on the stand-up time of unsupported Hong Kong University of Science and Technology,
vertical trench is investigated using the geotechnical Hong Kong.
modeling software, GeoStudio 2020. Numerical analyses Ileme, V. and Oh, W.T. 2019. Estimating the stand-up time
were carried out for a groundwater table at 3 m and a of unsupported vertical trenches in vadose zone, KSCE
tension crack whose distance and depth are 1.37 m and Journal of civil Engineering, 23(10): 4259-4273.
0.3 m, respectively under four different rainfall intensities Kwan, D. 1971. Observation of the failure of a
(i.e., 0.11, 0.36, 1.08 and 10.8 mm/hr). Tension crack was
simulated in two different approaches; i) a material with vertical cut in clay at Welland, Ontario, Canadian
zero strength and the hydraulic conductivity that is 10 folds Geotechnical Journal, 8(2): 283-298.
higher than that of unfractured IHT (khcf) and ii) a void that Lee, F.H., Lo, K.W. and Lee, S.L. 1988. Tension crack
allows rainfall infiltrated into the tension crack at the same development in soils, Journal of Geotechnical
rate as rainfall intensity (kvoid). Findings obtained from this Engineering, 114(8): 915–929.
study are as follow: Li, J.H. and Zhang, L.M. 2010. Geometric parameters and
REV of a crack network in soil, Computers and
i) Permeability of a tension crack affects the distribution Géotechnics, 37(4): 466–475.
of pore-water pressure in the vicinity of tension crack. Li, J.H., Zhang, L.M. and Li, X. 2011. Soil-water
However, insignificant difference in pore-water characteristic curve and permeability function for
pressure was observed between two scenarios, khcf unsaturated cracked soil, Canadian Geotechnical
and kvoid after two days of rainfall infiltration. This is Journal, 48(7): 1010–1031.
because volume of rainfall that infiltrates into a tension Oh, W.T. and Vanapalli, S.K. 2010. Influence of rain
crack is not significant. This result suggests that it is infiltration on the stability of compacted soil slopes,
more convenient to model a tension crack as a void Computers and Geotechnics, 37(5): 649–657.
space without taking account of hydraulic conductivity
of a tension crack in seepage analysis.
Oh, W.T. and Vanapalli, S.K. 2013. Interpretation of the
Bearing Capacity of Unsaturated Fine-Grained Soil
Using the Modified Effective and the Modified Total
Stress Approaches, International Journal of
Geomechanics, 13(6): 769–778.
Ping, F., Qingquan, L., Jiachun, L. and Jianping, S. 2005.
Numerical analysis of rainfall in filtration in the slope
with a fracture, Science China Technological Sciences,
48(1): 107–120.
Sasekaran, M. 2011. Impact of permeability and surface
cracks on soil slopes, Master’s Thesis, University of
Manchester.
Seki, K. 2007. SWRC fit - a nonlinear fitting program with a
water retention curve for soils having unimodal and
bimodal pore structure, Hydrology and Earth System
Science, 4(1): 407–437.
Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. and Clifton,
A.W. 1996. Model for the prediction of shear strength
with respect to soil suction, Canadian Geotechnical
Journal, 33(3): 379–392.
Vanapalli, S.K. and Oh, W.T. 2013. Uncertainties in
Interpreting the Scale Effect of Plate Load Tests in
Unsaturated Soils, Proceedings of the International
Symposium on Engineering under Uncertainty: Safety
Assessment and Management, Springer India, India,
pp. 141–163.
Wang, R., Zhang, G. and Zhang, J.M. 2010. Centrifuge
modeling of clay slope with montmorillonite weak layer
under rainfall conditions, Applied Clay Science, 50(3):
386–394.
Wang, Z.F., Li, J.H. and Zhang, L.M. 2011. Influence of
cracks on the stability of a cracked soil slope,
Unsaturated Soils: Theory and Practice 2011,
Thailand.
Yanamandra, G. and Oh, W.T. 2019. Influence of tension
cracks on the stability of unsupported vertical trenches
in vadose zone, Canadian Geotechnical Conference
’72, Saint John’s, NL, Canada.
Yanamandra, G. 2020. Influence of excavation rate and
tension crack on the stability of an unsupported vertical
cut in unsaturated soil, Master’s Thesis, University of
New Brunswick, Canada.
Yanamandra, G. and Oh, W.T. 2021. Influence of
excavation time, tension crack, and rainfall on the
stability of unsupported vertical cuts in unsaturated soil,
International Journal of Geomechanics, 21(10): 1−17.
Zhang, L.M., Li, J.L. and Fredlund, D.D. 2010. Hydraulic
Conductivity Function for Unsaturated Cracked Soil,
International Conference on Unsaturated Soils ’5,
Barcelona, Spain.
Zhang, S., Xu, Q. and Zhang, Q. 2017. Failure
characteristics of gently inclined shallow landslides in
Nanjiang, southwest of China, Engineering Geology,
217(1): 1–11.
Zhang, J., Zhu, D. and Zhang, S. 2020. Shallow slope
stability evolution during rainwater infiltration
considering soil cracking state, Computers and
Geotechnics, 117: 1–12.
Active earth pressure in unsupported
vertical trenches in vadose zone
Ashkan Shirzadi & Won Taek Oh
Department of Civil Engineering – University of New Brunswick,
Fredericton, NB, Canada

ABSTRACT
Accurate calculation of active earth pressure is vital in analyzing the stability of unsupported trenches. In conventional soil
mechanics, active earth pressure can be calculated using the earth pressure theories proposed by Rankine or Coulomb.
However, typically, unsupported trenches are excavated into soils that are in a state of unsaturated conditions. In this case,
the field conditions such as type of soil, matric suction profile, and rainfall events should be considered in calculating the
active earth pressure. The infiltration of rainfall into unsupported trenches excavated in the vadose zone can result in
sudden failure, which poses a serious risk to the lives of field workers. Therefore, it is of utmost importance to estimate the
behaviours of unsupported trenches under rainfall events to ensure the safety of field workers within unsupported trenches.
In the present study, a series of numerical analyses were carried out to estimate the variation of active earth pressure and
the optimal location of shoring and load that would maximize the stability of an unsupported vertical trench considering
rainfall infiltration by extending the conventional Rankine earth pressure theory. Analyses are carried out for multiple levels
of groundwater table and rainfall intensities.

RÉSUMÉ
Le calcul précis de la poussée active des terres est essentiel pour analyser la stabilité des tranchées non soutenues. En
mécanique des sols conventionnelle, la poussée active des terres peut être calculée à l'aide des théories de pression des
terres proposées par Rankine ou Coulomb. Cependant, les tranchées non soutenues sont généralement creusées dans
des sols qui se trouvent dans un état de conditions non saturées. Dans ce cas, les conditions sur le terrain telles que le
type de sol, le profil de succion matricielle et les événements pluvieux doivent être pris en compte pour calculer la poussée
active des terres. L'infiltration des précipitations dans les tranchées non soutenues creusées dans la zone vadose peut
entraîner une défaillance soudaine, ce qui constitue un risque sérieux pour la vie des travailleurs sur le terrain. Il est donc
d'une importance capitale d'estimer le comportement des tranchées non soutenues lors d'événements pluvieux afin
d'assurer la sécurité des travailleurs sur le terrain à l'intérieur de ces tranchées. Dans la présente étude, une série
d'analyses numériques a été réalisée pour estimer la variation de la poussée active des terres et l'emplacement optimal
des étançons et de la charge qui maximiserait la stabilité d'une tranchée verticale non soutenue en tenant compte de
l'infiltration des précipitations, en étendant la théorie de la pression des terres de Rankine conventionnelle. Les analyses
sont réalisées pour plusieurs niveaux de nappe phréatique et d'intensités de pluie.

1 INTRODUCTION report (deputy regional administrator of OSHA’s Region 6),

39 workers in the United States died in such accidents,
Unsupported trenches play an essential role in the exceeding the previous year's figure by more than double
construction of pipelines, electrical lines, stormwater, etc. and marking the highest number since 2017 (Philips,
since the placement of these crucial infrastructure 2023). As a result of this concerning trend, it is imperative
underground provides safeguards against physical to follow proper safety procedures in unsupported
damage and deterioration caused by weathering. trenching to prevent accidents and deaths.
However, it is also important to note that unsupported Unsupported trench failures can happen for various
trenching can pose greater risks for workers due to the causes. In regions with warm and dry climates, unsaturated
confined space, particularly concerning the potential trench soils are prevalent, and trench designs typically consider
wall collapse that could lead to workers being trapped the characteristics of these soils (Oh and Bajestani 2023).
(Jannadi, 2008). According to OSHA (Occupational Safety However, with the ongoing climate change, rainfall patterns
and Health Administration), failures or cave-ins of are becoming more intense and frequent, which can cause
unsupported trenches is among the most common causes significant decreases in the shear strength of soils and
of construction-related death and serious injury. Injuries eventually lead to failure in unsupported trenches. This is
can have varying consequences such as hospitalization, especially true when a project should be completed in a
long-term disability, and even fatality, impacting not only timely manner. This indicates that it is crucial to consider
the injured workers but also their families. In addition, these changing weather patterns and the effects of climate
excavation safety accidents can also negatively impact a change when designing and constructing unsupported
company's reputation, resulting in lost productivity and trenches to prevent accidents and injuries.
increased insurance costs. The fiscal year 2022 saw an The stability of an unsupported trench is governed by
unfortunate increase in fatalities resulting from trench the active earth pressure profile, which is essential for
collapses (OSHA, 2022). According to Stephen Boyd’s estimating the critical height of an unsupported trench.
When soil is fully saturated, the location of active thrust is s = −Se (ua − uw ) [3]
typically one-third of the height of the unsupported trench.
In contrast, for unsaturated or partially saturated soils,
determining the location of active thrust is not as
straightforward as in saturated soils. This is because the where uw = pore-water pressure, (ua – uw) = matric suction,
matric suction and overburden pressure profiles are a and Se = effective degree of saturation (Eq. [4])
function of the degree of saturation and continuously vary
under rainfall events. The primary benefit of determining
the depth of the resultant active thrust is that it allows for  − r
Se = [4]
the cost-effective improvement of unsupported trench s − r
stability without the need for trench boxes. However, the
process of determining the active thrust using analytical
modelling can be complex and time-consuming. In this where  = volumetric water content, and s, r = saturated
paper, an attempt is made to investigate the optimal depth and residual volumetric water content, respectively.
of load that can be applied against the unsupported vertical
trenches in unsaturated soils considering rainfall events. The horizontal stress in an unsaturated unsupported
For this, a series of numerical analyses were carried out vertical trench can then be estimated by substituting Eq. [2]
using geotechnical modeling software SLOPE/W and in Eq. [1], as shown in Eq. [5].
SEEP/W (GeoStudio 2020, Seequent Int. Ltd.). The
present study is anticipated to enhance the design of
unsupported vertical trenches by offering valuable insights
into the behavior of unsaturated vertical trenches for ( h − ua ) = ( v − ua ) K a − 2c K a − (K a − 1) s [5]
diverse rainfall intensities and groundwater levels.

where h, v = total horizontal and normal stress,

2 BACKGROUND respectively, and (h – ua), (v – ua) = net horizontal and
normal stress, respectively.
2.1 Active Earth Pressure in Unsaturated Soils
Eq. [5] shows that the active earth pressure in an
Excavation of an unsupported trench disturbs and alters unsaturated unsupported vertical trench can be estimated
the soil's natural state. As a result, active condition is considering three key components; namely, vertical stress,
developed in the vicinity of the unsupported trench, which effective cohesion, and suction stress. Vertical stress is an
decreases the stability of the unsupported trench. In this instability factor (i.e., positive active earth pressure), while
case, the horizontal effective stress in the active zone can effective cohesion and suction stress are stabilization
be calculated using Eq. [1] (Rankine 1858). factors (i.e., negative active earth pressure). The impact of
these factors on the trench wall is illustrated in Figure 1. If
the air pressure is assumed to be atmospheric pressure
h = vK a − 2c  K a [1] (i.e., ua = 0), Eq. [5] can be rewritten as Eq. [6].

where ’h = effective horizontal pressure, c’ = effective h =  vK a − 2c  K a − (K a − 1) s [6]

cohesion, ’ = effective internal friction angle, ’v = effective
vertical stress, and Ka = Rankine’s active earth pressure
coefficient [= tan2(45° – ’/2)].

According to Lu and Likos (2004, 2006), the effective

stress of unsaturated soils can be written in terms of
suction stress as shown in Eq. [2].


unsat = (  − ua ) − s [2]

where ’unsat = effective stress in unsaturated soils,  = total

stress, ua = pore-air pressure, ( – ua) = net normal stress,
and s = suction stress.
Figure 1. Illustration of Eq. [6], showing active earth
Suction stress for unsaturated soils can be defined as Eq.
pressure profile in an unsaturated unsupported vertical
[3] (Lu et al 2010).
trench
The depth of the tension zone is defined as the depth The variation of shear strength with respect to soil
where the net horizontal stress is zero as shown in Figure suction was calculated using the model proposed by
1. Infiltration of rainfall decreases the contribution of Vanapalli et al. (1996; Eq. [9]).
suction stress towards the negative active pressure, which
in turn, leads to the decrease in the depth of tension zone.
 (  − r ) 
unsat = c  + ( n − ua ) tan  + (ua − uw )  tan  [9]
3 SOIL PROPERTIES  ( s
 −  r) 

For the purpose of this study, the trenches were assumed

to be excavated into the glacial till (i.e., Indian Head till, where unsat = shear strength for unsaturated condition.
IHT) that is one of the most common soil type found
throughout Canada (Oh and Vanapalli, 2010, 2013). Basic
soil properties of IHT are summarized in Table 1.
0.5
0.45
Table 1. Characteristics of tested soils 0.4

Volumetric Water Content

0.35
Soil Property Value
0.3
Effective cohesion, c’ (kPa) 5
0.25
Effective internal friction angle, ’ (°) 23.1
20.7 0.2
Saturated unit weight, sat (kN/m ) 3

Saturated volumetric water content, s 0.468 0.15

Saturated hydraulic conductivity, ksat (m/sec) 10-7 0.1 a = 12.21

n = 1.47
0.05 m = 0.55
0
The soil-water characteristic curve (SWCC) was 0.01 0.1 1 10 100 1000
established using the nonlinear fitting model developed by Matric Suction(kPa)
Fredlund and Xing (1994) (Eq. [7]) based on the data
available in Oh and Vanapalli (2013) (Figure 2). Figure 2. SWCC of Indian Head till and fitting parameters

Se =
(  − r ) = 1
[7]
1.00E-07
( s − r ) ln e + (  a )n m
  1.00E-08
Water Conductivity (m/sec)

1.00E-09
where  = soil suction, a, n, m = fitting parameter, and e =
Euler’s number. 1.00E-10

The hydraulic conductivity function of IHT was 1.00E-11

estimated using the model proposed by Fredlund et al.
(1994; Eq. [8]), as shown in Figure 3. It was also assumed
1.00E-12
that the soil has an isotropic hydraulic conductivity.
1.00E-13
0.001 0.1 10 1000

(( ) ) ( )
b

 ln(  )  e −  (  ) e  e dy 
y y
 y Matric Suction (kPa)
k = k sat  b  [8]
  (( )
 ln( b )
) ( )
 e y − s e y  e y dy 

Figure 3. Hydraulic conductivity function of IHT used in the
numerical analyses

where k = hydraulic conductivity of the water phase, k sat = 4 METHODLOGY

saturated hydraulic conductivity, y = dummy variable of
integration representing the logarithm of the soil suction, ’ As previously stated, the objective of this study is to identify
= first derivative with respect to , b = air-entry value, and the optimal point of load placement that would maximize
b = upper limit of integration. the stability of an unsupported vertical trench considering
rainfall infiltration. For this, the geotechnical modeling
software, SEEP/W and SLOPE/W were jointly used for
seepage and stability analysis, respectively. Analyses
were carried out with the depths of groundwater table at 1.1 Step 3: Determination of Active Earth Pressure
m and 2.5 m from the ground surface to simulate field
conditions with low and high water levels within the For each seepage analysis time step, active earth pressure
unsupported trenches. The step-by-step procedures used distribution is established as shown in Figure 6. Based on
in the numerical analyses are as follow: the pore-water pressure distribution profile obtained from
SEEP/W and the SWCC, the contribution of suction stress
Step 1: Determining the Critical Height of Unsupported towards the active earth pressured was estimated. The
Vertical the Trench difference, then, between negative (i.e., stabilizing) and
positive (i.e., destabilizing) forces were determined by
The critical height (i.e., Factor of Safety = 1) of an calculating the areas of negative and positive active earth
unsupported vertical trench for a given depth of pressure distribution diagram.
groundwater table was first determined by utilizing
SLOPE/W. To define the potential slip surfaces, the 'Entry
and Exit' slip surface option was utilized. For the exit point,
a specific point located at the toe of the trench was chosen.
As for the entry point, a range was defined along the
ground surface to ensure comprehensive coverage of
possible slip surfaces (Figure 4). Critical heights were
determined to be 3.3 m and 2.6 m for the depth of
groundwater table at 1.1 m and 2.5 m, respectively. The
higher critical height with the groundwater table at 1.1 m is
attributed to the reason that the higher hydrostatic pressure
within the trench increases the stability of unsupported
trench. In practice, the water within an unsupported trench
is usually pumped out before construction operation,
Hence, in this study, the influence of hydrostatic pressure Figure 5. Applying influx boundary condition to simulate
on the stability of unsupported vertical trench was rainfall infiltration using SEEP/W
neglected to consider the worst-case scenario.

To analyze the non-linear active earth pressure

distribution line, it is essential to divide the area into
Range of entry smaller, manageable sections in order to determine the
positive and negative areas. To achieve this, the area is
divided into equal segments, and the area of each segment
is calculated using the trapezoidal rule. Area 1 gradually
reduces with time due to the decrease in matric suction in
association with rainfall event, which decreases the
stability of an unsupported vertical trench.

Exit point

Day 8

Active Earth Pressure (kPa)

-10 0 10 20
Figure 4. Determination of the critical height using 2.6
SLOPE/W 2.34
Area 1
2.08
Trench Elevation

1.82
Step 2: Simulation of Rainfall Event
1.56
Once the critical height was determined, rainfall infiltration 1.3
was simulated by assigning influx boundary conditions
1.04
along the ground surface using SEEP/W (Figure 5). In
order to depict the worst-case scenario, the maximum 0.78
Area 2
precipitation rate that corresponded to the saturated 0.52
hydraulic conductivity value was used in the seepage
0.26
analyses. It was also assumed that excess rainfall is not
accumulated on the ground surface (i.e., runoff). It should 0
be noted that it is out of scope to consider the regional
Figure 6. Example of active earth pressure distribution
precipitation rates in the numerical analyses.
based on the pore-water pressure distribution information
8 days after the rainfall infiltration
 m, respectively. Based on the active earth pressure profile
shown in Figure 7, negative (Area 1 in Figure 6) and
positive (Area 2 in Figure 6) thrusts were calculated. The
variation of difference between negative and positive active
thrusts with time under rainfall event is shown in Table 2.
For each time step, the calculated resultant force was
applied at various depths (i.e., 0, 20, 40, 60, 80, and 100%
of critical height) against soil mass to investigate the
variation of FOS.

GWT at 1.1 m
1.05
1
0.95
0.9

FOS
0.85
0.8
0.75
(a)
0.7
0 1 2 3 4
Time (Days)

GWT at 2.5 m
1.05
1
0.95
0.9
FOS

0.85
0.8
0.75
0.7
0 2 4 6 8 10 12 14 16 18
Time (Days)
Figure 8. Variation of FOS with time under rainfall event

Table 2. Difference between negative and positive active

(b) thrusts for different time steps under rainfall event
Figure 7. Variation of active earth pressure profile with time
Time The Resultant Force (kN/m)
under rainfall event: (a) GWT at 1.1 m; (b) GWT at 2.5 m
(days) GWT at 2.5 m GWT at 1.1 m
0 1.757 24.522
5 ANALYSES OF THE RESULTS 1 2.132 25.130
2 2.549 25.924
According to the results obtained from SEEP/W analyses, 3 2.961 26.850
it would take about 18 days to fully saturate the soil with the 4 3.390
groundwater table at 2.5 m, while only 4 days with the 5 3.835
6 4.295
groundwater table at 1.1 m. Hence, the stability analyses
7 4.770
were carried out for the period of 17 days and 3 days for
8 5.263
the groundwater table at 2.5 m and 1.1 m, respectively. 9 5.780
Figure 7 shows the variation of active earth pressure 10 6.333
profile with time under rainfall event (rainfall intensity = 10- 11 6.937
7 m/sec). As described earlier, rainfall infiltration gradually
12 7.612
decreases the magnitude of negative active thrust (i.e., 13 8.380
decrease in Area 1), which leads to increase in positive net 14 9.159
active thrust. This leads to the decreases in FOS with time 15 10.038
(Figure 8). FOS decreased by 20% after 3 and 16 days of 16 11.153
continuous rainfall with groundwater table at 1.1 m and 2.5 17 12.692
 In case of the trench with shallow groundwater, when a 0
force is applied at depths below 20% of the critical height
FOS was found to be greater than 1.2. This observation Initial
Day 1
was consistent regardless of the time step (Figure 9(a)). Day 2
With the deep groundwater table (Figure 9(b)), the Day 3
20
application of resultant force does not affect FOS
significantly at the early stage of rainfall infiltration. This is
because the magnitude of the resultant forces is relatively
small. 7 days after the start of rainfall infiltration, the
application of resultant force at any depths less than 60% 40

Depth (%)
of critical height increased FOS greater than 1.2 (i.e.,
shaded area in Figure 9(b)). This indicates that the stability
of unsupported vertical trench can be effectively improved
by applying the estimated resultant force at depth in the 60
range of 0 and 60% of the critical height.

6 SUMMARY AND CONCLUSIONS

80
This research focuses on the estimation of active earth
pressure and the stability analysis of unsupported vertical
trenches considering the infiltration of rainfall in
unsaturated soils. Traditional earth pressure theories were 100
extended to account for the effects of soil’s 1.0 1.1 1.2 1.3 1.4 1.5 1.6
unsaturated conditions. Also, geotechnical modeling FOS
software (SEEP/W and SLOPE/W) was used to conduct
(a)
numerical analyses, simulating different groundwater
levels and rainfall intensities. 0
The results demonstrated that rainfall infiltration affects
the stability of unsupported vertical trenches by altering the
active earth pressure profile. The depth of the tension
zone, where the net horizontal stress is zero, was found to
decrease under rainfall events. Based on the analysis 20
results, the outcomes of the numerical modeling conducted
in this paper reveal that the resultant force in unsupported
vertical trenches with shallow groundwater table can
significantly increase the FOS. However, in unsupported 40
Depth (%)

vertical trenches with deep groundwater levels, the overall

increase in FOS is comparatively lower. Additionally, it has
been observed that in trenches with a deep groundwater
table, the resultant force has a more noticeable effect when
60
applied against the soil mass at any depths less than 60%
of the critical height.
The findings of this study have important implications Initial
Day 1
for the design and construction of unsupported trenches, Day 2
especially in regions experiencing changing weather 80 Day 3
patterns and increased rainfall intensity due to climate Day 5
Day 7
change. By incorporating the effects of rainfall infiltration, Day 15
engineers and constructors can make better engineering Day 17
decisions in estimating the stability of unsupported vertical 100
trenches in terms of shoring and load placement. This will 1.0 1.2 1.4 1.6
contribute to ensuring the safety of field workers within
unsupported vertical trenches and reducing the risk of FOS
accidents and injuries associated with trench failures. (b)

Figure 9. Variation of FOS with time for different depth of

resultant force application: (a) GWT at 1.1 m; (b) GWT at
2.5 m
REFERENCES

Fredlund, D.G, Xing, A. and Huang, S. 1994. Predicting the

permeability function for unsaturated soils using
the soil-water characteristic curve, Canadian
Geotechnical Journal, 31(4): 533-546.
Jannadi, O. A. 2008 Risks associated with trenching works
in Saudi Arabia. Building and Environment, 43(5):
776-781.
Lu, N., Godt, J.W and Wu, D.T. 2010. A closed‐form
equation for effective stress in unsaturated soil,
Water Resources Research, 46(5): 1-14.
Lu, N and Likos, W.J. 2004. Unsaturated soil mechanics,
John Wiley & Sons. Inc., Hoboken, USA.
Lu, N. and Likos, W.J. 2006. Suction stress characteristic
curve for unsaturated soil, Journal of
Geotechnical and Geoenvironmental
Engineering, 132(2): 131-142.
Oh, W.T. and Vanapalli, S.K. 2010 Influence of rain
infiltration on the stability of compacted soil
slopes, Computers and Geotechnics, 37(5): 649-
657.
Oh, W.T and Vanapalli, S.K. 2013. Interpretation of the
bearing capacity of unsaturated fine-grained soil
using the modified effective and the modified total
stress approaches, International Journal of
Geomechanics, 13(6): 769-778.
Oh, W.T. and Bajestani, M. 2023. Effect of soil-water
characteristic curve on the stability of
unsupported vertical trenches in unsaturated
soils. In Proceedings of the 8th International
Conference on Unsaturated Soils (UNSAT 2023),
E3S Web of Conferences, 382, 13002.
OSHA. (2022, July 14). Alarming rise in trench-related
fatalities [Occupational Safety and Health
Administration].
Rankine, W.M.J. 1857. On the stability on loose earth,
Philosophic Transactions of Royal Society,
London, Part I: 9-27.
Philips, Z. (2023, January 31). 2022 was a bad year for
trench deaths. Construction Dive.
https://www.constructiondive.com/news/trench-
deaths-OSHA-2022-trends/641593/
Vanapalli, S.K., Fredlund, D.G. and Pufahl, D.E. 1996.
Model for the prediction of shear strength with
respect to soil suction. Canadian Geotechnical
Journal, 33(3): 379-392.
A framework for estimating the matric suction of
unsaturated soils using two artificial intelligence
techniques
Junjie Wang & Sai K. Vanapalli
Department of Civil Engineering – University of Ottawa, Ottawa, Ontario, Canada

ABSTRACT
A major obstacle in implementing the state-of-the-art understanding of the mechanics of unsaturated soils into geotechnical
and geo-environmental engineering practice is associated with a lack of quick, reliable, and economical techniques for
matric suction measurement. Matric suction is a key stress state variable that significantly influences the hydro-mechanical
behaviour of unsaturated soils. For this reason, in this paper, two artificial intelligence (AI) models were developed for
estimating matric suction based on the support vector regression (SVR) and multivariate adaptive regression spline
(MARS) algorithms. The results suggest that both models can reasonably estimate matric suction. Compared to the MARS
model, the SVR model is capable of achieving a higher accuracy. Nonetheless, the MARS model facilitates sensitivity
analysis and the selection of essential input parameters. An empirical equation is proposed based on the MARS model for
the matric suction estimation of low plastic soils with a plasticity index equal to or less than seven. Finally, a framework is
proposed for the estimation of matric suction, which combines the strengths of both SVR and MARS models. The study is
promising for engineers to implement the mechanics of unsaturated soils into practice because the input parameters used
in this study can be determined quickly from conventional soil tests in the laboratory.

RÉSUMÉ
Un obstacle majeur à la mise en œuvre de la compréhension de pointe de la mécanique des sols non saturés dans la
pratique de l'ingénierie géotechnique et géoenvironnementale est lié à un manque de techniques rapides, fiables et
économiques de mesure de succion matricielle. La succion matricielle est une variable d'état de contrainte clé qui influence
considérablement le comportement hydromécanique des sols non saturés. Pour cette raison, dans cet article, deux
modèles d'intelligence artificielle (IA) ont été développés pour estimer la succion matricielle en se basant sur les
algorithmes de régression par vecteur de support (SVR) et de régression par splines adaptatives multivariées (MARS).
Les résultats suggèrent que les deux modèles peuvent estimer raisonnablement la succion matricielle. Comparé au
modèle MARS, le modèle SVR présente une plus grande précision et efficacité. Néanmoins, le modèle MARS facilite
l'analyse de sensibilité et la sélection des paramètres d'entrée essentiels. Une équation empirique est proposée, basée
sur le modèle MARS, pour l'estimation de la succion matricielle des sols ayant un indice de plasticité égal ou inférieur à
sept. Enfin, un cadre est également proposé pour l'estimation de la succion matricielle, combinant les forces des modèles
SVR et MARS. Cette étude est prometteuse pour les ingénieurs afin de mettre en œuvre la mécanique des sols non
saturés dans la pratique, car les paramètres d'entrée utilisés dans cette étude peuvent être déterminés rapidement à partir
d'essais de sol conventionnels en laboratoire.

1 INTRODUCTION rainfall, drought, and the melting of glaciers as a result of

continued global warming effects alter the matric suction in
Matric suction, (ua – uw), and net normal stress, (n – ua), soils (Rahardjo et al., 2019). The variation in water content
are two independent stress state variables required for the of the soil in the vadose zone can lead to significant matric
reliable interpretation of the mechanical and hydraulic suction changes and influence the engineering properties
behaviour of unsaturated soils (Fredlund & Morgenstern, that include the shear strength (Vanapalli et al., 1996;
1976; Sun et al., 2007; Zhou et al., 2012). As the stresses Kizza et al., 2020), volume change (Fredlund et al., 1995;
are typically measured considering atmospheric pressure Liu et al., 2020) and the coefficient of permeability
as a reference, the net normal stress, (n – ua) can be (Fredlund & Rahardjo, 1993; Zhang et al., 2020), in an
easily calculated using the in-situ density and thickness of unsaturated soil as shown in Fig. 1. Several challenges
the soil. However, it is challenging to measure the matric associated with unsaturated soils behaviour are typically
suction in the field or laboratory. Matric suction, (ua – uw) is associated with the non-linear variation of matric suction
expressed as the difference between the pore-air pressure, (Fredlund et al., 2012). Due to this reason, rapid and
ua and pore-water pressure, uw, which is associated with reliable determination of the matric suction is critical for
the capillarity and the interaction between soil particles and monitoring and providing early warning of geohazards, as
pore water (Li et al., 2014; Lu, 2016; Zhang et al., 2022). well as protecting the geo-infrastructures.
Since matric suction is sensitive to changes in the water Several methods are available for the determination of
content of the surrounding environment, it varies the matric suction using both direct and indirect methods
significantly due to evaporation, infiltration, and rise of the from experiments (Fredlund & Rahardjo, 1993). The matric
groundwater table (Oh et al., 2016). Furthermore, intense suction can be measured in the laboratory using a suction
plate, pressure plate, and Tempe cell (Sawangsuriya et al., economical, and reliable estimation of matric suction of
2009; Power & Vanapalli, 2010; Tao et al., 2021) and in the unsaturated soils with a plasticity index, Ip ≤ 7.
field using a tensiometer, thermal conductivity sensor and
filter paper (Marinho et al., 2008; Power & Vanapalli, 2010;
Puppala et al., 2012; Saha & Sekharan, 2021). 2 METHODOLOGY

2.1 Artificial Intelligence (AI) algorithm

2.1.1 Support Vector Regression

The Support Vector Regression (SVR) method, which is

based on statistical learning theory, can be used to create
a model with a high capacity for generalization despite the
small amount of data sample required. (Vapnik & Chapelle,
2000; Pasolli et al., 2015). The SVR method is capable of
providing superior performance in processing data with
non-linear characteristics. The SVR method has been
increasingly widely used in geotechnical engineering,
where experimental data are non-linear, scarce and
valuable (Boadu, 2020; Nhu et al., 2020). The target of
SVR is to obtain a hyperplane (Fig. 2), as shown in Eq. 1.
Figure 1. The role of matric suction in analyzing the
behaviour of unsaturated soils

A range of matric suction from 0 to 1000 kPa in soils is

of significant interest in engineering practice (Vanapalli et
al., 2008; Sawangsuriya et al., 2009; Habasimbi &
Nishimura, 2019). As discussed earlier, engineering
properties are sensitive and highly non-linear in nature in
this matric suction range. However, presently, there is no
single device that is capable of reliably measuring the
matric suction for this range. Besides, the measurement of
matric suction is typically hindered by limitations that
include costly equipment, time-consuming measurement,
and sophisticated operation that requires highly trained Figure 2. Structure of support vector regression
professionals (Leong et al., 2007; Sreedeep & Singh, 2011;
Deka et al., 2021).
In recent years, machine learning and artificial 𝑓(𝑥) = 𝑤 𝑇 𝜙(𝑋) + 𝑏, 𝑤 ∈ 𝑅𝑛 , 𝑏 ∈ 𝑅 [1]
intelligence techniques (hereinafter termed AI techniques)
have gained popularity in the field of geotechnical
engineering to lower costs and increase efficiency. Many where, w is a weight vector; X is the input vector; 𝜙(𝑋) is
studies have demonstrated that AI techniques can be used
data after mapping into high-dimensional space through
as an efficient and reliable tool for estimating the non-linear
the kernel function; b is a bias; n is the number of inputs.
properties of unsaturated soils (Pham et al., 2019; Ly &
An empirical tolerance deviation value, 𝜀 is defined to
Pham, 2020; Pham et al., 2020; Gao et al., 2021; Jalal et
produce a margin area. The loss is calculated when the
al., 2021; Onyelowe & Shakeri, 2021). However, studies
absolute value of the difference between the observed
related to the estimation of matric suction of unsaturated
and estimated value is larger than 𝜀, as shown in Eq. 2.
soils using AI techniques are rather limited (Cheng et al.,
2020). Therefore, studies related to the estimation of matric
suction from AI techniques are of significant interest.
𝜉 = 𝑦𝑖 − (𝑓(𝑥𝑖 ) + 𝜀), if 𝑦𝑖 − 𝑓(𝑥𝑖 ) > 𝜀
In this study, models are developed using support { 𝑖
𝜉𝑖 = 0, otherwise
vector regression (SVR) and multivariate adaptive [2]
regression spline (MARS) algorithms based on AI 𝜉̂ = (𝑓(𝑥𝑖 ) − 𝜀) − 𝑦𝑖 , if 𝑦𝑖 − 𝑓(𝑥𝑖 ) < −𝜀
{
algorithms for estimating the matric suction in the range of 𝜉̂ = 0, otherwise
0 to 1500 kPa. The experimental data collected from
published literature were used to train and validate the
proposed models. A sensitivity analysis is conducted to where, 𝜀 is the tolerance deviation value; ξ and 𝜉̂ are the
provide some insight with respect to the significance of additional slack variables. The residual error of data points
essential input parameters. Finally, a framework that mapped inside the margin zone is error-free. However, a
combines the strengths of the two investigated AI models, residual error, ξ is generated based on the distance
namely SVR and MARS, is proposed to provide a rapid, between data points outside this area and the margin
area's border, which is defined by support vectors.
Therefore, the optimization goal of the SVR model is to The basis function in MARS is a tensor product of the
maximize the width of the area and minimize the loss spline function. The modelling process of MARS consists
function, as shown in Eq. 3. of three major steps: forward process, backward pruning
process, and model selection. In the forward process, data
is segregated by picking nodes in a self-adaptive manner.
𝑚𝑖𝑛
1
∥ 𝑤 ∥2 + 𝐶 ∑𝑚 ̂
𝑖=1 (𝜉𝑖 , 𝜉𝑖 )
When a knot is picked, two new basis functions will be
𝑤,𝑏,𝜉𝑖 ,𝜉𝑖 2 generated. The MARS will keep finding more knots based
s.t. 𝑓(𝑥𝑖 ) − 𝑦𝑖 ≤ 𝜀 + 𝜉𝑖 [3] on the input data until a highly non-linear overfitting model
𝑦𝑖 − 𝑓(𝑥𝑖 ) ≤ 𝜀 + 𝜉̂𝑖 is produced. A higher accuracy can be typically achieved,
𝜉𝑖 ≥ 0, 𝜉̂𝑖 ≥ 0, 𝑖 = 1,2, … 𝑚. if more knots are used with the training data. However, a
balance is needed to avoid overfitting. Once the full set of
knots has been found, the knots that contribute little to the
To estimate the matric suction, a high-dimensional data model can be removed, which is known as the "back
set that includes several inputs must be transferred through pruning" process. Finally, cross-validation can be used to
kernel transformation. Such a dataset can be used for find the optimal number of knots. Therefore, the MARS
linear separation in a higher-dimensional space. Finally, model can precisely handle the estimation problem
the SVR can be used to fit the data more efficiently. involving large amounts of non-linear data.

2.1.2 Multivariate Adaptive Regression Spline 2.2 Input parameter selection

The Multivariate Adaptive Regression Spline (MARS) The selection of inputs is critical in developing reliable
method can generate a piecewise linear model, as shown estimation models. In this paper, important state and
in Eq. 4, that can be used for modelling the nonlinearity of material variables such as the particle size distribution
matric suction. (PSD) comprising fractions of sand (x1), silt (x2), and clay
(x3), specific gravity (x4), the initial void ratio (x5), the
Atterberg limits comprising liquid limit (x6), plastic limit (x7),
𝑛 and plasticity index (x8), dry unit weight (x9), and degree of
(𝑢𝑎 − 𝑢𝑤 ) = 𝛼0 + ∑ 𝛼𝑖 ∗ 𝐵𝐹𝑖 [4] saturation (x10) were chosen as input for the estimation of
𝑖=1
matric suction.

2.3 Data preparation

where BFi is a basis function; 𝛼0 is a constant; 𝛼𝑖 is
constant coefficients of each basis function. The data were collected from the published literature to
create a dataset, as shown in Table 1.

Table 1. Experimental data gathered from the published literature

ID Classification1 References Testing method Location

1-9 SC, ML, CL Sawangsuriya (2006) Triaxial cell2 US
10-13 CL, CL-ML Khalili and Zargarbashi (2010) Triaxial cell2 AU
14-20 SC, SM Rahardjo et al. (2004) Pressure plate SG
21 SM Singh and Kuriyan (2003) Tensiometer IN
22-24 CL Vanapalli et al. (1996) Pressure plate CA
25 SM Hoyos et al. (2015) Pressure plate US
26 SM Muhammad and Siddiqua (2021) Filter paper CA
27-40 CL-ML, ML, CL Johari et al. (2011) Pressure plate IR
41-43 MH, CL, SC Goh et al. (2014) Pressure plate CA
44-46 SC-SM, SM Hoyos et al. (2014) Pressure plate US
47 CL Cunningham et al. (2003) Tensiometer FR
48-52 SP Malaya and Sreedeep (2010) Tensiometer IN
53 SW Saha and Sekharan (2021) Tensiometer IN
54-55 CL Tripathy et al. (2021) Axis translation technique UK
56-58 ML Ye et al. (2019) Pressure plate CN
59-60 SM Suwal and Kuwano (2019) Triaxial cell2 JP
1 The soils are classified according to the Unified Soil Classification System (USCS);
2 The matric suction is determined by extending the axis-translation technique in a triaxial cell.

The experimental data of matric suction and ten soil 11 is chosen for training and testing the AI model. The
properties from 60 soils that constitute dataset size 925× possible deviation of measured matric suction caused by
the use of various testing techniques can be controlled
within an acceptable range when tests are operated testing set. To reduce the possible effect of the different
appropriately (Muñoz-Castelblanco et al., 2012). ranges of inputs, normalization is employed to map the
data into the same range. Such preprocessing techniques
2.4 The procedure of AI modelling facilitate the efficiency and accuracy of AI models.

The modelling procedure mainly includes five steps: (i) 2.5 Evaluation criteria
data preparation, (ii) data preprocessing, (iii) model
training, (iv) model optimization, and (v) estimation. The statistical indices, as shown in Table 2, can be
The prepared dataset used for AI modelling is divided employed to conduct a comprehensive evaluation for the
into two sets: 70% for the training set and 30% for the estimation of matric suction results.

Table 2. Performance indices

Name Symbol Equations3

2

∑𝑛𝑖=1 (𝑂𝑖 − 𝑂̅𝑖 )(𝑃𝑖 − 𝑃̅𝑖 )

Coefficient of determination 𝑅2 𝑅2 =
√∑𝑛 (𝑂 − 𝑂̅𝑖 )2 ∑𝑛𝑖=1 (𝑃𝑖 − 𝑃̅𝑖 )2
( 𝑖=1 𝑖 )
Variance accounted for VAF VAF = (1 − var (𝑂𝑖 − 𝑃𝑖 )/var (𝑂𝑖 )) × 100

√1 ∑ni=1 (𝑂𝑖 − 𝑃𝑖 )2
Normalized root mean squared error NRMSE n
NRMSE =
Max(𝑂𝑖 ) − Min(𝑂𝑖 )
𝑛
1
Bias Bias Bias = ∑ (𝑃𝑖 − O𝑖 )
𝑛
𝑖=1
3 n is the number of data points; Oi is the observed values; Pi is the estimated values.

3 ESTIMATION OF MATRIC SUCTION AI models are trained to achieve optimal accuracy

unconstrained by the considerations of efficiency and
Fig. 3 presents the performance of SVR and MARS in the applicability.
estimation of matric suction. In this preliminary stage, the
 Figure 3. Estimation of matric suction using SVR and MARS models

Table 3. Performance indices of the preliminary estimation of matric suction

SVR MARS
Performance indices
Training set Testing set Training set Testing set
R2 0.98 0.94 0.96 0.89
VAF 98.06 93.79 96.36 88.59
NRMSE 0.02 0.05 0.04 0.07
Bias 0.47 -3.97 0.68 7.61

Typically, the estimation is considered accurate when soils were selected. Second, the MARS algorithm was
the R2 ≥ 0.80, which means the proposed model can be used as an efficient technique for selecting essential
used to explain 80% of the variance in matric suction via inputs. Finally, the SVR model was used to provide
the established relationship between matric suction and accurate estimation using the inputs provided by the
the inputs. The values of R2 of training and testing sets for MARS model.
all models are higher than 0.85. Therefore, the models are
well-trained and provide reasonable accuracy. Table 3 4.1 Data reprocessing for soils with different plasticity
reveals that the proposed SVR model achieves the best
performance for both the training and testing sets. The The estimation results of matric suction show a difference
MARS model performs better for the training set than for in the accuracy between soils with different plasticity (Fig.
the testing set. 3). Since the complexity and quality of the dataset have a
The performances of AI models for soils with different substantial impact on the upper bound of estimation
plasticity values are presented in Fig. 3. A boundary was accuracy, the dataset is divided into two groups (i.e., soils
set at Ip = 7 since soils with Ip ≤ 7 show sand-like behaviour with Ip ≤ 7 and soils with Ip > 7).
while others show clay-like behaviour (Boulanger & Idriss,
2006). The estimation accuracy for the soils with Ip ≤ 7 is 4.2 Identify essential inputs using the MARS model
better than that for the soils with Ip > 7. This may be
attributed to the variation of soil properties with respect to The MARS model can be simplified with fewer BFs to
the range of Ip. The fluctuation of matric suction in soils with conduct high-efficiency sensitivity analysis. As indicated in
Ip ≤ 7 is primarily influenced by the particle size distribution Table 4, the MARS model selects different essential inputs
(PSD). Conversely, in soils with Ip > 7, the presence of finer for different cases. For soil with Ip ≤ 7, sand fraction, which
particles and mineral composition not only contribute to but is useful in distinguishing the water-holding capacity, for
also complicate the variability in suction. soil with various pore size distributions is required. Soils
with Ip > 7 require dry unit weight and initial void ratio, which
is strongly correlated to the soil structure relating to fine-
4 FRAMEWORK FOR ESTIMATING MATRIC grained soils. Furthermore, the degree of saturation, silt
SUCTION USING TWO AI TECHNIQUES fraction, specific gravity, and plasticity index are required
by all the models for suction estimation.
In this section, a novel framework is proposed, combining
the strengths of the SVR and MARS models to provide a
more efficient and reliable estimation of matric suction. Table 4. Identify the essential input using the MARS model
First, potential inputs characterizing the features of studied
 𝑛
VAF
Soils Essential inputs (𝑢𝑎 − 𝑢𝑤 ) = max (0, 𝛼0 + ∑ 𝛼𝑖 × 𝐵𝐹𝑖 ) [5]
Training Testing
Ip ≤ 7 x10, x2, x8, x4, x1, x6 99.85 99.15 𝑖=1

Ip > 7 x2, x3, x4, x10, x8, x9, x5 92.04 82.37

Table 5. BFs and the coefficient of Eq.5
Since the MARS model performs better in estimating the
i Basis function Coefficient Value
matric suction of soils with Ip ≤ 7, an empirical equation 1 max(0, x10 -45.9) α1 -3.03
considering essential inputs was proposed, as shown in 2 max(0, x2 -14.4) α2 2.46
Eq. 4. However, the estimation results may occasionally be 3 max(0, x2-14.4)·max(0, 72.2-x10) α3 1.54
inconsistent with theoretical and practical expectations. For 4 max(0, x2-14.4)·max(0, 38.2-x1) α4 -0.34
instance, when the soil is fully saturated, the suction value 5
max(0, x2-14.4)·max(0, x10-72.2)
α5 -0.02
is zero (i.e., the lower limit of the dataset). Due to potential ·max(0, 38.2-x1)
fluctuations in the estimated value relative to the measured max(0, x2-14.4)·max(0, 72.2-x10)
6 α6 -17.14
·max(0, x4-2.71)
value, there exists a possibility of negative estimated
max(0, x2-14.4)·max(0, 72.2-x10)
suction values. Thus, it is necessary to enhance Eq. 4 by 7
·max(0, 2.71-x4)
α7 -20.79
incorporating appropriate constraints, leading to an max(0,x2-14.4)·max(0, 72.2-x10)
improved equation, Eq. 5. This modification ensures the 8 α8 -47.15
·max(0,x4-2.71)·max(0,19.3-x1)
non-negativity of the estimated suction value (The BFs and 9
max(0, x2 -14.4)·max(0, 38.2 -x1)
α9 0.02
coefficient of Eq. 5 are presented in Table 5). ·max(0, x10 -59.4)
max(0, x2-14.4)·max(0, 72.2-x10)
10 α10 0.14
4.3 Precise estimation and analysis using SVR ·max(0, x4 -2.71)·max(0, x1 -21.9)
max(0, x2 -14.4)·max(0, 72.2 -x10)
11 α11 11.84
·max(0, x4 -2.71)·max(0, 21.9-x1)
As shown in Fig. 4, the SVR model provides a reasonably 12 max(0, x4-2.66) α12 505.48
accurate estimation of matric suction for soils with Ip ≤ 7. 13 max(0, 2.66-x4) α13 3070.79
- - α0 -3.46

Figure 4 Estimation of matric suction based on SVR using essential inputs

Therefore, the proposed framework achieved efficient estimate the matric suction of soils with Ip ≤ 7. Based on
sensitivity analysis and precise estimation of matric suction the findings, the following key conclusions can be drawn:
for the soil with Ip ≤ 7 by combining the strengths of the (i) The two proposed AI models provide good results in
SVR and MARS models. estimating matric suction using fundamental soil
properties. The SVR model can reliably estimate matric
suction for various soils with high accuracy. Considering
5 SUMMARY only the essential inputs, the MARS model is a simple tool
for estimating the matric suction rapidly for use in
This study focuses on the development of AI models engineering practice applications.
utilizing the SVR and MARS algorithms for reliable (ii) Based on the MARS model, a practical and simple
estimation of matric suction in unsaturated soils. A novel empirical equation was proposed to estimate the matric
framework is proposed, which leverages the strengths of suction of soils with Ip ≤ 7 by utilizing essential inputs.
both AI algorithms to estimate matric suction. (iii) By employing the proposed framework that
Subsequently, the suggested framework is employed to combines the strengths of SVR and MARS models, it is
possible to estimate the matric suction of soils efficiently matric suction ranges and different stress conditions.
and accurately with Ip ≤ 7. International Journal of Geosciences, 10(1), 39-56.
The proposed model has some limitations that can be Hoyos, L. R.,Velosa, C. L., & Puppala, A. J. (2014).
summarized as follows: First, the proposed model is Residual shear strength of unsaturated soils via
inherently constrained by dataset limitations, such as suction-controlled ring shear testing. Engineering
difficulties in estimating soil properties for soils with Ip > 7 Geology, 172, 1-11.
in comparison to soils with Ip ≤ 7. Second, estimation Hoyos, L. R.,Suescún-Florez, E. A., & Puppala, A. J.
results can occasionally fail to align either with theoretical (2015). Stiffness of intermediate unsaturated soil from
or practical considerations that may occur at the simultaneous suction-controlled resonant column and
boundaries of the dataset range. It is crucial to incorporate bender element testing. Engineering Geology, 188, 10-
some restrictions into the model that are consistent with the 28.
state-of-the-art understanding of unsaturated soil Jalal, F. E.,Xu, Y.,Iqbal, M.,Javed, M. F., & Jamhiri, B.
mechanics. (2021). Predictive modeling of swell-strength of
expansive soils using artificial intelligence approaches:
ANN, ANFIS and GEP. Journal of Environmental
6 REFERENCES Management, 289, 112420.
Johari, A.,Habibagahi, G., & Ghahramani, A. (2011).
Boadu, F. K. (2020). A support vector regression approach Prediction of SWCC using artificial intelligent systems:
to predict geotechnical properties of soils from electrical A comparative study. Scientia Iranica, 18(5), 1002-
spectra based on Jonscher parameterizationSoil 1008.
electrical geotechnical properties. Geophysics, 85(3), Khalili, N., & Zargarbashi, S. (2010). Influence of hydraulic
EN39-EN48. hysteresis on effective stress in unsaturated soils.
Boulanger, R. W., & Idriss, I. (2006). Liquefaction Géotechnique, 60(9), 729-734.
susceptibility criteria for silts and clays. Journal of Kizza, R.,Leong, E., & Rahardjo, H. (2020). Matric suction
Geotechnical and Geoenvironmental Engineering, and shear strength of dynamically compacted soils. In
132(11), 1413-1426. Unsaturated Soils: Research & Applications (pp. 1323-
Cheng, Z.-L.,Zhou, W.-H., & Garg, A. (2020). Genetic 1327). CRC Press.
programming model for estimating soil suction in Leong, E.,Widiastuti, S.,Lee, C., & Rahardjo, H. (2007).
shallow soil layers in the vicinity of a tree. Engineering Accuracy of suction measurement. Géotechnique,
Geology, 268. 57(6), 547-556.
Cunningham, M.,Ridley, A.,Dineen, K., & Burland, J. Li, J.,Zhao, C.,Cai, G., & Guo, Y. (2014). A model
(2003). The mechanical behaviour of a reconstituted considering solid-fluid interactions stemming from
unsaturated silty clay. Géotechnique, 53(2), 183-194. capillarity and adsorption mechanisms in unsaturated
Deka, A.,Vadlamudi, S., & Sekharan, S. (2021). Effect of expansive clays. Chinese science bulletin, 59(26),
High Suction Measurement Technique on the SWCC of 3314-3324.
Expansive Soil. In Transportation, Water and Liu, G.,Toll, D. G.,Kong, L., & Asquith, J. D. (2020). Matric
Environmental Geotechnics (pp. 109-118). Springer. Suction and Volume Characteristics of Compacted
Fredlund, D.,Huang, S.,Clifton, A.,Wang, Q.,Barbour, Clay Soil under Drying and Wetting Cycles.
S.,Ke, Z., & Fan, W. (1995). Matric suction and Geotechnical Testing Journal, 43(2).
deformation monitoring at an expansive soil site in Lu, N. (2016). Generalized soil water retention equation for
southern China. Proc., Proceedings of International adsorption and capillarity. Journal of Geotechnical and
Conference on Unsaturated Soils. Rotterdam, Geoenvironmental Engineering, 142(10), 04016051.
Netherlands: AA Balkema, 835-862. Ly, H.-B., & Pham, B. T. (2020). Prediction of shear
Fredlund, D. G., & Morgenstern, N. (1976). Constitutive strength of soil using direct shear test and support
relations for volume change in unsaturated soils. vector machine model. The Open Construction &
Canadian Geotechnical Journal, 13(3), 261-276. Building Technology Journal, 14(1).
Fredlund, D. G., & Rahardjo, H. (1993). Soil mechanics for Malaya, C., & Sreedeep, S. (2010). A study on the
unsaturated soils. John Wiley & Sons. influence of measurement procedures on suction-water
Fredlund, D. G.,Rahardjo, H., & Fredlund, M. D. (2012). content relationship of a sandy soil. Journal of Testing
Unsaturated soil mechanics in engineering practice. and Evaluation, 38(6), 691-699.
John Wiley & Sons. Marinho, F.,Take, W., & Tarantino, A. (2008).
Gao, Y.,Li, Z.,Sun, D. A., & Yu, H. (2021). A simple method Measurement of matric suction using tensiometric and
for predicting the hydraulic properties of unsaturated axis translation techniques. In Laboratory and field
soils with different void ratios. Soil and Tillage testing of unsaturated soils (pp. 3-19). Springer.
Research, 209, 104913. Muhammad, N., & Siddiqua, S. (2021). Moisture-
Goh, S.,Rahardjo, H., & Leong, E. (2014). Shear strength dependent resilient modulus of chemically treated
of unsaturated soils under multiple drying-wetting subgrade soil. Engineering Geology, 285, 106028.
cycles. Journal of Geotechnical and Geoenvironmental Muñoz-Castelblanco, J.,Pereira, J.-M.,Delage, P., & Cui,
Engineering, 140(2), 06013001. Y.-J. (2012). The water retention properties of a natural
Habasimbi, P., & Nishimura, T. (2019). Soil water unsaturated loess from northern France.
characteristic curve of an unsaturated soil under low Géotechnique, 62(2), 95-106.
Nhu, V.-H.,Hoang, N.-D.,Duong, V.-B.,Vu, H.-D., & Tien Singh, D., & Kuriyan, S. J. (2003). Estimation of
Bui, D. (2020). A hybrid computational intelligence unsaturated hydraulic conductivity using soil suction
approach for predicting soil shear strength for urban measurements obtained by an insertion tensiometer.
housing construction: a case study at Vinhomes Canadian Geotechnical Journal, 40(2), 476-483.
Imperia project, Hai Phong city (Vietnam). Engineering Sreedeep, S., & Singh, D. N. (2011). Critical review of the
with Computers, 36(2), 603-616. methodologies employed for soil suction measurement.
Oh, W. T.,Vanapalli, S. K.,Qi, S., & Han, Z. (2016). International Journal of Geomechanics, 11(2), 99-104.
Estimation of the variation of matric suction with respect Sun, D.,Sheng, D.,Cui, H., & Sloan, S. (2007). A density‐
to depth in a vertical unsaturated soil trench associated dependent elastoplastic hydro‐mechanical model for
with rainfall infiltration. E3S Web of Conferences, 9, unsaturated compacted soils. International Journal for
15003. Numerical and Analytical Methods in Geomechanics,
Onyelowe, K. C., & Shakeri, J. (2021). Intelligent prediction 31(11), 1257-1279.
of coefficients of curvature and uniformity of hybrid Suwal, L. P., & Kuwano, R. (2019). Matric suction and
cement modified unsaturated soil with NQF inclusion. stiffness measurement on soil containing fines at low
Cleaner Engineering and Technology, 4, 100152. stress state. Japanese Geotechnical Society Special
Pasolli, L.,Notarnicola, C.,Bertoldi, G.,Bruzzone, Publication, 7(2), 297-304.
L.,Remelgado, R.,Greifeneder, F.,Niedrist, G.,Della Tao, G.,Li, Z.,Liu, L.,Chen, Y., & Gu, K. (2021). Effects of
Chiesa, S.,Tappeiner, U., & Zebisch, M. (2015). Contact Angle on the Hysteresis Effect of Soil-Water
Estimation of soil moisture in mountain areas using Characteristic Curves during Dry-Wet Cycles.
SVR technique applied to multiscale active radar Advances in civil engineering, 2021.
images at C-band. IEEE Journal of Selected Topics in Tripathy, S.,Al-Khyat, S., & Cleall, P. J. (2021). Impact of
Applied Earth Observations and Remote Sensing, 8(1), single and multiple specimen suction control
262-283. oedometer testing on the measurement of the soil–
Pham, B. T.,Nguyen, M. D.,Van Dao, D.,Prakash, I.,Ly, H.- water characteristic curve. Geotechnical Testing
B.,Le, T.-T.,Ho, L. S.,Nguyen, K. T.,Ngo, T. Q., & Journal, 44(6).
Hoang, V. (2019). Development of artificial intelligence Vanapalli, S.,Fredlund, D.,Pufahl, D., & Clifton, A. (1996).
models for the prediction of Compression Coefficient of Model for the prediction of shear strength with respect
soil: An application of Monte Carlo sensitivity analysis. to soil suction. Canadian Geotechnical Journal, 33(3),
Science of The Total Environment, 679, 172-184. 379-392.
Pham, B. T.,Qi, C.,Ho, L. S.,Nguyen-Thoi, T.,Al-Ansari, Vanapalli, S. K.,Nicotera, M., & Sharma, R. S. (2008). Axis
N.,Nguyen, M. D.,Nguyen, H. D.,Ly, H.-B.,Le, H. V., & translation and negative water column techniques for
Prakash, I. (2020). A novel hybrid soft computing model suction control. In Laboratory and field testing of
using random forest and particle swarm optimization for unsaturated soils (pp. 33-48). Springer.
estimation of undrained shear strength of soil. Vapnik, V., & Chapelle, O. (2000). Bounds on error
Sustainability, 12(6), 2218. expectation for support vector machines. Neural
Power, K. C., & Vanapalli, S. K. (2010). Modified null computation, 12(9), 2013-2036.
pressure plate apparatus for measurement of matric Ye, Y.-X.,Zou, W.-L.,Han, Z., & Liu, X.-W. (2019).
suction. Geotechnical Testing Journal, 33(4), 335-341. Predicting the entire soil-water characteristic curve
Puppala, A. J.,Manosuthkij, T.,Nazarian, S.,Hoyos, L. R., & using measurements within low suction range. Journal
Chittoori, B. (2012). In situ matric suction and moisture of Mountain Science, 16(5), 1198-1214.
content measurements in expansive clay during Zhang, C.,Hu, S.,Qiu, Z., & Lu, N. (2022). A poroelasticity
seasonal fluctuations. Geotechnical Testing Journal, theory for soil incorporating adsorption and capillarity.
35(1), 74-82. Géotechnique, 1-57.
Rahardjo, H.,Kim, Y., & Satyanaga, A. (2019). Role of Zhang, J.,Zhang, S., & Xu, T. (2020). Individual effect of
unsaturated soil mechanics in geotechnical matric suction on soil microstructure and hydraulic
engineering. International Journal of Geo-Engineering, conductivity. KSCE Journal of Civil Engineering, 24(5),
10(1), 1-23. 1448-1459.
Rahardjo, H.,Aung, K.,Leong, E. C., & Rezaur, R. (2004). Zhou, A.-N.,Sheng, D.,Sloan, S. W., & Gens, A. (2012).
Characteristics of residual soils in Singapore as formed Interpretation of unsaturated soil behaviour in the
by weathering. Engineering Geology, 73(1-2), 157-169. stress–saturation space, I: volume change and water
Saha, A., & Sekharan, S. (2021). Estimation of unsaturated retention behaviour. Computers and Geotechnics, 43,
hydraulic conductivity function: implication of low to 178-187.
high suction measurements. Acta Geophysica, 69(2),
547-559.
Sawangsuriya, A. (2006). Stiffness-suction-moisture
relationship for compacted soils. The University of
Wisconsin-Madison.
Sawangsuriya, A.,Edil, T. B., & Bosscher, P. J. (2009).
Modulus-suction-moisture relationship for compacted
soils in postcompaction state. Journal of Geotechnical
and Geoenvironmental Engineering, 135(10), 1390-
1403.
Wednesday, October 4, 2023

ADVANCED TESTING II
Effect of the degree of saturation and
mineralogy on acid generation critical time
Chevalier Corentin, Pabst Thomas
Research Institute of Mines and Environment (RIME), Canada
Department of Civil, Geological, and Mining Engineering –
Polytechnique Montréal, Montréal, Québec, Canada

ABSTRACT
Filtered tailings allow to increase geotechnical stability and minimize footprint of tailings storage facilities. Despite their
numerous advantages, they can also contribute to increase the risk for acid mine drainage (AMD) generation. In this
research, multicomponent reactive transport simulations were conducted to assess the critical time tailings can remain
exposed before AMD generation would start. Laboratory kinetics tests were performed on reactive mine tailings and the
evolution of pore water geochemistry was monitored with time. Numerical simulations were calibrated on laboratory tests
and then extrapolated for other mineralogies. Results showed that critical time could be predicted, as a function of tailings
mineralogy and degree of saturation. A predictive equation to estimate critical time is proposed and discussed in this paper.
Predicting critical time could be useful to optimize tailings management and improve the geochemical stability of filtered
tailings storage facilities.

RÉSUMÉ
Les résidus filtrés permettent d’améliorer la stabilité géotechnique et de minimiser l’empreinte des parcs à résidus miniers.
Malgré leurs nombreux avantages, la filtration des résidus peut aussi contribuer à augmenter le risque de génération de
drainage minier acide (DMA). Dans cette étude, des simulations de transport réactif multicomposants ont été effectuées
afin d'évaluer le temps critique avant que les résidus exposés génèrent du DMA. Des tests en laboratoires ont été réalisés
sur des résidus réactifs et l’évolution de la qualité des eaux interstitielle au cours du temps a été mesurée. Des simulations
numériques ont été calibrés sur les résultats de laboratoire, puis extrapolés pour d’autres minéralogies. Les résultats
montrent que le temps critique peut être prédit en fonction de la minéralogie et du dégrée de saturation. Une équation de
prédiction est proposée et discutée dans cet article. Prédire le temps critique pourrait permettre d'optimiser la gestion des
résidus et d'améliorer la stabilité géochimique des parcs à résidus filtrés.

1 INTRODUCTION Filtered tailings are typically characterized by a high solid

content (Pm > 85%; Ulrich (2019)) and are disposed of
The mining industry produces large amounts of wastes using trucks or conveyors at a degree of a saturation
(Aubertin et al. 2011), including mine tailings, which are between 60% and 80% (Davies 2011; Zorzal et al. 2020).
usually transported as a slurry (Simms 2017) and Filtered tailings have a significantly stronger properties, are
deposited in tailings storage facilities (TSF) surrounded by less prone to liquefaction, and thus typically do not require
containment structures (Oldecop and Rodari 2017). confinement (Williams 2021).
Tailings dams are susceptible to geotechnical failure risks However, exposed mine waste may negatively impact
(Bowker and Chambers 2017; Chambers and Higman the environment, due to their mineralogy (Hoffmann et al.
2011; Rico et al. 2008), primarily caused by the 1981; Nordstrom et al. 2015; Price 2009). Sulfide minerals
unconsolidated and low geotechnical properties of (e.g., pyrite or pyrrhotite) are, indeed, common in tailings
embankment materials (typically mine waste), and the high and their exposure to atmospheric condition may favor their
water pressure in the TSF (Azam and Li 2010; Rico et al. oxidation (Lindsay et al. 2015) and the generation of acid
2008; Strachan and Goodwin 2015). Tailings dam failures mine drainage (AMD) (Blowes et al. 2014). Dissolution of
remain frequent (Lyu et al. 2019; Williams 2021) and neutralizing minerals, such as carbonates (e.g., calcite and
regularly lead to loss of life and considerable environmental dolomite) can, however, temporarily neutralize the acid and
damages (e.g., Brumadinho, Brazil in 2019 ; Thompson et postpone the generation of AMD (Bouzahzah et al. 2014;
al. (2020)). Dold 2005; Lottermoser 2010). Prediction of tailings
The industry is therefore considering innovative acidification and neutralization potential is therefore crucial
management approaches to ensure TSF short- and long- for mining projects (Elghali et al. 2023). Static tests aim to
term stability (Williams 2021). Dewatering technologies, predict the neutralization potential (NP) and the
and particularly filtration, are an alternative to conventional acidification potential (AP) (Bouzahzah et al. 2015; Elghali
slurry tailings management (Davies 2011; McPhail et al. et al. 2018; Sobek et al. 1978), but give no indication of the
2019; Moreno et al. 2018). Decreasing water content delay before AMD would appear. Kinetic tests can typically
indeed contributes to improve tailings physical stability, be performed in the laboratory or in the field to quantify
minimizes TSF footprint and maximizes water recirculation leaching parameters, reactivity and chemical species
in ore processing (Gutierrez and Oldecop 2021; Lupo and leaching rates (Benzaazoua et al. 2004; Chopard et al.
Hall 2010; Schoenbrunn 2011; Ulrich and Coffin 2013).
2015), but these are typically very time consuming (from by gas pycnometer (ASTM-D5550 2014) and was around
several months to several years). Gs = 3.04.
This research therefore aimed to propose an approach
to evaluate filtered tailings critical time (i.e., the time 2.2 Laboratory kinetic tests
reactive filtered tailings can be left exposed before AMD
generation starts) as a function of their mineralogy and Tailings geochemical behaviour was evaluated using
degree of saturation. Laboratory kinetic cells were kinetic tests. The objective was to monitor the evolution of
performed and numerical simulations were calibrated on leachate quality and determine the critical time, i.e., the
experimental results. An analytical formula was derived time before pH became less than 6.0.
from numerical models to propose a simple approach to Tailings were compacted in layers of approximatively
predict critical time from tailings carbonate and pyrite 5.4 ± 0.2 cm thick and at a target porosity of around 0.44 ±
contents. 0.2 in small plexiglass cells (15 cm in height and 8 cm in
internal diameter ; Figure 1).

2 MATERIALS AND METHODS

2.1 Tailings properties

Acid generating mine tailings AGT were sampled from the

concentrator of a polymetallic mine, located in Quebec
Province, Canada. Tailings were kept under water to avoid
oxidation during transport and were then homogenized and
characterized in the laboratory.
Acid generation potential was evaluated by tailings total
sulfur and carbon contents measured on 3 duplicates using
induction furnaces (LECO CS744 Analyzer). The AP was
calculated by assuming that the total S content was
originating only from sulfides (pyrite, according to XRD
analysis; AP = %Stotal x 31.25) and sulfate concentration
was deemed neglectable. NP was estimated based on the
measured C content (NP = %C x 83.3), assuming that all
the carbon in the sample originated from calcite and
ankerite (Lawrence and Wang 1997). Mineralogical phases Figure 1. Kinetic cell setup. A total of 5 cells were used to
were identified by X-ray diffraction (XRD ; Bruker AXS D8 evaluate the effect of the degree of saturation on the critical
Advance diffractometer, precision 0.5-2%). Tailings time.
presented a high acid potential (AP = 382 kgCaCO3 eq/t),
with pyrite being the main sulfide (FeS2 ; 17.6 wt%) and
some trace of chalcopyrite (0.2 wt%). The buffering Flushes were regularly applied (every two weeks or so)
capacity was low (NP = 14 kgCaCO3 eq/t) due to the on top of the specimens with around 125 ml of distilled
limited content of carbonates (0.6 wt% of CaCO3 and 0.5 water (i.e., 1.25 times the tailings pore volume). Leachate
wt% of CaFe(CO3)2). Tailings also contained quartz (48 was collected at the bottom of the cells after drainage
wt%), muscovite (17 wt%), albite (5 wt%), anorthite (4 through tailings pores and characterized pH, EC, sulfates
wt%), chlorite (4 wt%), paragonite (3 wt%), orthoclase (1 (SO4), Ca and Mg concentrations. pH and EC were
wt%) and bassanite (1 wt%), but these did not contribute measured using Hanna HI5522 (pH and EC meter,
significantly to acid neutralization. Net neutralization accuracy of ±0.01 for pH and 1% for EC), SO4
potential (NNP = NP - AP) was very low with NNP = -368 concentration was determined using a spectrophotometer
kgCaCO3 eq/t, thus indicating that tailings were acid (HACH DR 3900, photometric accuracy 1%) and Ca and
generating (NNP < - 20 kgCaCO3 eq/t ; Morin and Hutt Mg concentrations using atomic absorption (PinAAcle 900
(2001)). F Perkin Elmer, relative error 1.5-2%). Analyses were
Tailings particle size distribution (PSD ; D10 = 0.0046 performed immediately after leachate had been collected.
mm, D60 = 0.038mm) was evaluated on 3 duplicates Both drainage and evaporation contributed to rapidly
(ASTM-D7928 2021). Tailings were typical of hard rock decrease the degree of saturation which was then
mine tailings (Bussière 2007; Qiu and Sego 2001) and maintained at the target value as described above.
were classified as non-plastic silts (ML ; ASTM-D2487 After the end of the experiments, cells were dismantled,
(2017)). Saturated hydraulic conductivity was estimated and total carbon and sulfur were measured using the same
using KCM model (Kozeny-Carman Modified; Mbonimpa et approaches before the tests.
al. (2002)), with a saturated hydraulic conductivity ksat =
5×10-7 m/s. Water retention curve was measured using a 2.3 Numerical simulation approach
pressure plate extractor (ASTM-D6836 2016) and results
were adjusted using van Genuchten (1980) mode (αvg = Numerical simulations were conducted using MIN3P, a
0.002 cm-1, nvg = 1.55). The specific gravity was evaluated multicomponent reactive transport model (Mayer 1999),
which can efficiently simulate the hydrogeochemical
behaviour of reactive tailings (Jurjovec et al. 2004; Kalonji-
Kabambi et al. 2020; Molson et al. 2008; Ouangrawa et al. A two-third power relationship was used for primary
2009; Pabst et al. 2017). MIN3P simulates the effect of time minerals effective rate constant (keff ; mol L-3bulk S-1),
on aqueous complexation, redox reactions and minerals allowing to update the mineral reactive surface area with
(Mayer et al. 2012). time (Lichtner 1996), while secondary mineral reactivity
Laboratory kinetic cells were simulated as one- was assumed to be constant.
dimensional (1D) models in MIN3P. The length of the
domain was 5.4 cm ± 0.2, and was discretized with 30 2.4 Calibration
control volumes (i.e., a mesh size of 1.8 ± 0.1 mm). The
surface of the exposed tailings was simulated as second Simulated hydraulic parameters were first calibrated to
type boundary (Neumann) for inflow (flushes) and mixed match leachate fluxes and degree of saturation observed
(Dirichlet/Cauchy) for reactive transport (e.g., O2 supply, in the laboratory. Then, geochemical system was calibrated
relative humidity and air temperature, 25°C). Constant with a particular focus on pH. Pyrite was responsible of the
degree of saturation in the tailings was controlled by the increase of sulfate concentration and the decrease of pH.
bottom boundary conditions with a negative pressure head. Calcite and ankerite were identified as the neutralizing
The reactive transport bottom boundary allowed free minerals and their depletion were linked to the continuous
advective mass outflux for aqueous phase (e.g., leachate acidification conditions. The relatively high reaction rate of
drainage). Tailings hydrogeological parameters carbonates allowed to maintain pH near neutrality, until
corresponded to laboratory characterization results. Initial their depletion. Sulfates, Ca and Mg cumulative
pore water chemical composition corresponded to the first concentrations and carbonate depletion (not shown here)
flush. were also used to calibrate simulations and ensure models
Mineral volume fractions were based on XRD analyses did reproduce well tailings geochemical behaviour.
and mineral dissolution-precipitation were kinetically The effective rate constant of each mineral was
controlled. Activity and equilibrium constants were iteratively adjusted until a good agreement was observed
obtained from MIN3P database and models included 12 between simulations and experimental results. Calibrated
primary and 3 secondary minerals, 2 gases (O2 / CO2) and simulations reproduced well measurements with time.
2 redox pairs (Fe2+/Fe3+ and HS-/SO42-) (Table 1). The calibrated reaction rates were usually in the same
order of magnitude as reported in other studies (Kalonji-
Kabambi et al. 2020; Mayer et al. 2012). Pyrite reaction rate
Table 1. Mineral reactions and equilibrium constants (Kim), was 6.0 x 10-9 mol L-3bulk S-1 and 2.0 x 10-8 and 2.0 x 10-9
from MIN3P databases. (am): amorphous; (aq): aqueous. mol L-3bulk S-1 for calcite and ankerite respectively. Calcite
reaction rate was higher than that of ankerite by one order
Minerals Reactions Log Kim of magnitude, which is common in mine waste (Amos et al.
2015; Mayer et al. 2012). Other reaction rates were 5.5 x
FeS2 + 7/2 O2 + H2O ↔ Fe 2+
+ 2SO42-
Pyrite
+ 2H+
-215.3 10-11 mol L-3bulk S-1 for chlorite, 5.0 x 10-7 mol L-3bulk S-1
for gypsum, 1.0 x 10-9 mol L-3bulk S-1 for ferrihydrite and 1.0
Calcite CaCO3 ↔ Ca + 2+
CO32- 8.5 x 10-11 mol L-3bulk S-1 for other minerals.
Ankerite CaFe(CO3)2 ↔ Ca2+ + Fe2+ + 2CO32- 17.1
(Fe(Mg,Mn)5Al)(Si3Al)O10(OH)8 + 28H+
Chlorite + 3O2(aq) → 5Mg2+ + Fe2+ + 5Mn2+ + 2Al3+ -2 3 TAILINGS OXIDATION AND CRITICAL TIME
+ 3H4SiO4 +12H2O
Quartz SiO2(am) + 2H2O ↔ H4SiO4 4.0 Calibrated simulations (Figure 2 ; lines) reproduced well pH
KAl2(AlSi3O10)(OH)2 +10H+ ↔ K+ + 3Al3+ variation with time for all cells (Figure 2 ; dots).
Muscovite -13.0 Initial pHs were between 7 and 8 for all cells and
+ 3H4SiO4(aq)
CaAl2Si2O8 + 8H+ → Ca2+ + 2Al3+ remained around neutrality for 58 to 77 days. The time
Anorthite -2 when pH started to decrease depended on the degree of
+ 2H4SiO4(aq)
saturation. For example, when Sr = 93% and Sr = 95%, pH
(NaAl)Si3O8 + 4H+ + 4H2O → Na2+ + Al3+ 2
Albite
+ 3H4SiO4(aq)
- remained around neutrality value over 155 days, but when
Sr = 79%, pH started to decrease after 58 days. Then pH
NaAl2 (AlSi3O10)(OH)2 + 10H+ ↔ Na+ rapidly decreased and reached within a few days values
Paragonite 17.5
+ 3Al3+ + 3SiO2 + 6H2O less than 4.0. After pH reached a value around 4, the
(KAl)Si3O8 + 4H + 4H2O → K + Al
+ + 3+
decrease was slower and more progressive, until the end
Orthoclase -2
+ 3H4SiO4(aq) of the tests. Final pH for all cells was relatively similar and
Chalcopyrite CuFeS2 + 4O2(aq) ↔ Cu2+ +Fe2+ + 2SO42- 35.3 around 3.0.
Bassanite CaSO4 ↔ Ca2+ + SO42- 4.4 The critical time (i.e., the time required for pH to
Gypsum1 CaSO4•2H2O ↔ Ca2+ + SO42− + 2H2O 4.6 decrease below 6.0) was reached between day 72 and 95.
Simulated and measured critical times were very close and
Ferrihydrite1 Fe(OH)3 + 3H+ ↔ Fe3+ + 3H2O -4.9 usually less than 10 days different. For example, critical
(H3O)Fe3(SO4)2(OH)6 + 5H+ ↔ 3Fe3+ time at Sr = 83% (duplicate) was measured around 93 days
Jarositeh1 5.4
+ 7H2O + 2SO42− (pH=6.08) and simulated at day 95.
1
Secondary mineral
2
Irreversible dissolution
 time was 121 days for AP = 21 kg CaCO3 eq/t (NP/AP =
0.5) and 235 days for 11 kg CaCO3 eq/t (NP/AP = 1.0).
The third step consisted of modifying the tailings NP.
Results indicated that critical time was strongly delayed
when NP increased. Critical time doubled when NP
doubled (Figure 3). For example, critical time was 40 days
for NP = 5 kg CaCO3 eq/t and 81 days for NP = 11 kg
CaCO3 eq/t.

Figure 2. Measured (dots) and simulated (lines) pH as a

function of time and degree of saturation. Critical time was
determined as the moment when pH decreased below 6.0.
Leachate remained around neutrality when Sr > 90%.

Critical time corresponded to the moment carbonates in

the tailings were depleted. For example, carbon content at
the end of the test (i.e., after 155 days) had decreased by
Figure 3. Simulated critical time as a function of NP and
71 wt%, 82 wt% and more than 95 wt% when degree of
degree of saturation.
saturation was 83%-1, 83%-2 and 79% respectively. When
Sr > 90%, the carbon content had decreased only by 23
wt%, indicating that the remaining content of carbonates
Critical time increased linearly with the content of
was significant and sufficient to maintain pH around 7.0.
neutralizing minerals in the tailings (Figure 4a and Figure
4c). NNP influence on the critical time was particularly
significant for elevated neutralizing potential (NP > 11 kg
4 CRITICAL TIME DETERMINATION
CaCO3 eq/t). For example, critical time was around 350
days for NNP = -328 kg CaCO3 eq/t (Figure 4b), and 650
Calibrated simulations were then used to evaluate the
days at NNP = -10 kg CaCO3 eq/t (Figure 4d) when
critical time at a larger scale and for various mineralogies.
NP = 43 kg CaCO3 eq/t while the critical time was 41 days
The effect of mineralogy was assessed by evaluating
for NNP = -328 kg CaCO3 eq/t and 49 days NNP = -10 kg
the geochemical behavior of tailings with various AP and
CaCO3 eq/t, when NP = 5 kg CaCO3 eq/t.
NP. The volumetric fractions of sulfide and carbonates
Critical time may be described from NP and NNP
minerals were adjusted to maintain a constant simulated
values. A simple model to predict critical time Tc (when Sr
mass.
< 90%) for tailings classified as acid generating tailings
The first step was to modify the degree of saturation,
(NNP < -20 kg CaCO3 eq/t) was determined (Eq. 1):
using the initial simulated values AP (339 kgCaCO3 eq/t)
and NP (11 kgCaCO3 eq/t). Critical time was strongly
influenced by the degree of saturation. Critical time was 8.40 2.11
over 1,000 days when Sr = 100%, 585 days when Sr = 95% Tc = NP (7.67 − ) + NP 2 (0.0065 − ) [1]
NNP NNP
and 81 days when Sr = 88%. The critical time was almost
independent of the degree of saturation when
Sr < 88% and around 80 days. Critical time could therefore The differences between the calculated and simulated
be significantly delayed if Sr > 90%. critical time never exceeded 21 days, with a median of 2
The second step aimed to evaluate the critical time days. Calculations were consistent with the experimental
evolutions with the AP variations. Critical time was only critical time which was observed in a period of 72 to 95
slightly influenced by AP. For example, critical time was 96 days.
days for AP = 61 kg CaCO3 eq/t and 88 days for AP = 761
kg CaCO3 eq/t. Critical time increased only when the
NP/AP ratio became greater than 0.5. For example, critical
Figure 4. Critical time as a function of NP and Sr, with NNP = -328 kg CaCO3 eq/t, and NNP = -10 kg CaCO3 eq/t)

5 DISCUSSION generate a variable amount of acidity and contamination

(Anawar 2015).
Results showed that the degree of saturation delayed the The model was developed, using experimental and
critical time when Sr > 90%. In practice, increasing (filtered numerical simulations results and has therefore been
tailings are disposed of at Sr = 80% ; Amoah et al. (2018)) validated only for mineralogies similar to that of the tested
and maintaining a high degree of saturation may be tailings (i.e., where calcite and ankerite are the main
complex in the field (Oldecop et al. 2017) and may have neutralizing minerals and pyrite the only sulfide). However,
negative impacts on TSF geotechnical stability. Ensuring the models could be adapted by using additional tests for
the regular deposition of a new tailings layers or starting other mineralogies to take in consideration the presence of
reclamation before critical time is reached is therefore minerals other than carbonates or the presence of
recommended. pyrrhotite which produces more acidity (Benzaazoua et al.
Extrapolations were performed to evaluate the critical 2001).
time with different AP and NP, based on tailings total
inorganic carbon and total sulfur contents. The critical time
increased linearly with the increase of the NP. At the same 6 CONCLUSION
time the critical time was also delayed by the increase of
the NNP. The prediction of the critical time using the Filtered tailings are an alternative to conventional slurry
proposed model therefore requires a precise determination tailings management which can contribute to increase
of the NP and AP in the field. However, critical time in a geotechnical stability in TSF. Their low degree of
large TSF will be fluctuating, due to their natural variability. saturation, however, expose them to the risk of AMD
Variable amount of sulfide and carbonate minerals generation. The objective of this study was therefore to
evaluate filtered tailings critical time, i.e., the time tailings
can remain exposed to atmospheric condition before they Aubertin, M., Bussière, B., and Zagury, G. 2011. La gestion
start producing acid (pH < 6). Results from laboratory des rejets miniers au Québec. Institut du nouveau
kinetic tests were simulated using MIN3P reactive transport monde , L’état du Québec: 225-232.
code and model extrapolations allowed to identify the key Azam, S., and Li, Q. 2010. Tailings dam failures: a review
parameters influencing the critical time : of the last one hundred years. Geotechnical news,
• Critical time is mainly governed by the neutralization 28(4): 50-54.
potential, and especially the carbonate contents. Benzaazoua, M., Bussière, B., and Dagenais, A.M. 2001.
Critical time increase linearly with the NP, and Comparison of kinetic tests for sulfide mine tailings. In
greater proportion of carbonates will contribute to Tailings and Mine Waste 2001. CRC Press. pp. 263-
delay the critical time. 272.
• Critical time is also influenced by the net Benzaazoua, M., Bussière, B., Dagenais, A., and
neutralization potential, especially when the Archambault, M. 2004. Kinetic tests comparison and
difference between the NP and the AP tend to the interpretation for prediction of the Joutel tailings acid
uncertainty zone and the ratio NP/AP is greater than generation potential. Environmental geology, 46(8):
0.5. 1086-1101.
• Critical time is also influenced by the degree of Blowes, D., Ptacek, C., Jambor, J., Weisener, C., Paktunc,
saturation. However, laboratory results and D., Gould, W., and Johnson, D. 2014. 11.5 - The
simulations both showed that critical time was geochemistry of acid mine drainage. . In Treatise on
significantly delayed only when Sr > 90%. geochemistry: Second edition. Edited by H.D. Holland
and K.K. T, Elsevier. pp. 131-190.
Results presented in this study show that critical time is Bouzahzah, H., Benzaazoua, M., Bussière, B., and Plante,
predictable and an adapted management in a filtered B. 2014. Revue de littérature détaillée sur les tests
tailings TSF will ensure a geochemical stability in the long statiques et les essais cinétiques comme outils de
term. Further studies need to evaluate the evolution of the prédiction du drainage minier acide. In Environnement,
critical time for other mineralogies and the influence of Ingénierie Développement. HAL. pp. 14-31.
wetting-drying cycles, water movement and frequency of Bouzahzah, H., Benzaazoua, M., Plante, B., and Bussiere,
fresh layer drop off. B. 2015. A quantitative approach for the estimation of
the “fizz rating” parameter in the acid-base accounting
tests: A new adaptations of the Sobek test. Journal of
7 REFERENCES Geochemical Exploration, 153: 53-65.
Bowker, L.N., and Chambers, D.M. 2017. In the dark
Amoah, N., Dressel, W., and Fourie, A.B. 2018. shadow of the supercycle tailings failure risk & public
Characterisation of unsaturated geotechnical liability reach all time highs. Environments, 4(4): 75.
properties of filtered magnetite tailings in a dry stack Bussière, B. 2007. Colloquium 2004: Hydrogeotechnical
facility. PASTE 2018: Proceedings of the 21st properties of hard rock tailings from metal mines and
International Seminar on Paste and Thickened Tailings. emerging geoenvironmental disposal approaches.
Australian Centre for Geomechanics. pp. 375-388. Canadian Geotechnical Journal, 44(9): 1019-1052.
Amos, R.T., Blowes, D.W., Bailey, B.L., Sego, D.C., Smith, Chambers, D.M., and Higman, B. 2011. Long term risks of
L., and Ritchie, A.I.M. 2015. Waste-rock hydrogeology tailings dam failure. Center for Science in Public
and geochemistry. Applied Geochemistry, 57: 140-156. Participation.
Anawar, H.M. 2015. Sustainable rehabilitation of mining Chopard, A., Benzaazoua, M., Plante, B., Bouzahzah, H.,
waste and acid mine drainage using geochemistry, and Marion, P. 2015. Kinetic tests to evaluate the
mine type, mineralogy, texture, ore extraction and relative oxidation rates of various sulfides and
climate knowledge. Journal of Environmental sulfosalts. ICARDS 10. Santiago, Chile.
Management, 158: 111-121. Davies, M. Filtered dry stacked tailings : the fundamentals.
ASTM-D2487. 2017. Standard Practice for Classification of Proceedings tailings and mine waste. Vancouver, BC,
Soils for Engineering Purposes (Unified Soil Canada.
Classification System). ASTM international. Dold, B. 2005. Basic Concepts of Environmental
ASTM-D5550. 2014. Standard Test Method for Specific Geochemistry of Sulfide Mine-Waste XXIV Curso
Gravity of Soil Solids by Gas Pycnometer. ASTM Latinoamericano de Metalogenia (UNESCO-SEG)
International. Lima, Perú. pp. 1-36.
ASTM-D6836. 2016. Standard test methods for Elghali, A., Benzaazoua, M., Bouzahzah, H., Bussière, B.,
determination of the soil water characteristic curve for and Villarraga-Gómez, H. 2018. Determination of the
desorption using a hanging column, pressure extractor, available acid-generating potential of waste rock, part I:
chilled mirror hygrometer, and/or centrifuge. ASTM Mineralogical approach. Applied Geochemistry, 99: 31-
International. 41.
ASTM-D7928. 2021. Standard Test Method for Particle- Elghali, A., Benzaazoua, M., Taha, Y., Amar, H., Ait-
Size Distribution (Gradation) of Fine-Grained Soils khouia, Y., Bouzahzah, H., and Hakkou, R. 2023.
Using the Sedimentation (Hydrometer) Analysis. ASTM Prediction of acid mine drainage: Where we are? Earth-
International. Science Reviews, 241
Gutierrez, G.J.R., and Oldecop, L.A. 2021. Test plot study Moreno, J.J., Kendall, S., and Ortiz, A. 2018. PASTE 2018:
of a tailings stack final cover prototype. MATEC Web Dewatering options for management of fine gold tailings
of Conferences. EDP Sciences. in Western Australian Goldfields. Proceedings of the
Hoffmann, M.R., Faust, B.C., Panda, F.A., Koo, H.H., and 21st International Seminar on Paste and Thickened
Tsuchiya, H.M. 1981. Kinetics of the removal of iron Tailings. Australian Centre for Geomechanics, Perth,
pyrite from coal by microbial catalysis. Applied Australia. pp. 413-424.
Environmental Microbiology, 42(2): 259-271. Morin, K.A., and Hutt, N.M. 2001. Environmental
Jurjovec, J., Blowes, D.W., Ptacek, C.J., and Mayer, K.U. geochemistry of minesite drainage: Practical theory and
2004. Multicomponent reactive transport modeling of case studies. Series. MDAG Publishing, Digital Edition.
acid neutralization reactions in mine tailings. Water Nordstrom, D.K., Blowes, D.W., and Ptacek, C.J. 2015.
resources research, 40(11) Hydrogeochemistry and microbiology of mine drainage:
Kalonji-Kabambi, A., Demers, I., and Bussière, B. 2020. an update. Applied Geochemistry, 57: 3-16.
Reactive transport modeling of the geochemical Oldecop, L.A., and Rodari, G.J. 2017. Unsaturated soil
behavior of highly reactive tailings in different mechanics in mining. In PanAm Unsaturated Soils
environmental conditions. Applied Geochemistry, 2017. American Society of Civil Engineers pp. 257-280.
122(104761) Oldecop, L.A., Rodari, G.J., and Muñoz, J.J. 2017.
Lawrence, R.W., and Wang, Y. 1997. Determination of Atmosphere interaction and capillary barrier in filtered
neutralization potential in the prediction of acid rock tailings. Geotechnical Geological Engineering, 35(4):
drainage. 4th International Conference on Acid Rock 1803-1817.
Drainage. Vancouver, Canada. pp. 449-464. Ouangrawa, M., Molson, J., Aubertin, M., Bussière, B., and
Lichtner, P.C. 1996. Continuum formulation of Zagury, G.J. 2009. Reactive transport modelling of
multicomponent-multiphase reactive transport. mine tailings columns with capillarity-induced high
Reviews in mineralogy & Geochemistry, 34: 1-82. water saturation for preventing sulfide oxidation.
Lindsay, M., Moncur, M.C., Bain, J., Jambor, J.L., Ptacek, Applied Geochemistry, 24(7): 1312-1323.
C.J., and Blowes, D.W. 2015. Geochemical and Pabst, T., Molson, J., Aubertin, M., and Bussière, B. 2017.
mineralogical aspects of sulfide mine tailings. Applied Reactive transport modelling of the hydro-geochemical
Geochemistry, 57: 157-177. behaviour of partially oxidized acid-generating mine
Lottermoser, B.G. 2010. Sulfidic mine wastes. In Mine tailings with a monolayer cover. Applied Geochemistry,
wastes : Characterisation, Treatment and 78: 219-233.
Environmental Impacts. Springer. pp. 43-117. Price, W.A. 2009. Prediction Manual for Drainage
Lupo, J., and Hall, J. 2010. Dry stack tailings–design Chemistry from Sulphidic Geologic Materials. Report
considerations. Proceedings Fourteenth International 1.20. 1. MEND, Natural Resources Canada.
Conference on Tailings and Mine Waste. CRC Press, Qiu, Y., and Sego, D.C. 2001. Laboratory properties of
Vail, Colorado, USA. pp. 327-334. mine tailings. Canadian Geotechnical Journal, 38(1):
Lyu, Z., Chai, J., Xu, Z., Qin, Y., and Cao, J. 2019. A 183-190.
comprehensive review on reasons for tailings dam Rico, M., Benito, G., Salgueiro, A.R., Díez-Herrero, A., and
failures based on case history. Advances in Civil Pereira, H.G. 2008. Reported tailings dam failures: a
Engineering, 2019(4159306) review of the European incidents in the worldwide
Mayer, K.U. 1999. A numerical model for multicomponent context. Journal of hazardous materials, 152(2): 846-
reactive transport in variably saturated porous media. 852.
Univertisy of Waterloo. Schoenbrunn, F. 2011. Dewatering to higher densities-an
Mayer, K.U., Amos, R., Molins, S., and Gerard, F. 2012. industry review. Proceedings of the 14th International
Reactive transport modeling in variably saturated Seminar on Paste and Thickened Tailings. Australian
media with MIN3P: Basic model formulation and model Centre for Geomechanics, Perth, Australia. pp. 19-23.
enhancements. In Groundwater Reactive transport Simms, P. 2017. 2013 Colloquium of the Canadian
Models. Edited by B. Science. Bentham Science Geotechnical Society: Geotechnical and
Publishers Sharjah, UAE. pp. 186-211. geoenvironmental behaviour of high-density tailings.
Mbonimpa, M., Aubertin, M., Chapuis, R.P., and Bussière, Canadian Geotechnical Journal, 54(4): 455-468.
B. 2002. Practical pedotransfer functions for estimating Sobek, A.A., Schuller, W.A., Freeman, J.R., and Smith,
the saturated hydraulic conductivity. Geotechnical and R.M. 1978. Field and laboratory methods applicable to
Geological Engineering, 20(3): 235-259. overburdens and minesoils. Series. United States
McPhail, G., Ugaz, R., and Garcia, F. 2019. Practical Environmental Protection Agency.
tailings slurry dewatering and tailings management Strachan, C., and Goodwin, S. 2015. The role of water
strategies for small and medium mines. PASTE 2019: management in tailings dam incidents. Tailings and
Proceedings of the 22nd International Conference on Mine Waste. Vancouver, BC, Canada.
Paste, Thickened and Filtered Tailings. Australian Thompson, F., de Oliveira, B.C., Cordeiro, M.C., Masi,
Centre for Geomechanics, Perth, Australia. pp. 235- B.P., Rangel, T.P., Paz, P., Freitas, T., Lopes, G., Silva,
243. B.S., and Cabral, A.S. 2020. Severe impacts of the
Molson, J., Aubertin, M., Bussière, B., and Benzaazoua, M. Brumadinho dam failure (Minas Gerais, Brazil) on the
2008. Geochemical transport modelling of drainage water quality of the Paraopeba River. Science of the
from experimental mine tailings cells covered by total environment, 705(135914)
capillary barriers. Applied Geochemistry, 23(1): 1-24.
Ulrich, B. 2019. Practical thoughts regarding filtered
tailings. PASTE 2019: Proceedings of the 22nd
International Conference on Paste, Thickened and
Filtered Tailings. Australian Centre for Geomechanics,
Perth, Australia. pp. 71-79.
Ulrich, B., and Coffin, J. 2013. Considerations for tailings
facility design and operation using filtered tailings.
PASTE 2013: Proceedings of the 16th International
Seminar on Paste and Thickened Tailings. Australian
Centre for Geomechanics, Perth, Australia. pp. 201-
210.
van Genuchten, M.T. 1980. A closed‐form equation for
predicting the hydraulic conductivity of unsaturated
soils. Soil science society of America journal, 44(5):
892-898.
Williams, D.J. 2021. Lessons from Tailings Dam Failures—
Where to Go from Here? Minerals, 11(8): 853.
Zorzal, R., Mansur Gomes, M., and Pinheiro, J. 2020.
Tailings Geotechnical Characterization from Cuiabá
Mine to Support a Dry Stacking Disposal Design in
Cuiabá Dam. PASTE 2020: 23rd International
Conference on Paste, Thickened and Filtered Tailings.
Gecamin Publications, Perth, Australia.
Assessing the Creep Properties of Filtered
Tailings in Continuous Permafrost Regions:
Methodological Developments and
Preliminary Results
Weber Anselmo dos Ramos Souza, Vincent Boulanger-Martel,
Bruno Bussiere, Mamert Mbonimpa & Mutaz Nujaim
Université du Québec en Abitibi-Témiscamingue - Research
Institute on Mines and Environment (RIME-UQAT), 445 University
Blvd, Rouyn-Noranda, QC, J9X 5E4, Canada

ABSTRACT

Filtered tailings are popular in the mining industry because they can reduce the risks associated with the physical stability
of filtered tailings storage facilities (FTSFs), especially for mine sites located in remote and Arctic regions. Understanding
the creep, long-term strength, and compressibility characteristics of frozen filtered tailings is essential in properly designing
FTSFs and reclamation covers in continuous permafrost regions. This article first aims to present the experimental
approach that was developed to assess the short-term creep properties of frozen filtered tailings using temperature-
controlled uniaxial constant-load tests. Preliminary results indicated that unsaturated specimens experienced more
significant displacement than saturated samples. The creep rate changed according to the stationary and non-stationary
creeps. This study also provides insights on the ability of existing creep constitutive models to describe and predict the
creep of frozen filtered tailings.

RÉSUMÉ

Les résidus filtrés sont populaires dans l'industrie minière car ils peuvent réduire les risques associés à la stabilité physique
des parcs à résidus miniers (PAR), en particulier pour les sites miniers situés dans des régions éloignées et arctiques. La
compréhension des caractéristiques de fluage, de résistance à long terme et de compressibilité des résidus filtrés gelés
est essentielle pour une conception appropriée de PAR situés dans des régions de pergélisol continu et de recouvrements
de restauration. Cet article vise d'abord à présenter l'approche expérimentale qui a été développée pour évaluer les
propriétés de fluage à court terme de résidus miniers filtrés gelés en utilisant des essais à uniaxiaux à charge constante
et température contrôlée. Les résultats préliminaires indiquent des déplacements plus importants pour les échantillons
non saturés que saturés. Le taux de fluage varie selon les régimes stationnaire et non stationnaire. Cette étude vise
également à évaluer la capacité des modèles constitutifs de fluage à décrire et à prédire le fluage des résidus filtrés gelés.

1 INTRODUCTION tailings dams (Lupo & Hall, 2010). FTSFs also provide the
advantage of facilitating progressive reclamation.
Mining plays a vital role in contributing to the Canadian However, this method has several challenges related to
economy, producing around 60 minerals in 200 mines and the hydro-geotechnical behavior of the filtered tailings, one
acting as one of the global leaders in the production of of which is compaction. Nujaim et al. (2022) pointed out
potash, cadmium, cobalt, diamonds, and gold (NRC, that material densification could be challenging because of
2022). Despite its contribution to the economy of around the freezing of tailings or the presence of excessive
CAN$48.2 billion in 2019 (NRC, 2022), the industry faces snow/water during compaction. Furthermore, ice lenses
multiple operational, geo-environmental, and social may be found in permafrost foundations, which can lead to
challenges. One of these challenges concerns the safe complex mechanical, thermal, and hydrogeological
storage of mine tailings. Filtered tailings, which have a high processes that can make the behavior more complex,
solids content (≈ 80–85% relative to the total mass), differ threatening long-term stability (Rykaart et al., 2018). For
considerably from conventional pulped tailings, which Arctic sites, these complexities could result in challenging
contain between 25–40% solids (Davies, 2011). Many conditions for mechanisms such as thaw consolidation,
mining companies have shifted to this tailings management thaw settlement, and creep, which depend on soil type,
method, which has increased the physical stability of structure, and density; amount of ice; temperature; and
filtered tailings storage facilities (FTSFs) compared with stress (Sayles and Haines, 1974).
conventional tailings, especially for remote and Arctic mine Understanding consolidation and creep and their
sites (Bussière, 2007; Davies, 2011). Filtered tailings impact on the thermal and hydrological behavior of FTSFs
contribute to reduce the consumption of fresh water and is critical for designing such structures in Arctic regions. It
the footprint of storage areas, and eliminate the need for also is now well established that creep depends on several
key factors: not only temperature and magnitude of the For uniaxial creep laboratory tests, there are two
applied stress but also soil type and structure, amount of approaches to load a sample: single-load and multi-load. A
ice, mineral type, strain rate, solute, density, and type and single-load test is the most common method for assessing
magnitude of loading (Sayles & Haines, 1974). French the creep of rocks, but this method requires several
(2017) pointed out that the frozen ground could deform homogeneous samples. The behavior of the material is
under gravity and stated that this deformation is mainly obtained via several tests performed on different samples
associated with the creep of pore ice and the migration of at selected loads. The multi-step loading approach is
unfrozen pore water. adapted to reduce the number of test specimens needed
The deformation of frozen soil can be divided into by utilizing the same sample for different stress levels and
instantaneous initial deformation and deformation over a initial strain levels. This approach is used widely nowadays
period of time (Equation 1). The first stage of deformation to assess the creep properties of geomaterials (Yang et al.,
(𝜀0 ) represents elastic and plastic deformation, and the 2018), especially through the use of standard creep tests
second (𝜀(𝑡) ) is associated with viscous deformation. like ASTM D5520 (ASTM, 2018) and GOST-12248 (GOST,
Viscous deformation can also be divided into stationary 2010). The major downfall of laboratory creep tests is that
and non-stationary deformations (Figure 1-a). they are time-consuming and costly. Therefore,
constitutive models are often used to predict the creep
𝜀𝑡 = 𝜀0 + 𝜀(𝑡) [1] properties of geomaterials (Yang et al., 2012) .
This study first aims to present the experimental
approach that was developed to assess the short-term
creep properties of frozen filtered tailings using a
where 𝜀𝑡 is the total strain, 𝜀0 is the plastic and elastic temperature-controlled uniaxial constant load testing
strain, and 𝜀(𝑡) is the viscous strain (creep). apparatus and a multi-step-loading approach. Another
Stationary creep is a type of creep in which the strain objective of this study was to investigate the ability of
rate approaches zero asymptotically, as shown in Figure 1- existing creep models to describe the creep behavior of the
b. This means that the material continues to deform over tested samples.
time, but the rate of creep slows down until it reaches zero.
Non-stationary creep is a type of creep in which the strain 2 BACKGROUND ON THE CREEP OF FROZEN
rate (rate of deformation) does not approach zero GEOMATERIALS
asymptotically but instead exhibits an S-like shape, as
shown in Figure 1-c. During primary creep, the creep rate Constitutive models are a mathematical representation of
decreases over time. In secondary creep, the creep rate the behavior of the material (Briaud, 2013). According to
remains constant. Finally, in tertiary creep, the creep rate Betten (2008), models can be analyzed in a microscopic or
increases until the failure of the material. macroscopic view. Microscopic models focus on describing
creep based on the laws of physics and established
equations (Ladanyi, 1972). On the other hand, the
phenomenological theory of creep (macroscopic models)
can be considered as a collection of laws found, by
experience, to adequately describe the observed
manifestations of creep in simple mathematical terms,
keeping the number of material parameters as small as
possible (Ladanyi, 1972). Microscopic models are typically
used to interpret material properties and macroscopic
models are rather used for the calculation of stress and
displacement of the same material (Scott, 1963). Several
creep models have been developed to simulate
experimental results of creep tests. These models include
visco-plastic (Foriero et al., 1998), viscoelastic (Perzyna,
1963, 1966), damage creep (Miao et al., 1995), hypoplastic
Figure 1. Types of creep and long-term strength curve (Xu et al., 2018), and endochronic models (Gopal, 1985).
function of temperature (T) and stress (σ) (a), stationary The following sections will introduce the Vialov (1969,
creep curve (b), and non-stationary creep curve (c) divided 1986b) model and the primary creep model, which have the
into primary (1), secondary (2), and tertiary (3) creeps. potential to accurately compute the creep of frozen filtered
tailings. The Vialov model is straightforward to fit and has
The two types of creep behavior differ in terms of their long- established reference values, making it a valuable tool for
term strength. The long-term strength is the stress at and this application.
below which the deformation rate decreases to zero over
time and no failure occurs within any practical period of 2.1 Primary creep model
load application (Vialov, 1986a). Vialov (1986a) observed
that the long-term strength of frozen soil is usually between One of the first models for soil is the primary creep model
1/5 and 1/15 of the uniaxial compression strength (UCS). (Vialov, 1969). This model combines the Voigt-Kelvin unit
Figure 1-a illustrates that long-term strength could be in series with a Maxwell unit and a blocking device, making
shown as a line, but it could also be represented as a single it an element assembly model. In other words, this model
value for a fixed temperature. has a physical background based on the Voigt and Newton
models for viscous materials. This creep model was initially 2.2 Modified Vialov model
developed to describe experimental data obtained from
single-load creep tests, as observed in Vialov (1969). Vialov (1986b) stated that the main characteristic of frozen
Equation 2 presents a primary creep model expression for soil deformation properties is the relationship between
single-load tests. The Yang et al. (2018) methodology to stress and total deformation, which is determined based on
convert single-load models to multi-load models was a time value. They also expressed that stress is directly
applied for the primary creep model. The Vialov model is related to time and deformation (Equation 4). Vialov
presented in Equation 3. This equation describes the creep (1986b) adapted the formulation of Equation 4 for step-load
response to an incremental increase in load over time. The tests, which reduced the residual between measured and
model is useful for step-load tests because it can modelled data. Equations 5 and 6 present these
significantly reduce the number of tests needed to obtain modifications in detail. As opposed to the primary creep
reliable results. Table 1 provides typical model parameters model, an estimation of the initial displacement is not
for selected materials. necessary for this model (Yang et al., 2018). In Equations
4 to 6, A and 𝑚 are empirical parameters.
𝜎𝑡 𝜆
1/𝑚 [2]
𝜀𝑡 = [ ] + 𝜀0
𝜔(𝜃 + 1)𝑘 σ= A (t). 𝜀𝑚 [4]
1/m
σb (t 1 )λ
εt = [ ] +ε0 𝜎𝑏 [5]
ω(θ+1)k 𝜀𝑡 = +𝜀0
1/𝑚 𝐴.𝑡 𝑚
𝜎𝑏+1 (𝑡1 −𝑡)𝜆
+[ ] + 𝜀1 [3]
𝜔(𝜃+1)𝑘 𝜎𝑏+1 [6]
𝜀𝑡 = +𝜀0
𝐴.𝑡 𝑚
where:
σb is the applied constant stress (kg/cm²), 3 MATERIALS AND METHODS
σb+1 is the applied constant stress in the following phase,
This study used filtered tailings from the Raglan mine after
t is the time (h) being prepared according to standardized procedures.
𝑡1 is the time when a load was increased (h) Step-load tests were conducted according to ASTM D5520
(ASTM, 2018) and GOST-12248 (GOST, 2010), and the
θ is the temperature below the freezing point (water) (°C) test results were analyzed to determine the long-term
strength. The ability of the Vialov (1969, 1986) models to fit
ε0 is the instantaneous strain (m/m) experimental results was investigated.
ε1 is an instantaneous strain in the following phase (m/m)
3.1 Test material
λ, m, ω and k are constants that are characteristic of the
material. Remolded filtered tailings from the Raglan mine were used
The main advantage of this model is that it provides model in this study. The Raglan mine is a nickel-copper operation
parameters for several types of materials (Table 1). located in the Nunavik region, Northern Quebec (61°N;
However, the practical use of the primary model (Equation 73°W) that started production in 1997. The ground thermal
3) can be challenging because it requires determination of regime at the mine site is characterized by continuous
a new set of empirical parameters for each geomaterial permafrost. The basic geotechnical properties and field
(Andersland & Ladanyi, 2004). This model also often properties of the Raglan filtered tailings were characterized
underestimates strain during the tertiary creep phase (Zhu by several authors (i.e., Coulombe, 2012; SNC, 2013;
& Carbee, 1987), which is unsuitable when important Nujaim et al., 2022). The basic geotechnical properties of
variations in creep stress occur (Xu et al., 2016). the tailings are presented in Table 2.

Table 1. Primary creep model parameters for selected Table 2. Basic geotechnical properties of the Raglan mine
materials (Sayles, 1968; Sayles & Haines, 1974). tailings (from Coulombe (2012) and Nujaim et al. (2022)).

Parameter Value
Specific gravity (-) 2.93
Optimum water content 17.3%
Maximum dry density 1,832 kg/m3
Field dry density (kg/m3) 1,650–1,700
Degree of compaction in situ 90–93%
Average water content in situ (-) 20.5% ± 3.8%
Coefficient of uniformity CU 16.7
Porosity in situ 42%
3.2 Preparation of specimens were placed at the surface of each sample and in the
sample housing to monitor temperature fluctuations: the
Samples were mixed at a water content of approximately first was fixed in the sample, and the second was fixed on
20% by weight. This represents the average in situ water the wooden box.
content at Raglan Mine. All samples were compacted to the Creep tests are commonly conducted by applying a single
desired density. Saturated samples were subjected to a constant load to a specimen, as shown in Figure 3-a. To
vacuum to obtain a high degree of saturation. reduce the number of samples required, a different
After this phase, the specimens were placed in a approach was used to obtain results regarding the creep
freezing cabinet with a membrane to prevent surface behavior of specific materials. One of these approaches
evaporation. They were then quickly frozen at –27 ºC from was a step-load test, in which the applied load was
top to bottom to avoid ice lens formation. After freezing, the incrementally increased several times for the same
mold was cut to facilitate the ejection the sample. The specimen, as shown in Figure 3-b. Creep tests were
sample was trimmed to ensure its flatness and parallel performed using ten loading stages, ranging from 300 kPa
alignment with the frozen plate. Each specimen had the to 1.8 MPa (Figure 4). Each loading step was applied for
nominal dimensions of 2 in in diameter and 4 in in length. 24 h. As required by ASTM (2018) and GOST (2010), each
Before the test, each sample was placed in the creep load increment was increased for no more than 30
apparatus for 24 h. Samples were left without any applied seconds. This study considered a creep strength failure
load to allow the sample to reach thermal equilibrium with criterion of 20% strain (GOST, 2010). Any test reaching the
the desired test temperature. Each test was assigned a failure criterion before the end of the loading sequence
number: unsaturated samples from 1–3 and saturated would be terminated. The termination of tests was also
samples from 4–5. Figure 2 shows the equipment used for considered if a brittle failure occurred, or if the 10th loading
step-load tests in frozen filtered tailings. stage was achieved. Surface area corrections (to adjust
calculated stress) were also conducted for samples
exhibiting changes in diameter exceeding 5% of the initial
value.

a b

Figure 3. Load during single-load test (a) and step-

loaded test (b).

Figure 2. Apparatus used for creep tests (Nujaim et al., Figure 4. Load used for creep test.
2022).
3.4 Calculation of long-term strength
3.3 Test method
Specimens were tested using the temperature-controlled According to Vialov (1969), the creep rate stabilizes when
uniaxial creep apparatus described in Nujaim et al. (2022). the creep rate for long-term strength is less than 0.0001/h.
All tests were performed in a controlled atmosphere cold This occurs after different time intervals depending on the
chamber in which the set-up and samples were stabilized type of soil: 6 h for sand, 12 h for sandy loam, and 24 h for
at –6 ºC using an insulated box and a circulating bath loam and clay. This indicates that the creep has reached a
(Nujaim et al. 2022). The samples’ axial displacements stationary phase, and a long-term strength value can be
were measured using a linear variable displacement determined.
transducer (LVDT) within a 50 mm range. Thermocouples
 GOST (2010) and Aksenov et al. (2016) describe a pore ice strength and weaker mechanical interaction
procedure to determine the long-term strength using short- between the ice and grains (Ting, 1981).
term tests (Equation 7). The time factor Kt is determined by
considering the duration of each phase and the stress
when the tertiary creep begins. The time factor for the 24 h
phase is 0.6, which is used to estimate the long-term
strength using Equation 7.

σlt =Kt σm
[7]

where:
σlt is the long-term stress,
Kt is the time factor; and
σm is the stress when minimum strain rate occurs.
Figure 3. Creep curves for all tested samples.
3.5 Model fitting
Figure 6 presents the creep rate from the last 6 h of
To evaluate the fitting process, Equation 8 was used, each loading step as a function of the applied stress in the
providing the possibility to observe the effect of parameter sample. Unsaturated samples presented a higher
changes. Lower R values suggest a better fit between variability in creep rate than saturated samples. Results
measured and computed data. For the primary creep also show that creep rates above 0.001 h-1 (ten times more
model, the initial parameters used for fitting were based on than the stabilization criteria of Vialov (1969)) resulted in
Table 1. Vialov’s (1986b) model empirical parameters were sample failure. These results suggest that all specimens,
estimated based on the Solver solution from Excel, either saturated or unsaturated, had at least 600 kPa of
minimizing the residual value. long-term strength at –6 °C.

R = Σ(𝑑1 − 𝑀)2
[8]

where R is the sum of difference square between values of

the model (𝑑1 ) for the time t and a point in a data (𝑀) for
the time t.

4 RESULTS AND DISCUSSION

The uniaxial stepwise load test was performed on five

frozen filtered tailings samples; the axial load and axial Figure 4. Creep rate of the last 6 h of each phase versus
displacement were recorded. Figure 5 shows strain curves applied stress in unsaturated and saturated samples.
for unsaturated (samples 1–3) and saturated (samples 4–
5) filtered tailings. The results show that at any given load 4.1 Model fitting
stage, the strain observed for unsaturated tailings was
higher than for saturated tailings. Both saturated samples 4.1.1 Primary creep model
(4 & 5) showed similar strain until the 5th loading stage (120
h). Such behavior was observed for unsaturated samples Table 3 shows the parameters that were used to fit the
(1 and 3) but only up to the 4th loading phase. Unsaturated primary creep model to the experimental data. Figure 7
sample failure occurred within a range of applied stress compares the results of the primary creep model to the
between 1,170 kPa (6th) and 1,480 kPa (8th). However, no experimental data for some selected samples (samples 3
saturated samples reached the 20% strain failure criterion, (7-a), 4 (7-b), and 5 (7-c)). First, the model requires four
indicating that 1.8 MPa was insufficient to induce failure. empirical variables, alongside an initial deformation
The short-term creep results suggest that the frozen parameter that increments with each subsequent load
saturated samples supported at least 22% more weight augmentation. This implies that it is challenging to use
than the unsaturated samples when compared to the methods like Solver from Excel to reach the desired results.
maximum stress in those tests. Based on these results, Some parameters are related, so increasing one variable
Equation 7 was used to estimate the long-term strength, requires a decrease or increase in the others. Even with
which varied between 750–960 kPa for the unsaturated the procedure from Yang et al. (2018) for transforming
samples and reached more than 1,080 kPa for the single-load models into multi-load models, this model did
saturated samples. This suggests that the long-term not fit, especially for test 4, the creep rate has multiple
strength of the frozen saturated tailings was greater than changes due to the introduction of additional load phases.
that of the unsaturated tailings. (Figure 6-b). However, Fig. 7-c shows the excellent fit for
The difference in behavior between the saturated and the saturated models when no changes in creep rate
unsaturated samples can mostly be attributed to the fact occurred in the sample. Table 3 lists the values of ω for the
that geomaterials containing less ice usually have weaker unsaturated and saturated samples. The values of ω can
be linked to the strain behavior during the test: a lower 4.1.2 Modifed Vialov creep model
value of ω indicates that the model experienced more
strain, as can be seen by comparing unsaturated and Generally, the modified Vialov model (1986b) showed a
saturated samples. good fit with experimental data (Figure 8). With only two
parameters to change, this model is easier to fit than the
primary creep model and presents good results. Because
this model cannot simulate the tertiary creep (Zhu &
Carbee, 1987), the last load step (the rupture) for curves of
unsaturated samples were not used for fitting using
residuals. Table 4 summarizes the values for A and m in
this model. The Vialov model provided a better fit for
a sample 4 than the primary creep model; however, sample
5 exhibited a better fit with the primary creep model than
the modified Vialov model at early times in the experiment.

Figure 5. Primary creep model and experimental data for

some selected experimental samples: 3 (a), 4 (b), and 5
(c). Figure 6. Modified Vialov model (1986b) and experimental
data for some selected samples: 3 (a), 4 (b), and 5 (c).
Table 3. Variables used in the primary creep model.
Table 4. Variables used in the modified Vialov model
Unsaturated Saturated (1986b).

𝑁𝑜 1 2 3 4 5
Unsaturated Saturated
m 0.64 0.64 0.64 0.64 0.64
𝜆 0.3 0.3 0.3 0.3 0.3 𝑁𝑜 1 2 3 4 5
ω 33 30 25 50 50 m 0.529 0.539 0.491 0.691 0.742
k 1 1 1 1 1 A 1.91 𝑥 10 8
1.02 𝑥 10 8
3.71 𝑥 10 8
4.43 𝑥 10 7
4.27 𝑥 107
A comparison of the residuals between the models and 7 REFERENCE
experimental data suggests that the modified Vialov model
computes creep better than the primary creep model. Aksenov, V., Iospa, A., Krivov, D., Ozeritskii, K., &
Although samples 3 and 5 yielded similar residuals, a huge Doroshin, V. 2016. Comparison of test results for
discrepancy was observed between the primary creep uniaxial compression of frozen soil in response
model and the modified Vialov model. This may have to constant load and to incremental loading. Soil
resulted from difficulty fitting the parameters: the primary Mechanics and Foundation Engineering, 53(2),
creep model has four parameters, making it difficult to find 91-98.
a unique solution. Andersland, O. B., & Ladanyi, B. 2004. Frozen ground
engineering (2nd ed. ed.). Wiley.
Table 5. Residuals of the primary creep model and
modified Vialov model. American Society of Civil Engineers.ASTM. 2018. ASTM
D5520, Standard Test Method for Laboratory
Determination of Creep Properties of Frozen Soil
Test Primary creep Modified Vialov
Samples by Uniaxial Compression.
1 7.883 0.177
Betten, J. 2008. Creep mechanics (3rd ed. ed.). Springer.
2 1.008 0.670 https://doi.org/10.1007/978-3-540-85051-9.
3 0.090 0.075 Briaud, J.-L. 2013. Geotechnical engineering: unsaturated
4 2.501 1.478 and saturated soils. John Wiley & Sons.
Bussière, B. 2007. Colloquium 2004: Hydrogeotechnical
5 0.069 0.082 properties of hard rock tailings from metal mines
and emerging geoenvironmental disposal
5 CONCLUSION approaches. Canadian Geotechnical Journal,
44(9), 1019-1052.
This study introduced creep behavior in frozen filtered Davies. 2011. Filtered dry stacked tailings–the
tailings, estimated the long-term strength of these fundamentals. Proceedings tailings and mine
materials, and discussed some models for predicting their waste (pp. 1-9).
behavior. Foriero, A., Ladanyi, B., Dallimore, S. R., Egginton, P. A.,
The results of this paper demonstrated that unsaturated & Nixon, F. M. 1998. Modelling of deep seated
samples exhibited failure for stress between 1,170 kPa and hill slope creep in permafrost. Canadian
1,480 kPa. On the other hand, saturated samples did not Geotechnical Journal, 35(4), 560-578.
show a strain of more than 20% for 1.8 MPa. A procedure French, H. M. 2017. The periglacial environment. John
to determine the long-term strength values based on the Wiley & Sons.
short-term tests was presented. The tested stresses were Gopal, R. K. 1985. Endochronic constitutive modeling of
not high enough to determine the long-term strength of the marine fiber reinforced concrete and frozen soil].
saturated samples. Two different models were tested to fit WorldCat.org.
the experimental creep curves. The number of fitting http://gateway.proquest.com/openurl?url_ver=Z3
parameters required for each model influenced the ability 9.88-
to achieve a good fit to the experimental data. The modified 2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertatio
Vialov model appeared to better represent creep than the n&res_dat=xri:pqm&rft_dat=xri:pqdiss:8606713
primary creep model. http://ufdc.ufl.edu/AA00047361
This work is part of an ongoing investigation on the creep
GOST. 2010. GOST-12248: Laboratory Methods for
of frozen filtered tailings. Additional experimental data are
Determining Strength and Deformation
required to understand the complex mechanisms
Characteristics.
governing the creep of frozen filtered tailings. Tests with
Ladanyi, B. 1972. An Engineering Theory of Creep of
different temperatures, degrees of saturation, and stress
Frozen Soils. Canadian Geotechnical Journal,
states will contribute to further knowledge. The suitability of
9(1), 63-80. https://doi.org/10.1139/t72-005
constitutive models to describe and predict the creep of
Lupo, J. F., & Hall, J. E. 2010. Dry stack tailings – design
filtered tailings should also be assessed.
considerations.
Miao, T., Wei, X., & Zhang, C. 1995. Creep of frozen soil
6 ACKNOWLEDGMENTS
by damage mechanics. Science in China
(Scienctia Sinica) Series B, 8(38), 996-1002.
This study was funded by the FRQNT as a part of the -
NRC. 2022. Minerals and the economy. Government of
Développement durable du secteur minier program and by
Canada. https://www.nrcan.gc.ca/our-natural-
the Research Institute on Mines and the Environment
resources/minerals-mining/minerals-metals-
(RIME UQAT-Polytechnique; http://www.irme.ca). The
facts/minerals-and-the-economy/20529
personnel from the Raglan mine’s environment department
Nujaim, M., Boulanger-Martel, V., Bussière, B., Anselmo,
is thanked for their collaboration.
W., & Pabst, T. 2022. Creep properties of filtered
tailings: design of a temperature-controlled
uniaxial constant load testing apparatus and
preliminary results.
Perzyna, P. 1963. The constitutive equations for rate
sensitive plastic materials. Quarterly of applied
mathematics, 20(4), 321-332.
Perzyna, P. 1966. Fundamental problems in
viscoplasticity. Advances in applied mechanics,
9, 243-377.
Rykaart, M., Barrero, A., & Lizcano, A. 2018. Frozen Core
Tailings Dam: Part 2, Long-Term Creep
Deformation. Proceedings of Tailings and Mine
Waste.
Sayles, F. 1968. Creep of Frozen Sands.
Sayles, F., & Haines, D. 1974. Creep of frozen silt and
clay.
Scott, R. F. 1963. Principles of soil mechanics. Addison-
Wesley Pub. Co.
Ting, J. 1981. The creep of frozen sands: Qualitative and
quantitative models, part 2. Massachusetts Inst.
of Tech. Report.
Vialov, S. S. 1969. Methods of determining creep, long-
term strength and compressibility characteristics
of frozen soils. National Research Council of
Canada.
Vialov, S. S. 1986a. Rheological fundamentals of soil
mechanics. Elsevier ;Distributors for the United
States and Canada, Elsevier Science Pub. Co.

Vialov, S. S. 1986b. Rheological fundamentals of soil

mechanics. Elsevier ;Distributors for the United
States and Canada, Elsevier Science Pub. Co.

Xu, G., Wu, W., & Qi, J. 2016. Modeling the viscous
behavior of frozen soil with hypoplasticity.
International Journal for Numerical and
Analytical Methods in Geomechanics, 40(15),
2061-2075. https://doi.org/10.1002/nag.2516
Yang, Q., Zhang, J.-M., Zheng, H., & Yao, Y. 2012.
Constitutive modeling of geomaterials: advances
and new applications. Springer Science &
Business Media.
Yang, W., Gamage, R. P., Huang, C., Luo, G., Guo, J., &
Wang, S. 2018. Loading history effect on creep
deformation of rock. Energies, 11(6), 1462.
Zhu, Y., & Carbee, D. L. 1987. Creep and strength
behavior of frozen silt in uniaxial compression.
Establishment of the mechanical and
hydraulic apertures of concrete fractures
using 3D scans and hydraulic tests
Khalil El Mekari1, François Duhaime1
1École de technologie supérieure (ÉTS), Montréal, Québec, Canada

ABSTRACT

Water infiltration through concrete fractures in tunnels causes deteriorations such as rebar corrosion. Chemical grout can
be injected to seal concrete fractures. The fracture aperture needs to be considered when planning the injection because
it affects the hydraulic characteristics of fracture such as the hydraulic conductivity. Two types of apertures can be defined:
mechanical and hydraulic apertures. The mechanical aperture is the average distance between the two bounding surfaces
of the fracture. The hydraulic aperture governs the liquid propagation and the head loss during the injection. This article
presents four methods to establish both apertures. For the mechanical aperture, the physical model of a concrete fracture
was built and scanned using the Polyworks software. The second method consisted of X-ray computed tomography (CT
scans) on core samples taken from tunnels of the Montréal Métro. For the hydraulic aperture, an in-situ test was developed
based on the injection of water-glycerol mixtures. A variable-head permeability test was also developed to establish the
hydraulic aperture.

RÉSUMÉ

Les infiltrations d’eau par les fissures du béton en tunnel provoquent des détériorations telles que la corrosion des
armatures. L’injection de résine est une technique employée pour sceller les fissures de béton. L’ouverture de la fissure
est un élément à considérer, car elle affecte les caractéristiques hydrauliques de la fissure telles que la conductivité
hydraulique. Deux types d’ouverture peuvent être définies: mécanique et hydraulique. L’ouverture mécanique est la
distance moyenne entre les deux surfaces qui délimitent la fissure. L’ouverture hydraulique régie la propagation du liquide
et les pertes de charge durant l’injection. Cette article présente quatre méthodes pour établir ces deux ouvertures.
L’ouverture mécanique a pu être établie à partir de modèle physique de fissure de béton et de scans avec le logiciel
Polyworks. La deuxième méthode consiste à effectuer des microtomographies aux rayons X (CT scans) sur des
échantillons provenant de tunnels du Métro de Montréal. Concernant l’ouverture hydraulique, un essai in-situ a été
développé à l’aide d’injection de mélanges eau-glycérol. Un essai à charge variable a également été développé pour établir
l’ouverture hydraulique.

1 INTRODUCTION establish the injection port drilling parameters and to

simulate the flow with numerical models. In rock fractures,
Water infiltrations through concrete fractures in tunnels can many authors showed that the aperture follows a normal or
be controlled or stopped to prevent important damages lognormal distribution (e.g. Berre et al., 2019; Huang et al,
such as rebar corrosion by collecting the water with gutters 2019; Li et al., 2019; Ye et al., 2015; Zhao et al., 2014). For
or by performing a chemical grout injection to seal the example, the numerical model results presented by Huang
fractures. Chemical grout injection is a common method et al. (2019) showed that the mean and standard deviation
that is performed by pumping a liquid grout into the of a normally distributed aperture influenced the flow
fracture. Thermosetting polymers such as polyurethane canalisation during liquid propagation in the fracture.
are often used (Panasyuk et al., 2013; von Fay, 2015). The mechanical aperture of fractured rock samples can
The fracture aperture is an important parameter to be established from X-ray computed tomography (CT
consider when an injection is performed in fractured porous scans) (Caulk et al., 2016; Crandall et al., 2017; Ramandi
media such as rock or concrete. It influences the fracture et al., 2017; Wanniarachchi et al., 2018). CT scans can be
hydraulic conductivity, the liquid propagation, and the used to reconstruct a 3D model of the fracture and can be
pressure distribution in time (Gao et al., 2023; Hao et al., used to establish the flow characterisation such as the
2018; Zhao et al., 2014). Two types of aperture definitions pressure distribution. For example, Wanniarachchi et al.
can be found in the literature: mechanical and hydraulic (2018) built a numerical model to simulate the flow using a
apertures (Berre et al., 2019; Gao et al., 2023; Huang et rock fracture reconstructed from CT scan results Such
al., 2021; Huo et al., 2014). models have not been presented in the literature for
The mechanical aperture is the average distance concrete fractures to our best knowledge.
between the two bounding surfaces of a fracture. It is The hydraulic aperture governs the liquid propagation
obtained after establishing the aperture distribution in and the head loss during the injection. In a fractured porous
space. It is used to estimate the injection volume, to media, it can be obtained from the cubic law and Darcy’s
law for laminar flow in porous media after performing a Scanning
hydraulic in-situ test (Cao et al, 2016; Maldaner, 2018;
Quinn et al., 2011). For example, Cao et al. (2016) used a device
variable-head in-situ test to establish the hydraulic aperture
and hydraulic conductivity of a rock fractured network.
Both types of apertures can be significantly different. Physical
Sun et al. (2020) did a review of the mathematical model
equations available in the literature to establish the
hydraulic aperture from the mechanical aperture and other
parameters. Some authors such as Olsson and Barton
(2001) and Cao et al. (2019) determined the relation
between both apertures from hydromechanical shear tests.
In rock fractures, it is possible to establish that the
mechanical aperture is higher than the hydraulic aperture
(Cao et al., 2016; Cao et al., 2019; Olsson and Barton, Figure 1. 3D scan of the entire physical model
2001; Sun et al., 2020; Zhao et al., 2014). To our best
knowledge, there is no such relation for concrete fractures. 2.1.2 CT scans on core sample
This article presents four methods to establish both
type of apertures. For the mechanical aperture, a 3D scan Two core samples were taken from two different tunnels of
of the physical model of a concrete fracture was performed the Montréal Métro. The first sample was taken from a
and analyzed with the Polyworks software. CT scans were concrete slab that was fractured on approximatively 120 m
done on two core samples from two different concrete and had a thickness of approximatively 40 cm. The sample
fractures in a tunnel to establish their aperture distribution length and diameter were approximately 20 cm and 5 cm,
in space. Two hydraulic tests were developed to establish respectively. The second core sample was taken from a
the hydraulic aperture. An in-situ test in a tunnel construction joint in a mechanical ventilation station. The
construction joint was performed with water-glycerol sample length and diameter were approximately 15 cm and
mixtures. Pressure reading ports allowed the pressure 7.5 cm, respectively.
distribution in time to be recorded and the hydraulic The CT scans were performed by the Shape Memory
aperture to be determined. The second test consisted of a Alloys and Intelligent Systems Laboratory of École de
variable-head permeability test on one of the core samples technologie supérieure. A Nikon XT-H-225 microfocus X-
used for CT scans. Both methods established the hydraulic ray source was used. The samples were scanned using a
aperture from the cubic law and Darcy’s law. The objective reflection scan source. The CT scan parameters were as
of this article is to propose multiple methods to establish follows:
both types of apertures. More information on fracture
aperture will allow for a more rational definition of injection
• 214 kV scan voltage.
parameters (e.g. pressure, volume, spacing of injection
• Effective pixel size of 51.9 µm.
ports).
• Copper filter of 2.5 mm to reduce the effect of beam
hardening.
2 METHODOLOGY
• 200 μA beam current.
2.1 Mechanical aperture • Gain of 24.02 dB.
• exposure time of 1.4 seconds.
2.1.1 Physical model 3D scan • 2634 projections.

The physical model is composed of two concrete slabs cast These specifications allowed reconstruct a 3D fracture
on top of each other. The fracture plan is a square with that has a length of approximatively 60 mm and a
sides of 61 cm. The model has a thickness of 12.7 cm. All- resolution of 60 µm. The Pro 3D and Dragonfly software
purpose concrete mix (Bomix brand) was used. A plastic packages were used for the segmentation and
film was inserted between the two slabs to be able to reconstruction. The mesh thickness evaluation module in
separate them after the concrete curing. Dragonfly was used to obtain the mechanical aperture (𝑏𝑚 )
The 3D scan was performed in the Metrology and its distribution in space. A total of 15,730,798
Laboratory of École de technologie supérieure. It consisted mechanical aperture values were calculated. Figure 2
in three steps. The first step was to scan the model with the shows one of the samples during CT scan preparation.
two concrete slabs on top of each other. The second step
was to separate the slabs and scan them separately. The
last step was to compare the scans of both slabs joined
together with the scan of the entire model using Polyworks.
Figure 1 shows a photograph taken during the scan of the
entire model. A total of 347,174 points were used for the
scan.
 Table 1. Injection and pressure reading ports parameters
Reflection
target
Parameters P3 P4 P7 P11
X-ray
source Distance from injection port 1632 1238 900 0
(mm)
Drilling distance (mm) 525 472 525 530
Drilling angle (°) 39 53 39 37

Core sample
Table 2. Properties of the injected mixtures

Figure 2. Core sample preparation for CT scans Mixtures Water Glycerol Dynamic viscosity Density
(%) (%) (Pa s) (kg/m3)
2.2 Hydraulic aperture A 18 82 7.510-2 1220
B 10 90 2.210-1 1225
2.2.1 In-situ test
C 5 95 5.210-1 1250
An in-situ test was developed in the same construction
joint used for CT scan sampling. It was performed before Equations 1 and 2 present the cubic law and Darcy’s
taking the sample. Surface capping was installed to seal law to obtain the hydraulic conductivity (K) and the flow rate
the joint. Acid and water injections were done to clean the (Q). K is influenced by the hydraulic aperture (𝑏ℎ ), the
joint before the in-situ test. Multiple injection and pressure gravitational acceleration (g), the liquid dynamic viscosity
reading ports were installed in the construction joint but (µ) and the liquid density (ρ). Q is influenced by the
only four were used. The P11 port was the injection port hydraulic gradient (i) and the section area (A) that is equal
and the P3, P4 and P7 ports were the pressure reading to the joint width (0.242 m) multiplied by bh. i is equal to the
ports. Figure 3 shows the construction joint before the in- head loss (dh) divided by the injection distance (L). ρ and
situ test. Table 1 presents the parameters of the injection 𝜇 were equal to 998 kg/m3 and 1.01 mPa s.
and pressure reading ports.
The injected liquid were mixtures of water and glycerol. 𝜌𝑔𝑏ℎ 2
Using Cheng’s (2008) equations, it was possible to 𝐾= [1]
produce three types of mixtures with different dynamic 12µ
viscosity and density. The injections were performed with
a P300 hydracell pump that used a constant injection flow 𝑄=𝐾𝑖𝐴 [2]
rate. The injection flow rate was between 2.61 and 3.91
mL/s. The injection duration was between 225 and 560 s. The next step was to combine Eq. 1 and 2 to have the
Both injection parameters depended on the injected liquid. fully developed equation of the flow shown in Eq. 3. It is
The injection pressure varied during the injection. Table 2 possible to isolate 𝑏ℎ in Eq. 3 as presented in Eq. 4.
presents the injected mixtures and their properties. The
water and glycerol proportions are mass proportions.
𝜌𝑔𝑏ℎ 2 𝑑ℎ
𝑄= 0.242𝑏ℎ [3]
12µ 𝐿
Pressure Injection
reading ports 3 49.39𝐿𝑄µ

ports 𝑏ℎ = √ [4]
𝑔𝜌𝑑ℎ

2.2.2. Variable-head permeability test

Capped A variable-head permeability test was developed with

constr. the first core sample used for CT scans. An initial water
joint level was set at 60 cm and the reduction of the water level
(𝐻) was monitored through time (𝑡) with transparent plastic
tubing. A constant head of 6.5 cm was applied at the
Figure 3. Capped construction joint before the in-situ test sample outlet. Surface capping was applied on the fracture
sample to get a unidirectional flow through the fracture
length. Figure 4 shows the test setup.
 PVC pipe
that contains
the water

Core
sample
Constant
head

Figure 4. Variable-head permeability test setup

Equation 5 presents another definition of the flow rate

from the cubic law and Darcy’s law. Equation 6 gives the
Figure 5. 3D scan of the physical model
flow rate that can calculated from the rate of change of the
water level in a pipe (∆𝐻⁄∆𝑡 ) during a variable-head test.
The mechanical aperture histogram for the concrete
The section of the pipe was 𝑆𝑖𝑛𝑗 = 19.6 cm² in this case. slab and the construction joint samples are presented in
Equation 7 presents the final equations to use when Figures 6a and 6b. Only the values between 0 and 2.5 mm
performing a permeability test. are shown for Figure 6a and 0 and 1 mm for Figure 6b.
Both fracture distributions in space were lognormal. Figure
𝜌𝑔𝑏ℎ 3 𝑤𝐻 6b displays additional information such as the median
𝑄= [5] value. For Figure 6a, the CT scans results established 𝑏𝑚
12𝜇𝐿
at 0.62 mm and a standard deviation of 0.35 mm. For
Figure 6b, 𝑏𝑚 was established at 0.59 mm with a standard
∆𝐻
𝑄 = −𝑆𝑖𝑛𝑗 [6] deviation of 0.25 mm. Figures 7a and 7b presents 3D
∆𝑡 reconstructed fracture of both samples. For the concrete
slab sample, the minimum and maximum apertures
𝜌𝑔𝑏ℎ 3 𝑤𝐻 ∆𝐻 detected by the CT scans were 0.16 mm and 3.72 mm. For
= −𝑆𝑖𝑛𝑗 [7] the construction joint sample, the minimum and maximum
12𝜇𝐿 ∆𝑡 apertures detected by the CT scans were 0.21 mm and
2.10 mm.
For this article, the Hvorslev (1951) method was used. The The methods presented to establish 𝑏𝑚 can be used for
hydraulic aperture (𝑏ℎ ) can be calculated from the slope different purposes. The 3D scan can cover a larger fracture
value (𝑆𝑉) obtained from the Hvorslev graph: surface and produce an entire map of aperture if needed.
However, the level of precision is lower due to the number
of points per surface scanned. The 3D scan produced
3 12𝑆𝑉𝜇𝐿𝑆𝑖𝑛𝑗
approximatively 93 apertures/cm2 and the CT scans
𝑏ℎ = √ [8] produced more than 500,000 apertures/cm2. Due to the
𝜌𝑔𝑤 small surface that can be scanned, the CT scans are better
suited for the study of the aperture at a precise location of
the fracture. Additionally, the application of the 3D scan
3 RESULTS AND DISCUSSION method is slightly different in tunnels for an injection. A
concrete fracture in tunnel is not accessible and cannot be
3.1 Mechanical aperture separated as it was for the physical model. It is only
applicable, for example, on samples that can be separated
Figure 5 shows the result of the 3D scan. The mechanical to perform the 3D scan. For both methods, multiple core
aperture varied from 0 to 1.5 mm. The aperture is higher samples are needed to establish the mechanical aperture
on the model extremities compare to the middle The of a concrete fracture with a high fracture surface.
aperture in the middle of the model was 0.50 mm according
to the map presented in Figure 5. The aperture on the
physical model extremities varied between 0.50 and 1.50
mm. A third area can be found on the model corners where
the aperture is greater than 1.5 mm. The cubic law
equation suggests that the fracture is more permeable
when the aperture is higher. Therefore, the hydraulic
conductivity should be lower in the middle and higher on
the extremities of the physical model.
 a) b) a)
20

15

Frequency (%)
10

0 0.5 1
Aperture (mm)
1.5 2 2.5
Aperture (mm)

14
b)

12

10

8
Frequency (%)

0.25 0.5 0.75 1

Aperture (mm)
Figure 6. Mechanical aperture histogram, a) Concrete slab sample, b) Construction joint sample

a) b)

Figure 7. 3D reconstruction of the fracture from CT scans, a) Concrete slab sample, b) construction joint sample
3.2 Hydraulic aperture Time (s)
0 5000 10000 15000 20000 25000
The pressure time series for port P7 were used to 0.00
establish 𝑏ℎ from the in-situ test. Figure 8 shows the -0.50
pressure time series for the three water-glycerol mixtures. Ln(Hr) = -0.000124s - 0.32
-1.00
The pressure magnitude is related to the dynamic viscosity
of the liquid. Higher dynamic viscosity values are -1.50

Ln(Hr)
associated with higher pressure values. The pressure -2.00
stabilized between 50 s and 150 s. Table 3 shows the 𝑏ℎ -2.50
values calculated from Eq. 4 for each mixture at 50, 100
-3.00
and 150 s. 𝑏ℎ varied between 0.46 mm and 0.51 mm. 𝑏ℎ
was set at 0.50 mm. The hydraulic conductivity obtained -3.50

from Eq. 1 was different for each injected mixture. It was Figure 10. Hvorslev graph obtained from the permeability
0.28, 0.11 and 0.05 cm/s for mixtures #A, #B and #C, test
respectively. The hydraulic conductivity of the fracture
decreases with increasing dynamic viscosity. The methods presented to obtain 𝑏ℎ are aimed for
different purposes. The values obtained from the variable-
2000
A
head permeability test are only applicable for the studied
1800 area that it was taken from. If the fracture surface is
B
1600 important, multiple core samples are needed to establish
1400
C
the proper 𝑏ℎ values. On the other hand, the in-situ test
captured the heterogeneity of the whole fracture media by
Pressure (kPa)

1200
1000 performing the test on the entire fracture surface.
800 Furthermore, the in-situ test took into account the in-situ
600
characteristics of an injection such as the ports volume and
400
volume losses that are not present during a variable-head
permeability test.
200
0
0 25 50 75 100 125 150
4 CONCLUSION
Injection time (s)
This article presented four methods to obtain the
Figure 8. Pressure time series for port P7 mechanical and hydraulic apertures. The aperture needs
to be considered because of its effects on the injection
Table 3. Hydraulic apertures (mm) for different injection results. It is the professional responsible of the injection to
duration (s) determine which method is better suited for his
requirements and needs. The mechanical aperture
Mixtures 50 100 150 obtained from the 3D scans can be performed on a larger
A 0.49 0.48 0.46 fracture surface but was less precise compared to the CT
scans method. Core sampling is needed to perform both
B 0.49 0.50 0.51 methods. The hydraulic aperture obtained from the in-situ
C 0.46 0.47 0.46 test requires more time, preparation and funding but
considers the in-situ characteristics of an injection
compared to the hydraulic apertures established from the
The velocity and Hvorslev graphs are presented in variable-head permeability test. Additionally, the in-situ test
Figures 9 and 10. The velocity graph permitted to establish permitted to establish the pressure distribution in time and
the piezometric error and adjust the values to produce the the head loss during the flow in the fracture.
Hvorslev graph. the piezometric error was 0.14 m. SV was
obtained from Figure 10 and 𝑏ℎ was calculated according 5 ACKNOWLEDGMENTS
to Eq. 7. 𝑏ℎ and K results were 0.11 mm and 0.93 cm/s for
water. The authors want to thank the Ministère des Transports
du Québec (MTQ), The National Sciences and Engineering
0.6 Research Council of Canada (NSERC) and the Société de
Hm= -11355(ΔH/Δt) + 0.14
0.5 transport de Montréal (STM) for the funding of this project.
The authors would also like to thank Jonathan Auger who
0.4
helped with the construction of the physical models, Joel
Grignon who performed the 3D cans and Sébastien
Hm (m)

0.3

0.2 Ménard and Richard Prowt whose helped with the in-situ
test.
0.1

0.0
-4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5
ΔH/Δt ( x 10-5 m/s)
Figure 9. Velocity graph obtained from the permeability test
6 REFERENCES Quinn, P. M., Parker, B. L., and Cherry, J. A. 2011. Using
constant head step tests to determine hydraulic
Berre, I., Doster, F., and Keilegavlen, E. 2019. Flow in apertures in fractured rock. Journal of Contaminant
fractured porous media: a review of conceptual models Hydrology, 126(1-2), 85-99.
and discretization approaches. Transport in Porous Ramandi, H. L., Mostaghimi, P., and Armstrong, R. T.
Media, 130(1), 215-236. 2017. Digital rock analysis for accurate prediction of
Cao, Y. B., Feng, X. T., Yan, E. C., Chen, G., Lü, F. F., Ji, fractured media permeability. Journal of
H. B., and Song, K. Y. 2016. Calculation method and Hydrology, 554, 817-826.
distribution characteristics of fracture hydraulic Sun, Z., Wang, L., Zhou, J. Q., and Wang, C. 2020. A new
aperture from field experiments in fractured granite method for determining the hydraulic aperture of rough
area. Rock Mechanics and Rock Engineering, 49, rock fractures using the support vector
1629-1647. regression. Engineering Geology, 271, 105618.
Cao, C., Xu, Z., Chai, J., & Li, Y. 2019. Radial fluid flow von Fay, K. F. 2015. Guide to Concrete Repair (2nd
regime in a single fracture under high hydraulic edition). U.S. Bureau of Reclamation. USA.
pressure during shear process. Journal of Wanniarachchi, W. A. M., Ranjith, P. G., Perera, M. S. A.,
Hydrology, 579, 124142.Caulk, R. A., Ghazanfari, E., Rathnaweera, T. D., Zhang, C., and Zhang, D. C. 2018.
Perdrial, J. N., and Perdrial, N. 2016. Experimental An integrated approach to simulate fracture
investigation of fracture aperture and permeability permeability and flow characteristics using regenerated
change within Enhanced Geothermal rock fracture from 3-D scanning: A numerical
Systems. Geothermics, 62, 12-21. study. Journal of Natural Gas Science and
Crandall, D., Moore, J., Gill, M., and Stadelman, M. 2017. Engineering, 53, 249-262.
CT scanning and flow measurements of shale fractures Ye, Z., Liu, H. H., Jiang, Q., and Zhou, C. 2015. Two-phase
after multiple shearing events. International Journal of flow properties of a horizontal fracture: The effect of
Rock Mechanics and Mining Sciences, 100, 177-187. aperture distribution. Advances in Water
Gao, M., Zhang, C., and Oh, J. 2023. Assessments of the Resources, 76, 43-54.
effects of various fracture surface morphology on single Zhao, Z., Li, B., and Jiang, Y. 2014. Effects of fracture
fracture flow: A review. International Journal of Mining surface roughness on macroscopic fluid flow and solute
Science and Technology, 33(1), 1-29. transport in fracture networks. Rock Mechanics and
Hao, M., Li, X., Zhong, Y., Zhang, B., Jin, D., and Chen, G. Rock Engineering, 47(6), 2279-2286.
2018. Numerical simulation of polymer grout diffusion
in a single fracture. AIP Advances, 8(10), 105329.
Huang, N., Jiang, Y., Liu, R., and Li, B. 2019. Experimental
and numerical studies of the hydraulic properties of
three-dimensional fracture networks with spatially
distributed apertures. Rock Mechanics and Rock
Engineering, 52, 4731-4746.
Huang, N., Liu, R., Jiang, Y., and Cheng, Y. 2021.
Development and application of three-dimensional
discrete fracture network modeling approach for fluid
flow in fractured rock masses. Journal of Natural Gas
Science and Engineering, 91, 103957.
Huo, D., Li, B., and Benson, S. M. 2014. Investigating
aperture-based stress-dependent permeability and
capillary pressure in rock fractures. SPE annual
technical conference and exhibition. Netherlands.
Hvorslev, M. J. 1951. Time lag and soil permeability in
ground-water observations (No. 36). USA. Waterways
Experiment Station, Corps of Engineers, US Army.
Li, X., Jiang, Z., and Couples, G. G. 2019. A stochastic
method for modelling the geometry of a single fracture:
spatially controlled distributions of aperture, roughness
and anisotropy. Transport in Porous Media, 128(2),
797-819.
Olsson, R., & Barton, N. 2001. An improved model for
hydromechanical coupling during shearing of rock
joints. International Journal of Rock Mechanics and
Mining Sciences, 38(3), 317-329.
Panasyuk, V. V., Marukha, V. I., and Sylovanyuk, V. P.
2013. Injection technologies for the repair of damaged
concrete structures. Ukraine. Springer.
A modified framework to describe stress-
strain behavior and volumetric response of
hydrate bearing sand

Maral Goharzay1*, Jeffrey Priest1, Richard Wan1

Department of Civil Engineering, University of Calgary, Calgary, AB,
Canada

ABSTRACT
Gas hydrate-bearing sands (GHBS) contain enormous reserves of methane gas, making them an attractive energy
resource. The mechanical properties of these sands are strongly influenced by the hydrate, with increases in hydrate
saturation (𝑆ℎ ) leading to higher strength, strain softening, and dilation. Field-scale tests for methane recovery from GHBS
have faced unexpected technical failures, emphasizing the need for numerical simulations to assess long-term feasibility
and minimize risks. Previous constitutive soil models modified the Mohr-Coulomb (MC) model by incorporating a
relationship between cohesion and hydrate saturation, however the occurrence of cohesion is disputed, and so the actual
stress-strain response was not truly captured. Recent researchers have utilized Rowe's stress-dilatancy theory showing
that strength increase in GHBS could be attributed to kinematics rather than cohesive influences. However, this assumption
may not be correct, especially at the initial stages of shearing. In this paper, a stress-dilatancy model is introduced that
captures the unique hydrate characteristics, soil density, and applied confining pressure. This model better represents the
geomechanical behavior of GHBS, allowing its implementation in elastoplastic models for realistic numerical simulations
and analysis.

Les GHBS, riches en méthane, offrent une source d'énergie prometteuse. Leurs propriétés mécaniques sont influencées
par les hydrates, augmentant la résistance, réduisant la déformation et provoquant une dilatation avec une saturation
accrue (Sh). Les tests de récupération de méthane ont rencontré des échecs techniques, nécessitant des simulations
numériques pour évaluer la faisabilité à long terme et réduire les risques. Les modèles antérieurs ont modifié le modèle
de Mohr-Coulomb (MC) en intégrant la relation entre cohésion et saturation des hydrates, bien que l'existence de la
cohésion soit contestée, limitant la capture précise de la réponse contrainte-déformation. Des chercheurs récents ont
proposé la théorie de dilatance de Rowe, suggérant que l'augmentation de résistance dans les GHBS est due à des
influences cinématiques plutôt que cohésives. Un modèle de dilatance contrainte, prenant en compte les caractéristiques
uniques des hydrates, la densité du sol et la pression de confinement, permet une meilleure représentation du
comportement géomécanique des GHBS et son utilisation dans des simulations numériques et des analyses réalistes.

1 INTRODUCTION Numerical simulations are required to assess long-term

viability and minimize risks of hydrate production (Collett,
Gas hydrates, particularly methane gas hydrates, have 2002) that are dependent on geomechanical models that
garnered international attention as a potential abundant can describe the stress-strain behavior of GHBS. Several
and valuable energy resource. Hydrates are ice-like geomechanical models have been developed to account
crystalline structures that form when low-density gases, for the behavior of GHBS, which typically integrate the
like methane, combine with water under specific Mohr-Coulomb failure criteria with relationships that link a
temperature and pressure conditions in deep-water marine cohesion parameter with hydrate saturation (𝑆ℎ ). However,
sediments and below the permafrost. The introduction of there is much debate in the literature as to whether hydrate
hydrate into the pore space of a sand leads to significant provides cementation (cohesion) to soil particles or can be
enhancement mechanical response of the sediment in considered as an additional ‘grain-like’ phase that is load-
which it resides, such as its strength and stiffness. As such, bearing.
the extraction of methane from hydrate, which involves the In this paper, the development of a conceptual model to
dissociation of hydrate, can lead to complex processes that better capture the stress-strain response of GHBS is
significantly affect the mechanical behavior of gas hydrate- presented. The model was developed using the stress-
bearing sediments (GHBS). These changes lead to strain response obtained from laboratory tests conducted
sediment instability, leading to potential slope failures and on water-saturated GHBS (Abbas, 2018). The improved
engineering problems associated with wellbore integrity. model incorporates the mobilization and loss of cohesion
Although field-scale tests have been conducted to evaluate during shearing as well as evaluating reduction in
the feasibility of methane recovery, these are typically of ‘apparent’ void ratio of the GHBS due to the presence of
short duration due to the high costs, and in some instances gas hydrate. This is crucial for numerical simulations that
prematurely ended due to production issues related to the seek to assess the long-term impact of hydrate production.
instability of the GHBS following hydrate dissociation.
2 BACKGROUND The Mohr-Coulomb (MC) failure criterion is the most
2.1 Soil Behavior widely used soil model, and represents the linear
relationship between the shear stress (𝜏𝑓 ) and the normal
Shear strength is a fundamental concept in soil mechanics, stress (𝜎𝑛′ ) at failure acting on a failure plane through a
representing the maximum shear stress that soil can granular soil, given by:
withstand before failure. The triaxial test (Figure 1a) is a
widely used method to assess soil strength, which involves 𝜏𝑓 = 𝜎𝑛′ 𝑡𝑎𝑛(𝜑 ′ ) [1]
subjecting a soil sample to a confining stress while
shearing the specimen under an increasing axial load. where 𝜑 ′ is the angle of internal friction of the material
Deviatoric stress (𝑞 = 𝜎1′ - 𝜎3′ , where 𝜎1′ and 𝜎3′ are mobilized at failure.
maximum and minimum principal effective stresses, As highlighted above most natural soils exhibit a degree
respectively), axial strain (𝜀𝑎 ), and volumetric strain (𝜀𝑣 ) are of bonding such that a shear stress can arise at 𝜎𝑛′ = 0.
measured after the test. Therefore, to account for this bonding a cohesion term (𝑐’)
is applied such that Eq. 1 is typically rewritten as:

a b 𝜏𝑓 = 𝑐 ′ + 𝜎𝑛′ 𝑡𝑎𝑛(𝜑 ′ ) [2]

The MC failure criterion is only strictly valid for soils that

are sheared to their critical state and cannot capture the
peak stresses associated with a dense sand compared to
the loose.
To address this weakness, Rowe (1962) proposed a
stress-dilatancy model for frictional soils, which is
applicable to the prepeak stress-strain behavior, as follows:

𝑅 = 𝐾𝐷 [3]

σ′1
where R = is the principal stress ratio.
σ′3

𝜀̇ 𝑣
𝐷 =1− [4]
𝜀̇ 𝑎
Figure 1. a) Schematic of triaxial test, b) Typical stress-
strain response of drained granular materials during triaxial 𝐾 = 𝑡𝑎𝑛2 ( +
𝜋 𝜑𝑐𝑣
) [5]
test. 4 2

𝜀̇𝑣 and 𝜀̇𝑎 are volumetric and axial strain rates, respectively
In granular soils the mobilized shear stresses are a (positive in contraction) and 𝜑𝑐𝑣 is the critical state friction
function of the frictional resistance at grain contacts that angle.
resist sliding of particles. Differences in stress-strain Also, Rowe (1962) proposed a second model to address
response occurs depending on whether the sand is loose cohesive materials. This model incorporates the concept of
or dense, which relates to the packing of soil grains in a true cohesion, represented by the parameter 𝑐, along with
unit volume. As a loose sand is subject to axial load the soil dilatancy and intergranular friction, as follows:
particles are able to rotate and slide into the available void
σ′1 π φcv 2c π φcv ε̇
space leading to compaction of the specimen and an = [tan2 ( + )+ tan ( + )] (1 − v) [6]
σ′3 4 2 σ′3 4 2 ε̇ a
increase in shear stress. This compaction process
continues until the sand reaches its critical state at which
Further modifications to Eqs. 3-5 have been proposed to
point shear stresses and specimen volume remain
better capture observed behavior. Wan and Guo (1998)
constant under increasing axial strain (Figure 1b). In
proposed a modification to account for the differences in
contrast, for dense materials after an initial compaction, soil
the initial void ratio of the specimen (𝑒) relative to the critical
particles need to ride up over each other to enable
void ratio (𝑒𝑐𝑟 ) given by:
shearing. This expansion (dilation) requires increased
shear stresses, until the specimen is at its critical state, 𝜎 1+𝑆𝑖𝑛𝜑∗
where shear stresses and volume change are constant. ( 1) ∗ 𝐷;
= 𝐾𝑐𝑣 ∗ =
𝐾𝑐𝑣 [7]
𝜎3 𝑚 1−𝑆𝑖𝑛𝜑∗
Dense sand therefore exhibits a peak stress due to the α
dilation of the sand and post peak strain softening behavior with Sinφ∗ = (e⁄ecr ) . Sinφcv
as the specimen reaches constant volume conditions
in which 𝑎 is a parameter to be determined. The angle 𝜑 ∗
(Figure 1b).
corresponds to the friction angle mobilized along a certain
In nature, most soils inherently possess a degree of
macroscopic plane which evolves during deformation.
cementation that enhances the strength of soil by
Zhang & Salgado (2010) introduced a modification to
promoting interparticle bonding and increasing resistance
capture the impact of both cohesive and frictional strength
to shear deformation. This leads to higher shear strength
components, as follows:
and a stiffer response with less compressibility for
cemented sand compared to uncemented sand.
𝜎1′ 𝜋 𝜑𝑐𝑣 𝜀̇ 2𝑐 𝜋 𝜑𝑐𝑣 𝜀̇ 𝑣 Various hydrate morphologies (Figure 3) have been
= 𝑡𝑎𝑛2 ( + ) (1 − 𝑣 ) + 𝑡𝑎𝑛 ( + )√1 − [8] considered to define the interaction between hydrate and
𝜎3′ 4 2 𝜀̇ 𝑎 𝜎3′ 4 2 𝜀̇ 𝑎
the host soil that account for differences in shear strength
observed in GHBS: pore filling, load bearing, and
2.2 Mechanical properties of GHBS cementation. Pore-filling occurs when hydrates nucleate
on the boundaries of soil particles and freely grow into the
Extensive research has been conducted to investigate the pore space, such that the soil behaves similar to a normal
impact of gas hydrate on the mechanical behavior of GHBS sand (Miyazaki et al., 2011) until 𝑆ℎ values > 25% are
(Ghiassian & Grozic, 2013; Hyodo et al., 2013; Li et al., encountered (Berge et al., 1999 ; Yun et al., 2005; Yun et
2016; Miyazaki et al., 2010; Priest et al., 2014; Winters et al., 2007) at which point the strength and stiffness of the
al., 2004). Figure 2a illustrates a typical stress-strain GHBS starts to slowly increase. Load-bearing considers
behavior of GHBS as a function of hydrate saturation (𝑆ℎ ). that hydrate bridges between soil particles and supports
It can be seen that as 𝑆ℎ increases, higher shear strength the soil matrix. This leads to a more rapid increase in
and stiffness are observed, with specimens exhibiting a strength and stiffness compared to the pore filling
peak strength at low axial strains, followed by post-peak morphology. It is suggested that load-bearing, in a practical
strain softening. Notably, the behavior of the soil transitions sense, is an extension of pore filling at higher hydrate
from contractive to dilative as the 𝑆ℎ increases. In contrast, saturations (Yun et al., 2005).Cementation is characterized
hydrate-free soils typically exhibit dominant strain by hydrate formation at particle contacts (Chaouachi et al.,
hardening behavior with maximum shear stresses 2015) that acts as a bonding agent within the soil matrix,
occurring at higher strains. Figure 2b depicts the stress- which results in a rapid increase in the strength and
strain response of highlighting that increasing the effective stiffness of the soil. Even a small level of hydrate saturation
stress for a given 𝑆ℎ , leads to an increase in shear strength can significantly enhance sediment stiffness and strength
while dilation is reduced. in this morphology (Jiang et al., 2014; Li et al., 2016; Priest
& Hayley, 2019; Priest et al., 2009).

Figure 3. Main types of hydrate morphology: a) pore filling;

b) load bearing; and c) cementation (Gai, 2018)

2.2.1 Numerical modeling of GHBS

The existing literature on the strength of soils containing

hydrates commonly uses the MC framework, with a focus
on including cohesion (Eq. 2) that is depend on hydrate
saturation 𝑆ℎ (Klar et al., 2010; Miyazaki et al., 2010;
Pinkert & Grozic, 2014; Rutqvist & Moridis, 2007). Another
perspective (Pinkert, 2017a; 2017b; 2019) considered
Rowe’s stress-dilatancy theory (Rowe, 1962) using Eqs. 3-
5 to explain the enhancement in strength that arises in
GHBS, suggesting that the impact of hydrate in the pore
space is kinematic in nature (function of friction angle of the
GHBS) and does not include any cohesion component.
This model was supported by experimental data from
triaxial tests conducted on water-saturated hydrate-bearing
Figure 2. Deviatoric stress (solid line) and volumetric strain soils (Hyodo et al, 2013, etc). Their conclusion was based
(dashed line) against axial strain for sand specimens with on the observation that when using Eqs. 6 and 8 (the
(𝑎) different hydrate saturations and (𝑏) different effectives modified versions of Rowe’s theory that included cohesion)
stresses (redrawn from Hyodo et al., (2013)) (Priest & the modeled results deviated from measured stress-strain
Hayley, 2019) behavior observed in the lab tests. However, it is known
that in cemented soils cohesion is mobilized at the
beginning of shearing and lost as shear strain increases. 𝑒𝑐𝑟0 , ℎ𝑐𝑟 and 𝑛𝑐𝑟 are 0.88, 75.86 MPa, and 0.3216,
Recent studies on water-saturated GHBS (Priest et al., respectively, that are found from experimental data.
2021) also observed such behavior. In addition, the stress-
dilatancy plots presented by Pinkert exhibited significant
deviations from a linear line at the early stages of tests on
a
GHBS.
It is evident that Rowe’s stress-dilatancy model does not
accurately reflect observed behavior, and more effective
modifications to Eqs. 3-5 that account for void ratio and
variations in cohesion should be considered to better
describe the dilatancy behavior of GHBS.

3. MODEL DEVELOPMENT
3.1 Application of Rowe’s theory to GHBS

As highlighted in Figure 2 the mechanical properties of a

sand are greatly affected by 𝑆ℎ and 𝜎 ′ . In order to
investigate the application of Rowe’s theory, the stress-
strain responses of a sand with different 𝑆ℎ and confining b
stresses were selected from laboratory tests conducted at
the University of Calgary (Abbas, 2018). Table 1 provides
a summary of the relevant details for each specimen,
including the initial void ratio, confining stress, and 𝑆ℎ .

Table 1. Specimens’ information

Confining
Hydrate Void
Specimen stress
saturation (Sh%) ratio (e)
(MPa)
WHS1 39.6 0.5 0.54
WHS2 37.6 1 0.55
WHS3 13.48 0.5 0.55
Figure 4. Investigation of the Effect of 𝑆ℎ on the stress-
strain-dilatancy behavior of GHBS
Figure 4a highlights the stress-strain response for the
three GHBS specimens highlighted in Table 1. It can be
observed that increasing 𝑆ℎ and confining stress results in
increased peak strength at lower axial strains. The stress
dilatancy plots (Figure 4b), derived from using the data
presented in Figure 4a, clearly deviate from the theoretical
line given by Rowe's theory (yellow line) assuming a
constant 𝜑𝑐𝑣 of 31° . The deviation from theory could be
related to cohesion being present or the reduction in
apparent void ratio of the GHBS specimen. Thus, the
consideration of other factors such as void ratio, cohesion,
and other relevant parameters is required to better map the
theory to lab results.

3.1.1 Wan and Guo (1998) modification

To investigate whether the change in void ratio of the Figure 5. Changes in void ratio and mean effective stress
specimen due to inclusion of hydrate is a factor, Wan & during shear tests for different sands when tested without
Guo (1998) void ratio modification (named W&G herein) hydrate. From this, a critical state line for sand can be
was considered. In using the W&G model (Eq. 7), the derived (red curve)
critical state line (CSL) of the host sand is required. Figure
5 shows the stress path followed for sand without hydrate
at various confining stress during triaxial shearing (Abbas, Applying Eq. 7 to the GHBS data, with the appropriate
2018; Hyodo et al. 2013), from which a CSL was developed 𝑒𝑐𝑟 value, an appropriate constant alpha value (α =
(red line). The values of CSL parameters are as follows: −0.1016), and using an apparent void ratio of GHBS
(eGHBS =
esand(1−𝑆ℎ )
), Figure 6 is obtained. Although the WHS1 and WHS2 for both scenarios are highlighted in
1+esand×𝑆ℎ Figures 7 and 8, respectively. Table 2 shows the variables
W&G modification appears to improve the mapping of for μ and 𝜅 using these different approaches, with the
theory to the lab results, the discrepancy still remains resulting fits shown in Figure 9 and Figure 10. It can be
noteworthy. This implies that assuming the influence of seen that mapping WHS1 to WHS2, or vice versa, results
hydrate in the pore space as purely kinematic in nature is in reasonably good fit, suggesting that cohesion is not
somewhat invalid. The deviation of the W&G-modified influenced by effective stress. The minor discrepancies that
theoretical line from the real data may suggest that can be seen mainly arise due to the different cohesion
cohesion does influence the behavior of GHBS. values mobilized for WHS1 and WHS2 (Figures 7-8).

Figure 6. Stress-dilatancy plots taking into account the

W&G (1998) modification of void ratio Figure 7. Cohesion variation and the related fitted curves
for two scenarios – PrePeak (WHS1)
3.2 Effect of cohesion on stress-dilatancy behaviour of
GHBS behaviour

To investigate the effect of cohesion on GHBS behavior, it

is necessary to determine a cohesion function that better
reflects the changes in cohesion that may occur during a
shear test. Two scenarios were considered. In the first
scenario, Eq. 8- Zhang and Salgado’s (2010) stress-
dilatancy theory (named Z&S herein)- was used to
calculate the changes in cohesion manifest during a shear
test. The resultant cohesion values were then fitted with a
function of the form shown in Eq. 9, where the unknown
variables μ and 𝜅 are obtained using the trial-and-error
method to achieve a suitable fit. This function was
consequently used to derive a theoretical curve for WHS2
using Eq. 8. This method was also adopted to fit WHS2
data to WHS1.
Figure 8. Cohesion variation and the related fitted curves
c(𝜀𝑎 ): [9]
for two scenarios – PrePeak (WHS2)
For 𝜀𝑎 < 𝜀𝑎(cmax) :
μTanh(−εa + 2εcmax ) × (slopeorigin to cmax ) × εa Table 2 Information of the model parameters (𝜇 and 𝜅)
Second
For 𝜀𝑎(cmax) ≤ 𝜀𝑎 ≤ 𝜀stress peak : scenario
First scenario
κTanh(εa − εcmax ) × (slope cmax to cstress peak ) × εa + cmax (Z&S with
Specimen Parameters (Z&S without
W&G
modifications)
modification
In the second scenario, the W&G modification was and 𝑒ℎ𝑟 )
initially applied to WHS1 before subsequently deriving a μ 2 2
WHS1
cohesion function that was then applied to WHS2. This 𝜅 1.1 1.1
method was also adopted to fit WHS2 data to WHS1. The μ 2.93 2.93
WHS2
variation in cohesion and fitted cohesion functions for 𝜅 1.6 1.6
Figure 9. Stress-dilatancy response of WHS2 using WHS1 Figure 11. Stress-dilatancy response of WHS3 using
cohesion function WHS1 cohesion function.

Figure 12 shows the application of this modification to

cohesion in mapping the stress-dilatancy to the test results.
The better alignment of these curves, compared to those in
Figure 11, is apparent suggesting that this modification to
𝑐 is appropriate.
This approach was also implemented based on utilizing
the WHS2 cohesion function (Figure 13) and subsequently
applying the same modification using Eq. 10 where 𝑆ℎ for
WHS2 is substituted for WHS1, with Figure 14 showing the
successful alignment between the fitted curves and the real
data.

Figure 10. Stress-dilatancy response of WHS1 using

WHS2 cohesion function

Figure 11 shows the stress-dilatancy plots for WHS3

estimated based on applying the WHS1 cohesion function
for both scenario 1 and 2. Given that 𝑆ℎ of WHS1 and
WHS2 is approximately 2.93 times that of WHS3, the
discrepancies observed in the fitted stress-dilatancy
responses shown in Figure 11 suggest that cohesion is a
function of 𝑆ℎ .
To better align the stress-strain responses for WHS3, a
modification was considered to scale the cohesion function
based on the ratio of 𝑆ℎ between the tests. The proposed Figure 12. Evaluation of the validity of the modifications to
modification is: map the fitted curves onto the real data (WHS3 data) based
on WHS1 cohesion function
𝑝
𝑆ℎ(𝑊𝐻𝑆3)
𝑐𝑊𝐻𝑆3,𝑀𝑜𝑑𝑖𝑓𝑖𝑒𝑑 = 𝑐(𝜀𝑎 )𝑊𝐻𝑆1 × ( ) [10]
𝑆ℎ(𝑊𝐻𝑆1)
4. CONCLUSION
where 𝑐(𝜀𝑎 )𝑊𝐻𝑆1 is WHS1 cohesion function and 𝑝 is a
variable to achieve an appropriate fit, which is determined The results of laboratory tests conducted on water-
through a trial-and-error method. The value of 𝑝 is 1.5 for saturated GHBS at the University of Calgary (Abbas, 2018)
the first scenario and 3 for the second scenario. were used to examine the influence of hydrate on the
stress-strain behavior of a sand. Utilizing Rowe’s stress-
 vice versa, showing that cohesion was a factor in the
strength of GHBS and not solely governed by kinematics,
with cohesion being strongly related to 𝑆ℎ and not confining
stress.
The comprehensiveness of the proposed model was
evaluated by applying the cohesion functions derived from
samples with higher 𝑆ℎ (WHS1 and WHS2) to estimate the
stress-dilatancy behavior of WHS3 with lower 𝑆ℎ . To
account for the difference in 𝑆ℎ between WHS1/WHS2 and
WHS3, a modification (to account for changes in 𝑆ℎ ) was
employed to align the fitted curves with the real data, with
modeling results confirming the effectiveness of the model
and the corresponding modification. However, due to the
limited dataset considered, in terms of 𝑆ℎ and confining
stresses, further studies are required to determine if the
developed model is appropriate for differing test conditions.
Figure 13. Stress-dilatancy response of WHS3 using
WHS2 cohesion function
6. REFERENCES

Abbas, M. (2018). Geomechanical Characteristics of

Hydrate-bearing Sands University of Calgary].
Chaouachi, M., Falenty, A., Sell, K., Enzmann, F., Kersten,
M., Haberthur, D., & Kuhs, W. F. (2015). Microstructural
evolution of gas hydrates in sedimentary matrices
observed with synchrotron X-ray computed
tomographic microscopy. Geochemistry Geophysics
Geosystems, 16(6), 1711-1722.
Chu, J., & Wanatowski, D. (2009). Effect of Loading Mode
on Strain Softening and Instability Behavior of Sand in
Plane-Strain Tests. Journal of Geotechnical and
Geoenvironmental Engineering, 135(1), 108-120.
Collett, T. S. (2002). Energy resource potential of natural
gas hydrates. Aapg Bulletin, 86(11), 1971-1992.
Coulomb, C. A. (1776). Essai sur une Application des
Rggles de Maximis et Minimis a Quelques ProblSmes
Figure 14. Evaluation of the validity of the modifications to de Statique RSatife a 1’Architecture. MSmoires de
map the fitted curves onto the real data (WHS3 data) based l’Academie Royale des Sciences 7, 343-382.
on WHS2 cohesion function. Gai, X. (2018). Geomechanical modeling of gas hydrate
bearing sediments and other complex soils Texas A&M
University].
dilatancy theory and assuming zero cohesion (𝑐 = 0), it Ghiassian, H., & Grozic, J. L. H. (2013). Strength behavior
was shown that the fitted model did not map to the stress- of methane hydrate bearing sand in undrained triaxial
dilatancy response of water-saturated GHBS. A testing. Marine and Petroleum Geology, 43, 310-319.
subsequent stress-dilatancy model was developed Helgerud, M. B., Dvorkin, J., Nur, A., Sakai, A., & Collett,
considering factors such as void ratio and cohesion in the T. (1999). Elastic-wave velocity in marine sediments
stress-dilatancy relationship for GHBS. with gas hydrates: Effective medium modeling.
By accounting for changes in void ratio resulting from the Geophysical Research Letters, 26(13), 2021-2024.
inclusion for hydrate in the pore space and applying Wan Hyodo, M., Li, Y. H., Yoneda, J., Nakata, Y., Yoshimoto,
and Guo (1998) void ratio modification factor, provided a N., Nishimura, A., & Song, Y. C. (2013). Mechanical
better approximation to the stress-dilatancy response. behavior of gas-saturated methane hydrate-bearing
However, this approach still failed to accurately capture the sediments. Journal of Geophysical Research-Solid
observed stress-dilatancy behavior, suggesting that Earth, 118(10), 5185-5194.
hydrate may also be active as a cohesive element during Jiang, M. J., Chen, H., Tapias, M., Arroyo, M., & Fang, R.
shearing. (2014). Study of mechanical behavior and strain
To assess whether hydrate contributes a cohesive localization of methane hydrate bearing sediments with
component to the shear strength, cohesion functions were different saturations by a new DEM model. Computers
determined for a given test (e.g., WHS1) and then used to and Geotechnics, 57, 122-138.
model the response of WHS2 (which had similar 𝑆ℎ but Klar, A., Soga, K., & Ng, M. Y. A. (2010). Coupled
different confining stresses). The results provided a deformation-flow analysis for methane hydrate
suitable fit when the WHS1 cohesion function was applied extraction. Geotechnique, 60(10), 765-776.
to estimate the stress-dilatancy behavior of WHS2, and
Li, Y. H., Liu, W. G., Zhu, Y. M., Chen, Y. F., Song, Y. C., methane gas hydrate. American Mineralogist, 89(8-9),
& Li, Q. P. (2016). Mechanical behaviors of permafrost- 1221-1227.
associated methane hydrate-bearing sediments under Yun, T. S., Francisca, F. M., Santamarina, J. C., & Ruppel,
different mining methods. Applied Energy, 162, 1627- C. (2005). Compressional and shear wave velocities in
1632. uncemented sediment containing gas hydrate.
Miyazaki, K., Masui, A., Sakamoto, Y., Aoki, K., Tenma, N., Geophysical Research Letters, 32(10), Article L10609.
& Yamaguchi, T. (2011). Triaxial compressive Zhang, J., & Salgado, R. (2010). Stress-dilatancy relation
properties of artificial methane-hydrate-bearing for Mohr-Coulomb soils following a non-associated flow
sediment. Journal of Geophysical Research-Solid rule. Geotechnique, 60(3), 223-226.
Earth, 116, Article B06102.
Miyazaki, K., Masui, A., Tenma, N., Ogata, Y., Aoki, K.,
Yamaguchi, T., & Sakamoto, Y. (2010). Study on
Mechanical Behavior for Methane Hydrate Sediment
Based on Constant Strain-Rate Test and Unloading-
Reloading Test Under Triaxial Compression.
International Journal of Offshore and Polar
Engineering, 20(1), 61-67.
Pinkert, S. (2017a). Rowe's Stress-Dilatancy Theory for
Hydrate-Bearing Sand. International Journal of
Geomechanics, 17(1), Article 06016008.
Pinkert, S. (2017b). The lack of true cohesion in hydrate-
bearing sands. Granular Matter, 19(3), Article 57.
Pinkert, S. (2019). Dilation Behavior of Gas-Saturated
Methane-Hydrate Bearing Sand. Energies, 12(15),
Article 2937.
Pinkert, S., & Grozic, J. L. H. (2014). Prediction of the
mechanical response of hydrate-bearing sands.
Journal of Geophysical Research-Solid Earth, 119(6),
4695-4707.
Priest, J. A., Best, A. I., & Clayton, C. R. I. (2005). A
laboratory investigation into the seismic velocities of
methane gas hydrate-bearing sand. Journal of
Geophysical Research-Solid Earth, 110(B4), Article
B04102.
Priest, J. A., & Hayley, J. L. (2019). Strength of Laboratory
Synthesized Hydrate-Bearing Sands and Their
Relationship to Natural Hydrate-Bearing Sediments.
Journal of Geophysical Research-Solid Earth, 124(12),
12556-12575.
Priest, J. A., Rees, E. V. L., & Clayton, C. R. I. (2009).
Influence of gas hydrate morphology on the seismic
velocities of sands. Journal of Geophysical Research-
Solid Earth, 114, Article B11205.
Rowe, P. W. (1962). The Stress-Dilatancy Relation for
Static Equilibrium of an Assembly of Particles in
Contact Proc. of the Royal Society of London,
Rutqvist, J., & Moridis, G. J. (2007). Numerical studies on
the geomechanical stability of hydrate-bearing
sediments Offshore Technology Conference,
Song, Y. C., Zhu, Y. M., Liu, W. G., Li, Y. H., Lu, Y., & Shen,
Z. T. (2016). The effects of methane hydrate
dissociation at different temperatures on the stability of
porous sediments. Journal of Petroleum Science and
Engineering, 147, 77-86.
Wan, R. G., & Guo, P. J. (1998). A simple constitutive
model for granular soils: Modified stress-dilatancy
approach. Computers and Geotechnics, 22(2), 109-
133.
Winters, W. J., Pecher, I. A., Waite, W. F., & Mason, D. H.
(2004). Physical properties and rock physics models of
sediment containing natural and laboratory-formed
Evaluation of the Relative Accuracies of
Two- and Four-Parameter Models for
Predicting Maximum and Minimum Void
Ratios for Sand-Silt Mixtures
Carmine P. Polito
Valparaiso University, Valparaiso, Indiana, USA

ABSTRACT le modèle à quatre paramètres ont produit des

estimations très précises. De plus, des
The maximum and minimum index void ratios of recommandations sont fournies pour choisir le
sand-silt mixtures can be predicted using the filling modèle à utiliser en fonction de plusieurs scénarios
coefficient, a, and embedment coefficient, b with an possibles.
appropriate model. There are three implementations
of these models: The four-parameter model, the emin-
two-parameter model and the emax-two parameter 1 INTRODUCTION
model. The four-parameter model uses two values of
the filling coefficient, a, and two values of the For predicting the maximum and minimum index void
embedment coefficient, b. Conversely, the two- ratios sand-silt mixtures, one of the more commonly
parameter models use one pairing of the filling used models was developed by Chang, Wang and
coefficient, a, and the embedment coefficient, b. Ge (2016). They developed a series of equations
for predicting the maximum and minimum index void
A study was performed to evaluate the relative ratios (emax and emin) for mixtures of sand and silt
accuracies of two-parameter models and four- based on binary packing. Implementing their
parameter model in predicting the maximum and methodology requires several common soil
minimum index void ratios of sand-silt mixtures. It parameters and either two or four empirical
was found that the emax-two parameter model was variables: the filling coefficient, a, and the
the least accurate model. Both the emin-two- embedment coefficient, b. The number of empirical
parameter model and the four-parameter model constants required is dependent on whether the
produced very accurate estimates. Additionally, predictions are based on the filling coefficients and
recommendations are provided for choosing which embedment coefficients from both the maximum and
model to use based on several possible scenarios. minimum index void ratio data (the four-parameter
model) or solely upon the filling coefficients and
RÉSUMÉ embedment coefficients from either the maximum
index void ratio data (the emax-two parameter model)
Les taux de vide d'indice maximum et minimum des or the minimum index void ratio data (the emin-two
mélanges sable-limon peuvent être prédits en parameter model).
utilisant le coefficient de remplissage, a, et le Using Chang, et al.’s models, if one has a
coefficient d'enfouissement, b avec un modèle collection of maximum and minimum index void
approprié. Il existe trois implémentations de ces ratios at various silt contents for the soil in question,
modèles : le modèle à quatre paramètres, le modèle one can determine the filling and embedment
emin à deux paramètres et le modèle emax à deux coefficients and then predict the maximum and
paramètres. Le modèle à quatre paramètres utilise minimum index void ratios at any other silt content
deux valeurs du coefficient de remplissage, a, et using the four-parameter model. It is also possible
deux valeurs du coefficient d'encastrement e, b. À to determine a and b based solely on minimum
l'inverse, les modèles à deux paramètres utilisent un density/maximum void ratio tests using one of the
couple du coefficient de remplissage, a, et du two-parameter models. If one does not possess such
coefficient d'enrobage, b. data, either the data must be obtained through
laboratory testing or correlations for predicting the
Une étude a été réalisée pour évaluer les précisions filling and embedment coefficients must be used
relatives des modèles à deux paramètres et du (Polito, 2021).
modèle à quatre paramètres dans la prédiction des If the intent of the user is to predict the index
taux de vide d'indice maximum et minimum des void ratios for a mixture of sand and silt with a
mélanges sable-limon. Il a été constaté que le specific silt content, the need to run multiple index
modèle à deux paramètres emax était le modèle le density tests at various silt contents negates the
moins précis. Le modèle emin à deux paramètres et main advantage of using the models to predict the
index void ratios (i.e. reducing the number of entirely contained in the voids between the sand
laboratory tests performed). For this reason, the grains, to a silt matrix that contains isolated sand
author developed a series of equations for predicting grains. Below the TFC, the soil behaves essentially
the filling and embedment coefficients. These as a sand; above the TFC the soil behaves
equations are based on the median grain sizes of the essentially as a silt (Polito and Sibley, 2020).
sand and the silt, which are easily obtained from a For soils with silt contents below the threshold
grain-size analysis. fines content, thus having a coarse-grain dominated
The goal of the study was to compare the soil matrix:
relative accuracy of the three models in predicting
emax and emin for mixtures of sand and silt using
parameters obtained through laboratory testing. 𝑚𝑎𝑥
𝑒𝑀1 = 𝑒1𝑚𝑎𝑥 𝑦1 + 𝑒2𝑚𝑎𝑥 𝑦2 − 𝑎𝑚𝑎𝑥 (1 + 𝑒2𝑚𝑎𝑥 )𝑦2 [1]

2 BACKGROUND 𝑚𝑖𝑛
𝑒𝑀1 = 𝑒1𝑚𝑖𝑛 𝑦1 + 𝑒2𝑚𝑖𝑛 𝑦2 − 𝑎𝑚𝑖𝑛 (1 + 𝑒2𝑚𝑖𝑛 )𝑦2 [2]

As increasing amounts of silt are added to a clean

sand, the classification of the soil changes from sand For soils with silt contents above the threshold
to silty sand to sandy silt and finally to silt. These fines content, thus having a fine-grain dominated soil
changes in soil composition lead to changes in both matrix:
the maximum and minimum index void ratios (emax
and emin).
Chang, et al. (2016) presented four equations 𝑚𝑎𝑥
𝑒𝑀2 = 𝑒1𝑚𝑎𝑥 𝑦1 + 𝑒2𝑚𝑎𝑥 𝑦2 − 𝑏 𝑚𝑎𝑥 𝑒1𝑚𝑎𝑥 𝑦1 [3]
for predicting the maximum and minimum index void
ratios for mixtures of a sand and a silt over the range
of possible silt contents, where the silt content is 𝑚𝑖𝑛
𝑒𝑀2 = 𝑒1𝑚𝑖𝑛 𝑦1 + 𝑒2𝑚𝑖𝑛 𝑦2 − 𝑏 𝑚𝑖𝑛 𝑒1𝑚𝑖𝑛 𝑦1 [4]
defined as being the ratio of the mass of silt in a
specimen to the combined mass of sand and silt in
the specimen. In the equations, silt content can Table 1: Soil Parameters and Empirical Constants
range from 0 (representing a pure sand) to unity Required for Each Model
(representing a pure silt).
The four-parameter model uses one pairing of
the filling coefficient, amax, and the embedment
coefficient, bmax, for predicting maximum index void
ratios, emax, and a second pairing of filling coefficient,
amin, and the embedment coefficient, bmin, for
predicting minimum index void ratios, emin.
Conversely, two-parameter models use one
pairing of the filling coefficient, and the embedment
coefficient, for predicting both the maximum index
void ratio and the minimum index void ratio.
Although Chang et al. only used data from maximum
void ratio tests for their two-parameter model, this
study also examined a two-parameter model based
on data from minimum void ratio tests. Henceforth,
the two-parameter model based on data from
maximum void ratio tests (i.e. Chang et al.’s model)
shall be referred to as the emax-two-parameter model
and the two-parameter model based on data from
minimum void ratio tests shall be referred to as the 3 MODEL EVALUATION
emin-two-parameter model.
The nature and number of soil parameters In order to evaluate the relative accuracies of the
required to implement the three models are three models, the models were used to predict
dependent upon the model chosen. The soil values of emax and emin using best-fit values of the
parameters and empirical constants required for filling and embedment coefficients determined
each model are outlined in Table 1. through computer analyses. The predicted values
Equations for the four-parameter model are were then compared to the corresponding values
provided as Equations 1 through 4. The emax-two- determined through laboratory testing and the
parameter model uses equations 1 and 3 and the goodness of fit determined. The coefficients of
emin-two-parameter model uses Equations 2 and 4. determination were calculated for emax and emin for
The threshold fines content (TFC) represents each of the 58 soil combinations for each of the three
the silt content at which the soil begins to transition models.
from a sand matrix, in which the silt particles are
 Polito (2021) provides details of how the parameter model against the values of e max
best-fit filling and embedment coefficients were measured in the laboratory. Figure 6 plots the values
developed for each combination of silt and sand and of emin predicted using the emax-two-parameter model
how a coefficient of determination, R2, was then against the values of emin measured in the laboratory.
developed to evaluate the ability of the best-fit Figure 7 presents a histogram of the
coefficients to predict the maximum and minimum relative frequencies of the R2 values calculated using
index void ratios for each sand-silt combination. the maximum index void ratios for the 58 sand-silt
The results of the analyses performed to combinations based on the best-fit filling and
evaluate the relative accuracy of the four-parameter
model and each of the two-parameter models are
discussed below.

3.1 The Four-Parameter Model

The results of the analyses performed using the four-

parameter model can be seen in Figures 1 and 2.
Figure 1 plots the values of emax predicted using the
four-parameter models against the values of emax
measured in the laboratory. Figure 2 plots the values
of emin predicted using the four-parameter model
against the values of emin measured in the laboratory.
Figure 3 presents a histogram of the
relative frequencies of the R2 values calculated using Figure 2: Measured versus predicted minimum index
the maximum index void ratios for the 58 sand-silt void ratios developed using the four-parameter
combinations based on the best-fit filling and model
embedment coefficients and the four-parameter
model.
Figure 4 presents a histogram of the
relative frequencies of the R2 values calculated using
the minimum index void ratios for the 58 sand-silt
combinations based on the best-fit filling and
embedment coefficients and the four-parameter
model.

Figure 3: Relative frequency of R2 values for

maximum index void ratios predicted using the four-
parameter model

Figure 1: Measured versus predicted maximum

index void ratios developed using the four-parameter
model

3.2 The emax-Two-Parameter Model

The emax-two-parameter model uses the best-fit

filling and embedment coefficients (amax and bmax, Figure 4: Relative frequency of R2 values for
respectively) determined from the results of minimum index void ratios predicted using the four-
maximum index void ratio tests. The results of the parameter model
analyses performed using the emax-two-parameter
model can be seen in Figures 5 and 6. Figure 5 plots embedment coefficients and the emax-two-parameter
the values of emax predicted using the emax-two- model.
 Figure 8 presents a histogram of the
relative frequencies of the R2 values on the best-fit
filling and the minimum index void ratios for the 58
sand-silt based on the best-fit filling and embedment
coefficients and the emax-two-parameter model.

Figure 8: Relative frequency of R2 values for

minimum index void ratios predicted using the emax-
two-parameter model

3.3 The emin-Two-Parameter Model

Figure 5: Measured versus predicted maximum The emin-two-parameter model uses the best-fit filling
index void ratios developed using the emax-two- and embedment coefficients (amin and bmin,
parameter model respectively) determined from the results of
minimum index void ratio tests. The results of the
analyses performed using the emin-two-parameter
model can be seen in Figures 9 and 10. Figure 9
plots the values of emin predicted using the emin-two-
parameter model against the values of e max
measured in the laboratory. Figure 10 plots the
values of emin predicted using emin-two-parameter
model against the values of emin measured in the
laboratory.
Figure 11 presents a histogram of the
relative frequencies of the R2 values calculated using
the maximum index void ratios for the 58 sand-silt
combinations based on the best-fit filling and
embedment coefficients and the emin-two-parameter
model.
Figure 12 presents a histogram of the
relative frequencies of the R2 values calculated using
the minimum index void ratios for the 58 sand-silt
Figure 6: Measured versus predicted minimum index based on the best-fit filling and embedment
void ratios developed using the emax-two-parameter coefficients and the emin-two-parameter model.
model

Figure 7: Relative frequency of R2 values for Figure 9: Measured versus predicted maximum
maximum index void ratios predicted using the emax- index void ratios developed using the e min-two-
two-parameter model parameter model
Figure 10: Measured versus predicted minimum
index void ratios developed using the e min-two-
parameter model Figure 13: Comparison of coefficients of
determination determined for emax using the two- and
four-parameter models

Similarly, Figure 14 presents a histogram of

the coefficients of determination for the minimum
index void ratios predicted using each of the three
models. Table 2 summarizes the results of the
comparison of the three models as well as providing
the percentage of R2 values greater than 0.95 and
greater than 0.80 produced by each model.
All three models predicted the maximum
index void ratios with mean R2 values above 0.96.
The four-parameter model and the emin-two
parameter model both predicted the maximum index
void ratios with mean R2 values above 0.94. The
emax-two parameter model was less accurate in
Figure 11: Relative frequency of R2 values for predicting the minimum index void ratios producing a
maximum index void ratios predicted using the e max- mean R2 value of 0.88.
two-parameter model

Figure 12: Relative frequency of R2 values for

minimum index void ratios predicted using the e min- Figure 14: Comparison of coefficients of
two-parameter model determination determined for emin using the two- and
four-parameter models
A histogram of the coefficients of
determination for the maximum index void ratios From the results presented in Table 2, it can
predicted using each of the three models is be seen that, while the emax-two-parameter model
presented in Figure 13. produces reasonably accurate predictions, the
predictions made with the four-parameter model are
more accurate. Both the mean and the median
coefficients of determination were approximately 4%
higher for the four-parameter model and the
standard deviation of the four-parameter model was 1. For determining the filling and embedment
about one-third the standard deviation of the emax- coefficients, maximum and minimum index
two-parameter model. Most significantly, the four- void ratio data for at least three silt contents
parameter model returned results that very closely in addition to the data for the pure sand and
matched (here defined here as an average R 2 value pure silt should be used.
greater than 0.95) the void ratios measured in the lab 2. If data from both maximum and minimum
97% of the time, while the emax-two-parameter model index data are available, the-four parameter
only returned such results 87% of the time. The four- model should be used.
parameter models returned R2 values that closely 3. If only data from minimum index data are
matched (here defined here as an average R 2 value available, the emin-two-parameter model
greater than 0.80) the void ratios measured in the lab should be used.
100% of the time, while the emax-two-parameter 4. If only data from maximum index data are
model only returned such results 94% of the time. available, the emax-two-parameter model
Similarly, in Table 2 it can be seen that the can be used with caution. The filling and
accuracy of the predictions made using the e min-two- embedment coefficients should be
parameter model very closely matched the accuracy confirmed using correlations such as those
of the predictions made with the four-parameter previously developed by the author (Polito,
model. Both the mean and the median coefficients 2021).
of determination are only 1% higher for the four-
parameter model and the standard deviation of the
two models is comparable. 5 CONCLUSIONS

Table 2: Comparison of the average coefficients of A study was performed to evaluate the relative
determination produced using the two-parameter accuracies of the four-parameter model, the emax-
and four-parameter models two-parameter model and the emin-two-parameter
model in predicting the maximum and minimum
index void ratios for mixtures of sand and silt with
different percentages of silt.
In the comparison of the models, the four-
parameter model was found to produce the most
accurate values for both the maximum and minimum
index void ratios. The emin-two-parameter model was
also found to produce accurate values for both the
maximum and minimum index void ratios. The emax-
two-parameter model was also found to produce
accurate values for maximum index void ratios but
was found to be less accurate for minimum index
void ratios.

6 ACKNOWLEDGEMENTS

The author would like to thank Valparaiso University

for the financial support provided through the Alfred
W. Sieving Chair of Engineering.

The main difference in accuracy between 7 References

the two models is that the four-parameter model
returned results that very closely matched (here Chang, C., Wang, L., and Ge, L. 2016, Maximum and
defined here as an average R2 value greater than minimum void ratios for sand-silt mixtures,
0.95) the void ratios measured in the lab 97% of the Engineering Geology, 211:7-18.
time, while the emin-two-parameter model only Polito, C. 2021, Regression Models for Estimating
returned such results 87% of the time. Both models Parameters a And b For Chang, Wang and
returned R2 values larger than 0.80 for all 58 sand- Ge’s Maximum And Minimum Void Ratio
silt combinations. Models, GeoNiagra 2021, Niagara Falls,
Ontario, Canada, Paper 288.
Polito, C. and Sibley, E. 2020, Threshold fines
4 MODEL SELECTION content and behavior of sands with
nonplastic silts, Canadian Geotechnical
Based on the results of this study, the author Journal, 57: 3.
recommends the following guidelines:
Study of fresh properties of cemented
tailings material with ternary cement blends
Aparna Sagade, Dr. Mamadou Fall
Department of Civil Engineering University of Ottawa, Ottawa, ON,
Canada

ABSTRACT
For the sustainable advancement of the mining sector, alternative binders,
such as ground granulated blast furnace slag (Slag) and fly ash, have been adopted to partially replace ordinary Portland
cement (OPC) in the preparation of cemented paste backfill (CPB). However, the supply of these materials is limited and
may not be sufficient for the industry's future demands. Limestone (LS), an inexpensive and abundant resource, has the
potential to complement these binders when used as a ternary mix. This paper presents experimental results on the effects
of ternary blended cement with slag and limestone powder on key engineering fresh properties of CPB: (i) rheological
properties and (ii) setting time. CPB samples with PCI/Slag ratios of 50/50 and 80/20 were considered, and the effect of
LS in the ternary mix was investigated by replacing the slag with increasing doses of LS (5, 10, and 20 wt. %). The findings
have shown that the replacement of slag by a higher dosage of LS in the ternary binder reduces the yield stress and
increases the viscosity of CPB. The effect of LS on the rheological properties of CPB has mainly been attributed to the
physical (e.g., filler effect, lubrication, particle size) and chemical (nucleation and changes in repulsive forces between
CPB particles) effects of LS. Further, an increase in the proportion of limestone in the CPB ternary binder system prolonged
the setting time (initial and final) of the CPB mixture due to the nucleation and dilution effects of LS. Overall, the optimal
usage of LS and slag in the ternary system can serve as a sustainable alternative to the widely used OPC or OPC/Slag
binary binder, thereby decreasing the energy consumption and carbon footprint associated with cement and CPB
technology.

Keywords: cemented paste backfill; tailings; rheology; setting time; ternary binder; sustainable mining

RÉSUMÉ
Pour l'avancement durable du secteur minier, des liants alternatifs, tels que le laitier granulé de haut fourneau (Slag) et
les cendres volantes, ont été adoptés pour remplacer partiellement le ciment Portland ordinaire (OPC) dans la préparation
du remblai en pâte cimentée (CPB). Cependant, l'offre de ces matériaux est limitée et pourrait ne pas être suffisante pour
les demandes futures de l'industrie. Le calcaire (LS), ressource peu coûteuse et abondante, a le potentiel de compléter
ces liants lorsqu'il est utilisé comme mélange ternaire. Cet article présente des résultats expérimentaux sur les effets du
ciment mélangé ternaire avec du laitier et de la poudre de calcaire sur les principales propriétés techniques fraîches du
CPB : (i) propriétés rhéologiques et (ii) temps de prise. Des échantillons de CPB avec des rapports PCI/Slag de 50/50 et
80/20 ont été considérés, et l'effet de LS dans le mélange ternaire a été étudié en remplaçant le laitier par des doses
croissantes de LS (5, 10 et 20 % en poids). Les résultats ont montré que le remplacement du laitier par une dose plus
élevée de LS dans le liant ternaire réduit le seuil d'écoulement (ou de cisaillement) et augmente la viscosité du CPB. L'effet
du LS sur les propriétés rhéologiques du CPB a été principalement attribué aux effets physiques (par exemple, effet de
remplissage, lubrification, taille des particules) et chimiques (nucléation et modifications des forces répulsives entre les
particules de CPB) du LS. De plus, une augmentation de la proportion de calcaire dans le système de liant ternaire CPB
a prolongé le temps de prise (initial et final) du mélange CPB en raison des effets de nucléation et de dilution du LS. Dans
l'ensemble, l'utilisation optimale du LS et du laitier dans le système ternaire peut servir d'alternative durable au liant binaire
OPC ou OPC/Slag largement utilisé, réduisant ainsi la consommation d'énergie et l'empreinte carbone associées à la
technologie du ciment et du CPB.

1 INTRODUCTION mine/rock drainage and tailings dam failures (Ercikdi et al.,

2009; Wu et al., 2013). Therefore, innovative mine waste
The mining sector contributes considerably to the economy management technologies, such as cemented paste
of numerous nations worldwide. Sustainable management backfill technology, have been developed over the recent
of huge amounts of mining waste (e.g., tailings) is one of decades to reduce and tackle the above-mentioned
the most critical issues and challenges faced by the mining problems related to the surface disposal of mine waste
sector (Liu & Fall, 2022; Jiang et al., 2020) Traditional (tailings).
practices such as surface waste disposal often cause Cement paste backfill (CPB) has proven to be effective for
geotechnical and geo-environmental hazards, such as acid mine waste management and ground support. Considering
its economic, environmental, and social benefits, CPB is compressive strength at young ages and a low rate of
widely used in the modern mining industry of several strength gain of CPB in comparison to those of CPB with
nations, such as Australia, Canada, China, Germany, and cement only (Snellings et al., 2022). Additionally, the fly
South Africa (Fall & Benzaazoua, 2005; Wu et al., 2013). ash and slag supply may be limited and insufficient to meet
CPB is a composite mixture of tailings (70-85% of total future industry needs (Bentz et al., 2017; Snellings et al.,
solids), water (processed or fresh), and a relatively low 2022). Moreover, Xiao (2021) noted an increase in yield
amount of binder (typically 2-8% by weight of total solids) stress and viscosity with increasing slag percentage, which
(Fall & Benzaazoua, 2005; Haruna & Fall, 2020). reduces the fluidity of CPB, ultimately negatively affecting
Important performance and technical properties of CPB flowability and delivery to stopes (Wu et al., 2013).
fresh CPB include flowability and setting time. The fresh Consequently, in the past few years, the use of
CPB mixture is typically prepared on the ground surface in limestone powder (LS) as a partial replacement material for
a backfill plant and then transported by means of gravity cement has received increasing interest. Limestone has
and/or pumping through the pipeline to the excavated the benefits of being abundant and cheap (Bentz et al.,
cavities underground (Liu & Fall, 2022). The successful 2017; Wang et al., 2018a). Several previous investigations
transport of CPB, which is a function of its fluidity, is key to on cementitious materials have reported that limestone has
the successful application of CPB technology in any mine, effectively improved the rheological behavior and
as poor fluidity can lead to pipe clogging, delayed transport, workability of cement paste, concrete, and mortar materials
disruption of mining operations, and ultimately financial (Courard and Michel, 2014; Jiang et al., 2020; Vance et al.,
losses (Simon & Grabinsky, 2013). The challenge of 2013a). This is mainly attributed to its filler and nucleating
practical and efficient transportation of CPB is growing as effect (Kumar et al., 2013; Vance et al., 2013b). Wang et
many mines are located at greater and greater depths, al. (2018a) recommended the addition of LS aluminum rich
often resulting in longer transport distances, which requires SCMs into cementitious systems for overall performance
fresh CPB to be prepared with excellent flowability. The improvement. These studies revealed that the addition of
fluidity of CPB is mainly related to its yield stress and LS to Portland cement systems leads to an increase in the
viscosity (Ali et al., 2021; Liu & Fall, 2022). Indeed, with rate of hydration at early ages, thus causing high early
the technological advances and the progressive strength, but it can diminish the subsequent strength
exhaustion of ore at shallow depths in numerous world because of the dilution effect when added at a higher
regions, mining activities are going deeper, with a replacement (Menéndez et al., 2003; Mounanga et al.,
maximum depth of up to 4350 m deep (Mponeng, South 2011).
Africa) (Ziegler et al., 2015). The setting time of CPB From the preceding discussion, it can be deduced that
provides crucial information on its early age strength the combined use of limestone powder and slag in a
development, in other words, its mechanical stability at ternary blended cement (LS, slag, OPC) has the potential
early ages. A rapid gain in early age strength is important to formulate a CPB binder with adequate early age strength
to reduce the risk of liquefaction of CPB to ensure or development and suitable rheological properties while
improve the safety of working conditions and increase mine decreasing the cost of CPB and the carbon footprint of the
productivity (Yin et al., 2012). mining industry. However, the effect of ternary cement mix
Ordinary Portland cement (OPC) is the predominant (Portland cement, slag, limestone) on fresh CPB properties
binder used in the production of CPB; however, it accounts is not well understood, and no studies on the development
for 75-80% of the cost of CPB (Grice, 2001; Haiqiang et al., of rheological properties (yield stress, viscosity) and setting
2016). In addition, clinker production accounts for 5-8% of time (initial, final) of CPB with ternary cement mixes have
the world’s man-made CO2 emissions (Gartner & Hirao, been conducted. All previous studies on the impact of this
2015). The above factors have driven mining firms to look ternary cement mixture on the fresh properties of the
for cement substitutes that decrease the Portland cement cementitious mix system such as concrete or mortar.
content (and thus the cost of CPB) and reduce the mining However, concrete and mortar are different from CPB and
industry's carbon footprint while keeping or enhancing the thus cannot be directly applicable to CPB. The aims of this
mechanical and rheological properties of CPB. Thus, in research are to experimentally investigate the effects of the
recent decades, the use of a binder consisting of Portland ternary cement mixture (Portland cement, slag, limestone)
cement and supplementary cementitious material (SCM), on the fresh properties of CPB. These properties include
such as blast furnace slag (slag), also known as binary rheological properties (viscosity and yield stress) and
cement, has been increasingly adopted in CPB technology setting time.
and practice. These SCMs are often utilized in CPB to
lower costs and improve material performance properties 2 EXPERIMENTAL PROGRAMS
(Haiqiang et al., 2020). For example, replacing Portland
cement with slag or fly ash in the binary mix format is an 2.1 Materials
effective way to improve the long-term mechanical strength
and microstructural properties of CPB (at older ages), Tailings: Silica tailings (ST) with 99.8 wt% quartz are the
consequently improving the mechanical stability and main constituents of the CPB preparation for this
environmental performance of the CPB structure at experimental program. Their physical characteristics are
advanced ages, thereby reducing the overall cost (Fall & presented in Table 1. Quartz is the most common mineral
Pokharel, 2010) and carbon footprint of CPB. However, in tailings from Canadian hard rock mines; therefore, the
due to the poor reaction kinetics of slag or fly ash, their use use of ST is most appropriate to correlate the results of the
as a binary binder with cement in CPB results in low experimental work with the actual Canadian mining
scenario. Moreover, its chemically inert nature reduces the of the CPB mix is presented in Table 4. The fresh samples
interference from additional chemical reactions that may were prepared by dry mixing predetermined proportions of
occur due to active minerals or impurities present in the the materials in the mixer for 1 min, followed by wet mixing
natural tailings. for a total of 7 minutes to ensure homogeneity of the mix.

Table 1. Physical characteristics of the tailings 2.2 Test methods

Element Gs D10 D30 D50 D60

2.2.1 Rheological tests
Unit - µm µm µm µm
The prepared samples were then used to perform viscosity
31.5
ST 2.7 1.9 9.0 22.5 and yield stress tests to investigate the rheology of fresh
CPB using a 10 cm x 10 cm mold for the yield stress and
Avg. of 9
--- 1.8 9.1 20.0 30.8 viscosity tests. The specimens were covered with a plastic
tailings
film throughout the test. In addition, to account for the
Gs: spec. Gravity; Ss: spec. surface area(m2/g); ST: silica continuous movement of the CPBs during transport and to
tailings; wt.: weight. ensure homogeneity, the samples were shaken for 1
minute before performing each yield stress or viscosity test.
and then stabilized for 30 seconds. This process mimics
the shearing of paste backfill during transportation and
avoids the settlement of tailings particles (due to self-
weight) during curing (Haiqiang et al., 2016). The yield
stress and viscosity tests were conducted twice to ensure
the accuracy of the results.

Table 2. Mix composition of the CPB samples

Sample Taili- Binder %PCI %Sl % W/B

nomenclature ng Conte- ag LP
nt (%)
50P-50S-0L 50 50 0
Figure 1. Grain size distribution of the silica tailings ST, 50P-45S-5L 50 45 5
PCI, Slag, and LS.
50P-40S-10L 50 40 10 7.35
Table 1. Physical characteristics of the binders ST 4.5
50P-30S-20L 50 30 20

Element RD Ss D10 D50 D90 80P-20S-0L 80 20 0

Unit µm µm µm
GGBS 2.8 2.10 0.77 8.4 27.6 80P-15S-5L 80 15 5
LS 2.93 2.69 0.28 2.31 7.94 80P-10S-10L 80 10 10
PCI 3.2 1.30 0.99 14.9 39.76
80P-0S-20L 80 0 20
Binders and Mixing water
Yield Stress Test:
In this experimental program, CPB specimens were made
using PCI as the main binder with varying weight Yield stress and viscosity are the primary parameters that
percentages of slag and limestone powder (CaCO3) in represent the flowability of fresh CPBs (Liu & Fall, 2022).
place of cement. Figure 2 displays the particle size Yield stress is the lowest stress needed to initiate flow,
distribution of ST and all binders. Two mix proportions indicating the magnitude of particle-particle forces (Simon
using 4.35 wt.% binder and 7.35 water/binder ratio were & Grabinsky, 2013). Since CPB is a non-Newtonian fluid,
chosen in this study, with PCI/Slag proportions of 50/50 the Wykeham-Farrance vane shear apparatus suitable for
and 80/20 by weight. In addition, slag was replaced with 5, soil and non-Newtonian fluids were used for this study. The
10, and 20% limestone in the ternary binder. The chemical apparatus consists of a vane arrangement with four 2.5cm
compositions and physical characteristics of the binders x 2.5cm blades powered by an electric motor. Then the
are listed in Tables 2 and 3, respectively. To study the torque was applied, and the resistance of the sample
rheological properties of the sample, distilled water is used against the blades of the vane was measured by the
to avoid the impact of impurities from tap water. attached deflection spring. The maximum torque reading
was measured on the comparator. The test was performed
Specimen preparation on all CPB specimens with different proportions of binder
mixture over a curing period of 0, 0.5, 1-, 2-, and 4 hours
Two groups of samples A and B, with 50 wt.% and 80 wt.% following ASTM D4648/D4648M-16 (ASTM 2011). The
CPB, respectively, were prepared using a binder content of specimens were covered with a thin plastic film. The yield
4.35 wt.% and a water-to-binder ratio of 7.35. The design stress was determined based on the following equation:
 τy =
2Tm
1 𝐻
into three phases. The phase between 50 °C and 200 °C
π𝐷3 [ + ]
3 𝐷 400 °C and 500 °C, and 600 °C and 800 °C indicates free
here τy is the yield stress, Tm is the maximum torque, H is water and calcium silicate hydrates, ettringite, monosulfate
the length, and D is the diameter of the vane, respectively. decomposition, Portlandite (CH) dihydroxylation, and
CaCO3 decarbonization, respectively (Deboucha et al.,
Viscosity test: 2017).

Viscosity is a crucial aspect of CPB behavior. It 2.2.4 Monitoring experiments

corresponds to the force required to maintain the flow of
the mixture. The viscosities of the samples were measured A set of monitoring experiments were performed on fresh
using the Brookfield digital viscometer (model LVDV-E, CPBs prepared in the same way as the samples used for
with a base spring torque of 673.7 dynes-cm). The principle rheological testing. An electrical conductivity (EC) test was
involves, measuring the viscous drag of the fluid measured done on fresh CPB specimens to examine the progress of
against the rotating spindle through the spring deflection; the cement hydration by monitoring changes in EC caused
finally, the viscosity was reported in cp (Brookfield, 2016). by ion movement and changes in water with time. Changes
The size and shape of the spindle and the spindle speed in EC were monitored using the 5TE sensor. The sensor
are chosen in rpm depending on the type of mixture. For has an accuracy of ± 10% for EC measurement, ranging
this study, spindle number 5 and a rotational speed of 50 from 0 dS/m to 23 dS/m. The freshly prepared samples
rpm were chosen. Fresh samples were poured into a 10cm were poured into a 20cm x 10cm mold, the sensors were
x 10cm mold and capped with thin plastic wraps to prevent inserted into the mold in the center and closed tightly with
water evaporation, as when the backfill was transported in the cap. EC and temperature changes were tracked by an
practice by pipes. The samples were tested for 0, 0.5, 1, 2, EM50 data logger.
4 hours curing time, 3 times for accuracy of results.
3 RESULTS AND DISCUSSIONS
2.2.2 Setting time measurements
3.1 Effect of ternary binder system (PCI-Slag-LS) on
Knowledge of setting time is important to the overall the rheological behavior of CPB
productivity, mechanical and economic performance of the
backfill operation. Initial set denotes the loss of workability It was noticed that the partial substitution of slag by LS
and the start of paste backfill stiffening or strength gain. In powder, with PCI as the ternary binder, has a considerable
addition, early strength gains are crucial for barricade impact on the overall rheology of CPB (yield stress and
opening and extraction of neighboring stopes to ultimately viscosity). Figure 2 illustrates the influence of the ternary
increase mining efficiency by reducing the mining cycle binder with varying dosages of LS on the yield stress
(Yin et al., 2012). Therefore, the impact of ternary binder evolution in the CPB samples. The yield stress of the
(PCI-Slag-LS) on setting time was investigated using the mixtures in both groups A and B, with PCI/Slag ratios of
Vicat apparatus. The test is based on the principle of 80/20, and B 50/50, respectively, decreased with
mixture resistance to needle penetration and is performed increasing dosage of limestone in place of slag.
according to ASTM C191-13. In this test, cylindrical molds Regardless of the cement and slag content, i.e., the yield
of 100 mm diameter and 40 mm height were used. The stress of the freshly prepared sample (0 hours), after the
initial setting time was recorded when the needle reached addition of 0-20 wt% LS, decreased from 230 Pa to 187 Pa,
25 mm penetration from the bottom, and the final setting in group A and from 216 Pa to 158 Pa in group B,
time was considered when the outer collar of the needle respectively. In addition, during 4 hours of curing, the yield
did not leave an impression on the specimens. stress reduced from 489 Pa to 374 Pa in group A and from
461 Pa to 360 Pa in group B, respectively. In other words,
2.2.3 Microstructural analyses the addition of LS to CPB improves its flowability, which is
beneficial in mine backfilling practice. This yield stress
To comprehend the evolution of the hydration process of reduction effect due to LS addition is attributed mainly to
CPB with the ternary binder at the microstructural level, by the following physical effects or factors (morphological,
measuring the sudden mass changes under the application filler, and dilution) of limestone, as explained below:
of heat, thermogravimetry (TG) and differential TG (DTG) (i) The lubrication provided by the limestone particles to
were performed on cement paste specimens with the mixture (Bentz et al., 2017); in other words, during the
considered binder proportions. The specimens were made hydration process, different hydration products are
with a binder-to-water ratio of 1:1 to accelerate the produced, and ions are released in suspension, the
hydration process, and samples were oven dried at 45 °C location of hydroxyl groups (OH-) on the surface of calcite
until the mass stabilized, then powdered and sieved (Ca2+), gives a means of electrostatic repulsion between
through a 40-micron sieve. The thermal analyzer, TGA the particles and thus, improve the flowability of the
Q5000 V20.13, was used to carry out a thermogravimetric suspension and reduces further flocculation (Bentz et al,
analysis on the cement paste samples up to a temperature 2017; Sekkal & Zaoui, 2013); ii) LS finer than cement acts
of 1000°C. The progression of cement hydrate as a filler, fills water-filled pores and expels excess water,
decomposition is represented by the TGA curve, and the thereby increasing the fluidity and decreasing the yield
endothermic peak by the derivative of the TGA curve. stress of the mixture (Zheng et al., 2016). However, further
Furthermore, the typical cement hydration curve is divided increasing the dosage of fine LS may increase the surface
area and water demand, which could increase the yield water film. The dominant phenomenon of these two effects
stress. An additional factor (chemical factor) should be decides the ultimate increase or decrease in viscosity
considered as contributing to the decrease in the yield (Hoshino et al., 2006). Indeed, the addition of LS dosage
stress of the CPB mixture when adding LS up to 20 wt.%. increased the viscosity of the samples in groups A and B
This factor is the change in the repulsive forces between with PCI/Slag (50/20 and 80/20). However, the influence of
particles in the CPB system due to the addition of LS. LS viscosity on the samples in group B with 80% PCI was
noted to be higher than on the samples in group A with 50%
PCI. This may be due to the greater availability of cement
and reactive alumina for hydration. In addition, more
cement in the system utilizes all available LS as nucleation
sites, enhancing the hydration reaction and potentially
forming more hydration products.

a) Group A CPB samples (binder with 50% PCI)

b)

a) Group A CPB samples (binder with 50% PCI)

b) Group B CPB samples (binder with 80% PCI)

Figure 2. Influence of LS on the yield stress of CPB

samples a) Group A b) Group B
b) Group B CPB samples (binder with 80% PCI)
On the contrary, the viscosity of the samples in both
groups, A with PCI/Slag 50/20 and B with 80/50, increased, Figure 3. Influence of LS on the viscosity of CPB samples
with the percentage of limestone replacing slag, regardless a) Group A CPB samples b) Group B CPB samples.
of the PCI/Slag ratio in the ternary binder as depicted in
Figure 3. The increase in viscosity upon the addition of LS The acceleration of the hydration rate due to the
is mainly attributed to its filler and nucleating effect. The nucleation effect of LS is also correlated with the
viscosity is mainly a function of solid content, particle size monitoring of EC; as shown in Figure 4a, it is seen that the
distribution in the mixture, and water demand (Vance et al., conductivity of the samples with 5 and 10 wt.% LS reached
2013; Xiao et al., 2021), which can be correlated with the a maximum value of 3.90 mS/cm and 3.76 mS/cm at 5
surface area of LS. Indeed, the SSA of LS is 2.69 (m 2/g) hours and 5.4 hours, while the EC of 50P-50S-0L peaked
and higher than that of slag and cement, which may at 6 hours with 3.56 mS/cm, which is lower compared to
increase the water demand of the mixture, increasing the the ternary mixture samples, indicating a slow hydration
viscosity. The finer limestone induces the nucleation effect rate. A similar pattern is observed in Group B samples,
in the mixture (Kumar et al., 2013; Oey et al., 2013), which from Figure 4b, the samples with 5 and 10 wt.% LS peaked
accelerates the degree of hydration and can produce more before 4 hours with peak values of 4.74 mS/cm and 4.26
hydration products (C-S-H, C-H). Thus, higher hydration mS/cm, respectively, while the EC of 80P-20S-0L peaked
products increase the volume and concentration of solids after 4 hours. This rise in hydration products due to LS
in the mixture and, ultimately, the viscosity of the samples. addition mentioned above is also observed and confirmed
However, this can also increase the amount of free water by the DTG/TG results, as shown in Figure 5. The first and
on the top surface, which can increase the thickness of the second peak, which represent the hydration products (e.g.,
C-S-H, ettringite, and CH) in the system, were found to be 3.2 Development of initial and final setting times of
slightly higher for the samples with 20% LS 80P-0S-20L CPB with binary and ternary cement blends
than the control sample 80P-20S-0L, respectively, this
indicates the acceleration of cement hydration in the The results of the measurements of the setting time (initial
presence of limestone. Also, as expected, an intense and final) of all the CPB samples with binary and ternary
increase during the 3rd peak, which represents the cement blends are shown in Figure 6. It indicates the trend
decomposition of CaCO3, was noted for the samples with of increase/retardation in the setting time with a decrease
LS because it added CaCO3 into it (Feng et al., 2022). in cement content, irrespective of the variation in slag and
limestone content. Control sample 80P-20S-0L from group
B (with 80% PCI) has a shorter initial and final setting time
(at 18 and 32 hrs) than that of the control sample of group
A 50P-50S-0L (with 50% PCI) (at 22 and 38 hrs). This
effect was attributed to the generation of more hydration
products in the samples with higher cement content,
leading to the early setting as already demonstrated with
DT-DTG results discussed above. Further,
hardening/solidification of the mix containing cement
mainly occurs due to the creation of hydration products of
the cement-like C-S-H, Portlandite, and ettringite (Klein &
Simon, 2006). As in the hardening pastes, the setting is the
process of connection of isolated or weakly bound particles
which form solid paths in paste (Yin et al., 2012).
a) Group A CPB samples (binder with 50% PCI)
Figure 6 also shows the influence of the ternary binder,
with a varying dosage of LS in replacement with slag, on
the setting time of the CPB mix. It is noticed that the
addition of an increasing percentage of limestone
prolonged/delayed the setting time of the CPB mix. The
initial and final setting time of group A control sample 50P-
50S-0L was 22.5 hrs and 38hrs respectively; upon addition
of 5% to 20% LS, the initial set of samples was noted as
23.5 and 26.5 hrs, and the final setting time at 39 hrs and
41 hrs respectively. The initial setting time was delayed by
4hrs and the final setting time by 3hrs, compared to the
control sample with 0% LS. The same pattern was
observed in the samples of group B; the initial and final
setting time of control sample 80P-20S-0L was observed
b) Group B CPB samples (binder with 80% PCI) at 18 and 31hrs, respectively. Further, upon addition of
incremental dosage of LS from 5% to 20%, the initial
Figure 4. Changes in EC of the CPB samples with 0%, 5%, setting time was delayed and noted at 18.5 hrs and 31.5
and 10% of LS a) Group A b) Group B hrs and the final setting time at 20.5 and 33 hrs. This infers
the maximum delay with 20% of LS in the initial and final
setting time was 2.5 and 2 hrs, respectively.

Figure 5. TG/DTG analysis results for specimens cured for

4hrs of curing time.
Figure 6. The effect of the addition of LS in the CPB
samples on the evolution of initial and final setting time a)
Group A CPB samples and b) Group B CPB samples
 v.) The setting time (initial and final) of the samples with
This delay in setting time upon the addition of LS is 80/20 PCI/Slag mix was found to be shorter than the
primarily due to the nucleation and dilution effects as sample with 50/50 because of formation of more
previously observed in cement-based materials. However, hydration products (C-S-H, ettringite, AFm) in the
from Figure 5 TG results of samples, it is seen that less sample with higher cement content.
product of hydration formed in 50P-30S-20L, with 20wt% vi.) The delay in setting time of CPB samples when LS was
of LS, than in 50P-50S-0L, no such formation of additional added in increasing dosage was observed and is
hydration products is reported. Indeed, in a sample from consistent with the DTG/TG results. This indicates the
group B, a sample with 20 wt.% LS, 80P-0S-20L, a nucleation effect of LS on cement hydration; however,
somewhat higher peak was observed than 80P-20S-0L. more hydration product was not formed in all samples.
However, the amount of the additional hydration products Despite the results obtained in this work, detailed chemical
was insufficient to shorten the sample's setting time. and microstructural studies are recommended to gain a
Further, as indicated in Figure 4, an increase in EC was deeper insight into the impact of LS on the microstructure
noted when LS powder was added to the control samples and chemical composition of the CPB system with a ternary
from both groups. However, higher EC may not be directly binder (PC, slag, LS). In addition, the use of a modeling
correlated with the formation of more hydration products as tool to optimize the use of LS in CPB with a ternary binder.
it may not be the result of only additional hydration products
but due to the finer nature of LS with higher surface area, 5 REFERENCES
which may influence the movement of ions in pore solution.
Hence it can be inferred that mainly nucleation and dilution Ali, G., Fall, M., & Alainachi, I. (2021). Time- and
effects are dominating and influencing setting time, as the Temperature-Dependence of Rheological Properties of
chemical effect of LS seems to have negligible or little Cemented Tailings Backfill with Sodium Silicate.
impact (delay) on the setting time of CPB as the chemical Journal of Materials in Civil Engineering, 33(3),
reaction between cement hydrate and LS does not occur 04020498, (ASCE)MT.1943-5533.0003605
before 24hours (Ren et al., 2020) . The Delay in setting Bentz, D. P., Ferraris, C. F., Jones, S. Z., Lootens, D., &
time observed upon the addition of LS is similar to Zunino, F. (2017). Limestone and silica powder
elsewhere (Meddah et al., 2014) replacements for cement: Early-age performance,
Cement and Concrete Composites, 78, 43–56.
4 SUMMARY AND CONCLUSIONS Courard, L., & Michel, F. (2014). Limestone fillers cement
based composites: Effects of blast furnace slags on
Extensive experimental work was performed on several fresh and hardened properties. Construction and
CPB specimens with variable PCI/Slag content (80/20 and Building Materials, 51, 439–445.
50/50) and increasing dosage of LS in place of slag were Cyr, M., Lawrence, P., & Ringot, E. (2006). Efficiency of
cast and cured, and their rheological properties were mineral admixtures in mortars: Quantification of the
evaluated over a curing period of 4 hours (0, 0.25, 1, 2, 4 physical and chemical effects of fine admixtures in
hours). The impact of LS on the flowability of CPB and the relation with compressive strength. Cement and
microstructure of CPB at different curing times is Concrete Research - CEM CONCR RES, 36, 264–277.
investigated through rheological tests (yield stress and Deboucha, W., Leklou, N., Khelidj, A., & Oudjit, M. N.
viscosity) and microstructural analysis. A summary of the (2017). Hydration development of mineral additives
major findings from the results is as follows, blended cement using thermogravimetric analysis
i.) Regardless of the different dosages of cement, slag, (TGA): Methodology of calculating the degree of
and LS, yield stress and viscosity with increased curing hydration. Construction and Building Materials, 146,
time due to the increase in hydration products, with the 687–701.
progress of cement hydration and the self-desiccation. Ercikdi, B., Cihangir, F., Kesimal, A., Deveci, H., & Alp, I.
ii.) The binary samples with PCI/Slag, i.e., 50/50, gained (2009).Utilization of industrial waste products as
higher yield stress with time than the samples with pozzolanic material in cemented paste backfill of high
PCI/Slag 80/20. The phenomenon is due to interparticle sulphide mill tailings.
friction resulting from the filler effect of the slag and a Journal of Hazardous Materials, 168(2–3), 848—856.
denser matrix due to its finer nature. Fall, M., & Benzaazoua, M. (2005). Modeling the effect of
iii.) The addition of LS with incremental dosage (5-20 wt%) sulphate on strength development of paste backfill and
in the form of a ternary binder impacts the rheology binder mixture optimization.
(flowability) by decreasing the yield stress and Cement and Concrete Research, 35(2), 301–314.
increasing the viscosity of fresh CPB. Interestingly, the Fall, M., & Pokharel, M. (2010). Coupled effects of sulphate
effect of LS is complementary to the effect of slag on and temperature on the strength development of
the rheology of the mix. cemented tailings backfills: Portland cement-paste
(iv.) This effect of LS on the rheology of CPB, the reduction backfill.Cement and Concrete Composites, 32(10),
in yield stress, was primarily attributed to the physical 819–828.
effects of LS, i.e., the finer nature., which provides Feng, K., Xu, Z., Zhang, W., Ma, K., Shen, J., & Hu, M.
additional lubrication to the mix, and additional (2022). Rheological Properties and Early-Age
repulsive forces. However, the viscosity increased due Microstructure of Cement Pastes with Limestone
to the nucleation effect of finer limestone and the higher Powder, Redispersible Polymer Powder, and Cellulose
water demand with higher surface area. Ether.Materials, 15(9), 3159.
Gartner, E., & Hirao, H. (2015). A review of alternative Paste Considering Fineness Effects. Journal of
approaches to the reduction of CO2 emissions Materials in Civil Engineering, 32(10), 04020289.
associated with the manufacture of the binder in Sekkal, W., & Zaoui, A. (2013). Nanoscale analysis of the
concrete. Cement &Concrete Research, 78, 126–142. morphology and surface stability of calcium carbonate
Grice, T. (2001). Recent mine developments in Australia. polymorphs. Scientific Reports, 3(1), 1587.
Proceedings of the 7th International Symposium on Simon, D., & Grabinsky, M. (2013). Apparent yield stress
Mining With Backfill, 351–357. measurement in cemented paste backfill. International
Haiqiang, J., Fall, M., & Cui, L. (2016). Yield stress of Journal of Mining, Reclamation and Environment,
cemented paste backfill in sub-zero environments: 27(4), 231–256.
Experimental results. Minerals Engineering,92,140-150 Snellings, R., Machner, A., Bolte, G., Kamyab, H.,
Haruna, S., & Fall, M. (2020). Time- and temperature- Durdzinski, P., Teck, P., Zajac, M., Muller, A., de
dependent rheological properties of cemented paste Weerdt, K., & Haha, M. B. (2022). Hydration kinetics of
backfill that contains superplasticizer.Powder ternary slag-limestone cements: Impact of water to
Technology, 360, 731–740. binder ratio and curing temperature. Cement and
Hoshino, S., Yamada, K., & Hirao, H. (2006). XRD/Rietveld Concrete Research, 151, 106647.
Analysis of the Hydration and Strength Development of Vance, K., Aguayo, M., Oey, T., Sant, G., & Neithalath, N.
Slag and Limestone Blended Cement. Journal of (2013a). Hydration and strength development in ternary
Advanced Concrete Technology, 4(3), 357–367. portland cement blends containing limestone, fly ash,
Jiang, D., Li, X., Lv, Y., Zhou, M., He, C., Jiang, W., Liu, Z., or metakaolin. Cement and Concrete Composites, 39,
& Li, C. (2020). Utilization of limestone powder and fly 93–103.
ash in blended cement: Rheology, strength and Vance, K., Kumar, A., Sant, G., & Neithalath, N. (2013).
hydration characteristics. Construction and Building The rheological properties of ternary binders containing
Materials, 232, 117228. Portland cement, limestone, and metakaolin or fly ash.
Jiang, H., Fall, M., Yilmaz, E., Li, Y., & Yang, L. (2020). Cement and Concrete Research, 52, 196–207.
Effect of mineral admixtures on flow properties of fresh Wang, D., Shi, C., Farzadnia, N., Shi, Z., Jia, H., & Ou, Z.
cemented paste backfill: Assessment of time (2018). A review on use of limestone powder in cement-
dependency and thixotropy. Powder Technology, 372, based materials: Mechanism, hydration and
258–266. microstructures. Construction and Building Materials,
Klein, K., & Simon, D. (2006). Effect of specimen 181, 659–672.
composition on the strength development in cemented Wang, Y., Wu, A., Zhang, L., Jin, F., & Liu, X. (2018).
paste backfill. Canadian Geotechnical Journal, 43(3), Investigating the Effect of Solid Components on Yield
310–324. Stress for Cemented Paste Backfill via Uniform Design.
Kumar, A., Oey, T., Kim, S., Thomas, D., Badran, S., Li, J., Advances in Materials Science and Engineering, 2018,
Fernandes, F., Neithalath, N., & Sant, G. (2013). 1–7.
Simple methods to estimate the influence of limestone Wu, D., Fall, M., & Cai, S. J. (2013). Coupling temperature,
fillers on reaction and property evolution in cement hydration and rheology of fresh cemented
cementitious materials. Cement and Concrete paste backfill. Minerals Engineering, 42, 76–87.
Composites, 42, 20–29. Xiao, B., Fall, M., & Roshani, A. (2021). Towards
Liu, S., & Fall, M. (2022). Fresh and hardened properties Understanding the Rheological Properties of Slag-
of cemented paste backfill: Links to mixing time. Cemented Paste Backfill. International Journal of
Construction and Building Materials, 324, 126688. Mining, Reclamation and Environment, 35(4), 268–290.
Meddah, M. S., Lmbachiya, M. C., & Dhir, R. K. (2014). Yin, S., Wu, A., Hu, K., Wang, Y., & Zhang, Y. (2012). The
Potential use of binary and composite limestone effect of solid components on cemented paste backfill's
cements in concrete production. Construction and rheological and mechanical properties. Minerals
Building Materials, 58, 193–205. Engineering, 35, 61–66.
Menéndez, G., Bonavetti, V., & Irassar, E. F. (2003). Zheng, J., Zhu, Y., & Zhao, Z. (2016). Utilization of
Strength development of ternary blended cement with limestone powder and water-reducing admixture in
limestone filler and blast-furnace slag. Cement and cemented paste backfill of coarse copper mine tailings.
Concrete Composites, 25(1), 61–67. Construction and Building Materials, 124, 31–36.
Mounanga, P., Khokhar, M. I. A., El Hachem, R., & Loukili, Ziegler, M., Reiter, K., Heidbach, O., Zang, A., Kwiatek, G.,
A. (2011). Stromeyer, D., Dahm, T., Dresen, G., & Hofmann, G.
Improvement of the early-age reactivity of fly ash and (2015). Mining-Induced Stress Transfer and Its
blast furnace slag cementitious systems using Relation to a Mw 1.9 Seismic Event in an Ultra-deep
limestone filler materials, 44(2), 437–453. South African Gold Mine. Pure and Applied
Oey, T., Kumar, A., Bullard, J. W., Neithalath, N., & Sant, Geophysics, 172(10), 2557–2570.
G. (2013). The Filler Effect: The Influence of Filler
Content and Surface Area on Cementitious Reaction
Rates. Journal of the American Ceramic Society, 96(6),
1978–1990.
Ren, Q., Xie, M., Zhu, X., Zhang, Y., & Jiang, Z. (2020).
Role of Limestone Powder in Early-Age Cement
Study of the Impact of Nano-particles on
strength development of cemented paste
backfill
Amirreza Saremi, Mamadou Fall
Department of Civil Engineering – University of Ottawa, Ottawa,
Ontario, Canada

ABSTRACT
This paper evaluates how adding various types of nanoparticles (NP) affects one of the key engineering parameters of
cemented paste backfill (CPB), which is strength development as a function of curing time. To achieve this objective,
varying doses of four distinct kinds of NPs, including nano-silica (SiO2), nano-calcium carbonate (CaCO3), nano-iron oxide
(Fe2O3), and nano-aluminum oxide (Al2O3), were utilized. To avoid NP agglomeration in the CPB microstructure, which
could have a detrimental impact on the CPB's mechanical performance, an ether-based Polycarboxylate Superplasticizer
(SP) was used to evenly disperse the NPs throughout the hydration reaction medium. Type I ordinary Portland cement
was used to prepare CPB and tap water was also utilized to blend the CPB components. A comprehensive testing and
monitoring programme was developed to evaluate the effect of NPs on the development of the strength of CPB. Uniaxial
compressive test (UCS) was done to measure the sample's strength over curing time. Nano-CPB samples or cement
pastes were subjected to several microstructural analyses or tests, including thermal analysis (thermogravimetry (TG),
differential thermogravimetry (DTG)), mercury intrusion porosimetry (MIP) tests, and X-ray diffraction (XRD) analysis to
gain insight into the fundamental mechanisms responsible for the behaviour observed. In the absence of the dispersing
agent, the aggregated NPs were found to degrade the strength development of nano-CPB samples. In contrast to the
control sample, nano-CPB samples combined with a dosage of SP exhibited greater UCS values. Microstructural analysis
results indicated that the addition of NPs increased the generation of hydration products, which primarily increased the
resistance to shear stress between tailings particles. The highest overall improvement in nano CPB's strength was reported
in samples containing a fixed concentration of SP (0.125%) and NPs (1%).

RÉSUMÉ
Cet article évalue comment l'ajout de différents types de nanoparticules (NP) affecte l'un des principaux paramètres
techniques du remblai en pâte cimentée (CPB), à savoir le développement de la résistance en fonction du temps de
durcissement. Pour atteindre cet objectif, des doses variables de quatre types distincts de NP, dont la nano-silice (SiO2),
le nano-carbonate de calcium (CaCO3), le nano-oxyde de fer (Fe2O3) et le nano-oxyde d'aluminium (Al2O3), ont été
utilisées. Pour éviter l'agglomération des NP dans la microstructure du CPB, qui pourrait avoir un impact négatif sur les
performances mécaniques du CPB, un superplastifiant polycarboxylate (SP) à base d'éther a été utilisé pour disperser
uniformément les NP dans le milieu de réaction d'hydratation. Le ciment Portland ordinaire de type I a été utilisé pour
préparer le CPB et l'eau du robinet a également été utilisée pour mélanger les composants du CPB. Un programme
complet d'essais et de surveillance a été mis au point pour évaluer l'effet des NP sur le développement de la résistance
du CPB. Des essais de compression uniaxiale (UCS) ont été réalisés pour mesurer la résistance de des échantillons au
cours du temps de durcissement. Les échantillons de nano-CPB ou les pâtes de ciment ont été soumis à plusieurs
analyses ou tests microstructuraux, notamment des analyses thermiques (thermogravimétrie (TG), thermogravimétrie
différentielle (DTG)), des tests de porosimétrie par intrusion de mercure (MIP) et des analyses de diffraction des rayons X
(XRD) afin de mieux comprendre les mécanismes fondamentaux responsables du comportement observé. En l'absence
d'agent dispersant, il a été constaté que les NPs agrégées dégradaient la résistance des échantillons de nano-CPB.
Contrairement à l'échantillon de contrôle, les échantillons de nano-CPB combinés à un dosage de SP ont présenté des
valeurs UCS plus élevées. Les résultats de l'analyse microstructurale ont indiqué que l'ajout de NP augmentait la
génération de produits d'hydratation, ce qui augmentait principalement la résistance à la contrainte de cisaillement entre
les particules de résidus. L'amélioration globale la plus importante de la résistance des nano CPB a été observée dans
les échantillons contenant une concentration fixe de SP (0,125 %) et de NP (1 %).

1 INTRODUCTION extraction and refinement. The most prevalent method for

storing mine waste, surface tailings disposal, presents
The mining industry serves as a vital driving force in the multiple potential hazards to the environment. Firstly,
economic progress of both developed and developing owing to the high water content in tailings and construction
nations. Over the past few decades, there has been a challenges, conventional tailings impoundments are
remarkable surge in the demand for essential minerals, vulnerable to geotechnical failures, such as liquefaction
prompting an increased extraction of ore bodies from the failure, dam collapse, and foundation failure. These failures
earth's crust. Mining operations generate a considerable frequently lead to severe consequences, including human
volume of tailings during the processes of mineral casualties, financial losses, and far-reaching social and
environmental impacts. Secondly, from an ecological Greater strength during CPB's early curing stages and
standpoint, the surface disposal of tailings heightens the reduced time between mining cycles contribute to
risk of environmental contamination by exposing enhanced mine productivity.
hazardous constituents, like heavy metals and acidic One practical approach to decrease the time needed
compounds, to the surrounding environment. for higher strength in the early stages of curing is to
Consequently, it is crucial to explore and adopt alternative increase the binder/cement content in the CPB, which
techniques for managing mine waste more effectively and results in more cement hydration products and accelerates
sustainably (Ghirian and Fall 2016, Yilmaz and Fall 2017, the self-desiccation process (Fall et al. 2008). However,
Haruna and Fall 2020b). this approach has drawbacks: increasing binder content
Backfilling has emerged as a more eco-conscious raises costs, as the binder can account for up to 75% of the
approach to tailings management and has been widely final CPB production cost, and using more cement
adopted in underground mines worldwide. This process contributes to carbon dioxide emissions, raising
entails the return of a substantial portion of mine waste, environmental concerns (Fall and Benzaazoua 2003).
such as tailings, to underground mine stopes. In recent Therefore, it is crucial to find an alternative method to boost
times, cemented paste backfilling has gained recognition the CPB strength increase rate.
as an innovative technique within the realm of mine Numerous studies have been conducted to enhance
backfilling. Cemented paste backfill (CPB) typically the strength and mechanical properties of cement during
consists of a blend of tailings (75-85 wt %), hydraulic binder early curing stages using chemical admixtures or additives,
(3-9 wt % of total dry paste weight), and water, offering a as cement is a widely-used construction material. With
sustainable solution to address the environmental significant advancements in nanotechnology,
concerns associated with tailings disposal (Fall and nanomaterials have been employed to improve concrete
Benzaazoua 2005, Sivakugan et al. 2006). The mechanical properties in construction. Research has shown that
resilience of CPB is fundamentally tied to its structural adding Nanoparticle additives (NP) increases the
stability. A CPB with greater mechanical strength ensures compressive strength of cement mortar (Ltifi et al. 2011,
improved ground support and enhances the safety of Nazari and Riahi 2011, Oltulu and Şahin 2013). However,
subterranean mining operations. Thus, the pace at which the high potential for agglomeration in nano-scale materials
CPB develops mechanical strength is of equal importance, necessitates the use of a suitable dispersing agent such as
as it dictates the effectiveness and productivity of the the superplasticizer.
mining procedure. A quicker strength-gaining rate enables Superplasticizers, when added to the CPB mixture,
mine workers to return to the backfilled areas promptly after have a significant impact on its performance, leading to
a stope is filled with CPB, optimizing work schedules and substantial enhancements in various properties. They
mitigating potential holdups that can impact the mine's considerably reduce the yield stress and viscosity of CPB,
financial efficiency (Yilmaz and Fall 2017, Roshani and Fall regardless of the tailings type used, and increase the
2020b). unconfined compressive strength (UCS) of the CPB
Additionally, an accelerated strength-gaining rate (Haruna and Fall 2020a, 2020b, Al-Moselly et al. 2022).
contributes to more efficient mine tailings management and Additionally, superplasticizers contribute to better
promotes eco-friendly mining practices by guaranteeing environmental performance by reducing the hydraulic
timely and safe containment of tailings materials. The conductivity and reactivity of the CPB (Lothenbach et al.
uniaxial compressive strength (UCS) of CPB is typically n.d.).
used to measure mechanical stability. Once CPB is Despite studies on adding NP additives to cement, little
introduced into the stopes, mining activities in adjacent research has been conducted on their use in CPB or on the
areas are suspended until the CPB achieves the necessary effects of nano-additives on CPB's mechanical
strength for stability. As a result, the rate of CPB strength performance. This experimental study examines the
development bears a substantial influence on mining impact of different dosages of four types of nanoparticles
productivity. (NPs), including nano-silica (SiO2), nano-calcium
To enhance mining effectiveness, it is crucial to carbonate (CaCO3), nano-iron oxide (Fe2O3), and nano-
decrease the curing time of CPB. In other words, during the aluminum oxide (Al2O3), on the mechanical strength of the
early stages of curing, strength should develop as rapidly CPB.
as feasible. Moreover, the structural integrity of CPB is of
utmost importance, as a collapse may lead to serious 2 EXPERIMENTAL PROGRAM
repercussions, including injuries or fatalities among
workers and damage to machinery. Consequently, the rate 2.1 Materials
at which CPB strength increases is an essential factor in
design considerations. 2.1.1 Tailings
Self-desiccation, which occurs due to water
consumption during cement hydration, is a key In this study, synthetic tailings called silica tailings (ST) are
performance factor in CPB (Li and Fall 2018). This utilized. Comprising 99.8% quartz, ST is the primary
phenomenon leads to a reduction in pore water pressure mineral found in tailings from Canadian hard rock mines
and/or the creation of matric suction in the CPB, positively (Fall et al. 2010). Employing STs aims to minimize material
impacting early-age strength development by increasing variability and uncertainties in test outcomes due to the
effective stress within the backfill structure. As a result, presence of reactive minerals in natural tailings. These
self-desiccation lowers the chances of CPB liquefaction. minerals may react with the binder during CPB hydration,
complicating the interpretation and analysis of test findings. Specific surface
The STs used in this research have a particle size area (m2/g) 40-80
distribution nearly identical to that of natural tailings, based Nano-Fe2O3
on tailings from nine hard rock mines in Eastern Canada, Appearance Red brown
Fe2O3 ≥98
as depicted in Figure 1. Table 1 presents the physical Particle size (nm) 20-40
properties and mineral composition of the tailings Specific surface
employed in this experiment. area (m2/g) 40–60
Nano-Al2O3
100 Appearance White powder
90
Al2O3 ≥99
ST 9 Mines Particle size (nm) <50
80
Specific surface
70 >40
area (m2/g)
60
Finer percent (%)

50 2.1.3 Cement and mixing waters

40
30 In this investigation, the fundamental binding agent
20 employed is Portland cement Type I (PCI), which is the
10 most prevalent cement type utilized in CPB. Table 3
0 outlines the chemical and physical attributes of the PCI.
0.01 0.1 1 10 100 1000
Particle size (µm)
Table 3. Chemical and physical properties of PCI
Figure 1. Grain size distribution of the silica tailings (ST)
and average grain size distribution of tailings from nine
mines in Eastern Canada Element PCI (%)

Table 1. Physical properties of tailings SO3 3.82

Fe2O3 2.70
Properties Gs D10 D30 D50 D60
Unit - µm µm µm µm Al2O3 4.53
ST 2.7 1.9 9.0 22.5 31.5
SiO2 18.03
Average of 9 types
2.9 1.8 9.1 20.0 30.8
of tailings CaO 62.82
Gs: specific gravity
MgO 2.65
2.1.2 Nanoparticles
(Relative Density) 3.1
In this investigation, four varieties of NP additives are
employed to prepare the requisite CPB specimens. These 2.1.4 Superplasticizer
NPs comprise Nano-Calcium carbonate (CaCO3), Nano-
Iron oxide (Fe2O3), Nano-Aluminum oxide (Al2O3), and Roshani and Fall (2020a) have demonstrated that
Nano-Silica (SiO2). Pertaining to Nano-silica, Tetraethyl- incorporating NP additives into CPB significantly
Orthosilicate (TEOS), the manufacturing precursor of diminishes its flowability by increasing both the yield stress
nano-SiO2, was utilized. Dow Corning, Inc. provided the and viscosity of the mixture due to various processes, such
TEOS under the brand name XIAMETER® OFS-6697 as the generation of more cement hydration products and
silane. TEOS was employed as a colloid, resulting in increased water demand. Furthermore, higher water
improved nano-SiO2 particle dispersion within the CPB consumption has been observed as a result of the
mixture (Thomas et al. 2009). Table 2 delineates the production of more cement hydration products like calcium
physical and chemical properties of the NP additives. silicate hydrate (C-S-H). NP additives also tend to form
agglomerated particles, which obstruct the cement
Table 2. Chemical and physical specifications of hydration process (Shaikh and Supit 2016) and
nanoparticles consequently impact the mechanical performance of CPB.
To mitigate these issues, several samples in this study are
Nano-SiO2 prepared with the addition of a superplasticizer (SP), as it
Appearance Clear liquid has been proven to enhance the flowability of CPB by
SiO2 (%) >99 reducing yield stress (Roshani and Fall 2020a) and
Particle size (nm) <12 promoting NP dispersion. During the CPB sample
Specific surface
area (m2/g)
350 preparation, multiple sets of samples, including a control
Nano-CaCO3 set (pure CPB without NP additives) and various nano-CPB
Appearance White powder sets, were created using the Master Glenium 7500
CaCO3 (%) ≥98 concrete admixture (Master G; Badische Andilin und Soda
Particle size (nm) 20–50 Fabrik (BASF)) as the SPF.
2.2 Sample preparation removed from the molds after designated curing periods
(12 hours, and 1, 3, 7, and 28 days), and placed in a
Various nano-CPB sample groups were created to mechanical press that applied a vertical load at a rate of
examine the impact of different NP additives on CPB 1.0 mm/min. The peak stress at which the specimens
properties when combined with distinct binders. Table 4 collapse can be gauged by documenting the compressive
provides more information on the CPB mix design. To stress exerted on them. Every test was performed on at
make the CPB samples, tailings and binder were mixed least two matching specimens to guarantee the
together for 3 minutes, resulting in a consistent dry blend consistency of the test outcomes.
of cement and tailings. However, introducing NP additives
straight into the mixer led to weak, unhydrated NPs 2.3.2 Microstructural analysis
because of agglomeration, which filled the gaps between
CPB particles and diminished the CPB's mechanical To gain deeper insights into the microstructural changes in
strength (Senff et al. 2009, Quercia et al. 2012). This issue the CPB due to cement hydration in the presence of NP
is due to the high reactivity of NPs and strong Van der additives and varying curing conditions, several
Waals forces between them, making them more prone to microstructural tests were performed. These tests included
agglomeration than other pozzolanic materials. thermal analysis (thermogravimetry (TG), differential
To address this issue, the additives were initially thermogravimetry (DTG)), mercury intrusion porosimetry
combined with water and mixed for a minimum of five (MIP), and XRD, and were conducted on the CPB or
minutes, after which the solution was added to the dry cement paste of CPB, either with or without NP additives.
tailings-cement mix and stirred for another 15 minutes. For For the XRD and TG/DTG tests, cement paste samples
samples containing SP, water and SP were mixed first for (with a water-to-cement ratio of 1), prepared similarly to the
one minute, then NP additives were introduced and mixed CPB samples mentioned earlier, were thoroughly dried in
for an additional three minutes. This mixture was then an oven at 45°C for 4 days to evaporate any free water.
added to the tailings and binder. After thorough mixing, the These samples were then ground into powder. TG/DTG
fresh CPB was poured into cylindrical molds measuring 5 analyses were carried out using a SDT 2960 simultaneous
cm in diameter and 10 cm in height for UCS testing at DSC-TGA thermogravimetric analyzer. The samples were
various curing periods. To eliminate any air bubbles within heated at a rate of 10°C/min up to a maximum temperature
the CPB, samples were agitated for a minute after being of 1000°C in a nitrogen (N2) environment at a flow rate of
poured into the molds. Finally, they were stored in a 100 mL/min.
temperature-regulated container until testing took place. The XRD tests were conducted using a Rigaku Ultima
IV diffractometer, equipped with cross beam optics, at a
Table 4. Summary of the mix composition of the prepared beam voltage of 40 kV and current of 44 mA. MIP tests
nano-CPB specimens were carried out on dried CPB samples using a
Micromeritics AutoPore III 9420 mercury porosimeter.
PCI
SP NP
Sample in the NP 2.3.3 Monitoring program
content contents
name binder type
(%) (%)
(%)
To understand the interaction between the NP and the
Effect of curing time and nano-particle content
CPB 100 0 0 None
binder, which is key to altering the CPB's mechanical
CPB-NS 100 0 1 NS properties, a 5TE electrical conductivity (EC) sensor was
CPB-NC 100 0 1 NC utilized. This sensor is capable of monitoring EC,
CPB-NF 100 0 1 NF temperature, and volumetric water content (VWC) with
CPB-NA 100 0 1 NA accuracies of ±10% for EC, ±3% for VWC, and ±1°C for
Effect of curing time and SP content temperature measurements. This sensor was positioned at
CPB 100 0.125 0 None the heart of each CPB sample and connected to a data
CPB-NS 100 0.125 1 NS logger to log the collected data. EC was assessed by
CPB-NC 100 0.125 1 NC applying an alternating current between two electrodes
CPB-NF 100 0.125 1 NF
within the sensor and gauging the electrical resistance
CPB-NA 100 0.125 1 NA
CPB 100 0.25 0 None between them. Changes in EC in a cement-based material
CPB-NS 100 0.25 1 NS serve as a useful measure of ion movement through the
CPB-NC 100 0.25 1 NC material due to the chemical evolution of cement hydration.
CPB-NF 100 0.25 1 NF
CPB-NA 100 0.25 1 NA
NS: Nano-silica particles; NC: Nano-CaCO3 particles; NF: 3 RESULTS AND DISCUSSION
Nano-Fe2O3 particles; and NA: Nano-Al2O3 particles.
3.1 Effect of nanoparticle additive on strength
2.3 Test methods and monitoring program development of Nano-CPB

2.3.1 Unconfined compressive strength test Figure 2 depicts how the strength of CPB samples with a
1% NP additive, without the use of any dispersing agent
Each CPB sample's compressive strength was measured (SP), evolves over the curing period. It is observed that all
following ASTM – C39 standards. Specimens were the samples consistently show an upward trend in strength
as the curing time increases, a phenomenon attributed to strength at 565 kPa. This might be attributed to the
the ongoing process of cement hydration. clustering of NPs.
The continuous increase in strength is attributable to Compared to larger-sized particles of the same
three fundamental elements. First, the ongoing process of material, interparticle forces such as van der Waals and
cement hydration throughout the curing period results in electrostatic forces primarily influence the NPs, while
the production of more hydration products, as depicted in gravity is the principal force acting on the same material
Figure 4 and previously discussed. This, in turn, enhances with a larger particle size. As a result, the attractive forces
the cementation, or bonding, between the tailings particles, of the electric charges on the NPs' surface hinder their
leading to an increase in the uniaxial compressive strength uniform dispersion, causing the formation of aggregated
(UCS) of the CPB. The second aspect, as visualized in particles within the mixture.
Figure 6, involves the utilization of pore water due to
cement hydration, which causes a reduction in water
content. The graph depicts how the ongoing hydration 0.40
process consistently diminishes the CPB sample's 0.35
volumetric water content throughout the curing period. A

Cumulative porosity (%)

lower water content is typically linked to increased strength 0.30
in porous or cementitious materials. 0.25
0.20
800
0.15
700
0.10
600
0.05
500
UCS (kPa)

0.00
400
0.01 0.1 1 10 100 1000
300 Pore size (µm)
200 CPB- 7 days CPB- 28 days

100 Figure 3. Changes in total pore refinement of CPB with

0 curing time
0 7 14 21 28
Time (days) It should be noted that improper dispersion of NPs
CPB (control) CPB-1%CaCO3 CPB-1%SiO2
negatively impacts the mechanical properties of
CPB-1%Fe2O3 CPB-1%Al2O3 cementitious materials or CPB in two distinct ways. Firstly,
the agglomerated NPs exhibit a much lower reactivity rate
Figure 2. Development of compressive strength of CPB than individual NPs due to the reduced specific surface
samples with different types of NP additives and without area (SSA) of the aggregated particles; hence, fewer NPs
SP partake in cement hydration, leading to lesser hydration
products at the end of the curing process. Secondly,
Lastly, the refinement of the pore structure, prompted without uniform dispersion of NPs, the filler effect of NPs
by the aforementioned mechanisms (such as changes in that typically boosts the packing density by filling the
water content and the precipitation of hydration products), micropores in the mixture becomes unattainable.
transforms the initial soft state of the CPB into a dense and In order to address the issue of NP agglomeration
hardened material with a reduced pore size and void ratio. affecting the compressive strength of the nano-CPB
The Mercury Intrusion Porosimetry (MIP) tests, performed samples, it's necessary to incorporate a dispersing agent
on two identical CPB samples that underwent different into the nano-CPB mixture. Polycarboxylate-based
curing durations (7 and 28 days), substantiate this pore superplasticizers (SPs) are found to be the most effective
refinement. As depicted in Figure 3, the CPB sample that at dispersing NPs, as per Kim et al. (2018) who
underwent a 28-day curing period exhibits a lower investigated the effects of various SP types on
cumulative pore volume (or reduced void ratio) compared nanomaterial dispersion in cementitious materials.
to the sample cured for 7 days. This is due to an increased Consequently, the second batch of nano-CPB samples,
precipitation of hydration products filling up the pores. The prepared with 1% NP additives, were treated with a 0.125%
outcomes from the uniaxial compressive strength (UCS) dosage of SP.
tests reveal that, except for the initial 24-hour period where This inclusion of SP improves the distribution of NPs within
the strength of backfill samples with and without the CPB mixture, which ultimately refines the pore structure
nanoparticle (NP) additives are relatively comparable, the of the hardened nano-CPB. This refinement is
strength of the samples with NP additives is consistently corroborated by the results of the MIP tests conducted on
lower than that of the ones without for all curing durations. two different nano-CPB samples with varying SP dosages,
For instance, after a 28-day period, the control sample's as shown in Figure 4.
UCS is the highest recorded strength at 680 kPa, while the Compared to the nano-sample prepared with SP, depicted
sample with 1% nano-Al2O3 additive presents the lowest in Figure 4, the nano-CPB sample prepared without SP
demonstrates a higher total pore volume and a larger pore
size. This further supports the observation of NP 2000
agglomerates that fail to fill the pores. Conversely, the
samples prepared with SP exhibit smaller pores, affirming 1500
that improved NP dispersion enables the NPs to effectively

UCS (kPa)
fill the pores. 1000

0.5 500
Cumulative porosity (%)

0.4
0
0.3 0 5 10 15 20 25 30
0.2 Time (days)
0.1 CPB-0.125SP CPB-1%CaCO3-0.125SP
CPB-1%SiO2-0.125SP CPB-1%Fe2O3-0.125SP
0 CPB-1%Al2O3-0.125SP
0.01 0.1 1 10 100 1000
Figure 5. Strength development of nano-CPB samples
Pore size (µm) with 0.125% SP

CPB-1%CaCO3 CPB-1%CaCO3-0.125SP The strength augmentation induced by the NP additive

can be attributed to two primary mechanisms. Firstly, the
Figure 4. Effect of SP on the pore structure of nano-CaCO3- NP additives serve as fillers, which contribute to a denser
CPB packing of the final backfill material. Consequently, a
reduced void ratio and denser microstructure are achieved,
The outcomes from the uniaxial compressive strength which naturally translate into heightened strength of the
(UCS) tests for the nano-Cemented Paste Backfill (CPB) backfill material. The results of the MIP test conducted on
samples treated with 0.125% superplasticizer (SP) are the nano-CPB samples support this conclusion,
presented in Figure 5. Consistent with expectations, the showcasing a refined pore structure of the CPB as shown
samples prepared with both 0.125% SP and 1% in Figure 6.
nanoparticle (NP) additive demonstrate superior strength The second mechanism concerns the active
compared to the samples devoid of SP, as represented in contribution of NP additives to the hydration reaction, which
Figure 5. This enhancement can be attributed to the accelerates the nucleation process during hydration,
dispersion effect of the repulsive forces acting on the solid leading to the generation of more hydration products.
particles of the CPB and NPs, an effect introduced by the Roshani and Fall (2020b) asserted that the introduction of
addition of SP. This process not only facilitates the active NPs provides numerous nucleation sites, promoting the
participation of NPs in the hydration reaction but also consumption of primary cement components such as C3S,
ensures their uniform distribution throughout the C2S, C3A, and C4AF.
cementitious medium. These observations are further However, despite these enhancements, the nano-CPB
supported by the results of the MIP tests depicted in Figure sample modified with nano-Al2O3 doesn't demonstrate any
4, which validate the efficient dispersion of NPs within the significant strength improvement compared to the control
nano-CPB samples treated with SP. This efficient sample, as shown in Figure 5.
dispersion, in turn, refines the pore structure of the Ultimately, this comprehensive analysis reaffirms the
hardened nano-CPB, ultimately leading to enhanced vital role of NP additives in enhancing the strength and
compressive strength. overall performance of cementitious materials, but also
The progression of strength increase in the samples underscores the need for careful selection and application
depicted in Figure 5 reveals that CPB samples of these additives, given their varying effects on material
incorporating nano-CaCO3, nano-Fe2O3, and nano-SiO2 properties.
exhibit enhanced strength compared to the control sample, As observed in Figure 5, the introduction of NP
while the strength of CPB adjusted with nano-Al2O3 is akin additives counteracts the negative impact of SP, which
to the control sample. The positive impact of nano-CaCO3 hinders early-stage strength development in CPB samples.
on the strength of CPB with SP aligns with the findings of Haruna and Fall (2020a) study found that SP addition
numerous studies investigating the effects of nano-CaCO3 slows the hydration process, thus impeding early-stage
on concrete and/or mortar, such as those conducted by strength development in CPB. Electrical conductivity (EC)
Kadri and Duval (2002), and Hefni and Ali (2021). These monitoring of the investigated nano-CPB samples provided
studies explored the effects of partially substituting cement additional evidence that NP additives can enhance cement
with nano-CaCO3 and highlighted the potential hydration, compensating for SP's slowing effect. Variations
enhancement of mechanical performance in concrete and in ion concentration in the backfill paste's pore water due
cement paste. to cement hydration can be tracked by observing changes
in EC values. Levita et al. (2000) stated that the initial EC
peak in cement-related substances signifies the onset of
the cement paste's solid phase transition. Therefore, a
shorter time to the first EC peak suggests potential for
greater early age strength. Furthermore, the size of the
capillary pores in a porous medium directly influences the
medium's EC, as it limits ion flow. It can be argued that
increasing the backfill microstructure's complexity by filling
pores with hydration products reduces the paste backfill's
EC. As a result, lower EC values indicate a CPB with a
reduced void ratio or increased packing density.

0.4
Cumulative porosity (%)

0.3

0.2

0.1 Figure 8. Effect of addition of nano-CaCO3 on the XRD

analysis result on 7 day cement paste.
0
0.01 0.1 1 10 100 1000
4 SUMMARY AND CONCLUSION
Pore size (µm)
CPB-0.125SP CPB-1%CaCO3-0.125SP This study investigates the impact of incorporating various
CPB-1%SiO2-0.125SP CPB-1%Fe2O3-0.125SP types of nanoparticle (NP) additives on the strength
CPB-1%Al2O3-0.125SP development of cemented paste backfill (CPB) which is a
Figure 6. Impact of NPs on pore size distribution of nano- key early-stage performance indicator. A polycarboxylate
CPB samples mixed with 0.125% of SP. ether-based superplasticizer (SP) is employed as a
dispersion agent to facilitate efficient nanoparticle
Figure 7 shows that nano-CPB with CaCO3 reaches its distribution. Various microstructural examinations and
first peak sooner than other nano-CPB and control monitoring schemes are executed to clarify the results
samples, which aligns with the highest strength noted for obtained in the Unconfined Compressive Strength (UCS)
CPB with CaCO3 in the UCS test. Moreover, after 11 days, tests performed on CPB samples mixed with NP additives.
nano-CPB with CaCO3 displays the lowest EC among all The salient findings of this research, informed by the
samples, which can be considered as an indication for experimental results, are summarized below:
increased production of hydration products. • The strength development rate in the CPB consistently
The XRD analysis demonstrates an increase in increase over the duration of the curing cycle, irrespective
hydration products due to the inclusion of NPs, as shown of the inclusion of NP additives.
in Figure 12. This analysis was conducted on cement paste • The mechanical performance of CPB can be either
both with and without nano-CaCO3. It is evident that when positively enhanced or negatively affected by the addition
mixed with nano-CaCO3 and cured for 7 days, the cement of NP additives, contingent on the existence of an effective
paste exhibits more hydration products than the control dispersing agent within the CPB mixture.
sample cured under the same conditions. For instance, the • CPB samples properly mixed with nanoparticle additives,
intensity (e.g., CPS) of C-S-H in the cement paste mixed such as nano-CaCO3, nano-Fe2O3, and nano-SiO2,
with CaCO3 is 367 at 32.16° at 2θ, while it is 363 at 2θ for demonstrate superior strength compared to both the
the control sample. Similarly, the intensity of CH for the control sample and the sample mixed with nano-Al2O3.
sample with nano-CaCO3 is 1993 at 18.06°, compared to • Among the NP additives evaluated, nano-CaCO3 exhibits
1862 at 2θ for the control sample. the most beneficial influence on the mechanical properties
of CPB, attributed to the development of a denser
microstructure and increased bonding products. As a
result, the highest UCS values are achieved compared to
other NP additives and control samples.

5 REFRENCES

Al-Moselly, Z., Fall, M., and Haruna, S. 2022. Further

insight into the strength development of cemented
paste backfill materials containing polycarboxylate
ether-based superplasticizer. Journal of Building
Engineering, 47: 103859. Elsevier.
doi:10.1016/J.JOBE.2021.103859.
Fall, M., and Benzaazoua, M. 2003. Advances in predicting
performance properties and cost of paste backfill.
Figure 7. Changes in EC of nano-CPB samples prepared
Tailings and Mine Waste,: 73–85.
with 1% NPs and 0.125% of SP
Fall, M., and Benzaazoua, M. 2005. Modeling the effect of undefined. (n.d.). The influence of superplasticizers
sulphate on strength development of paste backfill on the hydration of Portland cement.
and binder mixture optimization. Cement and researchgate.net,.
Concrete Research,. Ltifi, Guefrech, A., Mounanga, P., and Khelidj, A. 2011.
doi:10.1016/j.cemconres.2004.05.020. Experimental study of the effect of addition of nano-
Fall, M., Benzaazoua, M., and Saa, E.G. 2008. Mix silica on the behaviour of cement mortars Mounir.
proportioning of underground cemented tailings Procedia Engineering, 10: 900–905.
backfill. Tunnelling and Underground Space doi:https://doi.org/10.1016/j.proeng.2011.04.148.
Technology, 23(1): 80–90. Pergamon. Nazari, A., and Riahi, S. 2011. Improvement compressive
doi:10.1016/j.tust.2006.08.005. strength of concrete in different curing media by
Fall, M., Célestin, J.C., Pokharel, M., and Touré, M. 2010. Al2O3 nanoparticles. Materials Science and
A contribution to understanding the effects of curing Engineering: A, 528(3): 1183–1191.
temperature on the mechanical properties of mine doi:https://doi.org/10.1016/j.msea.2010.09.098.
cemented tailings backfill. Engineering Geology, Oltulu, M., and Şahin, R. 2013. Effect of nano-SiO2, nano-
114(3–4): 397–413. Elsevier. Al2O3 and nano-Fe2O3 powders on compressive
doi:10.1016/J.ENGGEO.2010.05.016. strengths and capillary water absorption of cement
Ghirian, A., and Fall, M. 2016. Strength evolution and mortar containing fly ash: A comparative study.
deformation behaviour of cemented paste backfill at Energy and Buildings, 58: 292–301.
early ages: Effect of curing stress, filling strategy and doi:https://doi.org/10.1016/j.enbuild.2012.12.014.
drainage. International Journal of Mining Science Quercia, G., Hüsken, G., and Brouwers, H.J.H. 2012.
and Technology, 26(5): 809–817. Elsevier. Water demand of amorphous nano silica and its
doi:10.1016/J.IJMST.2016.05.039. impact on the workability of cement paste. Cement
Haruna, S., and Fall, M. 2020a. Strength development of and Concrete Research, 42(2): 344–357. Pergamon.
cemented tailings materials containing doi:10.1016/J.CEMCONRES.2011.10.008.
polycarboxylate ether-based superplasticizer: Roshani, A., and Fall, M. 2020a. Flow ability of cemented
experimental results on the effect of time and pastefill material that contains nano-silica particles.
temperature. https://doi.org/10.1139/cjce-2019- Powder Technology, 373: 289–300.
0809, 48(4): 429–442. NRC Research Press 1840 doi:https://doi.org/10.1016/j.powtec.2020.06.050.
Woodward Drive, Suite 1, Ottawa, ON K2C 0P7 . Roshani, A., and Fall, M. 2020b. Rheological properties of
doi:10.1139/CJCE-2019-0809. cemented paste backfill with nano-silica: Link to
Haruna, S., and Fall, M. 2020b. Time- and temperature- curing temperature. Cement and Concrete
dependent rheological properties of cemented paste Composites, 114. Elsevier Ltd.
backfill that contains superplasticizer. Powder doi:10.1016/J.CEMCONCOMP.2020.103785.
Technology, 360: 731–740. Elsevier. Senff, L., Labrincha, J.A., Ferreira, V.M., Hotza, D., and
doi:10.1016/J.POWTEC.2019.09.025. Repette, W.L. 2009. Effect of nano-silica on rheology
Hefni, M., and Ali, M.A. 2021. The Potential to Replace and fresh properties of cement pastes and mortars.
Cement with Nano-Calcium Carbonate and Natural Construction and Building Materials, 23(7): 2487–
Pozzolans in Cemented Mine Backfill. 2491.
doi:10.1155/2021/5574761. doi:https://doi.org/10.1016/j.conbuildmat.2009.02.00
Kadri, E.H., and Duval, R. 2002. Effect of Ultrafine Particles 5.
on Heat of Hydration of Cement Mortars. Materials Shaikh, F.U.A., and Supit, S.W.M. 2016. Effects of
Journal, 99(2): 138–142. doi:10.14359/11705. Superplasticizer Types and Mixing Methods of
Kim, G.M., Nam, I.W., Yoon, H.N., and Lee, H.K. 2018. Nanoparticles on Compressive Strengths of Cement
Effect of superplasticizer type and siliceous materials Pastes. Journal of Materials in Civil Engineering,
on the dispersion of carbon nanotube in cementitious 28(2): 06015008. American Society of Civil
composites. Composite Structures, 185: 264–272. Engineers (ASCE). doi:10.1061/(ASCE)MT.1943-
Elsevier Ltd. 5533.0001373.
doi:10.1016/J.COMPSTRUCT.2017.11.011. Sivakugan, N., Rankine, R.M., Rankine, K.J., and Rankine,
Levita, G., Marchetti, A., Gallone, G., Princigallo, A., and K.S. 2006. Geotechnical considerations in mine
Guerrini, G.L. 2000. Electrical properties of fluidified backfilling in Australia. Journal of Cleaner
Portland cement mixes in the early stage of Production,. doi:10.1016/j.jclepro.2004.06.007.
hydration. Cement and Concrete Research, 30(6): Thomas, J.J., Jennings, H.M., and Chen, J.J. 2009.
923–930. Pergamon. doi:10.1016/S0008- Influence of Nucleation Seeding on the Hydration
8846(00)00282-9. Mechanisms of Tricalcium Silicate and Cement. The
Li, W., and Fall, M. 2018. Strength and self-desiccation of Journal of Physical Chemistry C, 113(11): 4327–
slag-cemented paste backfill at early ages: Link to 4334. American Chemical Society.
initial sulphate concentration. Cement and Concrete doi:10.1021/jp809811w.
Composites, 89: 160–168. Yilmaz, E., and Fall, M. 2017. Paste tailings management.
doi:https://doi.org/10.1016/j.cemconcomp.2017.09.0 In Paste Tailings Management.
19.
Lothenbach, B., Winnefeld, F., -, R.F., Dübendorf,
undefined, Switzerland, undefined, and 2007,
Wednesday, October 4, 2023

COLD REGIONS
Monitoring of permafrost beneath a stream
crossing along the Inuvik-Tuktoyaktuk
Highway

N. Kozachuk1, C.Cooper1, G.N.Eichhorn2, R.Beddoe2, G.Siemens2,

A.Wilson3, J.Humphries4
1Facultyof Engineering, Queen’s University, Kingston, ON, Canada
2Department of Civil Engineering, Royal Military College of Canada, Kingston, ON, Canada
3Northwest Territories Geological Survey, Inuvik, NWT
4Aurora Research Institute, Inuvik, NWT

ABSTRACT
With over 40% of Canada’s landmass situated in regions of continuous permafrost, climate change and its effects pose a
major concern for northern regions. Warming air temperatures, one driver of permafrost degradation, can lead to
geohazards such as slope instability, soil subsidence, and ground movement, all of which increase the risk of infrastructure
failure. The Inuvik-Tuktoyaktuk Highway (ITH) in the Northwest Territories was opened for use in 2017 as the only year-
round transportation link to the Arctic Coast in Canada. Since its opening, significant maintenance has been required to
keep the road operational as a result of changes in the underlying permafrost. At kilometer 131 of the ITH, approximately
14km South of Tuktoyaktuk, the highway crosses Gunghi Creek. The crossing was originally outfitted with a corrugated
steel pipe (CSP) culvert, but following significant deformation the crossing was rebuilt in early 2021 with an open bottom
arch concrete structure, founded on 20 adfreeze piles. A ground temperature monitoring program was implemented as
part of the rebuild to monitor the performance of the new structure. This has consisted of thermistor strings deployed within
the piles, along the road centreline, within the embankment, and other areas of interest. Additional climate data is being
collected including temperature, wind, precipitation, and snow depth to assist in numerical modelling to help predict future
performance of the structure and others like it.

RÉSUMÉ

Avec plus de 40% de la superficie du Canada située dans des régions de pergélisol continu, les changements climatiques
et leurs effets posent une préoccupation majeure pour les régions du Nord. Les températures de l'air en hausse, l'un des
facteurs de la dégradation du pergélisol, peuvent entraîner des géorisques tels que l'instabilité des pentes, l'affaissement
des sols et le mouvement du sol, tous augmentant le risque de défaillance des infrastructures. La route Inuvik-Tuktoyaktuk
(ITH) dans les Territoires du Nord-Ouest a été ouverte à la circulation en 2017 en tant que seul lien de transport toute
l'année vers la côte arctique au Canada. Depuis son ouverture, des travaux d'entretien importants ont été nécessaires
pour maintenir la route opérationnelle en raison des changements dans le pergélisol sous-jacent. Au kilomètre 131 de
l'ITH, à environ 14 km au sud de Tuktoyaktuk, la route traverse la crique de Gunghi. Le passage était à l'origine équipé
d'un tuyau de tôle ondulée (TTO), mais suite à une déformation importante, le passage a été reconstruit début 2021 avec
un pont-canal à fond ouvert, fondé sur 20 pieux adhérents. Un programme de surveillance de la température du sol a été
mis en place dans le cadre de la reconstruction pour surveiller la performance de la nouvelle structure. Cela consiste en
des chaînes de thermistors déployées dans les pieux, le long de l'axe central de la route, dans l'empilement et d'autres
zones d'intérêt. Des données climatiques supplémentaires sont collectées, notamment la température, le vent, les
précipitations et la hauteur de neige, pour aider à la modélisation numérique visant à prédire la performance future de la
structure et d'autres similaires.

1 INTRODUCTION Territories, Canada, it is estimated that the permafrost

damages to infrastructure will be over $1.3 billion over the
Climate change and its effects pose a major concern for next 75 years, with annual economic losses estimated at
northern infrastructure situated on permafrost (e.g Hjort et $51 million (NWT Association of Communities, 2017). One
al. 2022). The Canadian Arctic has been more susceptible example of this type of infrastructure is at Gunghi Creek
to climate change trends, with air temperatures warming at along the Inuvik-Tuktoyaktuk Highway (ITH) in the
more than twice the global rate (Bush et al., 2022) which in Northwest Territories. Gunghi Creek is a stream crossing
turn drives permafrost degradation. This degradation can 14 km south of Tuktoyaktuk and recently had its corrugated
lead to slope instability, soil subsidence, and ground culvert replaced with an arch founded 20 m piles. These
movement, all which increase the risk for weakening were designed as AdFreeze piles which rely on the
infrastructure integrity (WSP, 2021). In the Northwest interface friction and bond created between the steel pile
and adjacent frozen ground. A rigid buried arch design has
not been used in permafrost regions extensively, and it is
unclear how the performance of this piece of infrastructure
will perform over its design life. As such, an extensive
monitoring program is underway to evaluate the pile
performance through temperature monitoring, as this has
been proposed as an indicator of pile-structure interface
health (Stantec, 2018). This paper outlines the installed
instrumentation plan which includes thermal monitoring in
the subsurface along the AdFreeze piles, as well as along
the concrete arch , the embankment and road.

2 BACKGROUND INFORMATION

Before the ITH opened in 2017 as an all-season route

between Canada and the Arctic Ocean, ground access
between Inuvik and Tuktoyaktuk was via an ice-road
operational in the winter season only. In 2017 construction
was completed on the ITH, replacing the ice road, and
providing the only year-round ground transportation link to
the Arctic Coast in Canada. The 140 km two lane gravel
highway crosses over 8 major bridge crossings, 19 large
diameter corrugated steel pipe (CSP) culverts, and 340
standard stream size culverts (Government of Northwest
Territories, 2019). The ITH crossing at Gunghi Creek is Figure 2: Inuvik-Tuktoyaktuk Highway (ITH), with Gunghi
seen in Figure 1 in 2022. Creek Crossing at km 131.2.

2.1 Culvert replacement at Gunghi Creek

Following a hydrotechnical assessment, it was determined

that the CSP culvert at Gunghi Creek should be replaced
with a 7.3 m x 3.2 m open bottom buried concrete arch
(Stantec, 2018). The arch was constructed in winter 2020-
21, with a design life of 75 years, and a maximum allowable
settlement of 75 mm over this time. The arch was founded
on 40 AdFreeze piles, 20 on either side spaced
approximately 2 m apart. A photograph of the completed
concrete arch at Gunghi Creek is shown in Figure 3.

Figure 1: ITH Crossing at Gunghi Creek in 2022.

Gunghi Creek is located approximately 14 km south of

Tuktoyaktuk, at kilometer 131 of the ITH. A map of the local
area in the NWT showing the route of the ITH, and the
relative location of Gunghi Creek is shown in Figure 2. In
this area the permafrost active layer is estimated to be 3 m
deep beneath the creek crossing. The crossing was
originally outfitted with a 2 m diameter CSP culvert in 2013
as part of the ITH construction project (Kiggiak-EBA, 2019).
At this crossing, Gunghi Creek flows from the Northwest,
draining a 56 km2 watershed with significant summer flow Figure 3: New arch design, pictured during June 2022 field
(Stantec, 2018). Over time the culvert experienced trip.
significant differential settlement, resulting in deformations
causing problematic snow and ice blockages during winter. In anticipation of warming climate in the Canadian Arctic
By 2015, there was 800 mm of settlement at the centre of leading to degradation of the permafrost, the arch was
the culvert, at which point assessments began to determine installed with contingencies in place to ensure its long-term
an appropriate redesign of the culvert (Government of resilience. This included the installation of
Northwest Territories, 2020).
Figure 4: Arch and embankment thermal monitoring layout a) top down, and b) cut through the centre of the arch

thermosyphons, which are currently dormant but ready to Table 1: Bead depths for RST thermistor strings.
be activated in the future. These thermosyphons were
installed adjacent to the piles, and the radiators are located Bead Depth (m)
at each corner where the roadway intersects the arch. 1 0.5
When activated, they will extract heat from the ground 2 1
through the vaporization of carbon dioxide, which will then 3 1.5
rise to the surface and be cooled by the winter air before 4 2
5 3
condensing and cycling back into the ground. The
6 5
thermosiphons will provide passive cooling during the 7 8
winter months, and remain dormant in the summer 8 11
(Kiggiak-EBA, 2019). 9 14
10 18
3 THERMAL MONITORING
The embankment / arch thermistors were installed to
The construction and installation of the buried concrete capture the effects of the arch on the thermal conditions
arch at Gunghi Creek presented an excellent research within the embankment. This included thermistors, A1
opportunity for monitoring the short and long term ground (parallel to arch at crownline), and R1 to R3 (perpendicular
thermal regime. As such, a field monitoring program to arch at various depths below the road). These
consisting of thermistors, a weather station, snow depth thermistors were manufactured by LogR, with 8 thermistor
monitoring equipment, and a piezometer were installed at beads per string at varying spacings and a standard
the site. In total there are 15 thermistor strings, 8 of which accuracy of ±0.1° C. The data for the 7 LogR thermistor
were installed within the piles to a depth of 18 metres strings are being captured by 2 LogR ULog-C16-32
beneath the ground surface, the bead depths shown in loggers.
Table 1. These thermistors were manufactured by RST
Instruments, and are connected to RST DT2040 data Part of forecasting infrastructure resilience includes
loggers. Each thermistor string has 10 beads, and a numerical modelling of the climate-soil-infrastructure
standard accuracy of ± 0.1° C. interaction and performance. For permafrost, these
numerical models require inputs of soil thermal properties
Seven (7) additional thermistors were installed to capture and estimations of climate parameters. To aid in the
the thermal conditions surrounding the arch and beneath estimation of these inputs for future modelling purposes,
the road at several depths and orientations. A summary of additional instrumentation was installed at Gunghi Creek.
these non-standard thermistor layouts is summarized in The first of these additional instrumentation was a HOBO
Table 2, and the layout of these thermistors shown in the U30 climate station, equipped to measure temperature,
map view and cross-section of the crossing in Figure 4. wind speed and direction, precipitation, relative humidity,
and solar radiation. The climate station was attached to the
thermosiphon infrastructure immediately adjacent to the
roadway, on the southwest corner of the arch.
Table 1: Embankment / Arch thermistors

Name Location and length Number of beads Purpose

Hoop 1 (H1) Wraps around arch 8 Measure circumference
temperatures of arch

Arch 1 (A1) Along crown of arch 8 Temperature under road

centreline

Road 1 (R1) Horizontal under road, 8

perpendicular to base of arch Horizontal thermistors were
placed to capture the thermal
Road 2 (R2) Horizontal under road, halfway up 8 effect of the air within the arch
arch structure, and this affects the
thermal regime in the road.
Road 3 (R3) Horizontal under road, parallel to 8 (applies to R1, R2, & R3)
road

Radiator Pile (P1) Placed on a thermosyphon 10 Vertical thermal profile away from
the arch.

Embankment (E1) Adjacent to arch within the road 8 Capture temperatures under
embankment variable snow cover.

An additional factor required in numerical modelling of Additional instruments at Gunghi Creek include a vibrating
ground temperatures in cold regions is the snow thickness. wire piezometer installed in the thalweg of the creek on the
Seasonal snow cover has significant influence on the downstream side of the crossing, to monitor water
underlying thermal regime, as the snow acts as an temperature and depth. The presence of the stream
insulator, trapping heat (e.g. Zhang, 2005). A JUDD overlying the permafrost creates a warm source (relative to
Ultrasonic Distance Sensor was installed which transmits the seasonally frozen ground) that will impact the thermal
an ultrasonic pulse to the snow surface, outputting a snow response of the ground under the arch. Taken together, the
depth based on the transmission time. This sensor was water level height and the thermistor data aims to capture
installed on a tripod approximately 25 m adjacent to the the effect of creek on the pile performance. Over the longer
road surface. A second snow depth measurement term, this will aid in development of numerical models for
technique was also installed, provided by photographs of buried concrete arch structures with flowing water, and to
survey stakes by cellular connected ‘game’ cameras which evaluate the permafrost degradation in AdFreeze piles
transmit one photo per day. A photograph of the HOBO under these complex infrastructure / environmental
climate station and other thermal monitoring equipment is conditions.
shown attached to one of the thermosyphon stacks in
Figure 5 4 PRELIMINARY THERMAL MEASURMENTS

Since installation of the thermistor strings in the winter of

2021, there have been 3 instances of data collection.
While the data records are not all complete due to
equipment malfunctions and data acquisition challenges,
several preliminary findings and results can be inferred.
Shown in Figure 6 is the trumpet curve for the thermistor
located at the east end of the arch from winter 2021/22
through summer 2022. The active layer at this location
beneath the pile cap and creek extends to between 3-5
meters depth. The influence of seasonal variation in the
active layer in the permafrost zone extends to almost the
full depth of the instrumented piles. Significant variability in
sub-zero temperature fluctuations were observed to
approximately 2/3rds the length of the AdFreeze pile.
Owing to the latitude of this site, the permafrost base is not
encountered in the extent of the pile depth and
Figure 1: HOBO Climate station installed on instrumentation.
thermosiphon.

.
 Figure 7: Road thermistors R1, R2, and R3 for a single
reading in August 19, 2022.

The seasonal variation for another pile thermistor is shown

in Figure 8, which illustrates the temperature at the 10 bead
depths. These pile thermistors were designed to have
higher density in the near surface, to capture the active
layer. For this example, data collection of all 10 beads was
not initiated until June 2022. The time series shows large
fluctuations in the shallow zones and less variation at
Figure 2: Trumpet curve of temperatures for east end of depth, but still shows a 1 to 2 degree drop in temperature
arch AdFreeze pile(PS20) at the deepest (18 m) thermistors over the winter months
compared to summer.

The LogR thermistors installed to capture the road A compilation of 6 vertical thermistors along one row of the
infrastructure thermal response is more complex to AdFreeze piles is shown in Figure 9 and confirms that the
interpret owing to the different sources of warming / subsurface below the creek is frozen in the lateral direction
cooling, such as the arch. A selection of the road in all areas beneath the arch, and to the extent of the pile
thermistors R1, R2, and R3 and shown for one day in depth, with temperatures at the base of the piles of -8°C.
August 2022 in Figure 7. Each of the three instruments From the cross-section, preliminary observations show that
have bead number 1 located in the same location—in this the effect of the creek on the thermal regime of the
case very near the concrete arch. R1 and R2 are located subsurface is minimized under the centre of the road, but
perpendicular to the arch, with R2 located at a higher magnified at the arch portals where more solar energy
elevation, nearer to the road surface. R3 is oriented parallel reaches the ground. This was unexpected, as the presence
to the road, as shown in Figure 5. Interestingly, the three of a flowing stream was anticipated to have had a larger
strings show the thermal effect of the arch structure on the effect on the warming of the ground immediately beneath
lateral temperature gradient. Since these readings are the creek. At the centre of the road under the creek, the
shown for the warm summer, the beads closest to the arch active layer was found to reach a depth of approximately 2
are closer to the daytime air temperature but are likely metres below the pile cap (approximately the same
depressed owing to the soil burial. Moving laterally away elevation as the base of stream). However, at the portals
from the arch, the temperatures begin to drop at 3 m and 4 to the arch, the active layer reached a depth of 4 metres.
m for R1 and R2 respectively. Since R2 is elevated relative
to R1, it’s not surprising that this elevation is being warmed
both from the arch but also the proximity to the ground
surface. R3 shows an earlier temperature drop with
distance from the arch, which may reflect the different soil
conditions under the compacted road, compared to the
newer embankment with different soil conditions that
surround R1 and R1, at least at the beads farthest away
from the arch.
 Figure 8: Temperature string PS11, AdFreeze pile thermistor for April 2021 – Oct 2022

Figure 9: Thermal cross-section beneath arch, on August 19, 2022, with thermistors located on the south pile row at
locations PS1, 6, 11, 14, 17, and PS20
5 CONCLUSION NWT Association of Communities. (2017). The Technical
Opportunities & Economic Implications of Permafrost
The culvert replacement at Gunghi Creek along the ITH Decay on Public Infrastructure in the Northwest
has allowed a unique instrumentation opportunity to Territories. https://climatechange.toolkitnwtac.com/wp-
capture the thermal response of AdFreeze piles, not content/uploads/sites/21/2018/02/Permafrost-decay-
traditionally used for this type of road infrastructure. The Executive-Summary-with-map.pdf
AdFreeze thermistors have allowed the thermal response
of the subsurface to be studied, in combination with the Stantec. (2018). Gunghi Creek Hydrotechnical
streams thermal influence on the near surface warming / Assessment.
cooling. The arch concrete structure was instrumented with
a unique range of thermistors. The orientation of the arch WSP. (2021). Assessment of Climate Change Impacts on
thermistors aims to study the thermal influence of the Infrastructure in all NWT Communities (p. 61).
structure geometry, as well as the air gap created by the
arch. Preliminary results for the road thermistors shows
Zhang, T. (2005). Influence of the seasonal snow cover on
that the arch, which allows warm and cool air to flow under
the ground thermal regime: An overview. Reviews of
the road has a significant effect on the near surface and
Geophysics, 43(4).
soil temperatures. This effect is controlled by the proximity
https://doi.org/10.1029/2004RG000157
to arch as well as the ground surface. Gunghi Creek offers
a research test bed to capture both the thermal response,
as well as a suite of climate monitoring equipment. Future
work aims to link the climate conditions and the thermal
response of the infrastructure-subsurface interface.

6 ACKNOWLEDGMENTS

The authors gratefully acknowledge the on-site support

provided by the Aurora Research Institute. Funding for this
project was provided by the National Trades Corridors
Fund (“Climate change resilience study of transportation
infrastructure along the Inuvik-Tuktoyaktuk Highway”).

7 REFERENCES

Bush, E., Bonsal, B., Derksen, C., Flato, G., Fyfe, J., Gillett,
N., Greenan, B. J. W., James, T. S., Kirchmeier-Young,
M., Mudryk, L., & Zhang, X. (2022). Canada’s changing
climate report, in light of the latest global science
assessment. https://doi.org/10.4095/329703

Government of Northwest Territories. (2019, November 5).

Inuvik Tuktoyaktuk Highway: Opening a New Research
Corridor. NTAI Conference.

Government of Northwest Territories. (2020). Gunghi

Creek Culvert Replacement.

Heginbottom, A. . J. (1995). Canada-Permafrost. In

National Atlas of Canada.

Hjort, J., Streletskiy, D., Doré, G. et al. Impacts of

permafrost degradation on infrastructure. (2022)
Nature Reviews, Earth & Environment 3, 24–38
https://doi.org/10.1038/s43017-021-00247-8

Kiggiak-EBA. (2019). Adfreeze Steel Pipe Pile Design

Recommendations for Precast Concrete Bridge Culvert
Gunghi Creek, km 131.2 Inuvik-Tuktoyaktuk Highway.
Kiggiak - EBA.
A Review of Ground Temperatures at the
Hans Creek Bridge Foundation
T. Ed Hoeve, P.Eng.
HoevEng Consulting Ltd., Yellowknife, NWT, Canada

ABSTRACT
The Inuvik to Tuktoyaktuk Highway (ITH) includes eight bridges along its
140 km length. The highway crosses continuous ice-rich permafrost terrain
and bedrock is generally too deep to be accessed for foundation construction. The Hans Creek Bridge is at km 57.3 of the
ITH and is constructed on two abutments and two piers. Initial post-construction ground temperature monitoring essentially
confirmed the design assumptions at the abutment locations, but ground temperatures at the bottom of the pier piles were
found to be significantly warmer than assumed and the warmest observed at any of the ITH bridges. This paper
summarizes the ground temperature monitoring data from the initial five years following construction and compares it with
the design assumptions. The design and practical implications are discussed. It is expected that intervention to increase
the capacity of the pier piles may be required sooner than estimated at the time of design.

RÉSUMÉ
La route entre Inuvik et Tuktoyaktuk (ITH) comprend huit ponts sur une longueur de 140 km. La route traverse un terrain
continu de pergélisol riche en glace et le substrat rocheux est généralement trop profond pour être utilisé pour la
construction des fondations. Le pont Hans Creek se trouve au km 57.3 de l’ITH et est construit sur deux culées et deux
piles. La surveillance initiale de la température du sol après la construction a essentiellement confirmé les hypothèses de
conception aux emplacements des culées, mais les températures du sol en-dessous des piles se sont avérées beaucoup
plus chaudes que prévu, et ont représentées les temperatures les plus chaudes observées à tous les ponts de l’ITH. Le
présent document résume les données de surveillance de la température du sol des six premières années suivant la
construction et les compare aux hypothèses de conception. La conception et les implications pratiques sont discutées. On
s’attend à ce qu’une intervention visant à accroître la capacité des pieux supportant les piles soit nécessaire plus tôt que
prévu au moment de la conception.

1 INTRODUCTION

The Inuvik to Tuktoyaktuk Highway (ITH) was a nation-

building project that provides the first all-season road link
to the Canadian Arctic coast. It was constructed from about
2015 to 2018. The highway includes eight bridges along its
140 km length. Figure 1 shows an overview of the route
and bridge locations. The subject of this review is referred
to as the Hans Creek Bridge (Bridge 30a) and is situated
at km 57.3 of the ITH, south of Parsons Lake and almost
half-way along the route.
Rampton (1987) presents the surficial geology of the
project area. The highway alignment crosses two distinct
physiographic regions. The southern third of the alignment
(from Inuvik to south of Husky Lakes) crosses the