Proceedings 10th NUMGE 2023

10th European Conference on Numerical Methods in Geotechnical Engineering

Zdravković L, Konte S, Taborda DMG, Tsiampousi A (eds)
ISSN
© Authors: All rights reserved, 2023
doi: 10numge-2023-Y-XXXX

Seismic behaviour of a tailings dam in Peru

C. Salvador1, F. Robles1, C. Romanel1
1
Pontifícia Universidade Católica do Rio de Janeiro, Department of Civil and Environmental Engineering, Rio de
Janeiro, Brazil

ABSTRACT: This paper presents a study on the seismic behaviour of an upstream tailings dam situated in the province of
Arequipa, Peru. In 2013 seismic events caused cracks in the dam body, crest settlements and sand boils due to liquefaction. In
this investigation, the seismic hazard assessment was determined by probabilistic methods and design earthquakes were gener-
ated complying with the risk rating recommendations for the global industry standard on tailings management. A finite element
numerical analysis was performed considering the PM4Sand and PM4Silt constitutive elastoplastic models. Numerical results
were presented in terms of permanent displacements and pore pressure distribution immediately after the earthquake duration,
thus indicating that the Otapara tailings dam will suffer dynamic liquefaction for 1000 year return period earthquakes.

Keywords: Tailings dam; design earthquake; seismic behaviour; dynamic liquefaction

1 INTRODUCTION strain behaviour of sands, reflecting the physical

existence of a “semifluidized (Sf) state” for very low
Peru is located in an region of high seismic activity,
effective mean stress reached in the post-liquefaction
with different earthquakes faulting mechanism due to
stage. This model introduces a new internal degradation
interaction between the Nazca and the South American
variable that affects the plastic modulus and dilatancy
tectonic plates. Mining is a very important economic
within the Sf state, while leaving almost intact the
activity for the country which requires the construction response outside of it. Aside from the constants with
of civil engineering structures, such as tailings dams, in
default numerical values, SANISAND-Sf model
order to store large amounts of waste materials.
requires the calibration of 20 model constants: 15 of
In this paper, the seismic behaviour of the Otapara
them inherited from SANISAND and the remaining 5
tailings dam is investigated with emphasis on
related to the Sf state.
simulation of the dynamic liquefaction of the tailings by
The PM4Sand model (Boulanger and Ziotopoulou,
the finite element method (FEM).
2015) is a 2D plane-strain model that follows the basic
One important ingredient for FEM analysis is the
framework of the stress-ratio controlled, critical state
elastoplastic model incorporated in numerical
compatible, bounding-surface plasticity model for sand
simulations. In the literature, there are some suggestions
presented by Dafalias and Manzari (2004). The model
for dynamic liquefaction prediction of sand-like
parameters are grouped into two categories: a primary
materials, such as the UBCSand model (Beaty and
set of 3 parameters (apparent relative density DR, shear
Byrne, 1998) and its 3D generalization UBC3D-PLM
modulus coefficient 0, contraction rate parameter ℎ 0)
(Tsegaye, 2010) which incorporates a soil densification
that are most important for model calibration, and a
rule in order to predict a more realistic evolution of
secondary set of parameters (18 in total and optional)
excess pore pressures during cyclic loading. However,
that may be modified from their default values in special
this rule works consistently only in the case of
circumstances. By changing the three primary input
symmetric loading and for stress paths that start from
parameters it is possible to achieve reasonable
the isotropic axis. Also, UBC3D-PLM does not include
approximations of desired behaviour including pore
the influence of the effective confining pressure on the
pressure generation and dissipation, limiting strains,
cyclic resistance of the soil (Petalas and Galavi, 2013).
and cyclic mobility. The apparent relative density
The SANISAND models (Dafalias and Manzari,
has the most crucial effect in the cyclic strength and also
2004; Taiebat and Dafalias, 2008; Dafalias et al., 2016;
significant influence in the generation of excess pore
Petalas et al., 2019; Papadimitriou et al., 2019; Petalas
pressures, while the shear modulus coefficient 0 has
et al., 2020) are another class of models based on the
minor influence in the cyclic strength, and it not affects
framework of bounding surface plasticity and critical
the generation of excess pore pressures until a pore
state mechanics concepts. A model referred to as
pressure ratio of around 1.0 is reached. Since its
SANISAND-Sf (Barrero et al., 2020) has been recently
introduction, the PM4Sand model has drawn wide
proposed to capture the post-liquefaction cyclic shear

1 NUMGE 2023 - Proceedings

 Dams, embankments and slopes

attention in geotechnical engineering practice and 2.3 Field and laboratory tests
research communities due to its relatively easy
The geotechnical characterization of soils was carried
calibration process and good agreement with field
out through field and laboratory tests. The field tests (58
observations.
SPT and 10 CPT) were adequately distributed over the
area of interest, in order to estimate the shear strength
2 OTAPARA TAILINGS DAM
and consistency / compaction of soils that make up the
tailings deposit. In some selected samples, laboratory
2.1 Location tests were executed to determine the soil classification,
The Otapara tailings dam is situated in the province of according to the Unified Soil Classification System
Arequipa, Peru, at an average elevation of 430 meters (USCS), as well as the wet density and specific gravity.
above sea level. Figure 1 indicates the location of the To evaluate the resistance properties, direct shear tests
dam in relation to the Acarí River and the city of were carried out on samples of coarse tailings that form
Otapara-Acarí. the upper slope of the reservoir. Cyclic triaxial tests and
cyclic direct shear tests (CDSS) were also carried to as-
sess the dynamic properties of the tailings materials.

3 SEISMICITY AT THE OTAPARA DAM

SITE

3.1 Seismic classification

According to the Global Industry Standard on Tailing
Management Methodology (GISTM, 2020), the
Otapara tailings dam has a potentially significant risk
that should consider a 1000-year return period (Tr) for
the design earthquake.

Figure 1. Otopara dam, Otapara-Acari city and trajectory of 3.2 Probabilistic Seismic Hazard Assessment
dam-break flow. (PSHA)
2.2 Characteristics of the tailings dam The PSHA was investigated using the methodology pro-
posed by Cornell (1968) implemented in the software
The Otapara tailings dam consists of an embankment, R-CRISIS (Ordaz et al., 2020). It was estimated the seis-
built with borrow material, and subsequently raised by mic hazard for a probability of exceeding 5% in 50
the upstream construction method, using cyclonic tail- years, corresponding to a return period of 1000 years.
ings released from the crest where they were deposited The calculation of the peak ground acceleration (PGA)
in a fully saturated condition with low relative density. in the Otopara dam site was done using a regular mesh
Figure 2 shows the critical cross-section of the dam at of 0.1º in longitude and latitude.
the end of construction. The identification and characterization of the seismo-
genic sources were taken from Aguilar et al. (2017). For
subduction and continental earthquakes, some specific
ground motion prediction equations (GMPEs) were se-
lected and used in a logic tree framework, in order to
reduce the epistemic uncertainties inherent in the seis-
micity models.
From the estimated seismic risk curves for different
Figure 2. Critical cross-section of the Otapara dam. structural periods, it was possible to obtain the Uniform
Hazard Spectra (UHS) that presents acceleration values
The pulp-type tailings are pumped from the mining in different structural periods for the same annual ex-
plant to the disposal pond, forming a beach. The reser- ceedance probability considering an exposure time of
voir has a waterproofing geomembrane lining at the bot- 50 years of the structure (Figure 3).
tom, as well as surface and deep water drainage systems
and ponds for leachate treatment. According to the shear
wave velocity (Vs) the reservoir bedrock is classified as
a site class B (ASCE 7-16, 2017).

2 NUMGE 2023 - Proceedings

 Seismic behaviour of a tailings dam in Peru

1,6 mechanism were selected. In this investigation, the Lo-

Spectral acceleration, Sa(g) 1,4 mas (2018), Atico (2001) and Lima (1974) seismic rec-
UHS - Site class B (Tr=1000 years)
1,2 ords were chosen for seismic modeling (Table 1).
1,0
0,8 Table 1. Earthquakes considered for the present analysis.
0,6 Earthquake Lomas Atico Lima
0,4 Jan 14, Jun 23, Oct 03,
0,2 Date
2018 2001 1974
0,0 Magnitude
0,0 0,5 1,0 1,5 2,0 2,5 3,0 7.1 8.2 7.4
(Mw)
Period, T(s)
Rupture Subduction Subduction Subduction
Figure 3. Spectral acceleration response for 1000-year re- mechanism interface interface interface
turn period. Depth (km) 27 33 13
Caraveli Moquegua Lima
The seismic hazard disaggregation (Figure 4) also pro- Station
(IGP) (CISMID) (PRQ)
vides a useful insight into the main sources contributing Duration
to the hazard at the Otapara dam site. The interface 115.01 198.92 97.96
(sec)
earthquake of magnitude 6.3 Mw at 65 km distance gen- Acceleration
0.26 0.30 0.20
erates the greatest contribution, but it is also important (g)
to notice that there is another main contribution from a
5.4 Mw intraplate earthquake at a distance of 65 km. The spectral matching method (Al Atik & Abraham-
son, 2010) was used to generate the artificial accelero-
grams whose response spectra should be consistent with
the target spectrum obtained in the PSHA. Figure 5
shows the response spectra (before and after spectral
matching) while Figure 6 presents the artificial acceler-
ograms for each one of the earthquakes considered in
this analysis.
1,8
1,6 Lomas 2018 (orig.)
Spectral acceleration, Sa(g)

Atico 2001 (orig.)

1,4 Lima 1974 (orig.)
1,2 Lomas 2018 (match)
Figure 4. Disaggregation of probabilistic seismic hazard 1,0 Atico 2001 (match)
0,8 Lima 1974 (match)
(PGA) for 5% probability of exceedance in 50 years. UHS
0,6
0,4
3.3 Design Earthquakes 0,2
0,0
Considering the results obtained in the disaggregation
0,0 0,5 1,0 1,5 2,0 2,5 3,0
procedure, seismic records with similar spectral con- Period, T(s)
tent, magnitude, epicentral distance and focal rupture Figure 5. Response spectra before and after spectral match-
ing.

0,5 0,5 0,5

0,4 Lomas 2018 (match) 0,4 Atico 2001 (match) 0,4 Lima 1974 (match)
0,3 Lomas 2018 (Original) 0,3 Atico 2001 (original) 0,3 Lima 1974 (original)
0,2 0,2 0,2
Acceleration (g)

Acceleration (g)
Acceleration (g)

0,1 0,1 0,1

0,0 0,0 0,0
-0,1 -0,1 -0,1
-0,2 -0,2 -0,2
-0,3 -0,3 -0,3
-0,4 -0,4 -0,4
-0,5 -0,5 -0,5
0 10 20 30 40 50 60 70 0 20 40 60 80 100 120 0 10 20 30 40 50 60 70 80 90
Time (s) Time (s) Time (s)
(a) (b) (c)
Figure 6. Acceleration histories before and after spectral matching for the three selected earthquakes (a) Lomas 2018; (b) Atico
2001; (c) Lima 1974.

3 NUMGE 2023 - Proceedings

 Dams, embankments and slopes

4 SEISMIC RESPONSE OF THE OTAPARA tailings was prescribed as 2.0 m while for the other
TAILING DAM materials the values ranged from 5.0 m to 35.0 m.
The numerical results from finite element analysis were
compared with the values of safety factor against
liquefaction determined by the simplified stress-based
methods. In the numerical model, the mechanical
behaviour of the tailing materials was simulated with
the PM4Sand (Boulanger and Ziotopoulou, 2015) and
the PM4Silt (Boulanger and Ziotopoulou, 2018)
elastoplastic constitutive models, capable of estimating Figure 7. Materials in the Otopara dam cross-section.
the dynamic liquefaction triggering and softening of
materials when subjected to cyclic loading. 4.2 Material Properties
Based on field investigation, laboratory tests and
4.1 Cross-section geometry correlations from literature, the geotechnical parameters
The geometry of the representative cross-section of the for all materials were obtained. The linear elastic model
dam is shown in Figure 7, indicating the occurrence of was used to mechanically represent the bedrock
bedrock (R), embankment (E), alluvial deposit (AD) behaviour (E = 13,500 MPa, ν = 0.2), while the
and tailings (T0, T1, T2, T3, T4, T5, T6). elastoplastic HSM model (Schanz, 1999) was chosen to
According to Lysmer and Kuhlemeyer (1973) simulate the response of the embankment and the
recommendations, a suitable maximum mesh spacing is alluvial deposit (Table 2), and the PM4Sand and
usually determined by considering a tenth of the PM4Silt models were used for the tailings (Table 3 and
minimum relevant wavelength (or highest frequency 4).
fmax) in the input signal. The finite element size for the

Table 2. Parameters used for the seismic analysis of non-liquefiable materials.

Mate- γsat M cˈ ϕˈ γ0.7
SUCS
rial (kN/m3) (kPa) (kPa) (kPa) (-) (kPa) (°) (kPa) (-)
E GM 21.0 60,000 60,000 180,000 0.5 10 37.0 239,000 0.0001
AD GM 20.0 45,000 45,000 135,000 0.5 5 37.0 239,000 0.0002

Table 3. PM4Sand parameters for the tailings layers

Material SUCS γunsat γsat Dr Go hpo 4.3 Parameter calibration for cyclic models
T0 SM 16.5 20.0 0.40 524 0.27
The primary parameters for the PM4Sand and PM4Silt
T1 SM 16.5 20.0 0.40 524 0.27
constitutive models are based on correlation of the
T3 SM 16.5 20.0 0.40 524 0.20
relative density with SPT data. Specifically, the hpo
T4 SM 17.0 20.0 0.60 729 0.11
parameter was calibrated using the Plaxis 2D Soil Test
T5 SM 17.0 20.0 0.65 574 0.90
module, taking results from CDSS and cyclic triaxial
T6 SM 19.0 20.0 0.80 944 1.00
tests on tailings samples carried out at different
confinement stresses. Figure 8 shows the stress-strain
Table 4. PM4Silt parameters for tailings layers with softening
behaviour response between laboratory results (black curves) and
Material SUCS γunsat γsat Su,ratio Go hpo
the calibrated constitutive model (PM4Sand) for a
T2 ML 19.0 20.0 0.30 400 6.8
sample from tailings T5. The same procedure was
carried out for other samples, trying to approximate the
calibrated cyclical response (blue curves) to the results
An elastic analysis was previously performed to
obtained in the laboratory tests.
determine Rayleigh's damping coefficients in the
Figures 9 to 13 show the cyclical resistance curves. A
materials that make up the dam, although for the tailings
data conversion from cyclic triaxial tests to CDSS single
only hysteretic damping was considered. Table 5 shows
shear strain mode was performed following the proce-
the values of the damping coefficients and β.
dure suggested by Cappellaro et al. (2017).
Table 5. Rayleigh damping for seismic analysis.
Material β
E, AD, R 0.29 0.0048

4 NUMGE 2023 - Proceedings

 Seismic behaviour of a tailings dam in Peru

0,3

CSR ( = τ/σ' v0) for γ =3,75 %

0,2
y = 0,25x-0,28
R² = 1,00
y = 0,20x-0,21
R² = 0,98
0,1
Tailing T4 (SM), Cyclic TX, σ'vc= 200 kPa, Dr=60%
Cyclic TX Correlation - CDSS (Cappellaro et al. 2017)
Soil test Plaxis, σ'vc= 200 kPa, Dr=60%, Go=729, hpo= 0,11
0,0
1 10 100
Number of cycles, Nc
Figure 12. Cyclic resistance curves, calibration of sample
Figure 8. Calibration of CDSS test carried out in sample from from tailing T4.
tailing T5.
0,3
0,3 Tailing T5 (SM), CDSS test: γ =20 kPa, σ'vc = 1000 kPa
Tailing T0 and T1 (SM), Cyclic TX, σ'v,c= 100 kPa, Dr=40%
Soil test Plaxis, σ'vc =1000 kPa, Dr= 0,65, Go= 574, hpo= 0,9, Q=10, R=1,5

CSR ( = τ/σ' v0) for γ =3,75 %

Cyclic TX Correlation - CDSS (Cappellaro et al. 2017)
CSR (= τ/σ' v0) for γ =3,75 %

Soil test Plaxis, σ'vc= 100 kPa, Dr=40%, Go=524, hpo= 0,27

0,2 y = 0,30x-0,25
R² = 1,00
0,2

y = 0,13x-0,10
0,1 R² = 0,54 y = 0,29x-0,24
y = 0,21x-0,32 R² = 1,00
R² = 0,98
0,1
0,0 1 10 100
1 10 100 Number of cycles, Nc
Number of cycles, Nc Figure 13. Cyclic resistance curves, calibration of sample
Figure 9. Cyclic resistance curves, calibration of samples from tailing T5.
from tailings T0 and T1.

0,3
4.4 Permanent Displacements
The total permanent displacements induced by the Lo-
CSR (=τ/σ'v0) for γ =3,75 %

y = 0,35x-0,25 mas (2018), Atico (2001) and Lima (1974) earthquakes

R² = 0,91
0,2 may be seen in Figure 14.
y = 0,31x-0,21
R² = 1,00

0,1

Tailing T2 (ML), CDSS test: γ =19 kPa, σ'vc = 850 kPa

(a)
Soil test Plaxis, σ'vc=850 kPa, Suratio= 0,30, Go=400, ng=0,75, hpo=6,8
0m 1.50m
0,0
1 10 100 1000
Number of cycles, Nc
Figure 10. Cyclic resistance curves, calibration of sample (b)
from tailing T2.
0m 3.40m
0,2

(c)
CSR ( = τ/σ' v0) for γ =3,75 %

0m 2.0m
Figure 14. Total permanent displacements provoked by (a)
0,1
y = 0,15x-0,22 Lomas 2018; (b) Atico 2001; (c) Lima 1974 earthquakes.
R² = 0,87
y = 0,15x-0,26
Tailing T3 (SM), Cyclic TX, σ'vc= 200 kPa, Dr=40% R² = 1,00 4.5 Distribution of pore pressure ratio (ru)
Cyclic TX Correlation - CDSS (Cappellaro et al. 2017)
Soil test Plaxis, σ'vc= 200 kPa, Dr=40%, Go= 524, hpo= 0,2 In this research, it was used the definition of the pore
0,0
1 10 100
pressure raio ru given by Beaty and Byrne (2011):
Number of cycles, Nc
Figure 11. Cyclic resistance curves, calibration of sample
from tailing T3. r 1 (1)

5 NUMGE 2023 - Proceedings

 Dams, embankments and slopes

where σ'v is the current effective vertical stress and σ'v0 American Society of Civil Engineers – ASCE. 2017. Minimum De-
represents the initial effective vertical stress. Figure 15 sign Loads and Associated Criteria for Buildings and Other
shows the ru distribution immediately after the seismic Structures. ASCE Standard ASCE/SEI 7-16. Reston, Virginia.
excitation duration, where may be seen that large Barrero, A. R., Taiebat, M., & Dafalias, Y. F. 2020. Modeling cyclic
shearing of sands in the semifluidized state. International Journal
regions of the tailings reservoir suffer dynamic
for Numerical and Analytical Methods in Geomechanics, 44(3),
liquefaction with ru = 1 for all 3 earthquakes considered 371-388.
in the analysis. Beaty, M., & Byrne, P. M. 1998. An effective stress model for
pedicting liquefaction behaviour of sand. Geotechnical Earth-
quake Engineering and Soil Dynamics III (pp. 766-777). ASCE.
(a) Beaty, M. H.; Byrne, P. M. 2011. UBCSAND Constitutive Model
version 904aR. Itasca UDM Web Site, 69.
Boulanger, R. W.; Ziotopoulou, K. 2015. PM4Sand (version 3): A
sand plasticity model for earthquake engineering applications.
(b)
Report No. UCD/CGM-15/01, Center for Geotechnical Model-
ing, Department of Civil and Environmental Engineering, Uni-
versity of California, Davis, Calif.
(c) Boulanger, R. W.; Ziotopoulou, K. 2018. PM4Silt (version 1): A silt
plasticity model for earthquake engineering applications. Report
0 1 No. UCD/CGM-18/01, Center for Geotechnical Modeling, De-
Figure 15. Distribution of pore pressure ratio (ru) immediately after partment of Civil and Environmental Engineering, University of
the (a) Lomas 2018; (b) Atico 2001; (c) Lima 1974 earthquakes. California, Davis, CA, 108 pp.
Cappellaro, C., Cubrinovski, M., Chiaro, G., Stringer, M. E., Bray,
J. D.; Riemer, M. F. 2017. Undrained cyclic direct simple shear
5 CONCLUSIONS testing of Christchurch sandy soils. 20th NZGS Geotechnical
According to the GISTM (2020) methodology, the Symposium (pp. 1-8).
Otapara tailings dam was classified as a structure of Cornell, C. A. 1968. Engineering seismic risk analysis. Bulletin of
the Seismological Society of America, 58(5), 1583-1606.
significant potential risk for dynamic liquefaction, under
Dafalias, Y. F., & Manzari, M. T. 2004. Simple plasticity sand model
earthquakes with a return period of 1000 years. The accounting for fabric change effects. Journal of Engineering Me-
design earthquakes were generated with basis on a chanics, 130(6), 622-634.
probabilistic seismic hazard assessment for the Otapara Dafalias, Y. F., & Taiebat, M. 2016. SANISAND-Z: zero elastic
dam, considering GMPEs for class B (rock). The Lomas range sand plasticity model. Géotechnique, 66(12), 999-1013.
(2018), Atico (2001) and Lima (1974) seismic records GISTM Global Industry Standard on Tailings Management 2020.
were chosen for seismic modeling considering their International Council on Mining and Metals - ICMM.
magnitudes, epicentral distances, focal rupture Kuhlemeyer, R.L; Lysmer, J. 1973. Finite element accuracy for
mechanisms and frequency contents. wave propagation problems, Journal of the Soil Mechanics and
Simplified stress-based methods were initially used to Foundations Division, vol. 99(5), p. 421-427.
Ordaz, M., & Salgado-Gálvez, M. A. (2020). R-CRISIS v20 Vali-
estimate the potential of liquefaction triggering through
dation and verification document. Mexico City: ERN Technical
deterministic and probabilistic approaches. The results Report.
were confirmed by finite element analysis considering Papadimitriou, A. G., Chaloulos, Y. K., & Dafalias, Y. F. 2019. A
the PM4Sand and PM4Silt elastoplastic models to fabric-based sand plasticity model with reversal surfaces within
represent the seismic behaviour of the tailings materials. anisotropic critical state theory. Acta Geotechnica, 14(2), 253-
The earthquakes provoked maximum permanent dis- 277.
placements between 1.5 to 3.4 m and high pore pres- Petalas, A. L., Dafalias, Y. F., & Papadimitriou, A. G. 2019. SANI-
sures in the tailings reservoir, leading to dynamic lique- SAND‐FN: An evolving fabric‐based sand model accounting for
faction, as indicated by the pore pressure ratio ru =1 in stress principal axes rotation. International Journal for Numeri-
large regions of the Otopara dam. cal and Analytical Methods in Geomechanics, 43(1), 97-123.
Petalas, A. L., Dafalias, Y. F., & Papadimitriou, A. G. 2020. SANI-
SAND-F: Sand constitutive model with evolving fabric anisot-
ACKNOWLEDGEMENTS ropy. International Journal of Solids and Structures, 188, 12-31.
To the National Council for Scientific and Technologi- Petalas, A., & Galavi, V. 2013. Plaxis liquefaction model UBC3D-
cal Development (CNPq) of Brazil for the scholarships PLM. Plaxis Report.
Schanz, T. 1999. Formulation and verification of the Hardening-Soil
granted to the two first authors.
Model. RBJ Brinkgreve, Beyond 2000 - Computational Geotech-
nics, 281-290.
REFERENCES Taiebat, M., & Dafalias, Y. F. 2008. SANISAND: Simple aniso-
Aguilar, Z., Roncal, M., Piedra, R. 2017. Probabilistic Seismic Haz- tropic sand plasticity model. International Journal for Numerical
ard Assessment in the Peruvian Territory, 16 WCEE – World and Analytical Methods in Geomechanics, 32(8), 915-948.
Conference on Earthquake, 1-10. Tsegaye, A. 2010. Plaxis liquefaction model. external report.
Al Atik, L., & Abrahamson, N. 2010. An improved method for non- PLAXIS knowledge base.
stationary spectral matching. Earthquake spectra, 26(3), 601-617.

6 NUMGE 2023 - Proceedings