Missouri University of Science and Technology

Scholars' Mine

International Conference on Case Histories in (2008) - Sixth International Conference on Case

Geotechnical Engineering Histories in Geotechnical Engineering

13 Aug 2008, 9:00am - 9:45am

The Impact of Low Probability Ground Motions on Canadian

Geotechnical Engineering Practice
W. D. Liam Finn
University of British Columbia, Vancouver, Canada

Follow this and additional works at: https://scholarsmine.mst.edu/icchge

Part of the Geotechnical Engineering Commons

Recommended Citation
Finn, W. D. Liam, "The Impact of Low Probability Ground Motions on Canadian Geotechnical Engineering
Practice" (2008). International Conference on Case Histories in Geotechnical Engineering. 1.
https://scholarsmine.mst.edu/icchge/6icchge/session00d/1

This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

This Article - Conference proceedings is brought to you for free and open access by Scholars' Mine. It has been
accepted for inclusion in International Conference on Case Histories in Geotechnical Engineering by an authorized
administrator of Scholars' Mine. This work is protected by U. S. Copyright Law. Unauthorized use including
reproduction for redistribution requires the permission of the copyright holder. For more information, please
contact [email protected].
 THE IMPACT OF LOW PROBABILITY GROUND MOTIONS ON CANADIAN
GEOTECHNICAL ENGINEERING PRACTICE

W. D. Liam Finn
University of British Columbia
Vancouver, Canada, V6T 1Z4

ABSTRACT

The adoption of design motions with a 2% rate of exceedance in 50 years in the National Building Code of Canada (NBCC 2005) has had a
major impact on geotechnical engineering practice in Canada. The peak ground accelerations were doubled compared with the previous
motions which had an exceedance rate of 10% in 50 years. The increase in accelerations has had a huge effect on assessments of
liquefaction potential and slope stability, because the methods of assessment in common use depend on peak ground acceleration. This
paper describes typical problems encountered in Canadian practice with use of the low probability motions and describes some measures for
alleviating the impact on design, while maintaining the code objective of life safety.

INTRODUCTION building code in 2006, sites on slopes slated for residential

development failed to be approved for the use intended that
In 2005 the National Building Code of Canada (NBCC 2005) would have been considered safe under the previous code.
adopted design ground motions with a 2% chance of being Developers and municipalities were understandably upset by this
exceeded in 50 years. This change resulted in about a doubling abrupt turn of events and appealed to the BC government for
of peak ground accelerations (PGA) compared to the PGA relief. The government responded by issuing provincial
associated with the design ground motions in the previous code regulation M268 in December 2006 restoring the 10% in 50
(NBCC 1995) as shown in Table 1. years motions for slope stability assessment as a temporary
measure and setting up a task force on seismic slope stability
Table 1. PGA hazard in Canadian cities, NBCC 1995 and 2005 (TFSSS) under the direction of the Association of Professional
Engineers and Geologists of British Columbia (APEGBC) to
Median frequency Vancouver Toronto Montreal study the issues and make recommendations for future action.
of exceedance The writer is a member of the task force. The TFSSS approach to
10% in 50 yrs NBCC 1995 0.24 0.08 0.20 assessment of slope stability is described herein. A later
extension by the writer is described in which procedures
2% in 50 yrs NBCC 2005 0.46 0.20 0.43
recommended by the TFSSS are coupled with reliability analysis
to allow uncertainties in soil properties and seismic input to be
The impact of the increases in ground motions on geotechnical taken into account.
engineering practice depends on the type of design. Conventional
procedures for assessing liquefaction potential and slope stability The problems associated with liquefaction were studied by a
have been based traditionally on peak ground acceleration. Task Force of the Vancouver Geotechnical Society. A state of
Designs based on these procedures have been strongly and the art approach for dealing with liquefaction, especially under
directly affected by the increased peak ground accelerations. extreme motions is described in a comprehensive report
Sites and structures which would have been safe under the old (Anderson et al. 2007). The report advocates attention to the
code may now be considered unsafe for the new hazard levels. consequences of liquefaction, primarily expressed in settlements
Geotechnical engineers and their clients have been expressing and lateral displacements as the key to rational and cost–effective
concerns about the great impact of the changes in ground engineering. The report describes a state of the art level of
motions on projects. practice.

The impact on seismic slope stability makes an interesting case In more conventional practice, the Seed and Idriss (1971)
history. Following the adoption of the NBCC 2005 design simplified method for assessing liquefaction, as updated in Youd
motions by the province of British Columbia (BC) in the BC et al. (2001), is widely used, especially in Eastern Canada. The

Finn - Keynote Paper 1

method is based on a measure of soil resistance to liquefaction, a
earthquake magnitude as a surrogate for duration of shaking and

Acceleration, a
the peak ground acceleration. The probabilistic ground motions
in NBCC 2005 are the combined contribution to hazard of all 0

earthquakes in the seismic sources contributing to hazard. Which

magnitude should be associated with the code PGA for
k=PGA
implementation of the Seed-Idriss method? Finn and Wightman
(2006, 2007) suggested two approaches for dealing in a logical t
way with probabilistic motions. One method is based on the Time, t
concept of a hazard analysis based on earthquake magnitudes
weighted according to the relative contributions they make to Fig.2. k = PGA, can be a very conservative estimate of k.
liquefaction potential. This concept was first proposed by Idriss
(1985). The second approach is based on de-aggregating the
hazard and summing up the contributions of the individual SLOPE PERFORMANCE DURING SHAKING
magnitudes to liquefaction potential. These methods are
described later. Newmark (1965) revolutionized concepts of seismic slope
stability by pointing out that just because the factor of safety
occasionally fell below FS = 1.0 during earthquake shaking, it
SEISMIC SLOPE STABILITY: REVIEW OF CURRENT did not necessarily mean slope failure. He proposed that the total
PRACTICE IN BC displacement accumulated during the times when the factor of
safety was less than FS = 1.0 be used as the index of slope
In BC, the most common method currently used to carry out performance during an earthquake and he developed simple
seismic slope stability analysis is the pseudo-static limit procedures for calculating the displacements.
equilibrium method. In this method, earthquake loading is
represented by a constant horizontal force, kW, applied to the Permanent displacements can occur in a slope during an
centre of gravity of the potential sliding mass, as shown in Fig. 1. earthquake only if the shear stresses generated by the earthquake
W is the weight of the sliding mass and the coefficient, k, is exceed the shearing resistance of the slope. The horizontal force
called the seismic coefficient. required to bring the slope to the condition of incipient
displacement is shown in Fig. 3 as F = kyW where ky is the
seismic yield coefficient, a special value of the seismic
coefficient that just allows slip or yielding in the slope. The yield
coefficient, ky = ay/g, where, ay = yield acceleration and g = the
acceleration of gravity.

Fig.1. Pseudo-static method of seismic slope stability analysis

There is, however, no generally accepted method in BC practice

for selecting seismic coefficients for slopes. From a limited
survey of BC practice, the TFSSS found seismic coefficients in
the range 0.5(PGA) ≤ k ≤ 1.0(PGA), where PGA is the peak Fig.3. The condition of incipient displacement under ky .
ground acceleration.
Figure 4 is a segment of a typical earthquake shaking record to
The choice of k = 1.0(PGA) may be very conservative as shown an enlarged scale. Slope displacements can occur whenever the
by the acceleration time history in Fig. 2. The PGA occurs only ground acceleration, ‘a’, exceeds the yield acceleration, ay. In
for an instant and most of the record indicates accelerations Fig. 4, the time intervals during which displacements are initiated
much less than the maximum. The PGA has no significant impact have been shaded. The total slope displacement at the end of
on the response of the slope to shaking by the time history. earthquake shaking is the sum of the incremental slope
Therefore the TFSSS recommends the use of k = PGA only as a displacements generated by the ground acceleration being greater
preliminary screening tool. If FS ≥ 1.0, when k = PGA is used in that the yield acceleration. Newmark (1965) calculated these
a pseudo-static limit equilibrium slope stability analysis, no displacements by considering the sliding mass of soil to be rigid.
further stability analyses are required. He also provided charts for estimating the maximum
displacements. These charts were based on the small selection of

Finn - Keynote Paper 2

strong ground motion records available at the time. In present 2% probability of exceedance of the ground motions in 50 years,
practice, slope displacements are also estimated by direct the absolute probability of the median displacements being
calculation using design ground motions as input to the exceeded is 1% in 50 years (approximately 1/5000). The median
Newmark (1965) sliding rigid block computational model or by displacement is selected as the controlling slope displacement
using a model that takes the flexibility of the slope into account. because of the low absolute probability of exceedance.

Bray and Travasarou’s equation slope displacement, D, greater

Acceleration, a

A yield acceleration, ay than 1cm is:

Ln(D) = –1.10 – 2.83 ln(ky) – 0.333 (ln(ky))2

+ 0.566 ln(ky) ln (S(1.5Ts)) + 3.04 ln(S(1.5Ts))

Time, t – 0.244 (ln(S(1.5Ts)))2 + 1.5Ts + 0.278(M-7) ± ε (1)

Fig.4. Displacement is initiated when ground acceleration
The displacement D is due to shearing along the slip surface and
exceeds yield acceleration.
has both vertical and horizontal components.
Makdisi and Seed (1978) improved the Newmark model for
Ts is the initial fundamental period of the potential sliding mass
application to embankment dams by taking into account the
prior to the seismic event, (0.05s < Ts< 2.0s) and, for a slope
flexibility of the embankment and the amplification of ground
such as shown in Fig. 1, is estimated by:
motions on passing up through the embankment. They developed
charts relating slope displacement to earthquake magnitude and
the ratio of the seismic coefficient k to yield coefficient ky. On Ts = 4H/Vs (2)
the basis of Makdisi and Seed (1978) data, Seed (1979)
recommended values of k in the range 0.1-0.15 depending on where H is the average height and Vs is the average shear wave
earthquake magnitude, M, for the analysis of the slopes of earth velocity of the potential sliding mass. Site investigations for most
dams. For example the Seed procedure calls for k = 0.15 and a residential developments do not typically include measurements
factor of safety FS ≥ 1.15 for an earthquake with M = 8.25. This of shear wave velocity, but estimates can be inferred from
value of k is associated with a maximum allowable displacement standard penetration test or cone penetration test data (Sykora
of 100 cm. Los Angeles County subsequently modified this and Koester, 1988).
procedure to a single k value with k = 1.15 and FS ≥ 1.0 (Blake
et al. 2002). Hynes-Griffin and Franklin (1984) recommended In Eq. 1, ε is a normally distributed random variable with a mean
using k = 0.5(PGA). This value of k is also based on a maximum of zero and a standard deviation σ = 0.66 and M the moment
allowable slope displacement of 100 cm. It is important to note magnitude of the earthquake under consideration. The term
that these generally accepted methods for selecting a seismic S(1.5Ts) is the spectral acceleration at the site for the period of
coefficient in U.S. practice for embankment dams are based on (1.5Ts). It is given by S(1.5Ts) = F*Sa(1.5Ts) where Sa(1.5Ts) is
slope displacement criteria. The two procedures recommended the spectral acceleration for firm soil conditions and F is the
here by TFSSS are also based on a criterion of acceptable slope amplification factor for the site class. Values of F, as a function
response during an earthquake expressed in terms of allowable of site class and period, and Sa(1.5Ts), for periods T = 0.2, 0.5,
displacement. These methods, to be acceptable for general use, 1.0 and 2.0, are provided in NBCC 2005. Values of Sa(1.5Ts) for
had to be conceptually simple and easy to apply. other periods can be interpolated linearly from the values
provided in NBCC 2005. Bray (2007) suggested that a value of
Ts = 0.33, giving a spectral period (1.5Ts) of 0.5, would be
SLOPE DISPLACEMENT (METHOD 1) adequate for general use. The TFSSS recommends this value but
an engineer is not precluded using a slope specific Ts, when he
The TFSSS reviewed recent developments in methods of seismic considers it more appropriate. S decreases with increasing values
slope stability analysis and selected a new approach based on the of period and therefore the general value S = 0.5 will become
concept of tolerable displacements. The method is based on the more conservative as the slope period increases beyond Ts =
work of Bray and Travasarou (2007). They conducted 0.33. For periods shorter than 0.33s, S(T) increases. In such
approximately 55,000 Newmark type slope displacement cases the designer may wish to use a slope specific period.
analyses involving eight different slope configurations, ten
different yield accelerations for each slope configuration, and The ground motions specified by NBCC 2005 are probabilistic.
688 different recorded ground motions from the PEER (2005) Therefore the PGA is not associated with any particular
data base. From a regression analysis of the resulting slope earthquake magnitude but reflects the contributions of all
displacements, they developed an equation for estimating the earthquake magnitudes considered in the probabilistic seismic
median slope displacement along a slip surface with a hazard analysis. The designer has to select an appropriate
conditional probability of exceedance of 50%, if the design magnitude. The TFSSS recommends using the modal magnitude.
ground motion occurs. When this probability is combined with

Finn - Keynote Paper 3

This is the magnitude making the largest contribution to the stability for residential development than the factor of safety
PGA. Site specific values of modal magnitudes may be obtained approach.
from the Geological Survey of Canada, (GSC 2008). Since the
modal magnitudes for BC sites are rarely much larger than M =
7.0, it is suggested that M = 7.0 may be used for all sites. PSEUDO-STATIC ANALYSIS USING A SLOPE
DISPLACEMENT-BASED SEISMIC COEFFICIENT
The parameter ky is the yield coefficient (0.01< ky< 0.5) and is (METHOD 2)
best determined by iterative analyses using commercially
available computer programs. Simplified equations for To continue to allow the use of pseudo-static slope stability
calculating ky may be found in Bray et al. (1998). Bray and analysis and yet retain the advantages of using a displacement
Travasarou (2007) point out that “the primary issue in calculating criterion, the TFSSS asked Bray (2007) to provide a seismic
ky is estimating the dynamic strength of the critical strata within coefficient that would be compatible with the recommended
the slope.” Since ky is assumed to be a constant during limiting displacement of 15 cm of displacement, k15, (Fig. 5).
earthquake shaking, the earth materials in the slope cannot
undergo significant strength loss. The selection of appropriate
shear strength should follow best current practice. An extensive
discussion of the dynamic strength of soil, may be found in
Blake et al. (2002) and Duncan and Wright (2005).
The TFSSS recommends a displacement of 15cm or less as a
tolerable slope displacement along the slip surface for use with
the Bray and Travasarou (2007) method in most cases. This
guideline is based on experience with wood frame construction
and is predicated on the residential building being located back
from the slip surface. The objective is to avoid the slip surface
daylighting within or behind the building.
Fig.5. Pseudo-static analysis with a slope displacement-based
As examples of the use of Eq. 1, displacements were estimated
seismic coefficient.
for soil slopes located in Nanaimo, Duncan, and Victoria, BC,
being considered for development. Slope properties were
Bray estimated this value of k to be that given by:
provided by the geotechnical engineers involved in the projects.
As shown in Table 2, the calculated median slope displacements
k15 = (0.006 + 0.038 M) S(0.50) - 0.026 (3)
(D) are relatively small (2 cm to 13 cm). Using a maximum
allowable displacement of 15 cm, these slopes may be
with S(0.50) < 1.5g
considered suitable for residential development. Note that, in
these examples, site specific site periods, Ts, are used rather than
This regression equation is valid only for a spectral period of
the general value of 0.33. The applicable values for S(1.5Ts) are
0.5s. Therefore slope specific periods cannot be used with this
obtained from the listed NBCC 2005 values in Column 6 by
equation.
interpolation.
M is the moment magnitude of the earthquake. As in the case of
Table 2. Displacements estimated using Equation 1
Eq. 1, modal magnitude, M = 7.0, and spectral acceleration,
Slope H M T PG Sa (T) ky D S(0.50), are acceptable for general use but the designer is not
Location (m) (s) (g NBCC (cm) precluded from selecting a site specific period in determining the
0. 0. 1. period of the spectral acceleration S(T) and using a site specific
Nanaim 3 7 0.3 0.5 1. 0.6 0.3 0.1 1 modal magnitude, obtainable from the Geological Survey of
Dunca 2 7 0.3 0.5 1. 0.7 0.3 0.4 2 Canada (GSC).
Victoria 1 7 0.2 0.6 1. 0.8 0.3 0.5 2
Values for k15 were estimated for the three slopes in Table 3.
Conventional pseudo-static slope stability analysis, with 2% in The k15 values for 2% in 50 year ground motions, and k values
50 year ground motions and k = PGA, shows all three slopes to for k = 0.5(PGA) for 10% in 50-year ground motions, are also
have FS < 1.0 and therefore, typically would be considered shown in Table 3. The values of the slope displacement-based
unsuitable for residential development. Even for k = 0.5(PGA), seismic coefficient (the k15 values) corresponding to 2% in 50-
the Nanaimo slope would have a k = 0.25 which is greater than year ground motions are slightly larger, and therefore somewhat
the yield acceleration, and would be considered unsuitable for more conservative, for these cases, than the seismic coefficient
residential development. The use of displacement analysis in used in association with 10% in 50-year ground motions, when k
conjunction with a criterion for tolerable displacement provides a = 0.5 PGA.
more flexible and less conservative approach to evaluating slope

Finn - Keynote Paper 4

If the pseudo-static analysis, using the slope displacement-based The analyses were conducted using both first and second order
seismic coefficient k15 (Fig. 5) gives a factor of safety FS ≥ 1.0, statistical analysis to assess the impact of uncertainty in the
the slope may be considered suitable for residential development. independent variables controlling D. In this application the
difference between first and second order analyses was
Table 3. Comparison of k15 with k = 0.5(PGA) negligible.

Slope H k15 k=0.5(PGA) The observed displacement of the slope of the Austrian Dam was
Location (m) 2% in 50 yrs 10% in 50 yrs 50cm during the 1989 Loma Prieta Earthquake. The results of the
RELAN analysis are shown in Fig. 6 which shows the
Nanaimo 30 0.16 0.11
conditional probabilities of exceedance of prescribed
Duncan 22 0.18 0.15 displacements Dlim. The observed displacement of 50cm is
Victoria 13 0.20 0.18 . predicted to have a probability of exceedance of 38% for the
specified variations in seismicity and slope parameters. Bray and
Travasarou (2007) estimated the 84% displacement to be 70cm.
ANALYSIS WITH UNCERTAINTY IN VARIABLES RELAN estimates that this displacement has a probability of
exceedance of 30% for the specified uncertainties in the
controlling parameters.
Slope displacement D>1cm is given by Bray and Travasarou
(2007) as

ln D = f [S(T), ky, Ts, M] ± ε (4)

where S (T) = spectral acceleration at the period T = 1.5Ts ; ky =

yield coefficient; Ts = initial period of the potentially sliding
mass and M = earthquake magnitude. These variables are treated

Probability
as deterministic by Bray and Travasarou (2007) in their Eq.1 for
evaluating D. The error term, ε, is the uncertainty in the
displacements for deterministic values of the other independent
variables and has a normal distribution with a mean of 0.0 and a
standard deviation of 0.66.

The performance function for probabilistic analysis of the

likelihood that some limiting displacement Dlim is exceeded for
specified uncertainties in the variables is given by the
performance function: Dlim cm
Fig.6 Probability of exceeding displacement Dlim
G = Ln Dlim – LnD (5)

Dlim is some specified limiting displacement and D is the SUMMARY OF RECOMMENDATIONS

displacement calculated using the Bray and Travasarou (2007)
Eq. 3, taking into account the probabilistic variations in the Two methods for determining whether a slope is suitable for
controlling parameters. residential development, when subjected to 2% in 50-year design
ground motions are recommended.
Reliability analysis of the Austrian Dam in California was
conducted using the program RELAN (Foschi et al. 2007), in ƒ Method 1 involves calculating the median slope
which the performance function G had been inserted. This dam displacement with parameters that reflect slope
was one of many case histories analyzed by Bray and Travasarou properties and local seismicity (Eq. 1). This slope
(2007) in validating their method for estimating seismic displacement has an absolute probability of exceedance
displacement of slopes. of approximately 1/5000 for BC design ground motions.

The variations in slope parameters were prescribed as follows. ƒ Method 2 is based on pseudo-static (limit equilibrium)
The spectral values in NBCC 2005 have a lognormal distribution seismic slope stability analysis, similar to current
with a standard deviation of 0.7. A standard deviation of 0.3 was practice, but uses a slope displacement-based seismic
assumed for magnitude M. Standard deviations equivalent to coefficient, k15, given by Eq. 3, that is equivalent to a
20% of the deterministic values used by Bray and Travasarou tolerable slope displacement of 15cm, when the slope is
(2007) in their analysis of the dam were assumed for the other subjected to design ground motions.
variables, ky and Ts, reflecting the difficulty in defining shear
strength and slope period accurately. The latter three variables
were assumed to have normal distributions.

Finn - Keynote Paper 5

 ƒ Both methods provide the Qualified Professional with a the seismic demand is represented by the cyclic stress ratio, CSR,
basis for exercising his/her judgment as to whether the and the resistance by the normalized Standard Penetration
slope is suitable for residential development. Resistance, (N1)60. The curves shown in Fig. 7 separate
ƒ Based on experience with wood frame residential liquefiable from non-liquefiable sites for a given percentage of
construction a displacement of 15 cm is considered an fines in the sand for a duration corresponding to M = 7.5. Stress
acceptable slope displacement, when the sliding surface ratios on these lines are called cyclic resistance ratios, CRR. The
is between the building foundation and the face of the factor of safety against liquefaction is given by CRR/CSR.
slope.

The results of the above methods, when used in conjunction with

2% in 50-year design ground motions (NBCC 2005), appear to
be comparable to the results obtained by the current methods
using 10% in 50-year ground motions (Provincial Regulation
M268, December 2006) and k = 0.5(PGA).

The use of k = PGA with a factor of safety FS > 1.0 as a basis for
final judgment on slope stability is considered by the TFSSS as
too conservative and is recommended only as a preliminary
screening tool. The limiting displacement of 15 cm is proposed
as a guideline and is not intended to preclude the engineer of
record from selecting any other value that he judges appropriate.
The engineer should strive for a balance between desirable
locations for a building and the associated seismic displacements.

EVALUATION OF LIQUEFACTION POTENTIAL

The impact on the triggering of liquefaction is examined here and

suggestions are made for determining the appropriate compatible
input parameters (magnitude and acceleration) for evaluating the
potential for liquefaction, when probabilistic ground
Fig.7. Liquefaction chart (from Youd et al, 2001)
accelerations are used. These methods are shown to reduce
significantly the seismic demand in some environments. The new
The simplified method was originally used with scenario
seismic parameters are consistent with the hazard level for
earthquakes in California. The design earthquake was usually
seismic design of 2% in 50 years specified in NBCC 2005.
located on a fault and the outcrop acceleration at the site to be
used for site response analysis was determined by an attenuation
The generally accepted procedure in Canada for evaluating the
relationship. There was a direct link between the design
potential for triggering liquefaction is the updated Seed-Idriss
earthquake magnitude and the outcrop acceleration at the site.
(1971) procedure described by Youd et al. (2001). Whether
With the advent of probabilistic ground motion parameters, the
liquefaction occurs or not depends on the balance between the
direct link between site acceleration and design earthquake
resistance to liquefaction of the soil and the seismic demand on
magnitude was lost because the probabilistic site acceleration is
the site represented by the intensity and duration of shaking. The
composed of the contributions of many different earthquakes.
intensity of shaking is defined by the peak ground acceleration
For liquefaction assessment in Canada, the site acceleration has
and the duration is represented by earthquake magnitude.
been assigned to one, somewhat arbitrarily selected, single
Adopting the notation recommended by Youd et al. (2001), the
earthquake magnitude, without any assessment of how well the
seismic demand at a site is termed CSR, the cyclic stress ratio,
acceleration–magnitude pair simulated the combined effects of
and is defined by:
all the earthquakes affecting the site. As will be shown later, this
procedure results often in the probability of triggering
CSR = τav/ σ’vo = 0.65 (amax/g) (σvo/σ’vo) (rd) (6)
liquefaction being lower than the probability of the structural
where amax = peak horizontal ground acceleration at the ground design motions being exceeded and therefore there may be an
surface; g = the acceleration due to gravity; σvo, σ’vo = total and unintentional conservatism in evaluating the potential for
effective vertical overburden stresses respectively, rd = stress triggering liquefaction. The degree of conservatism depends on
reduction coefficient, and τav = average cyclic shear stress. The the seismic environment.
inherent resistance to liquefaction is represented in the Seed-
Idriss method by either penetration resistance or shear wave The duration of shaking depends on the magnitude of the
velocity. Liquefaction potential may be determined from a earthquake as was recognized by Seed and Idriss (1982) when
liquefaction assessment chart such as that shown in Fig. 7. Here they introduced Magnitude Scaling Factors, MSF, to relate the

Finn - Keynote Paper 6

contributions of different magnitudes in generating liquefaction This paper is an update of two previous reports (Finn and
relative to the base magnitude, M = 7.5, which anchors the Wightman, 2006a and 2006b) and incorporates updated
widely used liquefaction assessment chart shown in Fig. 7. deaggregation data for Vancouver and Toronto supplied by
These scaling factors can be applied in two different ways; either Halchuk and Adams (2006) of the Geological Survey of Canada
to the liquefaction resistance or the seismic demand, when (GSC).
assessing the potential for triggering liquefaction. Youd et al.
(2001) described a range of magnitude scaling factors that
geotechnical engineers may adopt for use in practice. In this WEIGHTED MAGNITUDE HAZARD ANALYSIS
paper the factors recommended by Idriss as reported in Youd et
al. (2001) are used. These factors are a lower bound to all the The weighted magnitude probabilistic analysis approach was first
factors recommended by Youd et al. (2001) and their use is more proposed by Idriss (1985). He demonstrated the need for
conservative. These factors for magnitudes M are given in Eq.7 weighting the magnitudes and showed how for the same
in terms of magnitude M = 7.5 which is the base magnitude in acceleration level the return period for the weighted response
Fig. 7. could be much longer depending on the seismic environment. As
noted above, the weighting factors, MWF, used in the present
MSF = 10 2.24/ M 2.56 (7) study are the inverse of the MSF proposed by Youd (2001) and
listed in Table 4.
Some examples of MSF are shown in Table 4.
The weighted magnitude probabilistic analysis is accepted in
Table 4. Idriss magnitude scaling factors (Youd 2001) California as a procedure for implementing the requirements of
the Division of Mines and Geology guidelines in DMG SP 117
Mag. 5.5 6.0 6.5 7.0 7.5 8.0 and the Seismic Mapping Act for projects requiring review under
MSF 2.2 1.76 1.44 1.19 1.0 0.84 the Seismic Mapping Act of California. DMG SP 117 states “The
alternative approach calculating “magnitude-weighted
In this paper the seismic demand is scaled using the magnitude accelerations” is considerably easier and it provides a unique
weighting factor, MWF, where MWF is the inverse of the scaling magnitude to be used with the probabilistically derived
factor. accelerations” (SCEC 1999).

The effect of the magnitude weighting factor on the CSR for a The weighted magnitude probabilistic analyses reported in this
given magnitude is given by Eq.8. paper were conducted to obtain the magnitude–acceleration pair
for evaluating liquefaction potential. In this context, the weighted
CSR = 0.65 (amax/g) (σvo/σ’vo) (rd) (MWF) (8) hazard curves are called liquefaction hazard curves. The seismic
hazard curve for Vancouver and the liquefaction hazard curve
When dealing with a scenario earthquake of magnitude M which weighted for magnitude M = 7.5 are shown in Fig. 8.
has a direct link to the PGA at the site, the MWF for M can be
Annual Frequency of Exceedance

applied directly in Eq.8 without any ambiguity. However, if a 10

-1

probabilistic PGA is used, which is the result of the contributions Vancouver

of many magnitudes, what magnitude and hence what MWF Seismic Hazard

should be used? In current practice a single magnitude is often 10-2 Liquefaction Hazard

selected which may be the maximum experienced earthquake or

tends towards the maximum magnitude expected in the 10
-3

governing seismic source zone and its weighting factor is used

with the NBCC 2005 PGA. Does this single magnitude represent
adequately the collective effects of the many different 10-4
magnitudes contributing to the probabilistic PGA? The answer
to this question is sought using two methods that logically -5
include the effects of weighting on the contributions of all 10
magnitudes to the probabilistic PGA. These methods are: (1) a 0 0.2 0.4 0.6 0.8 1
probabilistic seismic hazard analysis using weighted magnitudes PGA (g)
and (2) a weighted magnitude procedure based on a magnitude
deaggregation for the hazard level in NBCC 2005. The weighted Fig.8. Seismic hazard curves for Vancouver
magnitude probabilistic analyses were conducted using the The acceleration for assessing liquefaction potential for an
computer program EZ-FRISK 4.3 (Risk Engineering, 1997). exceedance rate of 2% in 50 years is 0.30g for M = 7.5 and the
Earthquake magnitudes in Eastern Canada are Nuttli magnitudes, site factor C = 1.0. For other values of C, the compatible
mN and are scaled to moment magnitudes M for liquefaction acceleration is 0.30Cg. The liquefaction hazard acceleration
hazard analysis. should be used directly with the liquefaction resistance curve for
magnitude M = 7.5 without further scaling. As pointed out by
Idriss (1985) the weighted probabilistic analysis can be done for

Finn - Keynote Paper 7

any normalizing earthquake magnitude other than M = 7.5 but
then the appropriate magnitude weighting factor for the chosen
magnitude must be used in assessing liquefaction resistance
using Fig.8. Therefore, when calculating liquefaction triggering
only, the magnitude-acceleration pair to be used is the
normalizing magnitude and the associated weighted acceleration.

The unweighted and weighted PGA are for firm ground and,
depending on the intensity of shaking, will be amplified or
deamplified at the surface by a site factor C on propagating
through the softer soils often associated with liquefaction. The
site factor C is often estimated from generalized amplification
data such in Idriss (1990), the short period amplification factors
in NBCC 2005 or from ground motion attenuation relations for
different soil types. Site response analysis should not be used to
get PGA for use with the simplified Seed-Idriss method. It would
be more reliable in this case to use the computed cyclic stress Fig.9. Magnitude-distance deaggregation for NBCC 2005
ratios from the analysis directly with Fig. 8 to assess liquefaction PGA in Vancouver
potential. The factors of safety against liquefaction presented in
the following table were calculated by the simplified method for
a range in (N1)60 values using the magnitude-acceleration pair
from the weighted magnitude probabilistic analysis. Generic site
conditions were assumed, consisting of sand, with unit weight
20 kN/m3, a water table at 2 m, and a range of (N1)60 values at
6 m depth. For these analyses the site factor was assumed to be C
= 1.0. The factors of safety are shown in Table 5.

Table 5 .Factors of safety against liquefaction in Vancouver

site with (N1)60 = 18 at 6m depth

(N1)60 Vancouver Liquefaction Safety Factors

Weighted
Current Practice Magnitude Analysis
M7.3: 0.46g M7.5: 0.30 Fig.10. Deaggregation matrix for NBCC 2005 PGA in
10 0.28 0.40 Vancouver
13 0.35 0.49
15 0.39 0.57 The total bin contributions to the NBCC 2005 PGA are given by
18 0.47 0.67 the row numbers outside the matrix boundary in Fig. 10. These
20 0.53 0.76 contributions per magnitude bin are shown in the 2-D plot in
25 0.72 1.02 Fig.11. The sum of the bin contributions is 100%.
30 1.15 1.64
The factor of safety against liquefaction at a site, taking into
account the magnitude weighting factors is calculated as follows.
MAGNITUDE DEAGGREGATION METHOD The factor of safety of the site at the code acceleration level is
computed for each binned magnitude and then multiplied by the
The magnitude deaggregation method will be explained with contribution of the magnitude to give the contribution to factor of
reference to the magnitude-deaggregation for Vancouver shown safety. The sum of all the bin contributions to the factor of safety
in Fig. 9 (Halchuk and Adams, 2006). In this case the gives the global factor of safety for the site. The calculation
magnitudes are collected in bins 0.25M wide and the central process for Vancouver is shown by the example in Table 6.
magnitude value is assigned to each of these bins. For example,
the bin labeled M = 5.125 contains all earthquakes in the range
5.0 ≤M <5.25. Contributions of the various bin magnitudes are
magnitudes are collected in bins 0.25M wide and the central
magnitude value is assigned to the bin. For example the bin
labeled M = 5.125 contains all earthquakes in the range
5.0≤M<5.25. The contributions of the bin magnitude are sampled
at various distances from the site. These contributions are shown
by the row numbers in the deaggregation matrix in Fig. 10.

Finn - Keynote Paper 8

% Contribution to Hazard 18 to the M = 7.3 used in Vancouver practice and it also
16 Vancouver Hazard underestimates the factors of safety by about the same amount.
14
Table 7. Factors of safety against liquefaction in Vancouver
12
for various triggering options
10
8 Liquefaction Triggering Safety Factors for Vancouver
6 Current Modal Mean Deaggregation Weighted
(N1)60 Practice Magnitude Magnitude Method Mag. Analysis
4
PGA = 0.46g PGA = 0.30g
2 M=7.3 M=7.1 M=6.3 M=7.25-4.75 M=7.5
0 10 0.28 0.30 0.40 0.41 0.40
13 0.35 0.37 0.50 0.51 0.49
4 5 6 7 8 15 0.39 0.42 0.57 0.58 0.57
Magnitude 18 0.47 0.50 0.68 0.69 0.67
20 0.53 0.56 0.77 0.78 0.76
Fig.11. Magnitude Contributions to NBCC 2005 PGA Hazard
25 0.72 0.76 1.04 1.05 1.02
in Vancouver 30 1.15 1.22 1.66 1.69 1.64

Table 6. Sample calculation of factor of safety against

A deaggregation study was also conducted for Toronto. The GSC
liquefaction for Vancouver site: (N1)60=18 at
magnitude deaggregation for Toronto is shown in Fig. 12 and the
a depth of 6m .
associated deaggregation matrix is shown in Fig. 13 (Halchuk
Magnitude Central Contribution Liquefaction F.S. and Adams, 2006). The equivalent 2-D plot is shown as Fig. 14.
Bins Magnitude Factor F.S. Contribution
4.75 - 5.0 4.875 0.033 1.33 0.044
5.0 - 5.25 5.125 0.045 1.17 0.052
5.25 - 5.5 5.375 0.058 1.03 0.06
5.5 - 5.75 5.625 0.074 0.92 0.068
5.75 - 6.0 5.875 0.091 0.82 0.075
6.0 - 6.25 6.125 0.109 0.74 0.08
6.25 - 6.5 6.375 0.126 0.67 0.084
6.5 - 6.75 6.625 0.143 0.6 0.086
6.75 - 7.0 6.875 0.157 0.55 0.086
7.0 - 7.25 7.125 0.163 0.5 0.082
Σ= 1.00 Total F.S. = 0.717

The factors of safety from the deaggregation method are

compared in Table 7 with the factors obtained using the
magnitude-acceleration pair from the magnitude weighted Fig.12. Magnitude-distance deaggregation for NBCC 2005
probabilistic analysis. The factors given by current practice in PGA in Toronto
Vancouver and those arising from using mean and modal
magnitudes with the code acceleration are also shown. The
weighted magnitude probabilistic method and the deaggregation
method give factors of safety within an average of 2% of each
other. Note that the mean magnitude combined with the NBCC
2005 peak ground accelerations gives results very similar to the
weighted magnitude probabilistic analysis in this seismic
environment.

Deaggregation gives additional information on the statistics of

the seismic environment. Of particular interest are the mean and
modal magnitudes. For Vancouver these are M = 6.32 and M =
7.125 respectively. The mean magnitude in conjunction with the
NBCC 2005 accelerations gives the same factors of safety as the
other methods described above for Vancouver. The modal
magnitude is the event that contributes most to the hazard even Fig.13. Deaggregation matrix for NBCC 2005 PGA in
though it usually contributes less than 25%. For Vancouver, for Toronto
example, it contributes about 16%. The modal magnitude is close

Finn - Keynote Paper 9

The factor of safety for each binned magnitude was calculated sampling data in the two methods. The weighted magnitude
for the previously prescribed range in (N1)60 values using the analysis does not account for epistemic uncertainty because it
Seed-Idriss simplified method. The contribution of each can not be included directly in the EZ-FRISK analyses.
magnitude bin to the total factor of safety was calculated using Therefore “best estimate” seismic parameters given by Halchuk
the contribution data given in Fig. 14. and Adams (2006) are used and the resulting data on liquefaction
hazard for Toronto are scaled by the proportion that the
22 acceleration hazard needs to be scaled to agree with code values.
20 Toronto Hazard It was not necessary to scale the results for Vancouver.
% Contribution to Hazard

18
16 The deaggregation method is based on site deaggregations given
14 by the Geological Survey of Canada (Halchuk and Adams,
12 2006). The analyses leading to these deaggregations include the
10 effects of epistemic uncertainty through the use of three sets of
8 seismic parameters, the best estimates and upper and lower
6
bounds on these estimates. The results from using these three sets
4
are weighted and summed to give the code values for PGA and
2
the associated deaggregations. The effects of epistemic
0
uncertainty vary with the seismic environment.
4 5 6 7 8
Magnitude
Fig.14. Magnitude Contributions to Toronto NBCC 2005 CONCLUDING COMMENTS ON LIQUEFACTION
PGA Hazard
There are two logical methods for incorporating probabilistic
The resulting factors of safety are given in Table 8. The ground accelerations into the Seed-Idriss simplified method for
magnitude deaggregation method gives factors of safety on evaluating liquefaction potential at a site. The most direct method
average 4% greater than the weighted magnitude analysis. The is a probabilistic seismic hazard analysis using weighted
mean magnitude in combination with the NBCC 2005 PGA gives magnitudes. The weighting factors quantify the contributions of
factors of safety for the Toronto site that are similar to the factors different magnitudes to liquefaction potential for a given ground
given by the deaggregation and weighted magnitude methods. In surface acceleration relative to a normalizing magnitude M. The
the Toronto seismic environment, the modal magnitude also normalizing magnitude is usually taken as M = 7.5. The
gives 15% - 20% less to liquefaction potential. These analyses weighting factors for liquefaction assessment may be any of the
were conducted with an amplification⁄deamplification factor C = sets recommended by Youd et al. (2001) as determined by the
1.0 as in the case of Vancouver. geotechnical engineer. In the analyses conducted for this study,
the weighting factors recommended by Idriss are used. These
Table 8. Factors of safety against liquefaction in Toronto for factors are a lower bound on the factors available in Youd et al.
various triggering options (2001).

Liquefaction Triggering Safety Factors for Toronto The weighted magnitude probabilistic analysis gives a unique
magnitude-acceleration pair for use with the Seed-Idriss
Modal Mean Deaggregation Weighted
simplified method. In this study the normalizing magnitude was
(N1)60 Magnitude Magnitude Method Mag. Analysis
taken to be M = 7.5. Any other normalizing magnitude can be
PGA = 0.20g PGA = 0.11g
selected and a compatible magnitude-acceleration pair can be
mN=5.875 mN=5.67 mN=7.0-4.75 mN=7.5 determined by simple proportion of the relative scaling factors
Mw=5.47 Mw=5.204 Mw=7.0-4.75 Mw=7.78 for the magnitudes. All compatible magnitude-acceleration pairs
10 1.33 1.52 1.59 1.5 determined by the weighted probabilistic analysis will yield the
13 1.67 1.9 1.96 1.89 same factor of safety against liquefaction. The probabilistic
15 1.9 2.16 2.24 2.2 acceleration from the weighted magnitude analysis must be
18 2.27 2.59 2.68 2.58 multiplied by the site amplification/deamplification factor, C, to
20 2.54 2.91 3.01 2.92 give the magnitude-acceleration pair to be used in evaluating
25 3.46 3.94 4.08 3.93
liquefaction potential.
30 4.61 5.05 5.29 4.92
The second logical approach is based on a magnitude-distance
deaggregation of the seismic hazard at a site. Here a 2-D
magnitude deaggregation is developed which gives the
ASSESSMENT OF LIQUEFACTION RESULTS
contribution of each magnitude to the probability of exceeding
the NBCC 2005 PGA. The code PGA is first multiplied by the
The factors of safety given by weighted magnitude analysis and
amplification/deamplification factor C. Then the factor of safety
deaggregation analysis are approximately the same. The minor
against liquefaction for each magnitude bin is calculated for the
differences result primarily from the different approaches to

Finn - Keynote Paper 10

modified acceleration and scaled by the contribution of that
magnitude to the hazard. The scaled contributions to the factor of Bray, JD and Travasarou, T [2007]. Simplified procedure for
safety are summed to give the total factor of safety against estimating earthquake-induced deviatoric slope displacements.
liquefaction. This process gives safety factors that are on average Journal of Geotechnical and Environmental Engineering, ASCE,
4% greater than the factors from weighted magnitude Vol. 153, No.4. pp 381-392.
probabilistic analysis. The differences are attributable primarily
to different approaches to sampling the relevant seismic Duncan JD and Wright, SG. [2005]. Soil strength and slope
parameters. stability, Wiley, Hoboken, N. J.

Finn, WD. Liam and Wightman, A. [2006a]. The application of

ACKNOWLEDGEMENTS probabilistic ground motions to liquefaction assessments, Proc.
59th Canadian Geotechnical Conference, CD format, Vancouver,
The critical appraisal and constructive suggestions of members of October 1-4.
TFSSS did much to improve the presentation of the proposed
methods for evaluating the seismic stability of slopes for Finn, WD.Liam and Wightman, A. [2006b]. The application of
residential development. Peter Mitchell of APEGBC chaired probabilistic ground motions to liquefaction assessments, Proc.,
TFSSS and Douglas Van Dine was editorial consultant. Jonathan New Zealand Workshop on Geotechnical Earthquake
Bray advised on all aspects of the Bray and Travasarou method Engineering, University of Canterbury, Christchurch, NZ,
and gave an illuminating address to the Vancouver Geotechnical November 20-21.
Society on the method. Adrian Wightman, BGC Engineering,
was outside technical reviewer of the slope stability work for Foschi, RO., Li, H., Zhang, J. , Folz, B. and Yao, F. [2007].
TFSSS. He also provided technical review and advice on the RELAN and IRELAN V8.0: Software Packages for Reliability
liquefaction section of this paper and, in addition, conducted the Evaluation and Performance-Based Design, Department of Civil
de-aggregation analyses. The reliability analyses were conducted Civil Engineering, University of British Columbia, Vancouver,
in collaboration with Ricardo Foschi. John Adams and Stephen B.C. Canada, V6T 1Z4.
Halchuk of the Geological Survey of Canada provided the input
files for the EZ-FRISK analyses and gave helpful details of the Halchuk, S. and Adams, J. [2006]. Updated deaggregation of
GSC approach to hazard analysis. Helpful discussions held with seismic hazard for selected Canadian cities, Personal
Roy Mayfield, University of Washington and with Gail Atkinson communication to the authors in advance of publication,
of the University of Western Ontario. Somasundaram Geological Survey of Canada, Ottawa.
Sriskandakumar helped with the figures and Juan Carlos Carvajal
helped with figures and text formatting. The author is very Hynes-Griffin, ME. and Franklin, AG. [1984]. Rationalizing the
grateful to these colleagues and friends. Seismic Coefficient Method. Miscellaneous Paper GL-84-13,
United States Army Engineers, WES, Vicksburg, MS.

REFERENCES Idriss, IM. [1985]. Evaluating seismic risk in engineering

practice, Proc. 11th Int. Conf. on Soil Mech. and Found. Engrg.,
Adams, J. [2004]. Personal communication to A. Wightman. Vol 1, 255-320.

Anderson, DL, Byrne, PM, DeVall, RH, Naesgaard, E, and Idriss, IM. [1990]. Response of soft soil sites during earthquakes,
Wijewickreme, D. [2007]. Task Force Report: Geotechnical Proceedings H.B. Seed Memorial Symposium, Editor J. Michael
Design Guidelines for buildings on Liquefiable Sites in Duncan, BiTech Publishers Ltd., Vancouver, BC Canada. Vol.2,
Accordance with NBC 2005 for Greater Vancouver Region, see pp 273-290.
http://www.civil.ubc.ca/liquefaction.
Kramer, SL. and Mayfield, RT. [2005]. Performance-based
BCBC (BC Building Code) [2006]. Published by the Office of liquefaction hazard evaluation, Earthquake Engineering and Soil
Housing and Construction Standards, BC Ministry of Forests and Dynamics, Geotechnical Special Publication GSP 133, ASCE.
Range.
Makdisi, F., and Seed, HB. [1978]. Simplified procedure for
Blake, TF, Hollingsworth, N, Bray, JD and Stewart, JP. [2002]. estimating dam and embankment earthquake-induced
Recommended Procedures for Implementing of DMG Special deformations, Journal of Geotechnical and Environmental
Publication 117, Guidelines for Analyzing and Mitigating Engineering, ASCE, Vol. 104, No.4 , pp 381-392.
Landslide Hazards in California, Southern California Earthquake Newmark, NM. (1965). Effects of earthquakes on dams and
Center, University of Southern California, Los Angeles, CA. embankments, Geotechnique, 15(2) p 139-160.
Bray, JD. [2007]. Private communication to TFSSS.

Bray, J.D., Rathje, E.M., Augello, A.J., and Merry, S.M. [1998].
Simplified seismic design procedures for geosynthetic-lined solid
waste landfills. Geosynthetic. Int., 5(1-2),203-235.

Finn - Keynote Paper 11

NBCC (National Building Code of Canada) [1995]. Published by the
National Research Council of Canada, Ottawa.

NBCC (National Building Code of Canada) [2005]. Published by the

National Research Council of Canada, Ottawa.

Peer [2007]. Pacific Engineering Research Center Database,

Web based at PEER NGA Database.

Risk Engineering. [1997]. EZ-Frisk 4.3, boulder, Colorado.

Seed, H.B.,and Idriss. I.M. [1971]. Simplified procedure for

evaluating soil liquefaction potential. Journal fo Soil Mechanics
and Foundation Engineering,ASCE, 97:SM9, 1249-1273.

Seed, HB. [1979]. Considerations in the earthquake-resistant

design of earth and rockfill dams, Geotechnique, Vol.29/3, pp.
215-263.

Seed, HB., and Idriss, IM. [1982]. Ground motion and soil
liquefaction during earthquakes. Earthquake Engineering
Research Institute Monograph, EERI. Oakland, California.

Sykora DW. and Koester JP. [1988]. "Review of existing

correlations between dynamic shear resistance and standard
penetrations in soils", Proceedings, Earthquake Engineering and
Soil Dynamics II - Recent Advances in Ground Motion
Evaluation, ASCE Geotechnical Engineering Division, Park
City, Utah.

T. L. Youd, Chair Member, ASCE, I. M. Idriss, Co-Chair

Fellow, ASCE, Ronald D. Andrus, Ignacio Arango, Gonzalo
Castro, John T. Christian, Richardo Dobry, W. D. Liam Finn,
Leslie F. Harder Jr., Mary Ellen Hynes, Kenji Ishihara, Joseph P.
Koester, Sam S. C. Liao, William F. Marcuson III, Geoffrey R.
Martin, James K. Mitchell, Yoshiharu Moriwaki, Maurice S.
Power, Peter K. Robertson, Raymond B. Seed, and Kenneth H.
Stokoe II [2001]. Liquefaction Resistance of Soils: Summary
Report from the 1996 NCEER and 1998 NCEER/NSF
Workshops on Evaluation of Liquefaction Resistance of Soils.
Journal of Geotechnical and Geoenvironmental Engineering,
ASCE, V127, No. 10.

Finn - Keynote Paper 12