See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/272012148

Liquefaction induced strength loss and deformation: Simulation and design

Article  in  Bulletin of Earthquake Engineering · June 2013

DOI: 10.1007/s10518-013-9549-x

CITATIONS READS

15 1,322

3 authors:

Ross Boulanger Ronnie Kamai

University of California, Davis Ben-Gurion University of the Negev
180 PUBLICATIONS   6,934 CITATIONS    38 PUBLICATIONS   1,151 CITATIONS   

SEE PROFILE SEE PROFILE

Katerina Ziotopoulou
University of California, Davis
62 PUBLICATIONS   517 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Evaluation of Generic Reference Rock Site Conditions for Israel View project

Ottawa F-65 Sand Characterization View project

All content following this page was uploaded by Katerina Ziotopoulou on 19 February 2015.

The user has requested enhancement of the downloaded file.

 Author's personal copy
Bull Earthquake Eng
DOI 10.1007/s10518-013-9549-x

ORIGINAL RESEARCH PAPER

Liquefaction induced strength loss and deformation:

simulation and design

Ross W. Boulanger · Ronnie Kamai ·

Katerina Ziotopoulou

Received: 17 December 2012 / Accepted: 29 October 2013

© Springer Science+Business Media Dordrecht 2013

Abstract Estimates of earthquake-induced deformations for geotechnical structures affected

by liquefaction can involve significant uncertainties stemming from the uncertainty in the in-
situ residual shear strengths and the effects of pore pressure diffusion (or void redistribution)
on those shear strengths. The results of physical model tests involving liquefiable sands
with lower-permeability interlayers have demonstrated how various factors can influence the
degree to which void redistribution can affect shear strength losses and slope deformations.
Numerical simulations using a critical-state compatible constitutive model are shown to
reproduce the patterns of void redistribution that were observed in two centrifuge model
tests. The constitutive model used in these simulations is described, followed by results for
each of the centrifuge model tests. Current limitations in our abilities to account for void
redistribution in nonlinear deformation analyses are described, followed by a discussion of
how this relates to current design practice for estimating residual shear strengths of liquefied
soils.

Keywords Liquefaction · Residual shear strength · Numerical modeling ·

Constitutive modeling · Physical modeling

1 Introduction

Performance-based design and evaluation of geotechnical structures affected by soil lique-

faction requires engineering procedures that can account for liquefaction-induced strength
loss and deformations. A major source of uncertainty in such analyses is the estimation

R. W. Boulanger (B) · K. Ziotopoulou

Department of Civil and Environmental Engineering, University of California,
Davis, CA, USA
e-mail: rwboulanger@ucdavis.edu

R. Kamai
Department of Structural Engineering, Ben-Gurion University of the Negev,
Beer-Sheva, Israel

123
 Author's personal copy
Bull Earthquake Eng

Fig. 1 Schematic of soil responses near the top and bottom of a liquefiable sand layer overlain by low
permeability soil (after Kulasingam et al. 2004)

of the in-situ residual shear strength (Sr ) for the liquefied soils. Laboratory tests on field
samples have been used to estimate undrained shear strengths (e.g., Robertson et al. 2000;
Poulos et al. 1985; Castro et al. 1992), but the application of such results to design implicitly
assumes that the void ratio of in-situ soils either stays constant or decreases during diffusion
of earthquake-induced excess pore water pressures. Empirical correlations based on back-
analyses of liquefaction flow slide case histories (e.g., Seed 1987; Seed and Harder 1990;
Wride et al. 1999; Olson and Stark 2002; Idriss and Boulanger 2007, 2008; Kramer 2009)
have instead been preferred for practice because they implicitly account for pore pressure dif-
fusion (and associated void redistribution) and other in-situ mechanisms of strength loss that
standard laboratory element testing cannot recreate. Nonlinear deformation analyses have
provided insights on the mechanisms of strength loss associated with pore water pressure
diffusion and void redistribution (e.g., Seid-Karbasi and Byrne 2007), but the ability of such
analysis methods to confidently predict void redistribution effects has not been adequately
demonstrated. For this reason, the application of nonlinear deformation analyses to practi-
cal problems still relies on case history experiences for estimating the in-situ residual shear
strength of liquefied soils.
The mechanism of strength loss due to void redistribution is schematically illustrated
in Fig. 1 for a liquefied soil layer within an infinite slope (Kulasingam et al. 2004). The
liquefied layer is overlain and underlain by lower permeability soils, as shown in Fig. 1a
(after Whitman 1985). The shear stress ratio, stress-path, and void ratio response at the top
and bottom of the sand layer (points A and C, respectively) are illustrated in Fig. 1b–d,
respectively. Earthquake shaking generates excess pore water pressures, which result in an
upward hydraulic gradient that causes pore water to flow up toward the interface with the
overlying low-permeability layer. A thin zone at the top of the sand layer dilates in response
to the net inflow of water and develops shear strains that are related to the imposed volumetric
strains through the sand’s dilation angle. The progressive loosening of this thin zone results
in a shear strain concentration and can result in the sand eventually reaching critical state
(point i). Any further inflow of water then results in the thin zone shedding shear stress and
the slope becoming unstable.

123
 Author's personal copy
Bull Earthquake Eng

The response of liquefied sand to loading conditions like those produced in an infinite
slope (Fig. 1), have been studied through a limited number of laboratory element tests, con-
stitutive modeling studies, and analytical studies. Laboratory element tests on saturated sands
have been performed using controlled volumetric strain paths (water injection or extraction)
to represent the effects of partial drainage conditions in the field (Uchida and Vaid 1994;
Boulanger and Truman 1996; Vaid and Eliadorani 1998; Tokimatsu et al. 2001; Sento et
al. 2004; Yoshimine et al. 2006). The effects of partial drainage on the monotonic, cyclic,
and post-cyclic behavior of liquefied sand was also studied at the element scale by Kamai
and Boulanger (2012) using a critical-state compatible constitutive model. Analytical studies
have related the responses observed in the laboratory to the deformations and strength losses
that would be expected under infinite slope conditions (Boulanger and Truman 1996; Sento
et al. 2004; Malvick et al. 2006; Yoshimine et al. 2006). The results of these laboratory,
constitutive, and analytical studies are consistent with the general interpretation of potential
behavior illustrated in Fig. 1.
This paper examines the mechanism of strength loss due to void redistribution within liq-
uefied soils using results from physical model tests and numerical simulations and discusses
how such effects may be incorporated in design. The first section of this paper discusses
lessons drawn from the results of physical modeling studies, including identification of the
main parameters which affect the potential for void redistribution to cause significant strength
loss. The second section presents results from numerical simulations that provide additional
insights on the void redistribution mechanisms observed in different physical models. Lim-
itations in our existing numerical models and the challenges in numerically simulating the
details of these mechanisms are discussed. The third section examines how the potential
effects of void redistribution on in-situ residual shear strengths may be accounted for in
design correlations and how those relationships may be incorporated in nonlinear numerical
simulations for performance-based design applications.

2 Physical modeling studies

A range of physical modeling studies have illustrated the various factors affecting pore
pressure diffusion and void redistribution in layered liquefied soil profiles (e.g., Liu and
Qiao 1984; Elgamal et al. 1989; Dobry and Liu 1992; Fiegel and Kutter 1994; Kokusho
1999, 2000, 2003; Kulasingam et al. 2004; Malvick et al. 2008; Kamai and Boulanger 2010).
Examples from four of these studies are used herein to illustrate some of the primary factors
and highlight some key features of behavior.
Kokusho (1998) presented 1-g shaking table tests of very loose sands that illustrated how
transient pore water seepage can cause dramatic strength loss and slope deformations after
shaking has ceased. The four images in Fig. 2 are from one of his videos of a fully submerged,
1-g sand model with two horizontal silt seams (the white layers) showing: (a) the geometry
before the start of shaking, (b) the geometry immediately after shaking has stopped and pore
water seepage from the sand below the silt seams is just starting to break up through the silt
seams along the glass side walls of the container, (c) the geometry as the lower silt seam
and overlying sand are weakened by the upward pore water flow and begin to flow both
laterally and down the adjacent slope, and (d) the geometry as deformations are nearing their
final values. These images suggest that the upward pore water seepage from the liquefied
sand layer below the silt seams, after shaking had ceased, was a major factor causing the silt
seams and overlying sand to flow laterally. The third image also shows the lower white silt
seam initially translating out over the adjacent muddy water, illustrating how hydroplaning

123
 Author's personal copy
Bull Earthquake Eng

Fig. 2 Images of shaking table test by Kokusho (1998) at different times: a before the start of shaking, and
b through d as time progressed

and intermixing with the adjacent water can accentuate the flow deformations experienced
by unstable, submerged slopes. These observed mechanisms of strength loss in liquefied
soils (i.e., pore water pressure and void redistribution; hydroplaning; and water entrainment)
are some of the factors that complicate the back-calculation of meaningful strengths in case
histories that involve submerged slide masses.
Kulasingam et al. (2004) reported results of twelve dynamic centrifuge model tests of
simple sand slopes with and without silt interlayers which demonstrated that the potential for
void redistribution to cause strain localizations and associated large deformations in liquefied
soil depends on the initial relative density (DR ), slope geometry (silt layer shape, sand layer
thickness), shaking duration, shaking amplitude, and shaking history. The models were tested
on a 1-m-radius centrifuge at a centrifugal acceleration of 80 g; results are presented in
equivalent prototype units unless otherwise stated. The photographs in Fig. 3 show post-
testing deformed geometries for two models from a set that had inclined silt planes within a
submerged slope of sand. The sand beneath the midpoint of the silt plane was 9 m thick for the
model in Fig. 3a (Geometry II) and only 3 m thick for the model in Fig. 3b (Geometry III). The
models were shaken with three different input motions; motion A was a short duration motion
with a peak acceleration of 0.26–0.33 g, motion B was a much longer duration motion with a
similar peak acceleration of 0.28–0.32 g, and motion C was the same as motion B except it was
scaled to a larger peak acceleration of 0.34–0.42 g. The model in Fig. 3a had an initial DR of
30 % and was shaken first with motion A and then with motion B. Concentrated shear strains
developed beneath the silt plane during shaking with motion B, whereas no concentrated
strains had developed during shaking with the shorter duration motion A. The model in
Fig. 3b had the same initial DR but did not develop any concentrated strains despite being
shaken by motions A, B, and C in sequence. The difference in behaviors between the models
in Fig. 3a, b was attributed to the effect of the sand layer thickness, which affects the amount
of pore water that can be transmitted toward the bottom of the silt interlayer. The results
from several tests with the thicker sand layer (Geometry II) and different combinations of

123
 Author's personal copy
Bull Earthquake Eng

Fig. 3 Post-shaking photos of centrifuge models having the same initial DR = 30 % but different underlying
sand layer thicknesses: a a test with Geometry II after motions A and B, and b a test with Geometry III after
motions A, B, and C (after Kulasingam et al. 2004)

Fig. 4 Displacements across the shear localization zone at mid-slope for centrifuge models with Geometry
II after each earthquake motion (after Kulasingam et al. 2004)

initial DR and shaking history are summarized in Fig. 4 showing the amount of displacement
that occurred across a 0.4-m thick (5-mm thick in model scale) interval immediately below
the silt plane at mid-slope. Large concentrations of strain occurred in this interval (i.e., large
displacements across the interval) for cases with initial DR of 35 % or less subjected to
the longer duration motion B (applied after motion A), whereas concentrated strains only
developed during motion A in one of these cases. Concentrated strains also occurred at the
higher DR = 50 % when the first shaking event was the stronger motion C alone. Concentrated
strains did not develop in: models without a silt layer; dense models with a silt layer shaken
by motion A, B, and C; and models with shallow depths of liquefiable sand beneath the
silt arc (Fig. 3b). An important feature of these results is that a wide range of models did
not develop shear strain concentrations despite the development of almost identically high
excess pore pressures. The combined data set provides a basis for evaluating the ability of
analysis methods to distinguish between conditions where void redistribution will or will not
be severe enough to significantly affect liquefaction-induced deformations.

123
 Author's personal copy
Bull Earthquake Eng

Fig. 5 Localization of shear deformations along a lower permeability interlayer within a saturated sand slope
tested in a 9-m-radius centrifuge (after Malvick et al. 2008)

Malvick et al. (2008) used back-analyses of time series data from a dense array of pore
pressure transducers within a saturated sand slope with embedded silt layers to quantify
several aspects of the pore pressure diffusion and void redistribution process. The test was
performed on a 9-m-radius centrifuge at a centrifugal acceleration of 45 g; results are pre-
sented in equivalent prototype units unless otherwise stated. The larger size of the centrifuge
model facilitated placement of dense arrays of pore pressure transduces at different loca-
tions in the model. The arrangement of pore pressure transducers and a photograph of the
post-testing deformed geometry of the model slope are shown in Fig. 5. The measured pore
pressures in the array were used to compute excess heads, hydraulic gradients, flow rates
and volumetric strains as a function of time and position throughout the slope. Figure 6
shows the pore pressures and volumetric strains calculated for Array 1, but similar analyses
were performed for other arrays as well. Hydraulic gradients were calculated by numerical
differentiation of the measured pore pressure distribution with respect to position along the
array. Flow rates were computed based on Darcy’s Law and integrated with respect to time
to obtain flow quantities, from which volumetric strains within specific intervals were com-
puted. Their analysis results showed that a dilating (loosening) zone of about 1 m or greater
thickness developed in the sand immediately beneath the embedded silt layer during diffusion
of the excess pore pressures (Fig. 6). They concluded that the dilating zone corresponded to
regions where the mobilized friction angle exceeded the critical state friction angle, and that
the dilating zone can be initially relatively thick before its size eventually diminishes to the
thickness of a thin shear band.
Kamai and Boulanger (2010) performed similar back-analyses for a dense array of pore
pressure transducers within a dynamic centrifuge model test involving liquefaction-induced
lateral spreading. The test was performed in a flexible shear beam container on a 9-m radius
centrifuge at a centrifugal acceleration of 15 g; results are presented in equivalent prototype
units unless otherwise stated. The model geometry, as shown in Fig. 7, had two symmetric
sides that sloped toward a central channel. Both sides had a 5-m thick liquefiable layer of
sand (DR = 40 %) overlain by a 1-m thick layer of clayey-silt. One side was treated with

123
 Author's personal copy
Bull Earthquake Eng

Fig. 6 Results of back-analyses of data from the dense array of pore pressure transducers: Excess pore
pressures and volumetric strains at different times (after Malvick et al. 2008)

Fig. 7 Configuration of model SSK01 (some instruments omitted for clarity) tested on a 9-m-radius centrifuge

model-scale plastic drains (Marinucci et al. 2008). The model was subjected to five main
shaking events separated from each other by sufficient time for full dissipation of any excess
pore water pressures. The five successive shaking events consisted of 20 uniform sinusoidal
cycles at 2 Hz with single-amplitudes ranging from 0.01 to 0.3 g. Shear strain concentrations
developed at the interface between the liquefied sand and overlying clayey-silt crust layer on
the untreated sides of these tests, as evidenced by the post-excavation photographs shown in
Fig. 8. The volumetric strain profiles obtained through back-analyses of the time series data
from the vertical array of pore pressure transducers located at the middle of the untreated
side of the model for the shaking event with a peak base acceleration of 0.11 g are shown
in Fig. 9. These back-analyses were performed using two different numerical procedures,

123
 Author's personal copy
Bull Earthquake Eng

Fig. 8 Examples of localizations observed between the liquefied sand and overlying clay crust in models
involving liquefaction-induced lateral spreading (Kamai 2011)

Fig. 9 Results of back-analyses of data from the dense array of pore pressure transducers on the untreated
side: excess pore pressures and volumetric strains at the end of shaking (after Kamai and Boulanger 2010)

one of which was similar to those used by Malvick et al. (2008). These results indicate that
loosening occurred over an approximately 0.6-m-thick interval at the top of the sand layer,
and that the greatest volumetric strains (up to a few percent) likely occurred immediately
near the interface with the clay crust.
The results of these and other physical modeling studies demonstrate that the potential for
void redistribution to cause strain localizations and contribute to deformations in liquefied
soil deposits depends on the initial DR , slope geometry (silt layer shape, sand layer thickness),
shaking duration, shaking amplitude, and shaking history. They furthermore demonstrate that
the potential for significant void redistribution can be expected to decrease as the soil becomes
denser because: (1) the consolidating portions of the deposit release less water because the

123
 Author's personal copy
Bull Earthquake Eng

post-liquefaction reconsolidation strains are smaller, and (2) the dilating portions of the
deposit can accommodate the influx of more water without localizing because the sand has a
greater dilatancy. The combined data set provides a basis for evaluating the ability of analyses
methods to distinguish between those conditions where void redistribution may or may not
be severe enough to significantly affect liquefaction-induced deformations.

3 Simulation of void redistribution

A range of numerical and analytical modeling studies have similarly demonstrated the poten-
tial effects of void redistribution under different conditions (e.g., Balakrishnan 2000; Yoshida
and Finn 2000; Yang and Elgamal 2002; Naesgaard et al. 2005; Seid-Karbasi and Byrne 2007).
Numerical simulations of the void redistribution process are complicated by a number of fac-
tors, including issues of localization length scale (i.e., the thickness of the localization zone)
and the suitability of the constitutive models to track the effects of loosening. Nonetheless,
numerical simulations of this process have provided a number of useful insights.
Results from numerical simulations of two centrifuge model tests are presented to illustrate
their general consistency with the void redistribution mechanisms as observed in the physical
modeling studies and quantified by back-analyses of pore pressure transducer arrays. The
constitutive model used in these analyses is described first, followed by examples of analysis
results for the centrifuge model tests.

3.1 The PM4Sand constitutive model

The formulation, calibration, and implementation of the constitutive model used in these
analyses, PM4Sand, is described in Boulanger and Ziotopoulou (2012, 2013) and Ziotopoulou
and Boulanger (2013). The model follows the basic framework of the stress-ratio controlled,
critical state compatible, bounding-surface plasticity model for sand presented by Dafalias
and Manzari (2004), with a number of modifications and additions as described in Boulanger
and Ziotopoulou (2013). In q–p′ space, the model has a narrow elastic cone and three other
key surfaces—the bounding, dilation and critical-state surfaces (Fig. 10). The locations of
the dilation and bounding surfaces are dependent on the relative state parameter index (ξR )
of the soil (Fig. 11), such that they move towards the critical state surface in a scissor-like
movement when the soil is sheared towards critical state, with all three surfaces coinciding
when the soil reaches critical state (i.e., ξR = 0).
The various modifications and additions included in the PM4Sand model were designed
to improve its ability to approximate the trends observed in empirical correlations commonly
used in practice. For example, Ziotopoulou and Boulanger (2012) describe how components
of the model were modified to improve the ability to simulate the effects of loading duration
and overburden stress on the cyclic resistance ratio of sand. The constitutive modifications
included: adding fabric history and cumulative fabric formation terms; modifying the plastic
modulus relationship and making it dependent on fabric; modifying the dilatancy relation-
ships to provide more distinct control of volumetric contraction versus expansion behavior;
providing a constraint on the dilatancy during volumetric expansion so that it is consistent
with Bolton (1986) dilatancy relationship; recasting the critical state framework to be in
terms of a relative state parameter index (i.e., ξR in Fig. 11) rather than a state parameter; and
providing default values for all but three primary input parameters. The constitutive equations
and model behaviors for the most current version of this model are provided in Boulanger
and Ziotopoulou (2012, 2013).

123
 Author's personal copy
Bull Earthquake Eng

Fig. 10 Schematic of yield, critical, dilatancy, and bounding lines in q–p space (after Dafalias and Manzari
2004)

Fig. 11 Definition of the relative state parameter index, ξR (Boulanger 2003)

There are three primary parameters that are most important for model calibration, and a
secondary set of 16 parameters that may be modified from their default values in special cir-
cumstances. The three primary model input parameters are: (1) the shear modulus coefficient,
Go , which should be calibrated to the estimated or measured in-situ shear wave velocity; (2)
an apparent relative density (DR ) which affects the peak drained and undrained strengths and
the rate of strain accumulation during cyclic loading; and (3) the contraction rate parameter,
hpo which is used to calibrate the model to the estimated in-situ cyclic resistance ratio. Sec-
ondary input parameters have default values that have been calibrated to produce reasonable
agreement with the trends in typical design correlations, although users still need to confirm,
through element loading calibrations, that the default parameters are appropriate for their
particular conditions.
The model was implemented as a user-defined material for use with the commercial
program FLAC (Itasca 2011). Details of the implementation for version 2 of the model,
which was the version used for the simulations presented herein, are given in Boulanger and
Ziotopoulou (2012).

123
 Author's personal copy
Bull Earthquake Eng

Fig. 12 Simulated response for undrained cyclic DSS loading of sand at DR = 55 %

A simulated stress–strain response for undrained cyclic simple shear loading is shown in
Fig. 12 for sand at a DR of 55 %, which would correspond to an SPT (N1 )60 of about 14.
The value for Go was obtained using a form of the correlation by Andrus and Stokoe (2000).
The value for hpo was obtained by calibrating the model to obtain the CRRM=7.5,σ =1 value
computed using the SPT-based liquefaction triggering correlation by Idriss and Boulanger
(2008). All secondary input parameters were assigned their default values. The stress–strain
response reasonably approximates the cyclic mobility behavior of sands, including the pro-
gressive accumulation of shear strains after peak excess pore pressure conditions have been
reached. The ability of the model to accumulate shear strains, rather than lock-up in a repeat-
ing pattern after triggering of liquefaction, was achieved by the addition of cumulative fabric
terms to the model.
Simulated responses for drained strain-controlled cyclic simple shear loading for sand at a
DR of 55 % under vertical consolidation stresses of 1.0, 4.0, and 16.0 atm with Ko = 1.0 are
shown in Fig. 13 with results also shown for the equivalent modulus reduction (G/Gmax ) and
equivalent damping ratio versus cyclic shear strain amplitude. Also shown are the modulus
reduction and equivalent damping ratio curves recommended for sands at different depths by
EPRI (1993). The simulated modulus reduction and equivalent damping ratio curves depend
on the effective confining stress in a pattern and magnitude that is consistent with empirical
design correlations, such as the ones proposed by EPRI (1993). The simulated responses
avoid the problem, common to some other plasticity models, of producing excessively high
equivalent damping ratios as shear strain amplitudes exceed about one percent.

3.2 Centrifuge test of a lateral spread

Numerical simulations of the dynamic centrifuge model test of liquefaction-induced lateral

spreading previously described in Figs. 7, 8 and 9 were performed using PM4Sand with
the program FLAC (Kamai and Boulanger 2011, 2013). Results are presented in prototype
units, as noted previously. The numerical simulations were performed using the two finite
difference meshes shown in Fig. 14. The aluminum and rubber rings of the container are
elastic materials, whose effective masses and properties were proportioned to achieve the
correct ratio of container and soil masses and the correct fundamental frequency of the
physical container (Armstrong 2010). The clayey-silt crust is assigned a Mohr–Coulomb
material model with measured values for its density and undrained strength. The sand layers,
both loose (DR = 40 %) and dense (DR = 80 %), are modeled using PM4Sand, calibrated
to results of in-flight shear wave velocity measurements and cyclic strengths determined
in laboratory element tests. The prefabricated vertical drains on the drain-treated side are
simulated by fixing the pore pressures at vertical columns of nodes to their hydrostatic value.

123
 Author's personal copy
Bull Earthquake Eng

Fig. 13 Simulated response for drained cyclic strain-controlled simple shear loading of sand at DR = 55 %
and corresponding modulus reduction and damping curves

Fig. 14 Finite difference meshes for SSK01: a coarse mesh, and b finer mesh

This approximation of the drains will produce an over prediction of drainage with respect to
the physical model test, but is sufficient for the purpose of this simulation which is focused
on the response of the non-treated side. Mesh node coordinates were updated after each
time step (referred to as large deformation in FLAC terminology) to account for changes in
geometry due to deformations.

123
 Author's personal copy
Bull Earthquake Eng

Fig. 15 Post-shaking deformed shapes and DR contours for the two finite difference meshes for SSK01: a
coarse mesh, and b finer mesh

Simulations of the largest four shaking events were performed (amplitudes of 0.03–
0.3 g) using a baseline set of properties, followed by a sensitivity study covering a range
of cyclic strengths, hydraulic conductivities, and other parameters. The numerical simu-
lations were largely successful in capturing the trends and magnitudes of the recorded
accelerations, pore pressures, and displacements, indicating that the primary features of
soil behavior were simulated reasonably well. In addition, the computed responses showed
a pattern of void redistribution (loosening and densifying zones) that is consistent with
observations from the centrifuge tests. Details of the numerical model, its calibration, and
the comparisons of recorded and computed responses are given in Kamai and Boulanger
(2013).
Computed patterns of void redistribution are illustrated by the results in Fig. 15 showing
the deformed meshes with contours of DR . Note that these more recent simulations were per-
formed using version 2 of PM4Sand (Boulanger and Ziotopoulou 2012) whereas the results
in Kamai (2011) were performed using version 1. The simulation results in Fig. 15a, b are
consistent in showing that, on the untreated side, the DR has decreased (loosened) over the
upper approximately 1.5 m (8 elements in the coarse mesh) of the liquefied sand layer and
increased (densified) over the lower approximately 3.5 m of the sand layer. Furthermore, the
uppermost row of sand elements immediately below the clay crust developed the greatest
volumetric expansion (loosening) and greatest shear strains, consistent with the discontinu-
ities that were observed at the sand–crust interface on the untreated side during excavations
of this and other models (photographs in Fig. 8). The DR of the uppermost row of sand
elements in the middle of the untreated side reduced from 40 % before shaking to about 25 %
after shaking with the coarser mesh (Fig. 15a) and to about 20 % after shaking with the finer
mesh (Fig. 15b). The uppermost row of elements was about 150 mm thick for the coarser
mesh and about 100 mm thick for the finer mesh, such that the differences in post-shaking
DR reflect the unavoidable mesh size effects for this type of problem. On the treated side, the
uppermost sand elements do not experience significant loosening because the drains preclude
the accumulation of water beneath the crust.

123
 Author's personal copy
Bull Earthquake Eng

The finite difference mesh sizes used in these analyses would be expected to underes-
timate the degree of loosening observed in the uppermost row of sand elements based on
comparing the element sizes to the expected dimensions of a shear band. The median particle
size for this sand is about 0.17 mm (model scale) and the thickness of a shear band is com-
monly considered to be approximated 10–20 grain diameters thick (e.g., Vardoulakis 1989).
Thus, the localization zone would be expected to be about 2–4 mm thick in model scale and
about 30–60 mm thick in the prototype scale used for the simulations. The thickness of the
shear localization along the sand-to-crust interface might be expected to be increased by any
irregularities in the contact surface, but the photographs in Fig. 8 suggest that the estimated
thickness of 10–20 grain diameters may be reasonable for these tests. The average thicknesses
of the uppermost row of sand elements in these two meshes are 150 and 100 mm, which are
both greater than the estimated thickness of the actual localization zone (either as estimated
above or observed in dissected sections such as shown by the photographs in Fig. 8). Thus,
the computed values for the volumetric strains and shear strains within the uppermost row of
sand elements are likely smaller than would be obtained if the uppermost elements were equal
in thickness to the eventual shear localization zone. Simulations using thinner elements were,
however, complicated by the fact that only moderate deformations could develop before the
mesh developed element geometry problems. It is noted that Seid-Karbasi and Byrne (2007)
avoided the problems of working with a very fine mesh in FLAC by incorporating a local-
ization length scale in their user-defined material model. This approach will be examined in
future analyses of these centrifuge model tests.
These simulation results are nonetheless consistent with the volumetric strain profiles
obtained through back-analyses of the time series data from the vertical array of pore pressure
transducers located at the middle of the untreated side of the model, as shown previously
in Fig. 9. Both the back-analyses and simulations indicate that loosening occurs over an
approximately 0.6–1.0 m thick interval at the top of the sand layer and that volumetric strains
of several percent likely occurred immediately near the interface with the clay crust.

3.3 Centrifuge test of a slope with embedded silt layers

Simulations were performed for the centrifuge test of a submerged slope of sand with an
embedded silt arc and two embedded horizontal silt planes (Malvick et al. 2008), as shown
previously in Fig. 5. The model was subjected to two shaking events, each being of long-
duration. Post-shaking photographs (e.g., Fig. 5) show that deformations were localized in a
thin shear band at the silt–sand interface.
Simulations were performed using the finite different mesh shown in Fig. 16 with the
PM4Sand model and a baseline set of properties, followed by a sensitivity study covering
a range of cyclic strengths, hydraulic conductivities, mesh size, and other parameters. The
numerical simulations captured the trends and magnitudes of the recorded accelerations and

Fig. 16 Model configuration and finite difference mesh for EJM02

123
 Author's personal copy
Bull Earthquake Eng

Fig. 17 Deformed mesh and contours of DR for two analysis cases that both start with an initial DR = 35 %:
a sand permeability of 0.012 cm/s, and b sand permeability of 0.06 cm/s

pore pressures during shaking reasonably well. The simulations were in reasonable agreement
with the final deformed geometry, but did not reproduce the delayed timing of the post-
shaking slope deformations that were observed in the experiment. Details of the numerical
modeling procedures and comparisons of the simulations with recordings are given in Kamai
(2011).
The sand elements immediately below the silt arc in this mesh were generally about
200 mm thick, which is comparable to the expected dimensions of a potential shear band in
this model. The median particle size for this sand is about 0.17 mm (model scale), such that
a 10–20 grain diameter thick shear band would be 2–4 mm thick in model scale and thus 90–
180 mm thick in the prototype scale used for the simulations. Analyses were also performed
using a slightly more refined mesh, with the differences in results being consistent with those
discussed previously. Considering the approximations involved in estimating the thickness
of the localization zone, the values of the strains within the localized shear zone computed
using the mesh in Fig. 16 are expected to be reasonable for illustrating the mechanism of
void redistribution.
Computed patterns of void redistribution for this model are illustrated by the results in
Fig. 17 for two different values of sand permeability. These more recent simulations were
also performed using version 2 of the PM4Sand model (Boulanger and Ziotopoulou 2012),
whereas the analyses in Kamai (2011) were performed using Version 1 (Boulanger 2010a,b).
The sand permeability for the case shown in Fig. 17a was set equal to the value obtained in
laboratory element tests, whereas the sand permeability was increased by a factor of five for
the case shown in Fig. 17b. This variation of permeability was chosen in part because some
studies have suggested permeability increases markedly when liquefaction is triggered (e.g.,
Haigh et al. 2012) and partly to examine the sensitivity of analysis results to the uncertainty
in input parameters. These two numerical simulations produce patterns of void redistribution
that are consistent with the patterns identified through back-analyses of the data from the
dense array of pore pressure transducers (Fig. 6). The simulation results are consistent in
showing that there is an approximately 1.0–3.0 m thick zone (varies along the arc) of sand
immediately beneath the silt arc that loosens as a consequence of a net inflow of pore water,

123
 Author's personal copy
Bull Earthquake Eng

Fig. 18 Stress and strain paths of two elements from one scenario simulation: Element 1 is in the sand just
beneath the silt interlayer and Element 2 is about 6 m below Element 1. Points A and B represent the end of
shaking for Elements1 and 2, respectively

and that the greatest degree of loosening occurs in a thin zone immediately beneath the silt
arc. For the case shown in Fig. 17a, the sand elements immediately beneath the silt arc near
the middle of the slope loosen from DR = 35 % before shaking to DR ≈ 29 % at the end of
shaking and then to DR ≈ 24.5 % about 300 s after the end of shaking. For the case shown
in Fig. 17b, the same sand elements loosen more rapidly until they reach critical state at
DR ≈ 18 % just before the end of shaking; these elements do not experience significant
changes in DR after shaking for this case, since most of the excess pore pressure dissipates
during shaking due to the higher permeability. Thus, the five-fold increase in permeability
between the models in Fig. 17a, b increased the volumetric strains from about 2.2–3.6 % at
this point in the slope, caused a greater proportion of the strains to occur during shaking
versus after shaking, and also increased overall slope deformations by about 18 %. These
effects are consistent with the simpler analytical and dimensional analysis results obtained
by Kulasingam (2004).
The stresses and strains that were computed for two elements in the mesh for the simulation
with the lower sand permeability are shown in Fig. 18. Element 1 is a sand element just beneath
the silt arc near the middle of the slope, whereas Element 2 is a few meters below Element 1
(below the zone of dilation). Points A and B correspond to the end of shaking for Elements
1 and 2, respectively. The sand at Element 2 developed high excess pore pressures and shear
strains of several percent during shaking, but then progressively contracted (densified) due
to the net outflow of pore water from this zone both during and after shaking. The sand
at Element 1 also develops high excess pore pressures during shaking, but it progressively
loosens under the net inflow of pore water and accordingly develops very large shear strains
under the combined effects of this pore water inflow and continued small levels of shaking.

123
 Author's personal copy
Bull Earthquake Eng

The simulation paths are very consistent with the expected patterns presented schematically
in Fig. 1 by Kulasingam et al. (2004).

3.4 Challenges for modeling void redistribution in practice

There are a number of technical challenges that need to be addressed for improving the ability
to numerically simulate the process of void redistribution and its effects on residual shear
strengths and deformations. The length scale for any eventual localization needs to be con-
sistently accounted for and better understood for conditions where geologic contacts are less
distinct and/or have gradually varying soil properties (e.g., permeability in a fining-upward
sequence). Constitutive models are also needed that can better simulate post-liquefaction
reconsolidation strains because these volumetric strains directly affect the volume of water
given off by consolidating zones. Existing elastic–plastic continuum models generally under-
estimate post-liquefaction reconsolidation strains, which can be expected to cause an under-
estimation of the magnitude of void redistribution and difficulties in simulating delayed slope
deformations. In addition, there are numerical challenges in simulating localization processes
using any continuum model, whether it is void redistribution or cracking of overlying crusts
with associated graben formation during the sliding process. It follows that there remains
an important need to evaluate the extent to which any numerical modeling procedure can
differentiate between cases (e.g., physical model tests) where void redistribution has and
has not led to significant concentrations of shear strains and contributed to the associated
deformations.
Simulating void redistribution effects for field conditions involves even further challenges,
including improving our ability to characterize geologic heterogeneities, geology contacts,
in-situ soil properties, and ground motion characteristics. Geologic details can be expected
to have a major influence on the thickness and/or continuity of any loosening zones and on
the formation of ground cracks or soil boils that may reduce the progression of loosening in
certain zones.

4 Residual strengths in design and evaluation

Engineering practice often requires that estimates be made for the residual shear strength of
liquefiable soils that are denser than those represented in the flow slide case history database.
For many practical problems, the potential failure plane for a slide mass may cut across
materials with a range of (N1 )60 values, such that the analysis of post-earthquake stability
requires that estimates be made for the residual shear resistance for all liquefiable zones
(e.g., Fig. 19). Since the case history database of flow slides only contains cases with (N1 )60
values less than about 14 (e.g., Olson and Stark 2002), the estimation of residual strengths
for liquefied soils with larger (N1 )60 values involves extrapolation.
One challenge in extrapolating our empirical residual strength correlations is quantifying
the effect that void redistribution may or may not have had on the in-situ residual shear
strength for a given case history or field condition. The degree to which void redistribution
may contribute to strength loss depends on numerous factors, as illustrated through the
previously described physical model test results and numerical simulations. Case history-
based relationships implicitly account for void-redistribution effects (Seed 1987), but the
correlation of back-calculated residual shear strengths to pre-earthquake SPT blow counts
or CPT penetration resistances does not provide a basis for differentiating between cases
where void redistribution may or may not be significant. Equally important, the ability of

123
 Author's personal copy
Bull Earthquake Eng

Fig. 19 Schematic of an earth dam with two zones in which liquefaction may be triggered by strong earthquake
shaking (Boulanger 2011)

′ < 400 kPa (Boulanger 2011)

Fig. 20 SPT-based correlation for cohesionless soils at σvc

our current nonlinear analysis procedures to reliably simulate the degree and extent of void
redistribution for realistic geologic conditions has also not been demonstrated. Until we
fully understand and can quantify the mechanisms of in-situ void redistribution and particle
intermixing reasonably well, the risks associated with important structures dictate that we be
cautious in our extrapolation of residual shear strength correlations.
In view of these practical challenges, Idriss and Boulanger (2007, 2008) proposed two
possible scenarios for guiding the extrapolation of residual shear strength ratio (Sr /σ ′ v )
correlations to (N1 )60 values greater than 14, as illustrated in Fig. 20. The first scenario cor-
responds to conditions where the effects of void redistribution are expected to be negligible;
e.g., the field conditions and stratigraphy are such that the post-earthquake pore water seep-
age will not be impeded by lower-permeability layers and thus the dissipation of excess pore
pressures is expected to lead to densification of soils at all depths (in-situ void ratios decrease).
In this case, the in-situ post-liquefaction shear resistance can be expected to be reasonably
represented by relationships developed with consideration of laboratory element test results.
This scenario provided guidance to the estimation of the upper curve, which bends strongly
upward at (N1 )60cs values of 15–17, which Idriss and Boulanger (2007) recommended only
for those cases where void redistribution effects are expected to be negligible.
The second scenario in Fig. 20 corresponds to conditions where the effects of void
redistribution can be significant; e.g., the field conditions and stratigraphy are such that

123
 Author's personal copy
Bull Earthquake Eng

the post-earthquake pore water seepage will, or could be, significantly impeded by an
overlying lower-permeability layer. In this scenario, the potential for void redistribution-
induced loosening and strength loss may be judged to represent a serious possibility. In
such a situation, the potential increases in void ratio near an impeded drainage boundary
would mean that relationships guided primarily by the results of laboratory element tests
could significantly over-estimate the in-situ shearing resistance. For this scenario, Idriss
and Boulanger (2007, 2008) recommended a lower relationship which curves more gently
upward, eventually approaching drained strengths at an (N1 )60cs of about 32; this latter point
on the residual strength curve was selected to agree with the limits of liquefaction triggering
curves.
Nonlinear deformation analyses of geotechnical structures such as embankment dams
often use case-history based correlations for estimating the in-situ residual shear-strengths
of liquefied materials, but the way that the residual shear strengths are introduced to the
numerical model depends on the constitutive model that is being used and certain engineering
judgments. For example, the dynamic response analysis for an earthquake motion may be
computed with the constitutive model first calibrated to produce a cyclic mobility response
consistent with the pre-earthquake in-situ SPT or CPT penetration resistances. At the end
of strong shaking, those elements which have exceeded some user-defined threshold (e.g.,
an excess pore pressure ratio or peak shear strain) may then be re-assigned a total stress
constitutive model (e.g., Mohr Coulomb) having the desired value of residual shear strength.
The analysis is then allowed to progress forward in time to see if the drop to residual shear
strengths results in an increase in post-shaking deformations (e.g., Naesgaard and Byrne
2007; Perlea and Beaty 2010). A thorough review of these and other practical aspects of
evaluating liquefaction effects for embankment dams in practice is provided by Perlea and
Beaty (2010).
The practice of incorporating empirical residual shear strength correlations in post-shaking
nonlinear deformation analyses was evaluated by Kamai (2011) as part of the simulations of
the centrifuge model test shown in Fig. 5. This approach produced deeper seated and larger
than observed deformations, and did not capture the shear strain concentration observed at
the silt–sand interface. These and other results suggest that the indirect procedure of using
empirically-based residual shear strengths in these simulations cannot directly simulate the
more localized effects of void redistribution. Nonetheless, this procedure remains a reasonable
engineering approach for evaluating the potential consequences of void redistribution until
more realistic analysis procedures have been proven effective.

5 Concluding remarks

The results of physical model tests involving liquefiable sands with lower-permeability inter-
layers have demonstrated how various factors can influence the degree to which void redistrib-
ution can affect shear strength losses and slope deformations. The potential for void redistrib-
ution to cause strain localizations and associated deformations in liquefied soil depends on the
soil properties (initial relative density, cyclic resistance ratio, permeability), slope geometry
(layer thicknesses, slope angle, continuity of interfaces), and ground motion characteristics
(shaking intensity, shaking duration, shaking history).
Numerical simulations using the critical-state compatible constitutive model PM4Sand
(Boulanger and Ziotopoulou 2012) with the commercial program FLAC (Itasca 2011) were
shown to reasonably reproduce the patterns of void redistribution that were observed in two
different centrifuge model tests. The numerical simulations provide additional insight on

123
 Author's personal copy
Bull Earthquake Eng

the mechanisms of void redistribution, although a number of factors that currently limit our
ability to simulate void redistribution effects in practice were also identified.
The estimation of earthquake-induced deformations for geotechnical structures affected
by liquefaction involves significant uncertainties, with the estimation of in-situ residual shear
strengths being a major contributor to those uncertainties. It is hoped that an improved under-
standing of void redistribution effects will eventually lead to reduced uncertainties in esti-
mating in-situ residual shear strengths and thus the performance of geotechnical structures.

Acknowledgments Funding for portions of this research was provided the U.S. Geological Survey through
Award G09AP00121. Support for K. Ziotopoulou was provided by the 2008 International Fulbright Science
and Technology Award from the Institute of International Education and the U.S. Department of State. The
authors appreciate the above financial support, the helpful comments and suggestions of I. M. Idriss, and the
assistance of numerous other colleagues with various aspects of the work presented herein.

References

Andrus RD, Stokoe KH (2000) Liquefaction resistance of soils from shear–wave velocity. J Geotech Geoen-
viron Eng ASCE 126(11):1015–1025
Armstrong RJ (2010) Evaluation of the performance of piled bridge abutments affected by liquefaction-induced
ground deformations through centrifuge tests and numerical analysis tools. Ph.D. dissertation, University
of California, Davis, California
Balakrishnan A (2000) Liquefaction remediation at a bridge site. Ph.D, Dissertation, University of California,
Davis, 261 pp
Bolton MD (1986) The strength and dilatancy of sands. Geotechnique 36(1):65–78
Boulanger RW (2010a) Sand plasticity model for nonlinear seismic deformation analyses. In: Fifth interna-
tional conference on recent advances in geotechnical earthquake engineering and soil dynamics. San Diego,
CA, paper IMI-6
Boulanger RW (2010b) A sand plasticity model for earthquake engineering applications. Report No.
UCD/CGM-10-01, Center for Geotechnical Modeling, Department of Civil and Environmental Engineer-
ing, University of California, Davis, California
Boulanger RW (2003) Relating Kα to relative state parameter index. J Geotech Geoenviron Eng ASCE
129(8):770–773
Boulanger RW, Idriss IM (2011) Cyclic failure and liquefaction: current issues. In: Proceedings of 5th inter-
national conference on earthquake geotechnical engineering, Santiago, Chile
Boulanger RW, Truman SP (1996) Void redistribution in sand under post-earthquake loading. Can Geotech J
33(5):829–833
Boulanger RW, Ziotopoulou K (2013) Formulation of a sand plasticity plane–strain model for earthquake
engineering applications. J Soil Dyn Earthq Eng 53:254–267
Boulanger RW, Ziotopoulou K (2012) PM4Sand (Version 2): a sand plasticity model for earthquake engineering
applications. Report No. UCD/CGM-12-01, Center for Geotechnical Modeling, Department of Civil and
Environmental Engineering, University of California, Davis, California, 100 pp
Castro G, Seed RB, Keller TO, Seed HB (1992) Steady-state strength analysis of Lower San Fernando dam
slide. J Geotech Eng ASCE 118(3):406–427
Dafalias YF, Manzari MT (2004) Simple plasticity sand model accounting for fabric change effects. J Eng
Mech ASCE 130(6):622–634
Dobry R, Liu L (1992) Centrifuge modeling of soil liquefaction. In: Proceedings of tenth world conference
on earthquake engineering. Madrid, Spain, pp 6801–6809
Electric Power Research Institute (1993) Guidelines for site specific ground motions, Rept. TR-102293, pp
1–5, Palo Alto, California
Elgamal AW, Dobry R, Adalier K (1989) Study of effect of clay layers on liquefaction of sand deposits using
small-scale models. In: Proceedings of 2nd U.S.-Japan workshop on liquefaction, large ground deformation
and their effects on lifelines. NCEER, SUNNY-Buffalo, Buffalo, pp 233–245
Fiegel GL, Kutter BL (1994) Liquefaction mechanism for layered soils. J Geotech Eng ASCE 120(4):737–755
Haigh SK, Eadington J, Madabhushi SPG (2012) Permeability and stiffness of sands at very low effective
stresses. Geotechnique 62(1):69–75
Idriss IM, Boulanger RW (2007) SPT- and CPT-based relationships for the residual shear strength of liquefied
soils. In: Pitilakis KD (ed) Earthquake geotechnical engineering, 4th international conference on earthquake
geotechnical engineering-invited lectures. Springer, Netherlands, pp 1–22

123
 Author's personal copy
Bull Earthquake Eng

Idriss IM, Boulanger RW (2008) Soil liquefaction during earthquakes. Monograph MNO-12, Earthquake
Engineering Research Institute, Oakland, CA, 261 pp
Itasca (2011) FLAC—Fast Lagrangian Analysis of Continua, Version 7.0, Itasca Consulting Group Inc.,
Minneapolis, Minnesota
Kamai R (2011) Liquefaction-induced shear strain localization processes in layered soil profiles. Ph.D. dis-
sertation, University of California, Davis
Kamai R, Boulanger RW (2010) Characterizing localization processes during liquefaction using inverse analy-
ses of instrumentation arrays. In: Hatzor YH, Sulem J, Vardoulakis I (eds) Meso-scale shear physics in
earthquake and landslide mechanics. CRC Press/Balkema, The Netherlands, pp 219–238
Kamai R, Boulanger RW (2011) Numerical simulations of a centrifuge test to study void redistribution and
shear localization effects. In: Proceedings of 8th international conference on urban earthquake engineering.
Tokyo Institute of Technology, Tokyo
Kamai R, Boulanger RW (2012) Single-element simulations of partial-drainage effects under monotonic and
cyclic loading. Soil Dyn Earthq Eng 35:29–40
Kamai R, Boulanger RW (2013) Simulations of a centrifuge test with lateral spreading and void redistribution
effects. J Geotech Geoenviron Eng ASCE 139(8):1250–1261
Kokusho T (1998) Video of shaking table tests. Geotechnical Engineering Laboratory, Chuo University, http://
www.civil.chuo-u.ac.jp/lab/doshitu/top/top_en.html
Kokusho T (1999) Water film in liquefied sand and its effect on lateral spread. J Geotech Eng ASCE
125(10):817–26
Kokusho T (2000) Mechanism for water film generation and lateral flow in liquefied sand layer. Soils Found
40(5):99–111
Kokusho T (2003) Current state of research on flow failure considering void redistribution in liquefied deposits.
Soil Dyn Earthq Eng 23(7):585–603
Kramer SL (2009) Evaluation of liquefaction hazards in Washington state. Report No. WA-RD 668.1, Wash-
ington State Transportation Center, Seattle, Washington, 325 pp
Kulasingam R (2004) Effects of void redistribution on liquefaction-induced deformations. Ph.D. thesis, Uni-
versity of California, Davis
Kulasingam R, Malvick EJ, Boulanger RW, Kutter BL (2004) Strength loss and localization of silt interlayers
in slopes of liquefied sand. J Geotech Geoenviron Eng 130(11):1192–1202
Liu H, Qiao T (1984) Liquefaction potential of saturated sand deposits underlying foundation of structure. In:
Proceedings of 8th world conference earthquake engineering, Vol. 3. San Francisco, pp 199–206
Malvick EJ, Kutter BL, Boulanger RW, Kulasingam R (2006) Shear localization due to liquefaction-induced
void redistribution in a layered infinite slope. J Geotech Geoenviron Eng ASCE 132(10):1293–1303
Malvick EJ, Kutter BL, Boulanger RW (2008) Postshaking shear strain localization in a centrifuge model of
a saturated sand slope. J Geotech Geoenviron Eng ASCE 134(2):164–174
Marinucci A, Rathje E, Kano S, Kamai R, Conlee C, Howell R, Boulanger RW, Gallagher P (2008) Centrifuge
testing of prefabricated vertical drains for liquefaction remediation. In: Zeng D, Manzari M, Hiltunen D
(eds) Geotechnical earthquake engineering and soil dynamics IV. Geotechnical Special Publication No.
181, ASCE, New York
Naesgaard E, Byrne PM (2007) Flow liquefaction simulation using a combined effective stress: total stress
model. In: 60th Canadian geotechnical conference. Canadian Geotechnical Society, Ottawa
Naesgaard E, Byrne PM, Seid-Karbasi M, Park SS (2005) Modeling flow liquefaction, its mitigation, and
comparison with centrifuge tests. In: Proceedings, performance based design in earthquake geotechnical
engineering: concepts and research, geotechnical earthquake engineering satellite conference. Osaka, Japan,
September 10, pp 95–102
Olson SM, Stark TD (2002) Liquefied strength ratio from liquefaction flow case histories. Can Geotech J
39:629–647
Perlea VG, Beaty MH (2010) Corps of engineers practice in the evaluation of seismic deformation of embank-
ment dams. In: Fifth international conference on recent advances in geotechnical earthquake engineering
and soil dynamics. San Diego, CA, paper SPL 6
Poulos SJ, Castro G, France JW (1985) Liquefaction evaluation procedure. J Geotech Eng ASCE 111(6):772–
91
Robertson PK, Wride CE, List BR, Atukorala U, Biggar KW, Byrne PM, Campanella RG, Cathro CD, Chan
DH, Czajewski K, Finn WDL, Gu WH, Hammamji Y, Hofmann BA, Howie JA, Hughes J, Imrie AS,
Konrad J-M, Kupper A, Law T, Lord ERF, Monahan PA, Morgenstern NR, Phillips R, Piche R, Plewes
HD, Scott D, Sego DC, Sobkowicz J, Stewart RA, Watts BD, Woeller DJ, Youd TL, Zavodni Z (2000) The
Canadian liquefaction experiment: an overview. Can Geotech J 37:499–504
Seed HB (1987) Design problems in soil liquefaction. J Geotech Eng ASCE 113(8):827–45

123
 Author's personal copy
Bull Earthquake Eng

Seed RB, Harder LF (1990) SPT-based analysis of cyclic pore pressure generation and undrained residual
strength. In: Duncan JM (ed) Proceedings of seed memorial symposium. BiTech Publishers, Vancouver,
British Columbia, pp 351–76
Seid-Karbasi M, Byrne PM (2007) Seismic liquefaction, lateral spreading, and flow slides: a numerical inves-
tigation into void redistribution. Can Geotech J 44(7):873–890
Sento N, Kazama M, Uzuoka R, Ohmura H, Ishimaru M (2004) Possibility of postliquefaction flow failure
due to seepage. J Geotech Geoenviron Eng 130(7):707–716
Tokimatsu K, Taya Y, Zhang JM (2001) Effects of pore water redistribution on post-liquefaction deformation
of sands. In: Proceedings of 15th international conference on soil mechanics and geotechnical engineering,
Vol. 1. Balkema, Rotterdam, pp 289–292
Uchida K, Vaid YP (1994) Sand behavior under strain path control. In: Proceedings of 8th international
conference on soil mechanics, and geotechnical engineering, Vol. 1. Balkema, Rotterdam, pp 17–20
Vaid YP, Eliadorani AA (1998) Instability and liquefaction of granular soils under undrained and partially
drained states. Can Geotech J 35:1053–1062
Vardoulakis I (1989) Shear-banding and liquefaction in granular materials on the basis of a Cosserat continuum
theory. Ingenieur-Archiv 59:106–113
Whitman RV (1985) On liquefaction. In: Proceedings of 11th international conference on soil mechanics and
foundation engineering. Balkema, San Francisco, pp 1923–1926
Wride CE, McRoberts EC, Robertson PK (1999) Reconsideration of case histories for estimating undrained
shear strength in sandy soils. Can Geotech J 36:907–933
Yang Z, Elgamal AW (2002) Influence of permeability on liquefaction-induced shear deformation. J Eng Mech
128(7):720–729
Yoshida N, Finn WDL (2000) Simulation of liquefaction beneath an impermeable surface layer. Soil Dyn
Earthq Eng 19:333–338
Yoshimine M, Nishizaki H, Amano K, Hosono Y (2006) Flow deformation of liquefied sand under constant
shear load and its application to analysis of flow slide of infinite slope. Soil Dyn Earthq Eng 26(2–4):253–
264
Ziotopoulou K, Boulanger RW (2012) Constitutive modeling of duration and overburden effects in liquefaction
evaluations. In: Proceedings of second international conference on performance-based design in earthquake
geotechnical engineering. Taormina, Italy, May 28–30, paper 03.10
Ziotopoulou K, Boulanger RW (2013) Calibration and implementation of a sand plasticity plane-strain model
for earthquake engineering applications. J Soil Dyn Earthq Eng 53:268–280

123
View publication stats