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

net/publication/373693925

Energy-Based Liquefaction Triggering Model

Article in Journal of Geotechnical and Geoenvironmental Engineering · November 2023

DOI: 10.1061/JGGEFK.GTENG-11402

CITATIONS READS
11 811

4 authors, including:

Kristin Ulmer Russell A Green

Southwest Research Institute Virginia Tech
22 PUBLICATIONS 91 CITATIONS 239 PUBLICATIONS 4,841 CITATIONS

SEE PROFILE SEE PROFILE

Adrian Rodriguez-Marek
Virginia Tech
161 PUBLICATIONS 4,734 CITATIONS

SEE PROFILE

All content following this page was uploaded by Russell A Green on 05 September 2024.

The user has requested enhancement of the downloaded file.

 Energy-Based Liquefaction Triggering Model
K. J. Ulmer, A.M.ASCE 1; R. A. Green, F.ASCE 2;
A. Rodriguez-Marek, M.ASCE 3; and J. K. Mitchell, Hon.M.ASCE 4

Abstract: The most commonly used approach for evaluating liquefaction triggering is via stress-based simplified models. Proposed herein is
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

a model for evaluating liquefaction triggering where the imposed loading and ability of the soil to resist liquefaction are quantified in terms of
0
normalized dissipated energy per unit volume of soil (ΔW=σvo ), computed within a total stress framework. The proposed model
overcomes limitations of many previously proposed energy-based triggering models. Additionally, the proposed energy-based model unites
concepts from both stress-based and strain-based procedures, overcoming some of their limitations, and in its simplified form is implemented
similarly to the simplified stress-based models. An updated field case history database is used to develop probabilistic limit-state curves.
0
These limit-state curves express ΔW=σvo required to trigger liquefaction as a function of corrected cone penetration test tip resistance (qc1Ncs )
for different probabilities of liquefaction (PL ) and have comparable predictive abilities to stress-based limit-state curves in terms of number
of correct predictions for the cases analyzed. However, because dissipated energy is a scalar quantity, multidirectional shaking and other
effects such as soil–structure interaction, nonvertical wave fields, and topographic site effects can readily be accounted for. Additionally, the
applicability of the proposed triggering curve is not limited to earthquake loading but, rather, can be used in relation to other sources of
vibrations (e.g., construction vibrations and explosive loading, among others). DOI: 10.1061/JGGEFK.GTENG-11402. © 2023 American
Society of Civil Engineers.
Author keywords: Liquefaction; Liquefaction triggering; Energy-based triggering model; Limit-state curves; Maximum likelihood.

Introduction simplification of the characteristics of the imposed loading makes

the stress-based approach attractive because of its relative ease of im-
The most commonly used approach for evaluating liquefaction trig- plementation, but to some extent, it inherently limits the implemen-
gering is via simplified stress-based models, where the imposed tation scalability of the procedure, particularly in terms of duration of
loading and the ability of the soil to resist liquefaction are expressed shaking, and limits the applicability of the procedure exclusively to
in terms of cyclic shear stress and resistance ratios for a set of refer- earthquake loading. In this context, implementation scalability refers
ence conditions. However, the simplified stress-based framework has to the ability of the procedure to be implemented using refined es-
some drawbacks, some of which result from the oversimplification timates of the imposed loading and/or the ability of the soil to resist
of the imposed loading and the imposed reference conditions. Spe- liquefaction triggering determined from numerical and/or laboratory
cifically, the imposed loading is expressed in terms of normalized studies. Quantifying ground motion duration in terms of MSF makes
cyclic stress ratio (CSR*), which can be divided into components it difficult to apply the procedure to evaluate liquefaction due to non-
that represent the amplitude and duration of the loading. The ampli- earthquake vibrations (e.g., construction vibrations).
tude of the loading is proportional to the peak horizontal acceleration The characteristics of earthquake motions range widely, even for
at the ground surface (amax ), and the duration of the loading is ex- motions having similar values of amax and M. As a result, alterna-
pressed in terms of the magnitude scaling factor (MSF). tive intensity measures (IM) to CSR* have been proposed for quan-
In turn, MSF is a function of the number of equivalent loading tifying the imposed loading for evaluating liquefaction triggering
cycles (neq;M ) for ground motions from an earthquake with mo- that better capture the characteristics of earthquake shaking. These
ment magnitude (M) relative to neq;M for a moment magnitude include Arias intensity (e.g., Kayen and Mitchell 1997), cumulative
7.5 (M7.5) earthquake, neq;M7.5 (i.e., the reference event). This absolute velocity (e.g., Kramer and Mitchell 2006), and other more
1
direct measures of the energy associated with the imposed load-
Research Engineer, Dept. of Geosciences and Engineering, Southwest ing on the soil [e.g., dissipated energy (Berrill and Davis 1985)].
Research Institute, San Antonio, TX 78238. ORCID: https://orcid.org
However, as elaborated upon subsequently in this paper, these
/0000-0001-8696-7447. Email: [email protected]
2
Professor, Dept. of Civil and Environmental Engineering, Virginia
alternative IMs and associated triggering models have their own
Tech, Blacksburg, VA 24061 (corresponding author). ORCID: https:// shortcomings, including the introduction of erroneous variable
orcid.org/0000-0002-5648-2331. Email: [email protected] dependencies, inconsistencies in the quantification of the imposed
3 loading in a total stress framework and the ability of the soil to
Professor, Dept. of Civil and Environmental Engineering, Virginia
Tech, Blacksburg, VA 24061. ORCID: https://orcid.org/0000-0002-8384 resist liquefaction in an effective stress framework, large uncer-
-4721. Email: [email protected] tainty associated with predicting the IMs, and more pronounced
4
University Distinguished Professor Emeritus, Dept. of Civil and issues with implementation scalability than the simplified stress-
Environmental Engineering, Virginia Tech, Blacksburg, VA 24061. Email: based procedure that they were developed to replace.
[email protected]
This paper proposes an energy-based liquefaction triggering
Note. This manuscript was submitted on October 7, 2022; approved on
July 6, 2023; published online on September 6, 2023. Discussion period model that quantifies the imposed loading and ability of the soil
open until February 6, 2024; separate discussions must be submitted for to resist liquefaction in terms of normalized dissipated energy per
individual papers. This paper is part of the Journal of Geotechnical unit volume of soil, both computed within a total stress frame-
and Geoenvironmental Engineering, © ASCE, ISSN 1090-0241. work. Dissipated energy better captures the characteristics of

© ASCE 04023105-1 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 ground motions than CSR* and has been shown to correlate well
with excess pore-water pressure generation. From a mechanics
perspective, the correlation of dissipated energy and excess pore-
water pressure generation is expected because the predominant
mechanism for energy dissipation in the soil is frictional, indi-
cating relative movements of soil particles (e.g., Green 2001;
Polito et al. 2013). This relative movement of soil particles is also
the mechanism for excess pore-water generation. Furthermore,
because dissipated energy is a scalar quantity, multidirectional
shaking and other effects such as soil–structure interaction, non-
vertical wave fields, and topographic site effects can readily be
accounted for (e.g., Ostadan et al. 1998).
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

In its simplified form, the implementation of the proposed

energy-based model is similar to the stress-based simplified mod-
els. However, implementation of the procedure is scalable, in that
refined estimates of the loading or ability of the soil to resist lique- Fig. 1. Shear stress–shear strain hysteresis loops at a depth of 2.9 m
faction can be made via numerical site response analyses and/or in the Imperial Valley Wildlife Site profile for the 1987 M6.6
laboratory testing. Such refinements are allowed as long as consis- Superstitions Hills earthquake. (Reprinted from Zeghal and Elgamal
tency is maintained in how the simplified form of the proposed 1994, © ASCE.)
energy-based model was developed and the refined estimates of
the loading and ability of the soil to resist liquefaction are made
(e.g., equivalent linear numerical analyses using the same shear
modulus reduction and damping curves). Additionally, similar to Berrill and Davis 1985; Figueroa et al. 1994; Green et al. 2000;
the stress-based simplified models, the ability of the soil to resist Davis and Berrill 2001; Jafarian et al. 2012; Polito et al. 2008;
liquefaction is derived from field case histories and correlated to Kokusho and Kaneko 2018; Kokusho and Tanimoto 2021; among
in situ test metrics (i.e., normalized cone penetration test (CPT) many others). Additionally, in its simplified form, the predictive
tip resistance, qc1Ncs ). This avoids issues regarding the representa- variables for estimating ΔW are similar to those used to compute
tiveness of laboratory test data from reconstituted samples to in situ the CSR*; hence all the advances that have been made in estimat-
soil fabric and the need for correlations relating the relative density ing the predictive variables for CSR* can be directly leveraged in
(Dr ) of laboratory specimens to in situ test metrics. Finally, because predicting ΔW.
normalized dissipated energy is a comprehensive and versatile IM, The energy dissipated per unit volume of soil (ΔW) is the differ-
the applicability of the proposed triggering curve is not limited ence between the shear strain energy input during the loading phase
to earthquake loading but, rather, can be used in relation to other of a cycle and the elastic shear strain energy returned during the
sources of vibrations (e.g., construction vibrations and explosive unloading phase of the cycle. Accordingly, ΔW can be computed
loading, among others). as the cumulative area bound by the shear stress–shear strain hys-
In the following, background information regarding energy- teresis loops such as those shown in Fig. 1.
based liquefaction triggering models and associated shortcomings As such, ΔW reflects the amplitude and duration of the motion, to
is presented, followed by specifics of the proposed energy-based include the variation of the amplitude of the motion over the duration
model. Toward the development of the triggering curve, modifica- of shaking, as well as the element-level soil response to the shaking.
tions made to an existing liquefaction field case history database ΔW fundamentally differs from other, less direct, measures of ground
and how the case histories were analyzed to develop probabilistic motion energy (e.g., Arias intensity I A and CAV). I A and CAV are
limit-state curves are detailed. The implementation scalability and proportional to the square of ground acceleration and the absolute
applicability of the proposed model are then discussed, as well as value of the ground velocity, respectively, integrated over the duration
comparisons of the proposed energy-based framework with that of of the shaking. Because I A and CAV are computed as integrals over
simplified stress-based models. time, they inherently introduce an erroneous dependency on ground
motion frequency into the liquefaction models, conflating profile and
soil element responses (Green and Mitchell 2003); this is elaborated
Background on in the “Discussion” section of this paper. In contrast, ΔW is com-
puted by integrating the imposed shear stress over the induced shear
Development of energy-based methods for evaluating liquefaction strain in a soil element to quantify the loading on an element of soil,
began in the 1970s as an alternative to stress-based procedures circumventing the frequency-dependency issue and the conflagration
(e.g., Nemat-Nasser and Shokooh 1979). Since then, liquefaction of profile and soil element responses.
triggering models that use varying IMs have been broadly classi- A final, but important consideration, is that liquefaction is in-
fied as energy-based procedures, to include models based on IMs herently a strain phenomenon (e.g., Martin et al. 1975; Dobry et al.
such as Arias intensity (I A ), cumulative absolute velocity (CAV), 1982); however, simplified strain-based procedures (e.g., Dobry
elastic strain energy, and energy dissipated per unit volume of soil et al. 1982) have issues with the proper estimation of ampli-
(ΔW). Although there are merits and shortcomings with any given tude of the induced shear strain and the ground motion duration
IM, the authors contend that ΔW offers advantages over other (Rodriguez-Arriaga and Green 2018; Green and Rodriguez-Arriaga
IMs, with few drawbacks. The reason for this is that both the 2019). The use of ΔW as an IM merges aspects of the simplified
generation of excess pore pressures and energy dissipation result stress- and strain-based liquefaction triggering concepts into
from the breakdown in the soil skeleton. As a result, dissipated one model, with the advantage that the use of ΔW as the IM over-
energy in soil can be directly correlated to the generation of excess comes issues with quantifying the ground motion characteristics in
pore pressures throughout the loading and deformation process terms of shear strain required by the simplified strain-based
(e.g., Nemat-Nasser and Shokooh 1979; Simcock et al. 1983; procedures.

© ASCE 04023105-2 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 Energy-Based IM other proposed relationships because it is derived using dissipated
energy concepts. The equation for the Lasley et al. (2017) neq;M
The authors propose using ΔW computed within a total stress relationship is detailed subsequently in this paper.
framework for their IM, where ΔW is computed by integrating Also, the relationship between ΔW and the more tradition-
the imposed shear stress over the induced shear strain in a soil ally used liquefaction triggering IM, CSR*, is discussed in the
element for the duration of earthquake shaking. However, in the “Discussion” section of this paper. However, it is emphasized that
spirit of implementation scalability, ΔW can be estimated via a ΔW is not a just a scaled variant of CSR*. The reason for this is
simplified approach. Toward this end, the ground motions param- because the damping and shear modulus factors in Eq. (5) [i.e., Dγc
eters that are used in simplified stress-based liquefaction triggering and Gmax ðG=Gmax Þγc  are not constants, but vary as a function of
models (i.e., amax and neq;M ) are used in conjunction with soil stiff- the induced shear strain, γ c , which in turn will be a function of the
ness and shear modulus reduction and damping (MRD) curves for imposed shear stress, stiffness of the soil, shear stress–shear strain
the soil (e.g., Ishibashi and Zhang 1993; Darendeli 2001) to esti-
response of the soil, and so on. As a result, ΔW inherently takes
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

mate ΔW. Specifically, amax and the MRD curves are used to es-
into account certain aspects of the influence of initial effective
timate the dissipated energy in one equivalent cycle (ΔW 1;eq ). An
confining stress and aging effects on liquefaction triggering, which
equation to estimate ΔW 1;eq can be derived in terms of the damping
need to be accounted for by additional relationships when CSR is
ratio for a viscous material (D) (Jacobsen 1930) as follows:
used as the IM.
ΔW 1;eq
D¼ ð1Þ
4πW stored
Case History Database
where W stored = elastic stored energy, equal to 0.5·τ ·γ, where τ
and γ are the imposed shear stress and the induced shear strain, In the absence of a liquefaction case history database that contains
respectively. Rearranging terms and letting τ ¼ τ avg , where τ avg ΔW information determined using a more rigorous approach, the
is the representative (or average) amplitude of shear stress of the simplified equation for estimating ΔW [i.e., Eq. (5)] is used in con-
imposed loading and γ ¼ γ c ¼ τ avg =Gγc , where Gγc is the secant junction the Boulanger and Idriss (2014) (BI14]) liquefaction case
shear modulus corresponding to γ c , the equation for ΔW 1;eq then history database to develop an energy-based triggering model. This
becomes database was originally compiled and used to derive the Boulanger
and Idriss (2016) simplified stress-based triggering model. Lique-
2πDγ c ðτ avg Þ2 faction case history databases have evolved over time, particularly
ΔW 1;eq ¼ ð2Þ
Gγ c following each significant earthquake in which liquefaction was
observed. In recent years, a collaborative effort by the next gener-
where Dγc = damping ratio corresponding to γ c , again determined ation liquefaction (NGL) project has led to the development of an
from the MRD curves. The value of τ avg can be estimated in the open database of liquefaction case histories (Brandenberg et al.
same way as in the simplified stress-based liquefaction models 2020; Zimmaro et al. 2019). At the time that the study herein was
(Seed and Idriss 1971) performed, the NGL database was not yet fully populated with case
amax histories nor vetted by the community. In addition, the objective
τ avg ¼ 0.65 · · σ v · rd ð3Þ data included in the NGL database do not provide recommenda-
g
tions for critical layer selections (i.e., the layer most likely to
where g = acceleration due to gravity in the same units as amax ; liquefy in a given profile). Accordingly, the authors have chosen
σv = total vertical stress at the depth of interest in the same units to use the BI14 liquefaction case history database in this study.
as τ avg ; and rd = dimensionless stress reduction factor that accounts The BI14 database contains 253 case histories with estimates
for the nonrigid response of the soil profile. Gγc can be estimated of earthquake parameters (e.g., amax and M), soil parameters asso-
as Gmax ·ðG=Gmax Þγc , where G is the secant shear modulus corre- ciated with the critical layer (e.g., CPT metrics, σv , initial vertical
0
sponding to γ c (i.e., Gγc ), and Gmax is the small-strain shear modu- effective stress σvo , and depth to the center of the critical layer z),
lus of the soil. The values of Dγc and ðG=Gmax Þγc can be estimated and observations of liquefaction response (Yes, No, or Marginal).
iteratively using MRD curves (e.g., Ishibashi and Zhang 1993; Traditionally, these labels have been interpreted as Yes indicating
Darendeli 2001), where γ c can be estimated as follows (e.g., Dobry that liquefaction having been triggered, No as indicating liquefac-
et al. 1982): tion not having been triggered, and Marginal as a case that falls
somewhere in between the Yes and No cases, although the criteria
amax
g · σ v · rd used to distinguish Marginal cases from Yes or No cases are ambig-
γ c ¼ 0.65 ð4Þ
Gmax ðG=Gmax Þγc uous across various studies (NRC 2016). Setting the ambiguity in
the definition of Marginal cases aside for now, Upadhyaya et al.
Finally, dissipated energy per unit volume of soil for the entire (2022) rightfully contended that the Yes and No cases can be more
duration of shaking can be estimated as ΔW 1;eq multiplied by the accurately interpreted as cases in which surficial liquefaction man-
number of equivalent loading cycles (neq;M ) for an earthquake of ifestations (e.g., sand boils and lateral spread displacements) were
magnitude M: and were not observed following an earthquake. This alternative
interpretation of the case history categories is important in evalu-
2πDγc ½0.65 · amax · σ v · rd 2 ating the efficacy of proposed liquefaction triggering models, as
ΔW ¼ ΔW 1;eq · neq;M ¼
gG  · neq;M ð5Þ
Gmax · Gmax γc discussed subsequently.
As detailed in the Supplemental Materials, for some of the case
where neq;M can be estimated for earthquake ground motions using histories compiled in BI14, we revised the interpretation of param-
established correlations with M, amax , tectonic setting, and so on eters for the critical layers: depth, in situ soil stresses, fines content
(e.g., Lasley et al. 2017). In this regard, the neq;M relationship pro- (FC), and qc1Ncs . The values of amax and M were also reviewed for
posed by Lasley et al. (2017) is specifically recommended over the compiled case histories.

© ASCE 04023105-3 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 Analysis of Case Histories stress-dependency term in the IZ93 relationship was in better ac-
cord with cyclic laboratory data than those for other MRD relation-
Based on the findings of Lasley (2015) and Ulmer (2019), ΔW is ships examined.
0
normalized by σvo to account for the influence of initial effective Use of the IZ93 shear modulus reduction and damping curves
confining stress on liquefaction. The sufficiency of this stress nor- require estimated values of the plasticity index (PI) and initial mean
malization is discussed in more detail in the “Discussion” section of 0 ) as input variables. For some case histories, the
effective stress (σmo
this paper. The following sections outline the relationships used to PI of the critical layer was measured, but for other case histories
0
compute ΔW=σvo for the case histories in the database. where PI is unknown but assumed to be zero (i.e., nonplastic), the
value of σmo0 was computed as [ð1 þ 2K Þ=3·σ 0 , where K is the
o vo o
Stress Reduction Factor, r d at-rest lateral earth pressure coefficient. A limited parametric study
showed that the choice of K o had negligible influence on the com-
The dimensionless stress reduction factor, rd , accounts for the non- puted values of ΔW, and as a result, for simplicity, K o was assumed
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

rigid response of the soil profile when subjected to earthquake to be 0.5 for all case histories.
shaking. Because the case histories in the database were predomi-
nantly located in shallow-crustal active tectonic regimes, the values
of rd are estimated as a function of z and M using a relationship Small-Strain Shear Modulus: G max
specifically developed for shallow-crustal events in active tectonic
The values of Gmax were estimated for the soil in situ as Gmax ¼
regimes [e.g., western US (WUS] as follows (Lasley et al. 2016):
ðγ t;soil =gÞ · V 2S , where γ t;soil is the total unit weight of the soil, g is
  the acceleration due to gravity, and V S is the small-strain shear-
z
rd ¼ ð1 − αÞ exp − þα ð6aÞ wave velocity. If V S of the soil is not measured in situ, it can be
β
estimated for young deposits using a correlation with standard pen-
α ¼ expð−4.373 þ 0.4491 · MÞ ð6bÞ etration tests (SPT) or CPT data. There are many such correlations
in the literature (e.g., Wair et al. 2012), but it is not certain whether
β ¼ −20.11 þ 6.247 · M ð6cÞ these provide consistent or biased values of V S if SPT versus CPT
data were used, or vice versa. Ulmer et al. (2020) regressed a rela-
where z is measured in meters. tionship such that stress-based liquefaction resistance curves based
on V S , CPT, and SPT metrics align. The relationship specific for
CPT data is
Number of Equivalent Cycles, n eq;M
 0 0.25
The value of neq;M is estimated using a relationship developed by σ
V S ¼ 16.88ðqc1Ncs Þ0.489 vo ð8Þ
Lasley et al. (2017). The relationship was developed using a low- Pa
cycle implementation of the Palmgren-Miner fatigue theory (Green
and Terri 2005) and accounts for plastic response of the soil and for where V S is in meters per second; and Pa = atmospheric pressure in
0 . For this study, G
the same units as σvo
multidirectional shaking. The recommended form of the equation max was estimated using this
used in this study is relationship assuming γ t;soil equal to 17.0 and 19.5 kN=m3 above
and below the groundwater table, respectively (e.g., Moss 2003;
lnðneq;M Þ ¼ 0.4605 − 0.4082 lnðamax Þ þ 0.2332M ð7Þ Green et al. 2014) for all case histories in the database (Table S1).
Table 1 provides a summary of how each input parameter was ob-
where amax is in g. As with the rd correlation, the relationship for tained for this study.
shallow-crustal active tectonic regimes proposed by Lasley et al.
(2017) is used to analyze the case histories.
Input Parameter Uncertainties
Dynamic Soil Properties: G γc and D γc To develop a probabilistic limit-state curve for the proposed
energy-based method, the uncertainties associated with each input
The values for Gγc and Dγc are solved iteratively using MRD curves.
parameter must be addressed, including an estimate of the uncer-
The iterative process entails assuming a value for ðG=Gmax Þγc in the 0 0
tainty in qc1Ncs and ΔW=σvo . It was assumed that ΔW=σvo was
first iteration and a value of γ c calculated using Eq. (4). In sub-
lognormally distributed with a standard deviation of σlnðΔW=σvo0 Þ .
sequent iterations, the ratio of ðG=Gmax Þγc is obtained from a 0
The equation for ΔW=σvo required to trigger liquefaction, written
shear modulus reduction curve (e.g., Ishibashi and Zhang 1993;
as a natural log, is
Darendeli 2001) corresponding to the γ c calculated in the previous
 
iteration. This process is repeated until the assumed ðG=Gmax Þγc ΔW
value converges to the MRD curve, similar to the iterative process ln 0 ¼ lnð2πg · 0.652 Þ þ lnðDγc Þ þ 2 lnðamax Þ þ 2 lnðσv Þ
σvo
used in equivalent linear site response analyses [e.g., SHAKE91 0 Þ
(Idriss and Sun 1992)]. Gγc is computed as Gmax ðG=Gmax Þγc . þ 2 lnðrd Þ þ lnðneq;M Þ − lnðγ t;soil Þ − lnðσvo
  
The value of Dγc is selected from a compatible damping curve at G
the same final value of γ c . − 2 lnðV S Þ − ln ð9Þ
Gmax γc
In this study, the Ishibashi and Zhang (1993) (IZ93) curves are
used to estimate (G=Gmax Þγc and Dγc . The reason for this is that the The value of σlnðΔW=σ 0 voÞ was estimated using a first-order,
functional form of the IZ93 equations captures the τ − γ response second-moment estimation method using the uncertainties of each
of the soil across all strains of interest better than the modulus re- 0
input parameter: Dγc ; lnðamax Þ, σv , rd , lnðneq;M Þ, γ t;soil , σvo ,
duction curves that use a hyperbolic function as their base equation lnðV S Þ, and (G=Gmax Þγc . The matrix notation for this first-order
(e.g., Darendeli 2001; Menq 2003). For these latter curves, a mod- approximation is
erate-to-large strain strength correction is often applied to better
capture the τ -γ response of the soil for larger strains (e.g., Yee σ2
ΔW  ¼ ∇T Σxx ∇ ð10aÞ
ln 0
et al. 2013). Moreover, Green et al. (2022) found that the effective σvo

© ASCE 04023105-4 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 2   3
∂ ΔW
6 ∂Dγc
ln 0
σvo 7
6 7
6 7
6  7
6 ∂ ΔW 7
6 ln 7
6 ∂ ln amax σvo0 7
∇¼6
6
7
7 ð10bÞ
6 .. 7
6 7
6 . 7
6  7
6 ∂ ΔW 7
4 5
ln 0
∂ðG=Gmax Þγc σvo

2 3
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

σ2D ρD;ln amax σD σln amax ··· ρD;G=Gmax σD σG=Gmax

6 7
6 7
6 .. 7
6 ρD;ln a σD σln a σ2ln amax ··· . 7
6 7
¼6 7
max max
Σxx 6 7 ð10cÞ
6 .. .. .. .. 7
6 . . . . 7
6 7
4 5
ρD;G=Gmax σD σG=Gmax ··· ··· σ2G=Gmax

where the value of ρ in each term is the correlation coefficient of The average COV from all of the soundings in the database was
the pair of variables represented in the term. All variables were as- reasonably constant with depth and was approximately 0.093.
sumed to be normally distributed except for V S, amax , and neq;M , When FC is estimated from I c , instead of measured in a laboratory,
which were assumed to be lognormally distributed. The value of the COV is increased by a factor of 1.5 (e.g., Boulanger and Idriss
γ t;soil was an assumed constant for all case histories, and thus 2014). This estimate of COV represents variability in qc1Ncs verti-
its uncertainty was not considered. Table 2 summarizes the stan- cally within the thickness of a single soil layer detected from a CPT
dard deviations and correlation coefficients required to estimate sounding and does not account for the uncertainties due to lateral
σlnðΔW=σ 0 voÞ , where all variables are assumed to be uncorrelated ex- variability or measurement error.
cept for the pairs of variables with values of ρ included in Table 2. Estimates of the standard deviation of lnðamax Þ [i.e., σlnðamax Þ ]
Justifications for each estimate of uncertainty are also provided, were extracted directly from USGS ShakeMaps (USGS 2019)
and uncertainties that require more detailed explanations are dis- when geographic coordinates were able to be determined for the
cussed next. case history sites. Some exceptions include the following:
The standard deviation of qc1Ncs (i.e., σqc1Ncs ) was estimated • When geographic coordinates were not able to be determined,
using a coefficient of variation (COV) of COV ¼ σqc1Ncs =qc1Ncs . the mean value of σlnðamax Þ within the map extents of each earth-
The adopted value of COV was determined from a set of over quake was adopted.
3,000 CPT soundings from the Christchurch, New Zealand, re- • In the case of the 1964 Niigata, Japan, event, no ShakeMap is
gion (data from Upadhyaya 2019; Geyin et al. 2021). Each sound- available. Instead, σlnðamax Þ was estimated based on an estimate
ing was divided into individual soil layers that were sufficiently of the standard deviation of amax (i.e., σamax ) given by Moss
thick (e.g., at least 1 m) to estimate σqc1Ncs within each layer. et al. (2006) for this specific event. Their assumption was that

0
Table 1. Required parameters to compute lnðΔW=σvo Þ
Parameter Equation Notes
σv γ A · ðzw Þ þ γ t;soil · ðz − zw Þ zw = depth to groundwater table (gwt) and γ A ¼ 17.0 kN=m3
0
σvo σv − γ w ðz − zw Þ γ w ¼ 9.81 kN=m3
lnðV s Þ Eq. (8) Ulmer et al. (2020)
γ t;soil 19.5 kN=m3 below gwt (z > zw ) Assumed value below the groundwater table (Moss 2003;
17.0 kN=m3 above gwtðz ≤ zw Þ Green et al. 2014)
0
ðG=Gmax Þγc Iterative computation: Eq. (4) Assume K o ¼ 0.5 to compute σmo and assume PI ¼ 0, unless given
0
Dγc Compatible with ðG=Gmax Þγc Assume K o ¼ 0.5 to compute σmo and assume PI ¼ 0, unless given
rd Eq. (6) Lasley et al. (2016) (WUS-specific)
lnðneq;M Þ Eq. (7) Lasley et al. (2017) (WUS-specific)

g Þ
lnðamax Varies for each case history From strong motion recordings, USGS ShakeMaps, or regional
ground motion models. Converted from peak to geometric
mean amax values when necessary

© ASCE 04023105-5 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 Table 2. Standard deviations and correlation coefficients required to compute σlnðΔW=σvo0 Þ
Parameter Equation Notes
σqc1Ncs 0.093 · qc1Ncs FC measured in the lab From analysis of CPTs in Christchurch,
0.1395 · qc1Ncs FC estimated from I c New Zealand
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2ffi
T
σ σv 0.56 þ 380.25 (kPa) T = thickness of critical layer (m). From assumed
6
uncertainties in depth to critical layer and depth to
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
T groundwater table
σσvo0 4.8 þ 93.89 (kPa)
6
   
0.489 2 0.25 2
σV s σln V s ¼ σqc1Ncs þ σσvo0 First-order approximation
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

qc1Ncs 0
σvo
σG=Gmax Varies (function of shear strain) Darendeli (2001)
σD Varies (function of shear strain) Darendeli (2001)
0.1506
σ rd Lasley et al. (2016) (WUS-specific)
1 þ expð−0.4975zÞ
0.5399 − ð2.928 × 10−3 Þz
σlnðneq Þ max Lasley et al. (2017) (WUS-specific)
0.4626
σlnðamax Þ Varies In general: from USGS ShakeMaps and increased
to account for uncertainty in conversion from peak
to geometric mean amax
ρD;G=Gmax −0.5 Rodriguez-Marek et al. (2014)
ρσvo0 ;σv −1.0 Based on uncertainty in groundwater table
ρrd ;lnðneq Þ −0.0390ðzÞ − 0.1251 for z ≤ 8.7 m Based on analyses of site-response analyses
−0.4644 for z > 8.7 m performed by Lasley et al. (2016, 2017)
ρlnðamax Þ;rd −0.0062ðzÞ − 0.1692 Based on analyses of site-response analyses
performed by Lasley et al. (2016, 2017)
ρlnðamax Þ;lnðneq Þ 0.0310ðzÞ − 0.3710 for z ≤ 12.5 m Based on analyses of site-response analyses
0.0165 for z > 12.5 m performed by Lasley et al. (2016, 2017)

amax ¼ 0.162g  0.03g. Assuming that the 0.03g represents Darendeli (2001) as a function of strain is reasonably applicable for
one standard deviation of amax (i.e., σamax ), the corresponding the IZ93 values as well.
σlnðamax Þ was computed as follows:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  
σamax 2 0.03 2 Regression of the Limit-State Function
σlnðamax Þ ¼ ln 1 þ ¼ ln 1 þ
amax 0.162
A limit-state surface separating the Yes liquefaction from the No
¼ 0.1836 ð11Þ liquefaction case histories is developed using maximum likelihood
regression. The limit-state function can be denoted
• Estimates of σlnðamax Þ for the 2010 M7.1 Darfield and 2011 M6.2
Christchurch events were obtained directly from the Bradley  
(2013) model to be consistent with the estimates of amax from ΔW
g ¼ fðqc1Ncs ; ΘÞ − ln 0 þε ð13Þ
the same model. σvo
If the value of amax for a given case history was converted from a
peak value to a geometric mean amax as discussed previously, then
its standard deviation was also increased to reflect the uncertainty where f = function of a chosen form (e.g., linear or power); Θ is a
in the conversion vector of coefficients; and ε = error term that is assumed to be nor-
mally distributed with a mean of zero and standard deviation of σε .
σ2ln GMamax ¼ σ2ln Peakamax þ σ2lnðPeak=GMÞ ð12Þ The limit state is obtained by setting g ¼ 0: if g ≤ 0, liquefaction is
predicted and if g > 0, liquefaction is not predicted. The liquefac-
where σln GMamax = standard deviation of the converted geometric tion case history database provides the vector of variables that de-
mean amax ; σlnPeakamax = original estimate of σln amax associated with fine the dissipated energy [i.e., Dγc ; ðG=Gmax Þγc , Gmax , V S , amax ,
the peak value amax ; and σlnðPeak=GMÞ = uncertainty in the conver- 0 , σ , and n
rd , σvo v eq;M ] and the resistance, qc1Ncs , where the vector
sion: 0.091 (Boore and Kishida 2017). of variables for a given case history is denoted by X i. The param-
Ishibashi and Zhang (1993) did not provide standard deviations eters of the limit-state surface (Θ), which include the uncertainty
for their values of Dγc and ðG=Gmax Þγc . However, it was assumed parameter σε, are obtained through regression. The likelihood func-
that the standard deviations of Dγc and ðG=Gmax Þγc estimated by tion can then be computed (Cetin et al. 2002)

© ASCE 04023105-6 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 Y Y  
LðΘ; σε Þ ¼ P½gðX i ; ΘÞ ≤ 0 × P½gðX i ; ΘÞ > 0 ΔW
ln 0 ¼ c1 ðqc1Ncs Þc2 − 7.52 þ ε ð18Þ
Liq: sites No-liq: sites σvo liq
ð14Þ
The intercept was fixed to a value of −7.52 to stabilize the re-
where the first term quantifies the probability of g being less than gression because the data are not sufficient to constrain c1 , c2 , σε ,
or equal to zero at liquefied sites and the second term quantifies and the intercept term. The value of −7.52 matches the intercept
the probability of g being greater than zero at nonliquefied sites. from a laboratory-derived liquefaction resistance curve (Ulmer
The values of the Θ vector, including the standard deviation σε , are 2019), which was assumed to be closely associated with a lower
then regressed to maximize the likelihood, L. Alternatively, the log- bound of the general field-based limit-state curve. Table 3 contains
likelihood function can be maximized using this equation: the regression coefficients for the limit-state curve for two sce-
X narios: excluding and including the uncertainties in the input
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

lnðLðΘ; σε ÞÞ ¼ lnðP½gðX i ; ΘÞ ≤ 0Þ parameters.

Liq: sites The uncertainty in the limit-state curve, σε , was smaller when
X the uncertainties in the input parameters were included [Eq. (17)].
þ lnðP½gðX i ; ΘÞ > 0Þ ð15Þ
No-liq: sites
This represents a case in which σε is an estimate of true model
error; the larger resulting uncertainty when ignoring uncertainties
in the input parameters [Eq. (16)] can be thought of as total
If uncertainties in the input parameters are not considered, then uncertainty (e.g., Franke and Olson 2021). Fig. 2 shows how
the probability of liquefaction, PL , can be written these limit-state curves fit the data from the case history database.
  A deterministic triggering curve is also shown in Fig. 2 and is
gðX i ; ΘÞ discussed subsequently. One case history with qc1Ncs of nearly
PL ¼ P½gðX i ; ΘÞ ≤ 0 ¼ 1 − Φ ð16Þ
σε 300 was treated as an outlier and ignored in the regression so
that it would not have undue influence on the resulting limit-state
where Φð·Þ = standard normal cumulative distribution function. curve.
If the uncertainties in the input parameters are considered, then At first glance, the number of No liquefaction cases above the
each input parameter has its own mean and error term. Thus, the curves and the number of Yes liquefaction cases below the curves
probability of liquefaction is estimated seem unusually high. These issues can be explained, at least par-
tially, by additional considerations as discussed subsequently in
PL ¼ P½gðX i ; ΘÞ ≤ 0 the “Discussion” section of this paper. Figs. S1–S4 show the same
0
ZZ    case histories split up into various intervals of FC, σvo , M, and
gðX i ; ΘÞ amax values, respectively. These figures illustrate the model per-
¼ 1−Φ fQ ðqÞfW ðωÞdqdω ð17Þ
σε formance for different ranges of applicability of the predictive
variables.
0
where ω ¼ lnðΔW=σvo Þ; q ¼ qc1Ncs ; and f Q ðqÞ and fW ðωÞ =
probability density functions of qc1Ncs and lnðΔW=σvo 0 Þ, respec-

tively. As indicated previously, f W was assumed to be lognormally Table 3. Regression coefficients for energy-based limit-state curves for
distributed with an uncertainty estimated using Eq. (10). Similarly, two scenarios: uncertainties in input parameters excluded and uncertain-
f Q was assumed to be normally distributed with the standard ties included
deviation given in Table 2. Scenario c1 c2 σε
As noted by Boulanger and Idriss (2014), outliers with very low
Uncertainties excluded [Eq. (16)] 1.224 × 10−7 3.352 1.590 (total)
probabilities of P½gðX i ; Θ; εi Þ ≤ 0 or P½gðX i ; Θ; εi Þ > 0 can have Uncertainties included [Eq. (17)] 1.223 × 10−7 3.335 1.400 (model)
a strong influence on the outcome of the regression. Thus, for prac-
ticality, it is reasonable to set a minimum probability for any one
case history, Pmin (effectively stating that any site on the liquefac-
tion side of the limit-state surface has a minimum probability of no-
liquefaction of Pmin , and conversely any site on the no-liquefaction
side of the limit-state surface has a minimum probability of lique-
faction of Pmin ). Boulanger and Idriss used a sensitivity analysis to
test the effect of a range of Pmin values on their regression and
found Pmin values between 0.05 and 0.075 to be realistic. In the
present study, a range of Pmin values was also considered, and
Pmin ¼ 0.05 was selected based on the stabilizing effect it had on
the regression while not altering the regression coefficients signifi-
cantly compared with lower values of Pmin .
The limit-state curve was regressed two ways, first assuming
that the input parameters are known (i.e., input parameter uncer-
tainties are not considered) and then assuming that the input param-
eters have some uncertainty (e.g., Cetin et al. 2002). Several
Fig. 2. Case histories from the updated database plotted as normalized
functional forms were considered for the regression. However, bal-
dissipated energy versus qc1Ncs . Also shown are median (PL ¼ 50%)
ancing the desires to limit the complexity of the functional form and
energy-based limit-state curves for two scenarios: uncertainties in input
having a shape that yielded the best predictive rates for the case
parameters are ignored and uncertainties are included. Bold line repre-
histories in our database, a power fit was chosen for the limit-state
sents deterministic curve.
curve as follows:

© ASCE 04023105-7 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 Implementation in Forward Analysis IM relates physically to the process that is being predicted (i.e., pore
pressure generation and liquefaction triggering). As discussed
To implement the proposed energy-based triggering model for previously, ΔW relates mechanistically to excess pore-water pres-
predicting the occurrence of liquefaction, the following equation sure generation (e.g., Nemat-Nasser and Shokooh 1979; Simcock
can be used to compute PL for a given layer of soil (assuming et al. 1983; Berrill and Davis 1985; Figueroa et al. 1994; Green
no uncertainty in the input parameters, i.e., using total uncertainty et al. 2000; Davis and Berrill 2001; Jafarian et al. 2012; Polito et al.
for the limit-state curve and best-estimated values for each input 2008; Kokusho and Kaneko 2018; Kokusho and Tanimoto 2021;
parameter): among many others).
"
# Another desired characteristic of liquefaction triggering models
ð1.224 × 10−7 Þðqc1Ncs Þ3.352 − 7.52 − ln ΔW
0
σvo is that these models should allow for correlations between the abil-
PL ¼ 1 − Φ
1.590 ity of the soil to resist liquefaction and in situ test metrics, circum-
venting issues about the representativeness of laboratory specimen
ð19Þ
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

versus in situ soil response (e.g., Peck 1979). Two additional de-
where lnðΔW=σvo 0 Þ can be computed using Eq. (5) or via more sired model characteristics relate to implementation scalability of
refined methods (e.g., equivalent linear numerical site response the model (i.e., the model should allow refined estimates of either
analyses using the IZ93 MRD curves and several input motions earthquake loading or ability of the soil to resist liquefaction to be
representative of the design event). readily incorporated) and to the applicability of the triggering curve
The recommended deterministic estimate of lnðΔW=σvo 0 Þ to for non-reference conditions (e.g., non-vertically-propagating wave
0
trigger liquefaction [i.e., lnðΔW=σvo Þliq  is fields or nonearthquake loading). These two attributes are further
  discussed subsequently as related to the proposed model. Also,
ΔW as elaborated subsequently, an attribute that relates to the overall
ln 0 ¼ 1.224 × 10−7 · ðqc1Ncs Þ3.352 − 8.133 ð20Þ
σvo liq
validity of the model, in addition to being a desired characteristic,
is that a model should consistently operate within either a total
which is associated with a PL ¼ 35% contour computed using the stress or effective stress framework. With all of this in mind, the
total uncertainty. The factor of safety against liquefaction trigger- authors’ proposed energy-based liquefaction triggering model is
ing, FSL , is then computed not burdened by these issues.


ln ΔW 0
σvo

ΔW liq Choice of ΔW as an Energy IM
FSL ¼ ð21Þ
ln σvo0 As mentioned in the “Background” section of this paper, various
0 0
IMs for liquefaction triggering evaluation models have been
where lnðΔW=σvo Þ and lnðΔW=σvo Þliq can be computed using broadly classified as energy-based procedures, to include I A and
Eqs. (5) and (20), respectively, or by more refined methods. The CAV. The justification for using these IMs is commonly based on
PL ¼ 35% contour was chosen as the deterministic curve because numerical effective stress site response analyses were I A or CAV of
it reasonably aligned the distribution of FSL from Eq. (21) with the the input motion is correlated to the computed excess pore-water
distribution of FSL from stress-based methods (e.g., Boulanger and response somewhere in the soil profile (e.g., Kramer and Mitchell
Idriss 2014; Green et al. 2019) and is consistent with the total un- 2006; Bullock et al. 2022). To estimate the I A associated with lique-
certainty PL used by Green et al. (2019) in defining their determin- faction triggering in cyclic laboratory tests, Green (2001) derived
istic triggering curve. Although a deterministic estimate of FSL is the following expression for soil samples subjected to cyclic simple
frequently the goal of liquefaction evaluations in current practice, shear, wherein the loading is sinusoidal, acting in one direction, and
the use of a full probabilistic approach allows the user to adopt a has a constant frequency and amplitude shear stress:
probability of liquefaction that is consistent with a desired or target
risk level. nπg τ 2
IA ¼ ð22Þ
4σ2v f

Discussion where σv = total vertical stress; g = acceleration due to gravity; n =

number of cycles of loading; and f and τ frequency and amplitude
As mentioned in the “Introduction” and “Background” sections, of the applied shear stress. When n equals the number of cycles
there are several characteristics that are desired in liquefaction trig- required to initiate liquefaction in the sample, I A computed using
gering models. Some of these characteristics relate to the IM used Eq. (22) equals the capacity of the soil (I A;liq ).
to parameterize the demand. For any alternative IM used in engi- Eq. (22) is revealing in two ways. First, I A;liq is independent of
neering analyses, the lack of maturity in models for predicting effective confining stress, and second, I A;liq is a function of the fre-
the IM is a concern (e.g., Is the benefit in refining the liquefaction quency of loading; numerous laboratory studies have shown trends
triggering prediction using a “new and improved” IM offset by the contradicting both of these (e.g., Lee and Seed 1967; Seed 1983;
increased uncertainty in predicting the IM?). Additionally, the IM Riemer et al. 1994). Focusing on the frequency dependency issue,
should relate mechanistically to excess pore-water pressure gener- Riemer et al. (1994) performed undrained stress-controlled cyclic
ation and should be consistent with seismic hazard maps issued by triaxial tests on Monterey 0 sand, where the applied sinusoidal
building regulators. In the simplified form proposed in this paper loading ranged in frequency from 0.1 to 20 Hz. One of the con-
the predictive variables for ΔW are similar to those used to com- clusions from their study is that the “effect of frequency on the
pute CSR* in the stress-based approaches, hence there is no addi- number of cycles to liquefaction at a given cyclic stress ratio was
tional uncertainties introduced in the prediction of the IM. not significant in the stress controlled loading.” Ignoring the differ-
Two other requirements for an IM are sufficiency and efficiency ence in the stress paths associated with cyclic triaxial versus cyclic
(Luco and Cornell 2007). Efficiency is evaluated by the predictive simple shear testing, Eq. (22) implies that the I A;liq for the tests
ability of the model, which is discussed subsequently in this sec- loaded at 0.1 Hz would be 200 times greater than those for the
0
tion. We postulate that sufficiency is addressed by ensuring that the tests performed at 20 Hz. In contrast, the ðΔW=σvo Þliq for these

© ASCE 04023105-8 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 0
tests have been shown to be approximately the same (e.g., Lasley Eq. (20) is based on the analysis of case histories where ΔW=σvo
2015). imposed on the soil was due to both components of horizontal
The frequency dependency of I A triggering models is an artifact motion during a seismic event, site response analyses need to be
of I A being computed by integrating the amplitude of the load- performed using both orthogonal components of horizontal input
ing over time; this same phenomenon is an issue any IM that in- motions and the ΔW for each summed together. Alternatively,
tegrates the amplitude of loading over time (e.g., CAV). Inherently, the ðΔW=σvo 0 Þ computed using Eq. (20) can be adjusted to allow
liq
the liquefaction triggering models that use I A or CAV and justify a comparison with ΔW computed from a single site response analy-
the use of the IMs based on numerical effective stress site re- 0 Þ
sis (i.e., one input motion). In this latter approach, ðΔW=σvo liq
sponse analyses conflate profile response characteristics, which are computed using Eq. (20) should be multiplied by the ratio of ΔW
frequency-dependent, and soil element liquefaction response char- for one dimension (1D) versus two dimensions (2D) given by
acteristics, which are frequency-independent. The conflagration  
of distinct phenomena will ultimately limit improvements in the ΔW 1D
¼ −0.8248 − 0.0023 lnðamax Þ þ 0.0221M ð24Þ
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

prediction efficacy of the models and limits the implementation ln

ΔW 2D liq
scalability of the models. In contrast, these limits are not inherent
to ΔW-based models.
This relationship is based on the neq;M relationships proposed by
Lasley et al. (2017) for 1D and 2D loadings.
Implementation Scalability Although, in theory, similar refinements can be made to the
CSR* for the stress-based triggering models, it would be inappro-
Specific to the proposed energy-based model implemented in its priate to compare the resulting CSR* with the stress-based limit-
0
simplified form, the IMs required to compute ΔW=σvo are the same state curves that use a biased rd relationship (e.g., Green et al.
as those required for simplified stress-based models (i.e., amax 2019). Furthermore, although some stress-based models (e.g., Cetin
and M), with these IMs used in conjunction with MRD curves et al. 2018; Moss et al. 2006; Kayen et al. 2013) use a rd relation-
(i.e., Ishibashi and Zhang 1993). However, refinements can be ship that is not biased, these models are still dependent on a generic
0
made to the loading imposed on the soil (i.e., ΔW=σvo ). Toward MSF to account for durational effects. In contrast, computation
this end, the refinements can be made by directly measuring V S of ΔW via site response analyses can be used to compute
in situ, developing site- or region-specific rd and/or neq;M relation- lnðΔW=σvo 0
Þ and compared with lnðΔW=σvo 0
Þliq , where the com-
ships for use in Eq. (5), and/or developing soil-specific MRD puted ΔW accounts for both the amplitude and duration of shaking
curves used to compute γ c [Eq. (4)]. However, the same general in a single parameter that has a physical link to liquefaction trig-
approaches used to derive the relationships used to develop the gering (e.g., Ostadan et al. 1998).
triggering model also should be used to develop new relationships Regarding refinements in estimating the ability of the in situ
(e.g., use of equivalent linear site response analyses using the IZ93 soil to resist liquefaction triggering, cyclic laboratory tests can be
MRD curves) to avoid the introduction of new biases due to com- performed on undisturbed samples. However, as may be surmised,
putational artifacts, and so on. applying Eq. (23) directly to compute ΔW from shear stress and
Additionally, if required inputs are available, equivalent linear shear strain time histories from undrained stress-controlled cyclic
site response analyses can be performed to estimate ΔW directly. laboratory test data does not result in a value that is directly com-
Toward this end, both the shear stress and shear strain time histories parable to ΔW computed using Eq. (5) or from equivalent linear
at the depth of interest should be output from site response analy- site response analyses. This is because the γ time history obtained
ses, and ΔW is calculated as the cumulative area enclosed by the from undrained stress-controlled laboratory tests inherently in-
stress-strain hysteresis loops (Green and Terri 2005). This area can cludes the effects of the degradation of soil stiffness (i.e., G) due
be estimated using the trapezoidal rule to excess pore-water pressure generation, whereas use of Eq. (4) to
estimate γ c does not (i.e., effective stress versus total stress frame-
1Xn−1
works). The inconsistent operation within a total versus effective
ΔW ¼ ðτ þ τ k Þðγ kþ1 − γ k Þ ð23Þ
2 k¼1 kþ1 stress framework is an inherent limitation of several previously pro-
posed energy-based triggering models.
where τ k and γ k (in decimal form) = kth increments of shear stress Fig. 3 shows an example set of hysteresis loops from a constant-
and shear strain, respectively; and n = total number of digitized volume cyclic simple shear test performed on Monterey 0=30 sand
points in a time history. and the associated relationship between normalized dissipated
0
Alternatively, the equivalent linear site response software energy within an effective stress framework (ΔW eff =σvo ) and the
ShakeVT2 (Lasley 2015; Thum et al. 2019) has an option to output number of cycles of loading (N). Initially, the hysteresis loops
0
ΔW directly. However, because the value for ðΔW=σvo Þliq given by are approximately the same size, and the relationship between

Fig. 3. Results from a cyclic direct simple shear test: (a) sample hysteresis loops; and (b) relationship between both effective and total normalized
dissipated energy and number of loading cycles.

© ASCE 04023105-9 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 0
ΔW eff =σvo and N is linear. Once the excess pore-water pressure element for each time step. Once the normalized dissipated energy
ratio (ru ) is sufficiently large to soften the soil, the hysteresis loops exceeds that required to trigger liquefaction, the properties of
grow larger with each cycle of loading. At this point, the relation- the soil of that element are switched to those corresponding to
ship between ΔW eff =σvo 0 and N is no longer linear and tends toward
the residual values of liquefied soil. Use of dissipated energy
an exponentially increasing shape. as the cumulative damage metric in these analyses is more direct
The initial linear portion of the ΔW eff =σvo 0 versus N plot is
than applying stress-based cycle counting methods (e.g., Naesgaard
0
equivalent to a ΔW=σvo versus N computed within a total stress and Byrne 2007; Beaty and Byrne 2008). However, in using
framework (i.e., ΔW tot =σvo 0 versus N). This is because the con-
Eq. (20) to estimate the normalized dissipated energy required
stant slope of the early portion of ΔW eff =σvo 0 versus N plot im-
to trigger liquefaction for 1D shaking in a 2D numerical analy-
plies that the stiffness of the soil is not significantly influenced by sis, adjustments for 1D versus 2D shaking need to be applied
the excess pore-water pressure and that size of hysteresis loops is per Eq. (24).
relatively constant from one cycle to the next. Accordingly, if this Outside the realm of earthquake analyses, the proposed trigger-
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

linear portion of the ΔW eff =σvo 0 versus N curve is extrapolated,

ing curve can be used to evaluate liquefaction due to construction
it is a reasonable representation of the ΔW tot =σvo 0 versus N curve;
vibrations [e.g., pile driving (Taylor 2011; Lamens and Askarinejad
Fig. 3(b) illustrates this relationship. Once the relationship be- 2021)], geophysical explorations (e.g., Hryciw et al. 1990), and soil
0
tween ΔW tot =σvo versus N is established, a liquefaction trigger- improvement (e.g., Green and Mitchell 2004, 2010), among other
ing criterion for interpreting cyclic laboratory tests (e.g., ru ¼ applications. However, for these applications, approaches for com-
0.85, single axial strain εsa equal to 3.5%, double axial strain εda puting ΔW associated with the vibratory loading may need to be
equal to 5%, and so on) is used to determine N corresponding to developed. Nevertheless, not linking the duration of the ground
0
liquefaction (i.e., N liq ). The value of ðΔW=σvo Þliq is equal to the vibrations to earthquake magnitude broadens the applicability of
0
value of ΔW tot =σvo corresponding to N ¼ N liq . Guiding rules the triggering curve to nonearthquake-type applications.
0
for extrapolating the initial portion of the ΔW eff =σvo versus N
0
curve to estimate ΔW tot =σvo are provided in the Supplemental
Materials. Not being able to refine the estimated energy required Comparison of Proposed Energy-Based and
to trigger liquefaction via undrained cyclic laboratory tests per- Stress-Based Triggering Model Frameworks
formed on undisturbed samples is an inherent limitation of several
previously proposed energy-based triggering models, particularly 0
Relationship between ΔW=σ vo and CSR*
for triggering models that use I A or CAV as an IM, as discussed
As detailed previously, for typical liquefaction evaluations,
previously. 0
ΔW=σvo can be computed using Eq. (5), which inherently esti-
mates the value of τ avg via Newton’s Second Law, the same as the
Applicability of Triggering Curve simplified stress-based liquefaction models (Seed and Idriss 1971).
0
One advantage of the proposed triggering curve is that its appli- However, this does not mean that ΔW=σvo is just a scaled value of
cability is wide ranging. Within the realm of earthquake analyses, CSR*. The reason for this is because the damping and shear modu-
the model can be used to analyze liquefaction triggering for free- lus factors in Eq. (5) [i.e., Dγc and Gmax ðG=Gmax Þγc  are not con-
field sites, similar to simplified stress-based models, to include stants, but vary as a function of the induced shear strain, γ c , which
liquefaction triggering of aged soils using directly measured values in turn will be a function of the imposed shear stress, stiffness of the
of V S to compute Gmax (e.g., Green et al. 2022). Furthermore, the soil, shear stress–shear strain response of the soil, and so on. The
relationship between ΔW=σvo 0 and CSR* is illustrated in Fig. 4,
model can be used to analyze flow liquefaction of earthen dams, for
example, resulting from earthquake shaking using an approach which shows ΔW 1 plotted as a function of CSR for varying soil
analogous to that outlined by Naesgaard and Byrne (2007) and densities and confining stresses. The significance of the non-one-
Beaty and Byrne (2008). In these types of analyses, the dynamic to-one relationship between the two IMs is discussed subsequently
response of the dam is numerically modeled within a total stress in the context of the influence of effective confining stress and
framework and the normalized dissipated energy is tracked for each aging on liquefaction triggering.

0 , of (a) 100 kPa (∼1 atm);

Fig. 4. Relationship between ΔW 1 and CSR for a range of soil densities and for initial vertical effective confining stress, σvo
and (b) 400 kPa (∼4 atm). The vertical scale of the two plots is different.

© ASCE 04023105-10 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 Ground Motion Duration
As stated in the “Introduction,” in simplified stress-based triggering
models, the influence of the duration of the loading is accounted
for via MSF, which is a function of neq;M normalized by neq;M7.5,
where neq;M7.5 is the duration of a reference earthquake having
a M7.5 expressed in terms of number of equivalent cycles. In con-
trast, ΔW directly accounts for the duration of the loading by
integrating the shear stress–shear strain hysteresis loops for the
duration of ground shaking. However, for the simplified implemen-
tation of the proposed energy-based triggering model, the duration
of the loading is accounted for via a proposed neq;M relationship
[i.e., Eq. (7)]. This same neq;M relationship was used by Green et al.
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

(2019) to develop a MSF relationship for their proposed stress-

based simplified procedure. However, in contrast to other neq;M
relationships used to develop MSF for stress-based triggering mod-
els, Eq. (7) is a function of amax , as well as M. Accordingly, a refer- Fig. 5. Comparison of the K σ relationship inherent in the proposed
ence value of amax is required, which Green et al. (2019) selected as energy-based triggering model, the K γ relationship proposed by Green
0.35g based review of the liquefaction case history databases. et al. (2022), and the K σ relationship used by Idriss and Boulanger
Nevertheless, the duration of the ground motion on liquefaction (2008) (IB08) for three different soil densities.
triggering is consistently accounted for in the simplified stress-
based triggering model proposed by Green et al. (2019) and the
simplified energy-based triggering model proposed herein. in Fig. 5 lie within the range of scatter of laboratory determined
values of K σ .
Normalization of ΔW for Effective Confining Stress
0
In the proposed energy-based model, ΔW is normalized by σvo to Aging
account for the influence of initial effective confining stress on The age of a soil deposit soil has been long recognized as having an
liquefaction triggering based on the findings of Lasley (2015) and influence on its susceptibility to liquefaction triggering (e.g., Youd
Ulmer (2019). Of most significance, Ulmer (2019) performed and Hoose 1977), where the soil fabric changes with the age of the
cyclic simple shear tests on similar samples with vertical effective deposit, generally resulting in an increased resistance to liquefac-
confining stresses 60, 100, and 250 kPa, and a unique correlation tion triggering. Seed (1979) proposed an early method for account-
between ðΔW tot =σvo 0 Þ
liq and Dr was shown to exist. This implies ing for aging on liquefaction resistance by computing the ratio of
that the additional normalization for effective confining stress from the CRR of an aged soil to that of a young deposit of the same soil
K σ used in computing CSR* is inherently accounted for in
0
ΔW=σvo . This is a direct result of ΔW being a function of the CRRaged
induced shear strain, γ c , which in turn is a function of the imposed K DR ¼ ð25Þ
CRRyoung
shear stress, stiffness of the soil, shear stress–shear strain response
of the soil, and so on. where K DR = liquefaction strength gain factor due to aging effects.
In this vein, the imposed shear stress required to result in the However, Andrus et al. (2009) and Hayati and Andrus (2009) have
0
same value of ΔW 1 =σvo in similar samples confined at different shown that the time since last disturbance is more relevant to lique-
initial effective confining stresses can be used to compute the K σ faction triggering susceptibility than geologic age. The two are the
relationship inherent to the proposed energy-based triggering same only if the deposit has not been significantly disturbed since
model. As an example, the K σ back-calculated by equating the deposition (e.g., if liquefaction has not been triggered in the deposit
normalized dissipated energies in two similar specimens having during a previous earthquake). Bwambale and Andrus (2019) de-
densities corresponding to qc1Ncs ¼ 105, with one specimen con- veloped the most recent relationship, at least that the authors are
fined at σvo 0 ¼ 100 kPa (∼1 atm) and the other at varying σ 0
vo aware of, for K DR as a function of time by regressing compiled
is shown if Fig. 5. The back-calculated values of K σ shown in this soil aging case histories
figure are for FSL ¼ 1, where the CRR* of the soil was estimated
using the relationship proposed by Green et al. (2019). K DR ¼ 0.13 · logðtÞ þ 0.82 ð26Þ
Opposed to K σ , the influence of initial effective confining stress
where t = time since last major disturbance (years).
inherent to the proposed energy-based triggering model is applied
0 , not to the ability of the soil to resist To estimate time since last disturbance, Andrus et al. (2009) pro-
to imposed loading, ΔW=σvo
0 posed using the ratio of measured to estimated V S (MEVR):
liquefaction ðΔW=σvo Þliq . The argument for this was articulated by
Green et al. (2022), based on liquefaction triggering being a strain Vsmeasured
phenomenon, not a stress phenomenon, and as a result, the influ- MEVR ¼ ð27Þ
Vsestimated
ence of effective confining stress on liquefaction triggering varies
as a function of the FSL , among other factors. In addition to the where Vsmeasured is directly measured, and the estimated V S
back-calculated K σ relationship from the proposed energy-based (i.e., Vsestimated ) is determined using correlations relating V S and
model, Fig. 5 also shows K σ relationship used by Idriss and penetration resistance [e.g., Eq. (8)]. The underlying premise of
Boulanger (2008) and the K γ relationship proposed by Green et al. the Andrus et al. (2009) approach is that the measurement of pen-
(2022) assuming FSL ¼ 1. etration resistance mobilizes intermediate to large strains that inher-
As may be observed from this figure, the K σ inherent to the ently disturb the soil fabric and, thus, is not that sensitive to aging
proposed energy-based model plots lower than the other relation- effects (i.e., penetration resistance correlates to the V S of the soil,
ships, although it is not too different from the K γ relationship pro- if the soil were young, regardless of the time since last distur-
posed by Green et al. (2022). However, all the relationships shown bance). In contrast, the measurement of V S directly in the soil is

© ASCE 04023105-11 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 0
model is applied to imposed loading, ΔW=σvo , not to the ability of
0
the soil to resist liquefaction ðΔW=σvo Þliq (i.e., aged soils have
higher Gmax, and therefore for a given imposed shear stress, the
induced γ c is less and the ΔW is less). As with the accounting
for the influence of initial effective confining stress, the argument
for this was articulated by Green et al. (2022), based on liquefaction
triggering being a strain phenomenon, not a stress phenomenon,
and as a result, the influence of aging effects on liquefaction trig-
gering varies as a function of the FSL , among other factors.

Initial Static Shear Stress

In the simplified stress-based triggering models, the influence of
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

initial static shear stress acting on a horizontal plane on liquefac-

tion triggering is accounted for by the inclusion of the K α factor.
Fig. 6. Comparison of the aging liquefaction strength gain factor
The authors hypothesize that the influence of this initial static
relationships inherent to the proposed energy-based triggering model
shear stress is also inherent to the proposed energy-based model.
and that proposed by Bwambale and Andrus (2019) (BA19).
However, this would require the computation of ΔW using modi-
fied Masing rules (e.g., Pyke 1979), for example, to account for
the change in the shear stress–shear strain hysteretic behavior of
a small-strain measurement and is sensitive to aging effects (i.e., it the soil. At this time, the authors have not developed a simplified
is the V S of the aged soil). Thus, the ratio of directly measured V S approach for computing ΔW using the modified Masing rules as a
to that estimated from penetration resistance should be able to serve function of α (i.e., initial static shear stress acting on a horizontal
as an index for the time since last disturbance. In this vein, Andrus plan normalized by the initial vertical effective stress).
et al. (2009) developed the following correlation relating MEVR Furthermore, and more importantly, the authors have not vali-
and t: dated their hypothesis that the influence of initial static shear stress
is inherent to the proposed energy-based model using laboratory
MEVR ¼ 0.082 · logðtÞ þ 0.935 ð28Þ data. Although these are planned future efforts in furthering the
development of the energy-based triggering model, in the interim,
where t is in years. the authors have developed energy-based K α relationships, K α ΔW ,
As with K σ , an inherent characteristic of ΔW being a function from the stress-based K α relationship used by Idriss and Boulanger
of the induced shear strain, γ c , is that the proposed energy-based (2008), where K α ΔW is given by
triggering model accounts for the influence of soil aging on lique-
ΔW 
faction triggering. An equivalent K DR relationship can be back- 0
σvo liq α
calculated from the proposed energy-based triggering model by Kα ΔW ¼
ΔW  ð29Þ
0
σvo
using Eqs. (27) and (28) to estimate the Gmax for an aged soil. liq α¼0
This can be done by taking the ratio of the shear stresses required
to result in the same values of ΔW in the aged and young sand. The resulting K α ΔW relationships are presented in Fig. 7 for two
A comparison of the back-calculated K DR relationship and that pro- initial effective confining stresses, with plots for initial effective con-
posed by Bwambale and Andrus (2019) [i.e., Eq. (26)] is shown fining stresses ranging from 100 to 1,000 kPa (1 to 10 atm) provided
in Fig. 6. in Fig. S5. As may be surmised from Eq. (29), Kα ΔW is applied to
0
As may be observed from this figure, the two relationships are in the ability of the soil to resist liquefaction [i.e., ðΔW=σvo Þliq ] similar
excellent agreement. However, as opposed to K DR , the accounting to how K α is implemented in the simplified stress-based models, and
0
for aging effects inherent to the proposed energy-based triggering not to the imposed loading (i.e., ΔW=σvo ) as would be the case if the

Fig. 7. Correction factors for initial static shear stress, K α ΔW , computed from the K α relationships used by Idriss and Boulanger (2008) (IB08) for
initial vertical effective confining stresses of (a) 100 kPa (∼1 atm); and (b) 400 kPa (∼4 atm).

© ASCE 04023105-12 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 modified Masing criteria were used to account for initial static shear liquefaction despite no observations of liquefaction manifestation
stress. at the site, and false negatives indicate that the procedure pre-
dicted no liquefaction despite observations of liquefaction man-
Comparison with the Boulanger and Idriss (2016) ifestations at the site. The proposed energy-based method had
Triggering Model nearly the same number of correct predictions as the two stress-
Two simplified stress-based liquefaction evaluation procedures that based methods, but the ratio of false positives to false negatives
use the original BI14 database as their basis are compared with the was higher.
proposed energy-based procedure. These stress-based procedures It is expected that there would be false negative predictions
use CSR* to quantify the seismic loading (Boulanger and Idriss because the lack of surficial manifestations of liquefaction does
2016; Green et al. 2019). Fig. 8 shows how the median (PL ¼ 50%) not necessarily indicate that liquefaction was not triggered at depth
stress-based limit-state curves fit the data in CSR* versus qc1Ncs (e.g., Upadhyaya et al. 2022). There are several potential issues that
space compared with how the median proposed energy-based limit- could prevent surficial manifestations, such as a thick unsaturated
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

0
state curve fits the data in ΔW=σvo versus qc1Ncs space. The stress- crust at the surface or thick layers of soil with high plasticity, i.e., I c
based CSR* values were computed using the data in the original greater than 2.6 (Green et al. 2018). Nine No liquefaction case his-
BI14 database (from which these procedures were regressed) and tories were identified as having this issue and are marked as having
values of ΔW=σvo 0 were computed using the data in the updated a possible capping layer in Fig. 8.
database outlined in the Supplemental Materials. As a result, some Another confounding issue could be interbedded layers of soil
of the noted trends may relate to the differences in the databases with I c > 2.6 within the critical layer (Maurer et al. 2014, 2015;
underlying the models. Upadhyaya et al. 2018; Cubrinovski et al. 2019). Eleven case his-
Table 4 summarizes the number of correct predictions, false tories with this issue are also marked in Fig. 8. These case histories
positives, and false negatives using the median curve from each were not removed from the database because of these issues;
procedure. False positives indicate that the procedure predicted however, it is important to note that such issues exist and may affect
the accuracy of these evaluation procedures when such issues
are present. If these were excluded, then the ratio of false positive
to false negative predictions given in Table 4 would be closer
to unity.
It is emphasized, however, that the comparisons presented in
Fig. 8 and Table 4 only represent data from a narrow range of
scenarios [e.g., earthquake magnitude, depth to liquefaction, and
so on (NRC 2016)]. Greater differences between the stress- and
energy-based procedures likely exist for scenarios different from
those represented by the case history database.

Conclusions
The objective of this research was to develop an energy-based
model for evaluating liquefaction triggering. Normalized dissipated
energy per unit volume of soil captures both the amplitude and
duration of the ground shaking, to include the variation of the
amplitude of the motion throughout the duration of shaking and
the soil response, and is mechanistically linked to excess pore-
water pressure generation. In its simplified form, the proposed
energy-based model is implemented similarly to the simplified
stress-based models, with the only additional inputs required rela-
Fig. 8. Case histories common to the BI14 database and the updated tive to stress-based models being soil stiffness and shear modulus
database plotted as CSR* versus qc1Ncs for stress-based procedures and reduction and damping (MRD) curves. The additional information
as normalized dissipated energy versus qc1Ncs for the proposed energy- resulting from introduction of soil stiffness and MRD curves inher-
based method. Solid lines represent median (PL ¼ 50%) limit-state ently merges aspects of the stress- and strain-based liquefaction
curves when uncertainties in input parameters are ignored. Non-black triggering concepts into one model, where liquefaction is inherently
coloring and stars indicate case histories with potential issues that affect a strain phenomenon.
their accuracy. However, use of normalized dissipated energy as the IM circum-
vents issues with quantifying the ground motion duration inherent
to simplified strain-based procedures. The energy-based limit-state
curves derived herein using maximum likelihood regression have
Table 4. Number of correct, false positive, and false negative predictions comparable predictive ability to stress-based limit-state curves in
for the proposed energy-based procedure and two stress-based procedures terms of number of correct predictions. As with existing stress-
based limit-state curves based on similar case history databases
False False of predominantly nonplastic sands and silty sands, the limit-state
Procedure Correct positive negative
curves derived herein are limited to these soil types and may not
Energy-based, uncertainties excluded 193 36 23 be applicable for soils with plastic fines or with significant fines
Energy-based, uncertainties included 192 37 23 content.
Boulanger and Idriss (2016) 204 23 25 The proposed energy-based model has desirable implemen-
Green et al. (2019) 207 23 22
tation scalability and applicability characteristics. Specifically,

© ASCE 04023105-13 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 refinements can be made to the imposed loading on the soil and to Supplemental Materials
the soil’s ability to resist liquefaction triggering. These refinements
can be made by using soil-, site- or region-specific relationships/ Table S1 and Figs. S1–S5 are available online in the ASCE Library
values in computing the imposed seismic loading or by performing (www.ascelibrary.org).
site-specific site response analyses (i.e., nonsimplified implemen-
tation of the model). Additionally, refinements to the ability of the
soil to resist liquefaction triggering can be made by performing References
undrained cyclic tests on undisturbed samples. Details on how to
Andrus, R. D., H. Hayati, and N. P. Mohanan. 2009. “Correcting liquefac-
determine the normalized dissipated energy per unit volume of soil
tion resistance for aged sands using measured to estimated velocity
within a total stress framework were presented. ratio.” J. Geotech. Geoenviron. Eng. 135 (6): 735–744. https://doi.org
The use of IMs common to other earthquake engineering analy- /10.1061/(ASCE)GT.1943-5606.0000025.
ses that are consistent with seismic hazard maps issued by regula- Beaty, M. H., and P. M. Byrne. 2008. “Liquefaction and deformation
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

tors, the implementation scalability characteristics of the proposed analyses using a total stress approach.” J. Geotech. Geoenviron. Eng.
model, and/or the consistent operation within a total stress frame- 134 (8): 1059–1072. https://doi.org/10.1061/(ASCE)1090-0241(2008)
work separate the proposed model from previously proposed 134:8(1059).
energy-based triggering models. Additionally, the mechanistic ro- Berrill, J. B., and R. O. Davis. 1985. “Energy dissipation and seismic lique-
bustness of the chosen IM is given further credence because it in- faction of sands: Revised model.” Soils Found. 25 (2): 106–118. https://
doi.org/10.3208/sandf1972.25.2_106.
herently accounts certain aspects of the influence of initial effective
Boore, D. M., and T. Kishida. 2017. “Relations between some horizontal-
confining stress and aging effects on liquefaction triggering, which
component ground-motion intensity measures used in practice.”
need to be accounted for by additional relationships when other Bull. Seismol. Soc. Am. 107 (1): 334–343. https://doi.org/10.1785
IMs are used. /0120160250.
Because dissipated energy is a fundamental engineering metric, Boulanger, R. W., and I. M. Idriss. 2014. CPT and SPT based liquefaction
the applicability of the proposed model is broader than that of triggering procedures. Rep. No. UCD/CGM-14/01. Davis, CA: Univ.
simplified stress-based models. Within the realm of earthquake of California.
analyses, the model can be used to analyze liquefaction trigger- Boulanger, R. W., and I. M. Idriss. 2016. “CPT-based liquefaction trigger-
ing for free-field sites, to include liquefaction triggering of aged ing procedure.” J. Geotech. Geoenviron. Eng. 142 (2): 04015065.
soils using directly measured values of V S to compute Gmax . Also, https://doi.org/10.1061/(ASCE)GT.1943-5606.0001388.
Bradley, B. 2013. “Site-specific and spatially-distributed ground-motion
because dissipated energy is a scalar quantity, multidirectional
intensity estimation in the 201–2011 Canterbury earthquakes.” Soil
shaking and other effects such as soil–structure interaction, non- Dyn. Earthquake Eng. 61 (Jun): 3–91. https://doi.org/10.1016/j.soildyn
vertical wave fields, and topographic site effects can readily be .2014.01.025.
accounted for. Furthermore, the model can be used to analyze flow Brandenberg, S. J., et al. 2020. “Next-generation liquefaction database.”
liquefaction of earthen dams, for example, resulting from earth- Earthquake Spectra 36 (2): 939–959. https://doi.org/10.1177/8755293
quake shaking using a hybrid total stress/effective stress approach. 020902477.
Use of dissipated energy as the cumulative damage metric in these Bullock, Z., S. Dashti, A. B. Liel, K. A. Porter, and B. W. Maurer.
analyses is more direct than applying stress-based cycle counting 2022. “Probabilistic liquefaction triggering and manifestation models
methods, which are currently used. Inherently, not linking the du- based on cumulative absolute velocity.” J. Geotech. Geoenviron.
ration of the ground vibrations to earthquake magnitude broadens Eng. 148 (3): 04021196. https://doi.org/10.1061/(ASCE)GT.1943-5606
.0002729.
the applicability of the model to nonearthquake-type applications.
Bwambale, B., and R. D. Andrus. 2019. “State of the art in the assessment
As a result, the proposed model can be used to evaluate, for exam- of aging effects on soil liquefaction.” Soil Dyn. Earthquake Eng.
ple, liquefaction due to construction vibrations. 125 (Oct): 105658. https://doi.org/10.1016/j.soildyn.2019.04.032.
Cetin, K. O., A. Der Kiureghian, and R. B. Seed. 2002. “Probabilistic
models for the initiation of seismic soil liquefaction.” Struct. Saf. 24 (1):
Data Availability Statement 67–82. https://doi.org/10.1016/S0167-4730(02)00036-X.
Cetin, K. O., R. B. Seed, R. E. Kayen, R. E. S. Moss, H. T. Bilge, M. Ilgac,
Some or all data, models, or code that support the findings of this and K. Chowdhury. 2018. “SPT-based probabilistic and determinis-
paper are available from the corresponding author upon reasonable tic assessment of seismic soil liquefaction triggering hazard.” Soil
Dyn. Earthquake Eng. 115 (Dec): 698–709. https://doi.org/10.1016/j
request.
.soildyn.2018.09.012.
Cubrinovski, M., A. Rhodes, N. Ntritsos, and S. Van Ballegooy. 2019.
“System response of liquefiable deposits.” Soil Dyn. Earthquake Eng.
Acknowledgments 124 (Sep): 212–229. https://doi.org/10.1016/j.soildyn.2018.05.013.
Darendeli, M. B. 2001. “Development of a new family of normalized
This study is based on work supported in part by the National modulus reduction and material damping curves.” Ph.D. dissertation,
Science Foundation (NSF) Grant Nos. CMMI-1435494, CMMI- Dept. of Civil, Architectural, and Environmental Engineering, Univ. of
1825189, and CMMI-1937984. The authors gratefully acknowl- Texas at Austin.
edge this support; however, any opinions, findings, and conclusions Davis, R. O., and J. B. Berrill. 2001. “Pore pressure and dissipated en-
or recommendations expressed in this paper are those of the authors ergy in earthquakes—Field verification.” J. Geotech. Geoenviron. Eng.
and do not necessarily reflect the views of NSF. The authors would 127 (3): 269–274. https://doi.org/10.1061/(ASCE)1090-0241(2001)
127:3(269).
like to thank Dr. Sam Lasley for his work (Lasley 2015) on which
Dobry, R., R. S. Ladd, F. Y. Yokel, R. M. Chung, and D. Powell. 1982.
this study builds. The authors also gratefully acknowledge Dr. Brett Prediction of pore water pressure buildup and liquefaction of sands
Maurer for providing many of the digitized CPT soundings and during earthquakes by the cyclic strain method: Building science series.
geographic coordinates for case histories in the database and Washington, DC: National Bureau of Standards.
Dr. Sneha Upadhyaya for providing updated amax values for the Figueroa, J. L., A. S. Saada, L. Liang, and N. M. Dahisaria. 1994.
Darfield and Christchurch events. “Evaluation of soil liquefaction by energy principles.” J. Geotech. Eng.

© ASCE 04023105-14 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 120 (9): 1554–1569. https://doi.org/10.1061/(ASCE)0733-9410(1994) Idriss, I. M., and J. I. Sun. 1992. User’s manual for SHAKE91: A computer
120:9(1554). program for conducting equivalent linear seismic site response analy-
Franke, K. W., and S. M. Olson. 2021. “Practical considerations regarding ses of horizontally layered soil deposits. Davis, CA: Univ. of California.
the probability of liquefaction in engineering design.” J. Geotech. Geo- Ishibashi, I., and X. Zhang. 1993. “Unified dynamic shear moduli and
environ. Eng. 147 (8): 04021061. https://doi.org/10.1061/(ASCE)GT damping ratios of sand and clay.” Soils Found. 33 (1): 182–191. https://
.1943-5606.0002561. doi.org/10.3208/sandf1972.33.182.
Geyin, M., B. W. Maurer, B. A. Bradley, R. A. Green, and S. van Jacobsen, L. S. 1930. “Steady forced vibration as influenced by damping:
Ballegooy. 2021. “CPT-based liquefaction case histories compiled from An approximate solution of the steady forced vibration of a system of
three earthquakes in Canterbury, New Zealand.” Earthquake Spectra one degree of freedom under the influence of various types of damping.”
37 (4): 2920–2945. https://doi.org/10.1177/8755293021996367. Trans. Am. Soc. Mech. Eng. 52 (2): 169–181. https://doi.org/10.1115/1
Green, R. A. 2001. “Energy-based evaluation and remediation of liquefi- .4057368.
able soils.” Ph.D. dissertation, Dept. of Civil and Environmental Engi- Jafarian, Y., I. Towhata, M. H. Baziar, A. Noorzad, and A. Bahmanpour.
neering, Virginia Tech. 2012. “Strain energy based evaluation of liquefaction and residual pore
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

Green, R. A., J. J. Bommer, A. Rodriguez-Marek, B. W. Maurer, P. J. water pressure in sands using cyclic torsional shear experiments.” Soil
Stafford, B. Edwards, P. P. Kruiver, G. de Lange, and J. van Elk. 2019. Dyn. Earthquake Eng. 35 (Apr): 13–28. https://doi.org/10.1016/j
“Addressing limitations in existing ‘simplified’ liquefaction trigger- .soildyn.2011.11.006.
ing evaluation procedures: Application to induced seismicity in the Kayen, R., R. E. S. Moss, E. M. Thompson, R. B. Seed, K. O. Cetin, A. Der
Groningen gas field.” Bull. Earthquake Eng. 17 (8): 4539–4557. https:// Kiureghian, Y. Tanaka, and K. Tokimatsu. 2013. “Shear-wave velocity-
doi.org/10.1007/s10518-018-0489-3. based probabilistic and deterministic assessment of seismic soil lique-
Green, R. A., A. Bradshaw, and C. D. P. Baxter. 2022. “Accounting for faction potential.” J. Geotech. Geoenviron. Eng. 139 (3): 407–419.
intrinsic soil properties and state variables on liquefaction triggering https://doi.org/10.1061/(ASCE)GT.1943-5606.0000743.
in simplified procedures.” J. Geotech. Geoenviron. Eng. 148 (7): Kayen, R. E., and J. K. Mitchell. 1997. “Assessment of liquefaction po-
04022056. https://doi.org/10.1061/(ASCE)GT.1943-5606.0002823. tential during earthquakes by arias intensity.” J. Geotech. Geoenviron.
Green, R. A., M. Cubrinovski, B. Cox, C. Wood, L. Wotherspoon, Eng. 123 (12): 1162–1174. https://doi.org/10.1061/(ASCE)1090-0241
B. Bradley, and B. Maurer. 2014. “Select liquefaction case histories (1997)123:12(1162).
from the 2010–2011 Canterbury earthquake sequence.” Earthquake Kokusho, T., and Y. Kaneko. 2018. “Energy evaluation for liquefaction-
Spectra 30 (1): 131–153. https://doi.org/10.1193/030713EQS066M. induced strain of loose sands by harmonic and irregular loading tests.”
Green, R. A., B. W. Maurer, and S. van Ballegooy. 2018. “The influence of Soil Dyn. Earthquake Eng. 114 (Nov): 362–377. https://doi.org/10
the non-liquefied crust on the severity of surficial liquefaction manifes- .1016/j.soildyn.2018.07.012.
tations: Case history from the 2016 Valentine’s Day earthquake in Kokusho, T., and S. Tanimoto. 2021. “Energy capacity versus liquefaction
New Zealand.” In Geotechnical earthquake engineering and soil
strength investigated by cyclic triaxial tests on intact soils.” J. Geotech.
dynamics V, liquefaction triggering, consequences, and mitigation,
Geoenviron. Eng. 147 (4): 04021006. https://doi.org/10.1061/(ASCE)
21–32. Reston, VA: ASCE.
GT.1943-5606.0002484.
Green, R. A., and J. K. Mitchell. 2003. “A closer look at arias intensity-
Kramer, S. L., and R. A. Mitchell. 2006. “Ground motion intensity mea-
based liquefaction evaluation procedures.” In Proc., 7th Pacific Conf.
sures for liquefaction hazard evaluation.” Earthquake Spectra 22 (2):
on Earthquake Engineering. Christchurch, New Zealand: Univ. of
413–438. https://doi.org/10.1193/1.2194970.
Canterbury.
Lamens, P., and A. Askarinejad. 2021. “Pile driving and submarine slope
Green, R. A., and J. K. Mitchell. 2004. “Energy-based evaluation and
stability: A hybrid engineering approach.” Landslides 18: 1351–1367.
remediation of liquefiable soils.” In Geo-Trans 2004: Geotechnical
https://doi.org/10.1007/s10346-020-01585-2.
Engineering for Transportation Projects, Geotechnical Special Publi-
cation 223, edited by M. K. Yegian and E. Kavazanjian, 1961–1970. Lasley, S. J. 2015. “Application of fatigue theories to seismic compression
Reston, VA: ASCE. estimation and the evaluation of liquefaction potential.” Ph.D. disserta-
tion, Dept. of Civil and Environmental Engineering, Virginia Tech.
Green, R. A., and J. K. Mitchell. 2010. “Comparison of energies required to
densify liquefiable soil.” In Proc., 5th Int. Conf. Recent Advances in Lasley, S. J., R. A. Green, and A. Rodriguez-Marek. 2016. “New stress
Geotechnical Earthquake Engineering and Soil Dynamics and Symp. reduction coefficient relationship for liquefaction triggering analyses.”
in Honor of Professor I.M. Idriss. Rolla, MO: Missouri Univ. of Science J. Geotech. Geoenviron. Eng. 142 (11): 06016013. https://doi.org/10
and Technology. .1061/(ASCE)GT.1943-5606.0001530.
Green, R. A., J. K. Mitchell, and C. P. Polito. 2000. “An energy-based ex- Lasley, S. J., R. A. Green, and A. Rodriguez-Marek. 2017. “Number
cess pore pressure generation model for cohesionless soils.” In Proc., of equivalent stress cycles for liquefaction evaluations in active tectonic
John Booker Memorial Symp. Developments in Theoretical Geome- and stable continental regimes.” J. Geotech. Geoenviron. Eng. 143 (4):
chanics, edited by D. W. Smith and J. P. Carter, 383–390. Rotterdam, 04016116. https://doi.org/10.1061/(ASCE)GT.1943-5606.0001629.
Netherlands: A.A. Balkema. Lee, K. L., and H. B. Seed. 1967. “Cyclic stress conditions causing lique-
Green, R. A., and E. Rodriguez-Arriaga. 2019. “Assessment of an alterna- faction of sands.” J. Soil Mech. Found. Div. 93 (1): 47–70. https://doi
tive implementation of the Dobry et al. cyclic strain procedure for evalu- .org/10.1061/JSFEAQ.0000945.
ating liquefaction triggering.” In Proc., 7th Int. Conf. on Earthquake Luco, N., and C. A. Cornell. 2007. “Structure-specific scalar intensity
Geotechnical Engineering (7ICEGE). London: International Society measures for near-source and ordinary earthquake ground motions.”
for Soil Mechanics and Geotechnical Engineering. Earthquake Spectra 23 (2): 357–392. https://doi.org/10.1193/1
Green, R. A., and G. A. Terri. 2005. “Number of equivalent cycles concept .2723158.
for liquefaction evaluations—Revisited.” J. Geotech. Geoenviron. Eng. Martin, G. R., W. D. L. Finn, and H. B. Seed. 1975. “Fundamentals of
131 (4): 477–488. https://doi.org/10.1061/(ASCE)1090-0241(2005) liquefaction under cyclic loading.” J. Geotech. Eng. Div. 101 (5):
131:4(477). 423–438. https://doi.org/10.1061/AJGEB6.0000164.
Hayati, H., and R. D. Andrus. 2009. “Updated liquefaction resistance cor- Maurer, B. W., R. A. Green, M. Cubrinovski, and B. A. Bradley. 2014.
relation factors for aged sands.” J. Geotech. Geoenviron. Eng. 135 (11): “Evaluation of the liquefaction potential index for assessing liquefac-
1683–1692. https://doi.org/10.1061/(ASCE)GT.1943-5606.0000118. tion hazard in Christchurch, New Zealand.” J. Geotech. Geoenviron.
Hryciw, R. D., S. Vitton, and T. G. Thomann. 1990. “Liquefaction and flow Eng. 140 (7): 04014032. https://doi.org/10.1061/(ASCE)GT.1943-5606
failure during seismic exploration.” J. Geotech. Eng. 116 (12): 1881– .0001117.
1899. https://doi.org/10.1061/(ASCE)0733-9410(1990)116:12(1881). Maurer, B. W., R. A. Green, M. Cubrinovski, and B. A. Bradley.
Idriss, I. M., and R. W. Boulanger. 2008. Soil liquefaction during earth- 2015. “Fines-content effects on liquefaction hazard evaluation for infra-
quakes. Monograph MNO-12. Oakland, CA: Earthquake Engineering structure in Christchurch, New Zealand.” Soil Dyn. Earthquake Eng.
Research Institute. 76 (3): 58–68. https://doi.org/10.1016/j.soildyn.2014.10.028.

© ASCE 04023105-15 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 Menq, F.-Y. 2003. “Dynamic properties of sandy and gravelly soils.” Ph.D. Seed, H. B. 1983. “Earthquake resistant design of earth dams.” In Proc.,
dissertation, Dept. of Civil, Architectural, and Environmental Engineer- Symp. on Seismic Design of Embankments and Caverns, 41–64. Reston,
ing, Univ. of Texas at Austin. VA: ASCE.
Moss, R. E. S. 2003. “CPT-based probabilistic assessment of seismic soil Seed, H. B., and I. M. Idriss. 1971. “Simplified procedure for evaluat-
liquefaction initiation.” Ph.D. dissertation, Dept. of Civil and Environ- ing soil liquefaction potential.” J. Soil Mech. Found. Div. 97 (9):
mental Engineering, Univ. of California. 1249–1273. https://doi.org/10.1061/JSFEAQ.0001662.
Moss, R. E. S., R. B. Seed, R. E. Kayen, J. P. Stewart, A. Der Kiureghian, Simcock, K. J., R. O. Davis, J. B. Berrill, and G. Mullenger. 1983. “Cyclic
and K. O. Cetin. 2006. “CPT-based probabilistic and deterministic as- triaxial tests with continuous measurement of dissipated energy.”
sessment of in situ seismic soil liquefaction potential.” J. Geotech. Geo- Geotech. Test. J. 6 (1): 35–39. https://doi.org/10.1520/GTJ10822J.
environ. Eng. 132 (8): 1032–1051. https://doi.org/10.1061/(ASCE) Taylor, O.-D. S. 2011. “Use of energy-based liquefaction approach to pre-
1090-0241(2006)132:8(1032). dict deformation in silts due to pile driving.” Ph.D. dissertation, Dept. of
Naesgaard, E., and P. M. Byrne. 2007. “Flow liquefaction simulation Civil and Environmental Engineering, Univ. of Rhode Island.
Thum, T. S., S. Lasley, R. A. Green, and A. Rodriguez-Marek. 2019.
using a combined effective stress–Total stress model.” In Proc., 60th
Downloaded from ascelibrary.org by VIRGINIA TECH UNIVERSTIY on 09/06/23. Copyright ASCE. For personal use only; all rights reserved.

ShakeVT2: A computer program for equivalent linear site response

Canadian Geotechnical Conf. Ottawa: Canadian Geotechnical Society.
analysis. Center for Geotechnical Practice and Research (CGPR)
Nemat-Nasser, S., and A. Shokooh. 1979. “A unified approach to densifi-
Rep. #98. Blacksburg, VA: The Charles E. Via, Jr., Department of Civil
cation and liquefaction of cohesionless sand in cyclic shearing.” Can.
and Environmental Engineering, Virginia Tech.
Geotech. J. 16 (4): 659–678. https://doi.org/10.1139/t79-076. Ulmer, K. J. 2019. “Development of an energy-based liquefaction evalu-
NRC (National Research Council). 2016. State of the art and practice in the ation procedure.” Ph.D. dissertation, Dept. of Civil and Environmental
assessment of earthquake-induced soil liquefaction and its consequen- Engineering, Virginia Tech.
ces. Committee on earthquake induced soil liquefaction assessment Ulmer, K. J., R. A. Green, and A. Rodriguez-Marek. 2020. “A consistent
(Edward Kavazanjian, Jr., Chair, Jose E. Andrade, Kandian “Arul” correlation between VS , SPT, and CPT metrics for use in liquefaction
Arulmoli, Brian F. Atwater, John T. Christian, Russell A. Green, evaluation procedures.” In Geo-Congress 2020: Geotechnical Earth-
Steven L. Kramer, Lelio Mejia, James K. Mitchell, Ellen Rathje, quake Engineering and Special Topics, Geotechnical Special Publica-
James R. Rice, and Yumie Wang). Washington, DC: NRC. tion 318, edited by J. P. Hambleton, R. Makhnenko, and A. S. Budge,
Ostadan, F., N. Deng, and I. Arango. 1998. “Energy-based method for 132–140. Reston, VA: ASCE.
liquefaction potential evaluation.” In Proc., 11th European Conf. on Upadhyaya, S. 2019. “Development of an improved and internally-
Earthquake Engineering. Rotterdam, Netherlands: A.A. Balkema. consistent framework for evaluating liquefaction damage potential.”
Peck, R. B. 1979. “Liquefaction potential: Science versus practice.” J. Geo- Ph.D. dissertation, Dept. of Civil and Environmental Engineering,
tech. Eng. Div. 105 (3): 393–398. https://doi.org/10.1061/AJGEB6 Virginia Tech.
.0000776. Upadhyaya, S., R. A. Green, A. Rodriguez-Marek, and B. W. Maurer. 2022.
Polito, C., R. A. Green, E. Dillon, and C. Sohn. 2013. “The effect of “True liquefaction triggering curve.” J. Geotech. Geoenviron. Eng.
load shape on the relationship between dissipated energy and residual 149 (3): 04023005. https://doi.org/10.1061/JGGEFK.GTENG-11126.
excess pore pressure generation in cyclic triaxial tests.” Can. Geotech. J. Upadhyaya, S., B. W. Maurer, R. A. Green, and A. Rodriguez-Marek.
50 (11): 1118–1128. https://doi.org/10.1139/cgj-2012-0379. 2018. “Effect of non-liquefiable high fines-content, high plasticity soils
Polito, C. P., R. A. Green, and J. Lee. 2008. “Pore pressure genera- on liquefaction potential index (LPI) performance.” In Geotechnical
tion models for sands and silty soils subjected to cyclic loading.” J. Geo- Earthquake Engineering and Soil Dynamics V: Liquefaction Trigger-
tech. Geoenviron. Eng. 134 (10): 1490–1500. https://doi.org/10.1061 ing, Consequences, and Mitigation, Geotechnical Special Publication
/(ASCE)1090-0241(2008)134:10(1490). 290, edited by S. J. Brandenberg and M. T. Manzari, 191–198. Reston,
Pyke, R. 1979. “Nonlinear soil models for irregular cyclic loading.” J. Geo- VA: ASCE.
tech. Eng. Div. 105 (6): 715–726. https://doi.org/10.1061/AJGEB6 USGS. 2019. “ShakeMap.” Accessed September 12, 2019. https://
earthquake.usgs.gov/data/shakemap/.
.0000820.
Wair, B. R., J. T. Dejong, and T. Shantz. 2012. Guidelines for estimation of
Riemer, M. F., W. B. Gookin, J. D. Bray, and I. Arango. 1994. Effects of
shear wave velocity profiles. Rep. No. 2012/08. Berkeley, CA: Univ. of
loading frequency and control on the liquefaction behavior of clean
California.
sands. Rep. No. UCB/GT/94-07. Berkeley, CA: Univ. of California.
Yee, E., J. P. Stewart, and K. Tokimatsu. 2013. “Elastic and large-strain
Rodriguez-Arriaga, E., and R. A. Green. 2018. “Assessment of the cyclic nonlinear seismic site response from analysis of vertical array record-
strain approach for evaluating liquefaction triggering.” Soil Dyn. Earth- ings.” J. Geotech. Geoenviron. Eng. 139 (10): 1789–1801. https://doi
quake Eng. 113 (Oct): 202–214. https://doi.org/10.1016/j.soildyn.2018 .org/10.1061/(ASCE)GT.1943-5606.0000900.
.05.033. Youd, T. L., and S. N. Hoose. 1977. “Liquefaction susceptibility and geo-
Rodriguez-Marek, A., E. M. Rathje, J. J. Bommer, F. Scherbaum, and P. J. logic setting.” In Proc., 6th World Conf. on Earthquake Engineering,
Stafford. 2014. “Application of single-station sigma and site-response 2189–2194. Tokyo: International Association for Earthquake
characterization in a probabilistic seismic-hazard analysis for a new Engineering.
nuclear site.” Bull. Seismol. Soc. Am. 104 (4): 1601–1619. https://doi Zeghal, M., and A.-W. Elgamal. 1994. “Analysis of site liquefaction using
.org/10.1785/0120130196. earthquake records.” J. Geotech. Eng. 120 (6): 996–1017. https://doi
Seed, H. B. 1979. “Soil liquefaction and cyclic mobility evaluation for level .org/10.1061/(ASCE)0733-9410(1994)120:6(996).
ground during earthquakes.” J. Geotech. Eng. Div. 105 (2): 201–255. Zimmaro, P., et al. 2019. “Next-generation liquefaction.” Earthquake
https://doi.org/10.1061/AJGEB6.0000768. Spectra 36 (2): 939–959. https://doi.org/10.1177/8755293020902477.

© ASCE 04023105-16 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2023, 149(11): 04023105

 Corrections

Erratum for “Energy-Based Liquefaction

Triggering Model”
2  31
K. J. Ulmer, A.M.ASCE ΔW 3
0
σvo
6 liq 7
Research Engineer, Dept. of Geosciences and Engineering, Southwest FSL ¼ 4  5 for qc1Ncs ≤ 180 ð21aÞ
Research Institute, San Antonio, TX 78238. ORCID: https://orcid.org ΔW
Downloaded from ascelibrary.org by 2001:468:c80:610e:7d17:ebeb:9b96:379a on 09/04/24. Copyright ASCE. For personal use only; all rights reserved.

0
σvo
/0000-0002-8384-4721. Email: [email protected]

R. A. Green, F.ASCE 0
where ðΔW=σvo Þliq is estimated using Eq. (20). For cases where
Professor, Dept. of Civil and Environmental Engineering, Virginia Tech, qc1Ncs > 180, qc1Ncs ¼ 180 is used in conjunction with Eq. (20)
0
Blacksburg, VA 24061 (corresponding author). ORCID: https://orcid.org to estimate ðΔW=σvo Þliq .
/0000-0002-5648-2331. Email: [email protected] Alternatively, FSL can be estimated from the probability of
liquefaction triggering (PL ) using the following equations:
A. Rodriguez-Marek, M.ASCE
1
Professor, Dept. of Civil and Environmental Engineering, Virginia Tech, 1 b
Blacksburg, VA 24061. Email: [email protected] FSL ¼ a × −1 for PL ≥ 10−4 ;
PL
 1
J. K. Mitchell, Hon.M.ASCE 1 b
FSL ¼ a × − 1 for PL < 10−4 ð21bÞ
University Distinguished Professor Emeritus, Dept. of Civil and Environ- 10−4
mental Engineering, Virginia Tech, Blacksburg, VA 24061. Email: jkm@
vt.edu where a ¼ 0.821; b ¼ 3.15; and PL is estimated using Eq. (19).
Eqs. (21a) and (21b) result in approximately the same values of
The following correction should be made to “Energy-Based FSL , with Eq. (21b) tending to yield slightly larger values of FSL
Liquefaction Triggering Model” by K. J. Ulmer, R. A. Green, than Eq. (21a) for FSL > 2.
A. Rodriguez-Marek, and J. K. Mitchell, https://doi.org/10 The authors apologize for the oversight of not correcting
.1061/JGGEFK.GTENG-11402: Eq. (21) before publication and assure readers that no other part
Eq. (21) for the factor of safety against liquefaction triggering of the paper’s content or conclusions is affected by this.
(FSL ) is incorrect and should be replaced by the following To preserve the published version of record, these details have
equation: been corrected only in this erratum.

© ASCE 08224002-1 J. Geotech. Geoenviron. Eng.

View publication stats J. Geotech. Geoenviron. Eng., 2024, 150(11): 08224002