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

net/publication/332544374

Evaluation of Earthquake-Induced Cracking of Embankment Dams

Conference Paper · April 2019

CITATIONS READS
0 420

2 authors, including:

Lelio H. Mejia
Geosyntec Consultants Inc, Oakland, California
80 PUBLICATIONS   513 CITATIONS   

SEE PROFILE

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

Chabot Dam Seismic Stability Evaluation View project

Panama Canal Expansion Project View project

All content following this page was uploaded by Lelio H. Mejia on 20 May 2020.

The user has requested enhancement of the downloaded file.

 Proceedings of 2019 USSD Annual Conference,
Chicago, Illinois

EVALUATION OF EARTHQUAKE-INDUCED CRACKING OF

EMBANKMENT DAMS

Lelio Mejia, PhD, PE, GE 1

Ethan Dawson, PhD 2

ABSTRACT

One of the most hazardous consequences of earthquakes on embankment dams is

cracking associated with embankment deformation induced by earthquake shaking.
Cracking is commonly observed at the crest of dams and can occur in various
orientations, but most often develops as fractures that are roughly parallel to the dam axis
(i.e. longitudinal cracking) or transverse to it. Transverse cracking at the crest is of major
concern and commonly represents a potential failure mode for embankment dams
because, if it extends below the reservoir level, it may lead to leakage from the reservoir.
If unimpeded, leakage through cracks can lead to continued internal erosion of
embankment materials, piping, and failure of a dam.

Considerable effort has been made by the dam engineering profession over the last few
decades to study the nature and effects of earthquake-induced cracking of embankment
dams and to develop defensive measures to minimize its occurrence and mitigate its
potential consequences. Extensive work has also been devoted to developing methods to
evaluate the location, extent, and depth of cracking in dams.

The purpose of this paper is to provide an overview of currently available methods to

assess the potential for earthquake-induced cracking of embankment dams and propose
an approach to the evaluation of potential embankment dam cracking. The paper provides
a summary of previous studies, a discussion of the mechanics of cracking in soils, and an
overview of available methods for evaluating the location, orientation, and depth of
earthquake-induced cracking in embankment dams. The evaluation of potential cracking
using three-dimensional (3D) nonlinear dynamic analysis is illustrated through
application of such procedures to the analysis of the response of Lenihan Dam during the
M 6.9, 1989 Loma Prieta Earthquake.

INTRODUCTION

One of the most hazardous consequences of earthquakes on embankment dams is cracks

induced by strong earthquake shaking. Foundation fault rupture during an earthquake can
also lead to cracking of an embankment dam. Co-seismic foundation rupture, however, is
a much less frequent occurrence in dams than strong shaking.

Cracks are caused by tensile stresses that exceed the tensile strength of soils. Because of
their temporary nature, transient tensile stresses that might be induced during earthquake
shaking are unlikely to leave open cracks in soils after the shaking ceases. In

1
Geosyntec Consultants, Oakland, CA, [email protected]
2
AECOM, Los Angeles, CA, [email protected]

1
embankment dams, open cracks (from here on simply referred to as cracks) observed
after an earthquake are most often associated with tensile stresses and strains resulting
from earthquake-induced permanent deformation of the embankment.

Dam embankment cracks can form in various orientations and shapes. Because of the
sustained stress field imposed by gravity, cracks will often be roughly vertical. However,
depending on soil tensile strength and on geometric constraints and stiffness
discontinuities that may lead to soil arching, cracks can take non-vertical orientations.

Earthquake-induced deformations often correspond to transverse spreading, bulging,

and/or settlement of embankment dams (e.g. see Swaisgood, 2014). Transverse spreading
of embankment dams often leads to tensile stresses normal to the dam axis near the crest,
whereas dam settlement can lead to tensile stresses parallel to the axis (particularly at the
crest near steep rock abutments and above irregularities in the longitudinal foundation
profile). Thus, earthquake-induced cracks in embankment dams typically develop near
the crest, in directions approximately parallel to the dam axis (longitudinal cracks) or
transverse to it (transverse cracks). However, cracks are also often observed on the
embankment upstream and downstream faces (e.g. see Fong and Bennett, 1995). Figure
1 illustrates some of the mechanisms of cracking often observed in embankment dams
after earthquakes (Sherard et al., 1963).

(a) Longitudinal cracking (b) Transverse cracking

Figure 1. Example mechanisms of earthquake-induced cracking of embankment dams

Cracking due to static deformation and hydraulic fracturing is known to be pervasive at

depth within dam embankments, as discussed in the landmark paper by Sherard (1973).
Permanent seismic deformations often exacerbate static deformations of an embankment
dam and enlarge cracking observed under static conditions at depth or near the surface.
As pointed out by Sherard, seismically-induced cracking is frequently similar in pattern
and location to cracking commonly caused by differential settlement under gravity
loading.

Whether at depth or near the crest, cracks can lead to increased seepage or leakage and
internal soil erosion, which if allowed to progress uncontrolled, can lead to piping failure
of embankment dams. Cracks located and oriented in such ways that allow concentrated
flows through an embankment dam, even under moderate hydraulic gradients, are
particularly dangerous. Transverse cracks that extend below the reservoir level and are
long enough or interconnect to cut through the embankment impervious zone and allow
flows to exit unfiltered into coarse pervious zones or at the downstream slope, may

2
readily initiate internal erosion and lead to failure (Mejia, 2013). Longitudinal cracks are
also hazardous, as they can interconnect transverse cracks and provide pathways for
leakage through a dam.

Clearly, assessment of the potential for earthquake-induced cracking is critically

important to the seismic safety evaluation of embankment dams. Reliable evaluation of
the likely location and geometric characteristics of potential cracks is essential for a
rigorous evaluation of potential failure modes related to cracking of an embankment dam.

Considerable effort has been made by the dam engineering profession over the last few
decades to study the nature and effects of earthquake-induced cracking of embankment
dams and to develop defensive design measures to minimize its occurrence and mitigate
its potential consequences. Extensive work has been devoted to developing methods to
evaluate the location, extent, and depth of cracking in dams (e.g. Pells and Fell, 2002).
However, few references are available in the literature (e.g. Fell et al., 2015; USBR,
2015) that provide an overview of available methods for the evaluation of earthquake-
induced cracking of embankment dams.

The purpose of this paper is to provide a summary of previous studies, a brief discussion
of the mechanics of cracking in soils, and an overview of available methods for
evaluating the location, orientation, and depth of earthquake-induced cracking in
embankment dams. The paper also presents an approach to the evaluation of potential
embankment dam cracking using three-dimensional (3D) nonlinear dynamic analysis
procedures. The use of such procedures for evaluating potential earthquake-induced
cracking of embankment dams is illustrated by applying them to the response analysis of
Lenihan Dam for the M 6.9, 1989 Loma Prieta Earthquake.

Cracking in embankment dams may also result from desiccation of the dam materials or
from hydraulic fracturing, in the absence of general embankment deformation. However,
discussion of these two latter causes of cracking is beyond the scope of this paper. So is
cracking of non-earth water barriers in embankment dams, including impervious
upstream facings.

OVERVIEW OF PREVIOUS STUDIES

Because of the similarities between various aspects of embankment dam cracking under
seismic and static loads, general research on cracking under static conditions is relevant
to the evaluation of seismically-induced cracking. Previous research related to the
evaluation of cracking of embankment dams under gravity loads includes several
underpinning studies dating back to the seminal work of Leonards and Narain (1963) and
of Covarrubias (1969), who studied tensile strains at failure in compacted clays and
cracking of earth and rockfill dams. Research on the use of finite element analysis
procedures to calculate stresses in dams under static conditions, which are necessary to
evaluate potential cracking, included insightful studies by Lefebvre et al. (1973),
Lefebvre and Duncan (1974) and Chirapuntu and Duncan (1975).

3
Lefebvre et al. (1973) performed two- and three-dimensional (2D and 3D) finite element
static stress analyses of embankment dams for various dam canyon geometries. They
showed that 2D plane-strain analyses of the crest longitudinal section of a dam yield
reasonably accurate estimates of the minor principal stresses and displacements for dams
with crest length to height ratios, L/H, as low as 2. This finding has been used by multiple
researchers since, to study the potential for transverse cracking of embankment dams by
means of 2D finite element stress analyses (e.g. see Hoeg, 1995).

Using 2D finite element models of the maximum longitudinal and transverse sections of a
hypothetical dam on a compressible foundation, Lefebvre and Duncan (1974) studied the
sensitivity of calculated stresses to various assumptions and input parameters. One of
their most noteworthy conclusions is that gravity turn-on analyses can significantly
overpredict calculated zones of tension relative to analyses that simulate the dam
construction sequence. They also concluded that for analysis interpretation, stress is
preferable to strain as a criterion for crack formation, because it is possible to have
extensional strains superimposed on compressive stresses with no resulting net tension.

Chirapuntu and Duncan (1975) studied the development of tension cracks in compacted
fills placed in layers over compressible foundations. They used finite element methods
with nonlinear soil stress-strain models and crack propagation techniques. A key finding
of their work was that once initiated, cracks propagate beyond the limits of tension zones
calculated in analyses with high tensile strength. As they propagate, cracks relieve tensile
stresses in surrounding and nearby areas. Thus, analyses of embankment dams that allow
the development of significant tensile stresses are only valid until cracking begins and
can significantly underestimate the extent of potential cracks.

Notable compilations of field observations of embankment dam cracking during

earthquakes have been published by Fong and Bennett (1995) and by Pells and Fell
(2002). Fong and Bennett reported on transverse cracking observed at 19 dams after the
1989 M 6.9 Loma Prieta, the 1992 M 7.4 Landers, and the 1994 M 6.4 Northridge
earthquakes. Pells and Fell significantly expanded the Fong and Bennett database to
include longitudinal cracking and other types of earthquake damage and compiled
observations from 305 dams during more than 70 earthquakes. In forward evaluations of
dam safety, those compilations often offer useful benchmarks for assessing potential
crack characteristics in an embankment dam, such as crack depths and widths, based on
estimates of seismically-induced dam deformations. An example of the use of such
compilations for assessing embankment cracking under strong earthquake shaking was
presented by Mejia and Dawson (2015).

Recent research has been carried out by Professors Robin Fell and Chongmin Song and
other researchers from the University of New South Wales in Australia and has focused
on numerical evaluation of the characteristics of cracking under static loading. For
example, Bui et al. (2005) performed 2D and 3D stress analyses of construction of
hypothetical embankment dams. They also performed 2D cross-valley stress analyses for
several existing dams known to have exhibited cracking in their past. Bui et al. postulated
that, for many purposes, 2D analyses of an embankment’s cross-valley longitudinal

4
section are sufficiently accurate to assess the likelihood of transverse cracking under
static loads. He et al. (2014) extended the work of Bui et al. (2005) and studied the
potential for cracking in embankment dams during construction using the results of 2D
stress and deformation analyses of dam cross-valley sections. They concluded that
abutment irregularities, such as large steps in a rock abutment, lead to zones of tensile
strain and transverse cracking due to differential settlement of the embankment as the
dam is constructed and after construction.

FIELD OBSERVATIONS AND CASE HISTORIES

Previous observations of dam cracking during earthquakes have been very useful in
developing an understanding of the potential characteristics of cracks and of the
conditions under which seismically-induced cracking may occur in embankment dams.
Cracks are often observed in dams after earthquakes and post-earthquake reconnaissance
reports commonly document observations of earthquake-induced cracking of
embankment dams (e.g. see http://www.geerassociation.org/reconnaissance-reports).

By far, the 12 May 2008 M 8.0 Wenchuan Earthquake in China affected a greater number
of dams than any other earthquake in recent history. As reported by Jing et al. (2011), the
Wenchuan Earthquake shook over 35,600 dams and damaged over 2,660 dams, of which
69 were severely damaged. Figure 2 shows examples of the types of observed damage in
embankment dams during the earthquake.

Extensive cracking was observed in 68 of the severely damaged embankment dams. The
dams were subjected to ground motions with seismic intensities in the range of VI to IX
(Jing et al., 2011). Other effects observed in the severely damaged dams included: sliding
(16 dams), settlement (35 dams), leakage (25 dams), and bulge deformation (15 dams).
Longitudinal and transverse cracks were observed at the crest and on the upstream and
downstream slopes, and transverse cracks were observed at the abutments.

(a) Longitudinal cracking of Xinyoufang Dam (b) Sliding of Yuejing Dam slope

Figure 2. Examples of severe damage in embankment dams during the 2008 M 8.0
Wenchuan Earthquake (From Jing et al., 2011)

5
Numerous other large earthquakes have occurred since the Pells and Fell (2002) database
was compiled. However, few newly published case histories include quantitative survey
data on the geometric characteristics of cracks or the engineering properties of the
cracked materials, and most of the published information on observed dam cracking is of
a qualitative nature.

The reasons for this perceived deficiency in published quantitative data on earthquake-
induced embankment dam cracking are not well known. The authors suspect, however,
that such deficiency is likely due to: a) insufficient emphasis within the profession on the
importance of disseminating quantitative survey data on embankment dam cracking, b)
lack of instrumentation specifically aimed at facilitating the measurement of embankment
deformations that lead to cracking, and c) lack of guidelines for the measurement and
documentation of embankment dam cracking after earthquakes. The authors believe that
this perceived deficiency could be corrected by development of guidelines and protocols
for appropriate survey and documentation of dam cracking after earthquakes.

Previous observations indicate that the interface between concrete walls (or other types of
rigid walls) and embankment fills are particularly vulnerable to cracking during
earthquakes. Such vulnerability was exemplified by the cracking observed near the
spillway structure at Austrian Dam during the 1989 M 7 Loma Prieta earthquake in
California. Figure 3 illustrates the locations where cracks were observed at the dam after
the earthquake and the extent of cracking seen near the spillway walls (Babbitt, 2014).

(a) Locations of observed cracking at the dam (b) Cracks near spillway walls

Figure 3. Cracking observed at Austrian Dam after the 1989 M 7 Loma Prieta Earthquake

Concrete or masonry walls founded on rock may be considered to offer a rigid interface
with a dam fill, which can crack readily if the fill tends to pull away from the wall. The
vulnerability of such interfaces to cracking is exacerbated by the low tensile strength of
the interface and the ability of the wall to stand unsupported. Thus, under those or similar
conditions, such as at the contact between a dam embankment and a steep rock abutment,
it is necessary to provide protection against cracking and internal erosion of the

6
embankment. An effective protection is often the placement of downstream filter zones
composed of cohesionless materials that cannot support open cracks.

CONSIDERATIONS FROM MECHANICS OF CRACKING

In concept, a soil will crack when subjected to tensile stress that exceeds the tensile
strength of the soil. This concept is consistent with the basic hypothesis of rupture for
solids, which postulates that fracture of a solid may be expected if the maximum tensile
stress exceeds a critical value, which defines the tensile strength of the material (Griffith,
1920).

Soils generally exhibit brittle behavior during tensile failure, such that they are unable to
sustain tensile stress once the tensile strength is exceeded. Figure 4 shows the results of
strain-controlled direct-pull tests by Tang et al. (2015) and of load-controlled beam
bending tests by Chirapuntu and Duncan (1975). It may be seen that once the peak tensile
load was reached, the soil specimens quickly lost their ability to carry load. Such brittle
behavior results from the fact that cracking in soils is a fracture-dominant type of failure
as opposed to a yield-dominant failure.

(a) Strain-controlled direct-pull tests (b) Load-controlled beam tests

(Tang et al., 2015) (Chirapuntu and Duncan, 1975)

Figure 4. Typical load-displacement relationships from tensile strength tests in clayey soil

Crack propagation is controlled by the principles of fracture mechanics (e.g. see Griffith,
1920; Janssen et al., 2006). Once a crack initiates, stress concentrations (singularities) at
the ends of the crack dictate propagation of the crack. In soils, cracking will propagate
until the driving stress or strain conditions change, or the capacity of the soil to support
an open crack is exceeded. As shown in Figure 4, in tensile strength tests with a sustained
driving stress or strain, cracking propagates quickly until the soil specimen fractures
completely. In the field, as cracks propagate within an embankment zone, the driving
stress can change spatially (and temporally), or the ability of the soil to support an open
crack may be exceeded, thus determining the extent and depth of cracking. As they
propagate, cracks themselves change the state of stress, relieving tensile stresses in

7
nearby areas. As observed by Terzaghi (1943), sustained tensile stresses lead sooner or
later to the formation of cracks in soils.

Notwithstanding the fundamental nature of cracking as a process controlled by fracture

mechanics, the initiation of a crack in soils considered as intact on a macroscopic scale
may be approached as a strength-of-materials problem. Thus, the potential for initial
cracking of a soil may be evaluated by comparing the tensile stress in the soil to its
tensile strength.

The tensile strength of compacted clayey soil in the laboratory has been shown to be a
function of soil type (including gradation and plasticity), specimen size, loading type and
rate, and density and water content at the time of fracture. Figure 5 shows the variation in
tensile strength with density (or compaction effort) and water content of compacted
clayey soils of medium plasticity (CL) from direct-pull tests by Tang et al. (2015) and
beam bending tests by Chirapuntu and Duncan (1975). For a given density, the soil
tensile strength varies significantly with placement water content and reaches a maximum
value at a water content of a few percent dry of optimum.

(a) Direct-pull tests – Silty Clay (b) Beam tests – Sandy Clay
(Tang et al., 2015) (Chirapuntu and Duncan, 1975)

Figure 5. Variation in tensile strength with density and water content for compacted
clayey soils

CONSIDERATIONS FOR ASSESSING FIELD BEHAVIOR

Because all soils have flaws and crack initiation and propagation is controlled by such
flaws, the tensile strength of soils is highly dependent on specimen size. Durelli and
Parks (1962) found the following relationship between the tensile strength of brittle
materials (σt) and the volume of that part of the specimen subjected to at least 95% of the
maximum tensile stress (V95):
σt = C(V95)n

where C and n are constants. The results of ring and Brazilian tests by Harison et al.,
(1994) correspond to a value of n ≈ -0.2 for a silty clay soil (CL). Such type of

8
information may be used to infer the tensile strength of intact soil in the field. It suggests,
for example, that the tensile strength of intact clayey embankment material at the scale of
small cracks often observed in the field may be 1/10 to 1/30 of the tensile strength
measured in laboratory tests such as direct-pull or beam bending tests. It follows that the
tensile strength of most embankment dam materials in the field is typically very low.

The potential for cracking in embankment dams depends on many factors including: a)
the material characteristics such as soil type and plasticity, degree of saturation, and
stress-strain-strength behavior under shear, b) the geometry of the embankment and
foundation (e.g. foundation irregularities), and c) the embankment deformation (e.g.
differential settlement). Cracking may also result from desiccation of the embankment
materials or from hydraulic fracturing, in the absence of overall embankment
deformation. However, discussion of these two latter causes of cracking is beyond the
scope of this paper.

Multiple factors determine the geometric characteristics of cracks in a dam, including

their spatial distribution, orientation, length, depth, and width. Such characteristics are
controlled by the pattern of dam deformation (which determines the distribution of
stresses and strains in the embankment) and by the principles of fracture mechanics that
control crack propagation. Crack widths are determined by the magnitude and spatial
extent of tensile strains and are roughly correlated to the depth of cracking.

The depth of cracking below the ground surface is limited by the cohesion component (or
apparent cohesion) of shear strength, which determines the ability of a soil to support an
open crack. The maximum depth of cracking below the surface, (Dcrack)max, in a soil is
given approximately by the following expression:

(Dcrack)max = 2c Tan (45°+φ/2) / γ

where c = cohesion, φ = friction angle, and γ = unit weight of the soil, applicable to the
pertinent load condition. It follows that coarse cohesionless soils, such as gravels, are
unable to support open cracks and thus, cannot crack. Partially saturated fine-to-medium-
grained cohesionless soils, however, may exhibit ‘apparent cohesion’ due to capillary
tension. As shown in Figure 6, medium sands in a moist condition, can support open
cracks temporarily to considerable depths. However, once they become saturated, the
materials lose their ‘apparent cohesion’ and collapse, closing any cracks (Mejia, 2013).

PROCEDURES FOR EVALUATION OF EARTHQUAKE-INDUCED

CRACKING

Although procedures for the evaluation of earthquake-induced cracking of embankment

dams are not well established, various approaches have been used in practice to evaluate
potential cracking for seismic safety evaluation and risk assessment of embankment
dams. These procedures may be generally classified into two broad categories: a) indirect
evaluation of the potential for cracking through uncoupled analyses of embankment
seismic deformations and crack development, and b) direct evaluation through numerical

9
analysis in which the extent of cracking is inferred directly from the analysis results or is
a direct output of the analysis.

Because seismic deformations of embankment dams often exacerbate static deformations,

seismically-induced cracking is frequently similar in pattern and location to cracking
under gravity loading. Thus, any evaluation of earthquake-induced cracking of an
embankment dam should be preceded by a thorough examination of its performance
(observed or anticipated) during construction and long-term loading. Observed or
anticipated differential settlement under static loads is often a good indicator of potential
differential settlement under seismic loads. In addition, static differential settlement may
reveal zones of weakness or discontinuities in the embankment and foundation that may
promote cracking. Even if previously repaired, cracking during construction and long-
term loading conditions will often create weaknesses in the embankment and/or
foundation, where earthquake-induced cracking may be expected.

(a) Trench excavation in moist sand (b) Trench collapse after flooding

Figure 6. Trench tests in moist medium sand before and after saturation

Indirect Procedures

In these procedures, the evaluation of cracking is generally done in two steps. Seismic
deformations of the embankment dam are calculated first using one or more of currently
available techniques. Using the calculated embankment deformations as input, the
potential for cracking is then evaluated using simplified methods or numerical analyses.

Methods to evaluate embankment dam seismic deformations can range from empirical
correlations (e.g. Swaisgood, 2014), to the Newmark sliding-block method of analysis, to
numerical analysis using non-linear finite element methods (e.g. see Perlea and Beaty,
2010). Procedures to assess potential cracking range from empirical use of case history
databases to numerical analyses of stresses and strains for the expected embankment dam
deformations.

10
Case history databases that may be used to empirically assess potential earthquake-
induced cracking include those published by Fong and Bennett (1995) and Pells and Fell
(2002). Mejia and Dawson (2015) presented a case history on the use of these databases,
together with calculated crest settlements from 2D nonlinear dynamic analyses, to assess
potential earthquake-induced crack depths and widths in the embankments of a large
canal.

Notwithstanding uncertainties inherent to calculated embankment dam seismic

deformations, the main issue with the empirical use of case history databases is the large
uncertainty associated with the resulting cracking estimates. Such uncertainty is due to: a)
the limited number of cases in the databases, and b) the few parameters used to represent
the many factors that affect cracking in any one case. The large uncertainty in the
resulting cracking estimates may be inferred from Figure 7, which illustrates data for
normalized dam crest settlement and crack depth compiled in the Fong-and-Bennett and
Pells-and-Fell databases. As shown in Figure 7, the number of case histories in the
databases is limited and the scatter in the data is large.

Figure 7. Normalized crest settlement and crack depth data in published databases

Basic assessments of potential transverse cracking have occasionally been made in

practice based on simplified kinematic analysis of the longitudinal strains associated with
differences between calculated deformations for various transverse sections of a dam
(e.g. Mejia et al., 2014). Conceptually, numerical plane-strain analyses of stresses and
strains in the longitudinal section of a dam can also be used to evaluate potential
transverse cracking due to seismic deformations calculated at transverse sections of the
dam. The main issues with this numerical approach, however, lie in the lack of analytical
rigor and inaccuracies associated with imposing on the longitudinal section of a dam the
deformations calculated for its transverse sections. Therefore, these types of numerical
analyses are not common in practice.

Direct Procedures

In these procedures, the extent of cracking is estimated directly from or is a direct output
of numerical dynamic response analysis of a dam. Two-dimensional analyses of

11
transverse sections of a dam may be used to examine the potential for longitudinal
cracking. However, 3D dynamic response analyses are necessary to examine the potential
for transverse cracking. Three-dimensional analyses also allow evaluation of longitudinal
cracking and of the potential for cracking at the dam abutments. The use of 3D numerical
analyses to examine the potential for cracking due to foundation fault rupture on
embankment dams has been discussed by Mejia and Dawson (2012).

Three-dimensional dynamic response analyses have seldom been used to evaluate

potential cracking of embankment dams because of their complexity. Key considerations
in performing these types of analyses include: a) the significant effort involved in
generating a representative analysis mesh, b) the need to appropriately model boundary
conditions that account for the unbounded nature of a dam system, c) the selection of
input motions that properly represent the complex wave field generated by an earthquake
and are consistent with the choice of system boundaries for analysis, and d) the use and
characterization of material constitutive models that adequately simulate 3D stress-strain
behavior under shear, compression, and tension loading.

Accurate simulation of crack propagation requires the use of techniques for analysis of
continuous media that allow for strain and fracture localization, or the use of
discontinuous models that allow cracks to be explicitly modeled. The latter include
numerical methods that incorporate adaptive meshing, to allow the mesh to conform to
the geometry of crack propagation, and the principles of fracture mechanics, and are
generally not practical for routine dam engineering applications involving earthquake
loading. Analysis methods for continuous media are more practical for such dam
applications.

If precise modeling of crack location and geometry is not required for analysis of a dam,
it is usually satisfactory to use a strength-of-materials approach. Although this approach
does not account for the process of crack propagation, it is acceptable for approximate
evaluation of cracking location, orientation, extent, and depth, provided suitable
nonlinear constitutive models with an appropriate tensile strength (i.e. tension cutoff) are
used in the analysis. Typically, a zero or small tension cutoff is appropriate for analysis,
considering that the field tensile strength of most embankment dam materials is expected
to be very low.

In the strength-of-materials approach, the location, extent, orientation, and depth of

cracks are given by the calculated extension strains and zones where the minor principal
stress reaches the tension cutoff. A useful index to identify zones of tension failure after
loading is the ratio Rt defined as:

Rt = (σ3final-σ3initial)/σ3initial

where σ3initial is the initial value of the minor principal stress and σ3final is the final value
after loading (Mejia and Dawson, 2012). Zones of tension failure correspond to values
close to Rt = -1 for this ratio.

12
For special dam applications, simulation of crack propagation with analysis methods for
continuous media may be achieved by using ‘crack band’ models (Bazant, 1982), or other
strain-softening-localization models, or by using interface elements. Such elements are
emplaced in an analysis mesh at pre-established locations for cracks. These may
correspond to the known locations of pre-existing cracks, or be selected, for example,
based on the results of strength-of-materials analyses.

Proposed Approach for Evaluation of Earthquake-induced Cracking

A practical phased approach to the evaluation of earthquake-induced cracking of existing

embankment dams might consist of the following steps. These steps should be
implemented as necessary, in sequence of increasing complexity and effort, depending on
the importance and characteristics of a dam project:

1. Examine the observed or anticipated performance of the dam under construction

and long-term loading conditions. For existing dams, this should include
examining the history and location of previous settlement, lateral deformation,
and known cracking, if any.
2. Estimate seismic settlements of the dam using simplified methods of deformation
analysis, such as: empirical correlations, analytically-based correlations, or
sliding-block analyses.
3. For the estimated seismic settlements, assess the potential for cracking
empirically (including crack depths and widths) using available case-history
databases.
4. Using 2D models of selected transverse sections of the dam and a strength-of-
materials approach, evaluate seismic deformations and the potential
characteristics of longitudinal cracking at the selected dam sections.
5. For the seismic deformations estimated from the 2D analyses, assess the potential
for transverse cracking using simple kinematic considerations and case-history
databases.
6. Using 3D models and a strength-of-materials approach, evaluate seismic
deformations and the potential characteristics of longitudinal and transverse
cracking in the dam.

For new dams, the approach would consist of the same steps, but be based on the
anticipated characteristics of the dam and its foundation. For large and important dams,
the potential for settlement during construction and long-term loading conditions would
typically be estimated using finite element analysis procedures and would include
consideration of cracking under static loads.

ANALYSIS OF LENIHAN DAM FOR THE 1989 LOMA PRIETA

EARTHQUAKE

The procedures for 3D analysis of earthquake-induced cracking of embankment dams,

using a strength-of-materials approach, are illustrated by applying them to the seismic
response analysis of Lenihan Dam for the 1989 Loma Prieta earthquake. The response of

13
the dam during the earthquake was previously analyzed by Makdisi et al. (1991) using
2D equivalent-linear analysis methods, and by Mejia et al. (1992) using 2D nonlinear
dynamic analyses procedures. The dam response to the earthquake was also recently
analyzed by Hadidi et al. (2014) and Armstrong et al. (2018) using similar methods.
However, none of these previous studies used 3D analysis models or included explicit
evaluation of potential cracking of the dam during the earthquake.

Dam Description

Lenihan Dam is located in the Santa Cruz mountains of California about 13 miles
northwest of the epicenter of the M 6.9 1989 Loma Prieta earthquake. The dam is a 207-
foot-high zoned earthfill with upstream and downstream slopes of 5:1 and 3:1,
respectively. The crest of the dam is 40 feet wide and about 810 feet long. A plan view
and the maximum cross section of the dam are shown in Figure 8.

(a) Plan view (b) Maximum section

Figure 8. Plan view and maximum cross section of Lenihan Dam

The dam is founded on Franciscan sandstone and shale bedrock. The downstream shell
consists of clayey sandy gravel with about 15 to 40% fines of low to medium plasticity.
The materials in the upstream shell and the upper 80 feet of the core are generally similar
and consist of clayey sands with about 15 to 35% gravel and 20 to 50% fines of medium
plasticity. The materials in the core below 80 feet from the crest are distinct from those
above and consist of a clay of medium to high plasticity. Additional information on the
characteristics and engineering properties of the embankment and foundation materials is
presented by Mejia et al. (1992) and Hadidi et al. (2014).

Recorded Response and Observed Cracking

Three accelerographs, located as shown in Figure 8(a), recorded the ground motions at
the site during the Loma Prieta earthquake and other earthquakes. The ground motions
recorded at these instruments during the Loma Prieta earthquake are described in
previous studies (e.g. Makdisi et al., 1991). The instrument at the left abutment recorded
peak ground accelerations of 0.44 g and of 0.41 g in the upstream-downstream and cross-
canyon directions, respectively. The maximum crest settlement measured after the

14
earthquake was about 10 inches. The extent and locations of surface cracks at the dam
were surveyed within days after the earthquake by the dam owner, the Santa Clara Valley
Water District. The pattern of earthquake-induced cracking at the dam was digitized from
these surveys by Hadidi et al. (2014) and is shown in Figure 9. Also shown are the
measured embankment displacements at the crest survey monuments.

Figure 9. Locations of surface cracks after the 1989 Loma Prieta Earthquake
(after Hadidi et al., 2014)

Three-Dimensional Dynamic Analysis

The 3D dynamic analyses of the dam were performed using the computer program
FLAC3D (Itasca, 2013). Constitutive soil models based on a rigorous mechanics
framework for dynamic elasto-plastic behavior of soils in 3D were used. Only the
analyses using the linear-elastic-perfectly-plastic Mohr-Coulomb model, built into the
FLAC3D code, are presented here. Further description of the analyses performed using
the Mohr-Coulomb model and more sophisticated constitutive models, and an in-depth
examination of the dam’s recorded response using 3D and 2D analysis models, are
presented by Dawson and Mejia (2019).

Figure 10 shows the 3D numerical mesh used in the analyses. The model includes the
dam embankment and a significant extent of the bedrock foundation and abutments,
beyond the limits of the embankment body, to allow for proper simulation of dynamic
interaction between the dam and the rock domain. Compliant and free-field boundaries
are included at the base and sides of the model. The earthquake motions recorded at the
left abutment (Figure 8(a)) were used to develop the input tractions applied at the base of
the model.

The Mohr-Coulomb model was used to simulate the stress-strain behavior of the
embankment materials. The foundation rock was assumed to be elastic. The key
parameters of the Mohr-Coulomb model are the elastic constants of the materials, which

15
control material behavior before yield, and the shear strength, which defines the limits of
elastic behavior and the stress state and plastic strains at yield. A key parameter for
analysis of potential cracking is the tensile strength of the materials. The analyses were
performed assuming a zero-tension cutoff for the embankment materials. This cutoff
properly represents the expected low field tensile strength of the embankment materials
and ensures that the extent of cracking is adequately captured by the analysis.

Figure 10. Three-dimensional model for dynamic analysis of Lenihan Dam

The analyses were used to calculate the earthquake ground motions at the locations of the
crest accelerographs, the embankment displacements at the locations of the crest survey
monuments, and the distribution of tensile stresses and strains in the dam after the
earthquake. Figure 11 shows the distribution of maximum horizontal tensile strains at the
embankment surface after the earthquake.

Figure 11. Plan view of maximum surface horizontal tensile strains after the earthquake

16
Near the ground surface, horizontal tensile strain reflects tensile stress because the static
horizontal stresses under gravity loading are small. Thus, near the surface, significant
tensile strains are directly associated with cracking. As shown in Figure 11, the
distribution of tensile strains closely resembles the distribution of surface cracking
observed at the dam after the earthquake, that is, mainly transverse cracking at the
abutments and along the groins and longitudinal cracking high on the dam upstream face
(Figure 9). Thus, it may be concluded that this type of analysis, even with a very simple
constitutive model, can be useful in evaluating the potential for and the locations of
earthquake-induced cracking in embankment dams.

CONCLUDING REMARKS

Transverse cracking at the crest commonly represents a potential failure mode for
embankment dams because it can lead to leakage from the reservoir, internal erosion and
piping, and dam breaching. An overview of previous studies, key aspects of the
mechanics of cracking in soils, and currently available methods to assess the potential for
earthquake-induced cracking of embankment dams has been presented.

It is noted that most of the published information on observed dam cracking during past
earthquakes is of a qualitative nature. Further, it is suggested this perceived deficiency
could be corrected by development of guidelines and protocols for appropriate survey and
documentation of dam cracking after earthquakes.

In addition, a practical, phased approach to the evaluation of potential embankment dam

cracking has been proposed. The approach consists of a series of steps to be implemented
in order of increasing complexity and effort, as necessary depending on the importance
and characteristics of a dam. The process begins with an examination of dam
performance under construction and long-term loading and, for important dam projects,
culminates with the use of 3D nonlinear analyses of dynamic response using a strength-
of-materials approach to estimate potential cracking.

The use of 3D nonlinear dynamic analysis procedures to evaluate potential cracking is

illustrated by analysis of the response of Lenihan Dam during the M 6.9, 1989 Loma
Prieta Earthquake. It is shown that this type of analysis, if properly implemented, can be
useful for evaluating the potential for earthquake-induced cracking in embankment dams,
and can indicate the likely locations of potential cracks.

ACKNOWLEDGMENTS

Impetus for this paper arose from the senior author’s work on a risk analysis sponsored
by Grant County Public Utilities District No. 2 of Washington for a major dam project.
The authors thank Dr. Jacquelyn Allmond for her assistance in collecting and reviewing
data from published case histories of embankment dam cracking. Presentation of this
paper was funded by Geosyntec Consultants. This support is gratefully acknowledged.

17
 REFERENCES

Armstrong, R.J., D.S. Park, and A. Gullen. 2018. “Cyclic soil behavior of common
constitutive models used in non-linear deformation analysis of embankment dams.”
Proceedings 2018 USSD Conference.

Babbitt, D.H. 2014. “Transverse Cracking of Embankment Dams Caused by Earthquakes

Revisited.” Proceedings of 34th Annual USSD Conference, San Francisco, California,
April.

Bazant, Z.P. 1982. “Crack Band Model for Fracture of Geomaterials.” Proceedings, 4th
Int. Conf. on Num. Meth. in Geomechanics, Edmonton, Alberta, Canada, June.

Bui, H.N., V. Tandrijana, R. Fell, C. Song, and N. Khalili. 2005. “Two and Three-
Dimensional Numerical Analysis of The Potential for Cracking of Embankment Dams –
Supplementary Report.” University of New South Wales, UNICIV Report R-438,
January.

Chirapuntu, S. and J.M. Duncan. 1975. “The Role of Fill Strength in The Stability of
Embankments on Soft Clay Foundations.” UCB Report TE 75-3, University of
California, Berkeley.

Covarrubias, S.W. 1969. “Cracking of Earth and Rockfill Dams.” Harvard Soil
Mechanics Series, No. 82, Cambridge, Massachusetts.

Dawson, E. and L. Mejia. 2019. “3-D Analyses of Lenihan Dam for the Loma Prieta
Earthquake.” In preparation.

Durelli, A.J. and V. Parks. 1962. “Relationship of size and stress gradient to brittle-
failure stress.” Proc., 4th U.S. Nat. Congr. Of Appl. Mech., ASME, New York, N.Y.,
931-938.

Fell, R., J.P. Macgregor, D.H. Stapledon, G. Bell, and M. Foster. 2015. Geotechnical
Engineering of Dams. Second Edition. Taylor Francis, London.

Fong, F.C. and W.J. Bennett. 1995. “Transverse Cracking on Embankment Dams Due to
Earthquakes.” Proceedings, ASDSO Western Regional Conference, Red Lodge,
Montana.

Griffith, A.A. 1920. “The phenomena of rupture and flow in Solids.” Transactions of the
Royal Society of London, Vol 221 (1921) 163-198.

Hadidi, R., Y. Moriwaki, J. Barneich, R. Kirby, M. Mooers. 2014. Seismic Deformation

Evaluation of Lenihan Dam Under 1989 Loma Prieta Earthquake. Tenth U.S. National
Conference on Earthquake Engineering, Frontiers of Earthquake Engineering,
Anchorage, Alaska, 21-25 July.

18
Harison, J.A., B.O. Hardin, and K. Mahboub. 1994. “Fracture Toughness of Compacted
Cohesive Soils Using Ring Test.” Journal of Geotechnical Engineering, 120(5): 872-891,
May.

He, K., C. Song, and R. Fell. 2014. “Assessing potential cracking zones in embankment
dams.” 23rd Australasian Conference on the Mechanics of Structures and Materials,
December, Southern Cross University, Lismore, New South Wales.

Hoeg, K. 1995. “Transverse Cracking in Embankment Dams.” Norwegian Geotechnical

Institute, Report No. 532060-1, March.

Itasca Consulting Group. 2013. “Fast Lagrangian Analysis of Continua in 3 Dimensions.”

User’s Guide Version 5.0, Itasca Consulting Group: Minneapolis, Minnesota.

Janssen, M., J. Zuidema, and R. Wanhill. 2006. Fracture Mechanics, VSSD, Delft, The
Netherlands.

Jing, L., Liang, H., Li, Y., and Liu, C. 2011. “Characteristics and Factors that Influenced
Damage to Dams in the Ms 8.0 Wenchuan Earthquake.” Earthquake Engineering and
Engineering Vibration, 10(3): 349-358. September.

Lefebvre, G., J.M. Duncan, and E.L. Wilson. 1973. “Three-Dimensional Finite Element
Analysis of Dams.” ASCE Journal of Soil Mechanics and Foundation Engineering, Vol.
99, No. SM7, pp 495-507.

Lefebvre, G. and J.M. Duncan. 1974. “Finite Element Analyses of Transverse Cracking
in Low-Embankment Dams.” U.S. Army Engineer Waterways Experiment Station,
Contract Report S-74-3.

Leonards, G.A. and J. Narain. 1963. “Flexibility of Clay and Cracking of Earth Dams.”
ASCE Journal of Soil Mechanics and Foundation Engineering, Vol. 89, No. SM2, pp 47-
68.

Makdisi, F.L. C.Y. Chang, Z.L. Wang, and C.M. Mok. 1991. “Analysis of the Recorded
Response of Lexington Dam During Various Levels of Ground Shaking.” Proceedings of
SMIP91 Seminar, Strong Motion Instrumentation Program, California Division of Mines
and Geology, May.

Mejia, L. 2013. “Field Testing of Crushed Ignimbrite for Dam Filter Material.”
Proceedings of United States Society on Dams (USSD) Annual Meeting and Conference,
Phoenix, Arizona. 11-15 February.

Mejia, L. and E. Dawson. 2012. “Analysis of Dam Response Under Foundation

Faulting.” Proceedings of United States Society on Dams (USSD) Annual Meeting and
Conference, New Orleans, Louisiana. 23-27 April.

19
 Mejia, L. and E. Dawson. 2015. “Analysis of Seismic Deformations and Cracking for
Embankments of the Tekapo Canal.” Proceedings of United States Society on Dams
(USSD) Annual Meeting and Conference, Louisville, Kentucky. 13-17 April.

Mejia, L.H., J.I. Sun, S. Salah-Mars, Y. Moriwaki, and M. Beikae. 1992. “Non-linear
dynamic response analysis of Lexington Dam.” Proceedings of SMIP92 Seminar, Strong
Motion Instrumentation Program, California Division of Mines and Geology, May.

Mejia, L., J. Wu, E. Newman, and M. Mooers. 2014. “Seismic Stability Evaluation of
Dam Underlain by Coarse-Grained Alluvium.” Proceedings of United States Society on
Dams (USSD) Annual Meeting and Conference, San Francisco, California. April.

Pells, S. and R. Fell. 2002. “Damage and Cracking of Embankment Dams by

Earthquakes, and the Implications for Internal Erosion and Piping.” Proceedings, 20th
International Congress on Large Dams, International Commission on Large Dams
(ICOLD).

Perlea, V. G. and M.H. Beaty. 2010. “Corps of Engineers practice in the evaluation of
seismic deformation of embankment dams.” Proc., Fifth International Conference on
Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics. San
Diego, CA, paper SPL 6.

Sherard, J.L. 1973. “Embankment Dam Cracking.” Embankment Dam Engineering, S.

Poulos ad R. Hirschfeld, Eds., John Wiley and Sons, New York, N.Y., pp. 272-353.

Sherard, J.L., Woodward, R.J., Gizienski, S.F. and Clevenger, W.A. (1963). Earth and
earth rock dams. John Wiley & Sons.

Swaisgood, J.R. 2014. “Behavior of embankment dams during earthquake.” Journal of

Dam Safety, 12 (2): 35-44.

Tang, C., X. Pei, D. Wang, B. Shi, and J. Li. 2015. “Tensile Strength of Compacted
Clayey Soil.” Journal of Geotechnical and Geoenvironmental Engineering, 141(4).
April.

Terzaghi, K. 1943. Theoretical Soil Mechanics. John Wiley and Sons, New York, N.Y.,
p. 37, 97.

United States Bureau of Reclamation (USBR). 2015. Reclamation: Managing Water in

the West, Design Standards Number 13, Embankment Dams. Chapter 13: Seismic
Analysis and Design. May.

20

View publication stats