Received: 9 March 2020 Revised: 22 July 2020 Accepted: 9 September 2020

DOI: 10.1002/eqe.3352

RESEARCH ARTICLE

Seismic performance of two classes of earth dams

Luca Masini1 Sebastiano Rampello1 Riccardo Donatelli2

1Department of Structural and

Geotechnical Engineering, University of In the last decades, Italy has experienced several strong earthquakes, triggering
Rome La Sapienza, Rome, Italy the attention on the high seismic risk associated with the potential instability of
2 Arup Italia s.r.l., Milan, Italy large existing earth dams. Under intense seismic events, dam response depends
Correspondence
on various factors including the geometry and the type of water retention
Luca Masini, Department of Geotechnical scheme. This paper explores the influence of these factors, comparing the
and Structural Engineering, University of seismic performance of two idealized earth dams: a homogeneous dam and a
Rome La Sapienza, Rome, Italy.
Email: [email protected] zoned dam. The simplified models adopted in the analyses are representative
of these two classes of earth dams, often adopted in the sixties to design various
dams located in southern Italy. The idealized models, here examined in detail,
have comparable heights and foundation soil conditions, differing in terms
of water retention scheme. The seismic response of the dams was evaluated
through time-domain nonlinear dynamic analyses in which the same input
ground motion, represented by real-time histories, was applied at the base of
the models. It is shown that for highly inelastic systems, such as those at hand,
duration of input motion should be accounted for in addition to compatibility
criteria with a design elastic response spectrum. Also, different slopes of dam
flanks, shear strength mobilization at the end of dam construction and high
stiffness contrast at the shells-core contacts are shown to result in different
deformation patterns at the end of earthquake loading. Some peculiarities in the
behavior of these classes of earth dams are highlighted in the paper, providing
a guidance for the rational assessment of their seismic performance.

KEYWORDS
dynamic numerical analysis, earth dam, earthquake, seismic performance, nonlinear soil
behavior

1 INTRODUCTION

Large earth dams are strategic infrastructure as they supply local communities with large water resources for several
purposes, promoting livelihood, particularly in developing countries. On the other hand, the safety and security of a dam
can affect life and property of persons across wide areas. One of the risk factors is related to the age of these structures:
from the ICOLD data collected by Foster et al.1–3 on earth dams built up to 1986 that had experienced problems, Fell et al.4
concluded that the number of failures of large earth dams built after 1950 was 1 in 200, while the occurrence of incidents
increases dramatically for dams built before 1950, being 1 in 25. According to this study, failure is most frequently related
to internal piping and overtopping, while only 1.7% occurred during, or soon after, seismic events. In 2014 the U.S. Society
of Dams published the results of a study on the performance of dams during earthquakes.5,6 About 30 case histories were
analyzed, 3 of them related to seismic events with a magnitude larger than 8. Only in 12 cases failure occurred, among

Earthquake Engng Struct Dyn. 2020;1–20. wileyonlinelibrary.com/journal/eqe © 2020 John Wiley & Sons Ltd. 1
2 MASINI et al

them being the Lower San Fernando dam7 and the Fujinuma dam.8 Most often, the key factor affecting the performance
of large earth dams is related to the materials employed for the construction and the quality of compaction. Old hydraulic
fill dams or loosely compacted dams are prone to experience severe damages during earthquakes, while well-compacted
dams with not liquefiable materials typically show relatively good performances.
In Italy there are more than 500 large dams, with height above 15 m or with a water reservoir larger than one million
cubic meters, 165 of them being earth dams. Those located in Southern Italy mainly consist of homogeneous or zoned earth
dams built between 1950 and 1960, their age giving rise to some critical issues that must be addressed for the safety of the
people living in the downstream area. First, the updating of the seismic hazard map of the Italian territory lead to a reclas-
sification of construction sites previously considered as nonseismic or to an increase of the design seismic loads. Second,
the analysis procedures originally adopted to design such structures are obsolete in light of the modern performance-based
design approach that is now adopted in most technical regulations or building codes, including the new Italian Building
Code for Dams—NTD-2014.9
The rare occurrence of failures induced by earthquakes may be somewhat misleading: on the one hand, it confirms
the high safety factors that were typically adopted for the design of large earth dams; on the other one, the small percent-
age of collapses may be attributed to the rare occurrence of seismic events stronger than the design seismic action. As a
consequence, the failure mechanisms against which seismic safety is assessed are often derived from analyses performed
assuming critical seismic scenarios, rather than from direct observations.
Nonlinear dynamic analyses performed with either the finite-element (FE) or finite-difference (FD) method can be
a very effective tool to analyze the seismic behavior of earth dams, as they enable to simulate complex geometries and
boundary conditions, realistically reproducing the soil behavior by the adoption of constitutive soil models characterized
by different possible levels of complexity.
The evaluation of the seismic performance of large earth dams through numerical analyses can be performed following
two approaches: in the first approach, a numerical model is conceived for an actual existing structure, based on a specific
characterization of soil parameters and, when available, on site measurements of real seismic events, usually of low ampli-
tude; this model is then subjected to scenario earthquakes associated to increasing return periods, to evaluate the portions
of the structure where damage occurs and the physical quantities that better represent this damage (e.g., Oroville dam,
USA10 ; Santa Felicia dam, USA11 ; Marana Capacciotti dam, Italy12–15 ; El Infernillo dam, Mexico16 ; Menta dam, Italy17 ; La
Villita dam, Mexico18 ; Yele dam, China19 ).
In the second approach, a conveniently simplified model is adopted to study the response of typical structural and
geotechnical systems.20–23 In this case, the adopted constitutive models are usually simpler and typological models of the
dams are subjected to pseudostatic or time-domain dynamic analyses to assess the patterns of deformation in the dam
body and to evaluate relevant quantities, such as permanent displacements, that should not be intended as a prediction
of actual performance of the earth structure, but rather as an index of seismic performance.
This latter approach is followed in this paper to compare the seismic performance of a homogeneous and a zoned earth
dam. To this aim, two numerical models were conceived, characterized by geometries inspired by two existing dams, the
Marana Capacciotti homogeneous dam and the San Pietro zoned dam, that represent well-documented case histories with
comprehensive geotechnical characterization and monitoring data collected during the construction and impoundment
stages. As such, these two idealized models are representative of two classes of earth dams differing for the water retention
scheme; they preserve the main features of the reference dams in terms of dimensions and soil properties but discard some
peculiarities which characterize each of them in detail. Here, the adoption of the simplified schemes of the dams is aimed
at highlighting the typical patterns of seismic behavior of these classes of earth dams, rather than at comparing the behav-
ior of two specific existing dams, as proposed elsewhere.24 At this scope, a hysteretic elastic-plastic constitutive model
was adopted, which represents a reasonable compromise between level of complexity and predictive capability. In fact,
alternative, more rigorous and accurate constitutive laws, which might be worth adopting to analyze the seismic behavior
of a specific case study, often require nontrivial parameters, calibration through nonstandard laboratory tests, and com-
parably not straightforward procedures to initialize their numerous internal variables, as such making them unsuitable in
the context of the research activity discussed in this paper. In detail, the adopted constitutive model is an effective stress-
based nonlinear elastic-perfectly plastic one, capable of reproducing the main aspects of the soil behavior during seismic
actions, such as shear modulus degradation, hysteretic damping development, and excess pore water pressure build-up.
In case of a severe earthquake loading, it accounts for the highly dissipative soil behavior stemming from the development
of the large plastic strains resulting from transient mobilization of shear strength during seismic loading.
Nonlinear dynamic analyses were carried out to compare the seismic performance of the two models assuming the
same height of the dam and impoundment level, as well as the same mechanical properties of the fine-grained soil of the
MASINI et al 3

(A)

(B)

FIGURE 1 Layout of the models adopted for the numerical analyses: (A) zoned dam and (B) homogeneous dam

embankments and the foundation soils, as inspired by real in-situ and laboratory investigations. Under these assumptions
it is shown that the steeper slope of the shells of the zoned dam results in higher shear strength mobilization at the end
of dam construction and higher ground motion amplification during seismic shaking, thus leading to larger permanent
displacements at the end of earthquake loading compared to those computed for the homogeneous dam, despite the
lower strength of this latter. The overall results of the analyses discussed in this paper provide an insight into the seismic
performance of the two examined types of dams, which can be generalized and used as a guidance in the preliminary
assessment of the seismic and postseismic behavior of similar real structures.

2 OUTLINE OF THE SIMPLIFIED MODELS

The idealized dam models discussed in the following were inspired by two existing earth structures located in Southern
Italy. The first is the San Pietro dam, a zoned dam with a vertical fine-grained core, along the Osento stream (Avellino).
With reference to its main cross-section, depicted with a thin line in Figure 1A, the dam is 49-m high and has a base
of 250 m; it overlays a foundation soil composed by a layer of alluvial gravels, with an average thickness of 7 m and a
deeper stiff and overconsolidated flysch deposit. The second embankment is a homogeneous earth dam along the Marana
Capacciotti stream (Foggia), built between 1970 and 1975. Its main cross-section is shown in Figure 1B with a thin line: the
dam, 48-m high with a base of 370 m, overlays a layer of medium stiff alluvial silt with a thickness of about 12 m, and a
deeper stiff and overconsolidated deposit of silty clay. These dams were selected as they are characterized by similar heights
of the embankment (H ≈ 50 m) and by comparable conditions of the foundation soil. They are also well-documented
case histories with a comprehensive geotechnical characterization for the embankment and the foundation soils with
monitoring data are available for both the construction and the impounding stages (San Pietro dam24–26 and Marana
Capacciotti dam12,13,27,28 ).
The main cross-sections of Marana Capacciotti and San Pietro dams were converted in two simplified plane-strain
numerical models, shown with a thick line in Figures 1A and 1B for the zoned and homogeneous model dams of height
4 MASINI et al

(A)

(B)

FIGURE 2 Finite difference grids adopted in the analyses. (A) Zoned dam and (B) homogeneous dam

H = 50 m. The adoption of 2D numerical schemes was deemed to be appropriate due to the large crest-length to dam-height
ratio, with L/H = 9.4 and 17.4 for the zoned and the homogeneous dam, respectively. In these conditions, the stiffening
effects observed for dams in narrow valleys, with natural periods decreasing and near-crest accelerations increasing as the
canyon becomes narrower, are expected to be small.11,29–31
The zoned dam has a width of 240 m and a constant slope of the flanks, α = 23◦ . The vertical fine-grained core
extends from the crest to the base of the embankment, with a maximum width of 30 m. An impervious diaphragm
is embedded into the lower firm soil to prevent seepage through the alluvial soil underlying the dam. The drainage
system at the downstream toe has been modeled as a 40-m-long horizontal drain. The homogeneous model dam has
a width of 400 m and a constant slope of the flanks, α = 14◦ . The drainage system was modeled with a subvertical
central drain about 47-m deep and with a horizontal drain, about 100-m long, located at the toe of the downstream
slope. As for the zoned dam, seepage through the foundation soil is prevented by an impervious diaphragm modeled
with a column of elements having the same stiffness of the adjacent soil and a permeability three order of magnitude
lower.
The two sites are characterized by similar foundation soils, with a shallow alluvial soil resting over a firm and deep
clayey deposit. Therefore, to compare the seismic performance of the idealized models of the homogeneous and zoned
earth dams, only differing for the water retention scheme while being characterized by the same other ingredients of
the problem, the same foundation soil was assumed, made by a 15-m-thick layer of alluvial deposit overlaying a 35-m-
deep layer of firm soil. The analyses were performed assuming the same water level of 45 m, which can be considered as
representative of the maximum storage level.
Figure 2 shows the FD grids adopted in the analyses. They are composed of the same number of quadrilateral zones
(13 224), with the minimum size of 1.59 m in the embankment and the maximum size smaller than one-eighth of the
smallest wavelength (lowest shear wave velocity over the highest frequency of the input motion). The grids extend 150 m
from the sides of the dam and 50 m below it to ensure a negligible interaction of the boundaries with the dam. During
the static analysis, displacements in both directions were restrained at the bottom boundary of the mesh and horizontal
displacements were restrained at the sides.
MASINI et al 5

TA B L E 1 Soil mechanical properties

Soil γ (kN/m3 ) c′ (kPa) φ′ (◦ ) k0 (−) k (m/s) A (MPa) B (MPa) n (−) ν (−)
Dam–fine grained 20.8 20 28 − 10−7 19.48 1.573 0.75 0.30
−6
Foundation soil 20.6 3 40 1.5 10 560 0 0 0.32
Firm soil 20.6 − − 1.5 10−9 2060 0 0 0.32
−6
Dam shells 23.0 5 38 − 10 3.2 56 0.50 0.20

A) B)

FIGURE 3 Model calibration for the fine-grained soil of the embankments: (A) shear strength envelope; (B) small-strain stiffness

3 CALIBRATION OF SOIL MODEL

The cyclic behavior of the soil was described through the hysteretic damping model Sigmoidal4 implemented in FLAC,32
coupled with the same Mohr–Coulomb plasticity criterion adopted in the pseudostatic analyses. The hysteretic damping
model is essentially an extension to two dimensions of the nonlinear soil models that describe the unloading–reloading
stress–strain cycles using the Masing (1926) rules.
The strength and stiffness parameters were obtained from the geotechnical investigations carried out at the sites of
the two existing dams, as reported by Calabresi et al.25 for the San Pietro dam and by Calabresi et al.27 for the Marana
Capacciotti dam (Table 1). Laboratory tests were carried out on undisturbed samples retrieved from the fine-grained soils
of the dam body of Marana Cappacciotti dam and the core of San Pietro dam. The strength envelope obtained from the
triaxial tests is plotted in Figure 3A; data points from both dams define a single strength envelope with φ′ = 28◦ and c′ =
20 kPa. The small-strain shear modulus G0 obtained from resonant column (RC) and bender elements tests are plotted
against p′ for both dams in Figure 3B; the results are in fair agreement, though slightly larger and more scattered values
are observed for the samples retrieved from the core of the zoned dam. In the analyses, the small-strain shear modulus
G0 was expressed as a function of the mean effective stress p′ using the equation33
)𝑛
𝑝′
(
𝐺0 = 𝐴 + 𝐵 ⋅ , (1)
𝑝ref

where pref = 1 kPa is a reference pressure. Values of parameters A, B, and n, selected to match the data of Figure 3B, are
listed in Table 1.
For the much stiffer coarse-grained material of the shells of the zoned dam, Equation (1) was calibrated using the results
of two cross-hole (CH) tests (Figure 4A), while the strength parameters c′ and φ′ listed in Table 1 were obtained from in-
situ and laboratory tests.25 Specifically, dynamic cone penetration tests and standard penetration tests were carried out
through the shells of San Pietro dam, showing a range of relative density DR = 50–70%34,35 and of the angle of shearing
resistance φ′ = 35◦ –40◦ , which are typical of dense coarse-grained soils. Laboratory triaxial tests were also performed on
large undisturbed samples (diameter D = 80 mm and height H = 160 mm) retrieved from the dam shells, providing values
of c′ = 0 kPa and φ′ = 37◦ . In this study, an angle of shearing resistance φ′ = 38◦ was adopted assuming a small effective
6 MASINI et al

A) B)

FIGURE 4 Small strain stiffness profile for (A) the shell soil of the zoned dam and (B) the foundation soil

(A) (B)

F I G U R E 5 Comparison between RC data and model simulations. Shear modulus and damping curves for (A) the fine-grained soil and
(B) the shell and foundation soil

cohesion c′ = 5 kPa to avoid development of unrealistically shallow plastic mechanisms at the dam shells during dynamic
calculations.
For the foundation layer underneath the dams, an equivalent alluvial deposit was assumed in the analyses with strength
and stiffness properties representative, on average, of the foundation soils of the above sites. This to focus the attention
solely on the influence of the water retention scheme, disregarding any effect due to different foundation soils. To this
aim, strength parameters c′ = 3 kPa and φ′ = 40◦ have been assumed (Table 1), while Equation (1) was calibrated against
the CH tests available for the foundation soil of San Pietro dam (Figure 4B).
The stiff deposit lying underneath the alluvial layer was regarded as an elastic bedrock characterized by a shear wave
velocity Vs = 1000 m/s, with small-train shear modulus G0 = 2060 MPa and Poisson’s ratio ν′ = 0.32.
The hysteretic damping model here adopted to describe the cyclic soil behavior requires the definition of the small-
strain shear modulus G0 and a shear modulus decay curve to describe the dependency of the secant shear modulus on the
shear strain. The former was expressed as a function of the mean effective stress by Equation (1), while the shear modulus
MASINI et al 7

TA B L E 2 Parameters of the hysteretic model

Soil a (−) b (−) x0 (−) y0 (−)
Dam – fine grained 0.99 −0.479 −1.249 0.01
Dam shells 1.00 −0.657 −1.400 0
Foundation soil 1.00 −0.657 −1.400 0

decay curve is written as

𝑎
𝑀S = 𝑦0 + log 𝛾(%)−𝑥0
, (2)
−
1+𝑒 𝑏

where Ms is the normalized shear modulus. Parameters a, b, x0, and y0 were calibrated to reproduce the modulus decay
curves obtained from the RC tests performed on undisturbed samples of the fine-grained soils, while for the coarse-
grained materials of the shells of the zoned dam and the foundation soil curves proposed by Rollins et al.36 for coarse-
grained soils were used. Figure 5 shows the comparison between the computed and the observed curves for both the
fine-grained soil (Figure 5A) and the coarse-grained one (Figure 5B). Table 2 lists the stiffness parameters adopted in the
analyses.
The dynamic analyses were carried out in terms of effective stresses assuming fully undrained soil behavior during seis-
mic loading. However, since the Mohr–Coulomb constitutive model does not provide any volumetric/deviatoric coupling,
it was combined with the hysteretic model in which the deviator-induced excess pore water pressure ∆u is evaluated using
an empirical relationship,37 as a function of the shear strains computed in the FD scheme; in the explicit time-marching
calculation, these ∆u are then used to update the effective stress state at the subsequent time-step. In this empirical model
the increment of permanent volumetric strain Δεvd induced by cyclic loading is expressed as

𝜀
( )
Δ𝜀𝑣𝑑 = 𝛾 ⋅ 𝐶1 exp −𝐶2 ⋅ 𝑣𝑑 , (3)
𝛾

where γ is the amplitude of the cyclic strain and C1 and C2 are two model parameters. Assuming the soil to behave as an
elastic material and the cyclic loading to occur in undrained conditions, the excess pore water pressure is ∆u = K∙∆εvd ,
where K is the bulk modulus of the soil. Finn–Byrne’s model predicts accumulation of ∆u until the effective stress state
reaches Mohr–Coulomb’s failure criterion. The two Finn–Byrne parameters C1 = 0.035 and C2 = 12 were calibrated to
reproduce the excess pore water pressure ratio ∆u/p′0 observed in the RC tests carried out on the undisturbed samples
retrieved from the fine-grained soils of the two dams.25,27 Figure 6 shows a comparison between the measured and the
computed values of ∆u/p′0 , as obtained after 10 and 100 cycles. The relation obtained by Matsuda et al.38 by best fitting
the results of cyclic triaxial tests on reconstitute kaolin is also plotted for comparison in the figure, for 100 cycles. Samples
from the two dams show similar trends, though the one retrieved from the zoned dam is seen to develop slightly larger
excess pore water pressures. The Finn–Byrne model is in fair agreement with the measured values of ∆u/p′0 for 10 cycles,
while overestimates the target response for 100 cycles and γ > 0.1%, with a very sharp increases of ∆u/p′0 , as also obtained
by Matsuda et al.38 In brief, the predictions of pore water pressure build-up obtained using Finn–Byrne’s empirical model
are fairly realistic for well-compacted soils as the ones at hand, which do not show high cyclic pore water pressures leading
to liquefaction conditions.

3.1 Static analysis

For both models, the staged construction of the dam and the subsequent impoundment, both assumed as drained pro-
cesses, were first simulated. To this aim, the operative shear stiffness adopted in the static calculations was calibrated to
reproduce the settlement profiles observed at the end of construction.13,14 Simulation of the construction and impound-
ment stages permits to initialize the state of effective stress, which, in turn, controls the small-strain shear modulus distri-
bution within the embankment and the foundation soil prior to the dynamic calculation phase. During each stage of the
dam construction, soil stiffness was then updated according to Equation (1) to be consistent with the new effective stress
state.
8 MASINI et al

FIGURE 6 Comparison between observed and computed values of the excess pore water pressure ratio

(A)

(B)

FIGURE 7 Contours of pore water pressure u at end of the impoundment stage: (A) Zoned dam; (B) homogeneous dam (values in kPa)

Impounding of the reservoir was simulated applying a hydrostatic pressure to the upstream slope of the dam and to the
bottom of the basin, reaching the maximum storage level of 45 m through three stages of 15 m each. Figure 7 shows the
contour lines of pore water pressure and the end of impoundment. Evaluation of steady-state seepage for the homogeneous
dam (Figure 7B) was carried out enabling negative pore water pressure to develop by gradually reducing the permeability
of the portion of the dam located above the uppermost flow line, on which u = 0. This procedure confines the seepage
flow below the line u = 0, while keeping the soil saturated above it, as observed in the undisturbed samples retrieved from
the fine-grained soil of the dam. A threshold value of −100 kPa was adopted to limit development of negative pore water
pressure above the seepage domain. For the zoned dam (Figure 7A), negative pore water pressures were not allowed to
develop in the flanks due to their high permeability of about 10−6 m/s. Figure 8 shows the contours of the shear strength
ratio t/tmax , where t is the radius of the current Mohr’s circle and tmax is the radius of a Mohr’s circle having the same
center s′, but tangent to the failure envelope; this ratio represents the degree of shear strength mobilization, being t/tmax =
1 when the available shear strength is attained. The pattern of strength mobilization is similar for both dams, with values
of t/tmax > 0.75 mainly computed in the upstream portion of the embankment and in the foundation soil below the center
of the dam. This can be attributed to the effect of the reservoir that reduces the effective stresses and the related available
MASINI et al 9

(A)

(B)

F I G U R E 8 Contours of mobilized shear strength ratio t/tmax at the end of the construction and impoundment phases: (A) Zoned dam;
(B) homogeneous dam

(A)

(B)

FIGURE 9 Pseudostatic analysis (kh = kc ). Contours of shear strain: (A) Zoned dam; (B) homogeneous dam

shear strength. In the zoned dam (Figure 8B) values of t/tmax > 0.90 extend to the surface of the upstream slope, due to
the steeper slope of the shells.

3.2 Iterative pseudostatic analysis

After static initialization of the models, pseudostatic analyses were carried out applying a uniform horizontal inertial
force expressed as a fraction kh of gravity. The seismic coefficient kh was increased progressively until convergence of
the computational algorithm was no longer achieved. In these conditions, the numerical model develops a well-defined
plastic mechanism, in which an unstable soil mass accelerates indefinitely with respect to the stable soil. The value of
kh that activates such mechanism is the critical seismic coefficient kc ; it depends on the strength properties only39 and
represents a measure of the overall seismic resistance of the system.
Calculations were conducted for both directions of the horizontal component of the inertial force kh ∙g, while the ver-
tical component was kept constant. Figure 9 shows the contours of shear strain computed at the critical conditions (kh =
kc ). For both dam schemes, a plastic mechanism is activated from the top to the toe of the embankment, on the upstream
10 MASINI et al

(A)

(B)

FIGURE 10 Contours of G0 : (A) Zoned dam; (B) homogeneous dam (values in MPa)

side, mainly consisting in the rotation of the unstable soil mass. The volume of soil involved in the pseudostatic plastic
mechanism delimited in the figure by the white dotted line is larger in the homogeneous dam, while a shallower slid-
ing surface is obtained for the zoned dam. For both dams, the lower critical seismic coefficient was obtained when the
pseudostatic force is oriented toward the upstream slope, with kc = 0.145 and 0.212 for the zoned and the homogeneous
dam, respectively. Therefore, one may infer that the seismic resistance of the zoned embankment is smaller than that of
the homogeneous dam. However, pseudostatic numerical analyses, in which the seismic action is treated as an equiva-
lent static force, provide only a first insight into the behavior of a dam when subjected to intense seismic loading, as they
are not capable to account for the transient nature of ground motion that induces inertial forces and internal states are
continuously changing in time and space.

3.3 Dynamic analysis

The dynamic analyses here discussed are based on a standard decoupled approach20,23 in which the static equilibrium is
solved first to compute an admissible effective stress field, to then switch to the hysteretic constitutive model for cyclic
loading, characterized by a small-strain stiffness depending on the effective stresses previously computed. Therefore, at
the end of the construction and impounding phases and prior to the dynamic calculations, the initial secant shear modulus
is set equal to the small-strain shear modulus G0 using Equation (1). Figure 10 shows the contour lines of G0 computed
for the two models. High stiffness contrasts are observed at the core-shells contacts for the zoned dam (Figure 10A), with
G0 reaching about 190 MPa in the core and about 1300 MPa in the stiffer right shell. Such large dynamic impedance may
induce significant amplifications of the seismic motion within the dam body. Conversely, a smooth transition is observed
in the homogeneous dam from the surface to the core of the dam with values of G0 slightly larger than 200 MPa at the
base of the embankment (Figure 10B).
In the dynamic calculations, the static boundary conditions were replaced by FLAC “quiet” (viscous) boundary con-
ditions at the base of the grid to simulate a compliant bedrock according to the formulation proposed by Lysmer and
Kuhlemeyer.40 Static fixities were instead replaced by FLAC “free-field” boundaries at the lateral sides, thus reproducing
the free-field condition through a parallel 1D free-field analysis. In addition to the hysteretic damping, a small amount of
Rayleigh damping was also introduced in the analyses to attenuate the soil response at very small strains and to reduce
spurious high-frequency noise. Rayleigh’s coefficients were calibrated to this purpose to obtain a maximum damping of
1% in the range of frequencies defined by the first and the last frequency excited by the input motion.41
Time-domain dynamic analyses were carried out in terms of effective stresses assuming undrained conditions; calcu-
lations were performed adopting a time step of 5 × 10−7 s and a water bulk modulus Kw = 1 GPa, which proved to be
high enough to reproduce undrained conditions while keeping low the noise associated to spurious high frequencies that
would be generated by the calculation algorithm in case of very stiff systems. Hydrodynamic pressures were not accounted
for in the analysis, as they are not relevant for earth dams with mildly steeped slopes.42–44
MASINI et al 11

FIGURE 11 Fourier amplitude spectra, elastic response spectra (5% damping), and acceleration time histories of the input motions

TA B L E 3 Properties of the input seismic motions

Record F amax (g) Ia (m/s) Tm (s) D5-95 (s)
Landers CLW LN 1.5 0.44 2.57 0.42 10.29
Landers CLW LN 1.5 0.25 1.39 0.23 17.12
Kobe TAZ000 0.9 0.63 2.56 0.80 4.6
Kobe TAZUP 0.9 0.38 0.87 0.50 2.2

Four real seismic records were used as input motion, selected among a set of records of real events that are compatible
with the seismicity of the original sites.22,28 The analyses discussed in the following refer to the two input motions that
induce the largest permanent displacements in the two dams. Figure 11 shows the elastic response spectra, the Fourier
spectra, and the acceleration time histories of the two records; these are the Takarazuka record of the 1995 Kobe earthquake
and the Coolwater record of the 1992 Landers earthquake (PEER, http://peer.berkeley.edu/smcat). The two input motions
were scaled by a factor F to obtain about the same Arias Intensity and to match the elastic response spectrum required by
the NTD-20149 ; the target spectrum was computed referring to the ultimate limit state of collapse prescribed for a newly
built large earth dam of relevant importance (return period TR = 2475 years). The same scaling factor was applied to the
horizontal and vertical components of the acceleration records. Table 3 reports some properties of the selected records,
where amax is the peak ground acceleration, IA is the Arias intensity, D5-95 is the significant duration computed between 5%
and 95% of the Arias Intensity, and Tm is the mean quadratic period as defined by Rathje et al.45 The Kobe input has a mean
period Tm close to the fundamental period of the homogenous dam (T0 = 0.74 s), while the Landers input is characterized
by a mean period Tm very close to the fundamental period of the zoned dam (T0 = 0.45 s). The Landers record has also a
significant duration about twice longer than that of the Kobe record. The analyses were carried out assuming a compliant
bedrock, applying the input motions as a time history of the shear stress τxy and the normal stress σyy at the base of the
model.46
12 MASINI et al

FIGURE 12 Sweep of frequency, amplification function (FA)

4 RESULTS AND DISCUSSION

4.1 Dynamic response at low strain level

Besides the dynamic analyses with the real seismic records, a low-intensity (0.001g) frequency sweep, from 0.1 Hz to
28 Hz, was applied to the base of the models as a time history of horizontal and vertical acceleration. The results of
these analyses are representative of the dynamic behavior of the dams at low strains, with a nearly elastic soil behavior.
The smoothed amplification function FA, defined as the ratio between the acceleration Fourier spectrum at the dam
crest and the input spectrum, is plotted in Figure 12 for both components. In the horizontal direction, as expected, the
large difference in stiffness distribution observed in Figure 10 results in sensibly different fundamental frequencies of
the two systems, being f0e,inp-crest = 2.20 Hz and 1.35 Hz for the zoned and the homogeneous embankment, respectively
(T0e,inp-crest = 0.45 s and 0.74 s). Differences are also observed for other fundamental frequencies that are higher for the
zoned dam. The above results depend on the dynamic properties of both the foundation soil and the dam body. The transfer
function between the base and the crest of each dam permits to isolate and analyze the response of the embankment: in this
case, the fundamental frequencies become f0e = 2.9 Hz and 1.5 Hz for the zoned and the homogeneous dams, respectively
(T0e = 0.35 s and 0.67 s).
Consistent with the horizontal component, the amplification functions in the vertical direction also show sensible dif-
ferences, the zoned dam being characterized by a larger focal amplification phenomena that appears to affect, to a larger
extent, the propagation of the vertical component of the input motion. The effect of assuming a bulk modulus of the water
Kw = 1 GPa has been also assessed by running additional analyses with a twice Kw = 2 GPa. Figure 12 shows the ampli-
fication function of the vertical component that could be affected by Kw to a larger extent. The curves obtained for Kw =
2 GPa (gray lines) plot closely around those computed for Kw = 1 GPa (black lines) up to frequencies of 10 Hz, providing
the same first natural frequency, while differences are observed for f > 10 Hz. However, the Fourier amplitude spectra of
Figure 11 show that a small energy content of both components of the input motion is associated to these frequencies that,
therefore, have a negligible influence on the dam response.

4.2 Dynamic response for high-intensity motions

Figure 13A shows the profiles of maximum horizontal and vertical accelerations computed along the dam axis using
the two seismic records. Both components are normalized by the values computed at the model base (ax,br and ay,br ,
respectively). Table 4 reports the values of ax and ay computed at the base of the model and at the base and the crest of
the dam. As expected, a low amplification is observed in the firm soil, with values of a/abr close to unity, mostly due to
the reflected waves propagating downwards, out of the model.
MASINI et al 13

(A) (B)

FIGURE 13 Acceleration profiles along the dam axis and elastic response spectra

TA B L E 4 Maximum accelerations
Bedrock Base Crest
Dam Input ax (g) ay (g) ax (g) ay (g) ax (g) ay (g)
Zoned Landers 0.20 0.13 0.37 0.33 0.77 0.84
Kobe 0.38 0.23 0.53 0.42 0.46 0.98
homogeneous Landers 0.21 0.13 0.41 0.39 0.39 0.58
Kobe 0.36 0.25 0.64 0.30 0.37 0.38

In the zoned dam, large amplification occurs for the Landers input, with the ratio ax /ax,br increasing sharply from
the intermediate third to the crest of the embankment, where a value of ax (0.77g), about four times higher than that
computed at the base of the model (0.20g) is observed (Table 4). Conversely, for the Kobe input, the ratio ax /ax,br decreases
from a value of about 2 at the lower third to about unity at the crest of the dam. These results can be attributed to the
near-resonance vibration conditions occurring for the Landers input, as its mean period Tm = 0.42 s is very close to the
fundamental period T0e,inp-crest = 0.45 s of the zoned dam, while the Kobe input is characterized by a much larger mean
period (Tm = 0.80 s).
For the homogeneous dam, the ratio ax /ax,br is equal to about 50% of that computed in the zoned dam. In this case,
though the mean period of the Kobe record is very close to the fundamental period of the dam, no sensible amplification
occurs, with a value of ax at the dam crest (0.37g) almost equal to that computed at the base of the model (0.36g). It
can be anticipated here that this is the result of transient activations of a plastic mechanism involving a large portion
of the dam, which exhibits a highly nonlinear and dissipative response. Conversely, in the lower third of the dam, the
ratios ax /ax,br computed for the two records are similar. These observations also apply to the vertical component of the
acceleration, though with values of ay /ay,br about twice higher, due to the higher volumetric stiffness of the system in
undrained conditions.
Figure 13B shows the elastic spectra computed at two points located in the upstream slope (point U) and at the upper
third along the dam axis (point A), while Table 5 reports the values of the maximum spectral accelerations Sa,x and Sa,y .
Higher spectral accelerations are computed in the zoned dam, that is more rigid and characterized by a large stiffness
contrast at the shells-core contacts. The maximum horizontal spectral acceleration Sa,x is computed close to the dam crest
(point A) where focal amplification phenomena are more pronounced. In the zoned dam, the Landers input results in
14 MASINI et al

TA B L E 5 Maximum values of the spectral acceleration

Point U Point A
Dam Input Sa,x (g) Sa,y (g) Sa,x (g) Sa,y (g)
Zoned Landers 0.97 0.94 1.59 1.90
Kobe 0.78 1.25 1.20 1.47
Homogeneous Landers 0.73 1.00 0.89 0.974
Kobe 0.63 1.16 0.96 1.22

(A) (B)

FIGURE 14 Time histories of relative displacements at the crest of the dam: (A) horizontal and (B) vertical component

largest amplification moving from point U to point A, with an increase of maximum Sa,x of about 64%, whereas the Kobe
record induces an increase of 54% (Table 5). Conversely, in the homogeneous dam the largest amplification occurs for the
Kobe record, with an increase of maximum Sa,x from point U to point A equal to about 52%, while the Landers record
produce an increase of 22%.
The vertical components of spectral acceleration Sa,y are larger than Sa,x : the Landers record is seen to double Sa,y
moving from point U to point A of the zoned dam; conversely, only slight changes are observed for both records in the
homogenous dam, with an increase of about 6% and a decrease of 3% for the Kobe and Landers record, respectively.

4.3 Deformative patterns

Figure 14 shows the time histories of displacements computed at the crest of the dams for both input motions. The hor-
izontal (∆ux ) and vertical (∆uy ) displacements are relative to the base of the dams; positive values indicate downstream
horizontal displacements and upwards vertical displacements. For the zoned dam, the largest permanent displacements,
∆ux = 115 cm and ∆uy = −121 cm, are computed at the end of the Landers record, characterized by a mean period Tm
closer to the fundamental period of the dam than the ones of the Kobe record that induces ∆ux = 100 cm and ∆uy =
−73 cm. Conversely, for the homogenous dam both components of permanent displacement at the end of the Kobe record,
∆ux = 106 cm and ∆uy = −84 cm, are about twice the values computed for the Landers record, ∆ux = 44 cm and ∆uy =
−36 cm, due to the closeness between periods Tm of the Kobe record and T0 of the dam. The zoned embankment under-
goes slightly larger horizontal displacements (+8.5%) and sensibly higher settlements at the crest (+44%). This may be
ascribed to two concurring factors: the first one is related to the initial stress state, in that a larger portion of the upstream
slope is close to the attainment of the available shear strength at the end of construction in the zoned dam (Figure 8), thus
leading to a lower seismic resistance (kc ); therefore, an intense seismic loading and near-resonance conditions, as it occurs
for the Landers record, are likely to induce larger plastic strains. The second factor is related to the well-known result that
a longer duration of the strong motion phase produces greater permanent displacements: although both embankments
are shaken close to resonance, the homogeneous dam by the Kobe record while the zoned one by the Landers record,
duration of the Landers record (D5−95 = 10.39 s) is more than two times larger than that of the Kobe record (D5−95 =
4.60 s). Then, it can be inferred that displacements of Figure 14 result from the transient activation of plastic mechanisms
during earthquake loading.
MASINI et al 15

(A)

(B)

F I G U R E 1 5 Kobe input motion: contour lines of mobilized shear strength ratio t/tmax at a time instance during the strong motion phase:
(A) zoned dam; (B) homogeneous dam

(A)

(B)

FIGURE 16 Contour lines of shear strain γ/2 (−) at the end of the Landers input motion: (A) Zoned dam; (B) homogeneous dam

In this respect, Figure 15 shows the contours of shear strength ratio t/tmax computed for the Kobe record, at a time instant
of the input motion (t = 6.5 s) when the largest displacement gradients were seen to develop, as shown in Figure 14: the
shear resistance is seen to be fully mobilized (t/tmax ≈ 1) along a wide band developing from the toe of the upstream slope to
the crest of both dams. This stress pattern is consistent with the plastic mechanisms observed in the pseudostatic analyses
(dotted white line in Figure 15), although the plastic volume engaged in the zoned dam is somewhat larger than that
involved in the pseudostatic mechanism. Values of the strength ratio close to unity were also computed at the downstream
toe as a result of a transient plastic mechanism developing also in the downstream portion of the embankment, though
that activating in the upstream slope is still the prevailing one, thus inducing larger displacements.
Figure 16 shows the contours of shear strain (γ/2) computed at the end of the Landers record: for both the dams, an
intensely shared zone (γ/2 > 2%) developed in the upstream slope, consistently with the stress patterns of Figure 15. This
may be ascribed to the lower effective stress state in the upstream slope due to the presence of the reservoir that also
results in higher t/tmax values at the end of the impounding stage, as shown in Figure 8. Larger strains are observed in
16 MASINI et al

(A)

(B)

(C)

(D)

FIGURE 17 Contours of (A and B) horizontal and (C and D) vertical displacements (Kobe record)

the zoned dam, particularly at the crest, where amplification phenomena are more pronounced due to the steeper slope
of embankment shells and the stiffness contrast at the shells-core contacts.
Figure 17 shows the displacement fields computed at the end of the Kobe record for both dams: the overall behavior con-
sists of a lateral spreading with settlements at the crest. Similar trends were obtained for the Landers record. The upstream
slope experiences the largest horizontal displacements, while vertical displacements mainly develop in the upper portion
of the embankment. These permanent displacements are essentially due to plastic strain accumulation throughout seismic
shaking and, again, are consistent with plastic mechanisms provided by the pseudostatic analyses at the critical condition
(kh = kc ) (Figures 17A and 17B). It can thus be inferred that the plastic mechanisms obtained from the iterative pseudostatic
analyses provide significant information on the actual dynamic behavior of earth dams; different earthquakes may induce
different displacements, though activating essentially the same plastic mechanisms. However, this activation only occurs
during seismic shaking as the computed displacements become constant soon after the end of earthquakes (Figure 14),
indicating a stable behavior of the dam once the seismic action is over. The computed seismic performance of the dams
can then be assumed as satisfactory given the very demanding input motions selected for the analyses (TR = 2475 years).
However, some damage can develop in the embankment that may impair its serviceability, possibly requiring some
rehabilitation works. An evaluation of earthquake-induced damage can be obtained by Figure 18, in which the nondi-
mensional crest settlement uy /Hdef is plotted against the input peak ground acceleration (PGA), where Hdef indicates the
dam height plus the alluvium thickness (65 m). The results of this study are compared with field observations collected by
Swaisgood.47 For the highly demanding seismic scenario (high values of PGA) assumed in this study, the two dams experi-
MASINI et al 17

FIGURE 18 Nondimensional crest settlement (uy /Hdef ) as a function of the input peak ground acceleration (PGA)

(A)

(B)

FIGURE 19 Contour lines of the nondimensional net excess pore water pressure (∆u − ∆p)/p′0 at the end of: (A)Landers; (B) Kobe input
motions

enced a level of damage from moderate to serious, the zoned dam suffering a more severe damage than the homogeneous
one.

4.4 Excess pore water pressure build-up

The evaluation of the excess pore water pressure (∆u) developed during undrained earthquake loading was obtained using
the Finn–Byrne model; for the zoned dam, this model was activated in the fine-grained core only.
The development of positive excess pore water pressure induces a reduction of the available shear strength, with a
potential increase in dam displacements. Shear strength reduction is proportional to the reduction of the deviator stress
at failure ∆qf , which can be expressed as

Δ𝑞f = 𝑀 ⋅ (Δ𝑢 − Δ𝑝) , (4)

18 MASINI et al

where M is the slope of the failure envelope in the p′−q plane, with p′ and q equal to the mean effective stress and the
deviator stress, and (∆u − ∆p) is the net pore water pressure build-up during seismic shaking. Figure 19 shows the contours
of the ratio (∆u − ∆p)/p′0 , where p′0 is the initial mean effective stress, computed for the homogeneous dam at the end
of the Landers (Figure 19A) and the Kobe (Figure 19B) seismic inputs. Larger values of the nondimensional ratio (∆u −
∆p)/p′0 are obtained when using the Kobe record as an input motion; this is consistent with the larger displacements
observed in Figure 14, where values of horizontal and vertical components at the end of the Kobe record were about
1.5 times higher than those computed using the Landers record. Higher values of (∆u − ∆p), equal to about 40% of p′0 , are
observed in the lower third of the dam, developing close to the potential sliding surface evaluated from the static pushover
analyses. Conversely, for the zoned dam, not shown here for the sake of space, a larger net excess pore water pressure is
observed for the Kobe record in contrast with the higher displacements computed applying the Landers record. This may
be attributed to the shear strains induced in a larger portion of the core by the Kobe record compared with those induced
by the Landers record, that developed in a narrow band, though being higher in modulus (Figure 16).

5 CONCLUSIONS

It is well known that different water retention schemes of earth dams imply the adoption of different materials for
the dam body together with different slopes of dam flanks, thus leading to different seismic responses: for example,
a greater amplification of ground motion is expected in a zoned dam, due to the steeper slopes of the shells and the
high stiffness contrast at the shells-core contact, with larger effects computed for the vertical component of the seismic
motion.
In this paper, the seismic performance of two idealized earth dams, a homogeneous and a zoned dam characterized by
the same foundation soil, height of embankment and water storage level, was studied through a series of time-domain
nonlinear dynamic analyses, accounting for the influence of soil plasticity activated by transient mobilization of shear
strength during earthquake loading.
The analyses showed that about 75% of the available shear strength is mobilized in the upstream half of both dams at
the end of the construction and reservoir impoundment stages. For the zoned dam, values of t/tmax nearly close to 90%
extend to the surface of the upstream slope: consistently, a larger settlement was computed at the crest of the zoned dam
at the end of seismic motion. It can thus be concluded that the initial degree of strength mobilization affects the plastic
mechanisms activated during earthquake loading, highlighting the need of properly simulating the construction stages of
the dam prior to any strong motion dynamic calculation.
As expected, larger displacements were computed using the input motion characterized by a mean period close to the
first fundamental period of the analyzed systems. For the zoned dam, the final horizontal displacement induced by both
records are about comparable, while very different crest settlements were computed. This highlights the importance of
selecting input motions for the seismic analyses of earth dams not only based on compatibility criteria with a design elastic
response spectrum, but also accounting for frequency coupling and duration characteristics. Specifically, the frequency
content should excite the system fundamental period, leading to near-resonance vibration conditions, while, as impor-
tantly, longer durations of events of high-intensity promote further activation of transient plastic mechanisms resulting
in larger permanent displacements.
Amplification of ground motion should be considered together with possible development of plastic strains: the first
increases as near-resonance vibration conditions are approached, while plastic strains occurring during seismic shaking
provide an important source of energy dissipation. For this reason, the lower values of the crest to bedrock acceleration
ratio, a/abr , computed for the homogenous dam not only depend on the milder slopes of the flanks, but also result from
the transient plastic mechanisms involving a wider portion of the embankment.
It has also been shown that iterative pseudostatic analyses, in which the seismic coefficient is progressively increased
until convergence is no longer attained, are an effective and easy tool to get a first insight into the overall seismic resis-
tance of the dam, quantified by its critical seismic coefficient, and into the plastic mechanisms that may be temporarily
triggered during seismic shaking, as these were seen to be consistent with the ones activated by the dynamic analyses.
Then, the dynamic behavior of the earth dams, resulting from the time-domain dynamic analyses, can be profitably inter-
preted on the basis of the plastic mechanisms computed by the iterative pseudostatic analyses at critical conditions (kh =
kc ); different earthquakes may induce different displacements, which, however, derive essentially from the same plastic
mechanisms. For both the dams, the critical mechanism develops in the upstream slope as the reservoir reduces the effec-
tive stress and the related available shear strength. The zoned dam is characterized by a shallower plastic volume and a
MASINI et al 19

lower critical seismic coefficient that, according to the result of the dynamic calculations, makes it prone to experience
larger permanent displacements during seismic loading. The plastic mechanisms obtained by the iterative pseudostatic
analyses can then be used to first estimate the portions of the dam body that may experience large deformation during
earthquake loading.

AC K N OW L E D G M E N T S
The research work presented in this paper was partly funded by the research project PRIN 2017 REDREEF (Risk Assess-
ment of Earth Dams and River Embankments to Earthquakes and Floods) supported by MIUR (Ministero dell’Istruzione,
dell’Università e della Ricerca).

ORCID
Luca Masini https://orcid.org/0000-0002-2970-1009

REFERENCES
1. Foster M, Spannagle M, Fell R. Report on the Analysis of Embankment Dam Incidents. UNICIV Report No. R-374. School of Civil and Envi-
ronmental Engineering, Sydney, Australia: The University of New South Walws, 1998. ISBN 85841 349 3, ISSN 0077-880X
2. Foster M, Fell R. Use of event trees to estimate the probability of failure of embankment dams by internal erosion and piping. In: Proceedings
of the 20th Congress on Large Dams, vol. 1, Question 76, 237-260, Paris: International Commission on Large Dams; 2000.
3. Foster M, Fell R, Spannagle M. A method for estimating the relatively likelihood of failure of embankment dams by internal erosion and
piping. Can Geotech. J. 2002; 39(2): 497-500
4. Fell R, MacGeregor P, Stapledon D, Bell G. Geotechnical Engineering of dams. Boca Raton, FL:CRC Press; 2005, ISBN 04 1536 440 x
5. Lanzo G. Alcune considerazioni sulla rivalutazione della sicurezza sismica delle dighe di materiali sciolti. In Proceedings of Conferenze di
Geotecnica di Torino - XXV Ciclo, Analisi e Progetto delle Opere Geotecniche in Zona Sismica. Politecnico di Torino (In Italian); 2018.
6. USSD. Observed performance of dams during earthquakes. Volume III, USSD Committee on Earthquakes, Denver, CO: United States Society
on Dams; 2014.
7. Seed HB, Lee KL, Idriss IM, Makdisi FI. The slides in the San Fernando dams during the earthquake of February 9, (1971) J Geotech Eng.
1975;101(GT7):651–668.
8. Pradel D, Wartman J, Tiwari B. Failure of the Fujinuma dams during the March 11, 2011 Tohoku Earthquake. Geotech Sp. 2013;1:1559-1573.
9. Ministero Infrastrutture e Trasporti. D.M. 26/06/2014 - Norme Tecniche per la Progettazione e la Costruzione degli Sbarramenti di Ritenuta
(dighe e traverse). Gazzetta Ufficiale Serie Gen, 2014, n. 156 del 08/07/2014.
10. Vrymoed JM. Dynamic FEM model of Oroville dam. J Geotech Eng. 1981;107(GT8):1057–1077.
11. Prevost JH, Abdel-Ghaffar AM, Lacy SJ. Nonlinear dynamic analysis of an earth dam. J Geotech Eng. 1985;111(7):882-897.
12. Elia G, Amorosi A, Chan AHC, Kavvadas MJ. Fully coupled dynamic analysis of an earth dam. Géotechnique, 2010;61(7): 549-563.
13. Rampello S, Cascone E, Grosso N. Evaluation of the seismic response of a homogeneous earth dam. Soil Dyn Earthquake Eng 2009;29(5):782-
798.
14. Elia G, Chan AHC, Kavvadas MJ, Amorosi A. Fully coupled dynamic analysis of an earth dam. Geotechnique 2011;61(7):549-563.
15. Elia G, Rouainia M. Seismic performance of earth embankment using simple and advanced numerical approaches. J Geotech Geoenviron.
2013;139(7):1115-1129.
16. Sica S, Pagano L, Modaressi A. Influence of past loading history on the seismic response of earth dams. Comput Geotech. 2008;35(1):61–85.
17. Albano M, Modoni G, Croce P, Russo G. Assessment of the seismic performance of a bituminous faced rockfill dam. Soil Dyn Earthquake
Eng. 2015;75:183-198.
18. Pelecanos L, Kontoe S, Zdravkovic L. A case study on the seismic performance of earth dams. Géotechnique. 2015;65:923-935.
19. Han B, Zdravkovic L, Kontoe S, Taborda DMG. Numerical investigation of the response of the Yele rockfill dam during the 2008 Wenchuan
earthquake. Soil Dyn Earthquake Eng. 2016;88:124-142.
20. Andrianopoulos CI, Papadimitriou AG, Bouckovalas GD, Karamitros DK. Insight into the seismic response of earth dams with an emphasis
on seismic coefficient estimation. Comput Geotech. 2014;55:195-210.
21. Masini L, Rampello S, Callisto L. Seismic behavior of large earth dams: From site investigations to numerical modelling. Paper presented
at: Proceedings of 1st IMEKO TC-4 Int. Workshop on Metrology for Geotechnics, March 17, 2016; Benevento, Italy; 288-294.
22. Masini L, Rampello S. Performance of large homogeneous earth dams during strong ground motions. Paper presented at: 19th International
Conference on Soil Mechanics and Geotechnical Engineering, Unearth the Future, Connect Beyond, September 17, 2017; Seoul, Korea.
23. Papadimitriou AG, Bouckovalas GD, Andrianopoulos KI. Methodology for estimating seismic coefficients for performance-based design
of earthdams and tall embankments. Soil Dyn Earthquake Eng. 2014;56:57-73.
24. Elia G, Amorosi A, Chan AHC, Kavvadas MJ. Numerical prediction of the dynamic behavior of two earth dams in Italy using a fully
coupled nonlinear approach. Int J Geomech. 2011;11:504-518.
25. Calabresi G, Rampello S, Callisto L, Cascone E. Diga S. Pietro sul fiume Osento. In: Ve-rifica delle condizioni di stabilità e analisi del
comportamento in condizioni sismiche. Research Rep., Univ. di Roma La Sapienza, Rome. 2004; (In Italian)
20 MASINI et al

26. Rampello S, Callisto L, Fargnoli P, Amorosi A, Elia G. Diga S. Pietro sul fiume Osento. In: Analisi dinamica del comportamento della diga
in presenza di sisma. Research Rep., Univ. di Roma La Sapienza, Rome. 2005.
27. Calabresi G, Rampello S, Sciotti A, Amorosi A. Diga sulla Marana Capacciotti: Verifica delle condizioni di stabilità e analisi del comporta-
mento in condizioni sismiche. In: Res. Rep.: Università degli Studi di Roma La Sapienza, Dip. di Ingegneria Strutturale e Geotecnica. Bastogi
Editrice, Foggia. (In Italian) 2000.
28. Cascone E, Rampello S. Decoupled seismic analysis of an earth dam. Soil Dyn Earthquake Eng. 2003;23(5):349-365.
29. Dakoulas P, Gazetas G. Vibration characteristics of dams in narrow canyons, J Geotech Eng. 1987;113(8), 899-904.
30. Mejia LH, Seed HB. Comparison of 2D and 3D dynamic analyses of earth dams. J Geotech Eng. 1983;109(11), 1383-1398.
31. Ohmachi T, Tokimatsu K. Simplified method for 3-dimensional dynamic analysis of embankment dams. Paper presented at: Fourth Inter-
national Conference on Numerical Methods in Geomechanics, March 31, 1982; Edmonton, Canada; 1:411-419
32. Itasca. FLAC Fast Lagrangian Analysis of Continua v. 7.0. User’s Manual. Minneapolis, MN: Itasca Consulting Group, 2011.
33. Viggiani G, Atkinson JH. Stiffness of fine-grained soil at very small strains. Géotechnique 1995;45(2):249–265
34. Kulhawy FH, Mayne PW. Manual on Estimating Soil Properties for Foundation Design, Final Report, Project 1493-6, EL-6800, Palo Alto, CA:
Electric Power Research Institute; 1990.
35. Skempton AW. Standard penetration test procedures and the effects in sands of overburden pressure, relative density, particle size, ageing
and overconsolidation. Géotechnique 1986;36(3):425-447
36. Rollins KM, Evans MD, Diehl NB, Daily III, WD. Shear modulus and damping relationships for gravels. J Geotech Geoenviron. 1998;124(5):
396-405.
37. Byrne P. A cyclic shear-volume coupling and pore-pressure model for sand. Paper presented at: 2nd International Conference on Recent
Advances in Geotechnical Earthquake Engineering and Soil Dynamics, March 11–15, 1991; St. Louis, Missouri; 47-55.
38. Matsuda H, Nhan TT, Ishikura R. Prediction of excess pore water pressure and post-cyclic settlement on soft clay induced by uni-directional
and multi-directional cyclic shears as a function of strain path parameters. Soil Dyn Earthquake Eng. 2013;49:75-88.
39. Masini L, Callisto L, Rampello S. An interpretation of the seismic behavior of reinforced-earth retaining structures. Géotechnique
2015;65(6):349-358.
40. Lysmer J, Kuhlemeyer RL. Finite dynamic model for infinite media. J Eng Mech. 1969;95(4):859-877.
41. Amorosi A, Boldini D, Elia G. Parametric study on seismic ground response by finite element modelling. Comput Geotech. 2010;37(4):515-
528.
42. Pelecanos L, Kontoe S, Zdravkovic L. Numerical modelling of hydrodynamic pressures on dams. Comput. Geotech. 2013;53:68–82.
43. Pelecanos L, Kontoe S, Zdravkovic L. Dam–reservoir interaction effects on the elastic dynamic response of concrete and earth dams, Soil
Dyn Earthquake Eng. 2016;82:138-141.
44. Pelecanos L, Kontoe S, Zdravkovic L. The effects of dam–reservoir interaction on the nonlinear seismic response of earth dams. J Earth-
quake Eng. 2018. https://org/10.1080/13632469.2018.1453409
45. Rathje EM, Abrahamson NA, Bray JD. Simplified frequency content estimates of earthquake ground motions. J Geotech Geoenviron.
1998;124 (2), 150-159.
46. Joyner WB, Chen ATF. Calculation of nonlinear ground response in earthquakes. Bull Seismol Soc Am. 1975;65(5):1315-1336.
47. Swaisgood JR. Embankment dam deformations caused by earthquakes. Paper presented at: 7th Pacific Conference on Earthquake Engi-
neering; February 13–15, 2003; Christchurch, New Zealand.

How to cite this article: Masini L, Rampello S, Donatelli R. Seismic Performance of Two Classes of Earth Dams.
Earthquake Engng Struct Dyn. 2020;1–20. https://doi.org/10.1002/eqe.3352