ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Evaluation of the seismic response of a homogeneous earth dam

S. Rampello a, E. Cascone b,, N. Grosso b,1
a
Dipartimento di Ingegneria Strutturale e Geotecnica, Università di Roma La Sapienza, Italy
b
Dipartimento di Ingegneria Civile, Facoltà di Ingegneria, Contrada di Dio, Università di Messina, 98166 Messina, Italy

a r t i c l e in f o a b s t r a c t

Article history: The response of an earth dam to seismic loading is studied through displacement-based analyses and
Received 8 April 2008 finite element, effective stress dynamic analyses. Displacement-based analyses are carried out using
Received in revised form both empirical relationships and the decoupled approach in which the deformable response of the soil is
30 July 2008
accounted for through ground response analyses, and the resulting accelerograms are used in the sliding
Accepted 19 August 2008
block analysis. The FE analyses are carried out using a constitutive model capable to reproduce soil non-
linearity, calibrated against laboratory measurements of the stiffness at small strains. The influence of
Keywords: the assumed input motion and bedrock depth on the seismic response of the dam is also studied.
Earth dam The FE analyses permitted evaluation of the overall patterns of the dam behaviour in terms of
Seismic response
displacements and of stress, strain and acceleration fields. For any of the seismic scenarios considered in
Effective stress analysis
the analyses the performance of the dam was satisfactory in that maximum computed settlements at
Displacement methods
Decoupled approach the crest were considerably smaller than the service freeboard of the dam and limited shear strains
were obtained within the dam during earthquake loading. Comparison of FE dynamic analyses with
displacement-based analyses allowed predictions from different approaches to be compared. A fair
agreement was obtained provided that ground response analysis was accounted for when using the
displacement-based approach.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction in which the soil is treated as a two-phase material with fully

coupled soil skeleton response and excess pore water pressure
Evaluation of stability conditions and crest displacements of development [1,2]. Inelastic soil behaviour under cyclic loading
earth dams is an issue of major concern in seismic regions. can be properly described through elastic–plastic constitutive
Seismic-induced deformations in an earth dam may lead to models developed within the framework of bounding surface
overtopping and consequently to severe losses in terms of plasticity, kinematic hardening plasticity or generalised plasticity.
property and human lives. Alternatively, non-linear soil behaviour can be modelled using
In recent years, a revision of seismic hazard maps and design given stress–strain relationships to represent the cyclic behaviour
criteria has led to the necessity of assessing the seismic safety of in simple shearing and empirical relationships to evaluate pore
earth dams previously built without consideration of seismic- water pressure build-up under earthquake loading [3]. Advanced
induced forces. As a consequence, design of proper remedial soil models can provide accurate predictions for static and
measures has become a challenging problem for geotechnical dynamic loading conditions and have been used successfully in
engineers. analyses of earth dams [4,5]. However, they often require
The study of the seismic response of earth dams is a complex parameters not usually measured in field or laboratory testing.
problem that generally requires the use of dynamic methods of Moreover, uncertainties are often related to seismic input when
analysis. These vary from highly simplified to rigorous methods, analysing real boundary value problems.
and consequently require different levels of sophistication in Due to the uncertainties mentioned above, the response of
terms of proper problem formulation, characterisation of material earth dams to earthquake loading is often evaluated through
properties and modelling of stress–strain soil behaviour. displacement-based approaches, decoupling the ground response
Nowadays coupled effective stress dynamic analyses can be analysis of the earth structure from the sliding block analysis
carried out using advanced constitutive models for soil behaviour [6–10].
In this paper, the response of a homogeneous earth dam under
seismic loading was studied through both displacement-based
 Corresponding author. Tel.: +39 090 397 7162; fax: +39 090 397 7480. analyses and FE effective stress dynamic analyses. The supple-
E-mail address: [email protected] (E. Cascone). mentary geotechnical investigation carried out at the site
1
Formerly post-doc student. provided a reliable set of stiffness properties, allowing the shear

0267-7261/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2008.08.006
 ARTICLE IN PRESS

S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798 783

stiffness at small strains to be evaluated for the earth dam and the tube samples for reliable evaluation of stiffness and strength
foundation soil. The effect of non-linear soil behaviour was then properties of dam and foundation soils.
accounted for in both the displacement-based decoupled ap- The foundation soil includes two layers: an alluvial soil
proach and in the FE analyses from the early beginning of deposit, about 12 m thick, made of medium-stiff silt and clay
earthquake loading. (IL ¼ 0.3) of low plasticity (IP ¼ 13.5%) with thin layers of sand and
Displacement-based analyses were carried out at different gravel and a stiff silty clay deposit (IL ¼ 0.1) of low plasticity
degrees of complexity. First, empirical relationships proposed in (IP ¼ 16%), which extends down to the deep bedrock formation.
the literature [11–13] were applied. Then, sliding block analyses The dam embankment is mainly formed by sandy silt and clay of
were carried out assuming a rigid-perfectly plastic soil behaviour. low plasticity (IP ¼ 18.7%). A detailed description of the geotech-
Finally the decoupled approach was applied carrying out 1D and nical characterisation of the site may be found in Calabresi et al.
2D equivalent linear ground response analyses followed by sliding [14] and Cascone and Rampello [10].
block analyses. Both artificial and real accelerograms were considered in the
The FE analyses were carried out using an elastic-plastic soil seismic analyses of the dam. A set of 10 artificial accelerograms
model with isotropic hardening, capable to reproduce some were generated to match the response spectrum provided by the
features of cyclic soil behaviour. Model parameters were cali- Eurocode no. 8 (EC8) for soil type A, that is for rock or rock-like
brated against consolidated undrained triaxial compression tests formations with equivalent shear wave velocity, in the first 30 m,
and resonant column tests. VS30X800 m/s, and for earthquake magnitude M45.5 [15]. The
Both real and artificial accelerograms were used as input accelerograms are characterised by a wide frequency content,
motions and four different depths of the bedrock zB were assumed peak ground acceleration amax of about 0.35g, duration of the
in the range of 25–300 m. This permitted to evaluate the influence event TD ¼ 30 s and values of the Arias intensity ranging in the
of factors such as bedrock depth, input motion and parameter interval IA ¼ 3.5–4.7 m/s. The value of amax ¼ 0.35g is consistent
calibration on the response of the dam to earthquake loading, with the results of a seismic hazard study that, for a return period
gaining some insight in its behaviour under seismic conditions. of 475 years, predicts at the site peak ground accelerations of
0.28–0.32g [16]. The real accelerograms selected for the analyses
are characterised by values of amax in the range of 0.3–0.5g, and by
plateau of the normalised response spectrum approximately
2. Site description and input motion similar to that given by EC8 for soil type A and M45.5. In the
analyses, the peak ground acceleration of each of them was scaled
The dam is located in Southern Italy and dikes the course of the to the same value of amax ¼ 0.35g adopted for the artificial
Marana Capacciotti stream. Fig. 1 shows a plan view of the dam accelerograms.
and its main cross section; Table 1 summarises the main These two sets of accelerograms were initially used to perform
geometrical features of the dam. The drainage system consists a decoupled seismic analysis of the earth dam, in which the
of a sub-vertical central drain discharging into a tunnel parallel to
the longitudinal axis of the dam and of a drain located at the toe of
the downstream slope. An impervious diaphragm extending into Table 1
Dam geometrical features
the lower clay deposit prevents seepage through the alluvial soils
underlying the dam. The figure shows the location of four Volume 3.71 M m3
extensometers installed at the crest and the downstream slope Water storage 49 M m3
to monitor the settlement of the dam during the construction and Crest length 837 m
after its completion for a period of about 10 years. Supplementary Crest width 7.5 m
Height 48 m
in situ investigation, recently carried out to assess the seismic
Freeboard 2.6 m
performance of the dam, involved three boreholes (BH), four cone Slopes 1 1
225
penetration tests (CPT) and laboratory tests on 21 undisturbed

Fig. 1. Plan view and main cross section of Marana Capacciotti dam (adapted from [14]).
 ARTICLE IN PRESS

784 S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

dynamic response analysis of the earth structure was performed

0.4 separately from the sliding block analysis. The ground response
Artificial No.6 analyses were carried out under both 1D and 2D conditions,
incorporating the non-linear soil behaviour through the equiva-
0.2
lent linear method, and assuming four different depths for the
bedrock. The accelerograms were assumed to be applied at the
a(t)/g

0 rock outcrop; analysis input motions were then obtained by 1D

deconvolution of the outcropping accelerograms to each assumed
bedrock depth. Sliding block analyses were then carried out to
-0.2
evaluate the dam displacements using the accelerograms ob-
tained at the centre of gravity of the critical failure surfaces. Using
-0.4 this approach, the seismic performance of the dam was evaluated
0 10 20 30 for a range of earthquake loadings, and the most critical seismic
t (s) scenarios were detected [10].
In the FE effective stress seismic analyses described in the
present study only the artificial accelerogram no. 6 and the
0.4 accelerogram recorded at Tolmezzo during the 1976 Friuli earth-
Tolmezzo
quake were used. These accelerograms were in fact among those
0.2 that yielded the largest displacements in the decoupled seismic
analysis of the dam [10].
a(t)/g

The artificial accelerogram no. 6, shown in Fig. 2a, is

0
characterised by a significant duration of the strong motion
phase; in fact, the bracketed duration, defined as the time
-0.2 between the first and last exceedances of a threshold acceleration,
set equal to 0.05g, is TDffi27 s. It has a large bandwidth (Fig. 2c),
-0.4 and an Arias intensity IA ¼ 3.96 m/s. Fig. 2b shows the Tolmezzo
0 10 20 30 accelerogram after scaling the peak ground acceleration
(amax ¼ 0.37g) to the same value of amax ¼ 0.35g adopted for the
t (s)
artificial accelerogram. The bracketed duration of the strong
motion phase is of 6.2 s and the Arias intensity is IA ¼ 1.42 m/s.
0.10 The Fourier amplitude spectra and the normalised response
Artificial No.6 spectra of the two accelerograms are compared in Fig. 2c–d. The
0.08 Tolmezzo spectrum of the artificial motion has the typical shape of a wide-
band motion, whereas the spectrum of the scaled real motion is
FA (m/s)

0.06 characterised by a well defined predominant period TP ¼ 0.68 s.

Fig. 2d shows that the spectral acceleration amplitudes of the
0.04 artificial and the real accelerograms are comparable in the range
of periods corresponding to the plateau of the EC8 spectrum for
0.02 soil type A. However, values of spectral acceleration for the real
motion are higher than those given by the EC8 spectrum for
0.00
periods in the range of 0.5–1.0 s; the opposite is observed for
0 1 2 3
T41.5 s.
t (s) Geotechnical investigation did not encounter a soil layer stiff
enough to be regarded as a bedrock (Vs4800 m/s). A geological
4 study of the site provides an estimate of the bedrock depth of
about 300 m. Therefore, a parametric study was performed, in
EC8
which the influence of the bedrock depth on the dam response
Artifical No.6 was evaluated. For this purpose, four different bedrock depths zB
3
Tolmezzo were selected: zB ¼ 25, 50, 100 and 300 m.
Sa/amax

2 3. Displacement-based methods

3.1. Simplified displacement methods

1
The simplest approach for evaluating the seismic performance
of earth structures is represented by the displacement-based
0 methods derived from the sliding block analysis [17]. According to
0 1 2 3 this method the earthquake-induced displacement of the poten-
t (s) tial sliding mass is evaluated following a two-step procedure:
first, the critical acceleration and the associated failure mechan-
Fig. 2. Input accelerograms: (a) artificial accelerogram no. 6; (b) scaled Tolmezzo ism are determined via the pseudo-static approach; then, the
accelerogram; (c) Fourier amplitude spectra and (d) normalised response cumulative displacement of the potential sliding mass is com-
spectra.
puted by double integration of the acceleration of the sliding soil
mass relative to firm ground. This approach is the theoretical basis
for a number of simplified methods proposed for evaluating the
 ARTICLE IN PRESS

S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798 785

Table 2
Empirical relationships used for simplified displacement evaluation

Reference Average displacement Earthqs. no. M No. of accels. Notes

[11] 37v2max 14 6.3–6.7 28 amax40.15g

u¼ e9:4ðac =amax Þ
amax
" 2:53  1:09 #
[12] ac ac 11 6.6–7.2 50 u (cm)
log u ¼ 0:90 þ log 1
amax amax

[13] ac 7 5.8–7.5 11 amax40.13g

log u ¼ 1:460 log IA  6:642 þ 1:546
g IA ¼ 0.2–9.96 m/s, u (cm)

artificial acc.
artificial acc.

10 real acc.
10 real acc.

1
1

s (cm)
u (cm)

0.1
0.1 kc 1D Response
0.15 0.20 & Newmark
Whitman & Liao
Ambraseys & Menu range of numerical results
Jibson
range of numerical results 0.01
0.01 100
0 0.2 0.4 0.6 0.8 1
ac /amax

Fig. 3. Displacements computed using simplified methods. artificial acc.

artificial acc.
10 real acc.

10 real acc.
s (cm)

2D Response
& Newmark
1

1
s (cm)

Kc zB 100 50 25 12
0.153
0.158
Sliding block
0.1
displacement analysis
0 0.2 0.4 0.6 0.8 1
0.1 kc = 0.153
ac /amax
kc = 0.158
Fig. 5. Displacements at the dam crest computed via decoupled approaches:
kc = 0.234
(a) 1D and (b) 2D ground response analyses.
kc = 0.285
range of numerical results

0.01 earthquake-induced displacements of slopes, derived by fitting

0 0.2 0.4 0.6 0.8 1 displacements computed for different values of the critical
ac /amax acceleration and for different sets of earthquake records. These
simplified methods predict permanent displacement u as a
Fig. 4. Displacement at the dam crest evaluated using the sliding block method. function of the critical acceleration ac and some parameters
 ARTICLE IN PRESS

786 S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

related to seismic shaking, such as peak acceleration amax, peak ac ¼ 0.15g the displacement is about 25 cm, for ac ¼ 0.2g it is
velocity vmax or Arias intensity IA. A comprehensive review of the reduced to about 12 cm.
available simplified displacement methods may be found in Cai In addition to the simplified methods, the sliding block
and Bathurst [18]. analysis was carried out considering rotational collapse mechan-
The methods used in this study to compute earth dam isms and assuming the potential sliding mass to behave as a rigid
displacements were those proposed by Whitman and Liao [11], plastic body. The crest settlement S, computed for the four values
Ambraseys and Menu [12] and Jibson [13]. Table 2 reports the of kc ¼ ac/g labelled in Fig. 4, is similar, in both trend and
empirical relationships providing average values of displacement magnitude, to that evaluated using the Whitman and Liao [11] and
together with some details concerning the seismic motions Ambraseys and Menu [12] relationships. In the figure the
considered in developing the equations. Only the artificial regression line and the 90% non-exceedance confidence levels
accelerograms were used in the computations. are also shown.
Two values of the critical acceleration were considered:
ac ¼ 0.15 and 0.20g that are close to the values of ac obtained
from the pseudo-static analyses [14]. The obtained permanent 3.2. Decoupled approach
displacements are plotted in Fig. 3 against the ratio ac/amax
together with the regression lines and the 90% non-exceedance Evaluation of earthquake-induced displacements can be also
confidence levels. The methods proposed by Whitman and Liao carried out following the decoupled approach, based on the
[11] and Ambraseys and Menu [12] provide similar results with assumption that the sliding block analysis can be decoupled from
maximum displacements less than about 10 cm. Larger displace- the ground response analysis of the earth structure. This approach
ments were obtained using the relationship proposed by Jibson represents a useful approximation for most earth structures and
[13]. In this method displacements are related to both Arias generally provides a conservative estimate of the seismic-induced
intensity and critical acceleration. Since artificial accelerograms permanent displacements [19,20].
are characterised by similar values of IA ¼ 3.52–4.67 m/s the Cascone and Rampello [10] carried out 1D and 2D ground
results are mainly affected by changes in the ratio ac/amax. For response analyses of the dam assuming four values for the
bedrock depth (zB ¼ 12, 25, 50 and 100 m). Ten artificial accel-
erograms were used for the 1D response analysis, and five
q strength envelope
artificial accelerograms were used for the 2D response analysis.
fs yield surface At the dam centre line, results from 1D and 2D analyses were in a
(deviatoric hardening)
fair agreement, with the exception of the top third of the dam
where the 2D analyses predicted higher values of amax. This was
αp'c ascribed to both the 2D geometry of the problem and the wave
fv yield surface
focusing effects that increase as the top of the dam is approached.
(volumetric hardening) Fig. 5a–b shows the displacements computed using the sliding
block method for two values of the critical seismic coefficient,
kc ¼ 0.153 and 0.158 [14], using the accelerograms provided by 1D
and 2D ground response analyses at the centre of the sliding mass.
The displacements obtained after 2D response analysis are greater
than those evaluated after 1D analysis, attaining maximum values
of about 40 cm. This is consistent with the larger amplification
p'c p' associated to the 2D geometry of the problem. The larger scatter
observed after the 1D response analysis can be attributed to the
Fig. 6. Yield surfaces of the hardening soil model and their evolution. larger set of input motions used in computations.

193.2 m a.s.l.

172.8
P 163.6
M

300 200
q (kPa)

q (kPa)

250 150

M P

200 100
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
εs (%) εs (%)

Fig. 7. Stress–strain response of selected soil elements of the dam during earthquake—real input accelerogram.
 ARTICLE IN PRESS

S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798 787

Among the simplified approaches, the method proposed by mechanical behaviour of the soil was described using the
Jibson [13] was in a fair agreement with the decoupled approach constitutive model Hardening Soil (HS) [22,23]. This is an
in which the seismic ground response analyses were carried out elastic–plastic rate independent model with isotropic hardening
under 2D conditions. Both methods yield permanent displace- in which the elastic behaviour is defined by isotropic elasticity
ments similar to those computed with effective stress FE analyses, through a stress-dependent Young’s modulus:
described in the next sections, which are shown for comparison in
 n
Figs. 3–5. 0 c0 cot j0 þ s03
E0 ¼ E ref (1)
c0 cot j0 þ pref

where s03 is the minimum principal effective stress, c0 is the

4. Coupled analysis of the earth dam cohesion, j0 is the angle of shearing resistance, pref ¼ 100 kPa is a
0
reference pressure, and E ref and n are model parameters.
4.1. Soil model The model has two yield surfaces fs and fv with independent
isotropic hardening depending on deviatoric plastic strains gp and
An additional analysis of the seismic behaviour of the earth on volumetric plastic strains epv , respectively (Fig. 6). The
dam was performed using the finite element code Plaxis [21]. The deviatoric hardening rule is related to parameter E050, while the
volumetric one is controlled by parameter E0oed. Both of them are
Table 3 given by expressions similar to Eq. (1) but, in contrast to E0 , they
Stiffness coefficients are not used within a concept of elasticity.
The flow rule is associated for states lying on the surface
OCR R m n S
fv, while a non-associated flow rule is used for states on the
Earth dam 1.0 1.0 – 0.75 1573 surface fs.
Alluvial silt 1.0 1.0 – 0.73 2155 A feature of HS model is that under monotonic loading it can
Silty clay 1.5 1.3 0.19 0.73 2000
account for non-linear stress–strain behaviour, and for the
occurrence of irreversible strains from the beginning of the
loading process; this implies that for undrained conditions an
Table 4 increase of deviatoric stress q produces excess pore water
Mechanical properties for HS model pressure. Under cyclic loading conditions the model predicts
c0 (kPa) j0 (deg)
0
E ref (kPa)
0 0
E ref/E50
ref
TX-CIU tests RC tests
plastic strains and excess pore water pressures only for cycles
0
E50ref 0 ref
/Eoed
0
E50ref 0 ref
/Eoed with increasing strain amplitude, that produce a progressive
isotropic enlargement of the yield surfaces fs or fv. Conversely,
Earth dam 30 30 119,400 10.9 1.25 2.43 for cycles of constant or decreasing amplitude, the model res-
Alluvial silt 7 33 149,200 18.3 1.16 2.49
ponse becomes mainly linearly elastic with negligible hysteretic
Silty clay 60 25 145,500 10.4 0.78 2.24
damping.

1000 p'0=343 kPa 1000 p'0=781 kPa

p'0=245 kPa p'0=380 kPa
p'0=147 kPa p'0=196 kPa
800 model 800 model

600 600
q (kPa)

q (kPa)

400 400

200 200
earth dam silty clay
z = 20.3 m
0 0

800 800

600 600
u (kPa)

u (kPa)

400 400

200 200
0 3 6 9 12 15 18 0 3 6 9 12 15 18
εa (%) εa (%)

Fig. 8. Comparison between stress–strain response from TX-CIU tests and model simulations.
 ARTICLE IN PRESS

788 S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

These features are shown in Fig. 7, in which the stress–strain values of g larger than about 0.2% in the earth dam and about 0.1%
response to earthquake loading is plotted for two soil elements of in the foundation soil.
the earth dam. The curves show non-linear soil behaviour for It is worth mentioning that the triaxial-based soil parameters
increasing values of the deviator stress q, starting from small provide a steeper decay of soil stiffness with shear strain than that
deviatoric strains es, while the response is stiffer and nearly linear obtained from RC tests. This may result from having calibrated the
0 0
ref ref
during the unloading–reloading cycles. The maximum value of q is model parameters Eoed and E50 on triaxial test results based on
attained at 6.5 and 7.1 s for soil elements M and P, respectively, external measurements of axial strain that are known to under-
that is towards the end of the strong motion phase of the input estimate the shear stiffness at small strains; as a consequence a
accelerogram. steep decrease is obtained when relating the small-strain shear
modulus G0 to values of G derived from TX-CIU tests. FE analyses
were carried out using the TX-based parameter calibration that
4.2. Parameter calibration should yield larger dam displacements. As discussed in the
following sections a few analyses were repeated using RC-based
Since E0 represents the tangent initial Young’s modulus of the model parameters to evaluate the influence of different model
stress–strain curve, it has been related to the shear modulus at
small strains G0, as obtained from RC tests carried out on
undisturbed samples retrieved at the site. Specifically, to evaluate
0
the Young’s modulus E ref at the reference pressure pref ¼ 100 kPa 1.2 0.4
the relationship proposed by Viggiani [24] was used: earth dam
 0 n 1
G0 p 0.3
¼S Rm (2)
pr pr 0.8

Δu/p'0
where p0 is the mean effective stress, pr ¼ 1 kPa is a reference
G/G0 0.6 0.2
stress, R is the overconsolidation ratio defined as the ratio
between the yield and the current mean effective stresses p0y =p0 ,
0.4
and S, n and m are non-dimensional material constants. The 0.1
stiffness exponents were estimated using the empirical relation- 0.2
model
RC data
ships relating n and m to the plasticity index IP [25], while the
stiffness multiplier S was computed by introducing in Eq. (2) 0 0
the experimental values of p0 , applied in the RC tests, and the 0.0001 0.001 0.01 0.1 1
corresponding measured values of G0. Once the stiffness coeffi- γ (%)
0
cients were known (Table 3), E ref was determined assuming
n0 ¼ 0.2 (Table 4).
0 0
ref ref
The remaining model parameters E50 and Eoed were calibrated 1.2 0.4
according to the results of both consolidated undrained triaxial alluvial silt
compression tests (TX-CIU) and consolidated undrained RC tests 1
carried out on undisturbed samples retrieved with standard thin- 0.3
walled open tube samplers. 0.8
Fig. 8 shows the comparison between model simulations and
G/G0

Δu/p'0
test results for triaxial tests carried out on samples retrieved from 0.6 0.2
the earth dam and the foundation soils. It was found that a good
agreement for the stress–strain curves and the generated excess 0.4
0 0 0.1
pore water pressures can be obtained using the ratios of E ref/E50 ref
model
0
ref 0 ref 0.2
and E50 /Eoed listed in Table 4 and a value for the angle of dilatancy RC data
at failure c ¼ 0. Table 4 also lists the values of effective cohesion c0
0 0
and angle of shearing resistance j0 adopted for calibrating the test
0.0001 0.001 0.01 0.1 1
results; these are the same values used in the displacement-based
γ (%)
approach. A fair agreement is obtained for the pre-failure
stress–strain and excess pore water pressure curves for the range
of deviatoric strain involved during the earthquake loading (p1%). 1.2 0.4
In Fig. 9 the experimental values of non-dimensional shear silty clay
modulus G/G0 and excess pore water pressure ratio Du=p00 1
obtained from RC tests are compared with the corresponding 0.3
values computed using the HS model and simulating a simple 0.8
shear test with increasing strain amplitude. Higher values of the
Δu/p'0
G/G0

ratio E50
0
ref 0 ref
/Eoed were needed for best fitting the experimental 0.6 0.2
results with respect to those required for calibration of monotonic
TX-CIU tests. The secant stiffness predicted by the model, shown 0.4
with a solid line in the figure, decreases progressively as strains 0.1
model
0.2
increase for values of es40.002%, plotting slightly above the RC data
experimental data. The computed values of excess pore water
0 0
pressure are in a fair agreement with the experimental values for
0.0001 0.001 0.01 0.1 1
the alluvial silt and the silty clay, while some overestimate of Du is
γ (%)
obtained for the earth dam for g40.06%. Assuming the volumetric
threshold gv to correspond to a ratio Du=p00 ¼ 0:05, significant Fig. 9. Comparison between stiffness decay observed in RC tests and model
pore water pressure build-up can be anticipated to occur for simulations.
 ARTICLE IN PRESS

S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798 789

calibration on the overall response of the earth dam to earthquake The time integration in Plaxis is carried out using the Newmark
loading. scheme: standard values of constants a ¼ 0.25 and b ¼ 0.50 were
adopted to obtain an unconditionally stable solution. Time
integration was carried out using the same time step Dt ¼ 0:01 s
4.3. FE model adopted for the input accelerograms.

The FE analyses were carried out under plane strain conditions

using triangular six-noded elements with three-point Gaussian 4.4. Back analysis of earth dam construction
integration, a second-order polynomial interpolation for the
displacements and a first-order interpolation for the pore water Prior to studying the dam response to earthquake loading, the
pressure, for a total of 3811 elements (Fig. 10). The FE mesh construction of the earth dam was simulated to reproduce the
extended horizontally to about 400 m from the axis of the earth initial state of effective stress. Comparison of dam settlements
dam and vertically down to a depth of 300 m, where the influence observed during and after construction with the computed ones
of the earth dam is negligible. permitted to check the constitutive model calibration.
In static analyses vertical boundaries were restrained in the The pore water pressure before dam construction was assumed
horizontal direction, while displacements of the bottom of the hydrostatic with the water table at the ground surface and the
mesh were restrained both horizontally and vertically. In seismic initial state of effective stress was computed using values of the
analyses absorbent boundaries were adopted along the vertical coefficient of earth pressure at rest K0 written in the form
and horizontal edges of the FE mesh and seismic loading was
K 0 ¼ ð1  sin j0 ÞOCRsin j
0

applied imposing the selected input accelerograms at the (3)

considered bedrock depth. The absorbent boundaries are based where OCR is the overconsolidation ratio before the earth dam
on the scheme described by Lysmer and Kuhlemeyer [26] in which construction. This was estimated using the OCR profile obtained
an increase in stress on the boundary is absorbed independently from the oedometer tests performed in the supplementary
of the frequency of the incident waves. geotechnical investigation, carried out about 25 years after the
To avoid numerical distortion of the propagating wave during end of the dam construction; values of K0 and OCR used in the
the dynamic analysis, the maximum height of elements of the analyses are listed in Table 5.
dam ðDlmax ¼ 3 mÞ and of the alluvial silt layer ðDlmax ¼ 6 mÞ was Modelling the earth dam history included the staged con-
smaller than 16 of the wavelength lmin associated with the highest struction of the dam, considered as a drained process, via the
frequency component of the input wave fmax [27]. A value of progressive activation of 19 rows of elements about 2.5 m thick,
fmax ¼ 10 Hz was fixed since negligible energy content is asso- the impoundment of the reservoir and the associated steady state
ciated to higher frequencies (Fig. 2); the input accelerograms were seepage flow through the dam.
then low-pass filtered removing frequency components higher Model calibration was checked against the settlements
than 10 Hz. measured during the dam construction via four extensometers
It was pointed out that, for strain cycles of constant or
decreasing amplitude, the soil model predicts negligible hysteretic
Table 5
damping. Therefore, the overall capability of the model to OCR and K0 values
dissipate energy was deemed insufficient and additional viscous
damping had to be introduced in the analysis using the Rayleigh Soil z̄ (m) Current Before dam construction
formulation.
s 0
v0 OCR s0v0 OCR K0
A viscous damping ratio x ¼ 10% was selected as an average
lower bound of the equivalent damping obtained by Cascone and Alluvial silt 6.0 752.3 1.5 60.0 18.8 2.31
Rampello [10] from 1D and 2D equivalent linear analyses of the Silty clay 18.5 847.3 2.0 185.0 9.2 1.49
earth dam. Values of Rayleigh coefficients aR ¼ 1 and bR ¼ 0.01 Silty clay 37.5 1010.0 1.5 375.0 4.0 1.03
Silty clay 74.6 1300.0 1.3 746.0 2.3 0.61
were obtained to ensure a nearly constant damping in the range of
Silty clay 199.6 2400.0 1.3 1996.0 1.6 0.59
predominant frequencies for the problem at hand, f ¼ 0.8–2.4 Hz.

374 m
48 m

zB= 25 m

50 m
300 m

100 m

300 m

800 m

Fig. 10. Mesh used in the numerical analyses.

 ARTICLE IN PRESS

790 S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

A1, A2 A3, A7
196 m a.s.l. 186
156 151

0 150 m
A3 - Hdam = 34 m
A3 - end of construction (eoc)
A3 - 5 years after eoc
A3 - 8 years after eoc
A3 - 12 years after eoc
A7 - Hdam = 34 m
A1 - Hdam = 34 m A7 - eoc
A2 - Hdam = 34 m model - Hdam = 34 m
model model - eoc

190 190

180 180
elevation (m a.s.l.)

170 170

160 160

150 Dam base 150 Dam base

axis downstream
140 140
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
settlements (m) settlements (m)

Fig. 11. Observed and computed settlement profiles: (a) extensometers A1 and A2 and (b) extensometers A3 and A7.

|ac,max| = 0.32 g at t = 25.20 s

200
Artificial acceler.
180

160

140
-200 -150 -100 -50 0 50 100 150 200
-upstream + downstream

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 (g)

|ac,max| = 0.40 g at t = 6.90 s

200
Real acceler.
180

160

140
-200 -150 -100 -50 0 50 100 150 200

Fig. 12. Contour lines of horizontal acceleration at the time when maximum acceleration is reached at the crest of the dam for artificial (a) and real (b) accelerograms
(zB ¼ 50 m).
 ARTICLE IN PRESS

S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798 791

installed at the centre line and in the downstream slope of Poulos et al. [28] in which it is assumed that the foundation soil
the dam (see Fig. 1). The observed settlement profiles are shown is incompressible. Fig. 11a shows the settlement profiles along two
in Fig. 11; they were obtained following the procedure by vertical axes (A1 and A2) through the crest of the dam at a stage
when the dam was not yet completed (H ¼ 34 m), while Fig. 11b
shows profiles along two vertical axes located on the downstream
Table 6 slope of the dam (A3 and A7). Measurements from extensometer
Maximum acceleration and amplification ratios at the dam crest A3 span over a long period of time during the dam construction
and after the dam was completed in 1975. The measured
Input zB abr (g) abase (g) ac,max (g) ac,max/abase ac,max/abr
settlement profiles attain the maximum value at approximately
Artificial 25 0.32 0.29 0.40 1.36 1.27 mid-height, consistently with 1D analytical and 2D numerical
50 0.30 0.28 0.32 1.14 1.08 results of homogeneous earth dams (e.g. [29]). The observed slight
100 0.29 0.19 0.24 1.25 0.83 deviation from the theoretical parabolic pattern might be
300 0.27 0.10 0.17 1.63 0.62
attributed to the presence of large banks on both sides of the
Real 25 0.28 0.23 0.42 1.83 1.51 embankment. In addition, the observed increase in settlements at
50 0.29 0.18 0.40 2.22 1.37 the top of extensometers can be attributed to the construction of
100 0.27 0.14 0.27 1.87 0.99
the overlying portion of the dam and to the consolidation process
300 0.27 0.08 0.21 2.67 0.77
of the dam after the end of construction.

2.5 zB = 25 m
2.5
50 m
100 m
2 300 m 2
Amplification

Amplification

1.5 1.5

Artificial acceler. Real acceler.

1 1

0.5 0.5

0 0
0 2 4 6 8 10 0 2 4 6 8 10
f (Hz) f (Hz)

Fig. 13. Amplification functions for artificial and real accelerograms.

200

180

160

140
-200 -150 -100 -50 0 50 100 150 200

0.00 0.15 0.30 0.45 0.60 0.75 0.90 1.0

200

180

160

140
-200 -150 -100 -50 0 50 100 150 200

Fig. 14. Contour lines of mobilised shear strength t/tmax for the artificial accelerogram (zB ¼ 50 m): (a) at the time when amax is reached at the crest and (b) at the end of
earthquake.
 ARTICLE IN PRESS

792 S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

The computed settlement profiles shown with solid lines in 5.1. Acceleration field
Fig. 11, embody the observed behaviour, capturing with a fair
agreement the overall deformation pattern of the dam during and Fig. 12 shows the acceleration contours obtained, for zB ¼ 50 m,
after construction. In addition to a fair estimate of the deforma- at the time when maximum acceleration is reached at the crest of
tion of the dam body, the settlement computed at the base of the the dam. For the artificial accelerogram, the acceleration field is
dam centre line wcomp ¼ 1.5 m is in good agreement with the gradually increasing with elevation from about 0.1 to 0.32g, with
settlement wmeasE1.6 m measured 11 years after the end of zones of higher acceleration extending towards the upstream
construction, indicating a satisfactory choice of the input stiffness slope (Fig. 12a); the direction of shaking is about uniform
parameters for both the dam and the foundation soils. throughout the dam. Conversely, for the real accelerogram, the
acceleration field is not in phase, acceleration being directed
downstream at the crest and upstream at the banks. Along the
5. Dynamic analysis centre line the acceleration attains low values from the base to
the upper third of the dam and increases abruptly thereafter
Coupled effective stress dynamic analyses were carried out for (Fig. 12b).
both the artificial and the real accelerograms. The analyses were Table 6 lists the values of the peak acceleration of the input
prolonged till 60 s, in order to allow the velocity field to vanish, motion at the bedrock, abr, the maximum acceleration at the base
regardless the actual duration of the accelerograms. Analysis and at the crest of the dam, abase and ac,max, as obtained from the
results are presented in the following sections in terms of numerical analyses, and the ratios of crest to base (ac,max/abase)
acceleration field and of stress, strain and displacement fields. and crest to bedrock (ac,max/abr) maximum accelerations. The
maximum accelerations at the base and at the crest of the dam
193.2 m a.s.l. decreases for increasing depth of the bedrock, and the values of
ac,max obtained using the real motion are slightly larger than those
172.8
O
M
P 163.6 obtained using the artificial accelerogram. The ratio ac,max/abase is
in the range of 1.8–2.7 and 1.1–1.6 for the real and artificial
motion, respectively. Maximum values of the amplification ratio
600
effective stress state at: (ac,max/abase) refer to weak motions (abaseo0.15g), propagating
start of input motion from deep bedrock (zB ¼ 100–300 m), that mostly deform the dam
M end of input motion

193.2 m a.s.l.
400
O P 172.8
σ'a (kPa)

M 163.6
K0 = 0.546
P

M
200 o 120
P

80
Δu (kPa)

40
0
0 200 400 600 0
σ'b (kPa)
-40
Artificial acceler.
200
σ' a (kPa) -80
0 5 10 15 20 25 30 35 40
σ'a > σ'b > σ'c σ'a > σ'c > σ'b time (s)

M
100 120
p'=247 kPa
P 80
O
σ'b > σ'a > σ'c σ'c > σ'a > σ'b
Δu (kPa)

40

0
0
p'=126 kPa

-40
Real acceler.

σ 'b σ'c -80

0 5 10 15 20 25 30 35 40

σ'b > σ'c > σ'a σ'c > σ'b > σ'a time (s)
Fig. 16. Pore water pressure build-up computed for artificial (a) and real (b)
Fig. 15. Undrained stress paths for selected soil elements in the dam. accelerograms in selected soil elements of the dam (zB ¼ 50 m).
 ARTICLE IN PRESS

S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798 793

in the range of small shear strains. Values of amplification ratio as the central portion of the embankment, to the left of the central
large as 1.8 and 2.2 were obtained using the real accelerogram for drain, while most of the earth body experiences values of t/tmax
zB ¼ 25 and 50 m; smaller values of the ratio were obtained using smaller than 0.6 (Fig. 14a). At the end of the seismic shaking the
the artificial accelerogram due to the significantly higher accel- mobilised shear strength reduces to 0.6–0.75 in the central
erations at the base of the dam. For a deep bedrock the ratio portion of the dam while it is t/tmaxo0.45 throughout the rest
ac,max/abr is minimum and smaller than unity, denoting a of the embankment (Fig. 14b).
global de-amplification of the original input motion; conversely, For the same bedrock depth, Fig. 15 shows the stress paths of
for shallower depths of the bedrock (zB ¼ 25–50 m) the ratio soil elements O and P, located at the elevation of the banks
ac,max/abr is greater than unity showing a net amplification of the (172.8 m a.s.l.) at about 29 m from the dam centre line, and
original input motion. element M, located along the centre line, about 10 m below
Fig. 13 shows the amplification functions obtained as the ratio (163.8 m a.s.l.).
between the Fourier spectra of the crest and the outcropping Subfixes a, b and c in the figure denote principal directions of
motions. Amplification functions relative to the artificial motion stress. Direction c is normal to the plane of deformation;
exhibit a plateau in the range of frequencies of 0.5–1.4 Hz for all directions a and b, contained in the plane of deformation, are
the considered bedrock depths (Fig. 13a). Maximum amplification about the vertical and the horizontal directions at the beginning
is about 2 for zB ¼ 25 and 50 m and 1.5 for z ¼ 100 m occurring for of earthquake loading, but can rotate during the seismic event.
a frequency of about 1 Hz; for zB ¼ 300 m no amplification is Then, principal stresses s0a , s0b and s0c are not ordered according to
observed. Amplification functions relative to the real motion their value.
exhibit a sharp spike, maximum amplification of 1.1–2.1 being During earthquake loading the soil elements experience a
attained for a frequency of about 0.6 Hz (Fig. 13b). Comparing the progressive decrease of horizontal effective stress with only minor
amplification functions with the Fourier amplitude spectra shown reduction of vertical stress, approaching the condition of active
in Fig. 2c it is apparent that in the range of frequencies of limit equilibrium (Fig. 15a). At the end of the seismic shaking, the
maximum amplification the artificial accelerogram is charac- stress state of the soil elements is characterised by lower effective
terised by large amplitudes, while the real accelerogram is stresses due to the excess pore water pressure developed during
characterised by low amplitudes. the earthquake loading. The same stress paths are also repre-
sented on the deviatoric stress plane (Fig. 15b) in which distances
from the origin are equal to the octahedral shear stress
5.2. Stress and strain fields
toct ¼ 13½ðs0a  s0b Þ2 þ ðs0b  s0c Þ2 þ ðs0c  s0a Þ2 1=2 (4)
The stress state within the dam embankment was evaluated In this plane, the initial stress state of elements M and P are close
using the mobilised shear strength t/tmax where t ¼ ðs01  s03 Þ=2 to axisymmetric conditions; during shaking the stress states of
and s01 and s03 are the maximum and minimum principal effective the soil elements undergo a progressive increase in the octahedral
stresses; as values of t/tmax approach unity, the full shear strength shear stress remaining in the sector s0a 4s0b 4s0c in which s0a and
of the soil is attained. s0c are the maximum and the minimum principal effective
Fig. 14 shows the contour lines of mobilised shear strength stresses. Stress path of soil element O plots within the sector
obtained using the artificial accelerogram at zB ¼ 50 m as input s0a 4s0c 4sb in which s0a and s0b are the maximum and minimum
motion. At the time when the maximum acceleration is reached at principal effective stresses. During the earthquake loading, the
the crest, values of t/tmax in the range of 0.9–1.0 are obtained in stress paths approach the Mohr–Coulomb strength envelopes

200
Artificial acceler.
180

160

140
-200 -150 -100 -50 0 50 100 150 200

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 (%)

200
Real acceler.
180

160

140
-200 -150 -100 -50 0 50 100 150 200

Fig. 17. Contour lines of deviatoric strain es at the end of earthquake motion for artificial (a) and real (b) accelerograms (zB ¼ 50 m).
 ARTICLE IN PRESS

794 S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

relevant to elements O and P (p0min ¼ 126 and 137 kPa) and to 5.3. Displacement field
element M (p0min ¼ 247 kPa) without reaching failure conditions.
Fig. 16 shows typical trends of pore water pressure build-up Fig. 18 shows vertical and horizontal displacements at different
computed in soil elements M and P during seismic shaking using elevations along the centre line of the dam cross section, as
both the artificial and the real accelerograms. Values of excess obtained by applying the input motion at the bedrock depth
pore water pressure obtained in the soil element M located in the zB ¼ 50 m. Vertical displacements decrease with elevation being
lower portion of the dam are higher than those computed maximum at the crest and minimum at the base of the dam: using
downstream, in point P. For any of the bedrock depths considered the artificial accelerogram as input motion the crest and base
in the analyses, a larger rate of pore water pressure build-up was displacements were 25 and 6.5 cm, respectively (Fig. 18a), while
obtained using the real accelerogram, consistently with the for the real accelerogram the corresponding values were 16 and
shapes of the applied acceleration time histories. 4 cm (Fig. 18b). Permanent displacement results from plastic
Fig. 17 shows the contours of the deviatoric strain strains that accumulate during the earthquake because of
pffiffiffi  progressive plastic loading of the yield surfaces fs and fv and of
1=2
2 3 the instantaneous mobilisation of the available shear strength.
s ¼ ðx  y Þ2 þ 2x þ 2y þ g2xy (5)
3 2 Therefore, the final permanent displacements are influenced by
the duration of the strong motion phase of the accelerograms.
obtained applying the input motions at a bedrock depth zB ¼ 50 m. The same differences in the dam response can be observed in
Both artificial and real accelerograms yield values of shear Fig. 18c–d showing the horizontal displacement time histories
strain esX0.8% in zones of limited extension located at the computed at the same four locations in the dam using both the
upstream toe and at the crest of the dam. In most of the artificial and the real accelerograms. The displacement response
embankment the shear strain is in the range of 0.1–0.3%, resulting to the artificial motion does not exhibit significant variations
greater than the volumetric threshold observed in the RC test within the dam body, horizontal displacement being directed
results (E0.2%) only locally. towards upstream with a maximum displacement of about 22 cm

193.2 m a.s.l. CC 195.8

CC
CU 179.4 CU
CL
CL 165.7
CB
148.2
CB

0 0
vertical displacement (cm)

vertical displacement (cm)

-10 -10

-20 -20

-30 -30
Artificial acceler. Real acceler.

-40 -40
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time (s) time (s)

20 20
horizontal displacement (cm)

horizontal displacement (cm)

10 10

0 0

-10 -10

-20 -20

-30 -30
Artificial acceler.
-40 -40 Real acceler.

-50 -50
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time (s) time (s)

Fig. 18. Vertical and horizontal displacements computed for artificial (a–c) and real (b–d) accelerograms along the centre line of the dam cross section for zB ¼ 50 m.
 ARTICLE IN PRESS

S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798 795

at the crest (Fig. 18c). The displacement time history is very horizontal displacement profiles for the two input motions
irregular and is characterised by sharp alternate peaks, denoting a are qualitatively similar with those typical of a second mode
back and forth motion until the end of the external excitation. The response of a linear elastic inhomogeneous shear beam [30].
displacement time histories obtained using the real accelerogram Differences in the horizontal displacement profiles are noticeable
are more regular and all the selected points move in the in the upper third of the dam (Fig. 20c–d). As a general rule,
downstream direction, with displacements in the range of horizontal and vertical displacements along the dam centre line
7–11 cm (Fig. 18d). are increasing with decreasing bedrock depth. However, it can be
Fig. 19 shows the influence of the bedrock depth on the vertical observed that for zB ¼ 25 and 50 m, the displacement profiles
displacements computed at the crest of the dam. In the case of the obtained for the real motion are almost coincident, possibly due to
artificial motion (Fig. 19a) the maximum permanent displacement the discussed coupling of predominant periods occurring for
is in the range of 22–30 cm and decreases for increasing zB. zB ¼ 50 m. The overall deformation pattern of the earth dam
Displacements of 8–16 cm are obtained when applying the real mainly consists in dam compaction associated with lateral
motion (Fig. 19b), with the maximum value computed for a deformation.
bedrock depth zB ¼ 50 m. This may be ascribed to a close coupling
occurring in this case among the predominant periods of the input
motion (E0.65 s), the 50 m thick soil layer (E0.74 s) and the earth 6. Discussion and conclusions
dam (E0.57 s).
Figs. 18 and 19 also show the long-term values of displacement In the different methods of analysis adopted to study the
computed at the end of the consolidation analyses. Increments of seismic performance of the Marana Capacciotti earth dam, the soil
vertical displacement at the crest of the dam, due to pore water behaviour was described using different constitutive assumptions,
pressure dissipation, are of 22–40%, with the maximum value from the simple rigid-perfectly plastic behaviour assumed in the
attained for the real motion (Fig. 19). The corresponding incre- sliding-block analyses, to the visco-elastic model adopted in the
ments of horizontal displacement are nearly negligible for both total stress ground response analyses, to the elastic–plastic model
the artificial and the real accelerograms. with isotropic hardening used in the effective stress FE dynamic
Vertical and horizontal displacements relative to the base of analyses, this requiring an increasing number of input parameters.
the dam were computed to obtain the displacement profiles of the The constitutive model used in the FE analyses is capable to
dam centre line (Fig. 20). Vertical displacement profiles evaluated reproduce to some extent non-linear soil behaviour with devel-
using the artificial motion (Fig. 20a) are about two times larger opment of irreversible plastic strains from the early beginning of
than those obtained using the real accelerogram (Fig. 20b). The the loading path.
Parameter calibration for FE analyses was based on TX test
results that yielded a steeper decrease of shear stiffness with
shear strain than that obtained from RC tests. Fig. 21 shows a
comparison between the earth dam displacements obtained using
0
TX-based (thick lines) and RC-based (thin lines) calibrations of
model parameters and using the real accelerogram applied at
vertical displacement (cm)

-10 zB ¼ 50 m. Fig. 21a shows the influence of calibration on vertical

300 m displacements of points located along the centre line of the dam.
100 m Using the results of RC tests to calibrate model parameters,
-20 50 m vertical displacements are 29–98% of those computed after TX-
based calibration; differences are larger for deeper points in the
300 m dam where the absolute values of displacement are smaller.
-30 100 m Fig. 21b shows the influence of calibration on horizontal
zB = 25 m 50 m displacements of points located on the slopes of the dam:
Artificial acceler.
25 m displacements are nearly coincident for points resting on the
-40 downstream slope, while for points located on the upstream slope
0 5 10 15 20 25 30 35 40 the RC-based displacements are larger than those obtained after
time (s) TX-based calibration, being however less than about 10 cm.
As it could be anticipated, different results are obtained
using different sets of experimental data to calibrate model
0 parameters, though, for the case considered in this study, the
overall pattern of behaviour of the dam was not altered
vertical displacement (cm)

300 m substantially.
-10 100 m
300 m The maximum acceleration profiles obtained from the effective
25 m stress FE analyses are shown in Fig. 22 together with the range of
100 m
zB=50 m results found in the literature [6–9,31–33], and with the results
-20 25 m
obtained from 2D equivalent linear analyses of the dam [10];
50 m a(z ¼ 12 m) in the figure is the maximum acceleration at the contact
between the alluvial soil and the underlying deposit of stiff clay
-30
(z ¼ 12 m). A fair agreement among results is observed, though
Real acceler. smaller amplification ratios are obtained for the coupled effective
stress analyses carried out using the artificial accelerogram,
-40
possibly due to more significant non-linear effects than those
0 5 10 15 20 25 30 35 40
reproduced in the equivalent linear method. For both real and
time (s)
artificial accelerograms, the acceleration profiles define a narrow
Fig. 19. Influence of bedrock depth zB on vertical displacements computed at the band. For the two profiles obtained assuming zB ¼ 300 m a
crest of the dam for artificial (a) and real (b) accelerograms. slightly different trend is observed with larger amplification in
 ARTICLE IN PRESS

796 S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

200 200

190 190

180 180

elevation (m a.s.l.)
elevation (m a.s.l.)

zB=25 m

170 170 50 m
100 m
300 m

160 160

150 150
Artificial acceler. Real acceler.

140 140
0 -10 -20 -30 0 -10 -20 -30
vertical displacement (cm) vertical displacement (cm)

200 200

190 190

180 180
elevation (m a.s.l.)

elevation (m a.s.l.)

170 170

160 160

150 150

Artificial acceler. Real acceler.

140 140
-10 -5 0 5 10 -10 -5 0 5 10
horizontal displacement (cm) horizontal displacement (cm)

Fig. 20. Vertical and horizontal displacement profiles computed along the dam centre line for artificial (a–c) and real (b–d) accelerograms.

the dam. In these cases, large amplifications are attributable to a response because of the larger participation of high frequency
mainly elastic response of the dam to the weak motions resulting motion components [34].
from propagation from a deep bedrock, the maximum accelera- The artificial accelerograms used in the analyses were
tion at the crest being smaller than those obtained for the other characterised by a long duration of the stationary strong motion
bedrock depths (Table 6). phase and by a broadband frequency content, resulting in larger
The acceleration field computed by the FE analyses (Fig. 12) is displacements of the earth dam compared with those induced by
in a fair agreement with the contour lines of amax evaluated by the the real input motions. In fact, spectrum compatible time histories
2D total stress FE analyses carried out via the linear equivalent match the entire spectrum within a prescribed tolerance, so that
method [10]. Acceleration contours in Fig. 12 refer to the time all periods are subjected to the full design seismic action and
instant when acceleration is maximum at the crest. Consistently displacement response may therefore be overestimated [2].
with the acceleration fields predicted by equivalent linear Displacements of the earth dam generally decreased as the
analyses [6,10] and non-linear analyses [34], maximum accelera- bedrock depth increased, due to the filtering effects of the
tions, as large as 0.28 and 0.36g, occurred at the slopes and foundation soil, although some amplification was observed when
close to the banks; this result can be ascribed to lack of using real accelerograms applied at shallow depths (zB ¼ 25–50 m);
confinement and geometrical effects that amplify the dam for these cases some coupling among the fundamental
 ARTICLE IN PRESS

S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798 797

193.2 m a.s.l. CC 195.8 amax/az=12 amax/az=12

UU CU 179.4 0 1 2 3 4 0 1 2 3 4
DU
CL 165.7 0
148.2
CB

10
0.2
vertica ldisplacement (cm)

0
zB = 300 m zB = 300 m

0.4
-10
RC

z/H
TX
CC
-20 CU literature results
CL 0.6
CB
-30
Cascone and
0 5 10 15 20 25 30 35 40 Rampello (2003)
time (s)
0.8
model

30
Artificial Real
accelerograms accelerograms
horizontal displacement (cm)

20
1

10 Fig. 22. Comparison between computed maximum acceleration profiles and

published results.

0 TX

-10 RC

UU Table 7
-20 DU Post-seismic stability evaluation

zB Artificial Real
-30
0 5 10 15 20 25 30 35 40 25 1.83 1.96
time (s) 50 1.91 1.97
100 2.01 2.11
Fig. 21. Influence of parameter calibration on vertical (a) and horizontal 300 2.12 2.19
(b) displacements computed using the real accelerogram at zB ¼ 50 m.

frequencies of input motions, foundation soil and earth dam deviatoric strain was generally smaller than 0.8%, with the
possibly occurred. exception of confined zones located at the upstream toe of the
Effective stress FE analyses permitted to evaluate post-seismic dam; horizontal and vertical displacements through the dam were
stability of the earth dam. To this purpose, undrained analyses were smaller than the values at the crest; at several instants during the
carried out in terms of effective stresses, accounting for the pore earthquake, available shear strength was mobilised in a confined
pressure build-up induced by earthquake loading. Stability conditions zone located in the upper third of the dam.
were evaluated by progressively reducing the effective cohesion c0 and Comparison between the effective stress FE analyses and the
the angle of shearing resistance j0 by a factor F until large displacement-based methods was carried out in terms of vertical
displacements of the dam were obtained. Using this procedure the displacements computed at the crest of the dam. Coupled
factor of safety was estimated as the value corresponding to a change effective stress dynamic analyses provided values of vertical
in the curvature of the F-displacements curves. The safety factor displacements of 22–30 cm and of 8–16 cm using the artificial and
increased with zB, being in the range of 1.96–2.19 and of 1.83–2.12 for real accelerograms as input motions, respectively. Using the
the real and the artificial accelerograms, respectively (Table 7). The accelerograms obtained from the 2D ground response analysis,
maximum reduction of F with respect to the static condition the sliding block analysis yielded dam displacements of 10–40 cm,
(F ¼ 2.33) was of about 21.5%, and occurred using the artificial the higher values being in a fair agreement with effective stress FE
accelerogram at zB ¼ 25 m as input motion. For each of the assumed analysis results.
bedrock depths and for both artificial and real input motions, the For any of the seismic scenarios considered in this study and
computed failure mechanisms occurred in the downstream slope of for each of the methods of analysis adopted, the estimated seismic
the dam, involving its upper third. performance of the dam was deemed satisfactory in that the
Through the FE dynamic analyses some insight was gained on maximum crest displacement was considerably lower than the
the seismic behaviour of the dam. Specifically, the computed service freeboard (2.6 m).
 ARTICLE IN PRESS

798 S. Rampello et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 782–798

References [16] CNR-GNDT. L’attività del GNDT nel triennio 1993–1995, a cura di A.
Corsanego, E. Faccioli, C. Gavarini, P. Scandone, D. Slejko, M. Stucchi; 1996.
[17] Newmark NM. Effect of earthquakes on dams and embankments. Géotechni-
[1] Finn LWD, Ledbetter RH, Marcusson WF. North American practise for
que 1965;15(2):139–60.
evaluating the seismic safety of embankment dams. In: Ishihara K, editor.
[18] Cai Z, Bathurst RJ. Deterministic sliding block methods for estimating seismic
Earthquake Geotechnical Engineering. Rotterdam: Balkema; 1997. p. 1227–52.
displacements of earth structures. Soil Dyn Earthquake Eng 1996;15:255–68.
[2] Finn LWD. State-of-the-art of geotechnical earthquake engineering practice.
[19] Rathje EM, Bray JD. An examination of simplified earthquake-induced
Soil Dyn Earthquake Eng 2000;20:1–15.
[3] Finn LWD, Yogendrakumar M, Yoshida N, Yoshida H. TARA-3: a program to displacement procedures for earth structures. Can Geotech J 1999;36:72–87.
compute the response of 2-D embankments and soil-structure systems to [20] Rathje EM, Bray JD. Nonlinear coupled seismic sliding analysis of earth
seismic loadings. Canada: Department of Civil Engineering, University of structures. J Geotech Geoenviron Eng ASCE 2000;120(11):1002–14.
British Columbia; 1986. [21] Vermeer P, Brinkgreve RBJ. Plaxis V7, Material models manual. Rotterdam:
[4] Griffiths DV, Prevost JH. Two- and three dimensional dynamic finite element Balkema; 1998.
analyses of the long valley dam. Géotechnique 1988;38:367–88. [22] Schanz T. Zur modellierung des mechanischen verhaltens von reibungsma-
[5] Elgamal A-WM. Three dimensional seismic analysis of La Villita Dam. J terialen. Habilitation, Stuttgart Universität; 1998.
Geotech Eng ASCE 1992;118:1937–58. [23] Schanz T, Vermeer PA, Bonnier PG. Formulation and verification of the
[6] Seed HB, Seed RB, Lai SS, Khamenehpour B. Seismic design of concrete faced hardening-soil model. In: Brinkgreve RBJ, editor. Beyond 2000 in computa-
rockfill dams. In: Concrete face rockfill dams—design, construction and tional geotechnics. Rotterdam: Balkema; 1999. p. 281–90.
performance. ASCE; 1985. p. 459–78. [24] Viggiani G. Small strain stiffness of fine grained soils. PhD thesis, City
[7] Gazetas G, Dakoulas P. Seismic analysis and design of rockfill dams: state-of- University, London; 1992.
the-art. Soil Dyn Earthquake Eng 1992;11:27–61. [25] Rampello S, Viggiani GMB. Pre-failure deformation characteristics of
[8] Özkan MY, Erdik M, Tunc- er MA, Yilmaz C - . An evaluation of Sürgü Dam geomaterials. Discussion leader report on session 1a: laboratory tests. In:
response during 5 May 1986 earthquake. Soil Dyn Earthquake Eng 1996; Proceedings of the second international symposium on ‘pre-failure deforma-
15:1–10. tion characteristics of geomaterials’, vol. 2, Torino; 2001. p. 1279–89.
[9] Troncoso JH, Krause AJ, Corser PG. Seismic design of lined face earth dams. In: [26] Lysmer J, Kuhlemeyer RL. Finite dynamic model for infinite media. J Eng Mech
Proceedings of the second international conference on earthquake geotech- ASCE 1969;95(4):859–77.
nical engineering, Lisbon; 1999. p. 703–9. [27] Kuhlemeyer RL, Lysmer J. Finite element method accuracy for wave
[10] Cascone E, Rampello S. Decoupled seismic analysis of an earth dam. Soil Dyn propagation problems. J Soil Mech Found ASCE 1973;99(5):421–7.
Earthquake Eng 2003;23:349–65.
[28] Poulos HG, Booker JR, Ring GJ. Simplified calculation of embankment
[11] Whitman RW, Liao S. Seismic design of retaining walls. In: Proceeding of the
deformations. Soils Found 1972;12(4):1–17.
eighth world conference on earthquake engineering, San Francisco; 1984. p.
[29] Pagano L, Desideri A, Vinale F. Interpreting settlement profiles of earth dams. J
533–40.
Geotech Geoenviron Eng ASCE 1998;124(10):923–32.
[12] Ambraseys NA, Menu JM. Earthquake-induced ground displacements. Earth-
[30] Gazetas G. Shear vibration of vertically inhomogeneous earth dams. Int J
quake Eng Struct Dyn 1988;16:985–1006.
[13] Jibson RW. Predicting earthquake-induced landslide displacements using Numer Anal Methods Geomech 1982;6:219–41.
Newmark’s sliding block analysis. Transportation Research Record 1411, [31] Makdisi FI, Kagawa T, Seed HB. Seismic response of earth dams in triangular
Transportation Research Board, National Research Council, Washington DC; canyons. J Geotech Eng ASCE 1982;108(10):1328–37.
1993. p. 9–17. [32] Prato CA, Delmastro E. 1-D seismic analysis of embankment dams. J Geotech
[14] Calabresi G, Rampello S, Sciotti A, Amorosi A. Diga sulla Marana Capacciotti: Eng ASCE 1987;113(8):904–9.
Verifica delle condizioni di stabilità e analisi del comportamento in [33] Gazetas G. Seismic response of earth dams: some recent developments. Soil
condizioni sismiche. Res. Rep. Dip. di Ingegneria Strutturale e Geotecnica, Dyn Earthquake Eng 1987;6:2–47.
Università di Roma La Sapienza; 2000. [34] Abouseeda H, Dakoulas P. Non-linear dynamic earth dam-foundation
[15] EN 1998-1 Eurocode No. 8: design of structures for earthquake resistance of interaction using a BE-FE method. Earthquake Eng Struct Dyn 1998;
structures, CEN European Committee for Standardisation; 2003. 27:917–36.