ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

www.elsevier.com/locate/soildyn

Mitigation of the seismic motion near the edge of cliff-type topographies

C.A. Stamatopoulosa,, M. Bassanoua, A.J. Brennanb, G. Madabhushic
a
Stamatopoulos and Associates Co, 5 Isavron Street, Athens 11471, Greece
b
Division of Civil Engineering, University of Dundee, UK
c
School of Engineering, University of Cambridge, UK
Received 14 August 2006; accepted 28 January 2007

Abstract

Concentration of damage of buildings near the edge of hard or dense cliff-type topographies has been observed during a number of
recent earthquakes. These observations have been interpreted by numerical dynamic analyses that illustrate the amplification of
accelerations near the edge of slopes. The paper studies the effect of mitigating these accelerations using anchors both experimentally and
numerically. The main issue is that if the part of the slope in which topographic amplification occurs is connected to that in which the
acceleration is less, then the accelerations have to become more uniform. The experimental study involves dynamic centrifuge tests at
the Cambridge University facility both with and without anchors. The numerical procedure was verified by the seismic response of the
centrifuge tests. It was then applied to the study of the effect of anchors of a typical field case, the Aegion slope, under two different input
motions. In all cases anchors were found effective in mitigating the seismic motion near the edge.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Cliff-type topographies; Mitigation; Centrifuge tests; Numerical analyses; Anchors; 2D wave propagation

1. Introduction horizontal acceleration and the generation of a vertical

acceleration at the crest and near the edge of slopes under
Concentration of damage of buildings at or near the horizontal excitation (Refs. [1,2]). Parametric studies have
edge of hard or dense natural slopes has been observed also been performed for the influence of different
during a number of recent earthquakes. The Lambesc parameters on the amplification of horizontal and vertical
Earthquake in France, 1909, the San Fernando Earth- accelerations near the tip of slopes (Ref. [3]).
quake in 1971, the Friuli Earthquake in Italy, 1976, the Methods mitigating the topographic effect described
Irpinia Earthquake in Italy, 1980, the Chile Earthquake in above do not exist in the literature. Such methods are
1985, the Whittier Narrows Earthquake in 1987, the Eje- needed in cliffs with existing buildings that are not designed
Cafetero Earthquake in Colombia, 1998 and recent earth- to sustain the very large anticipated total and differential
quakes in Greece in Kozani, 1995 and in Athens, 1999 and accelerations in the case of a near seismic event.
in Turkey in Bingöl, 2003 are only some examples of The purpose of this work is to demonstrate the effect of
catastrophic events, during which severe structural damage anchors as a mitigation measure to decrease the severity of
has been reported on hilltops or close to steep slopes (Ref. this seismic motion. The main issue is that different regions
[1]). The collapse of a building in the central part of the of the slope have different acceleration histories. Therefore,
Aegion slope in Greece as a result of the 1995 earthquake is if the part of the slope in which topographic amplification
another example (Ref. [2]). occurs is connected to that in which the acceleration is less,
These observations have been interpreted by numerical then the accelerations have to become more uniform.
dynamic analyses that illustrate the amplification of the Alternatively, it can be stated that the specific use of
anchors increases the rigidity of the surface soil introducing
Corresponding author. Fax: +30 2103616919. an ‘‘equivalent’’ stiffer ‘‘reinforced soil’’ extending at some
E-mail address: [email protected] (C.A. Stamatopoulos). depth.

0267-7261/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2007.01.012
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1083

Nomenclature Sax-max, Say-max maximum horizontal, maximum verti-

cal spectral acceleration
ax, ay horizontal, vertical acceleration Sax-max-1D maximum horizontal spectal acceleration far
ax-max, ay-max maximum horizontal, maximum vertical away from the edge of the slope, where the
acceleration topographic effect does not exist
ax-max-1D maximum horizontal acceleration far away Sax-max-2D, Say-max-2D maximum horizontal, vertical
from the edge of the slope, where the topo- spectral acceleration near the edge of the slope
graphic effect does not exist for the case without mitigation
ax-max-2D, ay-max-2D maximum horizontal, vertical ac- Sax-max-anchors, Say-max-anchors maximum horizontal,
celeration near the edge of the slope for the case vertical spectral acceleration near the edge of
without mitigation the slope for the case with anchors
ax-max-anchors, ay-max-anchors maximum horizontal, verti- Sfric anchor friction angle
cal accelerations near the edge of the slope for Sbond cohesive strength of the grout of anchor
the case with anchors T period
c, j cohesion and friction component of peak soil TE-ax, TE-ay, TE-Sax, TE-Say factors defined by Eq. (4)
strength Vs shear wave velocity
E Young modulus of anchors T1, T2, T3, T4 locations of measurements of horizontal
g acceleration of gravity accelerations in centrifuge tests
Gel, G elastic shear modulus, secant shear modulus Uy vertical displacement
Kbond stiffness of the grout of anchors Uy-final final vertical displacement
M1, M2, M3, M4, M5 locations of measurements of Yield tensile yield strength of anchor
horizontal accelerations in centrifuge tests
ME-ax, ME-ay, ME-Sax, ME-Say factors defined by Greek letters
Eq. (5)
Nomit case without mitigation evp, gp volumetric and shear plastic strain.
S1, S2 locations of measurements of settlement in t, s0 shear and effective normal stress at failure plane
centrifuge tests fcs, b, Ep, gmob, b, sco model parameters of Eqs. (1) and
Sax, Say horizontal, vertical spectral acceleration (2)

The paper studies the effectiveness of this mitigation The effectiveness of the considered mitigation method is
method both experimentally, in the centrifuge, and numeri- studied in the centrifuge of Cambridge University. The
cally. In particular, the study involved three steps: (a) in- container was rectangular with cross-section of 0.56 
vestigation in the centrifuge of the dynamic response of a 301 0.238 m2 and height 0.226 m. The model was contained in
slope without and with anchors, (b) numerical analysis of the an equivalent shear beam model container described in
seismic response of the slope in the centrifuge both without detail by Zeng and Schofield [4]. This container consists of
and with anchors primarily to validate the numerical alternate layers of metal and rubber such that box
procedure and (c) numerical study both with and without vibrations during shaking match the vibrations of dry
anchors of the response of the field Aegion slope. Numerical sand, so the soil should be unaffected by the vertical
simulation of the field case is needed because the centrifuge boundary during the test.
geometry, even at the prototype, has much smaller dimen- Two slope configurations were studied: with and without
sions than actual field cliff-type topographies. For simplicity, anchors, as shown in Fig. 1. In the first configuration, the
similarly to previous analyses (Refs. [2,3]), structures on soil was Hostun S28 sand, poured dry at a voids ratio of
the surface of the slope were not included in the models. In approximately 0.67. Corresponding relative density Dr is
the centrifuge, measurements were in terms of only the 60%. Testing was performed at an acceleration of 50 g
horizontal accelerations, while the Aegion slope numerical acting normal to the base of the model, simulating the
analysis was in terms of both the horizontal and vertical vertical direction. Slope inclination was 301. Prototype
acceleration. slope height is 8 m.
Fig. 1 gives the monitoring points: horizontal accelera-
2. Investigation in the centrifuge tions are measured at positions marked with rectangles
labeled B, M, T depending on depth. Vertical displace-
By applying high centrifugal acceleration small soil ments of the free surface are measured by instruments
models are made to behave similarly to natural shapes of S1–S3. The earthquake applied at the tests was parallel to
larger dimensions, by increasing the self-weight stresses. the base of the model simulating the horizontal direction. It
 ARTICLE IN PRESS
1084 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

Fig. 1. Layout of centrifuge tests (a) without anchors and (b) with anchors. At (a) the dimensions and the location of measurements are also given.

is given in Fig. 2. It has a peak acceleration of 7.5 g and 8.0

main frequency of 50 Hz. The corresponding ‘‘prototype’’ 6.0
values are 0.15 g and 1 Hz, respectively, typical for earth- 4.0
quakes in Europe. 2.0
For the second configuration, anchors had 8 cm length
0.0
ax [g]

and 0.6 cm diameter (or 4 and 0.3 m, respectively, at the

-2.0
prototype) and were constructed from brass rod. Along
their length, Hostun sand had been glued with Araldite -4.0
adhesive in order to increase the interface friction angle in -6.0
line with anchors that have been grouted in place. At each -8.0
Amax = -7.7g
end, the rods were bolted to a steel faceplate, 25 mm -10.0
(1.25 m prototype) square and 3 mm (150 mm) thick. Each 0.0 0.2 0.4 0.6 0.8 1.0
rod is angled at 601 to the horizontal. Three rods are Time [sec]
installed at the edge and three at the back of the slope. The 40
other geometry factors, soil type and density and applied
35
accelerations were similar to the first configuration.
For the cases: (a) without mitigation and (b) with 30
anchors the measured peak accelerations and final settle- 25
Sax [g]

ments at all monitoring points are given in Table 1. Fig. 3 20

gives the measured settlement with time at location S2, 15
Figs. 4 and 5 give the measured horizontal accelerations 10
histories and response spectra, respectively, at locations T2
5
and T4.
Locations T2 and T4 are near and relatively far from the 0
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
edge of the slope, respectively. Figs. 4 and 5 illustrate that
T [sec]
as a result of anchors, maximum horizontal acceleration
and maximum spectral horizontal acceleration decrease at Fig. 2. Centrifuge tests. Input acceleration.
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1085

Table 1 involve considerable permanent deformations. In the

Centrifuge tests: measured, computed and comparison of peak accelera- present study a simple elasto-plastic method is used
tions and final settlements at all the positions of measurements
instead. The reason is that as the centrifuge tests include
Location (Fig. 1) Measurement Estimation Estimation/ measurements of permanent settlements, a method that can
Measurement be verified by comparing both accelerations and settle-
ments is preferred. In addition, as mitigation measures are
(a) Case without mitigation
studied, the elasto-plastic analysis has the additional
(i) ax-max (g)
T1 10.1 13.2 1.31 advantage that it investigates the danger of increased
T2 10.2 10.8 1.06 permanent displacements that the mitigation measure may
T3 14.6 11.9 0.82 cause.
T4 15.3 13.9 0.91 The elasto-plastic model that is used in the present study
M1 9.35 10.2 1.09
M2 9.07 10.0 1.10
is a simplified version of the model developed at Ecole
M3 8.57 9.6 1.12 Centrale Paris and described by Aubry et al. [5]. It has been
M4 9.52 9.89 1.04 calibrated in terms of soil type by Modaressi and Lopez-
M5 9.53 8.99 0.94 Caballero [6]. Soil response under drained conditions is
(ii) Uy-final (mm) predicted by the following equations:
S1 0.076 0.060 0.79
S2 0.136 0.132 0.97 t=s0 ¼ F tan fcs ðgp =ðtan fcs =E p þ gp ÞÞ0:5 , (1)
S3 0.0454 0.063 1.39
0
where F ¼ 1b ln(s /sc ), 0
sc0 ¼ sco exp (bepv )
(b) Case with anchors
(i) ax-max (g) pv p
¼ g ðtan fcs þ t=s Þ 0
for gp 4gmob . (2)
T1 10.2 12.07 1.18
T2 9.43 10.07 1.07 In the above equations t and s0 are the shear and
T3 11.1 8.38 0.75 effective normal stress at the failure plane, fcs, b, Ep, gmob,
T4 10.2 9.80 0.96 b and sco are model parameters and epv and gp are the
M1 9.23 11.34 1.23
volumetric and shear plastic strain. The model is based on:
M2 10.3 10.27 1.00
M3 8.73 8.64 0.99 (a) the critical state concept (the factor F gives the distance
M4 10.6 8.66 0.82 from the critical state), (b) the fact that plastic shear
M5 9.11 8.94 0.98 depends on the t/s0 ratio and (c) a Roscoe type dilation
(ii) Uy-final (mm) equation.
S1 0.048 0.060 1.25
Typical values of the parameters fcs, b, Ep, gmob, b for
S2 0.187 0.225 1.20
S3 0.088 0.073 0.83 sands and clays are given by Modaressi and Lopez-
Caballero [6]. The parameter sco depends on sand density
or clay precompression value. The Eqs. (1) and (2) under
T4. To further illustrate the effect of anchors, Fig. 6 gives cyclic loading do not simulate kinematic hardening: the
the absolute value of the difference of horizontal accelera- response during unloading and reloading when the past
tions at locations T4 and T2 with time for the cases (a) peak stress ratio is not exceeded is assumed elastic. At the
without and (b) with anchors. If the slope did not exist, present studies, this is not very important because the
clearly the acceleration histories at T2 and T4 would have applied accelerograms have only one, or two, cycles
been identical and this factor zero. It can be observed that of large intensity. In addition, for this reason, when
this factor decreases versus time for the anchors case. dynamic analysis is performed, damping should not have
Indeed, the average of the factor, over the 1654 increments its elastic value, but a larger value corresponding to the
of the acceleration records, decreases from 1.0 g for the magnitude of average shear stress applied. In particular,
case without mitigation to 0.6 g for the case with anchors, for the one-dimensional (1D) case of level ground,
or by 40%. comparison of predictions of the proposed method with
Fig. 3 illustrates that movements are very small, those of the equivalent-linear dynamic method performed
insignificant for engineering purposes: Maximum settle- by the computer code Cyberquake [7] has illustrated that
ment is 5 mm for the case without mitigation and 9 mm the Rayleigh damping value of 0.07 gives similar results.
for the case with anchors, prototype scale. This difference The numerical code FLAC [8] was used in the analyses.
in movements is within the bounds of experimental The elasto-plastic model described by Eqs. (1) and (2) was
variability. applied at the soil layers using the SS (strain hard-
ening–softening) elasto-plastic model that is included in
3. Numerical analysis procedure the FLAC code. Application was made using the equiva-
lent relationship of (a) the friction ( ¼ arctan(t/s0 )) and
Usually dynamic analyses of the seismic amplification (b) the dilation angles in terms of the plastic shear strain.
near the edge of hard or dense slopes are performed using Anchors are simulated using the structural anchor
linear or equivalent-linear models (Refs. [2,3]). Such elements that are built-in in the FLAC code. In addition
models are acceptable because the problem does not to the length and cross-sectional area of anchors, the
 ARTICLE IN PRESS
1086 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

0.02

-0.01

-0.04
Uy at S2 - No mit [mm]

-0.07

-0.10

-0.13

MEASURED
-0.16
ESTIMATED
-0.19
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time [sec]

0.05

0.00

-0.05
Uy at S2 - Anchors [mm]

-0.10

-0.15

-0.20

-0.25

-0.30
MEASURED
ESTIMATED
-0.35
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time [sec]

Fig. 3. Centrifuge tests. Measured and estimated settlements with time at location S2 at the cases (a) without mitigation and (b) with anchors.

following anchor properties are needed (a) density, (b) effective stress equal to 800 kPa. Drained tests gave peak
Young modulus (c) tensile yield strength, (d) friction angle friction angles of 461, 451 and 371 at horizontal consolida-
at the interface, (e) cohesive strength of the grout, and tion stresses of 50, 100 and 200 kPa. Cyclic triaxial tests
(e) stiffness of the grout. Two-dimensional (2D) plane indicated that at a shear strain equal to 1.3(103) the shear
strain analyses are used. The spacing of anchors in the modulus is 10, 21, 29 MPa at octahedral stresses of 50, 100
direction not modelled is simulated by FLAC using and 200 kPa, respectively. The dry density of the sand was
appropriate scaling rules for the anchor sizes and proper- estimated as 1.6 t/m3.
ties (Ref. [8]). Based on the above, Table 2 shows the values of the
parameters of the constitutive model given by Eqs. (1) and
4. Numerical analysis of the centrifuge tests (2) that were selected for the numerical analyses. The
values of the parameters fcs, b, Ep, gmob, b were obtained
4.1. Soil parameters and numerical modeling from the set proposed by Modaressi and Lopez-Caballero
[6]. The value of sco was obtained from the undrained and
In the laboratory of Stamatopoulos and Associates Ltd, drained tests.
tests were performed on the same soil (i.e. the Hostun S28 From the shear modulus measured at a specific shear
sand) and void ratio as in the centrifuge tests. Monotonic strain, the elastic shear modulus Gel was estimated using
undrained triaxial tests gave a final large-strain octahedral the equations predicting the degradation of shear modulus
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1087

15 15

Estimated ax at T2 - No mit [g]

Measured ax at T2 - No mit [g]

10 10

5 5

0 0

-5 -5

-10 -10

-15 -15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time [sec] Time [sec]
15 15

Estimated ax at T4 - No mit [g]

Measured ax at T4 - No mit [g]

10 10

5 5

0 0

-5 -5

-10 -10

-15 -15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time [sec] Time [sec]

15 15
Estimated ax at T2 - Anchors [g]
Measured ax at T2 - Anchors [g]

10 10

5 5

0 0

-5 -5

-10 -10

-15 -15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time [sec] Time [sec]
15 15
Estimated ax at T4 - Anchors [g]
Measured ax at T4 - Anchors [g]

10 10

5 5

0 0

-5 -5

-10 -10

-15 -15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time [sec] Time [sec]

Fig. 4. Centrifuge tests. Measured and estimated accelerations with time at locations T2 and T4 for the cases (a) without mitigation and (b) with anchors.

with shear strain and confining stress proposed by respectively. It should be noted that these values are
Ishibashi and Zhang [9]. The obtained Gel equals 33, 57 consistent with predictions of the relationship proposed by
and 74 MPa at octahedral stress of 50, 100 and 200 kPa, Hardin [10] (Ref. [11]).
 ARTICLE IN PRESS
1088 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

50 50
MEASURED 45 MEASURED
45 ESTIMATED
ESTIMATED 40

Sax at T4 - No mit [g]

40
Sax at T2 - No mit [g]

35 35
30 30
25 25
20 20
15 15
10 10
5 5
0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10
T [sec] T [sec]

50 50
MEASURED MEASURED
45 45
ESTIMATED ESTIMATED
Sax at T2 - Anchors [g]

Sax at T4 - Anchors [g]

40 40
35 35
30 30
25 25
20 20
15 15
10 10
5 5
0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10
T [sec] T [sec]

Fig. 5. Centrifuge tests. Measured and estimated response spectra at locations T2 and T4 for the cases (a) without mitigation and (b) with anchors.

Fig. 7 gives the shear stress, the volumetric change and The two configurations without and with anchors, given
the shear modulus versus shear strain that the proposed in Fig. 1, were studied. In the second configuration, as the
numerical model with the parameters of Table 2 predicts anchors are simulated in the centrifuge model using brass
when s0 ¼ 125 kPa. The computed response predicts (a) the rods, the properties of the structural elements are taken as
measured peak shear strength, (b) the measured soil those of brass, i.e. density ¼ 8.7 t/m3, Young modu-
softening (i.e. degradation of soil strength with further lus ¼ 9.1(1010) Pa and tensile yield ¼ 2(108) N. The friction
straining), (c) the measured dilation and (d) the decrease angle at the interface is taken (according to the construc-
of the ‘‘equivalent’’ shear modulus, with the applied tion of the anchors at the model), as the peak friction angle
shear strain, for sands, as predicted by the relationships of sand, or 401. The two parameters related to grout, which
of [9]. is not present in the model, were taken zero. As the ends of
A number of different meshes were designed and the rods were bolted, the anchors were simulated as fixed.
adopted in order to validate the results. In the results The anchors spacing in the direction not shown in Fig. 1
presented, the mesh had about 700 elements. The size of the was simulated by setting appropriate scaling rules.
finite elements is, appropriately, smaller than one tenth of
the critical wavelength which is expected to be developed 4.2. Results and comparison with measurements
during the seismic excitation. Viscous dampers are used at
the lateral boundaries of the mesh. This avoids undesirable In both cases without and with the anchors, the
reflections of the outgoing waves. numerical analysis was performed in two steps: First
As both (i) the shear modulus, and (ii) the equivalent gravity was applied and the static stresses were computed.
relationship of (a) the friction and (b) the dilation angles in Then, the motion of Fig. 2 was applied horizontally at
terms of the plastic shear strain depend on the level of the base of the grid. For both cases, Table 1 presents the
stress, the grid was divided in three sublayers, each having computed peak acceleration and final settlement at all
a different range of octahedral stresses, as shown in Fig. 8. the monitoring points. For both cases, Fig. 3 gives the
The model parameters of Table 2 were used. In addition, computed settlements with time at location S2 and Figs. 4
similarly to practice in numerical analysis, a Poisson ratio and 5 give the computed acceleration histories and
equal to 0.3 was used. Finally, Rayleigh damping acceleration response spectra, respectively, at locations
coefficient of 0.07 was used. The value is consistent with T2 and T4. In addition, Fig. 9 gives the computed
the above discussion. maximum horizontal accelerations (in g) and final settlements
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1089

300

Shear stress (kPa)

6 250
Abs[ax(T4)-ax(T2)] No mit [g]

200
5
150
4 100
3 50
0
2 0 5 10 15 20
Shear strain %
1
Aegion Centrifuge
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.02

Volumetric strain (%)

Time [sec] 0
-0.02
6 -0.04
Abs[ax(T4)-ax(T2)] Anchors [g]

-0.06
5
-0.08
4 -0.1
0 5 10 15 20
3 Shear strain (%)
2 Aegion Centrifuge

1 1

0.8
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.6
G/Gel

Time [sec]
0.4
Fig. 6. Centrifuge tests. The absolute value of the difference of measured
0.2
horizontal accelerations at locations T2 and T4 with time for the cases (a)
without mitigation and (b) with anchors. 0
0.01 0.10 1.00 10.00
Shear strain %
Table 2 Aegion Centrifuge
The values of the model parameters of the soil used for the numerical
studies Fig. 7. Predictions of the model with the proposed model parameters. The
shear stress, volumetric strain and ratio G/Gel versus shear strain are given.
(a) Centrifuge case The case of the sand used at the centrifuge tests at s0 ¼ 75 kPa (middle
fcs 301 layer) and the case of the bottom clay layer of the Aegion geometry at
b 23 s0 ¼ 325 KPa are presented.
sco 800 kPa
EP 220
gmob 103
b 0.1
Gel 33, 57 and 74 MPa at soct0 of 50, 100 and 200 kPa.
(b) Aegion
case
Layer Fill (depth 0–2 m below the surface of the slope) Clay

f 301 301 Fig. 8. The geometry, mesh and different soil materials used to model the
b 23 23 centrifuge tests.
sco 250 kPa 400 kPa
EP 220 220 in centrifuge tests with the numerically predicted values at
gmob 103 103 all monitoring points for both cases without and with
b 1.0 1.0
anchors. It can be observed that predictions of both peak
Vs Fig. 9a Fig. 9a
accelerations and final settlements generally agree with
measurements: In all 13 monitoring points the ratio of
(in mm) at the surface of the slope in terms of horizontal predicted to measured values is between 0.8 and 1.3.
distance from the edge. The measurements are also given in Fig. 3 compares the measured with the predicted
this figure. accumulation of settlement with time at point S1 for both
Centrifuge tests are ideal for evaluating numerical the case with and without mitigation. It can be observed
methods (Ref. [12]). Table 1 compares the values measured that similarly to the measured response, most of the
 ARTICLE IN PRESS
1090 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

20 ness of mitigating the acceleration near the edge does

18 not practically depend on the anchor friction angle (Sfric),
16 the cohesive strength of the grout (Sbond), the stiffness
14 of the grout (Kbond) and the anchor diameter. In addition,
12 it is not affected by the Young modulus (E) and the tensile
ax-max [g]

10 yield strength (Yield) of the anchor once a limit value

8 of these parameters is exceeded. Furthermore, it is affected
6 somewhat by anchor length, anchor inclination and
No mit - measured
No mit - estimated 4 the presence of anchor at the toe. In particular, the
Anchors - measured
2 effectiveness of mitigating the acceleration near the edge
Anchors-estimated
0 increases as the anchor length increases, the anchor
-0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 inclination is about 621 and there is an anchor near the
Horizontal Distance from the tip [m] edge.
0.00
5. Case study. Numerical analysis at the Aegion slope
-0.05
5.1. Geometry and soil conditions
[mm]

-0.10
The case study considers the eastern part of the slope of
-0.15 the city of Aegion in Greece. The Aegion slope is described
by Athanasopoulos et al. [2]. As a result of the 1995
U

No mit - measured
No mit - estimated -0.20 earthquake, many buildings near the eastern part of the
Anchors - measured slope were severely damaged and either were abandoned or
Anchors - estimated
-0.25 required major repairing and strengthening (Ref. [13]).
-0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 Topographic mapping of the region was performed by the
Horizontal Distance from the tip [m] Institute of Geology and Mineral Exploration (IGME) of
Fig. 9. Centrifuge tests. Computed maximum horizontal accelerations Greece. Fig. 10a gives the 2D simplified cross-section of the
(in g) and final settlements (in mm) at the surface of the slope in terms of region under consideration.
horizontal distance from the edge of the slope. The measurements are also A geotechnical investigation was also performed by
given. IGME. It included borings near the toe of the slope and at
the base of the slope. The water table was found at depth
computed residual settlement accumulates initially. From about 40 m below the surface of the slope and 5 m below
Fig. 4 it can be observed that all the predicted acceleration the base. The geotechnical profile below the surface of the
histories at locations T2 and T4 for both the case with and slope consists of soil that extends to a depth of 32 m. It is a
without mitigation are similar to those measured. Fig. 5 low-plasticity clay layer, except from a surface fill layer 3 m
compares the measured with the predicted spectral deep. Below the base of the slope, the clay layer extends to
response at locations T2 and T4 for both the case with a depth of 20 m.
and without mitigation. It can be observed that both the In the laboratory of Stamatopoulos and Associates Ltd
period and the magnitude of the peak spectral response are tests were performed on soil samples retrieved from the
similar. From all the above it is inferred that the numerical borings. Classification tests revealed a mean plasticity
analyses predicted reasonably well the values of accelera- index (PI) value of 7 for the top fill layer and a mean PI
tions and settlements measured in the centrifuge tests. This value of 10 for the clay layer below. Drained triaxial tests
illustrates that the proposed method of numerical analyses gave for the top fill layer peak strength parameters of
can predict both the seismic response of slopes, and the j ¼ 341, c ¼ 45 kPa and for the underlying low-plasticity
effect of anchors in mitigating the seismic response. clay layer j ¼ 351, c ¼ 30 kPa.
Geophysical tests have been performed in the region by
4.3. Parametric analyses previous investigators. They are reported by Athanaso-
poulos et al. and Bouckovalas et al. [2,14]. Based on the
Parametric analyses were performed in order to inves- above, Fig. 10a gives the profile of shear velocity, Vs, of the
tigate the effect of different characteristics and locations of area.
the anchors on the seismic response. One parameter was
varied every time from the reference run of Section 4.2. The 5.2. Soil parameters and numerical modeling
peak acceleration at locations T2 and T4 (Fig. 1) and final
settlement at locations S1 and S2 (Fig. 1) was recorded at The geometry applied at the numerical analyses is shown
every run. The results obtained are given in Table 3. It can in Fig. 10a. It extends horizontally 600 m from the tip, in
be observed that the final settlement does not change order to capture the 1D case level ground. A number of
considerably from case to case. In addition, the effective- different meshes were designed and used in order to
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1091

Table 3
Parametric analyses performed for the centrifuge case using different properties and configurations of anchors and partial results

Case ax-max (g) at ax-max (g) at Uy-final (mm) at S1 Uy-final (mm) at S2

T2 (Fig. 1) T4 (Fig. 1) (Fig. 1) (Fig. 1)

Reference case 9.48 9.34 0.10 0.23

Variation
Length (m) 0.12 9.40 9.10 0.11 0.21
Length (m) 0.10 9.50 9.20 0.11 0.21
Length (m) 0.08 9.48 9.34 0.10 0.23
Length (m) 0.06 9.50 9.59 0.10 0.19
Length (m) 0.04 8.97 9.80 0.10 0.21
Inclination 201 11.18 10.40 0.08 0.11
Inclination 261 10.49 11.60 0.09 0.16
Inclination 421 8.22 11.19 0.11 0.20
Inclination 621 9.48 9.34 0.10 0.23
Inclination 821 9.50 10.60 0.10 0.12
Diameter (m) 1.00E03 9.47 9.84 0.09 0.15
Diameter (m) 4.00E03 9.48 9.29 0.10 0.19
Diameter (m) 6.00E03 9.48 9.34 0.10 0.23
Diameter (m) 1.00E02 9.46 9.47 0.11 0.22
Diameter (m) 1.20E02 9.43 9.50 0.12 0.24
Diameter (m) 6.00E02 9.53 9.78 0.55 0.19
Delete anchor at tip 9.47 10.60 0.10 0.05
E (Pa) 3.00E+08 9.51 10.50 0.98 0.12
E (Pa) 3.00E+10 9.56 9.44 0.10 0.20
E (Pa) 9.10E+10 9.48 9.34 0.10 0.23
E (Pa) 9.00E+11 9.66 9.35 0.11 0.20
E (Pa) 3.00E+12 9.47 9.35 0.11 0.11
Yield (Pa) 0 9.40 10.70 0.10 0.09
Yield (Pa) 2.00E+08 9.48 9.34 0.10 0.23
Yield (Pa) 2.00E+10 9.48 9.30 0.10 0.23
Sbond (Pa) 0.00E+00 9.48 9.34 0.10 0.23
Sbond (Pa) 1.00E+03 9.48 9.34 0.10 0.23
Sbond (Pa) 1.00E+05 9.61 9.35 0.10 0.20
Sbond (Pa) 7.00E+03 9.48 9.34 0.10 0.23
Sfric 01 9.48 9.34 0.10 0.23
Sfric 151 9.48 9.34 0.10 0.23
Sfric 301 9.48 9.34 0.10 0.23
Sfric 451 9.48 9.34 0.10 0.23
Kbond (Pa) 0.00E+00 9.48 9.34 0.10 0.23
Kbond (Pa)1.00E+10 9.48 9.34 0.10 0.23
Kbond (Pa)1.00E+12 9.48 9.34 0.10 0.23
Kbond (Pa)2.00E+11 9.48 9.34 0.10 0.23

validate the results. In the presented configuration, the value of sco was obtained from the peak strength measured
mesh had about 2000 elements and was finer near the toe of in the triaxial drained tests.
the slope. The vertical size of the finite elements was less Fig. 7 gives the shear stress, the volumetric change and
than 2 m, appropriately smaller than one tenth of the the shear modulus versus shear strain that the proposed
critical wavelengths that are expected to be developed model with the parameters of Table 2 predicts. The case of
during the seismic excitations. Viscous dampers are used at the bottom clay layer of the Aegion geometry at
the lateral boundaries of the mesh. s0 ¼ 325 KPa is considered. The computed response pre-
The elastic moduli were taken according to the Vs values dicts (a) the measured peak shear strength, (b) the
of Fig. 10a. Based on the above discussion, Rayleigh measured soil softening, (c) the measured dilation and
damping with damping coefficient of 0.07 was used. (d) the decrease of the ‘‘equivalent’’ shear modulus, with
Finally, similarly to practice in numerical analysis, a the applied shear strain, in terms of the PI of the soil, as
Poisson ratio equal to 0.3 was used. predicted by the relationships of [9].
Based on the elastic and strength parameters above, Similarly to previous analyses of the Aegion slope
Table 2 gives the values of the parameters of the (Ref. [2]) the horizontal input motion for the calculations
constitutive model given by Eqs. (1) and (2) that were is based on the record obtained during the 1995 earthquake
selected for the numerical analyses. The values of the (x–x component) at the basement of the building of OTE
parameters fcs, b, Ep, gmob, b were as proposed by [6]. The (given e.g. by [14]). The base motion should be such that
 ARTICLE IN PRESS
1092 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

when propagated through the 2D model generates a effects can be neglected. Therefore, the base motion was
response at the OTE site similar to the one recorded obtained by 1D equivalent-linear deconvolution. The code
during the 1995 earthquake. As the OTE site is located Cyberquake [7] was used for this purpose. The deconvo-
about 200 m from the edge of the slope and the recording luted excitation which is used as input is shown in Fig. 11a.
was at the basement of the OTE building, it is assumed that Numerical analyses using another base motion were also
wave propagation is 1D and soil–structure interaction performed. The Friuli-San Rocco-Italy 15/9/76 earthquake

Fig. 10. Aegion case: (a) site topography and Vs profile used in the analyses, (b) mitigation configuration at the surface of the slope.

0.7
0.15
0.6
0.10 0.5
0.05
Sax [g]

0.4
ax [g]

0.00 0.3

-0.05 0.2

-0.10 0.1

-0.15 0.0
0 5 10 15 20 25 30 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Time [sec] T [sec]

0.5
0.15

0.10 0.4

0.05
0.3
Sax [g]
ax [g]

0.00
0.2
-0.05
0.1
-0.10

-0.15 0.0
0.0 5.0 10.0 15.0 20.0 0.0 0.2 0.4 0.6 0.8 1.0
Time [sec] T [sec]

Fig. 11. Input accelerations: (a) decomvoluted Aegion 1995 earthquake and (b) Fruili 1976 earthquake.
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1093

(y–y horizontal component), shown in Fig. 11b, was

1.0
applied. The Friuli earthquake was selected because its ax, No mit
ax, Anchors 0.9
dominant period is 0.14 s, typical for Europe, and ay, No mit
0.8
considerably less than the dominant period of the Aegion ay, Anchors
0.7
earthquake (0.42 s).
0.6

ax-max , ay-max [g]

0.5
5.3. Results—no mitigation case
0.4

For the no mitigation case, Figs. 12 and 13 (No mit case) 0.3
give the computed maximum horizontal and vertical 0.2
accelerations and final settlements at the surface of 0.1
the slope in terms of horizontal distance from the edge of 0.0
-300 -250 -200 -150 -100 -50 0
the slope for the Aegion and Fruili earthquakes. Figs. 14a
Horizontal Distance from the edge [m]
and b (No mit case) give the computed maximum
horizontal and maximum vertical acceleration with depth
1.0
and horizontal distance from the edge for both earth- ax, No mit
ax, Anchors 0.9
quakes. Figs. 15a,b and 16a,b give the computed horizontal
ay, No mit 0.8
and vertical acceleration spectra of the seismic motion for ay, Anchors
0.7

ax-max , ay-max [g]

the Aegion and Fruili earthquakes, respectively, at the
0.6
surface of the slope. The horizontal spectral acceleration is
0.5
given at horizontal distances 20, 50, 80, 110, 140, 200, 300 m
0.4
from the edge of the slope. The vertical acceleration is given
0.3
only at horizontal distances 10, 20, 50 and 80 m from the
0.2
edge of the slope. The reason is that at large horizontal
0.1
distances from the edge, vertical acceleration is very small.
0.0
-300 -250 -200 -150 -100 -50 0
5.4. Results—mitigation case Horizontal Distance from the edge [m]

From Figs. 12, 14–16 it can be observed that of all Fig. 12. Aegion slope: (a) Computed maximum horizontal and maximum
components considered, the topographic effect of the vertical accelerations at the surface: (a) for the Aegion earthquake and (b)
for the Fruili earthquake. Both the no mitigation and mitigation with
component of ax-max for the Aegion earthquake extends anchors cases are given.
more, approximately 200 m from the edge. Along all these
200 m there are buildings. It is inferred that mitigation
must cover all of this region. 0
Figs. 14a and b (No mit case) illustrate that for both -5
earthquakes near the edge (a) the large maximum -10
horizontal accelerations develop at depths smaller than -15
5 m and (b) the large maximum vertical accelerations
[mm]

-20
develop at depths smaller than about 10 m. The above -25
indicate that anchors of about 5 and 10 m length applied at -30
U

the surface of the slope may be an effective measure -35

mitigating the large horizontal and vertical accelerations at AIGEIO-No mit -40
the crest of the slope, respectively. AIGEIO-Anchors
-45
FRIULI-No mit
Anchors were simulated by the ‘‘anchor elements’’ that FRIULI-Anchors
-50
are included in the FLAC code, described in Section 3. -100 -90 -80 -70 -60 -50 -40 -30 -20 -10 0
More than 100 parametric analyses were performed. The Horizontal Distance from the edge [m]
parametric analyses illustrated that the response does not
change considerably when the diameter and material Fig. 13. Aegion slope: final settlements in terms of horizontal distance
from the edge of the slope under the Aegion and the Fruili earthquakes.
properties of the anchors vary within the range of possible Both the no mitigation and mitigation with anchors cases are given.
values (diameter between 0.05 and 0.4 m, bond friction
between 01 and 451, tensile yield strength between 106 and
1010 N, stiffness of the grout between 0 and 1010 Pa, accelerations, consistently with Fig. 14, anchor length
cohesive strength of the grout between 0 and 104 Pa, should be about 5 m. Anchors must be present at all the
Young modulus between 3  1010 and 5  1011 Pa). region that mitigation is needed. For best cost-effective
The parametric analyses also illustrated that the results, anchors must be parallel to each other, with an
response depends on the inclination, the length, the spacing inclination of 551 to the horizontal and about 5 m spacing.
and the number of the anchors. To decrease horizontal Anchors of the larger length of 8 m are also needed to
 ARTICLE IN PRESS
1094 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

0.80 0.80
-10m -10m
0.70 -20m 0.70 -20m
0.60 -50m

ax-max Anchors [g]

0.60
ax-max No mit [g]

-50m

0.50 -80m -80m

0.50
-140m -140m
0.40 0.40
-300m -300m
0.30 0.30
0.20 0.20
0.10 0.10
0.00 0.00
0 10 20 30 40 0 5 10 15 20 25 30 35 40
Depth [m] Depth [m]
0.30 0.30
-10m -10m
-20m -20m
0.25 0.25
-50m -50m

ay-max Anchors [g]

-80m -80m
0.20 0.20
ay-max No mit [g]

0.15 0.15

0.10 0.10

0.05 0.05

0.00 0.00
0 10 20 30 40 0 10 20 30 40
Depth [m] Depth [m]

1.00 0.90
- 10m -10m
0.80
- 20m -20m
0.80
0.70
ax-max No mit [g]

ax-max Anchors [g]

- 50m -50m
- 80m
0.60
0.60 -80m
-140m 0.50
-140m
0.40 - 300m 0.40
-300m
0.30
0.20 0.20
0.10
0.00
0.00
0 10 20 30 40
0 5 10 15 20 25 30 35 40
Depth [m]
Depth [m]
0.35 0.35
- 10m
-10m
0.30 - 20m 0.30 -20m
ay-max No mit [g]

ay-max Anchors [g]

0.25 - 50m
0.25 -50m
- 80m -80m
0.20 0.20
0.15 0.15
0.10 0.10
0.05 0.05
0.00 0.00
0 10 20 30 40 0 5 10 15 20 25 30 35 40
Depth [m]
Depth [m]

Fig. 14. Aegion slope: computed maximum acceleration versus depth from the surface and horizontal distance from the edge of the slope for both the no
mitigation and anchors cases under: (a) the Aegion and (b) the Friuli earthquakes.
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1095

reduce some acceleration peaks. Yet, the above mitigation maximum horizontal and maximum vertical acceleration
geometry does not decrease vertical accelerations. To with depth and horizontal distance from the edge, for both
decrease these accelerations, consistently with Fig. 14, earthquakes. In these reference runs, the anchor properties
anchors must have larger length. The optimum configura- used were Young modulus ¼ 2  1011 Pa, density ¼ 7 t/m3,
tion obtained reducing both the horizontal and vertical diameter ¼ 0.05 m, tensile yield strength ¼ 2  106 N, stiff-
accelerations is given in Fig. 10b. Along the outer 100 m, ness of the grout ¼ 7  106 Pa, cohesive strength of the
anchors have 15 m length, two inclinations: +551 and 551 grout ¼ 102 Pa. These anchor properties are typical of
and about 5 m spacing. At the remaining 100 m, anchors those of the anchors used in practice.
have 6 m length, inclination +551 and about 5 m spacing. Regarding the parametric analyses, Fig. 17 presents
Anchors are assumed every five meters in the direction not results for the case when the anchor Young modulus (E) is
shown in Fig. 10. varied. Maximum horizontal accelerations at the surface in
For the configuration of Fig. 10b, Figs. 12 and 13 terms of horizontal distance from the edge for the Aegion
(Anchors case) give the computed maximum horizontal 1995 earthquake are considered. A useful measure here is
and vertical accelerations and final settlements at the the normalized factor PA, defined as
surface of the slope in terms of horizontal distance from the
PA ¼ ½ðaxmax  axmax2D Þ=ðaxmaxref  axmax2D Þ, (3)
edge of the slope for the Aegion and Fruili earthquakes.
Figs. 15c,d and 16c,d give the computed horizontal and where ax-max is the maximum horizontal acceleration at
vertical acceleration spectra of the seismic motion for the given distance from the edge and E value, ax-max-2D is the
Aegion and Fruili earthquakes, respectively, at the surface corresponding maximum horizontal acceleration for the
of the slope. Fig. 14 (anchors case) gives the computed case without mitigation and ax-max-ref is the corresponding

3.00
-20 m
-50 m
2.50 -80 m
-110 m
2.00 -140 m
Sax - No mit [g]

-200 m
-300 m
1.50

1.00

0.50

0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20
T [sec]

1.00
-10 m
0.90 -20 m
0.80 -50 m
-80 m
0.70
Say - No mit [g]

0.60
0.50
0.40
0.30
0.20
0.10
0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20
T [sec]

Fig. 15. Aegion slope: Aegion earthquake. Computed response spectra at the surface: (a) horizontal acceleration—no mitigation, (b) vertical
acceleration—no mitigation, (c) horizontal acceleration—anchors, and (d) vertical acceleration—anchors.
 ARTICLE IN PRESS
1096 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

Fig. 15. (Continued)

maximum horizontal acceleration for the reference mitiga- earthquake with ‘‘prototype’’ values of peak acceleration
tion case (E ¼ 2  1011 Pa). The factor PA equals to one if and main period of 0.15 g and 1 s, respectively, was applied.
the results are identical to the reference case and is less than The centrifuge geometry, even at the prototype, has
one when the acceleration decreases less than the reference much smaller dimensions than most actual field cliff-type
case run. It can be observed that when the Young modulus topographies. It is necessary to verify the mitigation
is larger than 3  1010 Pa, the factor PA equals to about method of anchors by considering a typical cliff-type
one. topography. Therefore, the seismic response of the eastern
part of the Aegion slope was studied numerically. The
6. Discussion slope has an average inclination of about 301, height of
40 m and rock at a depth of 140 m below the surface of
6.1. General the slope. Two earthquakes were applied at the bedrock:
(a) the Aegion 1995 earthquake decomvoluted at the
Under strong earthquakes, there occurs amplification of bedrock with ax-max ¼ 0.15 g and a dominant period of
the horizontal acceleration and generation of the vertical 0.42 s and (b) the Fruili earthquake with ax-max ¼ 0.14 g
acceleration near the edge of hard or dense slopes. The and a dominant period of 0.14 s.
present work studied the effect of anchors to mitigate this Similarly to previous studies (Refs. [1,3]), the topo-
effect. Two geometries were studied: First, in the centrifuge graphic effect is described by considering the amplification
device a slope with 301 inclination and prototype height of of the maximum horizontal acceleration (ax-max), the
8 m was considered. The prototype length of the surface of maximum vertical acceleration (ay-max), the maximum
the slope was 15 m and soil extended to a depth of 8 m. An spectral horizontal acceleration (Sax-max) and the maximum
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1097

spectral vertical acceleration (Say-max) at the surface using no mitigation case and the reference case with anchors,
the dimensionless factors respectively.
The ME factors equal to 0 if the mitigation method has
TE  ax ¼ axmax2D =axmax1D , no effect and 1 if the mitigation method completely
TE  ay ¼ aymax2D =axmax1D , eliminates the topographic amplifications. They take values
TE  Sax ¼ Saxmax2D =Saxmax1D , greater than one when the mitigation measures reduce the
response near the edge to values less than far away from
TE  Say ¼ Saymax2D =Saxmax1D , ð4Þ
the edge.
where the subscript 1D indicates the response far away from Below, for the centrifuge geometry, the 1D position
the edge where the topographic effect does not exist, while the taken at location T2 (Fig. 1), while the 2D position is taken
subscript 2D indicates the response near the edge of the slope. at location T4. For the Aegion geometry, according to the
To quantify the effect of the mitigation measure, the previous discussion, the 1D position is taken at the surface,
following dimensionless factors are defined: 200 m from the edge of the slope (for the case without
mitigation), while the 2D case is taken as the maximum
ME  ax ¼ ðaxmax2D  axmaxanchors Þ=axmax2D  axmax1D Þ, value of the response attained 0 to 200 m from the edge of
ME  ay ¼ ðaymax2D  aymaxanchors Þ=ðaymax2D  aymax1D Þ, the slope.
ME  Sax ¼ ðSaxmax2D  Saxmaxanchors Þ=ðSaxmax2D  Saxmax1D Þ,
ME  Say ¼ ðSaymax2D  Saxmaxanchors Þ=ðSaymax2D  Saymax1D Þ; 6.2. The centrifuge geometry
ð5Þ
The measurements at the centrifuge geometry without
where the subscripts max-2D and max-anchors indicate the mitigation indicated that at the ‘‘prototype’’, when base
maximum values attained near the edge of the slope for the maximum acceleration is 0.15 g, maximum acceleration at

Fig. 16. Aegion slope: Fruili earthquake. Computed response spectra at the surface: (a) horizontal acceleration—no mitigation, (b) vertical acceleration—
no mitigation, (c) horizontal acceleration—anchors, and (d) vertical acceleration—anchors.
 ARTICLE IN PRESS
1098 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

Fig. 16. (Continued)

1.2 factors are in general agreement with the findings of

1 previous analyses (Refs. [1,3]). Furthermore, the perma-
Factor PA (equation (3))

0.8 nent settlements near the edge were very small (less than
0.6 1 cm, prototype scale).
0.4
The second centrifuge test indicated that anchors can
0.2
prove very effective in reducing the peak value of
- 20m
- 40m
acceleration near the tip of the slope and produce a more
0
- 60m uniform acceleration pattern in the slope: Based on the
-0.2 - 80m measurements of points T4 and T2 (a) the average of the
-0.4 - 100m
- 120m
absolute value of the difference of horizontal accelerations
-0.6 at the two locations over the 1654 increments of the
1.E+09 1.E+10 1.E+11 1.E+12
records, decreases from 1.0 g for the case without mitiga-
Young modulus of anchors [Pa]
tion to 0.6 g for the case with anchors (Fig. 6) and (b) the
Fig. 17. Aegion slope: Aegion earthquake. Partial results of parametric factor ME-ax equals 81% and the factor TE-Sax equals
analyses. The factor PA (Eq. (3)) illustrating the effect of the mitigation 63% (Table 4). In addition, permanent settlements were
method on the horizontal accelerations at the surface 20, 40, 60, 80, 100 again very small.
and 120 m from the edge of the slope in terms of the anchor Young
modulus.
Concluding, the centrifuge tests (a) verified that the
potential problem of the seismic response of dry slopes
consisting of dense soil with an inclination of about 301 is
not necessarily one of permanent deformations, but of
T2 (ax-ax-D) equals 0.20 g. As illustrated in Table 4, the amplification of accelerations near the edge of the slope
measurements at T4 indicate that the factor TE-ax equals and (b) illustrated that anchors can mitigate the seismic
1.5 and the factor TE-Sax equals 1.2. These amplification motion near the edge.
 ARTICLE IN PRESS
C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100 1099

Table 4
Topographic effect (TE factors defined by Eq. (4)) and effect of mitigation measures (ME factors defined by Eq. (5)) on the centrifuge geometry (based on
measurements) and the Aegion Geometry under the Aegion and Fruili earthquakes (based on numerical analysis)

Case 1Da 2Db no mit 2D with anchors TE-ax factor ME-ax factor (%)

(a) axmax (g)

Centrifuge geometry (prototype value) 10.1 (0.20) 15.3 (0.31) 11.1 (0.22) 1.51 81
Aegion geometry—Aegion earthquake 0.50 0.76 0.53 1.52 88
Aegion geometry—Fruili earthquake 0.45 0.85 0.60 1.89 62
(b) Saxmax (g)
1Da 2Db no mit 2D with anchors TE-Sax factor ME-Sax factor (%)

Centrifuge geometry (prototype value) 40 (0.80) 48 (0.96) 43 (0.86) 1.20 63

Aegion geometry—Aegion earthquake 2.0 2.74 2.20 1.37 73
Aegion geometry—Fruili earthquake 2.0 2.70 1.60 1.35 137
(c) aymax (g)
1Da 2Db no mit 2D with anchors TE-ay factor ME-ay factor (%)

Centrifuge geometry – – – – –
Aegion geometry—Aegion earthquake 0.10 0.35 0.22 0.70 52
Aegion geometry—Fruili earthquake 0.06 0.35 0.18 0.78 59
(d) Saymax (g)
1Da 2Db no mit 2D with anchors TE-Say factor ME-Say factor (%)

Centrifuge geometry – – – – –
Aegion geometry—Aegion earthquake 0.27 0.90 0.52 0.45 60
Aegion geometry—Fruili earthquake 0.25 1.05 0.62 0.53 54
a
For the centrifuge case at position T2, for the Aegion case 200 m from the edge for the case without mitigation.
b
For the centrifuge case at position T4, for the Aegion case maximum response 0–200 m from the edge.

6.3. The Aegion geometry 62% for the Aegion and Fruili earthquakes, respectively,
while the factor ME-ay equals 52% and 59%, respectively.
The Aegion geometry numerical analyses indicated that Regarding the spectral response, the factor ME-Sax
when base maximum acceleration is 0.15 and 0.14 g, equals 73% and 137%, while the factor ME-Say is 52%
maximum acceleration at the surface 200 m from the toe and 59% for the Aegion and Fruili earthquakes, respec-
(ax-max-1D) equals about 0.50 and 0.45 g for the Aegion and tively. Finally, permanent settlements did not change
Fruili earthquakes, respectively. As illustrated in Table 4, considerably.
the factor TE-ax equals 1.52 and 1.70, while the factor Concluding, the numerical analysis of the Aegion slope
TE-ay equals 0.70 and 0.78 for the Aegion and for the illustrated that anchors can mitigate considerably the
Fruili earthquakes, respectively. Regarding the spectral amplification of the horizontal acceleration and the vertical
response, the factor TE-Sax equals 1.37 and 1.35 while the acceleration near the edge. In both earthquakes, the
factor TE-ay equals 0.45 and 0.53 for the Aegion and Fruili mitigation decreased the topographic effect (a) in terms
earthquakes, respectively. These factors illustrate the of horizontal accelerations by 62% or more, and (b) in
importance of the topographic effect in (a) increasing the terms of vertical accelerations by 52% or more.
horizontal acceleration and (b) producing a considerable
vertical acceleration component. It should be noted that 6.4. Parametric numerical analyses and further work
these amplification factors are in general agreement with
the findings of previous analyses (Refs. [1,3]). Furthermore, Parametric analyses in both geometries illustrated that
the permanent settlements near the edge were small for the effectiveness of anchors does not change considerably
both earthquakes (less than 5 cm). when the diameter and material properties of the anchors
The numerical analyses illustrated that the seismic vary within the range of possible values but depends on the
motion throughout the 200 m near the edge where the length, inclination, location and number of anchors. More
seismic motion increases due to the 2D effect can be analyses are needed to illustrate that anchors can be
mitigated if an adequate number of anchors is used. As effective to mitigate the amplification of accelerations near
illustrated in Figs. 12, 14–16, the peak value of acceleration the edge of slopes under any slope and any earthquake.
near the edge of the slope is reduced and a more uni- Validation of the method at more sites and seismic
form acceleration pattern is produced: In particular, excitations is needed prior to applying the method at new
according to Table 4, the factor ME-ax equals 88% and sites without previous analyses.
 ARTICLE IN PRESS
1100 C.A. Stamatopoulos et al. / Soil Dynamics and Earthquake Engineering 27 (2007) 1082–1100

7. Conclusions References

Concentration of damage of buildings near the edge of [1] Assimaki D, Kausel E, Gazetas G. Wave propagation and
hard or dense cliff-type topographies has been observed soil–structure interaction on a cliff crest during the 1999 Athens
Earthquake. Soil Dyn Earthquake Eng 2005;25:513–27.
during a number of recent earthquakes. These observations [2] Athanasopoulos GA, Pelekis PC, Leonidou EA. Effects of surface
have been interpreted by numerical dynamic analyses that topography on seismic ground response in the Egion (Greece) 15 June
illustrate the amplification of the horizontal acceleration 1995 earthquake. Soil Dyn Earthquake Eng 1999;18:135–49.
and the generation of vertical acceleration near the edge of [3] Bouckovalas GD, Papadimitriou AG. Numerical evaluation of slope
slopes. The paper studied the effect of mitigating these topography effects on seismic ground motion. Soil Dyn Earthquake
Eng 2005;August–October(7–10):547–58.
accelerations using anchors both experimentally and [4] Zeng X, Schofield AN. Design and performance of an equivalent
numerically. The main issue is that different regions of shear beam container for earthquake centrifuge modeling. Géotech-
the slope have different acceleration histories. Therefore, if nique 1996;46(1):83–102.
the part of the slope in which topographic amplification [5] Aubry D, Benzenati I, Modaressi A. In: Arulanandan K, Scott RF,
occurs is connected to that in which the acceleration is less, editors. Numerical predictions for model no. 1, verification of
numerical procedures for the analysis of liquefaction problems.
then the accelerations have to become more uniform.
Rotterdam: Balkema; 1993. p. 45–54.
The experimental study involved dynamic centrifuge tests [6] Modaressi A, Lopez-Caballero F. Global methodology for soil
at the Cambridge University facility both with and without behavior idenfication and its application to the study of site effects.
anchors. The numerical procedure was verified by the seismic Proceedings: fourth international conference of recent advances in
response of the centrifuge tests. It was then applied to the geotechnical earthquake engineering, San Diego, CA, 2001, March
(CD-ROM).
study of the effect of anchors of a typical field case, the
[7] BRGM (French Geological Survey) Software. Cyberquake, version
Aegion slope, under two different input motions. In all cases 1.1. User’s guide, Franc. 1998.
anchors were found effective in mitigating the seismic motion [8] ITASCA Consultants. FLAC (Fast Lagrangian analysis of continua),
near the edge: the maximum acceleration and maximum S.A., version 4.0, 2002.
spectal acceleration approaches the levels of the 1D case far [9] Ishbashi I, Zhang X. Unified dynamic shear moduli and damping
away from the edge of the slope, by about 55% or more. ratios of sand and clay. Soils Found 1993;33(1):182–91.
[10] Hardin BO. The nature of stress–strain behavior of soils. In:
Parametric analyses in both geometries illustrated that Proceedings, conference on earthquake engineering and soil dy-
the effectiveness of anchors does not change considerably namics, vol. 1. Pasadena, USA: ASCE; 1978. p. 3–90.
when the diameter and material properties of the anchors [11] Stamatopoulos C, Bassanou M, Madabhushi G, Brennan AJ.
vary within the range of possible values but depends on the Evaluation of an elastoplastic method for simulating the seismic
length, inclination, location and number of anchors. response of slopes using centrifuge test data. The 11th international
conference of IAMAG, the international association of computer
methods and advances in geomechanics, Torino, Italy, June 2005,
Acknowledgements 8pp.
[12] Arulanandan K, Scott RF, editors. Verification of numerical
The work was funded by the European Commission, procedures for the analysis of soil liquefaction problems. In:
Direction General XII for Science, Research and Devel- Proceedings of the international conference on the verification of
numerical procedures for the analysis of soil liquefaction problems,
opment (Project no G1RD-CT-2002-00702). Geotechnical
Davis, CA, USA. Rotterdam: Balkema; 1993. 1800pp.
and topographic data used for the Aegion site is based on [13] Stamatopoulos C, Bassanou M, Poyadji E, Kynigalaki M. Elasto-
investigations performed by the Institute of Geology and plastic dynamic analysis of the Aegion slope during the 1995
Mineral Exploration of Greece. Mr Hamelin and Mr Evers earthquake. The 11th international conference of IAMAG, the
of Soletance-Bachy Co and Mr Aris Stamatopoulos of international association of computer methods and advances in
geomechanics, Torino, Italy, June 2005, 8pp.
Stamatopoulos and Associates Co. assisted in designing the
[14] Bouckovalas GB, Gazetas G, Papadimitriou AG. Geotechnical
mitigation measures. Prof. Pitilakis of Aristotle University aspects of the 1995 Aegion, Greece, earthquake. Second international
of Thessaloniki assisted in the selection of accelerograms conference on earthquake geotechnical engineering, Seco e Pinto.
and provided the histories. Rotterdam: Balkema; 1999.