Soil Dynamics and Earthquake Engineering 69 (2015) 28–45

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Stability analysis of a slope subject to real accelerograms by finite

elements. Application to San Pedro cliff at the Alhambra in Granada
Antonio Morales-Esteban a,n, José Luis de Justo a, J. Reyes b, J. Miguel Azañón c,
Percy Durand a, Francisco Martínez-Álvarez d
a
Department of Building Structures and Geotechnical Engineering, University of Seville, Spain
b
TGT-NT2 Labs, Chile
c
Department of Geodynamics, University of Granada, Spain
d
Department of Computer Science. Pablo de Olavide University of Seville, Spain

art ic l e i nf o a b s t r a c t

Article history: The dynamic stability analysis of slopes is often conducted by the traditional method of slices, using
Received 31 January 2013 pseudo-static calculations. However, the response of a geotechnical structure subjected to seismic loads
Received in revised form can be studied through a dynamic finite element analysis, which can be considered one of the most
19 July 2013
complete available tools, as information about the stress distribution and the deformations can be
Accepted 26 October 2014
obtained. The dynamic analysis of the stability of San Pedro cliff at the Alhambra in Granada is studied in
this paper. The results have been compared with pseudo-static calculations worked out with the method
Keywords: of slices. Real accelerograms have been selected for the dynamic tests. Thorough in situ and laboratory
Dynamic calculation tests have been conducted in order to properly characterize the cliff. The soil constitutive model is also
Stability of slopes
explained in this paper. Finally, the influence of the sources of energy dissipation has been studied
Finite element
through the material damping, the integration scheme and the boundary conditions.
Accelerogram
& 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Hanging towns, built at the edge of a cliff and subjected to
instability of slopes, are one of the main problems for the
conservation of historic sites. An example of this problem is San
Pedro cliff, a dihedral 65.5 m high, which has progressed to place
itself at a distance of 23.8 m from the Alhambra palace-wall, a
World Heritage site (Fig. 1. View of San Pedro Cliff cutting the
Alhambra hill; at the foot of the hill runs Darro River; to the right
the scar of the 1985 slab fall).
Active normal faults (Fig. 2. Outline of the site showing the
river meander, the main cracks at La Alhambra palace and faults at
the hill slope) surround the cliff and have created an extensional
tectonic regime that loosens the ground and actives the fall of
slabs. A main fault matches the western face of the faces of the
dihedral. Pseudo-static stability analysis suggest that the Factor of
Safety (hereinafter, FOS) of this slope under 1000-year return
period earthquake loading may drop below 1.0 and the critical slip
surface could penetrate the Alhambra walls [17]. A restoration
project based on a high-yield-stress wire mesh, post-tensioned by
anchors and coloured to blend with the cliff is proposed. The mesh
n
Corresponding author. Tel.: þ 34 616360273; fax: þ 34 954541007.
E-mail address: ame@us.es (A. Morales-Esteban).
applies a pressure on the slope. This might be the only acceptable
solution and its visual impact is moderate.
A dynamic analysis of the stability of San Pedro cliff has been
carried out with Plaxis [27]. Accelerograms have been selected
with the method proposed by Morales-Esteban et al. [22]. The
calculations have been repeated for every accelerogram. The
information about the geotechnical properties has been obtained
from the site investigation conducted by Justo et al. [17].
2. The safety of slopes during earthquakes
This paper refers to ground which does not lose an appreciable
part of its strength due to the build-up of pore pressures induced
by the vibrations.
Many stability analyses still use pseudo-static calculations,
where horizontal forces are used instead of accelerograms. The
dynamic calculation by means of the finite element method
(hereinafter FEM) is accurate, versatile and requires fewer assump-
tions, especially the failure mechanism [11]. Moreover, it can be
considered one of the most complete tools to estimate the seismic
response of a geotechnical system, as the information about the
stress distribution and the deformations/displacements can be
obtained. Slope failure occurs naturally through the zones in

http://dx.doi.org/10.1016/j.soildyn.2014.10.023
0267-7261/& 2014 Elsevier Ltd. All rights reserved.
 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45
which the shear strength is insufficient to resist the shear stresses.
However, it requires an appropriate constitutive model of the soil,
a complete description of the soil characteristics and a proper
definition of the seismic data.
From the paper written by Ambraseys [2] the dynamic earth-
quake calculations were based upon the assumption that the
nonlinear hysteretic stress–strain response of soils could be
determined by an equivalent linear–elastic method of analysis,
based upon a damped linear–elastic model if the properties of that
model were satisfactorily chosen. This model will be called the
equivalent linear model.
It has been recognized by earthquake engineers long ago that
the usual concept of a FOS on shear strength does not properly
asses the behaviour of a slope during strong earthquakes. The FOS
is the factor by which the strength should be reduced so as to
bring the slope to a state of limiting equilibrium with the stresses
along a failure surface. If the FOS is less than one, a section of the
slope will slide along the failure surface. This state cannot be
permitted under static conditions, as the stresses will exist until
large displacements change the geometry of the structure. But
under dynamic conditions it may be possible to allow the FOS to
drop momentarily below one, as this state will exist only for a
short time, and the slope might come to rest again at a time when
the new stresses do not exceed the available strength. The
performance should rather be measured in terms of the relative
displacement that the sliding mass may undergo during the
earthquake.
Fig. 1. View of San Pedro Cliff.
Fig. 2. Outline of the site.
29
Since the final analysis with strain-compatible soil properties in
the equivalent linear model is purely elastic, the permanent
deformation caused by earthquake shaking cannot be computed
by this type of analysis. However, the stresses derived from these
strains are assumed representative of stresses in the ground and
the accelerations are also assumed to be reasonably representative
of field values [10]. Seismic slope deformation models are used to
make predictions of earthquake-induced permanent displace-
ments in natural slopes and constructed earth systems.
The first model was the rigid-block model established by
Newmark [25]. This model was extended to sliding along an
inclined plane by Sarma [31], which first examined the stability
of a rigid block resting on a plane surface. In this paper the
calculations will be limited to horizontal accelerations and zero
pore pressures. The angle of the acceleration has little influence on
the final result [31]. In Fig. 3 (Model of a rigid block on a sloping
surface), K is the seismic coefficient.
The driving shear force along the sloping plane is as follows:
D ¼ Wð sin β þK cos βÞ ð1Þ
If the contact is cohesionless, the resisting force is as follows:
0
R ¼ Wð cos β  K sin β Þ tan Φ ð2Þ
The critical acceleration is defined as the acceleration that
applied on the block will produce a state of limiting equilibrium.
Equating expressions (1) and (2) and operating:
tan ∅0 cos β  sin β
Kc ¼ ¼ tan ð∅0  βÞ ð3Þ
tan ∅0 sin β þ cos β
When the driving force exceeds the resisting force, the accel-
eration of the block relative to the plane surface is as follows:
2
Wd x
¼ DR ð4Þ
g dt 2
Fig. 3. Model of a rigid block on a sloping.
30 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45

Eliminating D and R by Eqs. (1) and (2), introducing Eq. (3) and For 0; 725 r KKmc r1
operating, Sarma [31] transformed Eq. (4) into:  

4xm cos∅0 K c =K m þ α  π þ cos 2 α=2 cotðα=2Þ

 ¼ ð9Þ
2
d x cos β  ∅0 K m gT 2 cos ðβ ∅0 Þ π2
¼ g ðK  K c Þ ð5Þ
dt 2 cos∅0
For 0 r KKmc r 0; 725
The solution of the above equation depends upon the variation The solution for earthquake records (Fig. 4d, earthquake
of K with time. record) is obtained by assuming piece-wise linear acceleration.
For a rectangular pulse of duration T/2 and seismic coefficient In this case the allowed movement is downhill only.
Km (Fig. 4a. Rectangular pulse), solution of Eq. (5) gives: Fig. 5 (Variation of dimensionless displacement with Kc/Km.
    Courtesy of the Institution of Civil Engineers) shows the relation-
4xm cos∅0 1 Km h i
¼  1 ð6Þ cos∅0
K m gT 2 cos ðβ  ∅0 Þ ship between K4xgT and Kc/Km for the four cases of Fig. 4.
m
2 Kc 2 cos ðβ  ∅0 Þ
m

For earthquake records, T is the predominant period obtained from

where xm is the total displacement
their acceleration spectra, and Km is the maximum acceleration.
For a triangular pulse of duration T/2, seismic coefficient Km and
Several records are chosen, including some obtained from nuclear
λ ¼1/2 (Fig. 4b, triangular pulse), the solution is as follows: blasts. As can be seen from the graph, the centre of gravity of
 




4xm cos∅0 4 1  K c =K m 1  ð1=2ÞðK c =K m Þ  1 þ 1=2 ðK c =K m Þ2 points for earthquake records corresponding to higher values of
¼

K m gT 2 cos ðβ  ∅0 Þ 24 K c =K m Kc/Km (4 0.4) is well represented by the half sine pulse and for low
ð7Þ to medium values of Kc/Km (0.1–0.4) by the triangular pulse. The

pffiffiffiffiffiffi rectangular pulse is an upper bound in all cases, and the triangular
1 1=2
For 0 r KKmc r pulse is an upper bound for Kc/Km Z 0.5.
2

 0

1 
K =K 3 2  2√1=2 
1=2 In a slope it is first necessary to select the critical slide surface
4xm cos∅ c m
¼ ð8Þ that will give the minimum critical acceleration, Kcg, in a pseudo-
K m gT 2 cos ðβ  ∅0 Þ 6
static calculation. Sarma [31] stated that a very simple way of

pffiffiffiffiffiffi
1  1=2 finding the equivalent angle of the sloping plane is to sum the
for r KKmc r1
2 shear forces along the failure surface vectorially. The direction of
For a half sine pulse of duration T/2 and seismic coefficient Km the resultant shear force vector will give the angle β. The average
(Fig. 4c, half sine pulse), the solution is as follows: Φ0 of the sloping plane-rigid block contact may be obtained from
 

 2 Eq. (3).
4xm cos∅0 K c =K m  sin q Usually, the acceleration input is applied at the base of the
¼

K m gT 2 cos ðβ  ∅0 Þ 2π 2 K c =K m slope, but the maximum acceleration in Eqs. (6)–(9) and Fig. 5

Fig. 4. Pulses and earthquake record.

 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45 31

magnitude of the displacement is evaluated by a double integra-

tion procedure. The direction of motion for a potential sliding mass
once yielding occurs was assumed to be along a horizontal plane.
This mode of deformation is not uncommon for embankments
subject to strong earthquakes, but is uncommon for slopes.
They present a simplified procedure for estimating the max-
imum crest acceleration and the natural period, T0, of an embank-
ment subjected to a given base motion [21]. The authors define
“maximum acceleration ratio” as the ratio of the maximum
average acceleration, kmax, to the maximum crest acceleration,
ümax. They produced a graph (Fig. 6. Variation of maximum
acceleration ratio with depth of sliding mass. Courtesy of the
ASCE) relating this ratio to the relative depth of the sliding mass.
This graph is obtained with several earthquakes acting on earth
dams and would allow obtaining the maximum average accelera-
tion from the maximum crest settlement for a given sliding
surface.
They present a graph relating the relative shear modulus
G/Gmax or the damping ratio with shear strain, similar to the graph
used by Plaxis. The maximum shear modulus, Gmax, is related to
the shear wave velocity, vmax, by the following relationship:

Gmax ¼ ρv2max ð11Þ

For the first iteration of computations, they assume any initial

value of shear modulus, G, and determine the ratio G/Gmax. From
the graph indicated above, the corresponding values of shear
strain and damping could then be determined.
Using several earthquake records they obtain one graph relating
u
kmax gT 0
with ky/kmax for different values of magnitude. This graph
would allow estimating the final displacement, u.
The dynamic module of Plaxis [6] uses the hardening soil
Fig. 5. Variation of dimesionless displacement. model with small-strain stiffness (hereinafter HS-small). This
model takes into account very small-strain stiffness and its non-
corresponds to the average mass inside the sliding surface (sliding linear dependency with strain amplitude. The HS-small is based
mass). Ambraseys and Sarma [3] divide this mass into horizontal
layers and obtain, at each instant, the force acting on the strip,
multiplying the mass of each horizontal strip by its acceleration
(obtained by the equivalent linear model) at the instant consid-
ered. Summing up the forces, the simultaneous force may be
found. For a given earthquake, by repeating this process at
different instants, the maximum value of the force, F can be found.
The maximum simultaneous seismic coefficient can be
obtained from the following expression:
F
Km ¼ ð10Þ
mg
where m is the mass of the sliding body.
For an earth dam and for a visco-elastic body this is an easy
process, because the acceleration at any level is given by a
theoretical expression [3]. For a slope, the maximum simultaneous
seismic coefficient must be obtained by a finite element
calculation.
Makdisi and Seed [20] propose a method that starts determin-
ing the sliding surface and the critical (yield) seismic coefficient,
ky, with the cyclic yield strength, as in Sarma's method. Earth-
quake induced accelerations in the slope are determined using
dynamic response analyses. Finite element procedures using the
equivalent linear model can be used for calculating time histories
of acceleration, or simpler one-dimensional techniques (shear slice
calculations) might be used for the same purpose. Integrating
along the sliding surface they obtain the average time history of
acceleration.
For a given potential sliding mass, when the induced accelera-
tion exceeds the calculated critical acceleration, movements are
assumed to occur along the direction of the failure plane and the Fig. 6. Variation of max acc ratio.
32 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45
upon the hardening soil model, with only two additional
parameters:
(1) The initial or very small-strain shear modulus, Gmax.
(2) The shear strain level at which the secant shear modulus, G,
is reduced to about 70% of Gmax.
Gmax
G¼   ð12Þ
1 þ aγ =γ 0:7 
where γ is the threshold shear strain. Using a ¼0.385 and γ ¼ γ0.7
gives G E0.722Gmax.
The module gives as output, the stresses, strains and accelera-
tions at any point and the permanent displacements, as this model
is not elastic, but elasto-plastic.
The dynamic response of a finite element model is conditioned
by the setting of several parameters that influence on the sources
of energy dissipation in time-domain analysis. The amount of
damping shown by a discrete numerical system is determined by
the material damping, the integration scheme of the equations and
the boundary conditions [34].
3. Background
This Section describes the evolution and the causes that have
modelled San Pedro cliff. First, a brief description of the geography,
the geology and the tectonics of the site is provided. Second, the
topography is described, followed by a seismicity analysis of the
Granada basin. Later, the origin, evolution and causes of the
present situation are discussed, followed by the results of the
stability analysis conducted by means of static and pseudo-static
calculations. Then, the important environmental and aesthetical
issues that surround San Pedro Cliff are described. Finally, a
comparative analysis between the proposed solutions and the
design solution is shown.
3.1. Geography, geology and tectonics
The Alhambra palace is located on the top of a red hill that
dominates the Granada basin (Fig. 1). San Pedro cliff cuts some
dense conglomeratic levels that constitute the Alhambra forma-
tion. The Alhambra conglomerates correspond to alluvial fans
coming from the erosion of Sierra Nevada, dated upper Pliocene-
lower Pleistocene [17].
Several normal active faults, with strike NW–SE are located in
the surroundings of La Alhambra (Fig. 2). The right face of the
dihedral (Fig. 1) is a fault-line scarp, as the cliff orientation here is
controlled by faults of this family. An extension tectonic regime
has created these faults.
3.2. Topography
The Eurocode 8, part 5 [9] recommends a specific study of the
stability of slopes whose slope is higher than 151. The stability can
be verified by means of pseudo-static calculations where the
topography and the stratigraphy have no many irregularities.
Fig. 7 (topographical map of San Pedro cliff) shows the topogra-
phical map of San Pedro cliff. The topography is dominated by the
Darro River and the Alhambra formation, between them, San
Pedro cliff has been formed. It can be observed that the construc-
tion of San Pedro church at the meander of the Darro River, placed
at the bottom of the cliff, might have pushed the Darro River
against San Pedro cliff. The cliff is almost vertical at some parts and
the height difference between San Pedro church and the Alhambra
wall is 70 m. Due to its topographical irregularities a specific study
must be done for San Pedro cliff. In this study, pseudo static and
dynamic calculations have been worked out.
3.3. Seismicity
[22] calculated that the annual rate of earthquakes by km2, in
the Granada Basin, is 1.34E-03, the highest in the Iberian Penin-
sula. There is evidence that important earthquakes occurred in
1431, 1526, 1806, 1911 and 1956 [4].
The Gutenberg–Richter law [12,13] relates the cumulative (or
absolute) number of events N(M) with magnitude greater or equal
to M with the seismic activity, a, and the size distribution factor, b.
A high b-value involves that the number of earthquakes of small
magnitude is predominant. Contrarily, a low value shows a smaller
difference between the relative number of small and large events.
The coefficient b usually takes a value around 1.0 and has been
widely used by researchers [33].
log 10 NðM Þ ¼ a  bM ð13Þ
To solve Eq. (13) the maximum likelihood method has been
applied Aki [1] and Utsu [32]:
log ðeÞ
b¼ ð14Þ
M  M0
where M is the mean magnitude and M0 is the cut-off
magnitude [29]. A value of 1.41 has been obtained for the
Granada Basin.
The earthquakes in the Granada Basin are frequent although,
normally, of moderate magnitude (Mr6).
3.4. Origin and evolution
San Pedro cliff has been created by the floods of Darro River, the
tectonics, erosion, earthquakes, and, maybe, seepage coming from
the palace. The extensive tectonic regime has created fractures
that have favoured the erosion generated by the River Darro
during floods, modelling a convex curve of the bed through the
cliff (Fig. 2).
Slides are documented since 1524 when a fire burned the
vegetation, leaving the hill unprotected. From Hüfnagel's engrav-
ing of 1564, it can be deduced that the dihedral was located at
60 m from the Alhambra walls and the total height was 33 m. The
cited causes of slides are: spills, soil softening produced by water
coming from Santa Ana's aqueduct, explosions, and hillside ero-
sion by seepage coming from the Alhambra [17].
3.5. Present situation
At present, a series of active faults with throws around 60 cm,
cross San Pedro cliff. One of these faults, whose strike is N1581,
constitutes the western face of the dihedral. The throw of the
eastern face is 7 m.
The extension neotectonics regime produces a reduction of the
horizontal stresses in the cliff, which may reach zero value. The
joints open in the cliff are now a preferential path for seepage
coming from the palace.
Azañón et al. [4] point out that the Alhambra has remained
standing during the past six centuries and that it has a good state
of preservation from a seismic point of view. Three reasons are
given:
1. Stability resulting from the mechanical behaviour of the con-
glomeratic formation on which the Alhambra is founded.
2. The moderate magnitude of the seismicity in the Granada Basin.
3. The lack of important seismic active faults in the neighbour-
hood of the Alhambra.
Justo et al. [16] have estimated that the average backward
displacement of the wedge might be 8 cm/year. Although it might
 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45 33

Fig. 7. Topographical mal.

take many years to reach the wall, history shows that the speed of
this process is not constant and an important earthquake might
accelerate it. Moreover, pseudo-static stability analyses suggest
that the critical slip surface could penetrate the Alhambra walls
(Fig. 8. Critical slip surface under pseudo-static conditions using
the Morgenstern and Price method for a pressure of 20 kPa).

3.6. Stability analysis

Justo et al. [17] have worked out static and pseudo-static

calculations of the stability of the slope. The program geo-slope
and the Morgenstern and Price method, with different pressures
applied to the hill, were used. The FOS, for the static calculation, is
1.35–1.42 and rises up to 1.48–1.61 for an applied pressure on the
mesh of 20 kPa. If no pressure is applied on the slope, the FOS
under pseudo-static conditions drops below one, and for a
pressure of 20 kPa the FOS is 1.12. Fig. 8 shows the critical slip
surface calculated with the geo-slope program. It can be observed
that the slip surface could penetrate inside the Alhambra wall
placed on the border of the upper part of the cliff. Fig. 8. Geoslope.
34 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45

3.7. Environmental and aesthetical issues of the project in relation to the aesthetic and cultural values of San
Pedro Cliff can be considered moderate.
San Pedro cliff is placed on the Alhambra hill, where the
Alhambra and Generalife rise. In front of it the Albaicín hill stands,
where the popular district of the same name is situated
4. Fundamentals
(Figs. 1 and 2). Both hills have been declared World Heritage sites
by UNESCO. Darro River runs between them, and the walk by the
4.1. Choice of the slope stability analysis method
border of the river, known as Carrera del Darro, is one of the most
romantic strolls of the world [16]. For these reasons, the solution
Traditional methods of slope stability analysis use limit equili-
should cause minimum visual impact and should keep the cultural
brium approaches involving methods of slices that have been
and aesthetic values of the surrounding unaltered.
progressing for years. Here the Morgenstern and Price [23] that
fulfils all equilibrium conditions have been used.
3.8. Proposed solutions
Pseudo-static calculations substitute real accelerograms, where
the value of acceleration changes instantly, for a static horizontal
The danger that implies the progress of the wedge for the Alhambra
force. Also the dynamic properties of a soil, for example Young's
wall has been foreseen long ago, and from 1520, the following
modulus, differ from static to dynamic conditions.
solutions have been proposed and in some cases executed [16]:
Jing [15] presents a very detailed work on techniques, advances
and outstanding issues in numerical modelling for rocks. Three main
1. Embankments or walls at the foot of the cliff provided to
groups can be distinguished: (a) continuum methods, (a1) finite
protect the cliff against the floods of the River. Currently, the
difference method (FDM), (a2) FEM and (a3) boundary element
problem of the floods is not that important, so the construction
method (BEM); (b) discontinuum methods, (b1) discrete element
of embankments is neither necessary nor convenient.
method (DEM) and discrete fracture network method (DFN); (c)
2. Forbid watering the Alhambra forest. This solution would
hybrid continuum/discontinuum models, (c1) hybrid FEM/BEM,
wither the vegetation.
hybrid DEM/BEM and (c3) other hybrid models. Jing [15] points out
3. To divert the river. This solution, unnecessary today, would have a
that there are no absolute advantages of one method over another.
tremendous environmental impact and would be very expensive.
The continuum criteria referred here is a macroscopic concept.
4. A reinforced earth wall and a double twisted steel wire mesh
Continuity implies that all material points cannot be torn open or
anchored at the head. This solution would cause a great visual
broken into pieces [15]. All material points should remain in the
impact on the slope.
same neighbourhood throughout the deformation. A microscopic
5. An ecological wall combined with Californian drains, slope
scale is neither necessary in practice nor mathematically advisable.
sewing, reinforcement micropiles and acrylic treatment of the
The FEM is perhaps the most widely applied numerical method
slope surface to avoid the erosion [30]. This solution would
in rock mechanics and rock engineering today. Compared with
modify considerably the current scenery and the visual impact
traditional methods, it presents some advantages: the failure
would be unacceptable.
surface occurs through the zones where the applied shear stresses
6. Grouting through a series of steel tubes sub parallel to the slope
exceed the soil shear strength, the failure occurs naturally and no
joined to River regulation. The medium coefficient of perme-
assumptions about its shape and location are previously made;
ability of the slope (2  10  7 m/s) makes the grouting of the
global equilibrium is kept until failure is reached, no slices are
conglomerate difficult. Even more, the grouting pressure might
supposed and no hypothesis about slide side forces is needed; the
lose slabs from the slope.
FEM provides information about deformations at working stress
levels; and progressive failure can be monitored up to and
including overall shear failure. The FEM is accurate, versatile and
3.9. Design solution
flexible in handling material heterogeneity, non-linearity and
boundary conditions. Finally, in a dynamic analysis the informa-
The solution should be effective and should have minimum
tion about the stress distribution and the deformation/displace-
environmental impact. The cliff is a unique geological anomaly
ments versus time can be obtained.
that should be preserved. To raise the FOS up to 1.0 under dynamic
The in situ tests [17], that will be described later in Section 5.2,
conditions and to counteract extensional stress of the cliff, a
show that the layers of the Alhambra conglomerate can be
solution consisting of a high-yield-stress wire mesh, post-
modelled as a continuum material. For all the aforementioned,
tensioned by anchors and coloured to blend with the cliff is
the continuum FEM has been used in this study.
proposed. High-yield-stress wire meshes are rhomboidal meshes
of galvanized steel lying directly on the slope. The yield stress is
1770–2020 MPa and the wire thickness 3–4 mm. The pressure on
the slope (10–30 kPa) is applied by post-tensioned anchorages 5. Finite element modelling
isolated or reinforced by 16–22 mm thick horizontal cables.
Anchored wire meshes provide a non-expensive method for slope A HS-small under drained conditions has been chosen, because
protection and stabilization. The mesh and the anchors should raise the Alhambra conglomerate is a granular material that may be
the FOS of the slope. Moreover, the mesh should avoid the erosion of qualified as a soft rock with a low degree of saturation, so that no
the conglomerate and should provide acceptable visual impact, important positive pore pressures will be set-up.
especially if the present vegetation is maintained. The basic element used in this analysis is the triangular
The structural elements can be coloured in ochre and brown to element of 15 nodes. It provides a fourth order interpolation for
blend with the tonalities of the Cliff. The mesh, unlike conventional displacement and the numerical integration involves 12 Gauss
galvanized steel meshes, does not reflect the light on the zinc coating. points (stress points). It is a very accurate element that produces
Moreover, the mesh adheres to the original profile of the slope. The high quality stress results. The mesh density is fine.
mesh thickness is only 3 mm thick and it is very difficult to perceive The average element size (AES) represents the average side
it. [16], analyzed the fields of vision from where the restoration is length of the element used and is an average dimension, which is
visible and concluded that, once the vegetation has grown, the impact representative of the mesh thickness.
 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45 35

Every time a numerical analysis is calculated, the influence of
the mesh must be checked. Kuhlmeyer and Lysmer [18] suggested
that the element size should be no larger than λ/8, where λ is the
wave length with the maximum frequency of interest, fr. In this
study, λ/8 ¼Vs/8fr ¼12.5 m, where Vs ¼800 m/s and fr ¼8 Hz. In
this analysis the AES is 6.56 m.
The ground anchors are modelled as a combination of a node-to-
node anchor and a geogrid. The geogrid simulates the grout body
(with a length of 3.1 m) whereas the node-to-node anchor simulates
the anchor rod. In reality there is a complex three-dimensional state of
stress around the grout body. Node-to-node anchors are two node
elastic spring elements with constant spring stiffness (normal stiff-
ness) that are used to model ties between two points. This element
can be subjected only to tensile forces for anchors. Node-to-node
anchors can be prestressed in a calculation phase. Geogrids are
composed of geogrid elements (line elements) with two translational
degrees of freedom (ux, uy). When 15-node soil elements are employed
then each geogrid element is defined by five nodes.
Tables 1 and 2 quantify the properties of the layers that form
the Alhambra conglomerate used in the dynamic analysis. Fig. 8
shows the layers of San Pedro Cliff with the mesh and anchors
proposed. Fig. 9 (discretization in finite elements and points used
for the dynamic analysis) plots the points used in the analysis and
the finite element discretization, automatically done by the
program, for a fine global coarseness.
5.1. Calculation parameters
The commercial program Plaxis [27] has been used for the
dynamic finite element calculations. As mentioned before, three

Table 1
Average properties for the layers of the Alhambra conglomerate. f.d. ¼ free drainage; est. ¼estimated value; pl* ¼ net pressuremeter limit pressure; EM ¼deformation
modulus; Ep ¼ pressuremeter modulus; Vp ¼ ongitudinal wave velocity; Vs ¼ transverse wave velocity; Ed ¼ dynamic modulus; sci ¼ uniaxial compressive strength.

Layer k (m/s) Classification tests Strength tests

o 0.08 mm (%) wL IP USCS sci (kPa) Direct shear test

c´ (kPa) F´

7
1 3  10 12.6 22 6 GC-GM
2 9  10  6 22.0 24 6 GC-GM
3 2  10  7 to f.d. 27.8 22 2 SM
4 6  10  10 to f.d. 35.2 21 7 SC-SM GM
4a 7.7  10  7 79.4 28 11 CL 394 34.5 33.41

Layer From plate loading test Pressuremeter tests Geophysical

Loading Unloading Ep (MPa) p1n (kPa) F´ est. (1) Vp (m/s) Vs (m/s) Ed (MPa)
EM (MPa)

1 397 1772 60 2700 38 1500 800 3700

2 601 2683 40 2000 960 5600
3 397 1772 33 2300 37 1500 800 3700
4 891 3976 115 7400 44 2400 1150 8300
4a 109 485 42 4500 41 1500 800 3700

Table 2
Calculation values. (1) Dense conglomerate; (2) very dense conglomerate; (3) moderately dense conglomerate; and (4) conglomerate with gravel.

Layer g (kN/m3) Eref (kPa) n Cref (kPa) F (1) Y (1) H (m) ρ (kg/m3) Gmax (kPa) G (kPa)

5
1 22 3.97  10 0.30 96 38 8 25 2242 1,435,270 1,036,296
2 22 6.01  105 0.35 92 40 10 7 2242 2,066,789 1,492,266
3 21 3.97  105 0.30 98 37 7 3 2140 1,370,031 989,192
4 23 8.91  105 0.35 84 44 14 95 2344 3,100,663 2,238,746
5 20 2.00  104 0.30 15 38 8 - 2038 – –

Fig. 9. Discretization and points.

36 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45

requirements must be fulfilled in order to conduct a realistic

dynamic analysis: first, a thorough soil characterization of the
geotechnical structure must be performed; second, the soil con-
stitutive model must be duly implemented; third, the seismic
input data selected must be introduced into the calculations.
Moreover, the sources of energy dissipation will be discussed.
Finally, the calculation hypotheses will be listed.

5.2. Ground characterization and water table

Ten boreholes were drilled in the upper, medium and lower

parts of the cliff, to a maximum depth of 45 m. In two of the
boreholes drilled at the top of the escarpment, pressuremeter,
down-hole, cross-hole, penetration, permeability and laboratory
tests were carried out [17]. Tables 1 and 2 present the properties of
the layers that appear from top to bottom in the geological profile:

1. Dense conglomerate, with 100 mm maximum particle size and

core recovery 100%, of brown to pale grey silty matrix.
2. Very dense conglomerate, with maximum particle size of
5–8 cm and core recovery 100%, of brown to reddish silty to
clayey matrix.
3. Moderately dense conglomerate, with core recovery 60%, of
brown to pale grey silty matrix.
4. Very dense, gravelly and sandy conglomerate, of brown to pale
grey silty fine matrix and variable permeability.
(4a) One meter thick clay layers interspersed in layer 4.Core
recovery 100%.
Talus appears at the foot of the slope. It is composed of
quartzose and phyllitic blocks, gravel and sand, with a predomi-
nance of the sand fraction.
With the exception of a 1-m layer of clayey silt, it was not
possible to take undisturbed samples in the conglomerate, owing
to the large size of the stones and the low cohesion. Owing to that
the variation of shear modulus and damping ratio with shear
strain were those given by Plaxis 2D as default for the HS-small.
The angles of internal friction summarized in Tables 1 and 2, were
obtained from correlations with the net pressuremeter limit pressure
[7] and the effective cohesions by the method indicated by Justo et al.
[17]. Gmax and G have been obtained from Eqs. 11 and 12, respectively.
Disturbed samples of material from the cliff have been
obtained. The standard proctor compaction test has given the
following average results:
wo ¼ 8.7%, ρmax ¼ 2.20 Mg/m3
The material was dynamically compacted in a 300 mm 
300 mm shear box with the optimum water content and up to
2.05 Mg/m3 dry density. Later on, it was statically compacted,
under a stress of 180 kPa, to reach 2.15 Mg/m3 dry density. Fig. 10
(shear stress and vertical displacement versus horizontal relative
displacement) indicates that there is no degradation of the
material after reaching the maximum shear stress. The results
give an angle of internal friction of 411, very near to the average
(401) of the angles of internal friction in Tables 1 and 2.
The water table is placed well below the base of the cliff, at the
level of the surface of River Darro, and has no influence in the
stability of slopes calculations.
The well-documented slab failure of 1985 indicates that slides
usually occur on one of the dihedral sides and are not three-
dimensional. Although there is clast orientation in the direction of
the fault, the granular nature of the conglomerate guarantees that
there is no great decrease in this direction, as would happen in
clay, shale or slate.
Boore et al. [5] classify the local geology according to the
average velocity of the shear waves measured or estimated at a
depth of 30 m. In the Alhambra conglomerate the lowest
Fig. 10. Shear stress and vertical displacement.
transverse wave velocity measured is 800 m/s for layers 1, 3 and
4a (Table 1) and corresponds to rock (Vs Z750 m/s).
5.3. Ground constitutive model
The dynamic modulus of Plaxis [6] uses the HS-small. As stated
above, in this ground the positive pore pressures are negligible so
that actually the analysis is carried out in total stresses. This
elasto-plastic model takes into account very small-strain stiffness
and its non-linear dependency with strain amplitude. The HS-
small is based upon the hardening soil model, with only two
additional parameters:
1. The initial or very small-strain shear modulus, Gmax.
2. The shear strain level at which the secant shear modulus, G, is
reduced to about 70% of Gmax.
The module gives as output, the stresses, strains and accelera-
tions at any point and the permanent displacements.
As all Plaxis models, this model uses the Mohr–Coulomb failure
envelope (see Fig. 11. Mohr–Coulomb failure surface in principal
stress space).
5.4. Seismic input
The method described in [22] has been used in order to select
accelerograms. This method is based on constructing a uniform
seismic hazard acceleration response spectrum (USHARS), accord-
ing to the type of soil and the required hazard level (exposure time
and exceedance probability) at the site the structure is located.
Then, the standard deviation between the response spectrum of
real accelerograms and the calculated response spectrum is
calculated by means of Eq. (12) in [22]. The five accelerograms
whose standard deviation is lower, and in which SMC format is
available for both axis, have been selected. Table 3 summarizes the
accelerograms used in the dynamic calculations and its standard
deviation (σ). The selected accelerograms were recorded in sta-
tions placed on rock (Vs Z750 m/s). Their response spectrum is
equivalent to a 5% probability of exceedance and a time of
exposure of 50 years, which is equivalent to a return period of
 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45 37

Table 3
Information about the seismic data of records whose standard deviation (σ) is
minor in relation to the uniform seismic hazard acceleration response spectra for
San Pedro cliff.

Earthquake record 128

Earthquake Friuli (northern Italy)
Date 15/09/1976
Magnitude (Mw) 6.0
Station Robic (Slovenia)
Fault distance (km) 19
σ 0.016
SAmax (m/s2) 3.96
Duration (s) 9.80

Earthquake record 361

Earthquake Umbria (centre of Italy)
Date 19/04/1984
Magnitude (Mw) 5.6
Station Nocera Umbra (Italy)
Fault distance (km) 19
Fig. 11. Mohr Coulomb.
σ 0.028
SAmax (m/s2) 7.49
Duration (s) 5.09
974 years. The applied ground motion corresponding to the Lazzio
Abruzzo aftershock (record 990, 07/05/1984) recorded at the
Earthquake record 365
Atina-Pretura Terrazza station in Italy is plot in Fig. 12 (Lazio
Earthquake Lazio Abruzzo (southern Italy)
Abruzzo aftershock, record 990, x-axis acceleration time-history). Date 07/05/1984
Magnitude (Mw) 5.9
Station Atina (Italy)
5.5. Sources of energy dissipation
Fault distance (km) 11
σ 0.130
Visone et al. [34] perform an analysis of the calibration of two- SAmax (m/s2) 3.83
dimensional finite elements models in geotechnical earthquake Duration (s) 9.93
engineering through the comparison between the one-dim-
ensional vertical propagation of S-waves in elastic layers, whose Earthquake record 990
theoretical solutions are available in literature, with a finite Earthquake Lazio Abruzzo aftershock (southern Italy)
element analysis. They conclude that in time domain seismic Date 07/05/1984
Magnitude (Mw) 5.5
analyses the response of FEM is conditioned by the sources of Station Atina-Pretura Terrazza (Italy)
energy dissipation and give some preliminary advances on how to Fault distance (km) 13
calibrate these parameters. σ 0.011
SAmax (m/s2) 6.11
Duration (s) 10.23
5.5.1. Material damping
In time domain analysis material damping includes viscous and
Earthquake record 5826
hysteretic soil damping and is frequency dependent. The material Earthquake Strofades (Jonic Sea)
damping is simulated by means of the well-known Rayleigh Date 18/11/1997
formulation [28]. The damping matrix, C, is assumed to be Magnitude (Mw) 6.6
proportional to the mass matrix, M, and to the stiffness matrix, Station Kyparrisia-Agriculture Bank (Greece)
Fault distance (km) 65
K, through the coefficients αR and βR according to the following
σ 0.129
relation: SAmax (m/s2) 2.76
Duration (s) 14.80
C ¼ αR M þ β R K ð15Þ
With respect to frequency, the dynamic response of a system is
strongly related to the value of these parameters. It should be process. Implicit and explicit methods are used. The advantage of
noted that this formulation is not the best to simulate damping the explicit integration is that is relatively simple to formulate.
but it is commonly used in numerical codes as it is straightforward However, the disadvantages are that the calculation process is not
to implement. Some studies have been performed in order to robust and several limitations are imposed in the calculation steps.
establish the damping of a layer of soil and to determine Rayleigh's Implicit methods are more complicated but produce a more
coefficients [26,34]. reliable calculation process and normally more accurate. In this
Park and Hashash [26] deeply studied the formulation of the study, Newmark's implicit time integration system is used. With
soil damping. For three columns of 100 m, 500 m and 1000 m, this method, the displacement and the velocity at the point in time
representative of the Mississippi Bay, they used two damping tþ Δt are expressed, respectively as follows:
ratios, one constant (1.8%) and another decreasing with depth.
  
Visone et al. [34] used a 2% constant damping ratio for a visco- 1 t þ Δt
ut þ Δt ¼ ut þ u_ t Δt þ  αN u€ þ αN u€ Δt 2
t
elastic homogenous layer that lies on rigid bedrock; following this ð16Þ
2
author, a 2% constant damping, D, has been used in this study.

 t

t þ Δt
5.5.2. Numerical damping (integration scheme) u_ t þ Δt ¼ u_ t þ 1  β N u€ þ β N u€ Δt ð17Þ
The formulation of the integration with time constitutes an
important factor in the stability and accuracy of the calculation where Δt is the time step.
38 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45

Fig. 12. Accelerogram 990  .

Fig. 13. Time x-acceleration.

The coefficients αN and βN control the accuracy of the numer- equal to 0.05 ms in order to respect the following rule on the
ical time integration. These coefficients should not be confused critical time step for a single mesh element [6]:
with the Rayleigh coefficients. In order to obtain a stable solution,
l
the following condition must be considered: Δt crit ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
e
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð20Þ
 2 h i
α ρð1 þEðν1Þð1υÞ 2νÞ 1 þ 4SB 2  B2S 1 þ 1 42ν 2S
4 2
1 1
βN Z0:5; αN Z þ βN ð18Þ B 2

4 2
Hilber et al. [14] proposed a change in the Newmark scheme,
named the Newmark HHT modification. The new Newmark para-
meters are now expressed as a function of the parameter γ, that is
a numerical dissipation parameter:

2
1þγ 1 c15 ¼4.9479  10  3 [35]. B and S are, respectively, the largest
αN ¼ ; βN ¼ þ γ ð19Þ
4 2
where γ varies between 0 and 1/3.
If γ ¼0, the coefficients coincide with those of the original
Newmark method with constant average acceleration. When γ 40,
the efficiency of the calculation is improved, but a numerical
source of damping is introduced into the model.
Despite the advantages of the implicit integration, it is sub-
jected to some limitations due to the time step used in the
calculation. If the time step is too large the solution will display
substantial deviation and the calculated response will be unreli-
able. The calculations have been done considering a time step, Δt,
The critical time step depends on the maximum frequency and
the coarseness of the finite element mesh. The parameter le
denotes the average length of an element. The factor α depends
on the element type. For a 15-node element α ¼ 1/19(c15)0.5, with
dimension and the surface area of the finite element.
Plaxis allows the input of accelerograms by means of a file with
ASCII or SMC format. For that purpose a displacement as a function
of time is applied to the basement, obtained from the integration
of the accelerograms through Newmark's integration method.
Fig. 13 (Horizontal acceleration versus time for a point placed at
the top of the cliff, subject to accelerogram 990) shows the
horizontal acceleration of a point at the top of the cliff and subject
to accelerogram 990 at the base of the discretization. Accelero-
gram 990 has been previously transformed into displacements by
means of the Newmark´s integration scheme.
 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45 39

This method has the difficulty of setting the appropriate equivalent layer over rigid bedrock has been used to obtain the
coefficients for the actual damping of the numerical model. Visone Rayleigh coefficients. An average shear wave velocity, Vsm, of the
et al. [34] suggest that a possible solution to limit this uncertainty soil profile is defined as follows:
is to set the minimum value for Newmark, γ, which allows N
1
stability, then fit the theoretical solution. To avoid introducing V sm ¼ ∑ hi V i ð24Þ
numerical dissipation in the numerical damping the following ∑N
i¼1 h i i ¼ 1

value has been used: γ ¼0 (αN ¼0.25 and βN ¼0.5). The fundamental period, T, of the equivalent soil profile is
In order to obtain the Rayleigh coefficients, the method calculated as follows:
proposed by Visone et al. [34] has been followed. They compare
N hi
the vertical one-dimensional propagation of shear waves in visco-
T ¼4 ∑ ð25Þ
elastic homogenous layer that lies on rigid bedrock, whose i ¼ 1 Vi
theoretical solution is known with the results obtained from the
In Table 4, the two first natural frequencies of the equivalent
FEM. They conclude that if no numerical dissipation is introduced
layer are listed. From the amplification function, plot in Fig. 14
in the time integration scheme, the damping of the system can be
(amplification function of the equivalent visco-elastic layer over
modelled by using the Rayleigh coefficients only. The following
bedrock) the natural frequencies of the seven first natural fre-
formulation in provided by Visone et al. [34], for a soil layer with a
quencies can be obtained. The Rayleigh coefficients shown in
constant damping ratio, D, to obtain αR and βR:
Table 4 have been used in all layers.
αR þ βR ω2ni ¼ 2ωni D ð21Þ

where ωni are the circular natural frequencies of the layer: 5.5.3. Boundary conditions
ωni ¼ 2π f ni ð22Þ In order to avoid the effects due to the reflection of waves on
the boundaries, absorbent boundaries have been defined. A
And the natural frequencies of the system can be obtained as damper is used instead of applying fixities in a certain direction.
follows: The damper ensures that an increase in stress on the boundary is
V si absorbed without rebounding.
f ni ¼ ð2n  1Þ ð23Þ The use of absorbent boundaries is based on the method
4hi
described by Lysmer and Kuhlmeyer [19]. The normal and shear
In which n is order number of the calculated natural frequency. stress components absorbed by a damper in x-direction are as
However, in this case study an additional difficulty appears as follows:
there are four layers superimposed on rigid bedrock. To solve
this limitation the EERA [8] code has been used and a unique σ n ¼  C 1 ρV p u_ x ð26Þ

Table 4
Raileigh coefficients and the two first natural frequencies for the equivalent San Pedro Cliff layer considering a 2% constant damping ratio.

Layer H (m) γ (kN/m3) Vs (m/s) ω1 (rad/s) f1 (Hz) ω2 (rad/s) f2 (Hz) αR βR

1 25 22 800
2 7 22 1150
3 3 21 800
4 95 23 960
Equivalent 130 22.70 935.76 11.30 1.79 33.92 5.39 0.3392 88.44  10  5

Fig. 14. AR frequency.

40 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45
τ ¼  C 2 ρV s u_ y ð27Þ
Where, ρ is the density, Vp and Vs, are the pressure wave and the
shear wave velocities, respectively. C1 and C2 are relaxation
coefficients that have been inserted in order to improve the effect
of the absorption. To the authors' knowledge there is not a well
established criterion in the literature to determine the value of the
relaxation coefficients. Brinkgreve [6] suggested the use of C1 ¼ 1
and C2 ¼0.25. However, it is not possible to state that shear waves
are fully absorbed so that in the presence of shear waves a limited
boundary effect is noticeable. To solve this issue, lateral boundaries
are placed sufficiently far away from the central zone. In this study,
the discretized zone has been enlarged to thrice the width used in
static or pseudo static calculations (compare Fig. 8 with Fig. 9).
5.5.4. Calculation hypotheses
The following calculation hypotheses have been considered:
The in situ stresses are modelled. Plastic calculations:
1. Own weight of the cliff.
2. Own weightþ mesh and anchors. It is considered that the mesh
and the anchors are installed.
3. Own weightþ mesh and anchors þpost-tension (125 kN). It is
considered that the mesh and a post-tension of 125 kN in the
anchors are applied.
4. Own weightþ mesh and anchors þpost-tension (250 kN).
5. Own weightþ mesh and anchors þpost-tension (375 kN).
6. Own weightþ mesh and anchors þpost-tension (448 kN).
7. FOS own weight. The FOS of the cliff is worked out. It is subjected
to the in situ stresses caused by its own weight. Static calculations.
The FOS of the cliff is worked out. It is subjected to the in situ
stresses caused by its own weight and the seismic acceleration given
by Spanish standard (hereinafter, SAS). Pseudo-static calculations:
8. FOS own weight þSAS.
9. FOS own weightþ mesh and anchors þSAS. It is considered
that the mesh and the anchors are installed.
10. FOS own weightþmesh and anchors þSASþpost-tension
(125 kN). It is considered that the mesh and post-tension of
125 kN in the anchors are applied.
11. FOS own weightþmesh and anchors þSASþpost-tension
(250 kN).
12. FOS own weightþmesh and anchors þSASþpost-tension
(375 kN).
13. FOS own weightþmesh and anchors þSASþpost-tension
(448 kN).
The in situ stresses are modelled. Later, the dynamic calcula-
tions caused by the USHARS are worked out:
14. Own weightþUSHARS accelerogram.
15. Own weightþ mesh and anchors þUSHARS accelerogram. It is
considered that the mesh and the anchors are installed.
16. Own weightþmesh and anchors þpost-tension (125 kN) þ
USHARS accelerogram. It is considered that the mesh and a
post-tension of 125 kN in the anchors are applied.
17. Own weight þmesh and anchorsþ post-tension (250 kN) þ
USHARS accelerogram.
18. Own weight þmesh and anchors þpost-tension (375 kN) þ
USHARS accelerogram.
19. Own weightþ mesh and anchors þpost-tension (448 kN) þ
USHARS accelerogram.
The anchors are separated 5 m. In the plane strain calculations,
the four anchor loads correspond to 25, 50, 75 and 89.6 kN/m.
In all cases the accelerograms were applied after tensioning the
anchors.
6. Results of numerical simulations using FEM
The most relevant results obtained from all the hypotheses
calculated are shown in this Section. First, the pseudo-static
results are exposed. Second, the stress is analyzed by means of
the relative shear stresses. Third, the plastic points and tension
cut-off are shown. Finally, an analysis of the deformation/max-
imum displacements is conducted.
6.1. Pseudo-static results
The FOS of San Pedro cliff at the Alhambra in Granada obtained
with a static calculation, using PLAXIS is 1.34. With the high-yield-
stress wire mesh and the post-tensioned anchors the FOS raises up
to 1.48 for p ¼20 kPa. For the pseudo-static calculations, the
ground acceleration applied to the slope corresponding to the
1000-year return period, following the recommendations of the
Spanish standard for monuments, is 0.28 g. Without the mesh and
the anchors the slope fails under an acceleration a ¼0.19 g; with a
pressure p ¼20 kPa, the FOS raises up to 1.12. The anchor force
corresponding to p¼ 20 kPa was 250 kN.
6.2. Dynamic results
6.2.1. Analysis of stress
Relative shear stresses indicate the proximity of stress to the
failure envelope. The relative shear stress, τrel, is defined as
follows:
τ τ ð28Þ
rel ¼
τmax
where τ is the maximum value of shear stress and τmax is the
shear strength.
Unless otherwise stated, all the graphs shown below corre-
spond to the following conditions:
1. When post-tension is employed, the anchor load is 89.6 kN/m.
2. The true dynamic calculations are for accelerogram 990, and
correspond to the instant with higher stresses or deformations.
The comparison between Fig. 15 (relative shear stresses of San
Pedro cliff, in a static calculation; no anchors, no mesh) and Fig. 16
(relative shear stresses of San Pedro cliff, with mesh and post-
tensioned anchors in a static calculation) shows that the area
where the relative shear stresses exceed 0.90 is fully covered by
the anchors. This shows the good distribution of anchors of the
anchors. It can also be observed that no slip surface is formed and
that if fall of slabs would happen it would only be superficial.
Fig. 17 (relative shear stresses of San Pedro cliff, with mesh and
post-tensioned anchors; dynamic calculation) plots the instant,
where the relative shear stresses are higher for the 5 accelerograms
calculated. The FE method allows analyzing at every instant a
geotechnical system subject to seismic loads. In this case, all the
images have been analyzed and the most relevant one is shown.
Although a slip surface is hinted, it does not reach the top of the
cliff and the anchors saw the area with higher relative shear
stresses. In Figs. 15–17, the higher relative shear stresses are
concentrated in the most superficial area of the slope.
The FE calculations also allow checking the anchors stress
variation versus time in order to verify that none of them exceeds
the allowable stress (100 kN/m).
A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45 41

Fig. 15. RSS no mesh-static.

Fig. 16. RSS mesh anchors-static.

Fig. 17. RSS mesh anchors-dynamic.

42 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45
Fig. 18. Anchors stress.
Fig. 19. PP no mesh-static.
Fig. 18 (Variation of the anchors' stress during the dynamic
calculation) shows the variation of the anchors' load with time. It
can be observed that the original post-tension is 89.6 kN/m and
that the anchors do not exceed the allowable stress (100 kN/m). It
should be noted that the variation of the anchors' load is less than
3% of its initial load.
6.2.2. Analysis of the Mohr–Coulomb points and tension cut-off points
Mohr–Coulomb points are the stress points that lie on the
surface of the Mohr–Coulomb failure envelope. Tension cut-off
points are the points where the tension cut-off criterion has been
reached (in this case for zero tension). The comparison between
Fig. 19 (Mohr–Coulomb points and tension cut-off points at San
Pedro cliff, in a static calculation; no anchors, no mesh) and Fig. 20
(Mohr–Coulomb points and tension cut-off points in San Pedro
cliff, with mesh and post-tensioned anchor load, in a static
calculation) shows how pressure applied to the slope reduces
the Mohr–Coulomb points for the static conditions. In Fig. 21
(Mohr–Coulomb points and tension cut-off points of San Pedro
cliff, with mesh and post-tensioned anchor load, in a dynamic
calculation), the Mohr–Coulomb points are represented for the
image of the five calculated accelerograms where the number of
them is higher. Similarly to Fig. 17, it can be observed that an
incipient slip surface is formed but not completed, and only
superficial fall of slabs might be expected.
6.2.3. Analysis of deformation
The FEM allows the user to observe the deformation versus time of
a geotechnical structure subjected to accelerograms. Fig. 22 (deformed
mesh of San Pedro cliff, with post-tensioned anchors and mesh, in a
dynamic calculation) shows the deformed mesh of the cliff in the
instant corresponding to Figs. 17 and 21, that is, when the relative shear
stresses are higher and when the number of Mohr–Coulomb points is
greater. It can be observed that the deformations in the vertical
direction are moderate. Fig. 23 (time history of horizontal displace-
ments at points A–C in dynamic calculation) plots the horizontal
displacements of points A–C (Fig. 9), point B is placed at the base,
under the application of accelerogram 990 to the base of the structure
for 20 s. It is important to study the displacements of the points in
relation to a point in the base in order to obtain the deformation as the
difference between the point considered and the point in the base. If
the dynamic analysis is worked out long enough, the damping
eliminates the movement produced by the accelerograms. Then,
permanent displacements can be measured. In this case, the perma-
nent horizontal displacement is 24 mm for point A and 27 mm for
point C. When p¼0 the permanent displacement at point C is 30 mm.
7. The use of simple methods for evaluating earthquake
induced deformations
The authors have tried to apply the simple methods included in
Section 2 for evaluating earthquake induced deformations. The
 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45 43

Fig. 20. PP mesh anchors-static.

Fig. 21. PP mesh anchors-dynamic.

Fig. 22. Deformed mesh.

Sarma and Makdisi and Seed methods are specially prepared for
homogeneous earth dams, where the accelerations at any height
are obtained by the well known equations based upon the
equivalent linear model. Moreover, the graph prepared by Makdisi
and Seed [20] (Fig. 6 in his paper) to obtain the deformation
depends upon the magnitude and the method presented in this
paper is a probabilistic method that takes into account all possible
magnitudes. Sarma's method does not make any reference to
magnitude, and, in principle, would be more adaptable to this
paper. In the case of a slope, as indicated by Makdisi and Seed,
finite element procedures using strain dependent soil properties
can be used for calculating time histories of acceleration above the
sliding body; but the FEM used in this paper, at the same time
provides the final induced deformation, so that there is no point in
using it only to find the time histories of acceleration and not the
final displacement.
In any case, the authors have tried to apply Sarma's method
(Table 5). The average critical seismic coefficients obtained from
the pseudo-static calculations are included in column (2). The
average earthquake induced seismic coefficients in the sliding
body have been obtained with this FEM and are placed in column
(3). It can be observed that they are smaller than the critical
44 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45
Fig. 23. X displacement ABC acc 990.
Table 5
Coefficients for Sarma's method.
(1) Pressure on (2) Critical seismic Peak seismic coefficient, K
the slope (kPa) coefficient, Kc
(3) Average (4) At the
from FEM, Km crest, Kpt
p¼0 0.165 0.104 0.432
p ¼ 20 0.27 0.124 0.366
accelerations. This is because the acceleration used in the pseudo-
static calculations (0.28 g) has been obtained according to the
Spanish standard [24] and the accelerograms obtained from the
probabilistic method have a smaller peak acceleration (0.15 g).
Column (4) includes the peak seismic coefficients at the top
obtained with the FEM presented in this paper. Column
(5) includes the average earthquake induced seismic coefficients
in the sliding body obtained from the data in column (4) using
Fig. 6; they are larger than the ones of column (3). The ratio Kc/Km
obtained with these values is larger than 1 for p ¼20 kPa and so
the displacement predicted by Sarma's method would be zero. For
p ¼0 the ratio is 0.86. The angles β obtained from the pseudo-
static calculations have been introduced in column (7) and the
average Φ0 of the sloping plane-rigid block contact, obtained from
Eq. (3) in column (8). The angle β–Φ0 is negative and this will be
the sign of coefficient C in the ordinates of Fig. 5. This shows the
Sarma's method cannot be applied to a slope so steep and with
important values of cohesion.
8. Conclusions
San Pedro cliff at the Alhambra in Granada is relevant due to
the aesthetical and cultural values that surround it. A historical
analysis has shown that, in the 16th century, it was situated at
60 m from the Alhambra walls and that its total height was 33 m.
However, nowadays it has approached up to 23.8 m and has grown
up to 65.5 m height. A solution is necessary to preserve the
Alhambra hill, the Alhambra walls and the Cubo Tower, placed
on the top of it. A high-yield-stress wire mesh, post-tensioned by
anchors and coloured is proposed. It may be one of the fewest
possible solutions to protect the slope and to raise the FOS with a
minimum environmental impact.
(6) Kc/Km (7) β (8) Φ0 from (9) β  Φ0
Eq. (3) from Fig. 5
(5) Average
from Fig. 6, Km
0.192 0.86 441 53.41  9.41
0.162 41 441 53.41  9.41
The analysis of the stability of the slope using the Morgenstern
and Price method has shown that the FOS for static conditions is
1.35–1.42, which rises up to 1.48–1.61 if a pressure of 20 kPa is
applied to the slope. The pseudo-static calculations have demon-
strated that the cliff could be unstable if the SAS is applied.
However, a pressure of 20 kPa would raise the FOS up to 1.12.
These calculations have shown that the slip surface could pene-
trate inside the Alhambra wall.
An analysis of the stability of the cliff has been conducted with
the FEM. The static calculations suggest that the FOS is 1.34. With
the mesh and the post-tensioned anchors it raises up to 1.55,
similarly to the results obtained with the Morgenstern and Price
method. The pseudo-static calculations under the SAS have
demonstrated that the FOS would only be 1.01 with the mesh
and the post-tensioned anchors.
Moreover, the response of the cliff under seismic loading has also
been studied by means of a dynamic finite element calculation. The
calculations have been repeated with five different real accelerograms
recorded over rock. The response spectrum of the five accelerograms
selected is equivalent to a return period of 974 years. The criteria used
to calibrate the dynamic parameters are shown in the text.
The analysis of the results (with mesh and post-tensioned
anchors) shows that an incipient slip surface is formed but it does
not reach the top of the cliff. This suggests, in concordance with
the pseudo-static calculation obtained by the Morgenstern–Price
and FEM, that the reinforcements of the slope is essential to
guarantee its stability. The dynamic calculations have shown that
the fall of slabs would only be superficial, similarly to the slab
failure of 1985. This result is slightly different to the critical slip
surface obtained with the Morgenstern–Price method that could
penetrate inside the Alhambra walls.
For all these reasons the solution of the mesh and the post-
tensioned anchors is essential to preserve the cliff. The mesh
would protect the surface of the slope from erosion and fall of
 A. Morales-Esteban et al. / Soil Dynamics and Earthquake Engineering 69 (2015) 28–45
slabs. The pressure applied would counteract the extensive tec-
tonic regime and reduce the formation of plastic points. It has also
been demonstrated that without reinforcement the slope could be
unstable under seismic loads.
Acknowledgments
The financial support given by the Spanish Ministry of Science
and Technology, Projects BIA-2004-01302, BIA 2010-20377 and
TIN2011-28956-C02-00 are acknowledged.
References
[1] Aki K. Maximum likelihood estimate of b in the formula log N ¼a  bM and its
confidence limits. Bull Earth Res Inst 1965;43:237–9.
[2] Ambraseys NN. The seismic stability of earth dams. In: proceedings of the 2nd
world conference on earthquake engineering; 1960, vol. 2. p. 1345–63.
[3] Ambraseys NN, Sarma SK. The response of earth dams to strong earthquakes.
Géotechnique 1967;17(3):181–213.
[4] Azañón JM, Azor A, Booth-Rea G, Torcal F. Small-scale faulting, topographic
steps and seismic ruptures in the Alhambra (Granada, southeast Spain). J Quat
Sci 2004;19(3):219–27.
[5] Boore DW, Joyner WB, Fumal T. Estimation of response spectra and peak
acceleration from western North American earthquakes: an interim report. US
Geological Survey. Open-File Report 93-509; 1993.
[6] Brinkgreve RBJ. Plaxis 2D. Lisse: AA Balkema Publisher; 2012.
[7] Centre d0 Études Ménard. Détermination de la poussée exercée par un sol sur
une paroi de soutènement. Publication D/60/75; 1970.
[8] EERA. Equivalent-linear earthquake site response analyses code; 2000. 〈http://
usc.edu/GEES/Software/EERA2000/Default.htm〉.
[9] Eurocode 8. Design of structures for earthquake resistance – Part 5: founda-
tions, retaining structures and geotechnical aspects. UNE-ENV 1998-5; 1998.
[10] Finn WDL. Dynamic response analysis of soil structures. (Chapter 3.2). In:
Kardestunger DH, Norrie DH, editors. Finite element hanbook. MacGraw-Hill;
1987. p. 3168–95.
[11] Griffiths DV, Lane PA. Slope stability analysis by finite elements. Géotechnique
1999;49(3):387–403.
[12] Gutenberg B, Richter CF. Earthquake magnitude, intensity, energy and accel-
eration. Bull Seismol Soc Am 1942;32:163–91.
[13] Gutenberg B, Richter CF. Seismicity of the Earth. 2nd ed. Princeton, New
Jersey: Princeton University; 1954.
[14] Hilber HM, Hughes TJR, Taylor RL. Improved numerical dissipation for time
integration algorithms in structural dynamics. Earth Eng Struct Dyn
1977;5:283–92.
45
[15] Jing L. A review of techniques, advances and outstanding issues in numerical
modeling for rock mechanics and rock engineering. Int J Rock Mech Min Sci
2003;40:283–353.
[16] Justo JL, Saura J, Durand P, Morales-Esteban A, Azañón JM, Castro D. The
restoration of San Pedro cliff at the Alhambra, an example of the preservation
of historic sites. Preserv Nat Stone Rock Weather 2007.
[17] Justo JL, Azañón JM, Azor A, Saura J, Durand P, Villalobos M, et al. Neotectonics
and slope stabilization at the Alhambra, Granada, Spain. Eng Geol
2008;100:101–19.
[18] Kuhlmeyer RL, Lysmer J. Finite element method accuracy for wave propaga-
tions problems. J Soil Mech Found Div 1973;99(5):421–7.
[19] Lysmer J, Kuhlmeyer RL. Finite dynamic model for infinite media. ASCE J Eng
Mech Div 1969;95(4):859–77.
[20] Makdisi FI, Seed B. Simplified procedure for estimating dam and embankment
earthquake-induced deformations. J Geotech Eng Div ASCE 1978;104
(GT7):849–67.
[21] Makdisi FI, Seed B. Simplified procedure for evaluating embankment response.
J Geotech Eng Div ASCE, 105; 1979; 1427–34.
[22] Morales-Esteban A, Justo JL, Martínez-Álvarez F, Azañón JM. Probabilistic
method to select calculation accelerograms based on uniform seismic hazard
acceleration response spectra. Soil Dyn Earthq Eng 2012;43:174–85.
[23] Morgenstern NR, Price VE. The analysis of the stability of general slip surfaces.
Géotechnique 1965;15(1):79–93.
[24] NCRS-02. Norma de construcción sismorresistente: parte general y edificación.
Updated up to October 2004. RD 997/2002, Madrid, Spain; 2004.
[25] Newmark NM. Effects of earthquakes on dams and embankments. Géotech-
nique 1965;15(2):139–60.
[26] Park D, Hashash YMA. Soil damping formulation in nonlinear time domain site
response analysis. J Earthq Eng 2004;8(2):249–74.
[27] Plaxis. Plaxis 2D Version 9.0 manual. Eds. RBJ Brinkgreve and W Broere, Delft
Univ and Plaxis BV, The Netherlands; 2012.
[28] Rayleigh JWS, Lindsay RB. The theory of sound. New York: Dover Publications;
1945.
[29] Ranalli G. A statistical study of aftershock sequences. Ann Geofis
1969;22:359–97.
[30] Rodríguez-Ortíz JM. El tajo de San Pedro en La Alhambra de Granada.
Unpublished report; 1998.
[31] Sarma SK. Seismic stability of earth dams and embankments. Géotechnique
1975;25(4):743–61.
[32] Utsu T. A method for determining the value of b in a formula log n¼ a  bm
showing the magnitude–frequency relation for earthquakes. Geophys Bull
Hokkaido Univ 1965;13:99–103.
[33] Utsu T. Representation and analysis of the earthquake size distribution:
a historical review and some new approaches. Pure Appl Geophys
1999;155:509–35.
[34] Visone C, Bilotta E, de Magistris FS. One-dimensional ground responses as a
preliminary tool for dynamic analyses in geotechnical earthquake engineering.
J Earth Eng 2009;14(1):131–62. http://dx.doi.org/10.1080/13632460902988950.
[35] Zienkiewicz OC, Taylor RL. The finite element method. 4th ed. Solid and fluid
mechanics, dynamics and non-linearity, vol. 2. London, UK: Mc-Graw-Hill;
1991.