Soil Dynamics and Earthquake Engineering 50 (2013) 106–116

Contents lists available at SciVerse ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Development of fragility functions for geotechnical constructions:

Application to cantilever retaining walls
Sotirios Argyroudis a,n, Amir M. Kaynia b,c, Kyriazis Pitilakis a
a
Department of Civil Engineering, Aristotle University of Thessaloniki, Greece
b
Department of Structural Engineering, Norwegian University of Science and Technology, NTNU, Trondheim, Norway
c
Norwegian Geotechnical Institute, Oslo, Norway

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

Article history: Fragility curves constitute an emerging tool for the seismic risk assessment of all constructions at risk.
Received 5 November 2012 They describe the probability of a structure being damaged beyond a specific damage state for various
Received in revised form levels of ground shaking. They are usually represented as two-parameter (median and log-standard
15 February 2013
deviation) cumulative lognormal distributions. In this paper a numerical approach is proposed for the
Accepted 18 February 2013
construction of fragility curves for geotechnical constructions. The methodology is applied to cantilever
Available online 9 April 2013
bridge abutments on surface foundation often used in road and railway networks. The response of the
Keywords: abutment to increasing levels of seismic intensity is evaluated using a 2D nonlinear FE model, with an
Fragility curves elasto-plastic criterion to simulate the soil behavior. A calibration procedure is followed in order to
Vulnerability
account for the dependency of both the stiffness and the damping on the soil strain level. The effect of
Retaining walls
soil conditions and ground motion characteristics on the global soil and structural response is taken
Bridge abutments
Site effects into account considering different typical soil profiles and seismic input motions. The objective is to
Numerical modeling assess the vulnerability of the road network as regards the performance of the bridge abutments;
therefore, the level of damage, is described in terms of the range of settlement that is observed on the
backfill. The effect of backfill material to the overall response of the abutment wall is also examined.
The fragility curves are estimated based on the evolution of damage with increasing earthquake
intensity. The proposed approach allows the evaluation of new fragility curves considering the
distinctive features of the structure geometry, the input motion and the soil properties as well as the
associated uncertainties. The proposed fragility curves are verified based on observed damage during
the 2007 Niigata-Chuetsu Oki earthquake.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction as lifeline systems (e.g. [32,35]). They relate the seismic intensity
to the probability of reaching or exceeding a damage state (e.g.
The experience of past earthquakes worldwide reveals that minor, moderate, extensive, collapse) for each element at risk.
geotechnical structures are quite vulnerable to earthquake shaking. Different approaches can be used to develop the fragility curves,
In addition to life and material losses, damage to roadway and including empirical, judgmental, analytical and hybrid methods.
railway elements (tunnels, embankments, trenches, levees, slopes, Analytical fragility curves adopt damage distributions simulated
retaining walls and others) can seriously affect the transportation from the analyses of structural models under increasing earth-
of people and products in both short-term (emergency actions) and quake loads as their statistical basis [40]. Recently, fragility curve
long-term periods. The examples of recent devastating earthquakes methodologies using numerical approaches have become widely
in Japan and New Zealand indicate that even in developed areas, adopted as they are more readily applicable to different structural
damage to geotechnical structures is often as visible and important types and geographical regions where damage records are insuf-
as the structural damage. Therefore, the vulnerability of earth ficient. Although such approaches are commonly used to describe
structures is of great interest in the seismic risk assessment of the seismic response of structures such as buildings (e.g. [1]) or
transportation networks and infrastructures. bridges (e.g. [33]), their use for the fragility analysis of geotech-
Fragility curves constitute one of the key elements of seismic nical structures is limited (e.g. [6], for tunnels; [4] for bridge
probabilistic risk assessment of the built-up environment as well abutments; [24], for waterfront structures). Fragility curves for
most of the lifeline elements and transportation infrastructures
are provided in HAZUS [32], however, retaining walls or bridge
n
Corresponding author. Tel./fax: þ 30 2310 995842. abutments are not included. Similarly, the REDARS methodology
E-mail address: [email protected] (S. Argyroudis). [44] that was developed for seismic risk analysis of highway

0267-7261/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.soildyn.2013.02.014
 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116 107

systems proposes damage threshold and repair costs for bridge soil dynamic properties and stratigraphy, can significantly affect
approach fills but not in the form of fragility curves. the response of the structure under consideration [15]. The
Generally, earthquake effects on geotechnical structures such selection of seismic input motion is an important step of the
as tunnels, retaining walls, bridge abutments, trenches or analysis. Among other procedures, the selection of accelerograms
embankments can be grouped into two categories: (1) ground is often performed on the basis of compatibility between their
shaking; and (2) ground failure such as liquefaction, fault dis- response spectra and a corresponding ‘target’ spectrum as defined
placement, and slope instability. The major factors determining by code provisions or computed directly through seismic hazard
the seismic performance of geotechnical structures include the analysis [25]. In the present application, different typical soil
geometry (shape, dimensions) and material properties of the profiles and seismic input motions are used to account for the
structure, the properties of the surrounding soil or rock and the effect of soil and backfill conditions and ground motion charac-
severity of the ground shaking or ground failure. Typical damage teristics in the global soil-structure response. The stiffness and
modes of bridge abutments include the lateral movements, tilting damping parameters are defined based on the results of 1D
and settlements of the wall as well as lateral movements and equivalent linear analysis, in order to be compatible with the
settlements of the backfill (Fig. 1). Although approach fill settle- expected strain levels during the earthquake. By defining the
ment does not typically result in extensive repair costs and traffic damage level through an appropriate damage index, the fragility
delay, it has been the most commonly occurring type of failure in curves are constructed as a function of the level and the type of
highway and railway systems during recent earthquakes. seismic excitation considering the primary sources of uncertain-
The objective of the present paper is to develop fragility curves ties as it is described later. This approach allows evaluating the
for bridge abutments of retaining cantilever wall type with fragility curves of geotechnical structures such as the type that is
surface foundation under seismic shaking. The effect of different considered herein, respecting the distinctive features of their
parameters on the overall performance of the backfill-abutment geometries, input motion characteristics and soil properties.
under seismic excitation is examined. In particular, the soil
conditions underneath the abutment, the geometry of the abut- 2.1. Definition of damage states
ment, as well as the properties of the backfill material are studied.
Fragility curves are derived for different soil types and abutment Different damage levels have been proposed for the vari-
heights. ous elements of the roadway system infrastructures based on
damage description, functionality criteria or repair costs (e.g.
[32,44,29,5]). The damage states (ds) that have been defined in
2. Methodology the European project SYNER-G [27] are used in this paper
(Table 1). These states are described in terms of induced perma-
The proposed procedure for the derivation of analytical fragi- nent vertical ground displacement (PVGD) of the backfill, which
lity curves for geotechnical structures is described in Fig. 2. The represents the damage index (DI) of the bridge abutment. In
procedure is similar to the one used by Argyroudis and Pitilakis particular, a mean value of PVGD is estimated for minor, moder-
[6] where numerical fragility curves were derived for shallow ate and extensive/complete damage based on a range of values
tunnels in alluvial deposits. In the aforementioned study the (min, max). These definitions are based on expert judgment and
response of the tunnel was calculated under quasi static condi- they are consistent with the existing ones (e.g. [32,44]).
tions, while in the present one the response of the coupled system
wall–soil, backfill is calculated through a 2D nonlinear dynamic Geotechnical Soil type Seismic input motion
numerical analysis. For a given seismic hazard intensity, the structure typology Typical soil profiles Accelerograms, intensity levels
Basic models
typology of the exposed structure is a key factor. The geometry,
material properties, soil and topography conditions, code design
level and structure details are important parameters which Damage index (DI), 2D dynamic analysis 1D equivalent linear analysis of
describe the typology of the element at risk as well as its capacity damage states (ds), of geotechnical the soil profiles - input motions
thresholds values of structures - soil models
to withstand seismic loads. For earth structures, the main typo- DI for each ds models
logical feature is the soil type which characterizes either a
construction (e.g. embankment, backfill, slope) or its foundation Soil stiffness
Evolution of damage with intensity and damping parameters
and surrounding material. Different soil classification systems are
measure (IM), definition of median
available based on various soil properties. A widely used classi-
threshold value of IM for each ds
Uncertainties
fication scheme is the one provided by Eurocode 8 [13], which is (seismic demand, element
based on the soil’s shear wave velocity, Vs30. Therefore, represen- capacity, definition of DI and ds)
Fragility curves for each structure
tative models should be selected for the development of adequate trypology and soil type
fragility curves.
The characteristics of the earthquake ground motion in terms Fig. 2. General procedure for deriving numerical fragility curves for geotechnical
of amplitude, frequency content and duration, in relation to the structures.

Subsidence

Backfill

Fig. 1. Settlement of backfill behind a bridge abutment (left) and example of bridge performance during 2007 Niigata-Chuetsu Oki, Japan earthquake (right).
108 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116

Table 1
Definition of damage states.

Damage state Permanent vertical ground displacement (m) Serviceability

Min Max Mean

ds1: Minor 0.02 0.08 0.05 Open reduced speeds or partially closed during repair
ds2: Moderate 0.08 0.22 0.15 Closed or partially closed during repair works
ds3: Extensive/complete 0.22 0.58 0.40 Closed during repair works

Damage Index
The selected thresholds are also related with the serviceability of
the road in terms of traffic conditions. The closure duration
depends on the length of the damage as well as on the number
of traffic lanes.

2.2. Estimation of fragility curve parameters

Fragility functions describe the probability of exceeding dif- βD

ln DI (dsi)
ferent limit states given a level of ground shaking. The level of
shaking can be defined using numerous earthquake parameters
depending on the structure under consideration. These include βD
peak ground acceleration, velocity, displacement, spectral accel-
eration, velocity, displacement or permanent ground displace-
ment. Fragility curves are usually described by a lognormal
probability distribution function, as follows:
  
1 IM
P f ðdsZ dsi 9SÞ ¼ F Uln ð1Þ
btot IM mi

where Pf(  ) is the probability of exceeding a particular damage ln IMmi Intensity Measure
state, ds, for a given seismic intensity level defined by the
Fig. 3. Example of evolution of damage with earthquake intensity measure (IM)
earthquake intensity measure, IM (e.g. Peak Ground Accelera- and definition of threshold median value (IMmi) for the damage state i (dsi).
tion-PGA), F is the standard cumulative probability function, IMmi Definition of standard deviation (bD) due to variability of input motion (demand).
is the median threshold value of the earthquake intensity mea-
sure, required to cause the ith damage state, and btot is the total
3. Application to cantilever retaining walls
lognormal standard deviation. Therefore, the development of
fragility curves according to Eq. (1) requires the definition of
A bridge abutment with a simplified model as a cantilever
two parameters, IMmi and btot.
retaining wall on surface foundation is considered for the appli-
The level of abutment-backfill damage is described by a
cation (Fig. 4). Two different heights (h), commonly used in
damage index (DI) expressing the exceedance of certain limit
practice are considered, equal to 6.0 and 7.5 m. It is assumed
states (Table 1) and the fragility curves are estimated based on
that the bridge deck is supported by bearings on the wall and that
the evolution of damage index with increasing earthquake inten-
it does not receive horizontal forces. A vertical load equal to
sity considering the associated uncertainties. An example is given
200 kN/m is applied on top of the wall in order to simulate the
in Fig. 3, where the different points indicate the results of analysis
total load of the deck. The deck lateral reaction is assumed
in terms of damage index for different levels of earthquake
negligible.
intensity. The solid line is produced based on a regression analysis
and the median threshold value of the intensity measure (IMmi)
required to cause the ith damage state (dsi) is estimated based on 3.1. Input motions
the definition of this damage state through the damage index. A
lognormal standard deviation (btot) that describes the total Five real acceleration time histories from different earth-
variability associated with each fragility curve has to be esti- quakes, recorded on soil similar to class A (rock) or B (stiff soil)
mated. Three primary sources of uncertainty are considered [32], of Eurocode 8 [13], were selected as outcrop motion for both the
namely the definition of damage states (bds), the response and 1D ground response and the 2D coupled soil-backfill-abutment
resistance (capacity) of the element (bC) and the earthquake input analyses. The earthquakes are:
motion (demand) (bD). The total variability is modeled by the
combination of the three contributors, assuming that they are
 Kocaeli (Gebze), Turkey, M ¼7.4, 1999;
statistically independent and lognormally distributed random
 Hector Mine (Hector), USA, M¼7.1, 1999;
variables:
 Parnitha (Kypseli), Greece, M¼6.0, 1999;
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 Loma Prieta (Diamond Height), USA, M ¼6.9, 1989;
btot ¼ b2DS þ b2C þ b2D ð2Þ
 Umbria Marche (Gubbio-Piana), Italy, M¼4.8, 1998.
Due to the lack of a more rigorous estimation, the uncertainty
parameters can be obtained from the literature (e.g. HAZUS, [32]). The records are selected to cover the inherent uncertainties
However, the last source of uncertainty, associated with seismic related to the seismic motion such as the seismotectonic envir-
demand, can be described by the standard deviation of the DIs onment, amplitude, frequency content and significant duration.
that have been calculated for the different input motions at each No specific soil amplification factors are applied, as this is
level of PGA. explicitly taken into consideration through the numerical
 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116 109

200kN/m

Bridge deck

Backfill

t1 h

t3

t2

H Soil

Bedrock EQ

L (m) h (m) H (m) t1 (m) t2 (m) t3 (m) γc*(kN/m3) Ec (kN/m2)

5.5 6.0 50.0 1.0 1.0 1.5 24.0 3.5E+07
6.5 7.5 50.0 1.0 1.0 2.5
* γc: unit weight of concrete

Fig. 4. Geometry and properties of the abutment under study.

4.0 care is needed for fault characteristic except of course the case
that the potential seismic fault is crossing the bridge abutment,
which is a rather different problem.
3.0
3.2. Soil profiles and geometry

Two ideal soil deposits of 50 m depth were considered for the

PSA/PGA

2.0 numerical analyses. Soil properties based on common engineering

practice were selected [3], in order to obtain soil type C and D
Mean according to the Eurocode 8. In particular, an initial value for the
Mean+σ undrained shear strength at the top layer, Su0, was assumed
1.0
Mean-σ (equal to 50 kPa for soil type C and 20 kPa for soil type D), while a
ΕC8-Α gradient model for Su is defined for the next layers
(Sun ¼Su0 þ0.25svn0 , where svn0 is the average vertical effective
0.0
stress in layer n). The variation of Gmax with depth is estimated as
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Gmaxn ¼ASun (where A was taken as equal to 1000 for soil type C
T (sec) and 800 for soil type D). The values of the other soil parameters
were taken as v¼0.35, g ¼19–19.5 kN/m3 for soil type C,
Fig. 5. Mean acceleration spectra 7 1 standard deviation of the input time
histories for the 1D ground response analyses and comparison with the spectrum
g ¼18–18.5 kN/m3 for soil type D. The variation of initial shear
provided by EC8 (soil type A). modulus (Gmax) and shear wave velocity (Vs) with depth is shown
in Fig. 6a.
Two different sandy materials have been considered for the
analyses and the different soil models where the seismic motion backfill, corresponding to different soil and compaction conditions
is applied in outcrop conditions at the base of the model. The (backfill 1: g ¼18.0 kN/m3, j ¼361; backfill 2: g ¼ 19.0 kN/m3,
normalized mean and mean7one standard deviation of the j ¼401). An average value of Gmax and Vs has been evaluated for
acceleration spectra of the selected motions, are plotted together each layer based on the corresponding properties as described before
with the EC8 spectrum for soil class A in Fig. 5. Spectral matching for the soil deposit (type C or D) and using Hardin [20] expressions for
is the most commonly proposed earthquake record selection sands. To avoid modeling the backfill in detail, the additional
method by seismic codes [25,21,22]. Previous studies have shown simplification was made that these profiles represent average proper-
that the use of spectrum-compatible ground motion provides ties for the whole soil layer behind the wall. The variation of Gmax and
least variations in the response parameters [17]. In the dynamic Vs with depth for the two backfills is shown in Fig. 6b.
analyses, the time histories are scaled from 0.1 to 0.5 g in order to
calculate the response of the backfill-abutment to increasing 3.3. 1D equivalent linear ground response analysis
seismic intensity. The user can simply estimate the PGA at the
surface and then use directly the relevant set of fragility curves to In order to account for the soil nonlinearity in the low to
estimate the vulnerability for the particular PGA value. No specific medium strain range, 1D equivalent linear analyses were carried
110 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116

Shear wave velocity (m/s) Gmax (MPa)

0 100 200 300 400 500 0 100 200 300 400
0 0
Soil type C Soil type C
Soil type D Soil type D
10 10

20 20
Depth (m)

Depth (m)
30 30

40 40

50 50

Shear wave velocity (m/s)

0 100 200 300
0

3
Depth (m)

6 backfill 1
backfill 2
7

Fig. 6. Variation of shear wave velocities (Vs) and initial shear modulus (Gmax) with depth for the soil profiles underneath the abutment (a) and the backfills (b).

out to derive realistic parameters for the soil modulus and As an example, Fig. 7 shows the variations with depth of
damping at the expected shear strain levels during the increasing maximum shear strain (gmax), normalized shear modulus
levels of seismic excitation. The resulting average stiffness and (G/Gmax), damping ratio (D) and maximum acceleration (PGA)
damping parameters were adopted in the finite element analyses for the input motion Kypseli scaled at 0.1 g.
in order to be compatible with the shear deformation level A calibration procedure was followed in order to account for
expected during each utilized earthquake. Similar approach was the dependency of both the stiffness and the damping on the
adopted by Gil et al. [18], Hashash et al. [19], Gazetas et al. [16], strain level. In this context, the results from the 1D analyses for all
Amorosi and Boldini [2] for the analysis of buried structures. the input motions scaled at 0.1 g were employed. This assumption
The 1D ground response analyses were performed using the was made because the equivalent linear methods are more
code EERA [8], which is based on the assumption of equivalent reliable only for lower strain levels. For higher strain levels the
linear soil behavior. The variation of the shear modulus G/Gmax effect of nonlinearity is captured through the Mohr–Coulomb
and damping ratio D with the shear strain level g was defined yield criterion used for the soil behavior in the 2D model.
according to the typical results available in the literature. In For each layer of the numerical FE model a single value of G
particular, the curves provided by Darendeli [11] as a function of was used based on the average computed G/Gmax ratio, equal to
plasticity index (PI) and effective stress were used. Curves with an 0.68 for soil type C and 0.55 for soil type D (Fig. 8). The reduction
average PI ¼30% were selected in this study. For the bedrock, the of the Gmax value for the backfill material was estimated in a
curves proposed by Schnabel et al. [41] were applied. A total similar way. The average aR and bR Rayleigh coefficients were
number of 14 layers were assumed to model the 50 m soil profile. computed based on the average value of damping for all layers.
 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116 111

Maximum Shear strain (%) Maximum Acceleration (g) G/Gmax Damping Ratio (%)
0 0.1 0.2 0.3 0 0.5 1 1.5 0 5 10 15
0 0.05 0.1
0 0 0 0

5 5 5 5

10 10 10 10

15 15 15 15

20 20 20 20
Depth (m)

Depth (m)

Depth (m)

Depth (m)
25 25 25 25

30 30 30 30

35 35 35 35

40 40 40 40

45 45 45 45

50 50 50 50

Maximum Shear strain (%) Maximum Acceleration (g) G/Gmax Damping Ratio (%)
0 0.1 0.2 0.3 0 0.5 1 1.5 0 10 20 30
0 0.5 1
0 0 0
0
5 5 5
5

10 10 10 10

15 15 15 15

20 20 20 20
Depth (m)

Depth (m)

Depth (m)

Depth (m)
25 25 25 25

30 30 30 30

35 35 35 35

40 40 40 40

45 45 45 45

50 50 50 50

Fig. 7. Results of the 1D ground response analysis performed with EERA for soil profile type C (a) and D (b). Input motion: Kypseli—0.1 g.

G (MPa)

0 100 200 300

0

10 Soil D
Soil C
15
Fig. 9. FEM employed in the analyses with detail of the finite element mesh near
the structure.
20
Depth (m)

25 The Rayleigh coefficients were chosen for the frequency interval

1.4–4.2 Hz for soil profile C and 1.0–3.0 Hz for soil profile D,
30 covering approximately the first and third modes of the profile. In
particular, the average values of damping were 5.7% and 8.5% for
35 soil type C and D accordingly. The average values of Rayleigh
coefficients were aR ¼0.7475, bR ¼ 0.0032 for soil type C and
40 aR ¼0.8047, bR ¼0.0068 for soil type D.

3.4. FE numerical analysis

45
The coupled soil–wall interaction analyses were performed
50
with the 2D (plane strain) finite element code PLAXIS (v9.02 [36]).
Fig. 8. Average computed values of G with depth for soil type C and D compatible The model is shown in Fig. 9. The basis was assumed to be rigid
to the developed ground strains for PGA¼ 0.1 g. and the lateral sides were characterized by absorbent boundaries.
112 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116

The total width of the model is 500 m, which is sufficient to avoid input motions. It is clear that in all the cases of soil types and
boundary effects. The domain was discretized in a total of 2702 abutment geometry the response of the abutment-approach fill is
15-node plain strain triangular elements. In the area around the improved when the friction angle of the backfill material is
abutment wall the mesh was refined as shown in Fig. 9. Con- increased. In particular, the maximum settlement of the backfill
sistently with the 1D ground response analyses, the foundation is reduced on average by 25% for the soil type C, while the
soil domain was modeled with 14 horizontal layers to account for reduction in soil type D is approximately 20%. Consequently, the
the variable stiffness with depth. The backfill was modeled in backfill material influences the overall vulnerability of the abut-
4 horizontal layers. ment as it is expressed in the present work.
All analyses were started by performing a set of initial stage
analyses to simulate initial geostatic stresses as well as the
construction of the wall and the backfill. Then the dynamic 5. Fragility curves
analyses followed where the seismic input is applied uniformly
at the basis of the model. To ensure proper parameters selection The derivation of fragility curves (i.e. definition of the median
for the soil and wall, an evaluation of the static safety factor was threshold value of PGA for each damage state) is based on the
made. All phases of the analysis were characterized by the construction of a curve for the computed damage indices in terms
assumption of elasto-plastic soil behavior (i.e. Mohr–Coulomb of permanent vertical ground displacement (i.e. settlements of
criterion). Proper interface elements with a realistic friction the backfill, PVGD), versus increasing seismic intensity in terms of
coefficient of Rinter ¼0.70 were used to model the interface PGA at the ground surface under free field conditions. The curves
between the backfill and foundation soil with the wall. A constructed in that way represent the evolution of damage with
representative example of the analysis output is given in Fig. 10 increasing earthquake intensity. The curves are established by a
where the vertical displacements in the backfill and the deformed regression analysis, considering PVGD as the dependent variable
mesh are illustrated. and PGA as the independent variable. Similar approaches are used
for the derivation of the fragility functions in other studies (e.g.
[33,34,6,15]). The median threshold value of PGA can be obtained
4. Effect of backfill material for each damage state using this curve and the definitions of
damage states given in Table 1. When the computational results
The influence of backfill strength properties on the perfor- do not reach complete damage threshold, the fragility curves for
mance of retaining walls and bridge abutments under seismic this damage state are derived based on extrapolation of the
conditions has been recently addressed in a number of studies available results.
based on numerical models (e.g. [7,31,39]), shaking table and A lognormal standard deviation (b) that describes the total
centrifuge tests results (e.g. [42,28,14,43]). The reference case for variability associated with each fragility curve has to be estimated
the backfill in this study is the case of backfill 2 in Fig. 6b. To (see Eq. (2)). Due to the lack of a rigorous estimation, the
investigate the influence of the backfill material on the results, a uncertainty associated with the definition of damage states (bds)
different backfill with lower strength and stiffness was considered is set equal to 0.4 following the approach of HAZUS [32] for
(backfill 1, Fig. 6b). buildings. The uncertainty due to the capacity (bC) is assigned
Fig. 11 shows the variation of the maximum vertical ground equal to 0.3 based on engineering judgment. The last source of
displacements in the two backfill cases under different seismic uncertainty, associated with the seismic demand, is described by

Fig. 10. Distribution of the backfill permanent vertical displacements (max 0.16 m) (a) and deformed mesh (b). Case: h ¼ 6 m, soil type C, input motion Gebze—0.4 g,
backfill 2.
 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116 113

0.6 Soil type C, h =6m

backfill 1
Vertical Displacement -Backfill (m)

0.5
backfill 2
Average (backfill 1)
0.4
Average (backfill 2)

0.3

0.2

0.1

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
PGA free field (g)

0.6 Soil type D, h= 6m

backfill 1 Fig. 12. Evolution of damage in retaining wall (settlement in the backfill) with
intensity measure (PGA) for soil type C (up) and D (down) and h ¼ 6.0 m.
Vertical Displacement -Backfill (m)

0.5 backfill 2
Average (backfill 1)
0.4 Average (backfill 2)
Table 2
Parameters of numerical fragility curves for cantilever retaining walls.
0.3
Soil Wall Median peak ground acceleration (g) Lognormal
0.2 type height standard
(m) ds1: ds2: ds3: Extensive/ deviation (b)
Minor moderate complete
0.1
C 6.0 0.38 0.64 1.02 0.70
7.5 0.26 0.52 0.97 0.70
0.0
D 6.0 0.20 0.45 0.93 0.90
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 7.5 0.18 0.39 0.78 0.90
PGA free field (g)

Fig. 11. Variation of vertical ground displacement for different backfill materials.

the variability in response (i.e. settlement of the backfill) due to

the variability of ground motion. In particular, an average stan-
dard deviation of the damage indices (settlements) that have been
calculated for the different input motions at each level of PGA is
estimated. The total variability is calculated with Eq. (2).
Fig. 12 shows the PVGD–PGA diagrams for the soil profiles
type C and D, abutment wall with h¼6.0 m and backfill 2. The
estimated parameters and the corresponding fragility curves for
backfill 2, which is considered to be the reference case, are
illustrated in Table 2, Figs. 13 and 14. The comparison of the
fragility curves shows that, as expected, the wall with h¼7.5 m is
more vulnerable than the one with h ¼ 6.0 m. The results also
indicate that for most of the damage states the vulnerability of
both walls in soil type D is higher than in soil type C. Fig. 13. Numerical fragility curves for cantilever retaining walls—soil type C.
114 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116

6. Verification of the proposed fragility curves

Settlements of approach backfills behind abutments have been

widespread in recent earthquakes. This is related to response of
soft soil as well as inadequately compacted backfill material [38].
This type of damage was common in recent earthquakes as for
example in 1999 Kocaeli, Turkey [23], 2007 Niigata-Chuetsu Oki,
Japan [26], 2010 Maule, Chile [9], 2011 Christchurch, New
Zealand [10] and others. Generally, damage in approach fill is
not of life safety concern, and repairs and reopening to traffic are
usually made within a short period. However, the emergency
traffic can be seriously affected. In addition, in some cases, the soil
movement is accompanied by movement/rotation of the wall
which could require considerable repair cost. Although such
damage has been widely observed, a complete description of
Fig. 14. Numerical fragility curves for cantilever retaining walls—soil type D. the displacement extent, backfill material and abutment geome-
try is rarely reported. Moreover, information related to the
seismic intensity and soil type at the damage site location is
usually incomplete.
Table 3
In order to verify the proposed fragility curves, the reported
Observed damage in bridge abutment-approach fills after the 2007 Niigata- damages in bridge approach fills after the 2007 Niigata-Chuetsu
Chuetsu Oki earthquake. Oki earthquake in Japan are used [26]. The reports and records
from this earthquake were found to be the most complete set for
Bridge Latitude, Length Observed Vs30 (m/s) PGV PGA
the present verification. The damage state has been defined based
Longitude (m) damage at (Soil type/ (cm/s) (g)
backfill EC8) on the description and extent of settlement in the backfill. The
soil type is defined based on Vs30 values provided by the Japan
Ansei 37.38639 106 Moderate 195 (C) 89.47 0.68 Seismic Hazard Information Station (www.j-shis.bosai.go.jp/en/).
138.56932 The seismic intensity in terms of PGA at the bridge location has
Kai-Un 37.39232 111.2 Extensive 195 (C) 82.54 0.63
138.58063
been estimated based on the distribution of peak ground
Nagomi 37.39040 102.5 Minor 195 (C) 78.22 0.59
138.59080
Toyota 37.38994 105.5 Extensive 195 (C) 78.22 0.59 0.30
138.59337 Soil C
Uehara Ni 37.39004 65 Extensive 157 (D) 91.91 0.70
0.25
138.59595
Uehara Ichi 37.38950 100 Minor 157 (D) 94.41 0.72
138.59662 0.20
Genji 37.38439 95 Extensive 158 (D) 91.91 0.70
P(n ds=x)

138.60083
Tsurugi 37.39212 43 Minor 158 (D) 89.47 0.68 0.15
138.60075 Minor
Heishi Hashi 37.38940 22 Moderate 195 (C) 84.79 0.64 0.10 Moderate
138.5742
Heishi O-Hashi 37.38896 118 Extensive 195 (C) 84.79 0.64 Extensive
138.57581 0.05
Rinko Yasaka 37.36774 81 Moderate 230 (C) 72.15 0.71
138.54372
0.00
Yasaka 37.36575 62 Minor 230 (C) 72.15 0.71
138.54353 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Ukawa 37.36405 67 None 195 (C) 84.79 0.64 number of damaged bridges
138.54608
0.50
Osu 37.36350 66 Extensive 195 (C) 87.10 0.66 Soil D
138.5478 0.45
Hokuriku 37.42252 130 Unknown 406 (B) 32.21 0.32
Expway 138.64122 0.40
Nishiura 37.42478 19 Extensive 195 (C) 66.56 0.50 0.35
(Route 352) 138.60757
Hachisaki 37.3150 185 Minor 260 (C) 26.68 0.26 0.30
P(n ds=x)

138.4370 Minor
0.25
Omigawa 37.343 300 None 420 (B) 24.61 0.24 Moderate
138.482 0.20 Extensive
Yoshiigawa 37.399 200 Extensive 161 (D) 78.22 0.59
0.15
138.625
Route 522 37.364 10 Moderate 196 (C) 91.91 0.70 0.10
138.570
Ugawa river 37.3605 60 Minor 195 (C) 84.79 0.64 0.05
138.5490 0.00
Yoneyama 37.346 280 None 456 (B) 23.96 0.23
0 1 2 3 4 5
138.488
Yonahime 37.31680 120 None 260 (C) 26.68 0.26 number of damaged bridges
138.4348
Highway 116 37.446 20 Minor 289 (C) 41.02 0.40 Fig. 15. Distribution of estimated damage through the proposed numerical
138.638 fragility curves for abutments in soil type C and D for the 2007 Niigata-Chuetsu
Oki earthquake.
 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116 115

Table 4
Comparison between observed and estimated by the proposed fragility curves damage for the 2007 Niigata-Chuetsu Oki earthquake.

Soil C Soil D

Minor Moderate Extensive Minor Moderate Extensive

Observed number of bridges in each damage state 5 4 5 2 0 3

Probability that the number of bridges in each damage state is equal to the observed 0.19 0.23 0.13 0.23 0.17 0.23
Most probable number of bridges in each damage state 4 4 3 1 1 2

velocities (PGV) at the ground surface. In particular, data calcu- 7. Conclusions

lated by Matsuoka [30] for the specific earthquake event using
records at seismic stations and considering the amplification A comprehensive methodology is described to construct
characteristics of surface layers was downloaded and imported numerical fragility curves for geotechnical constructions under
in GIS together with the bridge locations. The PGA values are seismic loading. The proposed approach is applied to bridge
assessed based on ratios of peak ground velocity to peak ground abutment with cantilever retaining walls of ordinary heights.
acceleration at surface as provided by Power et al. [37] for stiff The response of the wall is evaluated based on dynamic 2D
(type B or C) and soft (type D) soil, considering M ¼6.5 and source numerical analysis using appropriate 2D finite element code.
to site distance 20–50 km. Table 3 summarizes the observed The model is properly considering the dependency of both soil
damage states, Vs30, soil type, PGA and PGV values for each stiffness and damping on the strain level. Seismic inputs with
bridge site. different frequency content, scaled to different levels of seismic
Due to lack of information regarding the abutment height for loading (PGA outcrop) and typical soil profiles, classified as C and
each bridge, the verification test is performed for the average D according to EC8, are employed. Defining different damage
fragility curves of the two different heights (h¼6.5 m and 7.0 m) states describing the backfill vertical permanent displacements
in Figs. 13 and 14. In particular, the probabilistic distribution of (settlements), the fragility curves could be derived as a function of
number of bridges in each damage state is estimated based on a the level of peak ground acceleration in free field conditions,
Monte Carlo approach by applying the proposed fragility curves considering the related uncertainties in the definition of damage
for the two soil types (Fig. 15). Each simulation case represents a states, the demand and the capacity of the wall.
deterministic damage pattern that is a possible outcome of the The available fragility models for the vulnerability assessment
earthquake, and it is produced by sampling from the probability of geotechnical structures are generally limited; they are mainly
of each damage state given the observed intensity at each site. based on empirical data, without properly considering soil and
Collectively they represent the probabilistic damage outcome of structure characteristics. With the proposed numerically derived
the event which can be used in order to quantitatively assess the fragility curves, the distinctive features of the construction
compatibility of each fragility curve with the observed data. typology, as well the input motion characteristic and the soil-
Bridges with no damage are excluded from the analysis as the backfill properties are considered in a more systematic way. The
list for undamaged structures in the area is incomplete. Bridges in effect of the foundation soil as well the influence of the backfill
soil type B are also excluded from the analysis as fragility curves material on the performance of the wall is investigated. The
are not developed for this soil type. In total, 14 bridges in soil type vulnerability of the retaining wall is modified for different soil
C and 5 in soil type D are considered. Fig. 15 shows the probability types, while a notable variation in the settlement of the backfill is
distribution of expected damage for the specific earthquake event. observed when the backfill material is modified. The reliability of
The observed number of bridges in each damage state is com- proposed fragility curves is validated using a damage dataset of
pared with the distribution, in order to see if the observed bridge abutments from the 2007, Niigata-Chuetsu Oki earth-
number is compatible with the distribution. A good agreement quake. A good agreement is obtained between the observed and
is obtained between the estimated damage using the fragility estimated damage for the specific event, although several uncer-
curves and the observed damage (Table 4). In particular, a tainties are included in this verification study.
satisfactory agreement is observed when the most probable
number of bridges (i.e. distributions0 mode) is compared with
the reported number in each damage state (minor, moderate, Acknowledgments
extensive). In addition, the probability that the expected number
of bridges in each damage state is equal to the observed one is This work has been developed in the framework of the
reported. These probabilities are always greater than 0.05, and research project SYNER-G: Systemic seismic vulnerability and
thus the null hypothesis (i.e. that the proposed model is true) risk analysis for buildings, lifeline networks and infrastructures
cannot be rejected [12]. safety gain, funded by the European Commission (FP7-ENV-2009-
There are a number of uncertainties related to the above 1-244061).
comparison. Namely, the possible error in the estimation of PGA The authors would like to acknowledge Dr. Kei Horie, Visiting
at each location and especially in soft soil, the definition of Professor in Kobe University for his assistance in finding
damage states based on the qualitative description of reported supplementary data for the Niigata earthquake.
damage and range of observed settlements in the backfill, the
characterization of soil type and the unknown abutment features. References
In addition, the number of data is rather limited. Nevertheless,
these results are encouraging as the comparison is quite satisfac- [1] Akkar S, Sucuoglu H, Yakut A. Displacement-based fragility functions for low-
tory, proving that the proposed fragility curves for bridge abut- and mid-rise ordinary concrete buildings. Earthquake Spectra
ments, despite the various uncertainties involved, are capable to 2005;21:901–27.
[2] Amorosi A, Boldini D. Numerical modelling of the transverse dynamic
evaluate the seismic vulnerability of ordinary bridge abutments behaviour of circular tunnels in clayey soils. Soil Dynamics and Earthquake
expressed in terms of backfill settlements. Engineering 2009;29:1059–72.
116 S. Argyroudis et al. / Soil Dynamics and Earthquake Engineering 50 (2013) 106–116

[3] Andersen KH, Lunne T, Kvalstad T, Forsberg CF. Deep water geotechnical [25] Katsanos EI, Sextos AG, Manolis GD. Selection of earthquake ground motion
engineering. In: Proceedings of the XXIV national conference of the Mexican records: a state-of-the-art review from a structural engineering perspective.
society of soil mechanics, Aguasclientes, 26–29 November, 2008. Soil Dynamics and Earthquake Engineering 2010;30(4):157–69.
[4] Argyroudis S, Kaynia MA, Pitilakis K. Seismic fragility curves of cantilever [26] Kayen R, Collins B, Abrahamson N, Ashford S, Brandenberg SJ, Cluff L.
retaining walls. In: Proceedings of the 2nd international conference on Investigation of the M6.6 Niigata-Chuetsu Oki, Japan, Earthquake of July 16,
performance-based design in earthquake geotechnical engineering, 2007, US Department of the Interior & US Geological Survey, Open File Report
Taormina, Italy, 28–30 May, 2012. 2007-1365, 2007.
[5] Argyroudis S, Monge O, Finazzi D, Pessina V. Vulnerability assessment of [27] Kaynia MA, Argyroudis S, Mayoral JM, Johansson J, Pitilakis K, Anastasiadis A.
lifelines and essential facilities (WP06): Methodological Handbook—Appendix D3.7: fragility functions for roadway system elements. Report for the
1: Roadway Transportation System, Risk-UE Final Report, Report no. GTR-RSK European Project SYNER-G: Systemic Seismic Vulnerability and Risk Analysis
0101-152av7, February 2003. for Buildings, Lifeline Networks and Infrastructures Safety Gain (FP7-ENV-
[6] Argyroudis S, Pitilakis K. Seismic fragility curves of shallow tunnels in alluvial 2009-1-244061), 2011.
deposits. Soil Dynamics and Earthquake Engineering 2012;35:1–12. [28] Ling HI, Leshchinsky D, Mohri Y, Wang JP. Earthquake response of reinforced
[7] Athanasopoulos-Zekkos A, Lamote K, Athanasopoulos GA. Use of EPS geofoam
segmental retaining walls backfilled with substantial percentage of fines. Journal
compressible inclusions for reducing the earthquake effects on yielding earth
of Geotechnical and Geoenvironmental Engineering ASCE 2012;138(8):934–44.
retaining structures. Soil Dynamics and Earthquake Engineering 2012;41:59–71.
[29] Maruyama Y, Yamazaki F, Mizuno K, Tsuchiya Y, Yogai H. Fragility curves for
[8] Bardet JP, Ichii K, Lin CH. EERA: a computer program for equivalent-linear
expressway embankments based on damage datasets after recent earthquakes
earthquake site response analyses of layered soil deposits, University of
in Japan. Soil Dynamics and Earthquake Engineering 2010;30:1158–67.
Southern California, Department of Civil Engineering, 2000, 40p.
[30] Matsuoka M. Estimation of earthquake ground motion of the 2007 Niigata-ken
[9] Bray J, Frost D, editors. Geo-Engineering Reconnaissance of the February 27,
2010 Maule, Chile Earthquake Version 2, GEER Association Report no. GEER- Chuetsu-oki, 2007. /http://staff.aist.go.jp/m.matsuoka/zoning2/20070716Niigata
022, 2010. ChuetsuOki/20070716.htmlS.
[10] Cubrinovski M, Green RA, Wotherspoon L, editors. Geotechnical Reconnais- [31] Mitoulis SA, Tegos IA, Sextos A. An alternative proposal for a ‘‘movable’’
sance of the 2011 Christchurch, New Zealand Earthquake, Version 1, GEER abutment for integral bridges. In: Proceedings of the first European con-
Association Report no. GEER-027, 2011. ference on earthquake engineering and seismology (1st ECEES), Geneva,
[11] Darendeli MB. Development of new family of normalized modulus reduction Switzerland, 2006, Paper No 1377.
and material damping curves, PhD thesis, University of Texas, Austin, 2001. [32] National Institute of Building Sciences (NIBS). HAZUS-MH: Users’s Manual
[12] Davis JC. Statistics and data analysis in geology. New York: John Wiley & and Technical Manuals. Report prepared for the Federal Emergency Manage-
Sons; 2002 p. 656. ment Agency, Washington, DC, 2004.
[13] EC8 Eurocode 8: Design of Structures for Earthquake Resistance, Brussels, [33] Nielson BG, DesRoches R. Seismic fragility methodology for highway bridges
Belgium, European Committee for Standardisation, The European Standard using a component level approach. Earthquake Engineering and Structural
EN 1998–1, 2004. Dymamics 2007;36:823–39.
[14] Elgamal A, Wilson P. Chapter 4: Full scale testing and simulation of seismic [34] Pinto PE, editor. Probabilistic methods for seismic assessment of existing
bridge abutment-backfill interaction, Special topics in earthquake geotech- structures. LESSLOSS Report No. 2007/06, Istituto Universitario di Studi
nical engineering, geotechnical, geological and earthquake engineering, Sakr Superiori di Pavia, IUSS Press, ISBN: 978-88-6198-010-5, 2007.
MA, Ansal A, editors, Springer, vol. 16, ISBN: 978-94-007-2059-6, p.109–127. [35] Pitilakis K, Alexoudi A, Argyroudis S, Monge O, Martin C. Chapter 9:
[15] Fotopoulou S, Pitilakis K. Vulnerability assessment of reinforced concrete Vulnerability and risk assessment of lifelines, Assessing and managing
buildings subjected to seismically triggered slow-moving earth slides. Land- earthquake risk, geo-scientific and engineering knowledge for earthquake
slides 2012. http://dx.doi.org/10.1007/s10346-012-0345-5 Published online: risk mitigation: developments, tools, techniques, Goula X, Oliveira CS, Roca A,
10 July 2012. Springer, ISBN 1-4020-3524-1. pp. 185–211.
[16] Gazetas G, Gerolymos N, Anastasopoulos I. Response of three Athens metro [36] Plaxis 2D, Reference Manual, version 9, 2011.
underground structures in the 1999 Parnitha earthquake. Soil Dynamics and [37] Power MS, Rosidi D, Kaneshiro J. vol. III Strawman: screening, evaluation, and
Earthquake Engineering 2005;25:617–33. retrofit design of tunnels, Report Draft, National Center for Earthquake
[17] Ghosh B, Bhattacharya S. Selection of appropriate input motion for founda- Engineering Research, Buffalo, New York, 1996.
tion design in seismic areas. In: Proceedings of the 14th world conference on [38] Priestley MJN, Seible F, Calvi GM. Seismic design and retrofit of bridges. New
earthquake engineering, Beijing, China, October 12–17, 2008. York: John Wiley & Sons; 0-471-57998-X.
[18] Gil LM, Hernandez E, De la Fuente P. Simplified transverse seismic analysis [39] Psarropoulos PN, Klonaris G, Gazetas. G. Seismic earth pressures on rigid and
of buried structures. Soil Dynamics and Earthquake Engineering
flexible retaining walls. Journal of Soil Dynamics and Earthquake Engineering
2001;21:735–40.
2005;25:795–809.
[19] Hashash YA, Hook J, Schmidt B, Chiang Yao J. Seismic design and analysis of
[40] Rossetto T, Elnashai A. Derivation of vulnerability functions for European-
underground structures. Tunnelling and Underground Space Technology
type RC structures based on observational data. Engineering Structures
2001;16:247–93.
2003;25:1241–63.
[20] Hardin BO. The nature of stress–strain behavior of soils. In: Proceedings of
[41] Schnabel PB, Lysmer J, Seed HB. SHAKE: a computer program for earthquake
the ASCE geotechnical engineering division specialty conference, ASCE, New
York, vol. 1, p. 3–90. response analysis of horizontally layered sites. Report No. UCB/EERC-72/12,
[21] Iervolino I, Galasso C, Cosenza E. REXEL: computer aided record selection for Earthquake Engineering Research Center, University of California, Berkeley,
code-based seismic structural analysis. Bulletin of Earthquake Engineering 1972, 102p.
2010;8:339–62. [42] Scotto di Santolo A, Penna A, Evangelista A, Kloukinas P, Mylonakis GE,
[22] Iervolino I, Galasso C, Paolucci R, Pacor F. Engineering ground motion record Simonelli AL, et al. Experimental investigation of dynamic behaviour of
selection in the ITalian ACcelerometric Archive. Bulletin of Earthquake cantilever retaining walls. In: Proceedings of the 15th world conference on
Engineering 2011;9:1761–78. earthquake engineering, Lisbon, Portugal, September 24–28, 2012.
[23] Imbsen RA, Roblee CJ, Yashinsky M, Berilgen MM, Toprak S. Impact on [43] Siddarthan R, Ganeshwara V, Kutter B, El-Desouky M, Whitman R. Seismic
highway structures. Earthquake Spectra 2000;16(S1):411–35. deformation of bar mat mechanically earth walls I: centrifuge tests. Journal
[24] Kakderi K, Pitilakis K. Seismic analysis and fragility curves of gravity water- of Geotechnical and Geoenvironmental Engineering 2004;130:14–25.
front structures. In: Proceedings of the 5th international conference on recent [44] Werner SD, Taylor CE, Cho S, Lavoie J-P, Huyck C, Eitzel C, et al. REDARS 2:
advances in geotechnical earthquake engineering and soil dynamics, San Methodology and Software for Seismic Risk Analysis of Highway Systems,
Diego, California, May 24–29, 2010, paper no. 6.04a. MCEER-06-SP08, 2006.