Earthquake Geotechnical Engineering for Protection and Development of

Environment and Constructions – Silvestri & Moraci (Eds)

© 2019 Associazione Geotecnica Italiana, Rome, Italy, ISBN 978-0-367-14328-2

Vulnerability assessment of buildings exposed to seismically-

induced landslides, liquefaction and tsunamis

K. Pitilakis, S. Fotopoulou & S. Karafagka

Aristotle University, Thessaloniki, Greece

ABSTRACT: There is an increasing requirement for effective evaluation, management and

mitigation of the seismic risk of buildings and infrastructures considering not only the seismic
ground shaking itself but also some of the most destructive collateral hazards associated with
earthquakes, namely landslides, liquefaction and tsunami. In this respect an analytical vulner-
ability assessment methodology is proposed for buildings subjected to earthquake induced
phenomena like landslides, liquefaction and tsunamis. The methodology is based on the use of
nonlinear static analyses of representative structural typologies of reinforced concrete (RC)
buildings. The proposed methodology results in the development of probabilistic fragility
curves and vulnerability indices as function of the permanent ground displacement or the
maximum ground acceleration for the landslide hazard, the differential displacements for the
liquefaction hazard while the inundation depth is selected as intensity measure for the tsunami
case. The proposed fragility functions could be used within a probabilistic risk assessment
framework to assess the vulnerability of RC buildings exposed to different seismically induced
hazards.

1 INTRODUCTION

Ground shaking is widely recognised as the primary cause of damage to structures, loss of life
and injuries due to earthquakes. However, experience gained from recent strong seismic
events has demonstrated the high vulnerability of buildings to earthquake related hazards
associated not only to the ground shaking itself but also to the induced phenomena, namely
landslides,
liquefaction and tsunami, resulting in severe physical damages and important economic and
societal losses. For instance, Bird and Bommer (2004) reported five catastrophic earthquake-
induced landslides, namely in Ardebil, Iran, 1997; El Salvador, January 2001; Luzon, Philip-
pines, 1990; Manjil, Iran, 1990 and Paez, Colombia, 1997, causing hundreds of fatalities and
frequently destroying entire villages. The earthquake sequences in Christchurch have been
associated with massive liquefaction damages (Cubrinovski 2013). More recently, Lai et al.
(2015) observed widespread liquefaction phenomena (lateral spreading and after-effects) on
structures and infrastructures caused by the main shock that hit the Emilia-Romagna Region
(2012) in Northern Italy. Moreover, recent seismically induced tsunami events have caused
extremely severe impact to buildings, infrastructures, population and economy (e.g. the 2004
Indian Ocean tsunami, the 2006 Java tsunami, the 2007 Solomon Islands tsunami, the 2009
South Pacific tsunami, the 2010 Chilean Tsunami in Dichato, and the 2011 Great East Japan
tsunami) highlighting the need for effective evaluation of the tsunami hazard and
vulnerability.
Despite their importance only recently the development of adequate vulnerability functions
for these induced hazards has attracted the interest of researchers. To this end, this study aims
at proposing a coherent analytical vulnerability assessment methodology for buildings sub-
jected to different earthquake induced phenomena, namely landslides, liquefaction and tsu-
namis. The methodology is based on the use of nonlinear static time-history parametric

282
analyses of representative structural typologies of reinforced concrete (RC) buildings. The
proposed methodology results in the development of probabilistic fragility curves and vulner-
ability indices that could be used within a probabilistic risk assessment framework to assess
the vulnerability of RC buildings exposed to the above different seismically induced hazards.

2 METHODOLOGY

The layout of a generic methodology for the vulnerability assessment of RC frame buildings
subjected to seismically induced phenomena is illustrated in Figure 1 involving a comprehen-
sive set of nonlinear parametric numerical computations of representative structural typolo-
gies of RC buildings and adequate statistical analysis.
For the cases of landslides and liquefaction the structures are subjected to different patterns
of permanent differential earthquake induced ground displacements (vertical and horizontal)
applied quasi-statically directly at their foundation system. For the tsunami case appropriate
combinations of static tsunami forces based on FEMA P646 (2008) recommendations for
gradually increasing tsunami inundation depths are applied. More details regarding the esti-
mation of the input histories applied for the different earthquake induced hazards are given in
the following sections.
Nonlinear static time history analyses of the selected representative structure typologies are
conducted using an appropriate finite element code, which allows both geometric nonlineari-
ties and material inelasticity to be captured. Distributed inelasticity elements are used that
may fully account for the spread of inelasticity along the element length and across the section
depth. The so-called “fiber approach” is used to represent the cross-section inelastic behaviour
where each fiber is associated with a uniaxial stress-strain relationship. The uniaxial nonlinear
constant confinement model that follows the constitutive relationship proposed by Mander
et al. (1988) and the cyclic rules proposed by Martinez-Rueda and Elnashai (1997) is employed
for the concrete material, while for the steel reinforcement a uniaxial bilinear stress-strain
model with kinematic strain hardening is utilized.
The selection of realistic damage states is fundamental in the proposed vulnerability assess-
ment methodology. Different damage mechanisms are examined, including a flexural damage

Figure 1. Proposed methodological framework.

283
of the building members and a shear failure of the columns. The maximum structural flexural
response (or engineering demand parameter EDP) is defined in terms of maximum material
strain calculated from the nonlinear static analysis. In addition to the flexural response of the
beam/column elements, the shear forces in the columns are also extracted from the nonlinear
static analyses and compared to the shear capacity of the column calculated according to
EC8-3 (CEN 2005). In this way, the possibility of local failure of the columns due to shear is
also examined, especially for the liquefaction and tsunami hazard cases where low code
moment resisting frame (MRF) structures, which have not been designed to withstand seismic
loads, have been analyzed. In the latter cases shear failure might occur prior to the attainment
of the flexural failure (Mwafy and Elnashai 2008).
As far as the flexural damage mechanism is concerned, four damage limit states (LS1, LS2,
LS3, LS4) are defined, describing the exceedance of slight, moderate, extensive and complete
structural damage of the building in terms of limit values of steel and concrete material strains
based on the available literature (e.g. Crowley et al. 2004, Bird et al. 2005, Fotopoulou et al.
2018). The selected damage states and the adopted limit state values for both high- and low-
code MRFs (given in bold) are presented in Table 1.
As shear failure of columns is also likely to occur, one damage state is assigned describing
the complete shear failure of the structures; it is defined as being attained when at least 50% of
the columns in a storey reach their shear capacity (Kappos et al. 2006, Celarec and Dolsek
2013).
The results of the nonlinear static time history analysis are used to derive fragility curves for
different damage states expressed as two-parameter lognormal distribution functions accord-
ing to Equation 1:


 i
f ðDSjIM Þ ¼ ½ lnðIM Þ  ln IMÞ =β ð1Þ

where, Φ is the standard normal cumulative distribution function, IM is the intensity meas-
ure properly selected for each hazard, IM and β are the median and log-standard deviation
values respectively. Therefore, for the development of fragility curves according to Equation 1,
two parameters, namely (IM) and β, need to be defined. The peak ground acceleration (PGA)
and permanent ground displacement (PGD) are considered as IMs for the landslide hazard;

Table 1. Structural damage state description and definition of damage limit states for “low” and “high”
code design RC MRF buildings (after Crowley et al. 2004).
Structural damage
band Description

None to slight Linear elastic response, flexural or shear type hairline cracks (<1.0 mm) in some
members, no yielding in any critical section
Limit State 1 (LS1): steel bar yielding
Moderate Member flexural strengths achieved, limited ductility developed, crack widths reach
1.0 mm, initiation of concrete spalling
Limit State 2 (LS2): εs=0.0125, εc=0.0045
Extensive Significant repair required to building, wide flexural or shear cracks, buckling of
longitudinal reinforcement may occur
Limit State 3 (LS3):
“High” code frames: εs=0.04, εc=0.010
“Low” code frames: εs=0.025, εc=0.006
Complete Repair of building not feasible either physically or economically, demolition
required, could be due to shear failure of vertical elements or excessive displacement
Limit State 4 (LS4):
“High” code frames: εs=0.06
“Low” code frames: εs=0.045

284
the maximum differential displacement is selected for the liquefaction hazard while the inun-
dation depth is considered as the most appropriate IM for the tsunami case. The proposed
IMs have shown to adequately correlate with structural deformation and damage. The log-
standard deviation parameter β describes the total dispersion related to each fragility curve. It
contains three components of uncertainty associated with the definition of the limit state value
βLS, the capacity of each structural type βc, and the demand βD (NIBS 2004). These are
defined either empirically (i.e. on the capacity and on the damage states) or analytically (i.e.
on the demand). A common β value, which ensures that the fragility curves do not intersect, is
assumed for all limit states.
Once the fragility curves of the considered building typologies subjected to the different
earthquake induced hazards are available, a vulnerability curve can be constructed, which
provides a unique damage or vulnerability index dm for each level of IM. It is defined accord-
ing to the following expression:
X4
dmj ¼ i¼1
Pij  di ð2Þ

where dmj is the damage index corresponding to the intensity level j, Pij is the discrete
damage probability for each damage state and di stands for the damage index at each of the
considered damage states. The damage index is a convenient measure to quantify the struc-
tural losses and finally to assess the risk at building or community level when properly aggre-
gated. It may be quantified in cost terms as the ratio of cost of repair to cost of replacement,
taking values from 0: no damage (cost of repair equals 0) to 1: complete damage (cost of
repair equals the cost of replacement). An application is provided below for liquefaction
movements.
The proposed methodological framework is specialized for each of the specific earthquake
induced hazards in the following sections.

3 LANDSLIDE VULNERABILITY

3.1 Vulnerability assessment method

The physical or structural vulnerability of a building is defined as the degree of loss, in terms
of percentage of structural damage, to a given element or set of elements subjected to a land-
slide hazard (e.g. Corominas et al. 2014). In this study the focus is on the seismically triggered
landslides and in particular the co-seismic slope ground movements in the absence of flow-
type massive landslides or lateral spreading after liquefaction. The essential of the method-
ology has been developed in the framework of the EU research project SAFELAND (www.
safeland.no) and has been published in a series of papers (e.g. Fotopoulou and Pitilakis 2013a;
2013b; Mavrouli et al. 2014).
In the following the methodology used to derive fragility curves for low-rise RC buildings
located next to the crest of a potentially unstable slope is briefly presented. Contrary to earth-
quake engineering, where the structure is subjected to ground shaking, now it is subjected to
forced differential displacement due to the earthquake-triggered landslide and subsequently to
structural distress and damage.
The method is principally based on numerical computations performed in two consecutive
steps and statistical analysis. First, the total and differential permanent displacements at the
foundation level of a reference building located next to the slopecrest are estimated using a
plane strain finite difference, non-linear dynamic slope model and different progressively
scaled real accelerograms recorded at outcropping conditions as input motion. Then, non-
linear static time history analyses of the building are employed to assess its response to the
differential ground deformation. More specifically, the derived differential displacement
demand at the foundation level computed in the previous step is applied quasi-statically at
one of the building supports.

285
 The results of the nonlinear static analysis (PGA or PGD- maximum strain pairs) of the
structure are statistically correlated with the corresponding predefined damage limit states
(see Table 1) to derive probabilistic fragility curves. Two different procedures have been
applied to estimate the fragility parameters (median and log-standard deviation) given the
simulated damage data. The first one is based on a regression analysis method (e.g. Cornell
et al. 2002; Argyroudis and Pitilakis 2012; Fotopoulou and Pitilakis 2017), whereas the second
is based on a purely statistical approach, i.e. the Maximum Likelihood Method (e.g. Shino-
zuka et al. 2000; Fotopoulou and Pitilakis 2013b).

3.2 Vulnerability assessment of RC buildings subjected to co-seismic slope displacements

An extensive parametric investigation has been performed for a reference building standing in the
vicinity of the crest based on the proposed method (Figure 2). In particular, 24 model configur-
ations were analysed that vary on the geometry of the slope (slope height Η=20m, 40m and inclin-
ation θ=15ο, 30ο, 45ο), the soil properties of the slope material (soft clayey (φ=15ο, c=5KPa) and
sand (φ=25ο, c=0KPa) soils, relatively stiff clayey (φ=25ο, c=10KPa) and sand (φ=36ο, c=0KPa)
soils and stiff clayey (φ=27ο, c=50KPa) and sandy (φ=44ο, c=10KPa) soils corresponding to soil
categories B (Vs=500m/s), C (Vs=250m/s) and D (Vs=150m/s) according to EC8 respectively) and
the relative position of the building with respect to the slope crest (L=3, 5 m). It is noticed that
the soil properties of the slope materials were selected to be consistent with the considered inclin-
ation slope angles to ensure static slope stability. To fulfil this, it was necessary to consider a non-
zero cohesion value for the steeper (45°) sandy soils (i.e. φ=44ο, c=10KPa). Steep slopes with very
soft soil were not analysed as the slope would be already unstable in static conditions. It is also
worth noting that only the cases that result in the highest susceptibility to landsliding were con-
sidered. For instance, slopes with small inclination (θ=15ο) on stiff soil were not investigated as
the resulting permanent deformation and consequently the building expected damage due to seis-
mically induced slope displacements would be negligible.
Finally, it is noted that dry soil conditions were considered in all models and the influence
of groundwater in altering the slope seismic response and consequently the expected structural

Figure 2. Idealized single bay-single storey RC frame building with flexible and stiff foundation system
and displacement loading pattern considered for the non-linear quasi-static analysis (top), Soil paramet-
ric dynamic model (bottom).

286
damage was not investigated at this stage without ignoring its importance in slope stability.
Details regarding the parametric soil models and the characteristics of the outcropping accel-
erograms can be found in Fotopoulou and Pitilakis (2013b). The considered reference building
is a simplified single bay-single storey RC bare frame structure with flexible foundation
system, which was found to be adequately representative of the performance of real low-rise
RC frame buildings designed with “high” seismic code provisions (e.g. Bird et al. 2005; Negu-
lescu and Foerster 2010).
Seven sets of fragility curves in terms of PGA and PGD at the slope area were finally pro-
posed for the different investigated models based on the most influential features, i.e. the
slope inclination in conjunction with the slope soil material as well as the slope height for the
45° inclined sand slopes. Tables 2 and 3 present the median (in terms of PGD) and the disper-
sions of the suggested curves when using the regression analysis method and the Maximum
Likelihood Method respectively. The differences on the fragility curve parameters when apply-
ing the two different approaches are due to the different assumptions adopted in each method
and clearly identify the influence of epistemic uncertainty on the fragility analysis. In particu-
lar, the two methods estimate medians in terms of PGD that are generally in good agreement
while the estimated β values are slightly higher for the regression analysis method.
The derived fragility curves can be used to assess the vulnerability of low-rise, “high” code
RC frame buildings on flexible foundations subjected to seismically induced slope displace-
ments for a variety of slope configurations and soil conditions. Similar curves generally associ-
ated with increased vulnerability values could be derived for “low-code” RC frame structures.

3.3 Combined vulnerability assessment of RC buildings to ground shaking and co-seismic slope
displacement
In the previous section, the evaluation of building’s vulnerability is carried out for the effect
of co-seismic slope displacement due to the co-seismic landslide ignoring possible initial

Table 2. Parameters of the proposed fragility functions in terms of PGD based on the regression ana-
lysis method
Median PGD (m)

Parametric models LS1 LS2 LS3 LS4 Dispersion β

sand_θ30 0.17 0.46 1.09 1.53 0.57

clay_θ30 0.24 0.81 1.97 2.69 0.56
sand_θ15 0.25 0.70 1.50 1.96 0.57
clay_θ15 0.33 1.07 2.62 3.61 0.55
sandy_θ45_Η20 0.07 0.22 0.57 0.80 0.55
sandy_θ45_Η40 0.05 0.16 0.45 0.67 0.50
clay_θ45 0.80 2.82 - - 0.73

Table 3. Parameters of the proposed fragility functions in terms of PGD based on the Maximum Likeli-
hood Method
Median PGD (m)

Parametric models LS1 LS2 LS3 LS4 Dispersion β

sand_θ30 0.15 0.39 0.87 1.64 0.43

clay_θ30 0.22 0.67 1.70 3.43 0.43
sand_θ15 0.24 0.60 1.29 2.19 0.40
clay_θ15 0.28 0.94 2.32 - 0.43
sandy_θ45_Η20 0.09 0.18 0.47 0.91 0.43
sandy_θ45_Η40 0.04 0.16 0.45 0.67 0.50
clay_θ45 0.77 1.72 - - 0.47

287
damage due to ground shaking. In this section, it is proposed to tackle the complex issue of
combined damages due to ground shaking and landslide hazard. In particular, the vulnerability
of a typical low rise, high code RC building standing next to the cliff of a 40m high and 45°
inclined sandy slope (φ=40ο, c=15KPa, Vs=400m/s) subjected to the combined effect of ground
shaking and permanent co-seismic ground deformation due to landslide is assessed. More details
regarding the adopted slope and structural models as well as to the analyses performed can be
found in Fotopoulou and Pitilakis (2017). Two different scenarios were considered to account
for the combined vulnerability due to ground shaking and landslide hazards:
I. The two hazards do not interact, i.e. the damage due to ground shaking is independent and
does not affect the damage due to earthquake induced landslide (NIBS, 2004) and
II. The two hazards do interact, i.e. the landslide hazard is conditioned on the earthquake that
is the initial trigger of the landslide.
Fragility curves were derived by combining the damage state probabilities due to the land-
slide (PL) and ground shaking (PGS). Thus, the combined probability (PCOMB) of exceeding a
given limit state i is given as:
h i h i h i
PCOMB LS LSi ¼ PGS LS LSi þ Plandslide  PL LS LSi
h i h i ð3Þ
 PGS LS LSi  Plandslide  PL LS LSi

where Plandslide is the probability of landslide occurrence given the earthquake. It is noted
that the probability of landslide occurrence (Plandslide) in scenario I was taken equal to 1.0 as
the two hazards are considered independent. It should be noted that although this consider-
ation is a rather “simplified” assumption, it is reasonably acceptable taking into account the
uncertainties and complexities involved in the interaction between the two modes of loading
and failure (ground shaking and permanent slope displacement).
Scenario II is more realistic as some strong shaking is generally expected to take place
before the permanent displacements increase sufficiently for landslide to initiate. The estima-
tion of the probability of landslide occurrence involves a great deal of uncertainty. To this
end, in this study the probability of landslide occurrence (Plandslide) in scenario II was esti-
mated equal to 0.75 based on the maximum horizontal displacements computed from the total
number of the soil dynamic analyses assuming a threshold (critical) horizontal displacement
of 7.0 cm to initiate the landslide movement. It is noted that the selected threshold displace-
ment value pertains principally to thin, shallow landslides, as the ones expected for the stiff
sandy slope soil materials (Fotopoulou and Pitilakis 2017). Making use of the above equation
for the given structure located at different distances from the slope crest, i.e. 3m, 5m and 10m,
the combined probabilities for each damage limit state due to occurrence of ground shaking
or seismically induced landslide and the corresponding fragility curves were calculated for
scenarios I and II (Figure 3).
From the results of this case study it has been shown (Fotopoulou and Pitilakis 2017)
that for scenario I, where the two hazards are considered independent, the vulnerability of
the building standing in the vicinity of the slope crest exposed to the combined effect of
ground shaking (including the topographic effect) and earthquake induced slope displace-
ment is always higher compared to the vulnerability of the building where each mode of
failure is considered separately. It is also shown that for the same PGA the building stand-
ing at 3m and 5m from the crest would suffer higher combined extensive and complete dam-
ages due to combined ground shaking and landslide compared to the vulnerability of the
same building located at 10m from the crest. On the other hand, scenario II, which assumes
that the landslide hazard is conditioned on the earthquake, is less conservative compared to
scenario I, resulting to lower vulnerability values for the post-yield limit states (LS2, LS3
and LS4) for the building resting at 3m and 5m from the crest. The vulnerability of the
building located at 10m from the crest and exposed to the combined effect of ground shak-
ing and earthquake induced slope displacement is almost the same for all damage states

288
Figure 3. Fragility curves in terms of PGA for the building located at 3m, 5m and 10m from the crest
subjected to the combined effect of ground shaking and seismically induced landslide for scenarios I and II.

irrespective of the scenario. Thus, as we move away from the slope crest the effect of the
landslide hazard is less important and the influence of the scenario (I or II) becomes
insignificant.

4 VULNERABILITY DUE TO LIQUEFACTION DISPLACEMENTS

4.1 Vulnerability assessment method

The proposed vulnerability assessment method described in Fotopoulou et al. (2018) is shortly
described in this paper. In the proposed method the focus is on the differential component of
liquefaction-induced ground deformation, which commonly occurs due to the heterogeneity in
soil stiffness and stratigraphy both laterally and with depth and are considered as the major
cause of damage (Bird et al. 2005, 2006). Within an earthquake risk assessment framework,
the evaluation of liquefaction demand in terms of permanent ground displacements is far
from straightforward (Bird et al 2006). A number of methodologies are available for the pre-
diction of absolute settlements and lateral movements due to liquefaction, most of which rely
principally on empirical data (e.g. Tokimatsu and Seed 1987, Ishihara and Yoshimine 1992,
Youd et al. 2002, Zhang and Zhao 2005). They are generally accepted to be accurate only
within a factor of 2 or 3 and their ability to predict displacements tends to reduce for small-to-
moderate (0.3–0.75 m) deformations.

289
 To illustrate the proposed vulnerability assessment methodology three moment resisting
frame (MRF) RC buildings (without masonry infills) of various heights have been selected as
reference structures. They have been designed with a low seismic code according to the 1959
Greek seismic regulations (‘Royal Decree’ of 1959) (Kappos et al. 2006) in which the ductility
and the dynamic features of the structures are neglected. The first building is a two-storey
three-bay frame structure that represents low-rise (1-3 stories) MRF RC buildings. The
second one is a four-story three-bay frame structure that is considered representative of mid-
rise (4-7 stories) MRF RC buildings while the third one is a nine-story three-bay frame model
that is considered typical of high-rise (8+ stories) MRF RC buildings. It is reasonably
assumed that the selected poorly designed buildings are likely to have poorly designed founda-
tions with isolated footings allowing columns to move differentially. In this case, the various
modes of differential deformation can produce structural damage (e.g., in terms of cracks) to
the building members (Bird et al. 2005, 2006); hence the applied differential displacement
(settlement or lateral spreading) may be reasonably assumed equal to the forced displacements
at the bottom of the columns. The main characteristics and representative cross-sections and
the floor plan of the studied RC building typologies can be found in Fotopoulou et al. (2018).
Nonlinear static analyses of representative RC buildings are performed to assess the struc-
ture response and fragility due to the liquefaction-induced differential displacement using an
appropriate finite element code. Different combinations of differential displacements are
imposed as quasi-static loads at the structure supports, considering only the vertical compo-
nent of the differential displacement (i.e. to represent settlements in level ground) or taking
also into account the horizontal differential displacement with a vertical component (i.e. to
represent lateral spreading in gently sloping ground). In the latter case, two different cases are
analysed: (i) the horizontal component is assumed equal to the vertical one and (ii) the hori-
zontal component is assumed equal to twice the vertical one. For each of the previous differ-
ential displacement profiles the maximum differential displacement is assumed to be applied
either at the middle or edge column at the foundation level while a triangular distribution of
the differential displacement is considered in all cases.
Large differences in the maximum applied differential settlements and the maximum com-
puted strain values are expected depending on whether the maximum differential settlement
is applied at the edge or the middle column. In particular, in the former case, the building is
subjected to significantly higher maximum differential settlements until structural collapse
compared to the latter. In addition, it is observed that for the same level of maximum differ-
ential settlement (e.g. 0.5m) imposed at the edge or the middle column, the expected response
of the structure might change dramatically, implying that depending on the differential dis-
placement profile, the structure may either be in the elastic range or suffer significant dam-
ages. These observations are principally due to the different deformation modes and failure
mechanisms taking place in each case. For instance, for the 4-storey RC frame building sub-
jected to differential settlements with the maximum differential settlement applied at the
right edge column (Figure 4), deformation is initially expected to be concentrated at the
bottom of the left ground floor column rather than to the beams, with the higher storeys
rotating as rigid bodies. This failure mechanism, where the formation of plastic hinges takes
place at the base of the building, is due to the fact that the columns are free to rotate laterally
and therefore act as cantilevers in simple bending. This mechanism has also been observed in
Bird et al. (2005) for the case of a two-storey MRF building. On the other hand, when the
maximum differential settlement is applied at a middle column of the structure, as there is a
restraint to keep the columns vertical, a failure mechanism involving deformation and
damage of the beams and consequently damage of the columns occurs, resulting in much
higher deformation levels. In the following, taking into account this important observation
on the failure mechanisms as well as the significant differences in the structural response
depending on the considered displacement profile and recognizing the difficulty to predict
where the maximum differential displacement is expected to occur, fragility curves will be
derived for the most adverse cases where the maximum differential displacement is applied at
the middle column.

290
Figure 4. Failure mechanisms for the 4-storey RC frame building subjected to settlements with the max-
imum differential settlement equal to 1.35 m applied at the edge (left) column and equal to 0.15 m applied
at the middle (right) column (the green colour indicates steel bar yielding members, while the red one
post-yield damaged members).

4.2 Derivation of fragility curves and vulnerability index

The results of the nonlinear static analysis, in terms of pairs of differential displacement –
maximum material strain and differential displacement – shear forces to account for the max-
imum flexural and shear response respectively of the structures, are statistically correlated
with the corresponding predefined damage limit states (see for example Table 1 for the flex-
ural damage state description) to derive probabilistic fragility curves. A regression analysis
method is applied to estimate the fragility parameters (median m and log-standard deviation
beta) given the simulated damage data. In particular, a quadratic fit of the logarithms of the
differential displacement- maximum strain simulated data, which minimizes the regression
residuals, is adopted.
Figure 5 illustrates the derived sets of fragility curves for the 2-storey, 4-storey and 9-storey
MRF RC buildings subjected to differential settlements (left) and lateral spreading (right)
respectively which account for the flexural failure mechanism. It is noted that for the cases
where the structures are subjected to lateral spreading with the maximum differential displace-
ment vector applied at the middle column, the minimum of the two median differential displace-
ments estimated for the horizontal component of the differential displacement being equal to
and twice the vertical one is considered for each damage state. Thus, taking also into account
the difficulty of actually knowing the proportion of the horizontal component with respect to
the vertical one and to be on the safe side, instead of two different sets of curves, an upper
bound set of fragility curves is finally developed for each building for the lateral spreading case.
It is seen that for the studied typical low code MRF RC buildings the total height of the
structure may considerably influence its vulnerability to the liquefaction-induced differential
displacement, with the studied high-rise building being more vulnerable compared to the low-
and mid-rise ones. This last observation is more pronounced for the lateral spreading cases. It
is also observed that for the studied buildings their vulnerability is generally increased when
they are subjected only to liquefaction settlements compared to the case where the applied dif-
ferential displacement has also a horizontal component due to lateral spreading with the
increase being more noticeable for the complete damage state. This could be due to the fact
that in the latter case damage is generally more distributed at the building members while in
the former case damage is mainly concentrated at beams and columns at one side of the build-
ing resulting in a higher deformation demand and hence higher vulnerability.
Figure 6 presents the comparison between the new fragility curves for the complete damage
state due to shear failure (SF) with the corresponding ones due to flexural failure (FF). It is
shown that complete shear failure is anticipated only for the 2-storey and 9-storey buildings
subjected to lateral spreading. For the 4-storey building subjected to lateral spreading, even
though some shear failures in the columns distributed at the different floors are observed,
complete shear failure (as defined in this study) does not occur. It is also noted that for the

291
Figure 5. Fragility curves for the 2-storey, 4-storey and 9-storey frame buildings subjected to differen-
tial settlements and lateral spreading due to liquefaction.

studied low-code high-rise building shear failure would be the prevailing damage mechanism
of the structure while for the low-rise building the flexural failure is expected to come first.
This could be probably due to the P-delta effects that may adversely affect the performance of
the high-rise frame building at high deformation levels amplifying the seismic shear demand
and therefore resulting in a more fragile (local) failure mechanism (e.g. Fenwick et al. 1992).
Once the probabilities of exceeding the specified damage states are estimated, a damage (or
vulnerability) index dm may be evaluated for each differential displacement according to equa-
tion 2. The central value of the damage index for each damage state (i.e. di) is assumed equal
to 0.05, 0.20, 0.45 and 0.80 for the LS1, LS2, LS3 and LS4 respectively according to Kappos
et al. (2006). A vulnerability curve is then constructed for each structural typology which pro-
vides a unique damage or vulnerability index dm for each level of differential displacement.
Figure 7 illustrates the derived vulnerability curves for the 2-storey, 4-storey and 9-storey
MRF RC buildings subjected to differential settlements and lateral spreading respectively,
which include also the possibility of shear failure. With the development of these curves, losses
in the RC structural members of the buildings are accounted for. However, due to the fact
that the cost of the RC structural system totals less than the cost of a (new) building (which
also includes non-structural members), the computed damage indices could then be multiplied
by an empirically derived reductive coefficient to derive the final global damage index. For
example, according to Kappos et al. (2006) for the low, mid and high-rise residential RC
MRF buildings this coefficient varies from 0.33 to 0.38 (values that have been calibrated

292
Figure 6. Fragility curves for the complete damage state due to flexural (FF) or shear (SF) failure for
the 2-storey (left) and 9-storey (right) buildings subjected to lateral spreading.

Figure 7. Vulnerability curves for the 2-storey, 4-storey and 9-storey frame buildings subjected to differ-
ential settlements (left) and lateral spreading (right) due to liquefaction.

against cost of damage data from Greece). Thus, it is implicitly shown that the total cost of
non-structural damage (e.g. damage to ceilings, mechanical and electrical equipment that is an
integral part of the structure, piping and elevators, partitions, exterior walls, ornamentation,
glass) may be higher compared to the cost of repairs required by the RC structural system
itself. If an average replacement cost is assumed for the considered low-code RC buildings
then the final global damage index multiplied with these values returns the total cost of repairs
in each building due to structural and non-structural damage.

5 VULNERABILITY DUE TO TSUNAMI LOADING

Inspired from the methods used for the development of seismic fragility curves for ordinary
buildings, a methodology to construct fragility curves and assess the vulnerability of buildings
subjected to tsunami forces presented initially in Karafagka et al. (2018) is shortly discussed
herein. Low-code moment resisting frame (MRF) and dual reinforced concrete (RC) buildings
of various heights with and without masonry infills have been considered. All models used in
the analysis assume fixed-base conditions. Other co-seismic effects like scouring, liquefaction
or effects of a compliant foundation are not taken into account. Moreover, in this single-risk
assessment approach, far-source generated tsunami is considered, where the epicentre of the
earthquake is assumed to be at long distances from the structure and hence it is assumed that
the structure has not sustained any initial damage prior to the tsunami due to ground shaking.
Tsunami loading is determined based on FEMA P646 (2008) recommendations and proper
engineering judgment. Each structure is subjected to buoyant, hydrostatic and hydrodynamic
forces combined with forces due to debris, all of which constitute tsunami load effects. The
considered combinations of tsunami forces differ for the various building typologies whereas
the amplitude of the resultant force increases with increasing tsunami inundation depth. The
analyses are performed for different levels of inundation depth (at least 20 levels), varying

293
Figure 8. Tsunami loading conditions for the MRF 4-storey (a) bare-frame where d is lower than the
height of the first floor, (b) bare-frame, where d is higher than the height of the first floor (uplift forces
are applicable) and the (c) infilled RC building (where Fh, Fb, Fd, Fs, Fi and Fu are the hydrostatic, buoy-
ant, hydrodynamic, impulsive, debris impact and uplift tsunami-induced force components respectively).

from very small values (e.g. h=0.5m), which generally result in negligible structural damage to
large ones (e.g. h=10m), which may lead to significant structural damages and potential col-
lapse. Indicatively all possible combinations of tsunami loading for the analysed 4-storey MRF
(bare and infilled) building are presented in Figure 8. The computed forces are then directly
applied as input time-variant static loads to appropriate two-dimensional nonlinear structural
models for gradually increasing inundation depths and the structures’ response in terms of
material strain (i.e. the EDP) for the different statically applied tsunami loads is estimated. Flex-
ural and shear (the attainment of shear capacity in a structural member) failures are considered
and the structures demand in terms of shear forces (i.e. EDP) is estimated. One damage state is
defined in the latter case describing the complete shear failure of the structures.
To minimize the uncertainties related to the definition of damage limit states, tsunami non-
linear static analyses are performed and appropriate tsunami capacity curves are derived for
the considered structures. Structural limit states are defined on tsunami capacity curves in
terms of threshold values of material strain. Figure 9 illustrates the definition of limit states
on the corresponding tsunami capacity curves derived from the tsunami nonlinear static ana-
lyses for the considered 2-storey RC MRF bare and infilled structure. Similar tsunami

Figure 9. Limit states definition on tsunami capacity curves for the 2-storey RC MRF bare and infilled
structure.

294
capacity curves have been derived for the other structural typologies (Karafagka et al. 2018).
It is noted that the tsunami capacity curves are not extracted from a single nonlinear static
analysis as for a seismic capacity curve (derived from a pushover analysis) but from the total
number of the tsunami nonlinear static time history analyses.
The results of the nonlinear numerical analysis (inundation depth or shear stress- steel
strain values) are used to derive log-normally distributed fragility curves based on the regres-
sion analysis method. The derived curves describe the probability of exceeding a certain limit
state of the structure as a function of the inundation depth (Figure 10).
A general observation is that the high-rise RC buildings present lower vulnerability compared
to the low-rise ones, which is in accordance with the recent empirical observation (Suppasri
et al. 2015), where buildings taller than two stories were found more resistant than buildings
with one or two stories under the same inundation depth. It is also shown that the low-rise and
midrise buildings with infills are more vulnerable compared to the corresponding models with
bareframes. This trend also holds true for the high-rise MRF for the exceedance of none to
slight and moderate damage. It seems that the existence of infill walls plays an important role

Figure 10. Tsunami fragility curves for the different RC building typologies.

295
increasing the vulnerability in case of low-rise buildings. This is due to the increase of their
deformation demand and it is in accordance with the FEMA guideline, which recommends the
design of vertical evacuation buildings with breakaway walls or open construction in the lower
levels to allow water to pass through with minimal resistance. In contrast, when extensive or
complete damage of the high-rise MRF structures is anticipated, the bare frame is expected to
sustain larger damages in comparison with the corresponding infilled one. This could be due to
the fact that, for the high-rise building, the infills of the upper floors are less affected by the
tsunami waves making the remaining resistance of the infilled building higher compared to the
bare frame for extensive and complete damage states. Moreover, it is seen that the low-rise dual
models are more vulnerable compared to the corresponding MRFs. The latter could be related
to the concentration of large tsunami forces in shear walls. This is the reason why FEMA
recommends for the design of vertical evacuation structures, shear walls be oriented parallel to
the anticipated direction of tsunami flow to reduce associated tsunami forces. An important
observation is that the shear failure is the prevailing failure mechanism of the structure for all
the MRF RC buildings with bare frames, regardless of their height. Moreover, the infilled high-
rise MRF and low-rise dual models also collapse due to shear. Those results indicate that the
consideration of masonry infills may have a significant impact on the response and finally on
the failure mechanism of the structures (i.e. flexural or shear).

6 CONCLUSIONS

An analytical vulnerability assessment methodology has been presented for RC buildings sub-
jected to some of the most destructive earthquake induced phenomena, like landslides, liquefac-
tion and tsunamis. The methodology was based on the use of nonlinear static analyses of
representative structural typologies of reinforced concrete (RC) buildings and adequate statis-
tical analysis. The proposed methodology results in the development of probabilistic fragility
curves for different structural damage states as well as vulnerability indices, which could be used
in a probabilistic risk assessment framework to assess the vulnerability and risk of typical RC
buildings in Greece and in Southern Europe exposed to different seismically induced hazards.
The methodology presented herein offers a general framework for assessing the vulnerability of
structures subjected to seismically induced phenomena like landslides, liquefaction and tsunami.
Comparison with observed damage data is also necessary. Among other future developments
should be the study of many more structural typologies, the consideration of aging and soil-
foundation structure effects, and more cases describing the complexity of induced phenomena.
Further extension of the methodology to account for multiple induced hazards, accounting also
for the uncertainties in the computation of the considered intensity measures, is also pertinent.

REFERENCES

Argyroudis, S. & Pitilakis, K. 2012. Seismic fragility curves of shallow tunnels in alluvial deposits. Soil
Dyn Earthq Eng 35:1–12.
Bird, J.F. & Bommer, J.J. 2004. Earthquake losses due to ground failure. Engineering Geology 75
(2):147–179.
Bird, J.F., Bommer, J.J., Crowley, H., Pinho, R. 2006. Modelling liquefaction-induced building damage
in earthquake loss estimation. Soil Dyn Earthq Eng 26(1):15–30.
Bird, J.F., Crowley, H., Pinho, R., Bommer, J.J. 2005. Assessment of building response to liquefaction-
induced differential ground deformation. Bull NZ Nat Soc Earthq Eng 38(4):215–234.
Celarec, D. & Dolsek, M. 2013. Practice-oriented probabilistic seismic performance assessment of infilled
frames with consideration of shear failure of columns. Earthq Eng Struct D 42:1339–1360.
CEN, 2005. EN 1998-3. Eurocode 8: Design of structures for earthquake resistance, Part 3: Assessment
and retrofitting of buildings. European Committee for Standardization, Brussels.
Cornell, C.A., Jalayer, F., Hamburger, R.O., Foutch, D.A. 2002. Probabilistic basis for 2000 SAC federal
emergency management agency steel moment frame guidelines. J of Struct Eng 128: 526–533.

296
Corominas, J., Van Westen, C., Frattini, P., Cascini, L., Malet, J.-P., Fotopoulou, S., Catani, F., Van
Den Eeckhaut, M., Mavrouli, O., Agliardi, F., Pitilakis, K., Winter, M.G., Pastor, M., Ferlisi, S.,
Tofani, V., Hervás, J., Smith, J.T. 2014. Recommendations for the quantitative assessment of land-
slide risk. Bull Eng Geol Environ 73: 209–263.
Crowley, H., Pinho, R., Bommer, J.J. 2004. A probabilistic displacement-based vulnerability assessment
procedure for earthquake loss estimation. Bul. Earthq Eng 2(2):173–219.
Cubrinovski, M. 2013. Liquefaction-Induced Damage in the 2010-2011 Christchurch (New Zealand)
Earthquakes. In Proceedings of the 7th International Conference on Case Histories in Geotechnical
Engineering, 29 Apr–4 May, Chicago, Illinois.
FEMA, 2008. Guidelines for Design of Structures for Vertical Evacuation from Tsunamis. FEMA P646
prepared by Applied Technology Council for Federal Emergency Management Agency, June, Wash-
ington, D.C.
Fenwick, R.C., Davidson, B.J., Chung, B.T. 1992. P-delta actions in seismic resistant structures. Bull NZ
Nat Soc Earthq Eng 25:56–69.
Fotopoulou, S.& Pitilakis, K. 2013a. Vulnerability assessment of reinforced concrete buildings subjected
to seismically triggered slow-moving earth slides. Landslides 10(5): 563–582.
Fotopoulou, S.& Pitilakis, K. 2013b. Fragility curves for reinforced concrete buildings to seismically trig-
gered slow-moving slides. Soil Dyn Earthq Eng 48: 143–161.
Fotopoulou, S.& Pitilakis, K. 2017. Vulnerability assessment of reinforced concrete buildings at precar-
ious slopes subjected to combined ground shaking and earthquake induced landslide. Soil Dyn Earthq
Eng 93:84–98.
Fotopoulou, S., Karafagka S., Pitilakis, K. 2018. Vulnerability assessment of low-code reinforced con-
crete frame buildings subjected to liquefaction-induced differential displacements, Soil Dyn Earthq
Eng 110: 173–184.
Kappos, A.J., Panagopoulos, G., Panagiotopoulos, C., Penelis G. 2006. A hybrid method for the vulner-
ability assessment of R/C and URM buildings. B Earthq Eng 4:391–419.
Karafagka, S., Fotopoulou, S., Pitilakis, K. 2018. Analytical tsunami fragility curves for seaport RC
buildings and steel light frame warehouses, Soil Dyn Earthq Eng 112: 118–137.
Lai, C.G., Bozzoni, F., Mangriotis, M.-D., Martinelli, M. 2015. Soil liquefaction during the May 20,
2012 M5.9 Emilia earthquake, Northern Italy: field reconnaissance and post-event assessment. Earthq
Spectra 31(4):2351–2373.
Mander, J.B., Priestley, M.J.N., Park, R. 1988. .Theoretical stress-strain model for confined concrete, J
Struct Eng 114(8):1804–1826.
Martinez-Rueda, J.E. & Elnashai, A.S. 1997. Confined concrete model under cyclic load. Materials and
Structures 30(3):139–147.
Mavrouli, O., Fotopoulou, S., Pitilakis, K., Zuccaro, G., Corominas, J., Santo, A., Cacace, F., De Gre-
gorio, D., Di Crescenzo, G., Foerster, E., Ulrich, T. 2014. Vulnerability assessment for reinforced con-
crete buildings exposed to landslides, Bull. Eng. Geol. Environ. 73: 265–289.
Mwafy, A., Elnashai, A. 2008. Importance of shear assessment of concrete structures detailed to different
capacity design requirements. Eng Struct 30(6):1590–1604.
National Institute of Building Sciences, 2004. Direct Physical Damage - General Building Stock. HAZUS-
MH Technical manual, Chapter 5. Federal Emergency Management Agency, Washington D.C.
Negulescu, C. & Foerster, E. 2010. Parametric studies and quantitative assessment of the vulnerability of
a RC frame building exposed to differential settlements, Nat Hazards Earth Syst Sci 10: 1781–1792.
SeismoSoft, 2016. SeismoStruct – A computer program for static and dynamic nonlinear analysis of
framed structures, (online): Available from URL: www.seismosoft.com
Shinozuka, M., Feng M.Q., Lee J., Naganuma T. 2000. Statistical analysis of fragility curves, J. Eng.
Mech.-ASCE 126: 1224–1231.
Suppasri, A., Mas, E., Charvet, I., Gunasekera, R., Imai, K., Fukutani, Y., Abe, Y., Imamura, F. 2013.
Building damage characteristics based on surveyed data and fragility curves of the 2011 Great East
Japan tsunami. Nat Hazards 66:319–341.
Tokimatsu, K. & Seed, H.B. 1987. Evaluation of settlements in sands due to earthquake shaking. ASCE
J Geotech Geoenviron Eng 113(8):861–878.
Ishihara, K., Yoshimine, M. 1992. Evaluation of settlements in sand deposits following liquefaction
during earthquakes. Soils Found 32(1):173–188.
Youd, T.L., Hansen, C.M., Bartlett, S.F. 2002. Revised multilinear regression equations for prediction of
lateral spread displacement. ASCE J Geotech Geoenviron Eng 28(12):1007–1017.
Zhang, J. & Zhao, J.X., 2005. Empirical models for estimating liquefaction-induced lateral spread dis-
placement. Soil Dyn Earthq Eng 25:439–450

297