This article was downloaded by: [York University Libraries]

On: 10 August 2014, At: 22:54

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Structure and Infrastructure Engineering:

Maintenance, Management, Life-Cycle Design and
Performance
Publication details, including instructions for authors and subscription information:
[Link]

Seismic risk assessment of reinforced concrete bridges

in flood-prone regions
a a
Swagata Banerjee & Gautham Ganesh Prasad
a
Department of Civil and Environmental Engineering , The Pennsylvania State University ,
University Park , PA , 16802 , USA
Published online: 11 Jan 2012.

To cite this article: Swagata Banerjee & Gautham Ganesh Prasad (2013) Seismic risk assessment of reinforced concrete
bridges in flood-prone regions, Structure and Infrastructure Engineering: Maintenance, Management, Life-Cycle Design and
Performance, 9:9, 952-968, DOI: 10.1080/15732479.2011.649292

To link to this article: [Link]

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
[Link]/page/terms-and-conditions
 Structure and Infrastructure Engineering
Vol. 9, No. 9, September 2013, 952–968

Seismic risk assessment of reinforced concrete bridges in flood-prone regions

Swagata Banerjee* and Gautham Ganesh Prasad
Department of Civil and Environmental Engineering, The Pennsylvania State University, University Park,
PA 16802, USA
(Received 1 July 2011; final version received 14 September 2011; published online 11 January 2012)

This article presents a conditional seismic risk evaluation framework of bridges located in seismically active flood-
prone regions. Flood-induced bridge scour causes loss of lateral support at bridge foundations and thus the effect of
seismic hazard on bridge performance gets amplified. Two example reinforced concrete bridges located in
Sacramento County in California are considered. The regional multihazard scenario is characterised by combining
scour resulted from regional flood events of different intensities with a suite of earthquake ground motions that
Downloaded by [York University Libraries] at 22:54 10 August 2014

represents regional seismicity. Uncertainties in the hazard models are discussed and their influences on bridge
performance are investigated. A separate set of analysis is performed to evaluate the bridge performance only under
earthquake ground motions. Seismic fragility curves and risk curves for the example bridges are generated. Result
shows higher seismic risk of bridges when the impact of regional flood hazard on bridges is considered in the analysis
framework. This suggests the use of a combined seismic and flood hazard model for reliable seismic risk evaluation
of bridges located in flood-prone regions.
Keywords: concrete bridges; seismic effect; risk and probability analysis; sensitivity analysis; floods

knowledge-base is currently available. NCHRP Report

1. Introduction 489 (Ghosn et al. 2003) calculated reliability indices of
The present state-of-the-art practice of bridge engineer- bridges subjected to various combinations of extreme
ing considers discrete hazard models to evaluate the events (e.g. earthquake, wind, scour and vessel colli-
performance of bridges. Consequently, the seismic loss sion). The service life of a bridge was divided into
estimation methodologies and risk mitigation techni- several equal time intervals DT. These intervals were
ques for bridges are developed based on their failure decided based on the duration and occurrence of
probabilities only under regional seismic hazard. independent extreme loads expected within the bridge
However, discrete seismic hazard model is not adequate service life. Probabilities of occurrence of possible
for reliable seismic risk evaluation of bridges located in extreme events in time interval DT were calculated and
seismically active flood-prone regions. Flood-induced combined to obtain total load for the duration DT.
soil erosion, commonly known as scour, causes loss of This methodology cannot be applied for load combi-
lateral support at bridge foundations. This imposes nations involving bridge scour, as scour itself does not
additional flexibility to bridges which amplifies the represent a load rather it is a consequence of flood
adverse effect of seismic ground motions on bridge hazard. The loss of lateral support at bridge founda-
performance. Hence, regional flood events may sig- tions due to flood-induced scour contributes towards
nificantly increase the seismic risks associated with the amplification of the adverse effect of other natural
bridges and thus should be considered in their seismic hazards on bridges. Therefore, load combination or
risk evaluation framework. Such a conditional risk load factor design as proposed in NCHRP Report 489
evaluation will closely capture the safety and service- (Ghosn et al. 2003) cannot provide a reliable estima-
ability of bridges under multiple natural hazards such tion of bridge performance under any natural hazard
as earthquake and flood and will eventually aid the in the presence of flood-induced scour. Rigorous
reliable prediction of seismic performance of regional region-based analysis is required for this purpose.
highway transportation systems. The present study develops a conditional seismic
There is a growing recognition for the importance risk evaluation framework that integrates the impacts
of considering multiple regional hazard scenarios for of regional seismic and flood hazards on bridge
risk assessment of bridges (MCEER-AEI 2007, Alam- performance. Note that the joint probability of
palli and Ettouney 2008); however, no comprehensive occurrence of earthquake and flood within the service

*Corresponding author. Email: swagata@[Link]

ISSN 1573-2479 print/ISSN 1744-8980 online

Ó 2013 Taylor & Francis
[Link]
[Link]
 Structure and Infrastructure Engineering 953

life of a bridge is relatively small. Nevertheless, past hazard. Uncertainties in regional hazard models are
experience indicates that one natural event can occur considered to examine their influence on bridge seismic
just after another (even before the aftermath of the first behaviour. A sensitivity study is performed for this
event is taken care of). For example, an earthquake of purpose to identify major uncertain parameters in-
magnitude 4.5 struck the state of Washington on 30 volved in the calculation of scour depth. Time history
January 2009. This seismic event occurred within three analyses of the example bridges with and without scour
weeks after the occurrence of a major flood event in the are performed. Based on the results, bridge seismic
same region. Such successive occurrences of extreme damage is categorised in four different damage states
events can significantly enhance bridge seismic vulner- namely minor, moderate, major and complete collapse.
ability from that under discrete hazard events. To Seismic fragility curves are developed for all these
measure this enhanced vulnerability under multihazard damage states. These curves show the probability of
condition, two sets of analyses are performed (a) bridge failure under certain intensity of ground motion
considering regional seismic hazard only and (b) in the presence and absence of flood-induced scour.
considering regional seismic hazard in the presence of Developed fragility curves are further used to generate
scour expected from regional flood hazards (Figure 1). seismic risk curves that represent the expected risks of
Two example reinforced concrete bridges with different bridges under regional seismic hazard in the presence
Downloaded by [York University Libraries] at 22:54 10 August 2014

span numbers are analysed for each of the above of flood induced scour.
scenarios. Sacramento County in California is chosen
as the example bridge site. It is assumed that prior to
any earthquake event the example bridges are pre-
2. Regional hazard models and uncertainties
exposed to regional flood hazard that caused scour at
bridge piers. 2.1. Seismic and flood hazards of the study region
The conditional (i.e. the bridge foundations are Identification of region-specific seismic and flood
pre-exposed to flood induced scour) seismic risk hazard levels is important because structural safety
evaluation framework presented in this article involves depends on maximum demands from multiple hazards.
(i) modelling of seismic and flood hazard of the study The seismic and flood hazard maps provided by the
region, (ii) estimation of flood-induced scour at bridge United States Geological Survey (USGS) indicate that
piers based on bridge geometry, subsurface condition the states of California, Washington and partly
and flood hazard levels, (iii) seismic analysis of Oregon in the Western US and the New Madrid
example bridges in the absence and presence of Seismic Zone in the Eastern US are regions with high
flood-induced scour and (iv) evaluation of seismic seismic and flood hazards. Bridges located in these
risk under different flood conditions. The multihazard regions are equally exposed to both of these natural
effect on the example bridges is assessed by combining hazards. Failure of a large number of highway bridges
flood-induced scour depths with a suite of earthquake is observed in California over the last four decades due
ground motions that represents the regional seismic to the 1971 San Fernando, 1989 Loma Prieta and 1994
Northridge earthquakes. These extreme events signifi-
cantly disrupted the normal functionality of the
regional highway transportation networks in Califor-
nia. Besides seismic events, several damaging flood
events are also recorded in this region. The 1995
California flood (total casualty $1.8 million) and the
1997 Northern California flood (total casualty $35
million) are two examples of such events that caused
notable damage in California. Given the high seismi-
city and flood hazard associated with the state of
California, this article considers Sacramento County in
California as the study region.
Seismic hazard of this region is modelled through
60 ground motions which were originally generated by
the Federal Emergency Management Agency (FEMA)
for Los Angeles in California ([Link]
edu/data/strong_motion/sacsteel/ground_motions.html).
These ground motions include both recorded and
Figure 1. Framework for the seismic risk evaluation of synthetic motions and are categorised into three sets
bridges located in flood-prone regions. having annual exceedance probabilities of 2%, 10%
 954 S. Banerjee and G.G. Prasad

and 50% in 50 years. Each set has 20 ground motions; ^ collected over years for a
annual peak discharges (Q)
LA01 to LA20 represent moderate motions with specific watershed. The relations are given as
annual exceedance probability of 10% in 50 years,
LA21 to LA40 represent strong motions with annual
logðQÞ ¼ X þ Ks ð1Þ
exceedance probability of 2% in 50 years, and LA41 to
LA60 represent weak motions with annual exceedance  
probability of 50% in 50 years. The range of peak ^
X ¼ E log Q ð2Þ
ground accelerations (PGAs) of ground motions in rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h  i
each of these three sets are given in Table 1. ^ 2  X2
s ¼ E log Q ð3Þ
Regional flood hazard is expressed in the form of
flood hazard curve that provides probability of
exceedance of annual peak discharges in a region. To where K, the frequency factor, represents the property
develop such region-specific flood hazard curve, flood- of the probability distribution under consideration at
frequency analysis is performed through empirical specified annual occurrence probability (or return
method, analytical method and/or graphical method period T) of flood events. For Gamma-Type distribu-
(Gupta 2008). The present study used analytical tion, values of K can be obtained using the coefficient
Downloaded by [York University Libraries] at 22:54 10 August 2014

method for this purpose. The annual peak discharge ^

of skewness (g) of sample annual peak discharges (Q).
data recorded for Sacramento County, CA over the For a sample of size N, g is obtained as
past 104 years (1907–2010) are collected from USGS
PN    
National Water Information System (USGS 2011) and N ^ iX 3
log Q
i¼1
shown in Figure 2(a). In the analytical method, flood g¼ ð4Þ
ðN  1ÞðN  2Þs3
hazard curve is developed using a probability distribu-
tion function. The present study used Gamma-Type
distribution as this distribution has been adopted by For the present set of 104 data of annual peak flood
US Federal agencies for flood frequency analysis discharge Q^ collected for the study region (Figure 2a),
(Gupta 2008). Other distributions such as Gaussian, X; s; and g are calculated to be equal to 2.37 m3/s,
Lognormal and Gumbel can also be used for this 0.50 m3/s and 70.65, respectively. Using these statis-
purpose. The annual peak discharge (Q) of a specified tical parameters, values of K for various flood
flood event is expressed in terms of the mean (X) and occurrence probabilities are estimated (IACWD 1982,
standard deviation (s) of logarithmic values of sample Gupta 2008). Corresponding values of Q are obtained

Table 1. Range of PGA values of ground motions having various hazard levels.

Ground motions LA01–LA20 LA21–LA40 LA41–LA60

Annual probabilities of exceedance 10% in 50 years 2% in 50 years 50% in 50 years
Range of PGA values 0.23–1.02 g 0.42–1.33 g 0.11–0.79 g

Figure 2. Regional flood hazard for Sacramento County in CA; (a) historic flood data and (b) flood hazard curve.
 Structure and Infrastructure Engineering 955

using Equation (1). Figure 2(b) plots these Q values response of bridges under these motions is expected to
with their respective annual exceedance probabilities. have a wide variation. It should be noted that the
This figure is known as flood hazard curve. uncertainties associated with primary seismic para-
meters (such as magnitude, epicentre and occurrence
rate of earthquakes) and ground motion attenuation
2.2. Uncertainties in hazard models model are not considered here. This is because, not
Parameters involved in the seismic and flood hazard only these parameters are uncertain but also many
models are associated with uncertainties. For reliable other source mechanism parameters are not explicitly
performance evaluation of bridges seismic risk in the included in the analysis presented here. This requires
presence of flood-induced scour, it is therefore an extensive future study.
important to quantify these uncertainties and estimate
their influences on bridge seismic performance. Para-
metric uncertainty may get introduced in the analyses 2.2.2. Flood hazard
from various sources such as the calculation of Developed flood hazard curve through analytical
discharge rate of flood flow, occurrence rate and method (Figure 2b) provides the most expected (or
peak intensity of seismic motions (Johnson and Dock mean) estimates of annual peak discharge (Q) for flood
Downloaded by [York University Libraries] at 22:54 10 August 2014

1998, Perkins 2002). Approximate characterisation of events with various hazard levels. Thus, this curve
underlying mechanism for different physical processes corresponds to 50% statistical confidence. Uncertain-
induces model uncertainty. Statistical uncertainty, ties associated with the annual peak discharge Q are
which is due to the limited number of available measured here by developing the 90% confidence
observation data, also has a high potential to introduce interval (between 5% and 95% confidence levels) of
significant uncertainty in the final outcome (Banerjee the mean hazard curve. For this, error limits at 5% and
et al. 2009). For example, annual peak discharge rates 95% confidence levels for various flood events with
corresponding to certain flood hazard are estimated annual exceedance probabilities of 0.01, 0.10, 0.50,
from regional flood hazard curve. This curve is 0.90 and 0.99 are taken from Gupta (2008) and listed
generated based on the measured data. Hence, the in Table 2. Annual peak discharges (Q) corresponding
values of annual peak discharge obtained for given to these confidence levels are calculated and given in
flood hazard levels may be associated with significant Table 2. Thus, flood hazard curves of the study region
statistical uncertainty. For reliable estimation of peak with 5%, 50% and 95% statistical confidence levels are
discharges, it is therefore important to develop the developed (Figure 3). For a given Q, lower statistical
confidence bounds of flood hazard curves (Gupta confidence corresponds to higher annual exceedance
2008). The following paragraphs discuss the uncertain- probability. Similarly, for a particular flood event,
ties associated with hazard models. annual peak discharge corresponding to 5% statistical
confidence is always higher than that corresponding to
other two confidence levels.
2.2.1. Seismic hazard
As previously mentioned, three different seismic
3. Example bridges
hazard levels in the study region are considered. These
include seismic events with annual exceedance prob- 3.1. Description of bridges
abilities of 2%, 10% and 50% in 50 years. Due to this The example bridges used in this study are adapted
wide hazard range of selected ground motions, seismic from a five-span reinforced concrete bridge model

Table 2. Peak flood discharges with 5% and 95% statistical confidence.

Error limits
Q with 50% with confi- Error limits 6 log(Q) with Q with confidence
confidence dence Standard deviation (s) confidence levels levels (m3/sec)
Prob of in 5% 95% 5% 95% 5% 95% 5% 95%
exceed. m3/sec log(Q) conf conf conf conf conf conf conf conf
A B1 B2 C1 C2 D1 ¼ C1 6 s D2 ¼ C2 6 s E1 ¼ B2 þ D1 E2 ¼ B2 – D2 F1 F2
0.01 1946 3.29 0.36 0.29 0.18 0.15 3.47 3.14 2945.44 1393.64
0.10 925 2.97 0.25 0.21 0.13 0.11 3.09 2.86 1233.20 726.16
0.50 263 2.42 0.17 0.17 0.09 0.09 2.51 2.34 320.05 216.38
0.90 28 1.45 0.21 0.25 0.11 0.13 1.55 1.32 35.58 20.95
0.99 9 0.97 0.29 0.36 0.15 0.18 1.12 0.79 13.13 6.21
 956 S. Banerjee and G.G. Prasad

presented by Sultan and Kawashima (1993) which was

designed following the California Department of
Transportation (Caltrans) bridge design specifications
(Caltrans 1988). The bridge has two 39.6 m-long
exterior spans and three 53.3 m-long interior spans.
Bridge deck is composed of 2.1 m deep and 12.9 m
wide hollow box-girders. Bridge piers are 19.8 m long
and have circular cross sections (2.4 m diameter).
Keeping all structural conditions as the same, the
present study developed a three-span example bridge
by omitting two interior spans from the original five-
span bridge model. The schematics of this three-span
and the original five-span example bridges and cross-
sections of bridge piers and girders are shown in
Figure 4.
Downloaded by [York University Libraries] at 22:54 10 August 2014

3.2. Modelling of bridges

Figure 3. Flood hazard curve with 90% confidence Nonlinear analyses of the two example bridges are
interval. performed in SAP2000 nonlinear computer code

Figure 4. Schematics of example bridges (not to scale); (a) three-span bridge, (b) five-span bridge, (c) cross-section of bridge
pier and (d) cross-section of bridge girder (reproduced after Sultan and Kawashima 1993).
 Structure and Infrastructure Engineering 957

(Computer and Structures, Inc. 2000). Bridge girders material, Ieq is the moment of inertial for the
are modelled using linear elastic beam-column ele- equivalent pile cross-section and IGroup is the moment
ments as these components are expected to respond of inertia for the entire group. For lateral sway motion,
within the elastic range during earthquakes. The beam- Yin and Konagai (2001) calculated the bending
column elements are aligned along the centre line of stiffness of the equivalent pile as
bridge decks. Degrees of freedom with respect to the
longitudinal translation (along the longitudinal axis of EIeq ¼ EIGroup ¼ np EIp ð5Þ
the bridge) and in-plane rotation at the abutment
locations of bridge girders are allowed. Longitudinal
translation of bridge girders are used to measure where np is the number of piles in a pile group and Ip is
displacement ductility (i.e. the ratio of measured the moment of inertia of the cross section of a single
displacement to yield displacement) based on which pile. Using this equation, the diameters of equivalent
bridge performance is categorised at different perfor- piles deq under lateral sway motion is calculated to be
mance levels (or, bridge damage states). The degrees of equal to 0.97 m. These piles are considered elastic and
freedom for vertical movement of bridge girders at assumed not to have failure during seismic excitation.
abutment locations are fully constrained. Bridge piers As identical foundation is considered for the two
Downloaded by [York University Libraries] at 22:54 10 August 2014

are modelled as double column bents. During seismic example bridges, the above calculations of deq remain
excitation, the maximum bending moment generates the same for these two bridges. The length of the
at pier ends. This often leads to the formation of equivalent pile is considered to be the same as that of
plastic hinges at these locations when the generated the individual piles.
moment exceeds the plastic moment capacity of these To model the soil–foundation interaction, non-
sections. To model such non-linear behaviour at linear p–y springs at an interval of 0.3 m are assigned
bridge pier ends, nonlinear rotational springs are to the nodes along the entire length of the equivalent
introduced in the bridge model at the top and pile. Stiffness of p–y spring at any depth is calculated
bottom of each pier. Rigid elements are assigned at following the recommendations provided by the
girder–pier connections. This ensures rigid connectiv- American Petroleum Institute (API 2000). In the
ity at girder–pier connections of monolithic concrete presence of flood-induced scour at bridge piers, a
bridges. part of the bridge foundation (or the equivalent pile)
A group of forty 0.38 m-diameter, 18.3 m-long piles loses lateral support from soil. To model this loss of
was taken in Sultan and Kawashima (1993) as the lateral support, p–y springs are removed down to a
foundation below each pier of the five-span bridge. depth equal to scour depth measured from the top of
The same foundation is used in this study for the two the equivalent pile. As suggested by Priestley et al.
example bridges. The movement of ground due to (1996), hinge supports are assigned at the tips of the
seismic shaking imposes lateral load to the pile equivalent piles.
foundation. For laterally loaded pile groups, Brown
et al. (2001) suggests the use of reduction factors that
should be applied to the lateral soil resistance p versus
lateral pile deflection y curve (p–y curve) for each 4. Flood-induced bridge scour
single pile to obtain a set of p–y curves for piles acting The extent of the impact of flood on bridges is
as a group. Values of these reduction factors (known as commonly measured in terms of scour depth at bridge
p-multipliers) depend on the location of a pile in the foundation locations. Scouring is not a prominent
pile group with respect to the point of load application. phenomenon in fine-grained soils (e.g. clays, plastic
However, successful application of this method will silts); however, for coarse-grained soils (e.g. sand or
require a detailed pile group–soil interaction analysis silty sand) scouring is very common. Therefore, to
involving realistic soil constitutive models. Such a account for the worst possible scenario of flood-
specific pile–soil interaction analysis is beyond the induced scouring, the present study considers that the
scope of the present study. Randolph (2003) outlined example bridge site is consisted of sand and silty sand
an approach to calculate the stiffness of an axially down to a great depth. Local scour at bridge piers is
loaded pile group using an equivalent pile that will considered in this study to be the expected bridge scour
represent the functional behaviour of the pile group. resulting from flood events.
Yin and Konagai (2001) adapted the same approach
for laterally loaded pile groups. The pile group is
replaced by an equivalent pile with bending stiffness 4.1. Calculation of local scour
EIeq equal the bending stiffness EIGroup of the pile Depth of local scour Ys at bridge piers is
group, where E is modulus of elasticity for the pile calculated using the HEC-18 guideline (Richardson
 958 S. Banerjee and G.G. Prasad

and Davis 2001). According to this guideline, Ys where Vc represents critical velocity of flood water that
is expressed as is required for initiating the motion of subsurface
materials (Richardson and Davis 2001). This critical
a0:65
velocity is expressed as
Ys ¼ 2hK1 K2 K3 K4 Fr1 0:43 ð6Þ
h
Vc ¼ 6:19h1=6 D50 1=3 ð10Þ
where h is the upstream flow depth, a is the pier width,
K1, K2, K3 and K4 are correction factors representing
pier nose shape, angle of attack of flow, bed condition Note that the scour depth calculation discussed
and particle size of subsurface soil,
pffiffiffiffiffirespectively. The here provides a rapid estimation of scour at bridge
Froude number Fr1 is defined asV gh; V and g are the piers caused by regional flood events. For exact
mean velocity of the upstream flow and acceleration calculation of bridge scour, separate hydraulic analysis
due to gravity, respectively. Values of K1, K2, K3 and K4 at each bridge pier is required.
are directly obtained from HEC-18. All bridges
considered in this study have circular pier cross-
sections, and thus K1 is equal to 1.0. K2 is equal to 1.0 4.2. Sensitivity analysis
Downloaded by [York University Libraries] at 22:54 10 August 2014

for zero angle of attack of flow considered herein. K3 is It is recognised that the uncertainties associated with
taken as 1.1 considering clear-water scour. K4 is 1.0 for input parameters for the scour depth calculation may
the size of subsurface soil particles (D50 5 2 mm and introduce significant variability in the calculated scour
D95 5 20 mm, where D50 and D95 are particle sizes depths. It is therefore important to perform a
corresponding to 50% and 95% finer) considered in sensitivity study to identify major uncertain para-
this study. For coarser particles (i.e. D50 4 2 mm and meters that have great influence on the calculated scour
D95 4 20 mm), K4 is less than 1.0 and this results in less depth. For this purpose, the five-span example bridge
scour. The depth of upstream flow h for a given flood and a 100-year flood event are considered. This
discharge rate Q is calculated as (Gupta 2008) particular bridge is chosen as it is directly adopted
from the literature without any alteration. It is
Q expected that the three-span example bridge will
h¼ ð7Þ
bV behave in the same manner as of the five-span bridge
under the same regional hazard condition. Expected
where b is the passage width. The mean velocity of the variations of different input parameters are discussed
upstream flow V is given by here.

1 bh 2=3 1=2
V¼ S ð8Þ 4.2.1. Annual peak discharge (Q)
n b þ 2h
Uncertainties associated with Q for any particular
where n and S represent manning’s roughness coeffi- flood hazard level can be estimated from the flood
cient and slope of the stream bed, respectively. The hazard curve with 90% confidence interval (Figure 3).
present study considers n ¼ 0.08 (FEMA 2008) and This may have a significant impact on the expected
S ¼ 0.1%. For a given flood event, the annual peak bridge scour depth. To demonstrate this impact, the
discharge Q is the only known quantity. To calculate present study considered a 100-year flood event.
corresponding values of flow velocity V and flow depth Values of Q corresponding to 5%, 50% and 95%
h, the passage width b is assumed to be equal to the total confidence levels are estimated to be equal to
length of the example bridges. Hence, example bridges 2720 m3/s, 1946 m3/s and 1280 m3/s, respectively
will have different flow velocities even for a same flood (Figure 3). Respective values of flow depth and flow
event (i.e. same Q) as their passage widths are different. velocity are calculated using Equations (7) and (8).
Johnson and Torrico (1994) suggested another These two variables (i.e. flow velocity and flow depth)
correction factor Kw in Equation (6) for h/a 5 0.8 in are considered here as independent variables although,
a subcritical flow and uniform non-cohesive sediments in reality, some correlation may exist among them.
with a/D50 4 50. In such cases, Kw should be multi- These two variables change according to the variation
plied with the value of scour depth calculated using in of annual peak discharge obtained for flood events.
Equation (6). Kw is calculated as

4.2.2. Correction factors for HEC equation
2:58ðh=aÞ0:34 Fr1 0:65 for V < Vc This study considered K1 and K4 as deterministic
Kw ¼ ð9Þ
1:00ðh=aÞ0:13 Fr1 0:25 for V  Vc variables and K2 and K3 as random variables. The
 Structure and Infrastructure Engineering 959

correction factor K1 corresponds to the shape of the according to their sensitivity to the bridge scour, a
bridge piers which are considered to be circular. tornado diagram is developed. At first, all uncertain
Therefore K1 has no variation. K4 corresponds to the parameters are first kept at their mean values and the
size of subsurface soil particles and remains constant scour of the example five-span bridge for a 100-year
(equal to 1.0) for D50 5 2 mm and D95 5 20 mm. flood event is calculated. This scour depth is estimated
Any variation in soil particle size distribution is to be 3.4 m and referred to as the most expected scour
assumed not to alter this criterion. K2 and K3 depth. Following this, the upper and lower bounds of
correspond to the angle of attack and bed condition, one random variable are used at a time in the scour
respectively. During flooding, the angle of attack may depth calculation while keeping the values of other
vary depending on the direction of propagation of random variables at their respective mean values.
flood water to the bridge site. Therefore, some Thus, two scour depths are calculated for the two
variability may be present in K2. This variation is extreme values (at upper and lower bounds) of each
taken here following the recommendation of random variable. The absolute difference of the
Johnson and Dock (1998). K3 depends on the estimated scour depth between two bounds is referred
bed condition. Many a times it is very difficult to as the swing of the scour corresponding to the
to accurately predict the type of bed condition, selected random variable. This procedure is repeated
Downloaded by [York University Libraries] at 22:54 10 August 2014

particularly during flood events. Such prediction for all random variables considered herein, and
uncertainty is accounted here by considering a varia- corresponding swings of the scour is calculated and
tion in K3. plotted in Figure 5. This figure, known as the Tornado
As presented in Johnson and Dock (1998), K2 is diagram, shows the relative influence of the random
assumed to follow a normal distribution with mean variables on the calculated pier scour. The longer the
and coefficient of variation equal to 1.00 and 0.05. swing, the higher the influence of corresponding
Similarly, K3 is considered as a random variable variable on the estimated bridge scour.
having a mean equal to 1.10 with 5% variation In Figure 5, the vertical line in the middle of
(Johnson and Dock 1998). In previous sensitivity tornado diagram indicates the most expected scour
studies (Porter et al. 2002, Lee and Mosalam 2006), with a value of 3.4 m. The length of each swing
values at 10th and 90th percentiles of probability (horizontal bar) represents the variation in the
distributions are considered as lower and upper expected scour due to the variation of random
bounds of the input variables. Na et al. (2008) variables. As this figure shows, Q, K2 and K3 have
considered 16th and 84th percentiles (i.e. mean + one almost equal influences on the scour depth, while the
standard deviation) as these two bounds. The present same is slightly lower for a. Swings are slightly skewed
analysis considers a wider range between these two towards right with respect to the vertical line. This
bounds than that reported in these literatures. 2nd indicates that the same amount of positive and
percentile (i.e. two standard deviations below the negative variations of the random parameters will
mean) and 98th percentile (i.e. two standard devia- not result in the same amount of change in scour on
tions above the mean) are respectively taken here as both sides of the most expected scour depth (i.e.
the lower and upper bounds of these two random 3.4 m). For all parameters, higher values resulted in
variables K2 and K3. Thus, these bounds are more scour. Note that this analysis considered all
calculated to be equal to 0.9 and 1.10 for K2 and bridge piers have the same diameter irrespective of the
0.99 and 1.21 for K3. variability considered in its value.

4.2.3. Effective pier width (a)

The mean diameter of all bridge piers is 2.4 m. For
analysis purpose, it is assumed that the diameter of
bridge piers has 10% variation due to some construc-
tion errors. Hence the 2nd and 98th percentile values
of ‘a’ becomes 2.16 m and 2.64 m which respectively
correspond to the assumed upper and lower bounds of
this parameter.
Other input parameters and coefficients associated
with scour depth calculation (such as n and S) are
considered to be deterministic and their values kept
fixed at their best estimates (reported earlier) for the Figure 5. Tornado diagram for the scour depth of the five-
entire analysis. To rank the uncertain parameters span example bridge under a 100-year flood event.
 960 S. Banerjee and G.G. Prasad

5. Seismic response of bridges in the presence of than that is observed here by varying one parameter at
flood-induced scour a time. As the present study is focused on the
To estimate the combined effect of regional seismic and estimation of parameter sensitivity, random combina-
flood hazards on bridge performance, 1.1-year, 2-year, tions of input parameters are not considered.
10-year, 20-year, 50-year and 100-year flood events
with annual exceedance probabilities of 0.90, 0.50,
0.10, 0.05, 0.02 and 0.01, respectively are considered. 5.1. Time history analysis
Annual peak discharges for these flood events at 5%, Nonlinear time history analyses of example bridges
50% and 95% confidence levels are estimated from under 60 LA ground motions are performed in
Figure 3. Calculated scour depths for the example SAP2000 in the presence of flood-induced scour. As
bridges are presented in Table 3. The result depicts that mentioned previously, nonlinear p–y springs are
for any particular level of flood hazard, calculated removed up to a depth of YS (i.e. scour depth) from
scour depths for the example bridges at 5% and 95% the ground surface to model scour condition at bridge
confidence levels vary within 11% from that obtained piers. An additional set of time history analysis is
at 50% confidence level. Experience from a parallel performed under no flood (i.e. no scour) condition.
analysis indicates that such a variation of scour depth From modal analysis, it is observed that the funda-
Downloaded by [York University Libraries] at 22:54 10 August 2014

will not produce any notable variation of bridge mental time periods of example bridges increase with
seismic response. Thus, the uncertainty in predicting increase in scour depth (i.e. with flood hazard level) as
annual peak discharge for a flood event does not given in Table 4. This is due to the increased flexibility
produce any significant variation in the seismic induced by the loss of lateral resistance offered by the
performance of the example bridge. This is equally subsurface soil in the presence of scour. Time histories
true for other random variables (i.e. K2, K3 and a) in of bridge response are recorded at various bridge
scour depth calculation as the observed swings in locations. The primary outcome of the finite element
calculated scour depth due to the randomness of these analyses is the displacement of bridge girders measured
parameters are found to be either equal to or less than in the longitudinal direction (along bridge axis). Bridge
that observed for the variation of Q. Hence, the most girder displacements recorded at different stages of
expected scour depths (calculated using the mean analysis are used to express bridge performance in the
values of input parameters) for the above six flood form of bridge fragility curves.
events are used for the seismic performance evaluation Measured girder displacements are converted to
of example bridges. displacement ductility mD which is further used as a
It should be noted here that if randomness of all signature that represents bridge seismic performance.
uncertain parameters are considered simultaneously, By definition, displacement ductility mD is the ratio
the expected scour may have much higher variation of displacement of the bridge pier to the yield

Table 3. Calculation of scour of the example bridges under fix different flood events.

Statistical confidence Resulted scour depth (m)

of annual peak Example
discharge (Q) bridges 1-yr flood 2-yr flood 10-yr flood 20-yr flood 50-yr flood 100-yr flood
5% Three-span 0.74 2.54 3.54 3.79 4.15 4.29
Five-span 0.60 1.32 3.05 3.29 3.61 3.74
50% Three-span 0.68 2.41 3.30 3.55 3.81 3.97
Five-span 0.56 1.22 2.85 3.08 3.30 3.45
95% Three-span 0.62 2.30 3.1 3.23 3.49 3.58
Five-span 0.50 1.14 2.68 2.79 3.03 3.11

Table 4. Fundamental time periods (in seconds) of example bridges.

Fundamental time periods (in seconds)

Example bridges No floods 1-year flood 2-year flood 10-year flood 20-year flood 50-year flood 100-year flood
Three-span 2.19 2.72 3.38 3.58 3.64 3.69 3.73
Five-span 2.09 2.559 2.87 3.32 3.359 3.40 3.44
 Structure and Infrastructure Engineering 961

displacement measured at the same location (Caltrans which the curvature ductility (mj) is converted to
2006). Displacements at the top of bridge piers should displacement ductility (mD). Figure 6, (a) and (b)
be the same as that measured at the corresponding respectively, represents the shear capacity and shear
node at bridge girder due to the monolithic pier–girder demand at plastic hinge regions of bridge piers. This
connections of example bridges. mD corresponding to clearly indicates that the example bridges are safe from
the ultimate state (i.e. complete collapse) is calculated the occurrence of premature shear failure as the
to be equal to 5.0. This value is in accordance with capacity remains higher than the demand under any
Caltrans recommendation for target displacement of the ground motions considered here.
ductility (Caltrans 2006). Beyond yielding and before
complete collapse, three intermediate states of bridge
damage namely minor (or slight), moderate and major 5.2. Bridge fragility curves
(or extensive) damage (definition follows HAZUS 1999 Fragility curves for example bridges are developed
physical descriptions of bridge seismic damage) are using bridge damage states determined for all ground
considered. Banerjee and Shinozuka (2008a, 2008b) motions. These monotonically increasing curves re-
quantified these intermediate bridge damage states in present the probability of bridge failure in a particular
terms of rotational ductility of bridge piers. In this damage state under certain intensity of ground
Downloaded by [York University Libraries] at 22:54 10 August 2014

study, the rotational ductility values at different motions (e.g. PGA). Numerous studies have been
damage states are converted to their corresponding performed throughout the last few decades and
displacement ductility values. Values of mD correspond- significant advances are achieved in the state-of-the-
ing to minor, moderate and major damage states are art research on seismic fragility characteristics of
thus calculated to be equal to 2.25, 2.90 and 4.60,
respectively. These values represent the threshold limits
(lower bound) of their corresponding damage states.
The state of bridge damage under a specific ground
motion is decided by comparing the calculated
displacement ductility mD with the threshold limits of
all damage states (i.e. minor, moderate and major).
Thus, bridge damage state under a ground motion can
be decided as

mD < 2:25; No damage

2:25  mD < 2:90; Minor damage
2:90  mD < 4:60; Moderate damage ð11Þ
4:60  mD < 5:0; Major damage
mD  5:0; Complete collapse

Elastic displacement response spectra of 60 ground

motions considered here show an increasing trend of
spectral displacement as the structural time period
increases (Figure A1). This indicates that bridges with
scour at foundations due to flood events are expected
to have higher displacements at their girder levels
during seismic events.
Developed shear force at bridge piers may also lead
to premature shear failure of bridges during seismic
events (Priestley et al. 1996). Such failure occurs when
the developed shear force at plastic hinge regions of
bridge piers exceeds the shear capacity of these regions.
Priestley et al. (1996) provided an analytical frame-
work to evaluate the possibility of having premature
shear failure in bridge piers under seismic excitation. In
this, the shear capacity and shear demand of a bridge
pier are expressed as functions of its curvature ductility Figure 6. Check for premature shear failure at bridge piers;
(mj). These functions are used in the present study in (a) shear capacity and (b) shear demand.
 962 S. Banerjee and G.G. Prasad

bridges. Bridge fragility curves can be developed by where PGAj represents PGA of a ground motion j
using past earthquake damage data (Basöz and (¼LA01, LA02, . . . , LA60). Fragility parameters ck
Kiremidjian 1998, Shinozuka et al. 2000, Tanaka and zk for the damage state k can be estimated using
et al. 2000, Yamazaki et al. 2000, etc.) and/or through maximum likelihood method. The likelihood function
numerical simulation of bridge dynamic characteristics L is given as
(Hwang et al. 2000, Shinozuka et al. 2000, Karim and
Yamazaki 2003, Elnashai et al. 2004, Banerjee and Y
60   r   1rj
Shinozuka 2007, 2008a, Nielson and DesRoches 2007, L¼ F PGAj ; ck ; zk j 1  F PGAj ; ck ; zk
j¼1
Pan et al. 2010). In most of these studies, fragility
curves are expressed as two-parameter lognormal ð13Þ
distributions. The distribution parameters, referred to
as fragility parameters, are median c and log-standard where rj ¼ 0 or 1 depending on whether or not the
deviation z. These respectively represent PGA corre- bridge sustains the damage state k under jth ground
sponding to 50% probability of exceeding a damage motion. HAZUS (1999) suggested an uncertainty
state and the dispersion of fragility curve. At a damage factor for seismic demand to be equal to 0.5 based
state k (¼minor, moderate, major and complete on the studies performed by Pekcan (1998). Following
Downloaded by [York University Libraries] at 22:54 10 August 2014

collapse), the analytical form of the fragility function this, zk is taken here as 0.5 for all example bridges and
is expressed as damage states. This common zk for all bridges and
damage states will restrict the intersection of any two
   
  ln PGAj ck fragility curves.
F PGAj ; ck ; zk ¼ F ð12Þ Figures 7 and 8, respectively, show the seismic
zk
fragility curves for the three-span and five-span

Figure 7. Seismic fragility curves of the three-span example bridge (a) at minor damage, (b) at moderate damage, (c) at major
damage and (d) at collapse state.
 Structure and Infrastructure Engineering 963
Downloaded by [York University Libraries] at 22:54 10 August 2014

Figure 8. Seismic fragility curves of the five-span example bridge (a) at minor damage, (b) at moderate damage, (c) at major
damage and (d) at collapse state.

example bridges at the minor, moderate, major frequent flood (such as a 10-year flood) is capable of
damage and complete collapse states. At all damage causing degradation in the seismic performance of
states, fragility curves shift from right to left indicating bridges to the same level as of a major flood event
higher seismic vulnerability with increase in flood (such as 50-year and 100-year flood).
hazard level. This is due to the fact that flood events Although estimated scour depths at bridge piers
with higher hazard level results in higher scour depth under the six flood events are different for the three-
at bridge piers (as Q increases with hazard level; Figure span and the five-span example bridges (Table 3), their
2b). Therefore, seismic damageability of bridges seismic fragility characteristics at different damage
increases in the presence of flood-induced scour. states are found to be very similar (Figures 7 and 8).
At lower damage states (such as at minor and More than 5% difference in their median fragility
moderate), a major degradation in seismic fragility parameter (c) is observed in only four cases out of total
characteristics is observed from ‘No flood’ to ‘1.1-year 28 cases (combinations of seven flood scenarios and
flood’ for both example bridges. Beyond ‘2-year flood’, four seismic damage states). Developed fragility curves
the change in bridge seismic fragility characteristics are used further for calculating the seismic risk of the
with increased level of flood hazard is insignificant. At example bridges in the presence and absence of flood-
higher damage levels (i.e. at major damage and induced scour at bridge piers.
collapse), median seismic fragility parameter (c)
becomes constant on and beyond ‘10-year flood’ for
all cases except for the five-span bridge at collapse 6. Risk evaluation
state. This indicates that the seismic damageability of Risk curves of example bridges due to the regional
example bridges does not change with flood-induced seismicity are developed assuming that the bridges are
scour depth more than 3 m (approximately). From this already exposed to regional flood events. ‘System risk’
observation it can be concluded that a relatively less indicates the probability that the system will suffer
 964 S. Banerjee and G.G. Prasad

from a degraded performance equal to or below a event m. This value under different LA ground motions
certain level due to external loading condition. For is obtained from fragility curves of example bridges as
regional highway transportation systems in Los shown in Figures 7 and 8. Following Equation (14),
Angeles in California and Shelby County in Tennessee, bridge restoration costs are calculated for 60 LA
risk curves due to regional seismic hazard have been motions in the absence and presence of scour resulted
developed in several past studies (Chang et al. 2000, from various frequency flood events. These are further
Werner et al. 2000, Shiraki et al. 2007, Banerjee et al. used in developing risk curves. Figure 9 shows the
2009, Zhou et al. 2010). In these literatures, network restoration cost of the five-span example bridge under
delay or the total social cost are taken as an index to 20 strong ground motions (LA21 to LA40; annual
represent the degradation of system performance. probability of exceedance 2% in 50 years) for four
Following this, the present study uses societal loss flood scenarios (No flood, 1.1-year flood, 2-year flood
measured in terms of post-event bridge repair or and 10-year flood). As can be seen from this figure,
restoration cost as the performance degradation index. bridge restoration cost increases with increase in flood
Hence, the present study develops risk curves that hazard level for any seismic event. Across these 20
represent the annual probability of exceeding different motions, a variation in expected bridge seismic
levels of societal loss arising from system degradation restoration cost for a particular flood hazard level is
Downloaded by [York University Libraries] at 22:54 10 August 2014

due to regional hazards. Losses due to network observed even if the motions have the same hazard
downtime during bridge restoration are not considered level (i.e. occurrence probability).
as a part of the loss estimation. Figures 10 and 11, respectively show the seismic
Loss due to bridge damage under the combined risk curves of the three-span and five-span example
effect of earthquake and flood-induced scour is
calculated according to the extent of damage that the
example bridges suffer from these natural hazards.
Post-event bridge repair cost is taken as proportional to
its replacement value (Zhou et al. 2010). This propor-
tionality factor, known as the damage ratio rk (HAZUS
1999), depends on the state of bridge damage k. Ranges
of damage ratios for different bridge seismic damage
states and their best estimates within each range are
given in HAZUS (1999). These values are listed in
Table 5. The total replacement cost (C) is estimated by
multiplying unit deck area replacement cost with the
total area of a bridge deck (i.e. bridge length 6 deck
width). Caltrans recommended that the unit area
replacement cost can be taken as $1292/m2. Therefore,
the total bridge restoration cost (CRm) after an earth- Figure 9. Restoration cost of the five-span example bridge
quake event m can be written as (Zhou et al. 2010). under LA21 to LA40.

X
4
CRm ¼ pm ðDS ¼ kÞCrk ð14Þ
k¼1

where pm (DS ¼ k) is the probability that the example

bridge suffers from damage state k during the seismic

Table 5. Damage ratios for RC bridges as given in HAZUS

(FEMA 1999).

Seismic damage Range of Best

state damage ratio estimate
Minor damage 0.01–0.03 0.03
Moderate damage 0.02–0.15 0.08
Major damage 0.10–0.40 0.25
Complete collapse 0.30–1.00 2/Number
of spans* Figure 10. Risk curve of the three-span example bridge
under regional seismicity in the presence of flood-induced
Note: *For bridges with number of spans equal to or more than 3. scour.
 Structure and Infrastructure Engineering 965

identical structural attributes and configurations as of

the example bridges and spans over a similar site as of
the study region, developed fragility curves and risk
curves can be used to describe the damageability and
associated risk of these highway bridges under regional
multihazard. The present risk assessment framework
can be also used to investigate the adequacy of
currently existing seismic retrofit techniques for bridges
located in seismically active, flood-prone regions.

7. Conclusions
Seismic risk of two example reinforced concrete bridges
Figure 11. Risk curve of the five-span example bridge under located in a seismically active, flood-prone region is
regional seismicity in the presence of flood-induced scour. estimated in this study. It is assumed that prior to any
seismic event example bridges were exposed to regional
Downloaded by [York University Libraries] at 22:54 10 August 2014

bridges. As expected, seismic risk of both example flood events that caused scour at bridge foundations.
bridges increases with increase in flood hazard level. Observed result indicates higher seismic risk of example
For example, expected bridge restoration cost equal to bridges when the impact of regional flood hazard on
$250,000 has an annual probability of exceedance bridges is considered in the analysis framework. A low
equal to 0.10 for the five-span example bridge under intensity (i.e. very frequent) flood event that could
regional seismicity with no flood (hence, no scour) cause small amount of bridge scour may result in a
scenario. This annual probability becomes 0.12 and significant enhancement of bridge seismic risk. Hence,
0.14 due to the presence of bridge scour resulted from a consideration of only seismic hazard model for the risk
1.1-year and a 2-year flood event, respectively. Beyond assessment of bridges located in seismically active
10-year flood, risk curves for example bridges have flood-prone regions may underestimate the maximum
either no (for the three-span bridge) or very small (for demand from regional multihazard. Analysis with a
the five-span bridge) increment with increased flood combined seismic and flood hazard model is required.
hazard level. This is due to the fact that bridge scour Other notable observations that can be made from
resulted from flood events does not produce any this study are (i) expected scour depths for the two
further degradation to the seismic performance of example bridges are found to be different for a given
example bridges when the scour depth reaches to a flood event, however, no much variation is observed in
particular range (*3 m). Hence, no change in bridge their seismic fragility characteristics and (ii) five-span
fragility characteristics is found beyond this flood example bridge has higher seismic risk than the three-
hazard level (Figures 7 and 8). span bridge even if their seismic fragility characteristics
Comparison of Figures 10 and 11 indicates that the are very close. A few assumptions are made here; (a)
five-span example bridge has higher seismic risk than bridge abutments are not modelled and the end of the
the three-span bridge even if they have almost the same bridge girder is allowed to translate without any
seismic fragility characteristics at different damage constraint in the longitudinal direction (i.e. along
states. This is true irrespective of the hazard level and bridge axis), (b) piles are assumed to remain elastic
occurrence of flood events. As the five-span bridge during seismic excitation without any failure, (c)
comprises higher deck area than the three-span bridge, passage width during flood is considered to be the
expected restoration cost due to a seismic event is more same as overall bridge length and (d) the same amount
for the five-span bridge when the both suffer from the of local scour around all piers of an example bridge is
same level of disaster after an earthquake. considered, however in reality, various scour depths
Developed risk curves are the useful tools that serve are generally observed at different bridge piers based
as a foundation for future research related to the on the local hydraulic characteristics around piers.
development of strategic plans for bridge repair/retrofit Clearly, future research is needed to incorporate
prioritisation under similar multihazard conditions. possible bridge failure at abutment locations and in
Due to retrofit, seismic performance of bridges gets the transverse direction of the bridge to enhance the
enhanced which in turn results in the reduction of post- performance evaluation of bridges located in seismi-
event bridge restoration cost. Therefore, the risk curves cally active flood-prone regions.
will move towards left when bridges are retrofitted The overall framework presented in this study is
prior to the event. For a highway transportation transportable and can be adopted for any other
network that consists of bridges with statistically structure and region of interest. Risk curves such as
 966 S. Banerjee and G.G. Prasad

the ones developed in this study can be used as a Computer and Structures, Inc. 2000. SAP2000 nonlinear users
foundation for future research on the development of manual, V. 8. Berkeley, CA: Computer and Structures, Inc.
Elnashai, A.S., Borzi, B., and Vlachos, S., 2004. Deforma-
strategic plans for repair/retrofit prioritisation of tion-based vulnerability functions for RC bridges.
bridges for the similar multihazard scenarios. For a Structural Engineering and Mechanics, 17, 215–244.
highway transportation network that consists of FEMA. 2008. Flood insurance study for Sutter County,
bridges with statistically identical structural attributes California. Federal Emergency Management Agency
and configurations as of the example bridges and spans (FEMA) Flood Insurance Study Number 060394V000A.
Washington, D.C.: Federal Emergency Management
over a similar site as of the study region, developed Agency.
fragility curves and risk curves can be used to describe Ghosn, M., Moses, F., and Wang, J., 2003. Highway bridge
the damageability and associated risk of these highway design for extreme events. National Cooperative Highway
bridges under regional multihazard. Furthermore, the Research Program, NCHRP Report 489. Washington,
analysis framework discussed herein can be used to DC: Transportation Research Board, National Academy
Press.
investigate the effectiveness of currently available Gupta, R., 2008. Hydrology and hydraulic systems. Illinois,
seismic retrofit techniques for bridges under the IL: Waveland Press, Inc.
combined seismic and flood hazards. HAZUS. 1999. Earthquake loss estimation methodology.
Technical Manual SR2. Washington, DC: Federal
Downloaded by [York University Libraries] at 22:54 10 August 2014

Emergency Management Agency through agreements

References with National Institute of Building Science.
Alampalli, S. and Ettouney, M., 2008. Multihazard applica- Hwang, H., Jernigan, J.B., and Lin, Y.-W., 2000. Evaluation
tions in bridge management. Transportation Research of seismic damage to Memphis bridges and highway sys-
Circular, Number E-C128. Washington, D.C.: Transpor- tems. ASCE Journal of Bridge Engineering, 5 (4), 322–330.
tation Research Board. IACWD, 1982. Guidelines for determining flood flow fre-
API, 2000. Recommended practice for planning, designing and quency, Bulletin 17B. Reston, VA: Interagency Advisory
constructing fixed offshore platforms. API Recommended Committee on Water Data, U.S. Dept. of the Interior,
Practice 2A-WSD (RP 2A). Washington, D.C.: Amer- Office of Water Data Coordination.
ican Petroleum Institute. Johnson, P.A. and Dock, D.A., 1998. Probabilistic bridge
Banerjee, S. and Shinozuka, M., 2007. Nonlinear static pro- scour estimates. Journal of Hydraulic Engineering ASCE,
cedure for seismic vulnerability assessment of bridges. 124 (7), 750–754.
Computer-Aided Civil and Infrastructure Engineering Johnson, P.A. and Torrico, E.F., 1994. Scour around wide piers
(CACAIE), 22 (4), 293–305. in shallow water. Transportation Research Board Record
Banerjee, S. and Shinozuka, M., 2008a. Mechanistic quanti- 1471. Transportation Research Board, Washington, DC.
fication of RC bridge damage states under earthquake Karim, K.R. and Yamazaki, F., 2003. A simplified method of
through fragility analysis. Probabilistic Engineering constructing fragility curves for highway bridges. Earth-
Mechanics, 23 (1), 12–22. quake Engineering and Structural Dynamics, 32, 1603–1626.
Banerjee, S. and Shinozuka, M., 2008b. Experimental veri- Lee, T.H. and Mosalam, K.M., 2006. Probabilistic seismic
fication of bridge seismic damage states quantified by evaluation of reinforced concrete structural components and
calibrating analytical models with empirical field data. systems. Report 2006/04. Berkeley, CA: Pacific Earthquake
Journal of Earthquake Engineering and Engineering Engineering Research Center, University of California.
Vibration, 7 (4), 383–393. MCEER-AEI. 2007. Symposium on emerging developments in
Banerjee, S., Shinozuka, M., and Sgaravato, M., 2009. multi-hazard engineering [online]. McGraw-Hill New
Uncertainty in seismic performance of highway network York City, 18 September. Available from: [Link]
estimated using empirical fragility curves of bridges. buff[Link]/meetings/AEI/[Link] [Accessed 2 Jan-
International Journal of Engineering under Uncertainty: uary 2012].
Hazard, Assessment, and Mitigation, 1 (1–2), 1–11. Na, U.J., Ray Chaudhuri, S., and Shinozuka, M., 2008.
Basöz, N. and Kiremidjian, A.S., 1998. Evaluation of bridge Probabilistic assessment for seismic performance of port
damage data from the Loma Prieta and Northridge, structures. Soil Dynamics and Earthquake Engineering, 28
California Earthquakes. Technical Report MCEER-98- (2), 147–158.
0004. Multidisciplinary Center for Earthquake Engineer- Nielson, B.G. and DesRoches, R., 2007. Seismic fragility
ing Research, State University of New York at Buffalo. methodology for highway bridges using a component
Brown, D.A., et al. 2001. Static and dynamic lateral loading of level approach. Earthquake Engineering and Structural
pile groups. National Cooperative Highway Research Dynamics, 36 (6), 823–839.
Program, NCHRP Report 461. Washington DC: Trans- Pan, Y., et al. 2010. Seismic fragility of multispan simply
portation Research Board, National Academy Press. supported steel highway bridges in New York state II:
Caltrans. 1988. Bridge design aids. Sacramento, CA: Cali- fragility analysis, fragility curves, and fragility surfaces.
fornia Department of Transportation, Division of ASCE Journal of Bridge Engineering, 15 (5), 462–472.
Structures. Pekcan, G., 1998. Design of seismic energy dissipation systems
Caltrans. 2006. Seismic design criteria, V.1.4. Sacramento, for concrete and steel structures. Dissertation (PhD).
CA: California Department of Transportation, Division State University of New York at Buffalo, New York.
of Structures. Perkins, D.M., 2002. Uncertainty in probabilistic seismic
Chang, E.S., Shinozuka, M., and Moore, J. 2000. Probabil- hazard analysis. Acceptable risk processes: lifelines and
istic earthquake scenarios: extending risk analysis meth- natural hazards. Monograph no. 21. In: C. Taylor and E.
odologies to spatially distributed systems. Earthquake VanMarcke, ed. Technical council on lifeline earthquake
Spectra, 16 (3), 557–572. engineering. Reston VA: ASCE, 19–60.
 Structure and Infrastructure Engineering 967

Porter, K.A., Beck, J.L., and Shaikhutdinov, R.V., 2002. Tanaka, S., et al. 2000. Evaluation of seismic fragility for
Sensitivity of building loss estimates to major uncertain highway transportation systems. In: Proceedings of the
variables. Earthquake Spectra, 18 (4), 719–743. 12th world conference on earthquake engineering, 30
Priestley, M.J.N., Seible, F., and Calvi, G.M., 1996. Seismic January–4 February 2000, Upper Hutt, New Zealand,
design and retrofit of bridges. New York: John Wiley and Paper No. 0546.
Sons, Inc. USGS. 2011. National water information system [online].
Randolph, M.F., 2003. Science and empiricism in pile Available from: [Link] [Ac-
foundation design. Ge´otechnique, 53 (10), 847–875. cessed 14 February 2011].
Richardson, E.V. and Davis, S.M., 2001. Evaluating scour at Werner, S.D., et al. 2000. A risk-based methodology for
bridges. Publication No. FHWA NHI 01–001. Hydraulic assessing the seismic performance of highway systems.
Engineering Circular (HEC) No. – 18. Washington, DC: Report MCEER–00–0014, Multidisciplinary Center for
Federal Highway Administration, US. Department of Earthquake Engineering Research, State University of
Transportation. New York, Buffalo.
Shinozuka, M., et al. 2000. Statistical analysis of fragility Yamazaki, F., Motomura, H., and Hamada, T., 2000. Damage
curves. ASCE Journal of Engineering Mechanics, 126 assessment of expressway networks in Japan based on
(12), 1224–1231. seismic monitoring. In: Proceedings of the 12th world con-
Shiraki, N., et al. 2007. System risk curves: probabilistic ference on earthquake engineering, 30 January–4 February
performance scenarios for highway networks subject to 2000, Upper Hutt, New Zealand, Paper No. 0551.
earthquake damage. ASCE Journal of Infrastructure Yin,Y. and Konagai, K., 2001. A simplified method for
Downloaded by [York University Libraries] at 22:54 10 August 2014

Systems, 13 (1), 43–54. expression of the dynamic stiffness of large-scaled

Sultan, M. and Kawashima, K., 1993. Comparison of the grouped piles in sway and rocking motions. Journal of
seismic design of highway bridges in California and in Applied Mechanics, JSCE, 4, 415–422.
Japan. Recent selected publications of Earthquake Zhou, Y., Banerjee, S., and Shinozuka, M., 2010. Socio-
Engineering Div. Japan: Public Works Research economic effect of seismic retrofit of bridges for highway
Institute (PWRI) (Technical Memorandum of PWRI transportation networks: a pilot study. Structure and
No. 3276). Infrastructure Engineering, 6 (1–2), 145–157.
 968 S. Banerjee and G.G. Prasad

Appendix 1
Downloaded by [York University Libraries] at 22:54 10 August 2014

Figure A1. Displacement response spectra at 5% damping of (a) LA01–LA20 (10% annual probability of exceedance in 50
years), (b) LA21–LA40 (2% annual probability of exceedance in 50 years) and (c) LA41–LA60 (50% annual probability of
exceedance in 50 years).