You are on page 1of 17

Journal of Earthquake Engineering

ISSN: 1363-2469 (Print) 1559-808X (Online) Journal homepage: http://www.tandfonline.com/loi/ueqe20

Regional Seismic Risk Assessment of Bridge


Network in Charleston, South Carolina

Jamie E. Padgett , Reginald DesRoches & Emily Nilsson

To cite this article: Jamie E. Padgett , Reginald DesRoches & Emily Nilsson (2010) Regional
Seismic Risk Assessment of Bridge Network in Charleston, South Carolina, Journal of Earthquake
Engineering, 14:6, 918-933, DOI: 10.1080/13632460903447766

To link to this article: http://dx.doi.org/10.1080/13632460903447766

Published online: 08 Jul 2010.

Submit your article to this journal

Article views: 198

View related articles

Citing articles: 14 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=ueqe20

Download by: [IIT Indian Institute of Technology - Mumbai] Date: 08 November 2017, At: 01:58
Journal of Earthquake Engineering, 14:918933, 2010
Copyright A.S. Elnashai & N.N. Ambraseys
ISSN: 1363-2469 print / 1559-808X online
DOI: 10.1080/13632460903447766

Regional Seismic Risk Assessment of Bridge


Network in Charleston, South Carolina

JAMIE E. PADGETT1, REGINALD DESROCHES2,


and EMILY NILSSON3
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

1
Department of Civil & Environmental Engineering, Rice University, Houston,
Texas, USA
2
School of Civil & Environmental Engineering, Georgia Institute of Technology,
Atlanta, Georgia, USA
3
Datum Engineers, Austin, Texas, USA

This article presents the results of a seismic risk assessment of the bridge network in Charleston,
South Carolina and the surrounding counties to support emergency planning efforts, and for
prioritization of bridge retrofit. This study includes an inventory analysis of the approximately
375 bridges in the Charleston area, and convolution of the seismic hazard with fragility curves
analytically derived for classes of bridges common to this part of the country. State-of-the-art
bridge fragility curves and replacement cost estimates based on region-specific data are used to
obtain economic loss estimates. The distribution of potential bridge damage and economic losses
are evaluated for several scenario events in order to aid in the identification of emergency routes
and assess areas for investment in retrofit. This article also evaluates the effect of uncertainty on
the resulting predicted economic losses. The findings reveal that while the risk assessment is very
sensitive to both the assumed fragility curves and damage ratios, the estimate of total expected
economic losses is more sensitive to the vast differences in damage ratio models considered.

Keywords Seismic Risk Assessment; Loss Estimation; Bridges; Fragility; Transportation Net-
work; Sensitivity Study

1. Introduction
Regional seismic risk assessments (SRAs) are becoming popular tools for evaluating the
performance of transportation networks under earthquake loading. The term seismic risk
refers to the potential for damage or losses that may be associated with a seismic event.
Such regional assessments provide a unique approach for estimating the risk to highway
infrastructure by evaluating potential bridge damage and consequences of the seismic
event, such as the estimated direct and indirect losses. This framework offers support to
decision-makers for pre-event planning and risk mitigation, emergency route identifica-
tion, retrofit selection and prioritization, among other critical tasks.
Methodologies for seismic risk assessment of transportation systems have been pre-
sented by many researchers in the field of lifeline earthquake engineering [Kiremidjian, et
al., 2007; Shinozuka et al., 1997; Luna et al., 2008; Werner et al., 2000]. These methodol-
ogies offer a potential framework for assessing likely bridge damage, direct losses due to
repair and replacement of the structures, and some extend this evaluation to include an

Received 23 March 2009; accepted 28 October 2009.


Address correspondence to Jamie E. Padgett, Department of Civil & Environmental Engineering, Rice
University, 6100 Main Street, MS 318, Houston, TX 77005, USA; E-mail: jamie.padget@rice.edu

918
Risk to Charleston Bridges 919

assessment of the impact of the event on network performance and the resulting indirect
economic losses [Kircher et al., 2006; Werner et al., 1997.]
In this article, a detailed seismic risk assessment of the bridge network in Charleston,
South Carolina is conducted. The assessment is performed for a range of hazard levels,
for an inventory of approximately 375 bridges. The seismic risk assessment uses bridge
fragility curves that represent the unique characteristics of bridges in the region, as well
as state-specific bridge repair and replacement cost data. Distribution of damage and loss
estimates are tabulated for the different hazard levels. There are numerous uncertainties
associated with the seismic risk assessment process, and the resulting damage and loss
estimation. The second half of the article will assess the effect of uncertainty on the
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

resulting bridge damage distribution and estimated losses in Charleston, South Carolina.

2. Risk Assessment Framework and Input Models


The seismic risk assessment framework previously proposed by researchers varies in the
extent to which hazards, damage, and losses are treated. However, the general methodolo-
gies have common threads, as was highlighted in Werner et al. [2000]. The risk assessment
approach in this study is limited to an assessment of the bridge damage due to ground
shaking, and considers only the economic losses due to physical damage, rather than indirect
losses due to operation losses or time delay in the transportation system. While these losses
are significant considerations for evaluating the consequences of an earthquake event, the
objective of the study is to assess the sensitivity of the estimated bridge damage and repair
costs to input model variation. Seismic risk assessments are sometimes classified as deter-
ministic or probabilistic, in reference to the hazard itself. Probabilistic analysis is often
carried out by developing loss estimation for multiple simulations and scenario earthquakes,
then aggregating their results. While an SRA may be deterministic in terms of assessing a
specific scenario event, the potential uncertainty in achieving different levels of damage,
economic losses, or other consequences may still be treated probabilistically in the analysis.
The general seismic risk assessment framework used in this study is presented in Fig. 1.
As illustrated in Fig. 1, the first phase of the SRA process for bridge networks is to
initialize the process and define the problem by identifying the characteristics and locations
of the bridge inventory. The bridge inventory is obtained from the National Bridge
Inventory, with supplementary data provided by the South Carolina Department of
Transportation. Scenario earthquake events are used for the example presented herein,
where the magnitude and location of the event must be specified. During the system
analysis, fragility curves for classes of bridges common to the region are utilized. These
fragility curves depict the probability of meeting or exceeding different levels of damage
conditioned upon the ground motion intensity. Thus, the level of ground shaking at the
location of each bridge in the spatially distributed region must be estimated. This facilitates
evaluation of the expected level of damage to each bridge. The bridge damage coupled with
information on the damage ratio (or fraction of replacement cost) and replacement cost data
for different bridge types permits an assessment of the losses. The following sections detail
the different input models and scenarios which will be evaluated as a part of this study.

3. Case Study

3.1. Region of Interest


Charleston, South Carolina (Fig. 2) is located in the southeast United States. Charleston
has a history of large, but infrequent earthquakes. On August 31, 1886, a large earthquake
920 J. E. Padgett, R. DesRoches, and E. Nilsson

Bridge and Roadway


Inventory Earthquake
(characteristics, location, etc.) Scenario

Estimation of Ground
Fragility Curves for
Shaking
Bridge Classes
at Bridge Locations
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

Bridge Damage Bridge Repair Cost Ratios


Replacement Cost Data
State Evaluation (fraction of replacement cost)

Seismic Performance and


Consequence Assessment
(Damage Summary,
Direct Losses, etc.)

FIGURE 1 General flow chart for seismic risk assessment of bridge network.

FIGURE 2 Case study region in Charleston, South Carolina.

(approximate magnitude of 7.0) struck the Charleston region. The earthquake resulted in
60 casualties, and widespread destruction of the built environment in Charleston
[Bollinger, 1977]. The earthquake was felt over a wide area, ranging from Milwaukee,
Risk to Charleston Bridges 921

Wisconsin to Boston, Massachusetts. Summerville, South Carolina, located to the north-


west of Charleston, was subjected to extremely large ground shaking, resulting in the
collapse of many homes and widespread foundation settlement. A repeat of the 1886
earthquake could have a devastating effect on the Charleston region, as well as the local
and global economy.

3.2. Bridge Inventory


Using the National Bridge Inventory (NBI) data for the state of South Carolina [FHWA,
2005], bridges were first filtered by county and bridge identification number to limit the
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

case study evaluation to the region of interest in Charleston, South Carolina. All of the
bridges in Charleston County, and a select few from Berkeley, Dorchester, and
Orangeburg counties, were filtered out using Microsoft Access. The select additions
include the bridges on the I-26 corridor, along I-26 from Charleston to the Bowman
exit, as well as bridges along US 17 from Beaufort, Colleton, Georgetown, Horry, and
Jasper counties. This yielded in a revised inventory containing 375 bridges out of the
overall 10,000 in the state.
The bridges studied in the Charleston region are classified with the methodology
used by Nielson [2005], according to material and construction type. The classifications
simply identify the bridges by both their span configurationsimply supported (SS),
multi-span simply supported (MSSS), multi-span continuous (MSC)as well as by their
girder material typeconcrete or steel. An overall distribution of the bridge classes is
shown in Table 1. The Other bridge category contains all additional bridges not falling
into one of the ten major classifications (i.e., truss, moveable, segmented box girder, and
box single/spread).

3.3. Seismic Hazard


One of the first steps in evaluating the seismic risk for any region is to assess the seismic
hazard or identify the events of interest. In this study, three deterministic scenarios are
used selected based on recommendations from SCDOT: earthquakes of magnitude Mw
4.0, 5.5, and 7.0 located at 32.9 N, 80.0 W, which is approximately 14.5 km outside of

TABLE 1 Distribution of bridge classes within the study area


Bridge type Quantity Percent
MSC_Concrete 1 0.27%
MSC_Steel 31 8.27%
MSC_Slab 14 3.73%
MSC_Conc Box 6 1.60%
MSSS_Concrete 61 16.27%
MSSS_Steel 62 16.53%
MSSS_Slab 118 31.47%
MSSS_Conc Box 2 0.53%
SS_Steel 26 6.93%
SS_Concrete 19 5.07%
Other 35 9.33%
Total 375 100.00%
922 J. E. Padgett, R. DesRoches, and E. Nilsson
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

FIGURE 3 Comparison of hazard for deterministic scenarios: (a) Mw 5.3 and (b) Mw 7.3.

the Charleston city center near Summerville, South Carolina. These hazards produce
maximum ground motion intensities of 0.28 and 0.62 g peak ground acceleration for Mw
4.0 and Mw 7.0, respectively, as shown in Fig. 3.

3.4. Input Models and Risk Assessment


Key input to the risk assessment, as previously indicated, include bridge fragility curves
and repair models. Bridge fragility curves offer the probability of meeting or exceeding a
level of damage given an intensity measure of the ground motion. For this study, the
levels of damage are qualitatively described as slight, moderate, extensive, and complete
damage. Each damage state is associated with an anticipated level of post-event function-
ality, as further discussed in Padgett and DesRoches [2007]. A brief description of the
damage states is presented in Table 2, corresponding to the fragility models incorporated.
The fragility curves adopted are those developed by Nielson and DesRoches [2007].
These fragility curves were developed specifically for nine bridge classes common to the
Central and Southeastern U.S. (CSUS) and are representative of the bridge inventory in
the Charleston region. Uncertainty in component stiffnesses, material strengths, and
geometry were propagated through the analysis. The fragility development considered
damage to multiple vulnerable components, including bearings, columns, and abutments
in the longitudinal and transverse directions. The CSUS fragility curves were developed
for evaluation of the vulnerability of general classes of bridges across a region rather than
bridge specific analysis, and are used in this study to evaluate the probability of the
bridges experiencing different levels of damage in Charleston and subsequent regional
loss estimation. Stochastic dependence between bridge failures in the spatially distributed
region is not considered in the present study. While likelihood of achieving each level of
damage is evaluated for all bridges in the region, the mean value of the damage state is
often presented graphically.
Repair cost models are also required for estimating direct losses due to repair and
replacement of the seismically damaged bridges. Bridge repair costs are assessed as a
fraction of the replacement cost using the damage ratios, D, presented by Basoz and
Mander [1999], as listed in Table 2. The normalized replacement costs for various bridge
types using historic, region specific construction data in South Carolina are show in Table 2,
as a replacement cost per area of bridge deck.
The damage and loss estimates are evaluated and aggregated for the Charleston
region using the seismic risk assessment package, MAEViz [MAEC, 2006]. Within this
Risk to Charleston Bridges 923

TABLE 2 Damage state definitions [Padgett and DesRoches, 2007] and damage ratios
[Basoz and Mander, 1999]
Damage state definition
[Padgett and DesRoches, 2007] Damage ratios [Basoz and Mander, 1999]

Damage Best mean damage Range of damage


state Functionality description ratio (D) ratio

None No reduction in functionality 0.005 00.01


Slight Fully functional within a day 0.03 0.010.03
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

Moderate Reduced functionality for a week 0.08 0.020.15


Extensive Closed for a week, with partial 0.25 0.10.4
functionality beyond 30 days
Complete Complete closure beyond 30 days 1.0 (if n < 3) 0.31.0
2.0/n (if n  3)
n = number of spans.

framework, the damage state is determined from a mean damage ratio, mD, found as
follows:
X
4  
D Dj P DSj ; (1)
j1

where j is the damage state, Dj is the damage ratio for damage state j, and P[DSj] is the
probability of damage state j from the difference in damage state exceedance probabilities
evaluated by entering the fragility curves at the site pga. Given the mean damage ratio, an
expected damage state is presented graphically for intermediate visual inspection. Addi-
tionally, the mean value of the losses for the bridges in the region is found in MAEViz as:
X
L Cn Dn ; (2)
n

where n is the number of bridges in the region, mDn is the mean damage ratio for bridge n,
Cn is the cost to repair the bridge computed as a function of the deck area and replace-
ment cost shown in Table 3. The replacement cost data shown in Table 3, given in dollars
per deck area, reflects the average cost of new construction in South Carolina for different

TABLE 3 Bridge replacement cost data based on South Carolina


statistics [SCDOT, 2007] in dollars per area of bridge deck
Type Cost ($/ft2)
Concrete Girder 67.71
Concrete Box Girder 67.98
Steel Girder 94.37
Slab 60.04
Other (truss, moveable, etc.) 72.53
924 J. E. Padgett, R. DesRoches, and E. Nilsson

bridge types per recent construction data [SCDOT, 2007]. Additionally the standard
deviation of the losses is found as:

rX
L Cn Dn 2 ; (3)
n

where the sD, the standard deviation of the damage ratio for each bridge is:
v
u 4
uX  2  
D t
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

Dj  D P DSj : (4)
j1

The input models and loss estimate approach presented above are subsequently used in
the case study risk assessment of the 375-bridge network in Charleston.

4. Results: Magnitute 5.5 Earthquake Event

4.1. Bridge Damage


The risk assessment is conducted for the Charleston case study to evaluate expected
damage and total direct losses for different scenario events. Figure 4 illustrates the
distribution of bridge damage in the downtown Charleston region due to the Mw 5.5
earthquake event. These types of maps of the anticipated spatial distribution of bridge
damage can be beneficial not only for assessing economic losses, as emphasized in this

Damage States
None
Slight
Moderate
Extensive
Complete

FIGURE 4 Spatial distribution of damaged bridges in downtown Charleston for the Mw


5.5 event.
Risk to Charleston Bridges 925

TABLE 4 Distribution of bridges by damage state and bridge type for the Mw 5.5 event
Damage state
Type None Slight Moderate Extensive Complete TOTAL
MSC Concrete 0 0 1 0 0 1
MSC Steel 12 0 12 7 0 31
MSC Slab 1 2 11 0 0 14
MSC Conc Box 0 1 5 0 0 6
MSSS Concrete 23 8 30 0 0 61
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

MSSS Steel 25 1 32 4 0 62
MSSS Slab 25 22 71 0 0 118
MSSS Conc Box 0 2 0 0 0 2
SS Concrete 6 6 7 0 0 19
SS Steel 26 0 0 0 0 26
Other 30 3 2 0 0 35
TOTAL 148 45 171 11 0 375

article, but can support the identification of viable emergency response routes and
identification of bridges in need of potential retrofit. While a majority of the damaged
bridges in this region may be expected to experience moderate damage, a limited number
of bridges are in the extensive damage state. A summary of the bridges by type and
damage state is shown in Table 4. The anticipated level of damage is a function of the
ground motion at the bridge site, as well as the relative vulnerability of the bridge. For
example, the MSC and MSSS Steel bridges have fragility models that reveal they are
among the most vulnerable bridge types in the region, and the results of the risk
assessment also indicate that the extensively damaged bridges are of these types. It is
also clear that there are a larger number of bridges in the higher damage state in the
location closer to the epicenter of the earthquake.

4.2. Economic Losses


The calculation of expected economic losses is based on the potential damage states and the
repair and construction data from the state of South Carolina, as described in the previous
section. For the Mw 5.5 event, the direct economic losses are approximately $40 million
(Table 5). It is interesting to note that one bridge type alone (MSC steel girder bridge)
accounts for over 64% of the total direct economic losses. This is due to several factors.
Although the MSC Steel girder bridge only accounts for less than 10% of the bridges, it
accounts for 63% of the bridges in the extensive damage state. The economic losses
associated with the extensive damage state are considerably higher than those in the
lower damage states. A bridge in the extensive damage state would have a repair cost
ratio that is three times as high as the moderate damage state, and eight times as high as the
slight damage state. The other reason for the large losses in the MSC steel bridge are due to
the fact that this bridge type tends to have longer bridge lengths and widths as compared to
the other bridge types, as well as the fact that the normalized cost to repair or replace the
steel bridges tends to be higher than other bridge classes. Since the total loss is proportional
to the area, this bridge type tends to have higher loss values. It is also observed that the
bridges that are more robust (i.e., SS steel, SS concrete, MSC concrete box) also contribute
926 J. E. Padgett, R. DesRoches, and E. Nilsson

TABLE 5 Summary of direct losses by bridge type for the


Mw 5.5 event
Type Direct losses
MSC Concrete $13,000
MSC Steel $26,000,000
MSC Slab $830,000
MSC Conc Box $200,000
MSSS Concrete $2,800,000
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

MSSS Steel $6,000,000


MSSS Slab $1,400,000
MSSS Conc Box $60,000
SS Concrete $510,000
SS Steel $94,000
Other $2,400,000
TOTAL $40,307,000

less to the total direct losses. The relative contribution of bridges to the loss estimate offers
one approach to help identify and prioritize bridges in need of retrofit.

5. Results: Comparison of Different Earthquake Magnitudes


The seismic risk assessment was performed for three different hazards, Mw 4.0, 5.5, and
7.0 (epicenter in Summerville, South Carolina), using the MAEViz platform [MAEC,
2006]. The distribution of expected damage for the three hazard levels is shown in Fig. 5.

FIGURE 5 Distribution of damage as a function of earthquake magnitude.


Risk to Charleston Bridges 927

120 106

100 106

80 106

Loss ($) 60 106

40 106

20 106
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

0
4.0 5.5 7.0
Earthquake Scenario (Mw)

FIGURE 6 Direct economic loss estimates for three scenario earthquakes (Mw 4.0, 5.5,
and 7.0).

The results show that for the Mw 7.0 event, over 85% of the bridges are damaged, with
73% of the bridges having moderate to complete damage. For the Mw 5.5 event,
approximately 60% of the bridges are damaged, with nearly 50% of the bridges having
moderate to complete damage. Finally, the Mw 4.0 earthquake results in only 17% of the
bridges having damage, and less than 9% have moderate or greater damage.
It is interesting to note that a Mw 4.0 scenario results in expected damage states of
only slight or moderate damage to 65 bridges, with the remaining bridges having no
damage. This is an indication that for pre-event planning purposes the Mw 4.0 earthquake
might be a viable threshold upon which inspection teams are mobilized following an
earthquake event. However, this would depend on the location of the epicenter for the
particular earthquake. It is also important to note that as previously highlighted in the
input model and risk assessment section, while expected damage states are presented
graphically there is probability of achieving each damage state even at the lower level
events, which is further propagated through the loss estimation.
As shown in Fig. 6, for a Mw 4.0 seismic event, direct economic losses are estimated
to be close to $6.3 million. In contrast, the more severe earthquake scenario, Mw 7.0,
produces direct losses of approximately $90 million. As the earthquake scenarios increase
in intensity, the direct economic losses increase exponentially, and the error about that
estimate increases as well. While outside of the scope of the current study, indirect losses
in a transportation network due to bridge damage are often orders of magnitude greater
than the repair and replacement costs alone. For example, past studies have shown that
the indirect losses due to rerouting may be roughly 720 times direct losses [ATC, 1991],
revealing that for an increase of 13 times, the total losses in the Charleston region may be
on the order of $90 million to over $1 billion for the Mw 4.0 and 7.0, respectively.
Refined total loss estimates would require transportation modeling, which is outside of
the scope of this study.

6. Uncertainties and Sensitivity Study


While there have been many studies that propose and illustrate the viability of the risk
assessment framework, the results may depend heavily on the availability and reliability
of utilized tools and input models. These include such items as ground motion models,
928 J. E. Padgett, R. DesRoches, and E. Nilsson

fragility information on the bridge vulnerability, repair cost information, among others.
Different modeling assumptions and input tools may be classified as epistemic uncertain-
ties. An epistemic uncertainty is often defined a knowledge-based uncertainty, which
stems from incomplete data, ignorance, or modeling assumptions. The adoption of
different input models in the SRA framework could potentially have a significant effect
on the overall results and conclusions of the study.
Past studies have evaluated the sensitivity of loss estimates to input model variation
in other systems, particularly buildings. Crowley et al. [2005] assessed the impact of a
number of uncertain parameters, including ground motion modeling, structural demand,
and capacity estimates, on regional building damage. Porter et al. [2002] evaluated the
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

sensitivity of loss estimates for a single concrete moment-frame building and found, like
Crowley observed for regional damage, that the building capacity (limit at which damage
is expected) was the most important uncertain parameter followed by ground motion
characteristics. In a study assessing the average annual losses to a regional inventory of
low-rise wood framed buildings, Grossi [2000] compared using default models in
HAZUS to updated input models for the seismic hazard as well as the inventory
square footage and fragility. She found that models defining the seismic hazard, such
as the recurrence model for the earthquake and attenuation relationship were the most
critical updates, followed by the square footage and fragility.
While the studies listed have offered insight on the relative importance of different
loss modeling parameters for building inventories, few have assessed the impact on the
regional seismic risk to transportation networks. The relative sensitivity of the highway
bridge damage and loss estimates to different input models are evaluated as a follow-up
phase of the study in Charleston. This helps to identify critical components of the risk
assessment framework that significantly impact the overall results of a regional transpor-
tation network assessment, including bridge damage and direct economic losses due to
repair and replacement. This study emphasizes the difference due to assumed input
models, rather than variation about the estimate due to uncertainty modeled by a
particular input model. The Charleston region previously presented is used as an example
to gain insight on the effect of different input fragility curves for evaluating the perfor-
mance of bridges common to the region, as well as different estimates of the damage ratio
for repair cost modeling and loss estimation.

7. Input Parameters
Two different scenario earthquake events are considered as a part of the sensitivity study.
This permits an evaluation of whether or not the conclusions of the study are dependent
upon the level of the hazard. The characteristic scenario events assessed for Charleston
are moment magnitude 5.3 and 7.3 located 14.5 km outside of the city center near
Summerville. In order to estimate the level of ground shaking at the location of each
bridge, a weighted average of different attenuation functions is used [MAEC, 2006]. This
is to acknowledge the findings of past work which has indicated the importance of
considering the epistemic uncertainty in ground motion models, particularly attenuation
of ground motion for spatially distributed systems. Thus, the ground motions models
themselves are not a focus of this study and the epistemic uncertainty associated with
them is captured and treated explicitly in each scenario, rather than evaluating the
sensitivity of the results to different models.
The two input models that are considered in this study are change in fragility model
and in repair cost model (specifically due to change in damage ratio). The fragility
models considered in the sensitivity study for bridge classes common to the Charleston
Risk to Charleston Bridges 929

region include the Nielson and DesRoches [2007] fragility curves developed for the
CSUS region as previously discussed in the case study, as well as those adopted in
HAZUS-MH [FEMA, 2005]. The bridge fragility curves currently used in HAZUS-MH
were developed using a nonlinear static approach in past work by Basoz and Mander
[1999] and Dutta [1999]. These sets of fragility models are subsequently termed CSUS
and HAZUS fragilities, respectively. A detailed discussion of the difference in the two
models is presented in Nielson and DesRoches [2007], which illustrated that for some
bridge types (i.e., multi-span simply supported steel or concrete girder bridges) the CSUS
fragility curves exhibit lower vulnerability than originally anticipated in the HAZUS
curves, while for other bridge types (i.e., multi-span continuous steel and concrete girder
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

bridges) the CSUS fragility curves indicate a much higher vulnerability than depicted in
the HAZUS curves.
The two damage ratios considered in the sensitivity study are those formerly pre-
sented in the case study [Basoz and Mander, 1999] termed Basoz, as well as the damage
ratios presented in REDARS [Werner et al., 2006] as shown in Table 6. Figure 7 shows a
comparison of the damage ratios for an example bridge with three spans, noting that the
Basoz damage ratios are a function of the number of spans, while the REDARS damage
ratios do not change depending upon number of spans. As illustrated in the plot, the
REDARS damage ratios imply a larger anticipated repair cost for the moderate, exten-
sive, and complete damage states in particular. Moreover, they indicate a more linearly
increasing damage ratio than exhibited in the Basoz damage ratios.

TABLE 6 REDARS repair cost ratios [Werner et al., 2006]


Damage state Best mean damage ratio (D) Range of damage ratio
None 0.00 00.01
Slight 0.03 0.010.05
Moderate 0.25 0.050.5
Extensive 0.75 0.50.8
Complete 1.00 0.81.0

1.0

Basoz
0.8 REDARS
(for n = 3)
Damage Ratio

0.6

0.4

0.2

0.0
None Slight Moderate Extensive Complete
Damage state

FIGURE 7 Comparison of Basoz and REDARS damage ratios for a three span bridge.
930 J. E. Padgett, R. DesRoches, and E. Nilsson

8. Results
The sensitivity study is performed by conducting the regional risk assessment for
Charleston with different input models. The experiment conducted is a full factorial
design with each factor (fragility curves and damage ratios) having two categorical levels
(22), and a replication to consider two different levels of earthquake (2 x 22), for a total of
8 runs. Table 7 lists the risk assessment runs (scenarios) for the magnitude 5.3 and 7.3
events. The total estimated direct losses and standard deviation of the losses are compared
in the Table, indicating a potential range in estimated direct losses between $71,400,000
and $267,000,000 for the upper level event, and between $27,900,000 and $125,000,000
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

for the lower level event for different input model combinations. Similarly, the standard
deviation about those loss estimates varies for each scenario.
Figure 8 shows the percent difference in the mean value and standard deviation of the
losses relative to the base case (CSUS fragility curves and Basoz damage ratios). It is
noted that the base case uses the same input models considered in the Charleston case
study previously presented. This figure reveals that regardless of event magnitude, the use
of the REDARS damage ratios results in larger economic losses, as anticipated, due to the
increase in damage ratio and repair cost estimate for each damage state. The expected
value of total losses increases by nearly 150% for each earthquake level when the same
fragility curves are used as the base case (CSUS). In fact, the change in damage ratios
results in the largest impact on the loss estimate and standard deviation about that
estimate.
The use of HAZUS fragility curves results in a decrease in expected direct economic
losses for a given damage ratio. This finding is potentially counter-intuitive given the
total number of bridges expected in each damage state shown for each run in Fig. 9 for
Mw 5.3 and 7.3. As these figures reveal, the use of the HAZUS fragility curves as
opposed to the CSUS fragilities for the same damage ratio (Basoz) result in a larger

TABLE 7 SRA runs for sensitivity study and results


Estimated total Standard
Run number Scenario direct losses deviation
7.3 Base Case 7.3, CSUS Fragilities, $105,000,000 $19,900,000
Basoz Damage Ratios
7.3.A 7.3, HAZUS Fragilities, $71,400,000 $17,700,000
Basoz Damage Ratios
7.3.B 7.3, HAZUS Fragilities, $197,000,000 $27,500,000
REDARS Damage Ratios
7.3.C 7.3, CSUS Fragilities, $267,000,000 $41,200,000
REDARS Damage Ratios
5.3 Base Case 5.3, CSUS Fragilities, $50,900,000 $14,700,000
Basoz Damage Ratios
5.3.A 5.3, HAZUS Fragilities, $27,900,000 $11,500,000
Basoz Damage Ratios
5.3.B 5.3, HAZUS Fragilities, $74,200,000 $10,700,000
REDARS Damage Ratios
5.3.C 5.3, CSUS Fragilities, $125,000,000 $24,900,000
REDARS Damage Ratios
Risk to Charleston Bridges 931

HAZUS Fragilities, REDARS Damage Ratios

Low Level Event


E[Loss]
stdev[Loss]

(Mw 5.3)
HAZUS Fragilities, Basoz Damage Ratios

CSUS Fragilities, REDARS Damage Ratios

HAZUS Fragilities, REDARS Damage Ratios


High Level Event
(Mw 7.3)

HAZUS Fragilities, Basoz Damage Ratios


Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

CSUS Fragilities, REDARS Damage Ratios

100 50 0 50 100 150 200


% Variation from Base Model

FIGURE 8 Comparison of the change in expected value of losses and standard deviation
of losses relative to the base case (CSUS fragilities, Basoz damage ratios).

250 300
Mw = 5.3 Mw = 7.3
CSUS/Basoz CSUS/Basoz
CSUS/REDARS 250 CSUS/REDARS
200
HAZUS/Basoz HAZUS/Basoz
Number of Bridges

Number of Bridges

HAZUS/REDARS HAZUS/REDARS
200
150
150
100
100

50
50

0 0
None Slight Moderate Extensive Complete None Slight Moderate Extensive Complete
Damage State Damage State
(a) (b)

FIGURE 9 Number of bridges by expected damage state for each sensitivity study
simulation at the (a) Mw 5.3 event and the (b) Mw 7.3 event.

number of bridges in the extensive and complete damage states; however, the expected
value of the losses is lower for the HAZUS fragility curves. This can be attributed to the
fact that: (1) The HAZUS fragility curves have been shown to underestimate the damage
of MSC bridges [Nielson and DesRoches, 2007], which are among the costliest bridges to
repair and replace and the bridges contributing the most to the economic losses (i.e.,
Tables 2 and 4); and (2) Damage to other bridge types, such as the MSSS concrete girder,
slab, and steel girder bridges, may be overestimated by using the HAZUS fragilities,
yielding more total bridges in the upper damage states, yet with insignificant net effect on
the direct losses relative to the contribution of other bridges.
Figure 8 also indicates the interaction effects of changing both the fragility curves
and damage ratios for a given earthquake scenario. The reduction in expected value of
losses due to using HAZUS fragility curves is countered and dominated by the increase in
932 J. E. Padgett, R. DesRoches, and E. Nilsson

losses due to using REDARS damage ratios, yielding a net increase in economic losses of
46% and 88% for the Mw 5.3 and 7.3 events, respectively. The findings reveal that while
the risk assessment is very sensitive to both the assumed fragility curves and damage
ratios, the estimate of total expected economic losses is more sensitive to the vast
differences in damage ratio models.

9. Conclusions
In this article, the risk assessment framework for evaluating bridge damage and economic
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

losses due to earthquake events is presented for application to a case study in Charleston,
South Carolina. The bridge network for the case study consists of 375 bridges of varying
types, and the risk assessment conducted for three different scenario events utilizes region
specific bridge fragility curves and construction cost data for damage and loss estimation.
The case study reveals expected damage states of moderate, extensive, or complete
damage for over 85% of the Charleston bridges due to a Mw 7.0 event located approxi-
mately 14.5 km outside of the city center, near Summerville, South Carolina.
Additionally, while noting the potential for achieving each damage state is assessed
using the fragility curves and propagated through the loss estimation, the mean damage
state alone indicates that nearly 20% of the bridges may suffer some level of damage for a
Mw 4.0 event. Hence, this low level event may still warrant immediate deployment of
inspection teams. The expected value of direct economic losses due to bridge repair alone
are on the order of $40 million for the Mw 5.5 event, with both the loss estimate and
standard deviation about the estimate increasing exponentially with increasing event
magnitude. For the regional inventory in Charleston, the more vulnerable bridge types,
such as the multi-span continuous steel girder bridges, are expected to contribute dis-
proportionately to the economic losses, despite their relatively small percentage of the
overall bridge inventory. These results indicate that such bridge types may be critical
priorities for retrofit.
A sensitivity study is conducted to evaluate the impact of assumed SRA input models
on the resulting loss estimates, assessing the effect of fragility models and damage ratios
for upper and lower level events. In a full factorial design, both the CSUS specific bridge
fragility curves relative to current HAZUS fragilities, as well as REDARS versus Basoz
(currently implemented in HAZUS) damage ratios are considered. The findings reveal a
strong sensitivity of the resulting loss estimates, and variability about the estimate, to
assumed fragility models and damage ratios. The expected value of losses differ on the
order of 150% for both the upper and lower level events considered in the sensitivity
study (Mw 5.3 and 7.3). The roughly linearly increasing damage ratio and repair cost
estimate for the REDARS model, as opposed to roughly exponential increase with the
Basoz ratios, yields the greatest impact on increasing the loss estimate. For the case study
inventory and cost figures considered, the use of the HAZUS fragility curves resulted in
lower loss estimates. However, this was found to be a function of the type of bridges
found in the region and relative contribution of different bridge types to total losses, since
for some bridges HAZUS fragilities indicate an increase in vulnerability relative to the
CSUS specific models, while for other bridge types they depict a lower fragility.

Acknowledgments
This study has been supported by the Earthquake Engineering Research Centers program
of the National Science Foundation under Award Number EEC-9701785 (Mid-America
Risk to Charleston Bridges 933

Earthquake Center). The South Carolina Department of Transportation (SCDOT) is


gratefully acknowledged for their input and data sharing throughout the research project.

References
ATC (Applied Technology Council) [1991] Seismic vulnerability and impact of disruption of
lifelines in the conterminous United States, Report No. ATC-25. Redwood City, California.
Basoz, N. and Mander, J. B. [1999] Enhancement of the lifeline transportation module in
HAZUS, National Institute of Building Sciences Report, Washington, D.C.
Bollinger, G. A. [1977] Reinterpretation of the intensity data for the 1886 Charleston, South
Downloaded by [IIT Indian Institute of Technology - Mumbai] at 01:58 08 November 2017

Carolina, earthquake, in Studies Related to the Charleston, South Carolina, Earthquake of


1886: A Preliminary Report, ed. D. W. Rankin, U.S. Geological Survey Professional Paper, pp.
1732.
Crowley, H., Bommer, J. J., Pinho, R., and Bird, J. [2005] The impact of epistemic uncertainty on
an earthquake loss model, Earthquake Engineering and Structural Dynamics 34, 16531685.
Dutta, A. [1999] On energy based seismic analysis and design of highway bridges, PhD
Dissertation, State University of New York at Buffalo, Buffalo, New York.
FEMA [2005] HAZUS-MH (Hazards US Multi-Hazard) Software, Federal Emergency Management
Agency, Washington D.C.
FHWA [2005] National Bridge Inventory Data, Federal Highway Administration, Washington,
D.C.
Grossi, P. A. [2000] Quantifying the uncertainty in seismic risk and loss estimation, PhD
Dissertation, University of Pennsylvania, Philadelphia, Pennsylvania.
Kircher, C., Whitman, R., and Holmes, W. [2006] HAZUS earthquake loss methodologies,
Natural Hazards Review 7(2), 4560.
Kiremidjian, A., Moore, J., Fan, Y. Y., Yazlali, O., Basoz, N., and Williams, M. [2007] Seismic
risk assessment of transportation networks, Journal of Earthquake Engineering 11, 371382.
Luna, R., Hoffman, D., and Lawrence, W. T. [2008] Estimation of earthquake loss due to bridge
damage in the St. Louis metropolitan area. I: Direct losses, Natural Hazards Review 9(1), 111.
MAEC (Mid-America Earthquake Center) [2006] MAEViz Software: Mid-America Earthquake
Center Seismic Loss Assessment System, http://mae.ce.uiuc.edu/software_and_tools/maeviz.html,
Urbana, Illinois
Nielson, B.G. [2005] Analytical fragility curves for highway bridges in moderate seismic zones,
PhD dissertation, Georgia Institute of Technology, Atlanta, Georgia.
Nielson, B. G. and DesRoches, R. [2007] Seismic fragility curves for typical highway bridge
classes in the Central and Southeastern United States, Earthquake Spectra 23(3), 615633.
Padgett, J. E. and DesRoches, R. [2007] Bridge functionality relationships for improved seismic
risk assessment of transportation networks, Earthquake Spectra 23(1), 115130.
Porter, K. A., Beck, J. L., and Shaikhutdinov, R. V. [2002] Sensitivity of building loss estimates to
major uncertain variables, Earthquake Spectra 18(4), 719743.
SCDOT [2007] Personal communication: South Carolina bridge construction cost data, with J.
Padgett and R. DesRoches.
Shinozuka, M., Chang, S., Eguchi, R., Abrams, D., Hwang, H., and Rose, A. [1997] Advanced in
earthquake loss estimation and application to Memphis, TN, Earthquake Spectra 13(4), 739758.
Werner, S. D., Taylor, C. E., Moore, III, J. E., Walton, J. S., and Cho, S. [2000] A risk-based
methodology for assessing the seismic performanceof highway systems, Technical Report No.
MCEER-00-0014, Multidisciplinary Center for Earthquake Engineering Research, State
University at Buffalo, Buffalo, New York.
Werner, S. D., Taylor, C. E., and Moore, J. E. [1997] Loss estimation due to seismic risks to
highway systems, Earthquake Spectra 13(4), 585604.
Werner, S. D., Taylor, C. E. Cho, S., Lavoie, J. P., Huyck, C. K., Eitzel, C., Chung, H., and Eguchi,
R. T. [2006] REDARS 2 methodology and software for seismic risk analysis of highway
systems, Report No. MCEER-06-SP08, Buffalo, New York.

You might also like