Learning Center
Plans & pricing Sign in
Sign Out

Measuring improvements in the disaster resilience of communities


									                     Submitted for publication in Earthquake Spectra

Measuring Improvements in the Disaster
Resilience of Communities

Stephanie E. Chang, M.EERI, and Masanobu Shinozuka, M.EERI

  Corresponding (first) author:    Stephanie E. Chang
  Mailing address:                 Dept. of Geography, Box 353550, Univ. of Washington,
                                   Seattle, WA 98195
  Phone:                           (206)616-9018
  Fax:                             (206)543-3313
  E-mail address:        

  Submission date for review copies: to be completed
  Submission date for camera-ready copy:
                          Submitted for publication in Earthquake Spectra

Measuring and Improving the Disaster
Resilience of Communities

Stephanie E. Chang, M.EERI, and Masanobu Shinozuka

          This paper proposes a conceptual and measurement framework to
      quantitatively assess the disaster resilience of communities. The concept of
      resilience suggests a much broader framework than simply reducing monetary
      losses. Resilience can be conceptualized along four interrelated dimensions:
      technical, organizational, social, and economic. Evaluating community resilience
      and the degree to which mitigation and preparedness activities further resilience
      goals requires a multidisciplinary approach and quantitative measures. This paper
      proposes measures that relate the impacts of potential earthquakes to multi-
      dimensional performance standards. The approach is demonstrated in a case
      study of the Memphis, Tennessee, water delivery system. An earthquake loss
      estimation model provides a starting point for quantifying resilience. The analysis
      compares two retrofit strategies and finds that only one improves resilience over
      the status quo. While the resilience framework can be useful for guiding
      mitigation and preparedness efforts, further research is needed to fully implement
      the concept.

    The policies, practices, and theories of disaster management have in recent years
increasingly promoted the goal of achieving disaster-resilient communities. In the late 1990s,
the Federal Emergency Management Agency (FEMA) instituted Project Impact, a program to
encourage communities to better understand their risks, engage in education, and invest in
mitigation activities through public-private partnerships (FEMA, 2000; Nigg et al., 2000). A
synthesis of scholarship in the hazards field also emphasized fostering community resilience
as a key goal in halting the rapid growth of disaster losses (Mileti, 1999). In that context, it
defined a disaster-resilient community as one that “can withstand an extreme natural event
with a tolerable level of losses” and one that “takes mitigation actions consistent with
achieving that level of protection.” (p.5)
    As a step toward implementing this idea, Bruneau et al. (forthcoming) proposed a
conceptual framework for defining and measuring disaster resilience. They defined
resilience as “the ability of social units (e.g. organizations, communities) to mitigate hazards,
contain the effects of disasters when they occur, and carry out recovery activities in ways that
minimize social disruption and mitigate the effects of future earthquakes.” (p.5) More
specifically, a resilient system should demonstrate three characteristics: reduced failure
probabilities, reduced consequences from failures, and reduced time to recovery.

(SEC) Dept. of Geography, Box 353550, Univ. of Washington, Seattle, WA 98195-3550
(MS) University of California, Irvine, Dept. of Civil and Environmental Engineering, Irvine, CA 92697-2175
                      Submitted for publication in Earthquake Spectra

    Bruneau et al. pointed out the need for quantitative measures of the resilience concept.
Quantitative measurement can help in understanding how various factors contribute to
resilience and why some communities are more resilient than others. Quantitative measures
would also be able to address questions such as: How disaster-resilient is a community? Is
the community becoming more resilient over time? What types of risk reduction efforts can
most effectively move the community towards disaster resilience?
    Earthquake loss estimation models, which broadly quantify the potential impacts of
earthquakes on a region, provide a natural starting point for attempts to measure community
resilience. These models combine extensive spatial databases (e.g., of regional soil
conditions, populations, and buildings) with computational algorithms for physical damage,
monetary loss, and often human casualties and economic disruption.
    Loss estimation models have gained substantial attention in recent years and are
becoming well-established in the literature and, increasingly, in practice. A well-known
example is HAZUS, FEMA’s nationally applicable loss estimation methodology and
software. Earthquake Spectra devoted a theme issue to loss estimation models in 1997
(vol.13, no.4). While these models generally focus on property damage, particularly to the
building stock, there have been several recent studies that emphasize lifeline infrastructure
systems and how their damage could lead to regional economic disruptions (Applied
Technology Council, 1991; Rose et al., 1997; Cho et al., 2001; Chang et al., 2002; Kim et al.,
2002). Because loss estimation models generally produce results in terms of expected dollar
losses, they do not directly measure resilience.
    This paper aims to develop and demonstrate a measurement framework for disaster
resilience. It builds on the conceptual framework developed by Bruneau et al., as well on
advances in loss estimation methodologies. It proposes a series of resilience measures and
illustrates, through a case study application of the Memphis water system, how they can be
used to support decision-making for strengthening community resilience. A secondary
objective of the case study is to explore the extent to which loss estimation models can be
used to measure resilience. This entails identifying linkages between loss estimation
modeling and resilience measurement, and identifying gaps that require further research.

                             CONCEPTUAL FRAMEWORK

        Bruneau et al. (forthcoming) suggest that resilience can be conceptualized along the
following four interrelated dimensions: technical, organizational, social, and economic
(TOSE). Technical resilience refers to how well physical systems perform when subjected to
earthquake forces. Organizational resilience refers to the ability of organizations to respond
to emergencies and carry out critical functions. Social resilience refers to the capacity to
reduce the negative societal consequences of loss of critical services in earthquakes.
Economic resilience refers to the capacity to reduce both direct and indirect economic losses
resulting from earthquakes. Of these four (TOSE) dimensions, the technical and
organizational dimensions are most pertinent to the performance and resilience of critical
systems such as electric power, water, hospital, and emergency response. The social and
economic dimensions are most relevant to the performance and resilience of the community
as a whole. This is illustrated in Figure 1.
                        Submitted for publication in Earthquake Spectra

                                                           Resilience Objectives…
                                                           …at community level:
                                                              SOCIAL – minimize casualties
                                                                 & social disruption
                                                              ECONOMIC – minimize
                                                                 economic disruption

                          systems                          …at individual system level:
        Hospital                      Electric power          TECHNICAL – minimize loss
        system                            system                 of facilities & equipment
                                                              ORGANIZATIONAL – minimize
        Local emerg.                    Water                    service disruption through
        mgmt. system                   system                    organizational response

Figure 1. Relationship between resilience objectives and scale of analysis.

        Bruneau et al. (forthcoming) identify four properties of resilience: robustness,
rapidity, redundancy, and resourcefulness (4 R’s). Robustness refers to strength, or the
ability of elements, systems, and other units of analysis to withstand a given level of stress or
demand without suffering degradation or loss of function. Rapidity refers to the capacity to
meet priorities and achieve goals in a timely manner in order to contain losses and avoid
future disruptions. Redundancy is the extent to which elements, systems, or other units of
analysis exist that are substitutatible, i.e., capable of satisfying functional requirements in the
event of disruption, degradation, or loss of functionality. Resourcefulness is the capacity to
identify problems, establish priorities, and mobilize resources when conditions exist that
threaten to disrupt some element, system, or other unit of analysis. Robustness and rapidity
can be viewed as the desired ends of resilience-enhancing measures. Redundancy and
resourcefulness are some of the means to these ends.

                              MEASUREMENT FRAMEWORK
    Because resilience is a multi-dimensional concept, developing measures of resilience that
are quantifiable, succinct, and meaningful remains a challenge. Bruneau et al. (forthcoming)
offer a series of 80 illustrative measures set out in 5 tables. They relate to 5 systems
(“global”, electric power, water, hospital, and response and recovery systems), the 4
dimensions of resilience (TOSE), and the 4 properties of resilience (4 R’s). For example, a
social performance measure for rapidity of the hospital system might be: “all injuries treated
in first day.” The measurement framework proposed here makes two significant refinements
to the Bruneau et al. framework: (1) it outlines a more succinct series of resilience measures,
                                               Submitted for publication in Earthquake Spectra

and (2) it reframes the measures in a probabilistic context. In addition, in the next section of
this paper, the measures are applied in a case study of the Memphis water system.
    Figure 2 sketches the measurement framework proposed here. System performance Q is
plotted over time. Note that the concept of system performance is applicable not only to
infrastructure systems such as the water delivery network, but also to the community as an
urban system.

                                                                                        Example where
                 System Performance (Q)

                                                                                          r0 > r *,
                                                                                          t1 < t *

                                          r*       r0

                                                                   e arth
                                                        t0                         t1        t*
Figure 2. Resilience measurement framework.

    Earthquake impacts refer to a “without-earthquake” timepath. A simplification that is
often made in practice is to compare pre- and post-disaster states, assuming that pre-disaster
conditions are “normal” and static. The proper comparison is between “with” and “without”
disaster scenarios. For some types of disruptions, such as economic disruptions that last for a
long period of time (e.g., years), this distinction may be important and would require a
forecast of what would have happened without the earthquake.
    Resilience in Figure 2 is defined by comparing loss of system performance to pre-defined
performance standards of robustness (r*) and rapidity (t*). The initial loss r0 is compared
with r*, an absolute level of loss which can be some pre-specified “maximum acceptable
loss.” The time to full recovery t1 is compared with t*, an absolute duration of loss which
can be some pre-specified “maximum acceptable disruption time.” Figure 2 illustrates the
case where, in the particular scenario earthquake, the system meets the rapidity performance
standard (t1<t*) but not the robustness standard (r0>r*).
    Figure 2 pertains to a particular earthquake scenario. Since resilience refers to a capacity
for dealing with potential future events, it is necessary to introduce the probabilistic element.
This consists of (1) the probability that the system will meet the performance standards r*
and t* in a given earthquake i, and (2) the probability of occurrence of various seismic events.
    We can define resilience, as a starting point, as the probability that the system of interest
will meet pre-defined performance standards A in a scenario seismic event of magnitude i, or
                                                         Pr( A i ) = Pr( r0 < r * and t1 < t * )               (1)
                       Submitted for publication in Earthquake Spectra

In other words, resilience is quantified as the probability of meeting both robustness and
rapidity standards in event i. Often, discussion and decision-making regarding “acceptable”
performance standards can most readily take place in the context of scenario events.
    Beyond this, it may also be useful to define broad system resilience ZA in reference to the
entire range of possible earthquake events:
                              Z A = ∑ Pr ( A i ) ⋅ Pr(i )                                  (2)

Alternatively, different performance standards may be defined for different classes of events.
In lifeline earthquake engineering, for example, it is quite common for different standards to
be specified for a more common, lower-magnitude “operating basis earthquake” as well as a
rare, higher-magnitude “design basis earthquake.”
    The definition of performance measures and standards is clearly central to the
quantification of resilience. Ideally, these definitions should be developed in consultation
with decision-makers, the public, and other potential end-users. Because both professional
and lay persons are not accustomed to thinking in these terms, this may a formal investigation
(e.g., a structured survey) and/or a consensus-seeking discussion. While this was beyond the
scope of the current study, examples were developed (shown in Table 1) for purposes of the
case study. These examples are also useful for stimulating discussion and could be used in a
more formal investigation of performance measures and standards.
    The examples shown in Table 1 pertain to a water lifeline system and its role in broader
community resilience. “Technical” and “organizational” performance standards are defined
at the level of the water system. Technical performance refers to the extent of physical
damage to the network, measured in this case by the number of major pumping stations lost
and the percentage of pipes broken. Organizational performance refers to the extent of
service disruption, measured here as the percentage of population losing water service. Note
that organizational performance depends not only on the extent of physical damage, but also
on network flow conditions. This in turn reflects the degree of network redundancy.
Moreover, in principle, network flow should also reflect the utility agency’s degree of
organizational resourcefulness. This refers to the ability of the utility to respond to the
emergency by rapidly detecting damage, efficiently deploying repair crews, using shutoff
valves to isolate damage, implementing mutual aid agreements to speed up repairs, and so on.
     For both technical and organizational performance, “robustness” measures indicate the
initial post-earthquake conditions while “rapidity” measures consider the repair and
restoration timeframe. As mentioned previously, the specific examples of r* and t* in Table
1 are provided for illustrative purposes only.
    “Social” and “economic” measures are defined at the level of the community as a whole.
Social performance could refer to the population displaced from their homes – that is, forced
to seek emergency shelter – due to the disaster. (Another possible measure might refer to the
population needing medical attention.) Loss of water service to residences could be a main
source of population displacement. A complete analysis should also consider other factors
such as housing damage that could also force people to seek emergency shelter. Similarly,
economic performance refers to the loss of gross regional product (GRP) due to the disaster
and should consider water outage to businesses along with other sources of economic
                           Submitted for publication in Earthquake Spectra

Table 1. Example performance measures and standards.
Dimension of                Performance                  Robustness Standard         Rapidity Standard
Resilience                  Measure                      (r*)                        (t*)
 (unit of analysis)
Technical                   Network physical             ≤1 major pump               <1 week until all
 (water system)              condition                    station loses               pump stations and
                                                          function                    99% of pipes intact
Organizational              Water service                <5% of population           <1 week until 99% of
 (water system)                                           loses water service         population has
                                                                                      water service
Social                      Population living at         <5% of population           <1 week until 99% of
 (community)                 home(a)                      displaced from              population living at
                                                          homes                       home
Economic                    Economic activity            <5% of GRP(b) lost          <1 week until return
 (community)                                                                          to 99% of GRP

Notes: (a) as opposed to staying in emergency shelters or other temporary housing; (b) gross regional product.

     The conceptual and measurement frameworks described above were applied in a case
study of the water delivery system serving Memphis, Tennessee. This application builds on
prior research that developed an earthquake loss estimation model for that system (Chang,
2002; Chang et al., 2002). An important goal of the application was to gain insight on the
utility of loss estimation models for measuring resilience.

   Memphis Light, Gas, and Water Division (MLGW) provides water to the City of
Memphis and, with the exception of a few unincorporated municipalities, the remainder of
Shelby County (2000 population 897,000). MLGW draws water from an underground
aquifer that is accessed by wells. The water network (Figure 3) includes approximately 1370
km of buried pipe, as well as several pumping stations, elevated tanks, and booster pumps.
    The network is modeled here as a series of some 960 demand nodes and supply nodes,
connected by 1300 links. Nine pumping stations, each of which includes both booster pumps
and treatment facilities, are represented by supply nodes.1 Because the elevated tanks have
very limited capacity, they are not included in the model. The demand nodes indicate how
much water is normally required from the network at locations throughout the system. The
links primarily represent 10-inch and larger water mains. While this constitutes only some
30 percent of the entire system (which is predominantly 6-inch distribution pipes), these
larger mains would be given priority in the event of a disaster (C. Pickel, personal
communications, June 3, 2003).
    The database includes attributes of the pipelines and pumping stations that affect seismic
performance. Pipeline damage in earthquakes depends upon pipe diameter, material, and soil

 Since the time the data was originally obtained for the model, MLGW has added a tenth pumping station in the
eastern portion of the county and a few major pipelines in the northwestern part of the county to enhance
network redundancy.
                       Submitted for publication in Earthquake Spectra

condition. Pumping stations are characterized by fragility curves that indicate the probability
of failure for different levels of ground shaking, in this case, peak ground acceleration (PGA).
The fragility curves are calibrated on estimates of seismic capacity as reported in Hwang et al.

Figure 3. MLGW water network and population density in Shelby County

    The Memphis region is at risk from earthquakes originating in the New Madrid Seismic
Zone (NMSZ) in the central United States (Johnston and Nava, 1985). Historically, the
NMSZ produced three events of Richter magnitude 8.0 or greater in the winter of 1811-12.
At that time, damages were relatively minor, as the region was sparsely populated. However,
the earthquakes reportedly caused the Mississippi River to flow backgrounds for a time and
rang church bells as far away as Boston.
    MLGW is well aware of the seismic risk and has been engaging in a seismic upgrading
program since the late 1980s (Pickel, communications). This program has focused on the
pumping stations, which are critical for maintaining the water supply. Other retrofit
strategies, such as large-scale replacement of seismically vulnerable cast iron pipes, were not
considered due to the great expense that would be involved. The seismic upgrading program
has been partly financed by MLGW’s own budget and partly by grants from the Federal
Emergency Management Agency (FEMA) on a cost-share basis.
   The pumping station upgrades initially involved installing backup electric power
generators. Later, new facilities were also designed with seismic considerations. While
some older pumping stations have been seismically retrofitted, it may be another decade or
more before the retrofit program is completed.
                           Submitted for publication in Earthquake Spectra

    The current study models two alternative mitigation strategies. Retrofit 1 represents an
actual strategy adopted by MLGW, in which the Morton and Davis pumping stations are
seismically upgraded. These two facilities are labeled with a “1” in Figure 3 above.
Mitigation is represented in the model by shifting the respective fragility curves. The
mitigated fragility curves were calibrated based on contractor reports of the upgrading
    Retrofit 2 represents a hypothetical strategy in which two alternative pumping stations,
Mallory and Sheahan, are seismically upgraded, instead of Morton and Davis.3 These two
facilities are labeled “2” in Figure 3 above. They were selected on the basis of their location
in densely-populated areas of the county, as well as their low seismic capacities (0.18g). In
this hypothetical case, the pumping stations are retrofitted to a level typical of other facilities
(0.30g). The seismic capacities for all 9 pumping stations in each scenario (i.e., no retrofit,
retrofit 1, and retrofit 2) are shown in Table 2.

            Table 2. Pump station seismic capacities in each retrofit scenario.
              Pump station                     No retrofit(1)     Retrofit 1(2)     Retrofit 2(2)
              Mallory                               0.18g             0.18g             0.30g *
              Sheahan                               0.18g             0.18g             0.30g *
              Morton                                0.22g             0.30g *           0.22g
              McCord                                0.30g             0.30g             0.30g
              Allen                                 0.30g             0.30g             0.30g
              Lichterman                            0.30g             0.30g             0.30g
              Palmer                                0.30g             0.30g             0.30g
              Davis                                 0.30g             0.36g *           0.30g
              L.N.G.                                0.45g             0.45g             0.45g
             Notes: (1) after Hwang et al. (1998); (2) * denotes pump stations retrofitted in

    To evaluate resilience, this study made use of a previously developed lifeline loss
estimation model that had been applied to the MLGW water system. The model is described
here in overview; details can be found in Chang (2002), Chang et al. (2002), Hwang et al.
(1998), and Shinozuka (1994). The left-hand side of Figure 4 shows the structure of the loss
estimation model in thick lines. Beginning with a scenario earthquake, the model integrates
engineering and economic analysis to produce estimates of expected economic disruption
loss. 4 Monte Carlo simulation is performed to estimate physical damage, repair and
restoration, hydraulic flow and water outage, and economic loss. The simulations are
performed for weekly time intervals following the earthquake. Policy and decision variables

  “Feasibility Study Report on Seismic Retrofit of Existing Buildings at Ray Morton Pumping Station for
Memphis Light, Gas & Water Division, Memphis, Tennessee,” (1992) and “Memphis Light, Gas and Water
Division – Davis Water Pumping Station – Phase II Seismic Vulnerability Assessment of Existing Buildings,
Equipment, and Systems” (1997).
  MLGW is currently considering upgrading these two facilities.
  The model simulates direct losses to regional businesses. For related research on modeling indirect losses, or
the “ripple effects” of direct losses, see Rose and Liao (2003).
                       Submitted for publication in Earthquake Spectra

influence outcomes through two channels: pre-disaster mitigation (which reduces damage)
and post-disaster response (which reduces outage).

                    earthquake                                                  Performance

   Mitigation         Damage                                       Resilience


                                                Population         Organizational
   Response           Outage
                                                Impacted           Resilience

   Econ.-GIS         Economic                                      Economic
     Data              Loss                                        Resilience


Figure 4. Relationship between loss estimation model and resilience measures. Thick lines denote
loss estimation model. Shaded area indicates Monte Carlo simulation portion of model. Dotted lines
show linkages between loss estimation model and resilience measures.

    A key element of the model consists of the link between the engineering and economic
portions. Results of hydraulic analysis on the intact and damaged representations of the
network provide estimates of water availability at each demand node on the network.
Demand nodes are associated with economic activity through spatial overlays and other
manipulations using a geographic information system (GIS). Note that data on the spatial
distribution of economic activity (i.e., employment by place of work) are available for spatial
units known as Traffic Analysis Zones (TAZs). There are 515 TAZs in the county.
    The dotted lines in Figure 4 indicate the linkages between the loss estimation model and
the resilience measurement. As shown in the figure, even though the loss estimation model
was not designed to assess resilience, many of its intermediate results are useful for this
purpose. It should be noted that these intermediate results must be disaggregated by time (i.e.,
week after earthquake). In addition to the model results, performance standards are needed
for the resilience assessment.
   Technical resilience was estimated in a straightforward manner from results on physical
damage to the network. The number of pumping stations rendered inoperational and the
                           Submitted for publication in Earthquake Spectra

percentage of pipes suffering breaks, as well as the time to repair the network to 99 percent,
were compared to the performance standards listed in Table 1. Note that in the restoration
model, pump stations are given the highest priority, followed by large-diameter transmission
    Organizational resilience required some additional data analysis. Recall that the
performance standards pertain to the percent of the population that is without water service.
Intermediate model results indicated the water available at each demand node on the network
for each time period. This was spatially related to population data, which was available for
the 133 census tracts in the county, using GIS overlay. Note that while this does measure
population without water, it does not completely reflect organizational resilience because the
loss estimation model does not explicitly model MLGW’s emergency response capability.5 It
is also limited to analysis of water flow, and does not assess water quality or pressure, which
are important considerations for potability and fire suppression.
    Social resilience is omitted in the current study because the loss estimation model did not
produce outputs that indicate disruption to society along the lines of Table 1. In order to
estimate social resilience, a behavioral model is needed of the propensity of households to
seek temporary shelter in the event of water outage and other sources of earthquake-related
disruption. Developing such a model was beyond the scope of this study.
    Economic resilience was estimated in a straightforward manner from results on disruption
to the regional economy. The initial loss of gross regional product (GRP) and the time to
return to 99 percent of baseline levels were simply compared to the performance standards in
Table 1. Because total restoration times in the earthquake scenarios were relatively short, it
was assumed for the sake of simplicity that the without-earthquake GRP baseline would be
identical to pre-disaster levels. Note that the analysis measures economic resilience
assuming that only the water system is damaged in an earthquake. A more complete analysis
would involve modeling disruption from all sources in the earthquake, including building
damage to businesses, housing damage (which would affect households and hence labor
input), electric power outages, traffic disruptions, and so on. It would also involve modeling
how businesses respond to multiple sources of disruption.

    The loss estimation model was run for 200 Monte Carlo simulations for each of the 3
retrofit cases (including “no retrofit”) and 2 earthquake scenarios. The scenarios pertain to
events of magnitudes 6.5 and 7.0, respectively, occuring in the New Madrid Seismic Zone
with epicenter at Marked Tree, Arkansas, some 55 km northwest of downtown Memphis.
The results of the loss estimation model for these two events are an expected economic loss
of $42.4 million for the M 6.5 earthquake and $136.4 million for the M 7.0 earthquake (see
Chang, 2002). In contrast, as discussed below, resilience results are much more informative
and insightful.
   Table 3 shows results for technical resilience, where values in the table indicate the
probability of meeting performance standards in Table 1. The probabilities refer to the

 To the authors’ knowledge, there are currently no models of utility response capability that include such
elements as damage to emergency response infrastructure (e.g., vehicles, equipment, communications),
availability of personnel, existence of an effective response plan, decision-making on valving off portions of the
network, etc.
                            Submitted for publication in Earthquake Spectra

percent of simulations where outcomes meet these standards. Results range from 0 (no
resilience) to 100 (complete resilience).
    Recall that the robustness standard is that no more than 1 of the 9 pumping stations is
rendered inoperational by an earthquake.6 The rapidity standard is that after one week’s
repair activity, all of the pumping stations and 99 percent of the pipes are intact. Table 3
shows that Retrofit 1 does not materially improve system performance over the unretrofitted
case.7 However, Retrofit 2 nearly doubles system robustness in the M 6.5 earthquake, to 23
percent. In the M 7.0 event, it increases robustness from 0 to 7 percent. Thus, Retrofit 2
appears to be much more beneficial than Retrofit 1. However, additional pumping station
retrofits beyond Retrofit 2 are needed to raise the system’s technical robustness to a high
level. In the M 7.0 event, rapidity is also an issue. Pipe retrofits (e.g., replacement of some
of the cast iron pipe in the system to reduce expected damage) or response measures such as
increased mutual aid are needed in order to improve MLGW’s capacity to make the
necessary repairs rapidly in such a seismic event.

       Table 3. Technical resilience results (percent of simulations meeting performance standards).
               Resilience criteria                             M 6.5                   M 7.0
                 Retrofit strategy                           earthquake              earthquake
               Robustness and rapidity
                 No retrofit                                      12                       0
                 Retrofit strategy 1                              10                       0
                 Retrofit strategy 2                              23                       0
               Robustness only
                 No retrofit                                      12                       0
                 Retrofit strategy 1                              10                       1
                 Retrofit strategy 2                              23                       7
               Rapidity only
                 No retrofit                                      100                      0
                 Retrofit strategy 1                              100                      0
                 Retrofit strategy 2                              100                      0

    Table 4 shows results in terms of organizational resilience. The robustness standard here
is that less than 5 percent of the population lose water service in the immediate aftermath of
the earthquake. The rapidity standard is that less than 1 percent of the population is still
without water one week later. Since rapidity standards are met in all cases, results are
dominated by robustness outcomes. Again, Retrofit 2 improves robustness to a much greater
degree than Retrofit 1. In the case of the M 7.0 event, it nearly doubles robustness over the
unretrofitted case, from 21 to 39 percent. Improving technical resilience, network
redundancy, or organizational factors such as repair prioritization plans could further improve
organizational resilience.

  According to MLGW engineers, the system would be able to withstand the loss of one pumping station, but
would have difficulty coping with the loss of more than one in a disaster (Pickel, communications).
  Results for Retrofit 1 in the M6.5 event are actually slightly worse than results for the Unretrofitted case, but
this is an artifact of the Monte Carlo simulation approach.
                       Submitted for publication in Earthquake Spectra

    Table 5 shows economic resilience results. The robustness standard is that the economic
disruption entail less than 5 percent of gross regional product (GRP) immediately after the
disaster, and less than 1 percent a week later. Again, Retrofit 2 affords noticeable
improvements in robustness over both Retrofit 1 and the unretrofitted cases. Nonetheless,
overall economic resilience in Retrofit 2 is still quite low in the M 7.0 event (16 percent).
Improving technical or organizational resilience, or implementing measures such as business
continuity plans, could raise economic resilience.

          Table 4. Organizational resilience results (percent of simulations meeting
          performance standards).
           Resilience criteria                       M 6.5                M 7.0
             Retrofit strategy                     earthquake           earthquake
           Robustness and rapidity
             No retrofit                                63                  21
             Retrofit strategy 1                        66                  26
             Retrofit strategy 2                        74                  39
           Robustness only
             No retrofit                                63                  21
             Retrofit strategy 1                        66                  26
             Retrofit strategy 2                        74                  39
           Rapidity only
             No retrofit                               100                  100
             Retrofit strategy 1                       100                  100
             Retrofit strategy 2                       100                  100

     Table 5. Economic resilience results (percent of simulations meeting performance standards).
           Resilience criteria                       M 6.5                M 7.0
             Retrofit strategy                     earthquake           earthquake
           Robustness and rapidity
             No retrofit                                68                  9
             Retrofit strategy 1                        72                   6
             Retrofit strategy 2                        79                  16
           Robustness only
             No retrofit                                68                  25
             Retrofit strategy 1                        72                  28
             Retrofit strategy 2                        79                  46
           Rapidity only
             No retrofit                               100                  34
             Retrofit strategy 1                       100                  26
             Retrofit strategy 2                       100                  30

    In sum, the example application has shown how the multi-dimensional concept of
resilience can be operationalized and quantified. It has demonstrated how mitigation
activities can be assessed and compared on the basis of their resilience improvements. In this
                      Submitted for publication in Earthquake Spectra

example, the selection of the Mallory and Sheahan pumping stations for retrofit in strategy 2
is shown to be preferable to the Morton and Davis pumping stations in strategy 1.
Nonetheless, additional measures beyond Retrofit Strategy 2 are needed to raise system and
community resilience to a high level. These might involve additional pump station retrofits,
pipe replacements, or such measures as improvements to emergency response planning.

                          AREAS FOR FURTHER RESEARCH
    One aim of this study was to identify further research that is needed in order to fully
implement the concept of community resilience for purposes of supporting disaster reduction
decision-making. The example application showed that while current loss estimation models
provide a useful starting point for quantifying resilience, important data gaps and modeling
issues remain. Three areas for further study merit particular attention.
     First, research is needed into appropriate performance standards – technical,
organizational, social, and economic – which are central to the concept and measurement of
resilience. Such research should identify stakeholder groups and their representatives (e.g.,
utility managers, elected officials). Appropriate performance standards could be identified
through such methods as interviews, surveys, and focus groups. Standards of robustness and
rapidity should be defined in ways that can best support mitigation decision-making.
     Second, a new class of models is needed that integrates engineering with organizational
response. Such models should focus on the delivery of infrastructure service. For a water
system, this would mean the delivery of potable water supply to end users. In contrast,
current loss estimation models estimate the available water supply at network nodes given
conditions of water leakage. They are unable to address such questions as: How can the
lifeline agency manage flows through such responses as using shutoff valves? How can it
prioritize restoration to best meet post-disaster water demands? What is the quality of the
water from the damaged network? New data on organizational decision-making will be
required to develop such models.
    Third, a new class of models is needed that integrates engineering with societal impacts.
Such models should focus on community-level disruptions (e.g., households being displaced
from their homes). This would entail modeling how individuals and households behave in
the face of disaster, and how this behavior differs across groups. These models should
consider the effects of community-level mitigation, preparedness, and response activities
(e.g., government deployment of water trucks to neighborhoods losing potable water).

    As an objective, improving the disaster resilience of communities calls for a broader
conceptual framework than simply reducing monetary losses. A resilience framework must
address the capacity of social units (e.g., individuals, organizations, and communities) to
prepare for and respond to disasters. A key challenge is how to measure resilience and
resilience improvements. This paper has proposed and demonstrated a quantitative
framework for measuring resilience. While implemented for the case of earthquake risk to
the water delivery system in Memphis, the methodology is broadly applicable to other critical
infrastructure systems, communities, and types of disasters. However, new research is
needed in several key areas in order to fully operationalize and realize the advantages of the
concept of resilience.
                        Submitted for publication in Earthquake Spectra

     The Memphis case study showed that resilience assessment goes beyond traditional loss
estimation and can better inform loss-reduction decision-making in some important ways.
First, it quantifies not simply the losses that might be expected in potential future earthquakes,
but how these losses relate to standards of acceptable performance. This addresses the
question of how much – and how much more – disaster mitigation and preparedness is
needed. It helps to determine whether community resilience is increasing over time. It also
encourages community-based discussion of appropriate seismic performance standards for
critical infrastructure systems.
    Second, resilience assessment focuses attention not only on the magnitude of losses, but
also on the speed of recovery. This places both pre-disaster mitigation and post-disaster
response within a common framework by which they can be evaluated and compared.
Comparative studies can thus be conducted to assess the effectiveness of various loss
reduction measures, ranging from structural retrofits to emergency response plans.
    Third, the approach addresses in a systematic fashion the multiple, interrelated
dimensions of resilience (technical, organizational, social, and economic). This clarifies the
role of different types of interventions in improving community resilience. It helps in
understanding why some systems are more disaster-resilient than others. Further, it
emphasizes that the seismic resilience of engineered systems (e.g., the water delivery
network) is not simply an engineering problem, but has important organizational aspects and
contributes significantly to the disaster resilience of the entire community.

   This work was supported in whole by the Earthquake Engineering Research Centers
Program of the National Science Foundation under Award Number ECC-9701471 to the
Multidisciplinary Center for Earthquake Engineering Research. The authors thank Ron
Eguchi (Imagecat, Inc.) and Chris Chamberlin (University of Washington) for their valuable
contributions. Information and cooperation provided by MLGW is gratefully acknowledged.

Applied Technology Council, 1991, Seismic Vulnerability and Impact of Disruption of Lifelines in the
   Conterminous United States, ATC-25, Redwood City, California, 440 pp.
Bruneau, M., Chang, S. E., Eguchi, R. T., Lee, G. C., O’Rourke, T. D., Reinhorn, A. M., Shinozuka,
   M., Tierney, K., Wallace, W. A., and von Winterfeldt, D., A Framework to Quantitatively Assess
   and Enhance the Seismic Resilience of Communities, Earthquake Spectra, forthcoming.
Chang, S. E., 2002, “Probabilistic Assessment of Economic Impacts of Earthquakes: Memphis Water
   System Case Study,” Proc. 7th National Conference on Earthquake Engineering, Boston.
Chang, S. E., Svekla, W. D., and Shinozuka, M., 2002, “Linking Infrastructure and Urban Economy:
   Simulation of Water Disruption Impacts in Earthquakes,” Environment and Planning B, Vol. 29,
   No. 2, pp.281-301.
Cho, S., Gordon, P., Moore, J. E. II, Richardson, H.W., Shinozuka, M., and Chang, S. E., 2001,
   “Integrating Transportation Network and Regional Economic Models to Estimate the Costs of a
   Large Urban Earthquake,” Journal of Regional Science, Vol. 41, No. 1, pp. 39-65.
Federal Emergency Management Agency (FEMA), 2000, “Planning for a Sustainable Future: The
   Link Between Hazard Mitigation and Livability,” Washington, D.C.
Hwang, H. H. M., Lin, H., and Shinozuka, M., 1998, “Seismic Performance Assessment of Water
   Delivery Systems,” Journal of Infrastructure Systems, Vol. 4, No. 3, pp. 118-125.
                        Submitted for publication in Earthquake Spectra

Johnston, A. C., and Nava, S. J., 1985, “Recurrence Rates and Probability Estimates for the New
    Madrid Seismiz Zone,” Journal of Geophysical Research, Vol. 90, No. B7.
Kim, T. J., Ham, H., and Boyce, D., 2002, “Economic Impacts of Transportation Network Changes:
    Implementation of a Combiend Transportation Network and Input-Output Model,” Papers in
    Regional Science, Vol. 82, No. 2, pp. 223-246.
Mileti, D., 1999, Disasters by Design: A Reassessment of Natural Hazards in the United States,
    Joseph Henry Press, Washington, D.C., 351 pp.
Nigg, J., Riad, J. K., Wachtendorf, T., and Tierney, K., 2000, “Disaster Resistant Communities
    Initiative: Evaluation of the Pilot Phase, Year 2,” Disaster Research Center, University of
    Delaware, Newark, Delaware.
Rose, A. and Liao, S.-Y., 2003, “Understanding Sources of Economic Resiliency to Hazards:
    Modeling the Behavior of Lifeline Service Customers,” in Research Progress and
    Accomplishments 2001-2003, Multidisciplinary Center for Earthquake Engineering Research,
    Buffalo, NY, pp. 149-160.
Rose, A., Benavides, J., Chang, S. E., Szczesniak, P., and Lim, D., 1997, “The Regional Economic
    Impact of an Earthquake: Direct and Indirect Effects of Electricity Lifeline Disruptions,” Journal
    of Regional Science, Vol. 37, No. 3, pp. 437-458.
Shinozuka, M, 1994, “GIS Applications in Lifeline Earthquake Engineering,” in Proceedings of the
    Second China-Japan-US Trilateral Symposium on Lifeline Earthquake Engineering, Xi’an, China,
    pp. 223-230.

To top