by S.E. Chang, H.A. Seligson, and R.T. Eguchi
NCEER's multi-year, multidisciplinary investigation of the seismic performance of Memphis Light, Gas and Water (MLGW) Division's lifeline systems culminates in an assessment of economic impact. This article presents results on the estimated direct and indirect economic losses that would follow from disruption to electric power, water, and natural gas service in Memphis/Shelby County, Tennessee in a scenario earthquake occurring in the New Madrid Seismic Zone. It focuses on methodologies developed to evaluate direct losses, including lifeline repair costs, revenue losses to the utility itself, and direct business interruption losses suffered by the utility customers. The evaluation builds upon results from related NCEER studies, including work by H. Hwang (hazard assessment), M. Shinozuka and S. Tanaka (systems performance), K. Tierney (business impact assessment), S. French (GIS), and A. Rose (indirect impact modeling). Further detail will be presented in a forthcoming NCEER technical report. Questions and comments should be directed to Stephanie Chang, EQE International, at (714) 833-3303; or email: email@example.com.
Recent disasters such as the Northridge earthquake and the Great Hanshin (Kobe) earthquake vividly demonstrated the seismic vulnerability of lifeline systems and the severe socioeconomic impact of lifeline service disruption. Previous NCEER studies have shown that a magnitude 7.5 New Madrid Seismic Zone (NMSZ) seismic event with an epicenter at Marked Tree, Arkansas would cause major disruption to many critical lifeline systems in Memphis/Shelby County, Tennessee (Shinozuka et al., 1992, 1994). The objectives of the current study were (1) to develop a methodological approach for estimating economic losses from damage to multiple urban lifeline systems in earthquake disasters, and (2) to demonstrate its application to Memphis/Shelby County and Memphis Light, Gas and Water's (MLGW) lifeline systems in the Marked Tree scenario event.
Figure 1 outlines the methodological framework adopted in this study. Beginning with analysis of the seismic hazard, information on utility lifeline facilities and applicable vulnerability models can be combined to provide a profile of utility risk and expected facility damage in the scenario earthquake. This enables estimation of the repair costs that would be incurred as a result of the physical damage. Next, expected lifeline service outage and restoration are evaluated based on the damage patterns. Together with information on utility customers and their usage of lifeline services, this provides a means for estimating the revenue losses that would be incurred by the utility company during the service interruption period. To model the direct economic losses or business interruption impacts, information is needed on the locations of businesses throughout the affected region and on their economic resiliency to disruption of various lifeline services. This model enables the estimation of the economic losses directly caused by lifeline service disruption. The final step consists of applying an Input-Output (I-O) economic modeling approach to evaluate the additional "upstream" and "downstream" or indirect economic impacts caused by this direct business interruption. Results from studies by several NCEER investigators on the Memphis area provided many of the elements shown in figure 1 (see also Rose et al., eds., forthcoming); a primary challenge of this study consisted of synthesizing these results in order to implement the overall framework.
The study area is shown in figure 2, together with the expected ground shaking pattern by census tract for the scenario M 7.5 event. Ground shaking is evaluated in terms of Modified Mercalli Intensity (MMI) and is based on work by H. Hwang and colleagues at the Center for Earthquake Research and Information at the University of Memphis. Figure 3 depicts the pattern of employment in Shelby County and indicates the geographic distribution of economic activity in the study area. The geographic information system (GIS) overlay of information on seismic hazard, lifeline facilities, utility customers, and economic activity on a census tract basis represented a central means for integrating the engineering and social science layers of the problem.
The scope of the analysis includes the impact of disruption to water, electric power, and natural gas systems independently, as well as simultaneously. However, while in this sense it does evaluate economic interaction, it does not consider functional interaction between the systems (e.g., between electric power and water systems). Furthermore, while the entire water delivery system is included in the analysis, for consistency with previous NCEER studies, the scope is limited to transmission substations and to distribution system pipelines in the cases of electric power and natural gas systems, respectively.
The following section provides a summary of the economic impact results of the research and discusses their significance. Subsequent sections describe the methodologies employed, data sources, and more detailed results of the analysis.
The results of the study led to several conclusions regarding the Memphis scenario event specifically and insights into the expected economic impact of lifeline system damage more generally:
Economic losses deriving from lifeline damage alone in an earthquake can be substantial.
In the case of multiple lifeline disruption, direct plus indirect economic loss (i.e., loss to gross regional product (GRP)) represents roughly the same magnitude of loss as repair costs.
The relative significance of repair costs, revenue loss, and economic impact differs substantially between lifelines. To a large extent, this results from the expectation that certain lifelines such as electric power would be restored much more quickly than others following a disaster.
Restoration patterns were found to critically influence the ultimate economic impact resulting from lifeline damage. While repair costs depend upon damage patterns, both revenue losses and business interruption losses are linked directly to the duration and extent of service outage. Restoration estimates for the MLGW lifelines were based on damage and initial outage estimates produced by M. Shinozuka et al., at Princeton University and on restoration models based on past California earthquakes with some adjustments for central U.S. conditions. Figure 4 shows the inferred restoration curves for electric power, water, and natural gas service for the scenario earthquake. Complete restoration times are estimated at approximately two weeks for electric power and water and four weeks for natural gas. It is reassuring that both the shapes and relative positions of the three lifeline curves are consistent with actual experience in the Northridge and Kobe earthquakes. For example, electric power is restored most quickly, followed by water and then natural gas.
Direct economic loss results consistent with these outage and restoration patterns for each of the lifelines individually are shown in table 1. In this analysis, effects of simultaneous disruption to several lifelines are not considered. Direct economic loss is estimated in GRP or final demand terms, rather than gross output terms, to avoid double-counting. The sum of repair costs, revenue loss, and direct economic loss ranges from $71.5 million for water to $450.9 million (over six times as large) for electricity, with natural gas in the middle range at $344.7 million.
Not only do the overall impacts differ substantially between the three lifelines, but so does the relative significance of different types of direct loss. Table 1 shows that for each of the utilities, revenue losses are relatively minor. In the case of natural gas, repair costs ($1.5 million) and revenue losses ($5.9 million) are eclipsed by direct economic or business interruption losses ($337.3 million), which represent 98% of the sum. Similarly, direct economic losses constitute 90% of the sum of direct impacts for water. However, in the case of electric power, repair costs ($401.1 million) dominate, while business interruption losses represent only 11% of the sum. That damage to electric power facilities entails such high repair costs is consistent with observation in the Northridge earthquake, for example, where the estimated $137 million damage to electric power facilities was much higher than the $49 million for water and $60 million for natural gas facilities (Eguchi 1995). The contrasts in the significance of direct economic or business interruption loss are closely related to the outage and restoration patterns summarized in figure 4, where restoration was relatively rapid for electricity and relatively slow for natural gas.
To the extent possible, it is useful to make order-of-magnitude comparisons between results in table 1 and estimated losses in the benchmark ATC-25 (1991) study on "Seismic Vulnerability and Impact of Disruption of Lifelines in the Conterminous United States." In the scenario NMSZ magntiude 7.0 earthquake used in that study, the ratio of direct economic loss to repair costs (referred to in ATC-25 as "indirect" and "direct" losses, respectively) for electric power transmission systems in the entire impacted area is about 3.0. Results in table 1 indicate that in the current study, the comparable ratio is about 0.1. This disparity probably derives in large part from the difference in expected restoration times - two weeks in the current study and about 14 weeks (for Tennessee) in ATC-25. In view of past earthquake experience, the shorter time period is expected to better reflect the time to restore electricity service. Service restoration time may be significantly less than the time required to complete repairs because in a disaster, emergency measures can be used to temporarily restore service while repairs are under way. In the case of water, the ratios, at 2.0 in ATC-25 and 9.5 from table 1, are more comparable. It was not possible to make the comparison for natural gas because distribution system losses were not evaluated in ATC-25.
In an actual disaster, the direct economic loss caused by disruption of all three lifelines is likely to be somewhat greater than for each of the lifelines individually and somewhat less than their sum. For example, a business that is forced to temporarily shut down because of electric power loss will not suffer any additional losses caused by disruption of water service, whereas one that was only forced to scale back production might. Table 2 shows the results for multiple lifeline disruption. Direct economic losses are estimated at $350 million in GRP terms. This leads to repercussions in the economy or indirect loss that amounts to an additional $70 million. Including repair costs, total direct and indirect loss suffered in Shelby County as a result of damage to electric power, water, and natural gas lifelines is estimated at $829 million. This represents about 3% of GRP for Shelby County in the study year, 1991. Note that revenue losses are excluded in order to avoid double-counting. Direct and indirect economic impacts comprise 42.2% and 8.4% of the total, respectively. Thus in this case, direct and indirect economic loss represent roughly the same magnitude of dollar loss as repair costs.
Damage and repair costs were evaluated using results from previous NCEER studies, existing methodologies, and facility exposure data from MLGW. Probabilistic damage estimates were provided by M. Shinozuka and S. Tanaka at Princeton University for the number of pipe breaks or repairs in the water delivery system. Combining this data with costs per repair, determined from expert consultation, yielded total repair costs. In the case of electric power, damage and repair cost estimates were made based on ATC-25 methodology and typical substation replacement costs based on data from the Los Angeles Department of Water and Power (LADWP). For natural gas, pipeline breaks were estimated using existing models (Eguchi 1991) based on observation in historic earthquakes. Costs per repair were determined from expert consultation.
Service outage estimates for the scenario earthquake were generally available from studies by M. Shinozuka et al. In the case of the water delivery system, their results on probabilistic estimates of initial outage by census tract in Shelby County were utilized directly. For electric power, similar results were provided by electric power service area (EPSA) and used to infer outage on a census tract level. Natural gas service was assumed to be disrupted throughout the county immediately after the disaster, a result that is consistent with the expected intensity patterns (MMI VIII and VIII ? and experience in previous earthquakes.
Restoration was modeled based on MLGW lifeline system and service area information and observations and data from previous earthquakes. For water, estimates of the density of pipe breaks were used to approximate the amount of time that would be required for work crews to restore service in a given census tract. Data from the San Fernando earthquake described in Seligson et al. (1991) provided the means to calibrate the model. Electric power restoration was estimated by first developing an overall restoration curve calibrated to the Northridge earthquake and scaled to the scenario event. Expected restoration times for individual EPSAs and census tracts were derived based on initial outage and information on customer distribution. For natural gas, total restoration time was estimated from ATC-25 models. Four recovery "zones" were then delineated based on knowledge of the different gas pressure systems in Shelby County and an assumed restoration sequence based on expected damage concentrations.
To estimate expected revenue losses to the utility company, lifeline service disruption results were combined with information on MLGW's customer base and average daily revenue per customer by residential, commercial, or industrial class. The geographic location of MLGW customers throughout the county's census tracts was estimated using census population data for residential customers. The number of nonresidential customers by census tract was approximated using unpublished industry employment data that was originally available on a traffic analysis zone basis. Revenue loss results by user type are shown in table 3. Figures for natural gas pertain to an average of winter and summer seasons.
Direct economic loss models were developed based upon existing methodologies described in ATC-25; however, several important refinements were made that take advantage of results from other NCEER studies focusing on Shelby County (see Rose et al., eds., forthcoming). These refinements include (1) utilization of empirical data on business dependency on lifelines in Memphis, (2) consideration of duration effects associated with lifeline outage, (3) evaluation at a geographically disaggregate (i.e., census tract) level, (4) consideration of multiple lifeline disruption, and (5) evaluation of indirect economic losses.
The ATC-25 methodology assumes that the first 5% of lifeline service disruption can be absorbed without economic loss to the user and that subsequent losses would be incurred proportionally to the extent of disruption up to some maximum level of impact. The maximum loss associated with complete lifeline outage ("importance" factor) or its converse ("resiliency" factor) varies according to a particular industry's dependency on the specific lifeline. Rather than utilize the expert-based factors in ATC-25, new values were estimated based on results from a study of business vulnerability to lifeline disruption in Shelby County conducted by K. Tierney at the Disaster Research Center (DRC) of the University of Delaware. In addition to reflecting business operators' own assessments, it is expected that this approach also captures differences in utility usage between the central U.S. and California (for which the factors in ATC-25 were developed). Further, time-dependent factors were developed to account for the deterioration of resiliency with the duration of outage.
Table 4 provides a comparison of ATC-25 and Shelby County resiliency factors by major industry and lifeline. For comparability, the former are averaged over sub-industries and the latter over the first month of outage. Shelby County factors take into account businesses that do not use natural gas. Resiliency is generally highest to gas outage and lowest to electricity disruption. Results show that while the factors are generally consistent for electric power, ATC-25 overestimates business resiliency to water, and in many industries, natural gas disruption, versus the Shelby County results. For example, for the finance, insurance and real estate (FIRE) industry, the electricity resiliency factors are similar (.10 and .07) and indicate low resiliency to outage. However, ATC-25 factors for water and gas are both .80, whereas the Shelby County results indicate much lower resiliency (.21 and .51, respectively). With all else equal, this suggests that previous models of business impact may substantially underestimate the direct economic losses from lifeline disruption.
Direct economic losses were evaluated on a census tract level by industry, lifeline, and day following the disaster. For simultaneous disruption of electricity, water and gas, it was assumed that a particular industry (for the specific day and census tract) would suffer direct economic loss equal to the greatest of the losses from the three lifelines individually. This assumption may lead to underestimates of the economic loss from multiple lifeline disruption.
The indirect impact caused by the direct economic loss was modeled using a methodology based on Input-Output (I-O) analysis that takes into account inter-industry linkages in the regional economy. This application of I-O analysis was developed by A. Rose et al. at Pennsylvania State University and makes two major adjustments to the lifeline disruption problem: first, model coefficients are adjusted to reflect business resiliency to lifeline outage; and second, indirect impact is modeled as those losses over and above direct economic loss that derive from the repercussions of "bottleneck" sector constraints (Rose et al., forthcoming). In the multiple lifeline disruption case, the manufacturing, services, and FIRE industries constituted the "bottlenecks" in the first two weeks, the third week, and the fourth week of disruption, respectively.
Table 5 shows the direct and indirect economic loss by major industry for the multiple lifeline disruption case. Of the total ($419.6 million), 83% consists of direct economic impact. The results demonstrate that the significance of indirect losses varies substantially by sector. At one extreme, virtually all (97%) of the economic losses suffered by the "bottleneck" industries (manufacturing, FIRE and services) are direct. As expected, indirect losses are much more sizable for the non-bottleneck industries.
The cooperation of MLGW, LADWP, and P. McDonough of Mountain Fuel Supply Co., as well as efforts by the NCEER investigators mentioned above, are gratefully acknowledged.
Applied Technology Council, (1991), Seismic Vulnerability and Impact of Disruption of Lifelines in the Conterminous United States, ATC-25, Redwood City, California: Applied Technology Council.
Eguchi, R.T., (1991), "Seismic Hazard Input for Lifeline Systems," Structural Safety, Vol. 10, pp. 193-198.
Eguchi, R.T., (1995), "Mitigating Risks to Infrastructure Systems Through Natural Hazard Reduction and Design," pp. 109-118 in C.-K. Choi and J. Penzien, eds., Public Infrastructure Systems Research, Taejon, Korea: Techno-Press.
Rose, A.Z., Shinozuka, M. and Eguchi, R.T., eds., forthcoming, "Engineering and Socioeconomic Analysis of a New Madrid Earthquake: Impacts of Electricity Lifeline Disruptions," NCEER monograph.
Rose, A.Z., Benavides, J., Chang, S.E., Szczesniak, P. and Lim, D., forthcoming, "The Regional Economic Impact of an Earthquake: Direct and Indirect Effects of Electricity Lifeline Disruptions."
Seligson, H.A., Eguchi, R.T., Lund, L. and Taylor, C.E., (1991), "Survey of 15 Utility Agencies Serving the Areas Affected by the 1971 San Fernando and the 1987 Whittier Narrows Earthquakes," report prepared for the National Science Foundation.
Shinozuka, M., Hwang, H.H.M. and Tanaka, S., (1992), "GIS-Based Assessment of the Seismic Performance of a Water Delivery System," Proceedings of the Fifth U.S.-Japan Workshop on Earthquake Disaster Prevention for Lifeline Systems, Public Works Research Institute (PWRI) Technical Memorandum No. 3198, pp. 233-249.
Shinozuka, M., Tanaka, S. and H. Koiwa, H., (1994), "Interaction of Lifeline Systems under Earthquake Conditions," Proceedings of the Second China-U.S.-Japan Trilateral Symposium on Lifeline Earthquake Engineering, pp. 43-52.
Some of the material reported herein is based upon work supported in
whole or in part by the National Science Foundation, the State of New York, the U.S.
Department of Transportation, the Federal Highway Administration, the Federal Emergency
Management Agency and other sponsors. Any opinions, findings, and conclusions or
recommendations expressed in this publication are those of the author(s) and do not
necessarily reflect the views of NCEER or its sponsors.
Click here to return to the Table of Contents
| Home | NCEER | Information Assistance | Publications | Databases, Software |
| Events | New! | Links | Data Resources |