SEISMIC ZONATION FOR LIFELINES AND UTILITIES T.D. O’Rourke1 and S.-S. Jeon2 ABSTRACT This paper focuses on the four principal uses of seismic zonation for lifelines and utilities: 1) hazard delineation, 2) physical loss estimation, 3) assessment of economic and social consequences, and 4) planning for emergency response and recovery. Emphasis is given to geographic information systems (GIS) and their application to pipeline networks in evaluating the spatial characteristics of earthquake effects. The paper examines the GIS databases for water supply performance obtained for the 1994 Northridge and 1995 Kobe earthquakes. Relationships among buried lifeline damage and various seismic parameters are examined, and the parameters that are statistically most significant are identified. Using GIS data from the Northridge earthquake, the relationships among pipeline repair rate, type of pipe, diameter, and various seismic parameters are assessed. The effects of permanent ground deformation (PGD) on buried lifelines are discussed. Air photo measurements of PGD and the spatial characteristics of ground deformation revealed by these measurements are described. Mapping procedures for PGD hazards, including probabilistic mapping techniques, are reviewed. Recent studies of economic losses from earthquake damage to lifelines in Shelby County, TN are used to explore the direct and indirect economic losses of lifeline systems. Introduction Seismic zonation has been defined in broad terms as the geographic delineation of variations in the potential for various earthquake hazards (Cluff and Pecker, 1995). From a geotechnical perspective, Finn (1991) refers to seismic microzonation as a procedure for improving estimates of seismic hazard for design by taking the effects of local site conditions into account. By virtually any definition, seismic zonation involves treatment of the spatial variability of seismic excitation, geotechnical conditions, and the characteristics of the built environment and communities affected by earthquakes. Lifelines are often grouped into six principal types of systems (in alphabetical order): electric power, gas and liquid fuels, telecommunications, transportation, wastewater facilities, and water supply. These systems share three common characteristics: geographical dispersion, interconnectivity, and diversity (O’Rourke, 1998). Lifelines are geographically dispersed over broad areas, and are exposed to a wide range of seismic and geotechnical hazards, community 1, 2 – Thomas R. Briggs Professor of Engineering and Graduate Research Assistant, respectively, School of Civil & Environmental Engineering, Cornell University, Ithaca, NY 14853-3501 uses, and interactions with other sectors of the built environment. They are interconnected and interdependent. Each lifeline system is composed of many interconnected facilities and is influenced by the performance of other lifeline systems. Lifeline performance is related to the characteristics of many diverse components; most lifeline networks have been built over many years and function with parts produced according to different construction and/or manufacturing techniques, standards, and design procedures. Given the characteristics of lifelines and the goals of seismic zonation, it is natural to combine the two and apply zonation for more effective engineering and management of critical transportation and utility systems. There has been substantial effort in recent years in applying seismic zonation to lifeline networks with many examples available in the proceedings of previous seismic zonation conferences (e.g., Earthquake Engineering Research Institute, 1991 and 1995). This paper focuses on the four principal uses of seismic zonation for lifelines and utilities: 1) earthquake hazard delineation, 2) physical loss estimation, 3) assessment of economic and social consequences, and 4) planning for emergency response and recovery. Pipeline systems, in particular, are used to illustrate current trends in seismic zonation and lifeline engineering. Pipeline networks are an essential part of water, wastewater, gas, and liquid fuel conveyance systems, and are used extensively in telecommunication and electric power networks. The paper begins with an examination of GIS, followed by a discussion of buried pipeline response to earthquakes. The principal uses of seismic zonation are covered by describing recent investigations that are relevant for each use and summarizing the important lessons learned for future applications. Geographic Information Systems GIS has been defined as “a set of tools for the input, storage and retrieval, manipulation and analysis, and output of spatial data” (Marble, et al., 1984). It is also a problem solving process, organized with respect to spatial and attribute data, that can be integrated with various geographic technologies (e.g., remote sensing, global positioning systems, computer-aided design, etc.) for support of decision making (Malezewski, 1999). GIS can be viewed, therefore, as a system for integrating data from various disciplines and sources to guide planning and management about a specific geographic area or site. It is advantageous to think of GIS in terms of the different types of databases used to characterize civil infrastructure systems (CIS). As proposed by Uy and O’Rourke (2000), CIS involve both the physical environment and the social and economic characteristics of communities that are located within the physical environment. The different aspects of CIS can be divided into the two broad categories of societal and physical environment, which are shown in Fig. 1. Data about the societal environment is divided into demographic and economic information. Demographic information includes data on political boundaries, population, housing, and human hazards/security/vulnerability. Economic information includes personal and CIVIL INFRASTRUCTURE SYSTEMS The Societal Environment Demographic Economic The Physical Environment Natural Environment Built Environment Figure 1. Civil Infrastructure Systems governmental income, personal and governmental spending, land and housing rent/value, trade (manufacturing, retail, wholesale), and other types of industries. The physical environment consists of the natural and built environments. The natural environment includes land use and land cover, topography, geology and seismology, water and air quality, sources of pollutants, wildlife/plant habitat, and natural hazards. Key elements of the built environment are public and private buildings and lifeline systems. The rapid development of computer mapping and visualization tools, embodied in GIS, provides a powerful basis for evaluating earthquakes effects on lifelines, as well as the consequences of these interactions on communities. It is not surprising, therefore, that GIS has become an engine for driving new methodologies and decision support systems focused on the spatial variation of potential earthquake effects. GIS, for example, is the backbone of the National Loss Estimation Methodology sponsored by FEMA and implemented through HAZUS computer software (National Institute for Building Science, 1997; Whitman, et al., 1997). GIS also has been harnessed to explore the engineering and socioeconomic impacts of earthquakes through multidisciplinary studies of the losses incurred by disruptions of water supply and electric power systems (Chang, et al., 1996; Shinozuka, et al., 1998). GIS Lifeline Earthquake Databases Advances in seismic zonation for lifelines and utilities have been influenced in a profound way by records of lifeline performance acquired after recent earthquakes, most notably the 1989 Loma Prieta, 1994 Northridge, and 1995 Kobe earthquakes. Data from the Northridge and Kobe earthquakes have been compiled in GISs of unprecedented size and complexity that allow for a detailed examination of the spatial relationships among lifeline damage, permanent and transient ground deformation, and the surface, subsurface, and groundwater conditions. A comprehensive investigation of water supply pipeline damage after the 1995 Kobe earthquake undertaken by the Japan Water Works Association (1996) has been described by Shirozu, et al. (1996). The study was concentrated on seven water distribution systems, totaling over 12,000 km of pipelines and 2,885 earthquake related repairs. Data on location, mode of damage, material, diameter, and year of installation were collected and entered into a GIS database with approximately 13,000 photos. Digitized maps of the water distribution pipelines were made part of the GIS. Data on surficial geology, degree of liquefaction, seismic intensity, and vectors of PGD determined by air photo measurements (Hamada and Wakamatsu, 1996) were also included. Statistics were compiled for each of six types of pipe composition, summarizing the number of repairs, length of affected pipeline, repair rate (repairs divided by affected pipe length), and damage mode observed in either the pipe body or joints. The highest repair rates were incurred by steel pipelines with threaded couplings, asbestos cement (AC) pipelines, and cast iron (CI) pipelines. The overall repair rates for ductile iron (DI) and welded steel pipelines were approximately one third of that for CI mains, and the predominant mode of failure for DI pipelines was pullout at mechanical joints. DI pipelines equipped with earthquake-resistant restrained joints were not damaged, even in areas of liquefaction-induced PGD. Pipeline repair rates were inversely proportional to diameter and increased in direct proportion to peak ground acceleration. Repair rates in areas of liquefaction-induced PGD were 6 to 10 times higher than repair rates in areas of comparable peak ground acceleration with no PGD effects. The GIS database for the Kobe earthquake provides an exceptionally detailed and comprehensive assessment of earthquake performance in a large, geographically dispersed system. GIS for the 1994 Northridge earthquake provides a similar database for U.S. water distribution performance, and is discussed in detail in forthcoming sections of this paper. Pipeline Response to Earthquakes Earthquakes cause transient ground deformation (TGD) and permanent ground deformation (PGD), both of which affect underground pipelines. TGD is the dynamic response of the ground, and PGD is the irrecoverable movement that persists after shaking has stopped. PGD often involves large displacements, such as those associated with surface fault rupture and landslides. TGD can cause soil cracks and fissures triggered by pulses of strong motion that develop localized shear and tensile strains exceeding the strength of surficial soils. The principal causes of PGD have been summarized and discussed by O’Rourke (1998). They are faulting, tectonic uplift and subsidence, and liquefaction, landslides, and densification of loose granular deposits. Liquefaction is the transformation of saturated cohesionless soil into a liquefied state or condition of substantially reduced shear strength (Youd, 1973). Liquefactioninduced pipeline deformation can be caused by lateral spread, flow failure, local subsidence, post-liquefaction consolidation, buoyancy effects, and loss of bearing (Youd, 1973; O’Rourke, 1998). It is widely accepted that the most serious pipeline damage during earthquakes is caused by PGD. Furthermore, it is well recognized that liquefaction-induced PGD, especially lateral spread, is one of the most pervasive causes of earthquake-induced lifeline damage (Hamada and O’Rourke, 1992; O’Rourke and Hamada, 1992). Ground displacement patterns associated with earthquakes depend on PGD source, soil type, depth of ground water, slope, earthquake intensity at a given site, and duration of strong ground shaking (O’Rourke, 1998). It is not possible to model with accuracy the soil displacement patterns at all potentially vulnerable locations. Neverthless, it is possible to set upper bound estimates of deformation effects on buried lifelines by simplifying spatially distributed PGD as movement concentrated along planes of soil failure. Various modes of pipeline distortion caused by PGD are illustrated in Fig. 2. Pipelines crossing a fault plane subjected to oblique slip are shown in Fig. 2a. Reverse and normal faults promote compression and tension, respectively. Strike slip may induce compression or tension, depending on the angle of intersection between the pipeline and fault. Fig. 2b shows a pipeline crossing a lateral spread or landslide perpendicular to the general direction of soil movement. In this orientation, the pipeline is subject to bending strains and extension. As shown in Fig. 2c, the pipeline will undergo bending and either tension or compression at the margins of the slide when the crossing occurs at an oblique angle. Fig. 2d shows a pipeline oriented parallel to the general direction of soil displacement. At the head of the zone of soil movement, the displacements resemble normal faulting; under these conditions, the pipeline will be subjected to both bending and tensile strains. At the toe of the slide, the displaced soil produces compressive strains in the pipeline. Strike slip Fault plane Pipeline subject mainly to bending ss sv Legend sd - Dip slip ss - Strike slip sv - Vertical displacement sh - Thrust displacement sd sh Dip slip b) Perpendicular Crossing a) Three-Dimensional View Pipeline subject to compression and bending Pipeline subject to tension and bending c) Oblique Crossing Pipeline subject to tension and bending Pipeline subject to compression and bending d) Parallel Crossing Figure 2. Principal Modes of Soil-Pipeline Interaction Triggered by Earthquake-Induced PGD (O’Rourke, 1998) TGD generally induces much smaller levels of pipeline strain and deformation than PGD. Neverthless, TGD covers a broader area than PGD. Pipeline systems involve many different components, some of which are invariably weakened by corrosion and/or residual stress concentrations. It follows, therefore, that the effects of widespread TGD, rather than localized PGD, are more likely to overlap with weaker components. At least four principal types of TGD have been identified: 1) traveling P and S waves, 2) surface wave generation in large sedimentary basins, 3) vibration of relatively narrow soil-filled valleys, and 4) ground oscillation (O’Rourke, 1998). This latter phenomenon was first described by Youd and Keefer (1994). It involves transient lateral shear strains and oscillatory horizontal movement of liquefiable soil relative to adjacent and underlying competent ground. Ground oscillation has been shown to be the principal cause of extensive pipeline damage in the Marina of San Francisco during the 1989 Loma Prieta earthquake (Pease and O’Rourke, 1997). One of the principal challenges for seismic zonation and lifeline engineering is the identification of landforms, subsurface conditions, and wave propagation scenarios that can trigger PGD and various levels of TGD. During post-earthquake investigations, it is often difficult to collect sufficient data over broad geographic areas to distinguish all PGD zones from zones predominately influenced by TGD. As a planning and predictive exercise, the identification of PGD vs. TGD zones is especially problematic because of the spatial variability of soil and groundwater and the lack of specific information at most locations about the subsurface nature of the ground. A practical way to learn about the spatial variability of TGD and PGD effects is to study the relationships among pipeline damage, strong motion, and locations of observed permanent deformation for large, geographically dispersed systems. Recently, severe earthquakes have occurred in urbanized areas with large pipeline systems and relatively dense networks of strong motion instruments. In the next two sections, data on pipeline system behavior during the 1994 Northridge earthquake are examined and analyzed to illustrate regional patterns of TGD and PGD, and their corresponding influence on water distribution system performance. Earthquake Hazard Delineation This section is divided into two subsections dealing with liquefaction hazards and pipeline damage patterns. Liquefaction is a principal cause of PGD, and therefore potentially disruptive for underground facilities. Mapping and characterizing liquefaction hazards allow for the prediction of lifeline damage as well as the distribution of damage at liquefaction areas throughout the system. Conversely, the pattern of post-earthquake damage allows for the identification of seismic hazards at the locations of concentrated pipeline repair. In this section, earthquake hazard delineation is examined through the identification of liquefiable soils and their potential for lifeline damage, and through actual lifeline damage patterns as a means of identifying local PGD hazards, including liquefaction. Liquefaction Hazards As mentioned previously, pipeline damage can be caused by various types of PGD triggered by liquefaction. The delineation of liquefaction hazards, therefore, is of primary interest in assessing the potential vulnerability of lifelines during an earthquake. Youd and Perkins (1978) developed procedures for identifying sedimentary deposits susceptible to liquefaction and combining maps of these deposits with the spatial delineation of liquefaction opportunity. To quantify opportunity, it is necessary to identify earthquake sources, provide estimates of the number and magnitude of earthquakes in the source zones, and apply source-distance relationships that link moment magnitude Mw, and distance from source with the occurrence of liquefaction. On the basis of historical evidence of liquefaction-induced ground deformation, Youd and Perkins (1987) also developed a technique for compiling liquefaction hazard maps by mapping a parameter called the liquefaction severity index (LSI). The LSI represents the general maximum differential ground movement (in inches) associated with lateral spread that can be anticipated in active flood plains, deltas, or other areas of gently sloping Holocene fluvial deposits. By means of statistical correlations, Youd and Perkins developed an equation relating LSI, earthquake magnitude, and distance from the seismic energy source for data pertaining to western U.S. earthquakes. The equation, a model of seismic sources, and a published seismic risk algorithm were used to compile probabilistic LSI maps for southern California. Bartlett and Youd (1995) developed an empirical model on the basis of multiple linear regression (MLR) analyses for predicting the horizontal ground displacement resulting from liquefaction-induced lateral spread. They used data from Japanese and U.S. earthquakes, and distinguished two general types of lateral spread: 1) lateral spread towards a free face, and 2) lateral spread down a gentle slope where a free face is absent. An equation was proposed for predicting the magnitude of horizontal displacement as a function of six variables. Using a similar database, Bardet, et al. (1996a and b) proposed a four-parameter MLR model of the form: Log (D + 0.01) = b0 + b off + b1Mw + b2 Log (R) + b3R + b4 Log (W) + b5 Log (S) +b6 Log (T15) (1) in which D is the horizontal displacement (m); Mw the moment magnitude; R the nearest horizontal distance (km) to seismic energy source or fault rupture; S the slope (%) of the ground surface; W the free face ratio (%) defined as height of the free face divided by the distance from the free face to the location of displacement, T15 the thickness (m) of saturated cohesionless soils (excluding depth > 20 m and clay content > 15%) with corrected standard penetration test values N160 < 15. In free-face cases, the term Log (S) is zero. In ground slope cases, the term Log (W) is zero. The b-coefficients were derived from MLR analyses, and are summarized by Bardet, et al. (1999a). In Fig. 3 the horizontal displacements predicted by the MLR model of Bardet, et al. (1999a) are compared with liquefaction-induced lateral movements measured by means of air photos for Kawagishi-cho area of Niigata after the 1964 Niigata earthquake. The two sitespecific variables of the four-parameter MLR model are the thickness of liquefiable deposits, T15, and the free-face ratio, W. The spatial distribution of T15 is evaluated by interpolation procedures, as explained by Bardet, et al. (1999a). Figs. 3d and e show the borings and contours of T15 determined from the borings, respectively. Fig. 3f shows the spatial variation of W, and Fig. 3c provides a three-dimensional plot of the ground surface taken from topographic maps of the area. The values of T15 and W were calculated at 50 by 34 grid points evenly spaced on a 10.5-m grid interval. The predicted displacement contours are illustrated in Fig. 3b, which can be compared with the displacement contours determined from air photo measurements in Fig. 3a. The irregular pattern of lateral movement at site, with a locally high concentration of displacement, is not duplicated by the MLR model. The model is influenced strongly by W, and this parameter in Eq. 1 tends to control the pattern of predicted displacement, resulting in contours of predicted movement that are characteristically parallel to the free face. Even though local concentrations of displacement were not predicted by the model, the average magnitude of predicted displacements are nonetheless consistent with the measured movements. The MLR four-parameter model predicts mean values, and probabilistic methods were applied to determine confidence limits on lateral movement and the probability of exceeding some level of ground deformation (Bardet, et al., 1999b). Fig. 4 presents a map of the predicted probability of liquefaction-induced lateral spread larger than 2 m in the Kawagishi-cho area for an earthquake with similar magnitude and source-to-site characteristics as the 1964 Niigata earthquake. The parallel pattern of probability contours reflects the influence of W. Probabilitybased maps for liquefaction-induced displacements are consistent with the probabilistic procedures frequently followed for seismic hazard characterization, and are valuable for supporting decisions by lifeline operators regarding the earthquake risk incurred at key facilities. Lifeline Damage Patterns Seismic zonation generally involves the identification of seismic and geotechnical hazards followed by an assessment of the potential damage associated with such hazards. The process can be inverted for pipeline systems. When there are many repair locations, the pipeline network can be used as a grid with which to characterize the local densities of repair so that subsurface hazards can be located. Pipeline repair patterns for the 1989 Loma Prieta and 1906 San Francisco earthquakes were used successfully to supplement soil boring information and characterize liquefaction hazards in San Francisco (Pease and O’Rourke, 1997; O’Rourke and Pease, 1997). In fact, the combination of pipeline repair locations and soil borings was quite effective in delineating zones vulnerable to liquefaction at a scale consistent with the size of a city block. As previously mentioned, pipeline repair records in the Marina of San Francisco were evaluated relative to liquefiable soils at the site. The pattern of pipeline repairs was instrumental in identifying ground oscillation as the principal cause of damage, and in developing simplified models to estimate the deformation that can be triggered by this type of liquefaction phenomenon. SHINAN RIVER (a) Observed Displacement (cm) (d) Borehole Location (b) Predicted Displacement (cm) (e) T15 (m) (c) Ground Surface (m) (f) Free-face Ratio H/L (%) Figure 3. Representation of Measured and Predicted Lateral Displacements and Key Surface and Subsurface Characteristics for MLR Model Prediction of Liquefaction-Induced Movement During the 1964 Niigata Earthquake (Bardet, et al., 1999b) ECHIGO RAILWAY LINE 211950 211900 211850 211800 ECHIGO RAILWAY BRIDGE KAWAGISHI CHO 211750 211700 SHINANO RIVER 211650 47000 47100 47200 47300 47400 47500 Figure 4. Probability of Liquefaction-Induced Lateral Spread Larger Than 2 m in the Kawagishicho Area Predicted by MLR Model (Bardet, et al., 1999b) In this paper, the GIS database for water distribution system performance during the 1994 Northridge earthquake is used to delineate geotechnical hazards in the Los Angeles region. The earthquake-induced damage to water pipelines and the database developed to characterize this damage have been described elsewhere (O’Rourke, et al., 1998; O’Rourke and Toprak, 1997), and only the salient features of this work are summarized herein. GIS databases for repair locations, characteristics of damaged pipe, and lengths of distribution (pipe diameter < 600 mm) and trunk (pipe diameter ≥ 600 mm) lines according to pipe composition and size were assembled with ARC/INFO software. Nearly 10,000 km of distribution lines and over 1,000 km of trunk lines were digitized. Fig. 5 shows the portion of the Los Angeles water supply system most seriously affected by the Northridge earthquake superimposed on the topography of Los Angeles. The figure was developed from the GIS database, and shows all water supply pipelines plotted with a geospatial precision of ± 10 m throughout the San Fernando Valley, Santa Monica Mountains, and Los Angeles Basin. The rectilinear system of pipelines is equivalent to a giant strain gage. Seismic intensity in the form of pipeline damage can be measured and visualized by plotting pipeline repair rates and identifying the areas where the largest concentrations of damage rate occur. The resulting areas reflect the highest seismic intensities as expressed by the disruption to underground piping. To develop a properly calibrated strain gage, it is necessary to select a measurement grid with material having reasonably consistent properties and a damage threshold sensitive to the externally imposed loads being measured. Fig. 6 presents charts showing the relative lengths of trunk and distribution lines according to pipe composition. As shown by the pie chart, the most pervasive material in the Los Angeles distribution system is CI. The 7,800 km of CI pipelines have the broadest geographic coverage with sufficient density in all areas to qualify as an appropriate measurement grid. Moreover, CI is a brittle material subject to increased rates of damage at tensile strains on the order 250 to 500 µε. It is therefore sufficiently sensitive for monitoring variations in seismic disturbance. Fig. 7 presents a map of distribution pipeline repair locations and repair rate contours for CI pipeline damage. The repair rate contours were developed by dividing the map into 2 km x 2 km areas, determining the number of CI pipeline repairs in each area, and dividing the repairs by the distance of CI mains in that area. Contours then were drawn from the spatial distribution of repair rates, each of which was centered on its tributary area. A variety of grids were evaluated, and the 2 km x 2 km grid was found to provide a good representation of damage patterns for the map scale of the figure (Toprak, et al., 1999). The zones of highest seismic intensity are shown by areas of concentrated contours. In each instance, areas of concentrated contours correspond to zones where the geotechnical conditions are prone either to ground failure or amplification of strong motion. Each zone of concentrated damage is labeled in Fig. 8 according to its principal geotechnical characteristics. In effect, therefore, Fig. 7 is a seismic hazard map for the Los Angeles region, calibrated according to pipeline damage during the Northridge earthquake. Figure 5. Map of Los Angeles Water Supply System Affected by Northridge Earthquake (O’Rourke and Toprak, 1997) Cast Iron 11% Concrete 18% Ductile Iron 1% Trunk Lines : 1014 km Distribution Lines : 10750 km Riveted Steel 14% Steel 56% a) Trunk Lines Asbestos 9% Steel 11% Length (km) 10000 Concrete Riveted Steel Cast Iron Steel Asbestos Cement Ductile Iron 1000 Ductile Iron 4% 100 Cast Iron 76% b) Distribution Lines LADWP LADWP MWD c) Combined Lines Figure 6. Composition Statistics of Water Trunk and Distribution Lines in the City of Los Angeles (O’Rourke and Toprak, 1997) Figure 7. Cast Iron Pipeline Repair Rate Contours for the Northridge Earthquake (O’Rourke and Toprak, 1997) Figure 8. Geotechnical Characteristics of the Areas of Concentrated Pipeline Damage After the Northridge Earthquake Of special interest is the location of concentrated repair rate contours in the west central part of San Fernando Valley (designated in Fig. 8 as the area of soft clay deposits). This area was investigated by USGS researchers, who found it to be underlain by local deposits of soft, normally consolidated clay (Holzer, et al., 1999). Field vane shear tests disclosed clay with uncorrected, vane shear undrained strength, Suvst = 20-25 kPa, at a depth of 5 m, just below the water table. USGS investigators concluded that the saturated sands underlying this site were not subjected to liquefaction during the Northridge earthquake. Newmark sliding block analyses reported by O’Rourke (1998) provide strong evidence that near source pulses of high acceleration were responsible for sliding and lurching on the soft, normally consolidated clay deposit. The results of GIS analysis and site investigations have important ramifications because they show a clear relationship between PGD, concentrated pipeline damage, and the presence of previously unknown deposits of normally consolidated clay. With GIS, it is easy to divide a spatially distributed data set into arbitrarily sized areas. If the areas are delineated by a framework of equally spaced, vertical and horizontal lines, the resulting grid can be characterized by a single dimension representing one side of each area, n, and the number, N, of areas comprising the total area of the system, Nn2. The choice of n can be regarded as a means of visually resolving the distribution of damage. Some practical questions emerge. Is there a useful relationship between n and the visualization of zones with high damage? What values of n represent the best choices for visualizing damage patterns? Toprak, et al. (1999) found a relationship between the area of the map covered by repair rate contours and the grid size, n, used to analyze the repair statistics. If the contour interval is chosen as the average repair rate for the entire system or portion of the system covered by the map, then the area in the contours represents the zones of highest (greater than average) earthquake intensity as reflected in pipeline damage. The area within the contour lines divided by an area closely related to the total area of the map, AC, is referred to as the threshold area coverage, TAC (Toprak, et al., 1999). Alternate thresholds may also be defined on the basis of the mean plus some measure of the variance. The relationship between TAC and grid size was explored for a variety of map sizes. Fig. 9 presents a map showing the location of pipeline damage in the City of Los Angeles on which are superimposed the rectangular outlines of several smaller areas. The visualization of damage patterns for the entire area of the LADWP system affected by the Northridge earthquake and for the smaller rectangular areas in Fig. 9 was investigated as a function of grid size. These data were supplemented by pipeline and damage databases for the San Francisco Marina after the 1989 Loma Prieta earthquake. A map of San Francisco and the Marina are shown at the same scale as the Los Angeles study area in Fig. 9. A hyperbolic relationship was shown to exist between TAC and the dimensionless grid size, defined as the square root of n2, the area of an individual cell, divided by the total map area, AT. This relationship is illustrated in Fig. 10, for which a schematic of the parameters is provided by the inset diagram. The relationship was found to be valid over a wide range of different map scales spanning 1,200 km2 for the entire Los Angeles water distribution system Figure 9. LADWP and Marina Study Area (Toprak, et al., 1999) 0.6 Threshold Area Coverage,TAC 0.5 0.4 Fit Curve 0.3 All LA All SFV 0.2 Section of SFV Sherman Oaks North of LA Basin 0.1 San Francisco TAC = DGS / ( 1.68 * DGS + 0.11) 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Dimensionless Grid Size, DGS 0.8 Figure 10. Hyperbolic Fit for Threshold Area Coverage and Dimensionless Grid Size (Toprak, et al., 1999) affected by the Northridge earthquake to 1 km2 of the San Francisco water distribution system in the Marina affected by the Loma Prieta earthquake (Toprak, et al., 1999). The data points refer to maps of various dimensions from which the relationship was developed. Physical Loss Estimation In this section, physical loss estimation will be examined with emphasis on the Northridge earthquake database. Loss estimation is addressed with respect to TGD and PGD effects. Transient Ground Deformation Effects The records from 241 Northridge earthquake strong motion instruments were examined, and the data from 164 corrected records were selected for regression analyses (O’Rourke and Torprak, 1997). In this paper, additional studies were performed with data from 142 selected Northridge earthquake records processed and catalogued by Silva (2000), and made available online by the Pacific Engineering Earthquake Center. All records were chosen to represent free field motion. Fig. 11 shows the CI pipeline repair rate contours (see Fig. 7) superimposed on peak ground velocity (PGV) zones. The PGV zones were developed by interpolating the larger of the two horizontal components associated with each of 164 corrected motion sites. Using the GIS database, a pipeline repair rate was calculated for each PGV zone, and correlations were made between the repair rate and average PGV for each zone. As explained by O’Rourke (1998), similar correlations were investigated for pipeline damage relative to spatially distributed peak ground acceleration, spectral acceleration and velocity, Arias Intensity, Modified Mercalli Intensity (MMI), and other indices of seismic response. By correlating damage with various seismic parameters, regressions were developed between repair rate and measures of seismic intensity. The most statistically significant correlations for both distribution and trunk line repair rates were found for PGV. The parameter, however, can be defined in several different ways. In attenuation relationships, PGV is commonly defined as the geometric mean of the two largest horizontal components (e.g., Campbell, 1997). PGV is also defined as the larger of the two horizontal components (e.g., Boore, et al., 1997), which is the value used in Fig. 11. PGV may also be defined as the maximum vector magnitude of the two horizontal components. Figs. 12a, b, and c show the CI repair rates for the Northridge earthquake regressed against the geometric mean PGV, maximum PGV, and maximum vector magnitude of PGV. The data from the 142 records processed and catalogued by Silva were used to develop these regressions. The plots indicate that the choice of PGV makes little difference in the statistical significance of the regressions. All are characterized by r2 that are comparable, although the highest r2 is associated with maximum PGV. Fig. 12d shows the maximum PGV regressed against the geometric mean for 162 strong motion records for the Northridge earthquake. In the remainder of this paper, the PGV cited will refer to the maximum PGV. If the forthcoming regression equations for PGV are to be used in conjunction with attenuation Figure 11. Pipeline Repair Rate Contours Relative to Northridge Earthquake Peak Ground Velocity (O’Rourke and Toprak, 1997) relationships based on the geometric mean, Fig. 12d can be applied to estimate maximum PGV, from predicted geometric means. As reported by O’Rourke and Jeon (1999), statistics were compiled for the repair rates of CI, DI, and AC distribution lines, and regressions were developed for repair rate vs. PGV. The compilation of statistics for steel distribution pipelines showed that there were many different types of steel pipeline grouped within this category. Figs. 13a and b present pie charts for the lengths and earthquake repairs associated with the different types of steel pipelines. At least six different kinds of pipeline were identified after a careful review of the records. Matheson and Mannesman steel pipelines were installed primarily in the 1920s and 1930s. Most were installed without cement linings and with minimal coating. In addition, their wall thickness is generally less than that of other steel pipes with similar diameter. Matheson and Mannesman steel pipelines are vulnerable to corrosion, as are steel pipelines with threaded couplings. In the latter case, corrosion tends to concentrate at the threaded cross-sections. These types of pipelines did not perform well during previous U.S. and Japanese earthquakes (Katayama and Isoyama, 1980; Eguchi, 1982; O’Rourke, et al., 1985). 4 3 3 Repair Rate (Number of Repairs/km) Repair Rate (Number of Repairs/km) 4 2 0.10 9 8 7 6 5 4 3 Fit Equation: log(Y) = 1.09 * log(X) - 6.12 R-squared = 0.82 2 0.01 2 0.10 9 8 7 6 5 4 3 Fit Equation: log(Y) = 1.21 * log(X) - 6.78 R-squared = 0.84 2 0.01 10 100 10 Geometric Mean PGV (cm/sec) Maximum PGV (cm/sec) (a) Geometric Mean 100 (b) Maximum 200 4 Fit Equation: Y = 1.21 X R-squared = 0.98 2 Maximum PGV (cm/sec) Repair Rate (Number of Repairs/km) 3 0.10 9 8 7 6 5 4 3 150 1 1 100 50 Fit Equation: log(Y) = 1.30 * log(X) - 7.21 R-squared = 0.77 2 n = 162 0 0.01 10 100 Maximum Vector Magnitude of PGV (cm/sec) (c) Maximum Vector Magnitude 0 50 100 150 200 Geometric Mean PGV (cm/sec) (d) Maximum vs. Geometric Mean Figure 12. CI Repair Rate Regression with Geometric Mean, Maximum, and Maximum Vector Magnitude of PGV, and Relationship Between the Maximum and Geometric Mean PGVs Matheson 3% Threaded Couplings 4% Victaulic 14% Mannesman Riveted 3% 1% Victaulic 7% Threaded Couplings 9% Riveted Unknown 1% 10% Mannesman 11% Welded 75% Total Pipeline Length: 1018 km Total Repairs: 205 Repair Rate a) Steel Pipeline Length 1.2 1.0 0.8 0.6 0.4 0.2 0.0 Welded 43% Matheson 19% b) Steel Pipeline Repairs 1.08 0.77 0.49 0.11 Welded 0.24 0.11 Victaulic Threaded Couplings Matheson Mannesman Riveted c) Steel Pipeline Repair Rate Figure 13. Pipeline Lengths, Repairs, and Repair Rates for Various Types of Steel Pipeline in Service During the Northridge Earthquake Victaulic couplings are bolted, segmental, clamp-type mechanical couplings whose housings enclose a U-shaped rubber gasket (American Water Works Association, 1964). The gasket tends to deform and lose its initial water-tight characteristics under prolonged service. Riveted steel pipelines are older installations, which are prone to corrosion. Contact between the rivets and laminated steel of the pipe body promotes galvanic action between the two dissimilar metals. A welded slip joint is fabricated by inserting the straight end of one pipe into the bell end of another and joining the two sections with a circumferential fillet weld. The bell end is created by the pipe manufacturer by inserting a mandrel in one end of a straight pipe section, and expanding the steel into a flared, or bell casing. The pie charts in Fig. 13 show that this type of steel pipeline (refered to as welded) was the predominant type operated during the earthquake. The histogram in Fig. 13c indicates very high repair rates associated with Matheson, Manneson, and threaded steel pipelines. The lowest repair rates are associated with welded steel pipelines. Because of their broader coverage and greater aggregate length, regression analyses were performed for this type of steel pipeline. Fig. 14a shows the repair rates for steel (Steel Distr.), CI, DI, and AC distribution lines regressed against PGV. The regressions indicate that the highest rate of damage for a given PGV was experienced by steel pipelines. This result at first seems surprising because steel pipelines are substantially more ductile than CI and AC pipelines. Steel distribution pipelines in Los Angeles, however, are used to carry the highest water pressures, installed in areas of relatively steep slopes susceptible to landslides, and are subject to corrosion that has been shown to intensify their damage rates in previous earthquakes (Isenberg, 1979). Comparatively high repair rates for steel pipelines were reported by Eidinger (1998) after the 1989 Loma Prieta earthquake. Fig. 14b compares the regression equations derived from the GIS database for the Northridge earthquake with the default regression currently used in HAZUS (National Institute for Building Sciences, 1997). The regression for the steel trunk lines (nominal diameter ≥ 600 mm) was developed according to similar procedures followed for the distribution pipelines. The steel trunk lines are constructed with welded slip joints. Virtually all the regressions developed from the Northridge GIS database plot significantly below the HAZUS default. This trend is especially true for steel trunk lines that show repair rates 10 to 20 times lower than those estimated with the HAZUS default. Fig. 14c presents the linear regression that was developed between CI pipeline repair rate and PGV on the basis of data from the Northridge and other U.S. earthquakes. By taking advantage of additional data, this regression provides a more comprehensive representation of repair rate trends. It is not significantly different from the regression in Fig. 12b, thereby implying a consistency between the Northridge earthquake trends and previous earthquake statistics. As explained by O’Rourke and Jeon (1999), The regression relationship for the repair rate, RR, can be expressed as Log RR = Log K + α Log VP – β Log DP (2) in which K, α, and β are constants, VP is PGV, and DP is the pipe diameter in cm. Eq. 2 can be rewritten as RR = K (VP/DPβ/α)α (3) in which VP/DPβ/α is referred to as the scaled velocity that represents PGV normalized with respect to diameter. The exponent of DP is a scaling factor that accounts for the statistical relationship embodied in the database. This relationship is, in turn, a function of the strong motion and pipeline system characteristics. Fig. 14d shows the linear regression between repair rate and scaled velocity. Combining the PGV and DP into a common parameter offers a new tool for loss estimation. Scaled velocity combines the effects of PGV and nominal diameter, and thus accounts for the two most important variables affecting earthquake damage to pipeline of a particular composition. Repair Rate (Number of Repairs/km) 4 Fit Equation (Steel Distr.): log(Y) = 0.88 * log(X) - 5.28 R-squared = 0.90 3 2 0.1009 Steel Distr. CI DI AC 8 7 6 5 4 3 2 Fit Equation (CI): log(Y) = 1.21 * log(X) - 6.78 R-squared = 0.84 Fit Equation (DI): log(Y) = 1.83 * log(X) - 9.39 R-squared = 0.73 Fit Equation (AC): log(Y) = 2.26 * log(X) - 11.02 R-squared = 0.71 0.0109 8 7 6 5 4 3 2 0.001 10 2 3 4 5 CI 0.10 fffdsfd 9 8 7 6 4 Steel Trunk 3 AC 2 0.01 100 2 10 3 4 5 6 7 89 2 100 PGV (cm/sec) (b) Regressions vs. HAZUS Default 4 4 Fit Equation: log(Y) = 1.55 * log(X) - 8.15 R-squared = 0.85 3 2 0.10 9 8 7 6 Repair Rate (Number of Repairs/km) Repair Rate (Number of Repairs/km) DI 5 6 7 8 9 (a) Steel Distr., CI, DI, and AC aaa 5 4 3 1994 Northridge 1989 Loma Prieta 2 Steel Distr. 2 PGV (cm/sec) 3 HAZUS 3 Repair Rate (Number of Reparis/km) 4 1987 Whittier Narrows Fit Equation: RR = 0.036 (VP/DP1.021)0.989 R-squared = 0.60 2 0.10 9 8 7 6 5 4 3 2 1971 San Fernando (south) 0.01 0.01 10 2 3 4 5 6 7 8 9 PGV (cm/sec) (c) Combined CI Database 100 0.1 1.0 Scaled Velocity 10.0 VP/DP1.021, ((cm/sec)/cm1.021) (d) Scaled Velocity Figure 14. Regression for the Repair Rates of Various Types of Pipelines vs. PGV and Scaled Velocity The regressions in Fig. 14 were developed after the data were screened for lengths of pipeline that represent approximately 1.5 to 2.5 % of the total length or population for each type of pipe affected by the earthquake (O’Rourke and Jeon, 1999). This procedure reduces the influence of local erratic effects that bias the data derived from small lengths of pipeline. The use of this filtering procedure leads to statistically significant trends. The regressions are applicable only for PGV ≤ 75 cm/s. For the Northridge earthquake, zones with PGV exceeding 75 cm/sec generally correspond to locations where PGD, from sources such as liquefaction and landsliding, was observed. Hence, this screening technique tends to remove damage associated with PGD, resulting in correlations relevant for TGD. Permanent Ground Deformation Effects Among the most notable research accomplishments in recent years is the work of Hamada and coworkers (Hamada, et al., 1986; Hamada and O’Rourke, 1992) in the use of stereo-pair air photos before and after an earthquake to perform photogrammetric analysis of large ground deformation. This process has influenced the way engineers evaluate soil displacements by providing a global view of deformation that allows patterns of distortion to be quantified and related to geologic and topographic characteristics. After the Northridge earthquake, pre- and post- earthquake air photo measurements in the Van Norman Complex were analyzed as part of collaborative research between U.S. and Japanese engineers (Sano, et al., 1999; O’Rourke, et al., 1998). Air photos taken before and after the earthquake were acquired by U.S. team members and analyzed through advanced photogrammetric techniques by Japanese team members. Ground movements from this initial set of measurements were corrected for tectonic deformation to yield movements caused principally by liquefaction and landslides. The area near the intersection of Balboa Blvd. and Rinaldi St. has been identified as a location of liquefaction (Holzer, et al., 1999) where significant damage to gas transmission and water trunk lines was incurred. Ground strains were calculated in this area from the air photo measurements of horizontal displacement by superimposing regularly spaced grids with GIS software onto the maps of horizontal displacement and calculating the mean displacement for each grid. Grid dimensions of 100 m x 100 m were found to provide the best results (Sano, et al., 1999). As illustrated in Fig. 15, ground strain contours, pipeline system, and repair locations were combined using GIS, after which repair rates corresponding to the areas delineated by a particular contour interval were calculated. Fig. 16 shows the repair rate contours for CI mains superimposed on the areal distribution of ground strains, identified by various shades and tones. In the study area, there were 34 repairs to CI water distribution mains and 2 for steel water distribution pipelines. There were 5 water trunk line repairs in the area. The repair rate contours were developed by dividing the map into 100 m x 100 m cells, determining the number of CI pipeline repairs in each cell, and dividing the repairs by the length of the distribution mains in that cell. The intervals of strain and repair rate contours are 0.001 (0.1%) and 5 repairs/km, respectively. The zones of high tensile (+) and compressive (-) strains coincide well with the locations of high repair rate. Surface Analysis Contour Interpolation Ground Strains Ground Strain Contours Overlay Length in Each Strain Range Repair Rate (Repairs/Length) vs. Ground Strain Pipeline System Overlay Repairs in Each Strain Range Pipeline Repairs Figure 15. Procedure for Calculating Repair Rate in Each Strain Range (O’Rourke, et al., 1998) Figure 16. Distributions of CI Repair Rate and Ground Strain (O’Rourke, et al., 1998) In Fig. 17, the relationship between the absolute values of the ground strains and repair rates is presented graphically using linear regression. The repair rate in each ground strain range, 0-0.1, 0.1-0.3, and 0.3-0.5%, was calculated as explained previously. Ground strain contours, obtained by both air photo measurements and surface surveys, were used. The regression analysis shows that repair rates increase linearly with ground strain. In some instances, anomalously high repair rates were determined. Such values do not represent the actual distribution of damage, but are a consequence of locally high concentrations of repairs within a given cell. The occurrence of anomalously high values depends on the cell size and positioning of cells with respect to pipeline repair locations. It is important, therefore, to incorporate screening procedures to filter such erroneous types of data. For example, an investigation of the locally high data point in Fig. 17 showed that this repair rate was calculated for one particular positioning of the grid of GIS cells and not for others. As such, this locally high repair rate was not used in the regression analyses. Figure 17. Correlation Between Ground Strain and CI Repair Rate (O’Rourke, et al., 1998) Hamada and Wakamatsu (1996) showed a similar strong correlation between frequency of pipeline repairs and ground strains evaluated by air photo measurements after the 1995 Kobe earthquake. Seismic design codes for gas pipelines in Japan have been developed on the basis of acceptable strain levels (Japan Gas Association, 2000). These values, in turn, can be related to the anticipated levels of ground strain triggered by an earthquake. Economic and Social Consequences Advances in computer technologies, particularly GIS, have led to significant developments in computerized methodologies for estimating expected losses from future earthquakes (e.g., National Research Council, 1999; Earthquake Engineering Research Institute, 1997). Of special interest are recent advances in estimating the economic consequences caused by earthquake-induced damage in lifeline systems. A multi-year, multidisciplinary research effort has been focused on the Memphis Light, Water, and Gas (MLWG) lifeline systems in Memphis and Shelby County, TN that are at risk from earthquakes originating in the New Madrid Seismic Zone (Chang, et al., 2000a; Shinozuka, et al., 1998; Rose, et al., 1997, and Chang, et al., 1996). The methodology involves the evaluation of individual earthquake scenarios for events with different magnitudes and distances. For each scenario earthquake, Monte Carlo simulations are performed during which earthquake effects are modeled in the lifeline system. In the case of the MLWG electric power system, the peak ground accelerations (PGAs) were computed at 424 different sites for a scenario earthquake, spatially interpolated, and linked with various damage states in electric substations. Ratios of power output in the damaged system to power output in the undamaged system were evaluated for various parts of Shelby County. Assessments were also made of the number of days for restoration of power, as illustrated in Fig. 18a. Restoration time is an important index with respect to the economic and social consequence of lifeline damage. The estimated restoration times are shown for various parts of Shelby County for a 7.5 Mw earthquake at Marked Tree, AR. The times are based on the average state of repair determined from 100 simulations. For the MLWG water distribution system, damage states were evaluated with a hydraulic network model specifically configured for the piping and facilities operated by MLWG. For each earthquake scenario, 100 cases of simulated system-wide damage were produced according to procedures similar to those followed for the electric power system. Fig. 18b shows the ratio of water flow in the damaged system, averaged over the 100 simulations, to water flow in the undamaged system for a 7.0 Mw earthquake at Marked Tree, AR. The water flow and associated damage states are for the week immediately following the earthquake. The water delivery system damage is linked with economic consequences through a methodology that correlates water losses with areas of economic activity, adjusts for business resiliency (remaining percentage output that can still be produced by a specific industry in the event of total water outage), and accounts for direct and indirect economic losses. Indirect economic loss is defined as business interruption arising from interaction among regional industries through changes in input demands or output sales. Indirect economic losses generated by seismic damage to lifelines were initially estimated with Input-Output (I-O) analysis (Rose, et al., 1997), which is based on tabulated economic data pertaining to purchases (inputs) and sales (outputs) among all sectors of the regional economy. The model is based on the assumption that a linear relationship exists between inputs and outputs. Recent modeling employs Computable General Equilibrium (CGE) methods to assess losses. This method incorporates nonlinear interactions through multi-market simulation models that optimize the behavior of consumers Number of Days for Restoration 0–1 2–3 4–7 8 – 14 (a) Electric Power Restoration Flow Ratio 0.00 – 0.10 0.10 – 0.20 0.20 – 0.30 0.30 – 0.40 0.40 – 0.50 0.50 – 0.60 0.60 – 0.70 0.70 – 0.80 0.80 – 0.99 0.99 – 1.00 (b) Water Losses S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S ## S S # # S S # S # S # S # S # S # S # S # S # S # S # # S # S S # S # S # S # S # S # S # S # S # S # S # S# # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # # S S # S # S # # S S # # S S # # S S# S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # # S # S S# # S # S # S S # S # S S # # S # S # S # S # S # # S S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # # S S # S S # # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # # S ## S S S # S # S # S # S # S # # S S S# # S # S # S # S # S # S # # S S # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S# # SS# S S # # S# S # S# # S S# S # S# S # # S# S S # S # S # S # S # S # S # S # S # S# # S S # S # S # S # S # S # S # S# # S S # S # S# # S S # S # S # S # S # S# # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # S # # S S # S # S # S # S# # S# S S # S # S# # S S # # S # S# S S # S # S # S # S # # S S # S # S # S # S # S # S # # S S # S # S # # # S S S # S # S # S # S # S # S # S # S # S # S # S# # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S# # S S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # S # S # S S# # S # S # S # S # S # S# # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # # S S # S # S # S# # S S # S # S # S # S # S # S # # S S # # S S # S # S # S # # S S # S # S # S # S # S # S # S # SS # #S # S # # S S # S # S # S # S # S # S # S # Analysis Zone S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S S# # S # S # S # # S S # S # S # S # S # S # S # S # S # S # S # S # S # # S S # S # S # # S S # S # S # S # # S Service Zone TAZ Demand Node S # S # # S S # S # S # S # # S # S # S S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # 0 S # S # 10 S # 20 Kilometers (c) Analyses Zones Figure 18. GIS and Seismic Zonation for Assessing the Economic Effects of Lifeline Losses in Shelby County, TN (after Chang, et al., 2000a; Shinozuka, et al., 1998) and businesses, subject to economic account balances and resource constraints (e.g., Shoven and Whalley, 1992). It is beyond the scope of this paper to expand on the economic models, and reference to other publications (e.g., Chang, et al., 2000a; Shinozuka, et al, 1998) should be made for further details. It is of interest for this paper, however, to explain briefly how lifeline damage is spatially correlated with economic activity. Economic data for various areas of the U.S. are summarized according to census tracts, which are divisions of land that are designed to contain 2,500 to 8,000 inhabitants with relatively homogeneous population characteristics, economic status, and living conditions. There are 133 census tracts in Shelby County. To provide a finer grain for economic analysis, Traffic Analysis Zones (TAZs) were used. There are 515 TAZs for Shelby County that contain information on local employment. Such zones represent aggregated census data developed jointly by the U.S. Bureau of Census and U.S. Department of Transportation. Fig. 18c shows how analysis zones were created by overlaying 515 TAZs with 971 polygons. The GIS Thiessen polygons (e.g., Burrough and McDonnell, 1998) were generated to represent the inferred service areas for all demand nodes in the water delivery system. Assuming that there is uniform labor distribution within each industry, the employment data for each TAZ can be linked with economic output. The amount of economic output in each TAZ affected by water losses was estimated as being directly proportional to the fraction of the TAZ overlapped by a polygon. In this way the economic activity and water losses in Shelby County were correlated in 3,400 analyses zones, thereby providing a spatial framework for the distribution of lifeline damage effects on the regional economy. Methodologies have been proposed for probabilistic risk analyses of lifeline systems that are suitable for estimating both physical damage and economic consequences. Site-specific probabilistic hazard assessments may be applied to lifelines if the lifeline system under study is treated as a single site (e.g. Ishikawa and Kameda, 1994), or if single sites within the lifeline system are analyzed. For a lifeline network, however, the spatial correlation among earthquake effects at many sites is critical for determining system functionality. A procedure has been advanced recently for developing probabilistic scenarios that account for spatial variations in earthquake-induced damage. The procedure extends traditional risk evaluation techniques to spatially dispersed systems (Chang, et al., 2000b), and can be used for benefit-cost analyses of lifelines. The application of this probabilistic approach has been demonstrated by Chang, et al. (2000b) by means of an example involving the Los Angeles highway system. Emergency Response and Recovery Eguchi, et al. (1997) report on how rapid damage assessment tools, based on GIS, were used to facilitate response and recovery decisions after the Northridge earthquake. Data from near real-time broadcasts of earthquake source information, including magnitude, epicenter and measurements of regional peak ground accelerations, served as input parameters for a computerized methodology referred to as Early Post-Earthquake Damage Assessment Tool (EPEDAT). Utilizing the input parameters in conjunction with ground motion and soil amplification models, EPEDAT estimated seismic intensities that were combined through GIS with computerized data on buildings and lifelines to estimate damage and casualties in the impacted area. The results were used by the California Office of Emergency Services to plan and manage emergency response and restoration activities. Modifications of the damage and casualty estimates were made continuously as new information was obtained and integrated into the GIS database. Shimizu, et al. (1996) and Shimizu (2000) report on a special system of strong motion instruments and water level monitors deployed throughout the Tokyo Gas Company network of pipelines and facilities. The system acquires data in virtual real time at about 3,600 locations, including acceleration, Spectrum Intensity (SI), and indirect evidence of liquefaction through changes in acceleration wave forms. SI is defined as the area under the pseudo-velocity response spectra curve between T = 0.1 and 2.5 seconds. It was first introduced by Housner (1952) as a measure of the intensity of an earthquake, and has subsequently been shown to correlate reasonably well with gas pipeline system damage during the 1995 Kobe earthquake. On the basis of the data acquired by the sensor system, the gas flow and pressures at district regulators can be controlled through a central command center. Seismic zonation can provide for the identification of vulnerable portions of lifeline systems and help delineate potential damage scenarios and operational constraints. These conditions constitute a “virtual” framework or baseline for planning and implementing emergency response and applying restoration strategies for system recovery. Given the ability to acquire and process information in real time, real information about post-earthquake damage can be compared rapidly with initial zonation scenarios. Improvements in system characterization can be implemented on an incremental basis as additional information is obtained. Kakumoto, et al. (2000) discuss various options that are available for collecting and wharehousing preearthquake information, and for updating pre-earthquake data with post-earthquake observations. If GIS is used as the vehicle for visualizing and communicating post-earthquake damage and restoration progress, then seismic zonation can be viewed as a dynamic, virtual real-time process for tracking and managing the evolution of spatial changes in critical systems. Recent advances in remote sensing technologies have been discussed by Eguchi, et al. (2000). Satellite imagery, global positioning systems, and synthetic aperture radar are among the promising technologies for rapid data acquisition and system characterization after an earthquake. Concluding Remarks In concluding this paper, it is appropriate to address both the limitations and opportunities associated with seismic zonation. A brief discussion of limitations and opportunities is provided under the subheadings that follow. Limitations As described previously, one of the most significant difficulties in characterizing seismic hazards for lifelines, especially buried lifelines, is the identification of PGD and TGD zones, and the characterization of TGD effects in areas where constructive interference of waveforms and ground strain concentrations may occur. TGD in the form of surface waves in sedimentary basins, vibration of relatively narrow soil-filled valleys, and liquefaction-induced ground oscillation may induce locally high ground strains, especially near basin and valley margins where soft sediments impinge on less deformable geologic media. In all PGD zones, there are some TGD effects. As a consequence, PGD and TGD effects tend to merge, making a clear distinction between the two very difficult at some locations. To resolve accurately the locations and areal coverage of PGD zones, sufficient data are required to characterize the subsurface deposits and delineate the potential boundaries of movement. The spatial variability of soil, surface gradients, subsurface geometry of deposits, and groundwater coupled with sparse information at most locations poses a substantial challenge for the seismic zonation of lifelines. Fortunately, increased use of GIS and computerized databases make it possible to collect and store large amounts of information for a particular region. As site explorations are performed in the future, the ensuing information can be added to existing databases to improve subsurface characterization. Regression analyses for predicting pipeline damage and liquefaction-induced PGD are performed by assuming that seismic and subsurface parameters are known without error. For example, when the pipeline repair rate is regressed against PGV, it is assumed as a premise of the regression analysis that PGV is known precisely. A similar assumption applies to the thickness of liquefiable soil in MLR analysis of liquefaction-induced horizontal movement. It should be recognized that procedures are available to account for the variability of the different parameters in regression analyses (e.g., Snedecor and Cochran, 1980). Moreover, methods, such as kriging (e.g., Ripley, 1987), inherently account for spatial variability, and can be adapted to delineate confidence limits and probabilities of exceedance for a particular parameter in two-dimensional space. In the future, it will be advantageous to explore the variability of all regression parameters to reflect properly the spatial uncertainties of seismic and site conditions. Lifeline systems are composed of diverse components and facilities that may differ from one system to another. To apply regressions or correlations that were developed with emphasis on a particular system, one must be satisfied that the characteristics of that system are representative of others for which the predictions are being made. The statistics of pipeline performance from GIS investigations of water distribution systems after the Northridge and Kobe earthquakes show several consistent trends. For both earthquakes, water main damage was inversely proportional to nominal diameter, directly proportional to PGA, and the ratios of repair rates for DI and CI pipelines are comparable for smaller diameter lines. For steel distribution pipelines, however, the repair rates are significantly different. The repair rates for steel distribution pipelines after the Northridge earthquake are higher than those of CI pipelines, whereas the opposite trend applies for the Kobe earthquake. The relatively high susceptibility of steel piping to damage during the Northridge earthquake may be explained in part by utility practices, such as using steel pipe for the highest internal pressures and deploying steel pipe in areas with relatively steep slopes. Increased susceptibility to corrosion also appears to play a role in steel pipeline performance. Pipeline system complexities apply to operational strategies and company procedures. It is necessary, therefore, to be aware that practices can vary among different utilities and that local differences may affect the suitability of empirically-based predictions. Opportunities Advances in seismic zonation for lifelines and utilities have been influenced to a very significant extent by records of lifeline performance acquired in recent earthquakes, most notably the 1989 Loma Prieta, 1994 Northridge, and 1995 Kobe earthquakes. Data from the Northridge and Kobe earthquakes have been compiled in GIS of unprecedented size and complexity that allow for a detailed examination of the spatial relationships among lifeline damage, permanent and transient ground deformation, and subsurface soil and groundwater conditions. Pipeline damage data from the Northridge earthquake are useful not only for developing predictive regressions for different types of piping, but for identifying areas of seismic hazards. When compared with default predictions currently embodied in the National Loss Estimation Methodology, the regressions presented in this paper show substantially lower repair rates. Damage statistics from recent earthquakes and improvements in GIS technology provide the opportunity for comprehensive assessment of system performance and the characterization of spatial variability at a level of detail previously unattainable. One of the most significant trends in seismic zonation for lifelines and utilities is the development of loss modeling techniques that link lifeline disruption with the economic consequences of such damage in a given region. Research focused on water and electric power systems in Shelby County, TN has resulted in techniques for spatially correlating lifeline damage with economic losses. Analytical procedures, such Input-Output analysis and Computable General Equilibrium, have been applied successfully to estimate economic losses from earthquake damage to lifelines. This approach is well suited for benefit/cost assessments to evaluate the most appropriate seismic retrofitting and policy initiatives. Ultimately, losses can be characterized by economic fragility curves that indicate the likelihood of exceeding different levels of economic loss, with and without mitigation measures. The resulting information will assist lifeline and utility agencies in deciding on suitable mitigation procedures. Acknowledgements The work presented in this paper is the result of research sponsored by the Multidisciplinary Center for Earthquake Engineering Research, Buffalo, NY; the Institute for Civil Infrastructure Systems, New York, NY; and the National Science Foundation. The authors gratefully acknowledge the assistance of the Los Angeles Department of Water and Power, M. Shinozuka, C. Chang, J. P. Bardet, and S. Toprak in acquiring information presented herein. References American Water Works Association (AWWA), “Steel Pipe Design and Installation”, AWWA Manual M11, American Water Works Association Inc., New York, NY, 1964. Bardet, J. P., N. Mace, T. Tobita, and J. Hu, “Large-Scale Modeling of Liquefaction-Induced Ground Deformation Part I: A Four-Parameter MLR Model”, Technical Report MCEER-990019, Multidisciplinary Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, 1999a, 155-173. Bardet, J. P., T. Tobita, N. Mace, and J. Hu, “Large-Scale Modeling of Liquefaction-Induced Ground Deformation Part II: MLR Model Applications and Probabilistic Model”, Technical Report MCEER-99-0019, Multidisciplinary Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, 1999b, 175-190. Bartlett, S. F. and T. L. Youd, “Empirical Prediction of Liquefaction-Induced Lateral Spread”, Journal of Geotechnical Engineering, ASCE, Vol. 121, No. 4, 1995, 316-329. Boore, D. M., W. B. Joyner, and T. E. Fumal, “Equations for Estimating Horizontal Response Spectra and Peak Acceleration from Western North American Earthquakes: A Summary of Recent Work”, Seismological Research Letters, 68, No. 1, 1997, 154-179. Burrough, P. A. and A. McDonnell, Principles of Geographical Information Systems, Oxford University Press, Oxford, U.K., 1998. Campbell, K. C., “Empirical Near-Source Attenuation Relationships for Horizontal and Vertical Components of Peak Ground Acceleration, Peak Ground Velocity, and Pseudo-absolute Acceleration Response Spectra”, Seismological Research Letters, 68, No. 1, 1997, 154-179. Chang, S. E., A. Rose, M. Shinozuka, W. D. Svekla, and K. J. Tierney, “Modeling Earhquake Impact on Urban Lifeline Systems: Advances and Integration”, Research Progress and Accomplishments: 1999-2000, Multidisciplinary Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, 2000a. Chang, S. E., H. A. Seligson, and R. T. Eguchi, “Estimation of the Economic Impact of Multiple Lifeline Disruption: Memphis Light, Gas and Water Division Case Study”, Technical Report NCEER-96-0011, National Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, 1996. Chang, S. E., M. Shinozuka, and J. E. Moore II, “Probabilistic Earthquake Scenarios: Extending Risk Methodologies to Spatially Distributed System”, Earthquake Spectra, 2000b, in press. Cluff, L. and Pecker A., “Foreword”, Proceedings of the 5th International Conference on Seismic Zonation, Nice, France, Earthquake Engineering Research Institute, Vol. I, October, 1995. Earthquake Engineering Research Institute, Proceedings of the 4th International Conference on Seismic Zonation, Stanford, CA, August 1991. Earthquake Engineering Research Institute, Proceedings of the 5th International Conference on Seismic Zonation, Nice, France, October 1995. Earthquake Engineering Research Institute, “Special Theme Issue on Earthquake Loss Estimation”, Earthquake Spectra, Vol. 13, No. 4, 1997. Eguchi, R. T., “Earthquake Performance of Water Supply Components During the 1971 San Fernando Earthquake”, Technical Report No. 82-1396-2a, J. H. Wiggins Company, Redondo Beach, CA, March, 1982. Eguchi, R. T., C. K. Huyck, B. Houshmand, B. Mansouri, M. Shinozuka, and G. Webb, “The Mormara, Turkey Earthquake: Using Advanced Technology to Conduct Earthquake Reconnaissance”, Research Progress and Accomplishments: 1999-2000, Multidisciplinary Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, 2000, 81-96. Eguchi, R. T., J. D. Goltz, H. A. Seligson, P. J. Flores, N. C. Blais, T. H. Heaton, and E. Bortugno, “Real-Time Loss Estimation as an Emergency Response Decision Support System: The Early Post-Earthquake Damage Assessment Tool (EPEDAT)”, Earthquake Spectra, Vol. 13, No. 4, 1997, 815-832. Eidinger, J. M., “Water-Distribution System”, The Loma Prieta, California, Earthquake of October 17, 1989 – Lifelines, Schiff, A. J., Ed., USGS Survey Professional Paper 1552-A, U.S. Department of Interior, Washington, D.C., 1998. Finn, W. D. Liam, “Geotechnical Engineering Aspects of Microzonation”, Proceedings of the 4th International Conference on Seismic Zonation, Stanford, CA, Earthquake Engineering Research Institute, Vol. I, August 1991, 199-259. Hamada, M., S. Yasuda, R. Isoyama, and K. Emoto, Study on Liquefaction Induced Permanent Ground Displacement. Association for the Development of Earthquake Prediction, Tokyo, Japan, November 1986. Hamada, M. and K. Wakamatsu, Liquefaction, Ground Deformation and Their Caused Damage to Structures. The 1995 Hyogoken-Nanbu Earthquake – Investigation into Damage to Civil Engineering Structures, Committee of Earthquake Engineering, Japan Society of Civil Engineering, June 1996. Hamada, M. and T. D. O’Rourke, Eds., “Case Studies of Liquefaction and Lifeline Performance During Past Earthquakes”, Technical Report NCEER-92-0001, National Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, February 1992. Holzer, T. L., M. J. Bennett, D. J. Ponti, and J. C. Tinsley III, “Liquefaction and Soil Failure During 1994 Northridge Earthquake”, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 125, No. 6, 1999, 438-452. Housner, G. W., “Spectrum Intensity of Strong-Motion Earthquakes”, Proceedings, Symposium on Earthquakes and Blast effects on Structures, Earthquake Engineering Research Institute, Oakland, CA, June 1952, 20-36. Isenberg, J., “Role of Corrosion in Water Pipeline Performance in Three US Earthquakes”, Proceedings of the 2nd US National Conference on Earthquake Engineering, Stanford, CA, August 1979, 683-692. Ishikawa, Y. and H. Kameda, “Scenario Earthquakes vs. Probabilistic Seismic Hazard, in Schueller”, Shinozuka and Yao, Eds., Structural Safety & Reliability, Balkema, Rotterdam, 1994, 2139-2146. Japan Gas Association, Recommended Practices for Earthquake Resistant Design of Gas Pipelines, Published in Earthquake Resistant Design Codes in Japan, Japan Society of Civil Engineering, Tokyo, Japan, January 2000. Japan Water Works Association, Report on Damage to Water Distribution Pipes Caused by the 1995 Hyogoken-Nanbu Earthquake, 1996, in Japanese. Kakumoto, S., M. Hatayama, and H. Kameda, “Advanced GIS Applications for Earthquake Disaster Mitigation”, Proceedings, Workshop on Mitigation of Earthquake Disaster by Advanced Technology, Multidisciplinary Center for Earthquake Engineering Research, Los Angeles, CA, March 2000, 174-179. Katayama, T. and R. Isoyama, “Damage to Buried Distribution Pipelines During the Miyagi-ken oki Earthquake”, PVP-43, ASME, New York, NY, 1980, 97-104. Malezewski, J., GIS and Multicriteria Decision Analysis, John Wiley & Sons, New York, NY, 1999. Marble, D. F., H. W. Calkins, and D. J. Peuquet, Basic Readings in Geographic Information Systems, SPAD Systems Ltd., Williamsville, NY, 1984. National Institute of Building Sciences, Earthquake Loss Estimation Methodology HAZUS 97: Technical Manual, prepared for Federal Emergency Management Agency, Washington, D.C., 1997. National Research Council, The Impacts of Natural Disasters: A Framework for Loss Estimation, National Academy Press, Washington, D.C., 1999. O’Rourke, T. D. “An Overview of Geotechnical and Lifeline Earthquake Engineering”, ASCE Geotechnical Special Publication No. 75, Pakoulis, P., M. Yegian, and D. Holtz, Eds., Reston, VA, Vol. II, 1998, 1392-1426. O’Rourke, T. D. and J. W. Pease, “Mapping Liquefiable Layer Thickness for Seismic Hazard Assessment”, Journal of Geotechnical Engineering, ASCE, Vol. 123, No. 1, 1997, 46-56. O’Rourke, T. D. and M. Hamada, Eds., “Case Studies of Liquefaction and Lifeline Performance During Past Earthquakes”, Technical Report NCEER-92-0002, National Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, 1992. O’Rourke, T. D. and M. D. Grigoriu, and M. M. Khater, “Seismic Response of Buried Pipes”, Pressure Vessel and Piping Technology – A Decade of Progress, ASME, Sundarajan, C. Ed., 1985, 281-323. O’Rourke, T.D. and S. Toprak, “GIS Assessment of Water Supply Damage from the Northridge Earthquake”, Spatial Analysis in Soil Dynamics and Earthquake Engineering, Frost, D. Ed., ASCE, 1997. O’Rourke, T.D., S. Toprak, and Y. Sano, “Factors Affecting Water Supply Damage Caused by the Northridge Earthquake”, Proceedings of the 6th US National Conference on Earthquake Engineering, Seattle, WA, June 1998, 1-12. O’Rourke, T. D. and S.-S. Jeon, “Factors Affecting the Earthquake Damage of Water Distribution Systems”, Proceedings of the 5th US Conference on Lifeline Earthquake Engineering, Seattle, WA, ASCE, Reston, VA, August 1999, 379-388. Pease, J. W. and T. D. O’Rourke, “Seismic Response of Liquefaction Sites”, Journal of Geotechnical Engineering, ASCE, Vol. 123, No. 1, 1997, 37-45. Ripley, B. D., Spatial Statistics, John Wiley & Sons, Inc., New York, NY, 1987. Rose, A., J. Benavides, S. Chang, P. Szczesniak, and D. Lim, “The Regional Economic Impact of an Earthquake: Direct and Indirect Effects of Electricity Lifeline Disruptions”, Journal of Regional Science, Vol. 37, 1997, 437-458. Sano, Y., T. D. O’Rourke, and M. Hamada, “Geographic Information System (GIS) Evaluation of Northridge Earthquake Ground Deformation and Wager Supply Damage”, Proceedings of the 5th US Conference on Lifeline Earthquake Engineering, Seattle, WA, ASCE, Reston, VA, August 1999, 832-839. Shimizu, Y., “Advanced Accelerometers and Liquefaction Meters for Damage Assessment of Gas Distribution Systems”, Proceedings, Workshop on Mitigation of Earthquake Disaster by Advanced Technology, Multidisciplinary Center for Earthquake Engineering Research, Los Angeles, CA, March 2000, 46-54. Shimizu, Y., S. Negishi, Y. Yoshikawa, and H. Kano, “SIGNAL – Real Time Pipe Damage Estimation System After Earthquake”, Technical Report NCEER-96-0012, Hamada, M. and T. D. O’Rourke, Eds., National Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, September 1996, 679-684. Shinozuka, M., A. Rose, and R. T. Eguchi, Eds., “Engineering and Socioeconomic Impacts of Earthquakes”, Monograph Series 2, Multidisciplinary Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, 1998. Shirozu, T., S. Yune, R. Isoyama, and T. Iwamoto, “Report on Damage to Water Distribution Pipes Caused by the 1995 Hyogo-ken-Nanbu (Kobe) Earthquake”, Technical Report NCEER-960012, National Center for Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, 1996, 93-110. Shoven, J. B. and J. Whalley, Applying General Equilibrium, Cambridge University Press, Cambridge, U.K., 1992. Silva, W., Personal Communication, Pacific Engineering Analysis, El Cerrito, CA, July 2000. Snedecor, G. W. and W. G. Cochran, Statistical Methods, Seventh Edition, Iowa State University Press, Ames, IW, 1980. Toprak, S., T. D. O’Rourke, and I. Tutuncu, “Geographic Information System (GIS) Characterization of Spatially-Distributed Lifeline Damage”, Proceedings of the 5th US Conference on Lifeline Earthquake Engineering, Seattle, WA, ASCE, Reston, VA, August 1999, 110-119. Uy, C. O. and T. D. O’Rourke, “Advanced GIS Applications for Civil Infrastructure Systems”, National Forum on Advanced GIS Applications and Database Needs for Civil Infrastructure Systems, Institute for Civil Infrastructure Systems, New York, NY, 2000, in press. Whitman, R. V., T. Anagnos, C. A. Kircher, H. J. Lagorio, R. S. Lawson, and P. Schneider, “Development of a National Earthquake Loss Estimation Methodology”, Earthquake Spectra, Vol. 13, No. 4, 1997, 643-661. Youd, T. L., “Liquefaction, Flow, and Associated Ground Failure”, U.S. Geological Survey Circular 688, US Department of the Interior, Washington, D.C., 1973. Youd, T. L., “Mapping of Earthquake-Induced Liquefaction for Seismic Zonation”, Proceedings of the 4th International Conference on Seismic Zonation, Earthquake Engineering Research Institute, Stanford, CA, Vol. I, August 25-29, 1991, 111-138. Youd, T. L. and D. K. Keefer, “Liquefaction During the 1977 San Juan Province, Argentina Earthquake (Ms = 7.4)”, Engineering Geology, Vol. 37, No. 3, 1994, 211-233. Youd, T. L. and M. Perkins, “Mapping Liquefaction-Induced Ground Failure Potential”, Journal of Geotechnical Engineering, ASCE, Vol. 104, No. 4, 1978, 433-446. Youd, T. L. and M. Perkins, “Mapping Liquefaction Severity Index”, Journal of Geotechnical Engineering, ASCE, Vol. 113, No. 11, 1987, 1374-1392.