Comparison of Mitigation Alternatives for Water Distribution Pipelines Installed in Liquefiable Soils Donald Ballantyne 1 William Heubach 2 Abstract Water pipeline distribution system mitigation measures are being evaluated: pipeline replacement, automated pipeline control systems, and planned manual valve actuation response. Seattle Public Utility s distribution system is being used as model for this evaluation. The most vulnerable part of the distribution system is located in the highly liquefiable Duwamish River Valley. A decade ago, it was estimated that it would cost in excess of one billion dollars to replace all of the cast iron pipe founded in liquefiable soils in Seattle s system. This cost was too expensive particularly when considering the other financial demands confronting the utility such as water treatment and reservoir upgrades. Seattle has joined forces with other water utilities in seismically active parts of the world to fund research under the American Water Works Association Research Foundation to evaluate alternate, less expensive, means of mitigating pipeline damage to improve the post-earthquake functionality of the system. The concern is that many pipelines will break, draining water from storage reservoirs, and also affect areas without significant pipeline damage. Following the Kobe earthquake, as well as many other earthquakes, this scenario has resulted in loss of water for fire suppression. Damage in the 2001 Nisqually Earthquake, as well as similar events in 1965 and 1949 was limited to less than 40 failures. However, a Cascadia subduction or Seattle Fault event will be much more damaging, and Seattle wants to be ready. The general strategy is to quickly isolate the damaged sections of the distribution network, or to replace pipe in the network so it will not fail. For each of these options, scenarios are being developed estimating pipeline performance and the resulting network hydraulic performance. The estimated costs and expected losses are being developed for a range of mitigation options. 1 Donald Ballantyne, Senior Consultant, MMI Engineering, Federal Way, WA. dballantyne@mmi Engineering.com 2 William Heubach, Formerly Senior Civil Engineer, Seattle Public Utilities, Current Bellevue, WA Utilities Department
Introduction This paper presents a project to develop an estimate of the damage from a future earthquakes and develop a cost effective mitigation strategy to minimize service disruption.. The project focuses on performance of the water distribution pipeline system in Seattle, Washington. One of the most significant issues following an earthquake is fire. Fires can be ignited by electrical shorts and fueled by natural gas leaks. It is a critical time for water to be available for fire suppression. Water systems have performed poorly in past earthquakes. In the 1989 Loma Prieta, 1994 Northridge and 1995 Kobe earthquakes, pipeline failures have been the primary reason for loss of water system pressure. Regionally there were approximately 40 pipelines failures resulting from the 2001 magnitude 6.8 Nisqually Earthquake, about half of which of which occurred in the Seattle water system. In this moderate earthquake, loss of service was limited to about 40 customers. Similar damage occurred in Seattle region earthquakes in 1949 and 1965. Hazard and pipeline damage data from these three events was collected. In larger events such as the 1995 Kobe Earthquake, and the 1994 Northridge Earthquake where there were in excess of 1,000 failures each, the systems drained and could not provide adequate water for fire suppression. In the 1994 Northridge Earthquake, approximately two-thirds of the San Fernando Valley water system drained. In 1995 Kobe Earthquake, the entire system serving the urban area drained within six hours. Post-earthquake performance and pipeline damage was gathered obtained for these events to augment information from the Seattle earthquakes. Methods have been developed and applied estimating expected damage from selected earthquakes, such as to the Seattle water system (Ballantyne et al, 1991). These methods are dependent on an accurate mapping representation of geologic hazards. In the Seattle case, it appears that there may be subtle features that may have resulted in pipeline damage that are not evident in geologic hazard mapping. Even if we have an understanding of the expected level of pipeline damage, mitigation can be expensive. One mitigation approach is to replace all cast iron pipe buried in soils susceptible to liquefaction, with restrained joint ductile iron. However, such an approach is prohibitively expensive. Isolation of heavily damaged pipeline, alternate system control strategies, and development of emergency plans for pipeline repair offer a promising ways to mitigate damage at more reasonable costs than pipeline replacement. There are many issues that still need to be resolved such as isolation and control strategy development, including inadvertently isolating an area from needed fire-suppression water, and selection of appropriate hardware. 2
Seismicity and Ground Motions The Seattle region is vulnerable to earthquake from three different source zones (see Figure 1): A Benioff Intraplate Zone, a Benioff Interplate Zone and a shallow fault zone. The three significant historic earthquakes, 1949, magnitude 7.1 Olympia; 1965, magnitude 6.5 Seattle-Tacoma; 2001, magnitude 6.8 Nisqually, were deep (50 to 60 kilometers) Benioff Intraplate Zone events. These events occur directly below the Puget Sound region, can recur as often as every 25 to 35 years and may produce firm ground accelerations as high as 0.3g in epicentral areas. A Benioff Interplate Zone extends from the Northern California coast to the Southern British Columbia coast. These Cascadia interplate subduction earthquakes may exceed Magnitude 9.0. However, because the epicenter would be located further away from the Puget Sound than intraplate events, peak firm soil ground accelerations are not expected to exceed 0.3g in the Puget Sound region. However, strong ground motion from these events may last more than two minutes and will likely trigger extensive liquefaction. The last Cascadia Subduction event occurred in 1700, and has an average recurrence of 500 to 600 years. There are also a series of active shallow faults in the Puget Sound region. The Seattle Fault runs east-west through Seattle and passes within a kilometer of the southern edge of downtown Seattle. A Seattle Fault event, potentially of Magnitude 6.5 or 7.0, is thought to have occurred about 1,100 years ago with a recurrence interval as short as 700 years. A moderate Seattle Fault event is expected to produce PGA s on the order of two-thirds times gravity on firm soils. Project Approach The following steps were used to develop the mitigation strategies: 1. Map peak ground velocity and permanent ground acceleration during the 1949, 1965 and 2001 Puget Sound earthquakes. 2. Document and map the pipeline damage that occurred during these earthquakes. 3. Compare the peak ground velocity and permanent ground displacement maps with the pipeline damage maps to verify pipeline vulnerability assumptions. 4. Modify pipeline vulnerability models, as necessary to reflect the findings in Step 3. 5. Prepare GIS-Based maps that predict peak ground velocity and permanent ground displacement for design-level ground motions (in this case, 10% probability of exceedance in 50 years). 3
6. Estimate the number and location of pipe breaks from the design-level earthquake. 7. Use hydraulic models to evaluate the system effect of the pipeline damage. 8. Identify mitigation measures to reduce damage effects on system operation. Peak Ground Velocity and Permanent Ground Displacement Hazards Ground motion mapping from the 1949 and 1965 events is limited. We used Modified Mercalli Index (MMI) data to estimate ground motions for pipeline damage in those events. In the 2001 Nisqually Earthquake, better PGA mapping was available due to the much higher density of ground motion instruments. However, Peak Ground Velocity (PGV) mapping was still too course to be of much value. PGV is the preferred ground motion intensity parameter used to evaluate pipeline damage due to wave propagation. Liquefaction and associated lateral spreading and landslides are both considered to be forms of permanent ground deformation (PGD). PGD results in significantly higher pipeline unit failure rates than wave propagation, so it becomes an important parameter when evaluating pipeline damage. Detailed liquefaction potential maps were developed by Shannon & Wilson, under subcontract to the U.S. Geological Survey for the City of Seattle, which have subsequently been published by the USGS (Grant, et al., 1998). The maps are based on an extensive subsurface boring database and more quantified liquefaction analysis of the data in the database. The hazard mapping and studies identify a large portion of south Seattle (the Duwamish River Valley) as underlain by extensive deposits of sandy Holocene alluvium and hydraulically placed fill that have a high susceptibility to liquefaction (see Figure 1). Reports of ground-deformation related damages to pipelines are known for the 1949 and 1965 events (Chleborad and Schuster, 1998). Reconnaissance work by Shannon & Wilson personnel and others have identified areas of liquefaction and ground deformations as a result of the recent 2001 Nisqually Earthquake. Liquefaction and lateral spreading occurred primarily in hydraulic fill and Holocene alluvium in the Seattle area during all three events. The majority of reported instances of liquefaction and permanent ground deformations occurred in the filled tideflats in and around the port areas in Seattle. During the 2001 Nisqually earthquake, liquefaction and lateral spreading was again observed in the hydraulic fills in the port and filled tideflats area in Seattle and farther south in the Duwamish Valley, particularly in the vicinity of Boeing Field. This information was used to correlate the cause of pipeline damage for these three events. The City of Seattle and the US Geological survey collaborated to map areas susceptible to landslides. Historically, these landslides are activated during extended 4
periods of rainfall. However, they did not appear to be a significant cause of pipeline failures in the three historic Seattle earthquakes. It should be noted that 2001 was a particularly dry year that may have limited the number of landslides that occurred in the Nisqually Earthquake. Historic Pipeline Damage Pipeline damage for the three earthquakes was gathered for analysis (refer to Figure 1). As expected, many of the pipeline failures occurred in liquefaction zones along the Duwamish River. However, there were clusters of failures in other areas where soils are not liquefiable. Refer to locations A, B, and C on Figure 1. It is known that the topography and geology of the Seattle Basin amplify ground motions in some of the hilly areas with firm soils adjacent to liquefiable lowlands. iat this time it is unclear whether these clusters of failures occurred due to an area of one type of particularly vulnerable pipe, ground motion amplifications, a result of unidentified/unmapped geotechnically vulnerable deposits or a combination of these factors.. Pipeline Seismic Vulnerability Model Development Although mechanical pipeline and soil models may be practical for site-specific studies, it is not practical to use theoretical analysis techniques to assess the seismic vulnerability of large pipeline networks. Researchers have developed relationships between pipeline damage and various seismic and geotechnical parameters using empirical data. The most commonly used are relationships between shaking intensity (PGA or PGV) and failures per unit length, and permanent ground deformation (PGD) and failures per unit length. Eguchi developed damage relationships for shaking in the early 1980 s. Relationships for PGD were developed for application to the San Francisco water system after the Loma Prieta earthquake (Harding and Lawson, 1991). These relationships are sometimes used in GIS-based computer models that are used to estimate the post-earthquake damage state of water system following an earthquake. The accuracy of these models are all limited by relatively small data sets, accuracy of the data, data interpretation by the researchers, and the inherent variability of pipeline condition and construction quality. Perhaps the best empirical data and pipe damage databases, are from the Northridge Earthquake and the Kobe Earthquake (T. O Rourke, et al, 1996; Shirozu, et al, 1996; Takada, et al, 1996). This data, and the damage relationships developed from this data were used to augment the information form the three Seattle events. 5
Figure 1. 1949 (triangles), 1965 (circles) and 2001 (squares) earthquake pipeline damage overlaid on liquefaction (vertical hatching) and landslide (horizontal hatching) susceptible areas. Pipeline failure clusters A, B and C are outside PGDsusceptible areas. 6
Peak Ground Velocity and Permanent Ground Displacement Maps Zipper-Zeman Associates prepared ground hazard maps for 10% probability of exceedance in 50 year ground motions. These maps were developed from supplementing USGS geologic hazard maps with soil boring data and from USGS ground shaking intensity maps. Peak ground velocity was computed by propagating the bedrock ground motions up through the soil column. Liquefaction-induced ground deformation was computed using the models developed by Bardet (Regional Modeling of Liquefaction-Induced Ground Deformation, Jean-Pierre Bardet, et al, Earthquake Spectra, February, 2002). Estimate Operational State in Hours Following Event Immediate post earthquake functionality of the water system is important to provide water for fire suppression. The operational state was modeled using a hydraulic network analytical tool, EPANET. The model results were supplemented with performance of water pipeline networks in previous earthquakes. The most relevant information was provided by representatives of the Los Angeles Department of Water and Power and the Kobe Water Department for the 1994 Northridge and 1995 Kobe earthquakes, respectively. They provided inside into the failures that limited system functionality, and the potential for use of the strategies considered in this project. For example, in the Northridge Earthquake about 2/3 of the San Fernando Valley (West Valley) was without water in the hours following the earthquake. This was not due to distribution pipe failures, but to transmission line failures primarily in and just downstream of the Van Norman Complex. Every pumped transmission line leaving the Complex had damage. The was compounded because the Los Angeles Aqueduct also went down. The trunk lines leaked but stayed in service. The result was problems with the pumped zones on the southern edge of the Valley, the opposite side from the source/treatment plant. Isolation of damaged sections of pipe would not have helped. Further, leaking water is critical for locating leaks. If the damaged lines are isolated (shut down), it would be nearly impossible to locate the leaks (interview with Marty Adams, LADWP). One strategy that worked in Kobe was to place seismic-actuated shutoff valves on storage tanks that operated in pairs. The strategy was to allow water for fire-fighting also save some water for post-earthquake use in the event of extensive pipeline failures. Although unvalved reservoirs did drain quickly after the earthquake, water in approximately 30 valved reservoirs was saved and used as a drinking water supply after the earthquake. 7
Mitigation Strategy The relative effectiveness of pipeline upgrade, isolation and control strategies, and emergency response to reduce the effect of pipeline failures on water system functionality is being examined. We are developing plans for: 1) pipeline replacement, 2) control valve implementation, and 3) emergency response. Specific hardware and strategies are being identified for typical water system configurations, and ranges of costs for the three alternatives provided. The pipeline replacement program would propose replacement for all vulnerable pipe (e.g. all cast iron pipe in liquefiable areas or at least increasing the replacement priority), and as an alternative, replacement of all backbone pipe in the same areas. Additional appurtenances may also be required if only the backbone was upgraded. Pipe replacement would minimize the recovery time. Another potential option may be to use more seismic-resistant pipe for new and replaced lines. This option could range from always using already available restrained joint ductile iron pipe to exploring the feasibility of importing the Japanese S-Joint which has performed extraordinarily well in earthquakes. The S-Joint permits both rotation and axial movement beyond the normal construction tolerances (see Figure 2). Rubber Gasket Lock Ring Maximum axial movement is 45 to 75 mm depending on diameter. Figure 2. Japanese S-II Joint (100-450 mm Diameter) permits both rotation and axial movement. Control valve implementation could include of a system to isolate vulnerable areas of the distribution system to keep the system from draining. The control valve system may be significantly less expensive to implement and would provide for water for fire suppression, but would still require a significant recovery period. The emergency response plan would require no capital improvements (stockpiling of spare parts and flexible hose that could be used to temporarily span broken mains is an option), and rely on manual operation of valves following the earthquake. Such a system may be too slow to effectively maintain system operation. The control valve system alternative would address immediate post-earthquake system operation, but would not help in reducing the overall system recovery time. The envisioned control valve system could make use of pressure reducing valves (PRVs) that feed water into the pressure zone. PRVs could be operated by installing a solenoid valve on the control loop. The decision to close could be based on an 8
earthquake ground motion threshold (e.g. 20% g), excess flow, rate of change of flow, etc. Each control valve could close automatically or could require interaction by an operator. Summary This paper discusses an ongoing Seattle Public Utilities project that is evaluating alternative methods for mitigation of earthquake damage to the pipeline distribution system. Empirical data from three historic earthquakes are used as input for the analysis. Mitigation alternatives include wholesale replacement of the pipe and/or incorporating seismic factors in pipeline replacement prioritization, automated system control to allow quick isolation of damaged areas in the system, use of more earthquake-resistant pipe joints such as the Japanese S-Joint, or application of an emergency plan where response staff would evaluate and isolate damaged system components as appropriate. The project is scheduled for completion by the end of July, 2003 Acknowledgements The authors thank the American Water Works Association Research Foundation for their interest and financial support of the project. We thank James Doane, David Lee, and Charles Pickel, members of the AWWARF Project Advisory Committee, for their direction. Further, the authors would like to express appreciation for the participating utilities that also provided financial support for the project: Seattle Public Utilities, Washington; City of Everett, Washington; Greater Vancouver Water District, British Columbia, Canada; Los Angeles Department of Water and Power, California; San Francisco Public Utilities, City of St. Louis, Missouri; Tacoma Public Utilities, Washington; and the Thames Water Company, Great Britain. References Ballantyne, D.B.; Taylor, C.; 1990; Earthquake Loss Estimation Modeling of the Seattle Water System, USGS Grant Award 14-08-0001-G1526, Kennedy/Jenks/Chilton Report No. 886005.00, Federal Way, Washington. Chleborad, A.F., and Schuster, R.L., 1998, Ground failure associated with the Puget Sound region earthquakes of April 13, 1949, and April 29, 1965, chap. earthquake hazards of Rogers, A.M., Walsh, T.J., Kockelman, W.J., and Priest, G.R., Assessing Earthquake Hazards and Reducing Risks in the Pacific Northwest: U.S. Geological Survey Professional Paper 1560, v. 2, p. 373-439. Grant, W.P., Perkins, W.J., and Youd, T.L., 1998, Evaluation of liquefaction potential in Seattle, Washington, chap. earthquake hazards of Rogers, A.M., Walsh, T.J., Kockelman, W.J., and Priest, G.R., Assessing Earthquake Hazards and Reducing 9
Risks in the Pacific Northwest: U.S. Geological Survey Professional Paper 1560, v. 2, p. 373-439. O Rourke, T.D, S. Toprak, and Y. Sano, Los Angeles Water Pipeline System Response to the 1994 Northridge Earthquake, Proceedings from the Sixth Japan- U.S. Workshop on Earthquake Resistant Design of Lifeline Facilities and Countermeasures Against Soil Liquefaction, edited by Hamada and T. O Rourke, National Center for Earthquake Engineering Research, Buffalo, N.Y., 1996 Shirozu, T., S. Yune, R. Isoyama, and T. Iwamoto, Report on the Damage to Water Distribution Pipes Caused by the 1995 Hyogo-ken-Nanbu (Kobe) Earthquake, Proceedings from the Sixth Japan-U.S. Workshop on Earthquake Resistant Design of Lifeline Facilities and Countermeasures Against Soil Liquefaction, edited by Hamada and T. O Rourke, National Center for Earthquake Engineering Research, Buffalo, N.Y., 1996 Takada, S., J. Ueno, and S. Goto, Damage Features of Buried Water Pipelines Related to Active Fault Geography During the 1995 Kobe Earthquake, Proceedings from the Sixth Japan-U.S. Workshop on Earthquake Resistant Design of Lifeline Facilities and Countermeasures Against Soil Liquefaction, edited by Hamada and T. O Rourke, National Center for Earthquake Engineering Research, Buffalo, N.Y., 1996. 10