HR: 17:30h
AN: U22C-07    [PDF]
TI: Estimating System Impact of Earthquakes on a Major Metropolitan Roadway Network
AU: * Perkins, D M
EM: perkins@usgs.gov
AF: U.S. Geological Survey, MS 966 Denver Federal Center Box 25046, Denver, CO 80225 United States
AU: Taylor, C E
EM: cetaylor@earthlink.net
AF: President, Natural Hazards Management Inc, 5402 Via Del Valle, Torrance, CA 90505 United States
AU: Werner, S D
EM: sdwerner@ix.netcom.com
AF: President, Seismic Systems & Engineering Consultants, 8601 Skyline Blvd., Oakland, CA 94611 United States
AB: The impact of an earthquake on the Memphis, TN, roadway system has been estimated using a prototype computer program, REDARS (Risks from Earthquake DAmage to Roadway Systems). For scenario earthquakes, the program computes ground motions and ground deformations at bridges and other components throughout the system. Then, estimates of costs and times to repair this damage, together with the component's ability to accommodate traffic flows during repairs, are estimated, and these are used to establish modified post-earthquake system states at various times after the earthquake. Transportation network analysis procedures are then applied to each system state, in order to estimate how post-earthquake traffic flows and travel times are affected by these various roadway closures. Consequences of this damage to the roadway system, in terms of economic losses, reduced access to key locations in the regions (e.g., hospitals, airports, etc.) are then estimated. Uncertainties are incorporated throughout. Using Monte Carlo simulation in earthquake occurrence and modeled uncertainties, the program was used to produce a 50,000-year history of annual estimated drive-time loss in dollars, with nearly 800 non-zero loss years. The ordered list of annual losses makes up an empirical annual-rate loss distribution function. From this list, likelihood functions for the 500-yr and 2500-yr losses can be obtained, yielding most likely values of about 300 million dollars and 600 million, respectively, with ranges of uncertainty (10th to 90th percentile) of about 10 and 20 percent. The ordered list of non-zero losses is nearly exponentially distributed. This simple structure permits estimation, through the employment of bootstrapping with variance-reduction techniques, of the average conditional loss around several tens of times more precise than the average of the simulated losses. The model for earthquake magnitude, locations, and occurrence frequency has been taken from the US Geological Survey sources used for national hazard mapping in the central United States.
DE: 6304 Benefit-cost analysis
DE: 6309 Decision making under uncertainty
DE: 6334 Regional planning
DE: 6344 System operation and management
DE: 7223 Seismic hazard assessment and prediction
SC: U
MN: 2003 Fall Meeting