Michal J. Orlikowski
July 30, 2001
When examining the seismic risk for a region, any analysis must focus both on the direct and indirect potential losses caused by an earthquake. “Direct losses” refer to casualties and property damages caused by the initial tremor, while “indirect losses” refer to the post-event losses in productivity and functionality due to the direct losses. The objective of my project while participating in this summer’s MCEER sponsored REU program is to focus on indirect losses caused by damages to highway network systems. As transportation systems are critical to the function of countless societal institutions, the question of their post earthquake performance is very important. This project is developed, under Professor Masanobu Shinozuka’s direction, as an extension of Nobuhiko Shiraki’s Master’s Thesis on the subject, Performance of Highway Network Systems Under Seismically Induced Traffic Delays. Taking Los Angeles and Orange Counties of the Los Angeles Metropolitan Area as the study region, methods of developing system damage and performance indices are examined, as are methods of modeling the repair progress in the region.
In order to approach this analysis, a model of approximating the damage caused to a highway network system is developed. Simplifying the problem somewhat, it is taken that bridges are the primary highway components at risk for damage due to ground motion. Thanks to the extensive data collected on highway bridge damage following the 1994 Northridge Earthquake, empirical fragility curves for CalTrans bridges have been constructed by Shinozuka et al. (2000). These curves represent four discrete damage states: minor, moderate, major, and collapse. (The state of “no damage” also exists, but is not represented with a curve). Given these fragility curves, it is possible, given a study region with network model and scenario earthquake with PGA distribution, to implement a Monte Carlo simulation of bridge damage, and thus, network damage. The region of interest in this study, as mentioned before, consists of Los Angeles and Orange Counties. The total number of bridges in the area is 2,727, and the highway network has 185 links connecting 118 nodes. Concerning scenario earthquakes, a comprehensive database is used; it consists of 47 scenario earthquakes, 13 of which are probabilistic, and 34 of which are deterministic. (These scenarios are developed from studies by the California Department of Conservation, Division of Mines and Geology, and translated to a useable form by Mr. Nobuhiko Shiraki). Using the given fragility curves, study region, and earthquake scenarios, indices of bridge damage and link damage can be calculated following the method in Mr. Shiraki’s thesis.
Furthermore, as is outlined in Mr. Shiraki’s work, by utilizing origin-destination data of commuters on the LA highway system, (most recently collected by survey, in 1991), a more convenient, though simplistic index can be calculated to estimate the network’s performance: driver’s delay. The driver’s delay is defined as the difference between the total daily commute time for all network travelers on the damaged system, and that for those on the intact system. Though simplistic, this seems to be a fairly adequate measure of system performance. Notably, as this index is based on daily origin-destination data, it displays a driver’s delay in hours of delay per day of travel. In order to develop a total driver’s delay, one must examine the period of time over which this driver’s delay persists. Repair efforts, however, cause the driver’s delay to decrease with time. Thus, to have a better understanding of the system’s post earthquake performance, a time-based analysis must be performed, modeling the region’s highway repair effort. For the time being, “reasonable” probabilistic models for repair progress are being implemented to simulate the post-earthquake network recovery. I say “reasonable” because sufficient data about the repair methods used by CalTrans is unavailable to develop a more accurate portrayal. An attempt is currently being made to remedy this situation. Regardless, the use of these models still provides some insight into the performance of the network.
Currently, the simulations being used present a fairly adequate model of earthquake related network damage, and the post event recovery. It is important to remember that the analysis regarded here is only a simulation, similar to HAZUS or REDARS. The practical applications of this work lie in the same realm. Further, additional data must be collected on the implementation of repair processes before this approach could possibly be considered for serious use in simulating the losses caused by seismic events.