Progress Report (1997)
Probabilistic Seismic Hazard Analysis of Southern California

Mehrdad Mahdyiar
VortexRock Consultants, Inc.

The primary objective of this study is to evaluate differenf regional seismicity models for southern California in terms of their ground motion hazards using different attenuation equations and taking into consideration the regional site conditions and basin effects information. Five different seismicity models for southern California are investigated: 1) The CDMG/USGS model, 2) the historic seismicity model; 3) fhe smooth historical seismicity model; 4) the strain rate model; and 5) a cascade earthquake model based on the CDMG/USGS fault

segments. Each model produces different regional ground motion hazard in southern California The probabilistic seismic hazard analysis (PSHA) is used to compare and evaluate these models against each other. The CDMG/USGS regional seismicity model is reconstructed based on the related information on their web-site. A detail report on the methodology and the results of the analyses was submitted to the Phase III group for publication. Following is a brief summary of the conclusions of this study.

The Inherent uncertainfies in the source models translate to large uncertainfies in seismic hazard values. The largest uncertainty, and the hardest to resolve, is the frequency of very large earthquakes, both on the major faults and in the regions between them. Allowing for very large earthquakes tends to reduce the ground motion hazards with typical return periods usedfor engineering design purposes. The seismicity models thaf balance a given moment budget are very sensitive to the choice of maximum magnitude of the seismic zones. Including large magnitude earthquakes, either by increasing the maximum magnitude or by cascading adjacent faults, will consume large amount of moment budget and would decrease the rate of moderate size earthquakes and thus the hazard.

Site classification matters, especially because models for attenuation and site effects are closely linked. The differences between available models are pronounced, sometimes exceeding a factor of 2, at very short periods and in Tertiary and Mesozoic rock, where site conditions are treafed differently in different models. Differences between models are less pronounced at longer periods and for sites on Quaternary sediments. Basement depth is also an importanf factor thaf is not well represented by surface soil conditions. Including basin depth using a very simple empirical correction can affect the predicted spectral acceleration by a factor of two in some cases.