PALEOSEISMIC DATA AND REGIONAL HAZARD MODELING IN CALIFORNIA

Paleoseismic records from individual sites have contributed, on a range of levels, to the enterprise of earthquake hazard estimation. Trenching is often used to confirm fault location and assess Holocene activity. We review some less well-known points at which paleoseismic studies bear on regional seismic hazard estimation, including the Uniform California Earthquake Rupture Forecast version 3 (UCERF3) and potential successors. In UCERF3, paleoseismic estimates of recurrence rates at over 30 sites were used to constrain rates of ground rupturing earthquakes. Using a Monte Carlo-based inversion, recurrence rates were fit, along with fault-slip-rate estimates and the assumption of a regional Gutenberg-Richter magnitude distribution, to arrive at a state-wide estimate of rupture rates on California faults. The final model has been difficult for geologists to evaluate because essentially all data they might use for testing were used as inputs to the model. We identify two cases where paleoseismic data can provide useful comparisons with the UCERF3 model. First, we consider joint probabilities among the paleoseismic site fits that were modeled independently by UCERF3. The joint probability suggests that UCERF3 - favors larger, less-frequent events, compared to the observed record. Second, the Mystic Lake site record on the San Jacinto fault was not available in time to contribute to UCERF3, and thus can be used as an independent test of modeled rates on the San Jacinto. In contrast with the Mystic Lake record, UCERF3 predicts that, by a factor of about 30, events of M7.4 or larger are more likely than smaller ground ruptures. Dynamic rupture models through the Claremont-Casa Loma step over, and the presence of nearby precarious rocks both make this earthquake distribution unlikely. We conclude with some ideas for the use of paleoseismic data in a potential UCERF4.