Validation of Regression-Based Event Selection for Seismic Risk Assessment of Distributed Levee System in California

We implement a regression-based event selection method presented in a companion paper to more efficiently assess seismic risk for a spatially distributed levee system surrounding Bacon Island in the Sacramento-San Joaquin river Delta region of California. We select a reduced earthquake event set consisting of only 44 events (24 events from mapped faults and 20 events from point sources) based on the regression-based event selection method from more than 400,000 events present in the Uniform California earthquake rupture forecast, version 3 (UCERF3) model (within 300 km of Bacon Island). To validate the effectiveness the method, we first compare hazard curves produced using all UCERF3 events and the reduced-event set. The similar hazard curves produced by the two sources confirm that, for the present application, the proposed methodology can capture the overall seismic hazard well. We then compute the probability of failure of the Bacon Island levee system applying a method based on level-crossing statistics using both all UCERF3 events and the reduced-event set. The comparable system risk results of two approaches further demonstrate the effectiveness of the event selection method.