Simulations of earthquakes in California with multiple velocity models

Abstract

Estimating seismic hazard using 3D earthquake simulations of moderate‐to‐large magnitude events is becoming a reality. In this study, we further this endeavor by simulating seismic wavefields for five earthquakes in California, United States, and evaluating their pseudospectral acceleration at periods of 2, 5, and 10 s. We compare results from the recently developed U.S. Geological Survey (USGS) National Crustal Model with three well‐established and validated regional community velocity models (CVMs): two CVMs for Southern California (CVMS4.26.M01 and CVMH15.1.0) and the USGS San Francisco Bay region 3D CVM. A minimum shear‐wave velocity of 500 m/s is imposed in all four of the velocity models. We use finite‐difference calculations to simulate the earthquakes and specify a grid and timestep capable of resolving periods down to 1 s. For our simulations, we find that for each earthquake, the available velocity models produce broadly similar amplitudes and patterns of earthquake ground motions. Use of the 3D velocity models generally perform better than ergodic ground‐motion models, and variability in the results between earthquakes may indicate a combination of factors including uncertainty in earthquake magnitude and source parameterization, signal‐to‐noise ratio in the observed time series, and the lack of fault damage zones, topography, and other features in the computational seismic velocity models. Of particular note, we find that for most cases, averaging the ground motions for all available velocity models outperforms the ground‐motion results from any single velocity model. Our results indicate that the 3D models considered in this study can be sufficiently independent and have enough predictive power that their combination could be used to improve estimates of earthquake ground motions.