Evaluation of the USGS 3D Seismic Model of the San Francisco Bay Area From Moderate Earthquake Waveform Modeling

Date: 4/26/2019

Time: 09:15 AM

Room: Vashon

The United States Geological Survey (USGS) three-dimensional (3D) model of the geology and seismic material properties in San Francisco Bay Area (SFBA) has been used for ground motion simulations of past large and moderate earthquakes as well as possible damaging scenario events (including simulations to 5 Hz and higher). Simulated intensities reveal significant variability caused by known geologic structure. In order to evaluate the USGS 3D model, we are modeling waveforms from recent moderate earthquakes. These events are appealing because they are well recorded by permanent seismic stations in the region, especially as station coverage has improved in the last few years, and permit sampling diverse propagation paths. We use routine source parameters: double-difference locations (Waldhauser, 2009) and point double couple focal mechanisms or moment tensors (Dreger and Romanowicz, 1994). We evaluate source parameters by modeling waveforms at the closest stations (within 10 km). This helps identify candidate events and problems with source parameters. We have modeled a few well-characterized events for paths sampling the broader SFBA to investigate path-dependent performance of simulated motions using the USGS 3D model and an average plane-layered (one-dimensional, 1D) model of the region. Motions are simulated with the SW4 finite difference code resolving frequencies up to 5 Hz and including surface topography with the 3D model. We have found that 3D effects are clearly seen, especially for longer paths (> 30 km) even at relatively low-frequencies (< 0.5 Hz). Matching waveforms is difficult for frequencies above about 0.5 Hz because of large phase errors that complicate waveform alignment. We are quantifying goodness-of-fit using several metrics including delay times, correlation coefficient, duration, peak motions and spectral measurements. Maps of path-specific goodness-of-fit can inform which areas in the model need improvement.

Presenting Author: Arthur J. Rodgers

Authors