Oral Presentation

Community Velocity Models (CVMs) for different regions in California exist at different scales and levels of complexity. These models are used in various studies that lead to improved seismic hazard assessment. Extending the CVMs outside their original geographic domain presents opportunities to enhance their development and representation. Harvesting models from local to regional scales and capturing meaningful complexity will ensure the next generation of CVMs is widely used. In Southern California, research has shown that the two popular regional scale CVMs integral to ground motion prediction studies and seismic hazard estimates fit low-frequency seismograms reasonably well. However, for higher frequencies, most important to structural engineers, they underperform. Improving the CVMs through the traditional method of full-waveform inversion remains computationally prohibitive due to their large spatial scale. At the same time, dense seismic arrays have made it possible to produce better images of active fault zones, high-resolution attenuation models of the shallow crust, and high-resolution velocity models of sedimentary basins near the southern San Andreas fault. This advancement in instrumentation has triggered the development of UCVM to produce hybrid models comprising the CVMs and high-resolution basin models seamlessly stitched together. Model hybridization can be performed using models of arbitrary shape and has two main advantages: 1. the exploitation of the fine spatial resolution of local models in ground motion predictions, and 2. a more comprehensive evaluation of the parent CVMs since the same local model improvements can be applied to any CVM. Therefore, the influence of other hyperparameters in model representation, including attenuation, anisotropy, surface topography, and geotechnical modifications, on ground motion predictions can be carefully evaluated to give insight into model performance.