Development of a Data-driven Near-surface Velocity Models for the San Francisco Bay Area: Stationary and Spatially Varying Approaches
Description:
In this presentation, we introduce two new sedimentary velocity models for the San Francisco Bay Area (SFBA) to improve near-surface shear-wave velocity (VS) representation in large-scale broadband numerical simulations. The goal is to enhance the accuracy of sedimentary layer characterization in the Bay Area community velocity model. The first model uses only the time-averaged shear-wave velocity of the top 30 m (VS30), while the second model applies location-specific adjustments to account for spatial variability. A dataset of 200 measured VS profiles supports the development of both models, which are formulated within a hierarchical Bayesian framework to ensure robust scaling.
The spatially varying model incorporates a slope adjustment term, modeled as a Gaussian process, to capture site-specific effects. Residual analysis shows that both models remain unbiased for VS values up to 1000 m/s. Along-depth variability models, derived from within-profile residuals, further refine the VS predictions. Compared to the USGS Bay Area community velocity model, the proposed models predict up to twice as high VS values in areas such as San Jose and the Livermore Valley.
Goodness-of-fit (GOF) comparisons, based on one-dimensional linear site-response analyses at selected sites, reveal that the proposed models outperform the USGS model by better capturing near-surface amplification across a broad frequency range. Incorporating along-depth variability further improves GOF scores by reducing over-amplification at high frequencies.
These results emphasize the importance of integrating data-driven shallow crust models, like those presented here, into regional community velocity models to enhance seismic hazard assessments in the SFBA.
Session: Accuracy and Variability of Physics-based Ground Motion Modeling - I
Type: Oral
Date: 4/15/2025
Presentation Time: 02:15 PM (local time)
Presenting Author: Grigorios
Student Presenter: No
Invited Presentation:
Poster Number:
Authors
Grigorios Lavrentiadis Presenting Author Corresponding Author glavrent@caltech.edu California Institute of Technology |
Elnaz Seylabi elnaze@unr.edu University of Nevada, Reno |
Feiruo Xia fxia@caltech.edu California Institute of Technology |
Hesam Tehrani hsalmanitehrani@unr.edu University of Nevada, Reno |
Domniki Asimaki domniki@caltech.edu California Institute of Technology |
David McCallen dbmccallen@lbl.gov Lawrence Berkeley National Laboratory |
|
|
|
Development of a Data-driven Near-surface Velocity Models for the San Francisco Bay Area: Stationary and Spatially Varying Approaches
Session
Accuracy and Variability of Physics-based Ground Motion Modeling