Multi-scale Statewide California Seismic Velocity Models: Version 1
Te-Yang Yeh, Yehuda Ben-Zion, & Kim B. OlsenSubmitted August 30, 2026, SCEC Contribution #15290, 2026 SCEC Annual Meeting Poster #075
A high-quality multi-scale description of subsurface seismic velocities is foundational for a range of topics including advancing simulations of ground motion and consistent derivation of earthquake source parameters. In this study, we construct multi-scale statewide P and S velocity models for California by merging existing models with varying spatial extent and resolution. We use the following two-step framework: (1) development of an updated statewide background model targeting frequencies of 0.1-0.5 Hz followed by (2) integration of local high-resolution models validated currently at 0.3-1 Hz. In the first step, we consider three regional subdomains - Southern, Central, and Northern California - and merge the regional-scale models into the starting model CANVAS (Doody et al., 2023). Candidate models are incorporated sequentially using a cosine-taper windowing approach (Ajala and Persaud, 2021). For each iteration, we conduct 3D numerical simulations of ground motion generated by 20-25 earthquakes (Mw 4.1-4.7) per subdomain, using models before and after merging. The simulation results are compared to data using an error function measuring misfits from both waveform envelopes and Fourier amplitude spectra. These results are translated into smooth 2D spatial weighting functions (refinement weights), which preserve portions of candidate models that outperform the previous iteration. Validation simulations confirm the robustness and effectiveness of the results, which significantly improve the ground motion prediction efficacy in various areas. Building upon the updated statewide background model, we incorporate local high-resolution velocity models. For each local model, we evaluate both the cosine-taper merging strategy and a dictionary learning approach (Zhang and Ben-Zion, 2024) and select the model version that yields the better agreement with observations. Lastly, we implement a low-velocity taper (Ely et al., 2010) with regionally optimized taper depths to improve the shallow seismic structure. The derived models can be improved in future studies by repeating the process for higher frequencies. The established workflow provides a robust framework for the long-term maintenance and systematic incorporation of future regional and local velocity models, enabling MUSCAL to evolve as new datasets become available.
Citation
Yeh, T., Ben-Zion, Y., & Olsen, K. B. (2026, 08). Multi-scale Statewide California Seismic Velocity Models: Version 1. Poster Presentation at 2026 SCEC Annual Meeting.
Related Projects & Working Groups
Community Earth Models (CEM)
