Group B, Poster #236, Ground Motions (GM)

Increasing the spatial resolution in physics-based site term estimates: results from southern San Andreas ruptures

John Rekoske, Alice-Agnes Gabriel, Dave A. May, & Scott Callaghan
Poster Image: 

Poster Presentation

2025 SCEC Annual Meeting, Poster #236, SCEC Contribution #14592 VIEW PDF
Recent advances in Probabilistic Seismic Hazard Analysis (PSHA) leverage physics-based simulations to estimate seismic hazard, such as the CyberShake project. However, computational costs quickly escalate when performing PSHA analysis for numerous faults and sites and can become prohibitively expensive. To reduce computational demands, hazard maps commonly interpolate ground shaking intensities from simulations conducted at fewer locations, but the accuracy of these interpolations remains poorly quantified. To address these questions, we generate a set of 100 earthquake point source simulations distributed along approximately 630 km of the non-planar UCERF3 fault geometry for the Southern Sa...n Andreas Fault (SSAF) extending from the Coachella section in the south to the Parkfield section in the north. Using SeisSol, we simulate 300 seconds of viscoelastic wave propagation for these sources and use a parallel HDF5 file format to store approximately 30 TB of three-component velocity seismogram data. From these data, we derive frequency-dependent site terms for 335 sites within Southern California, corresponding to the CyberShake study 22.12 locations. We use the same interpolation as CyberShake, and compare interpolated site terms against our high resolution site terms. We identify local discrepancies for our set of earthquake sources with peak ground velocities differing by up to a factor of approximately three. We identify areas where unexpectedly high or low ground motions could be missed when using the interpolated dataset to create a hazard map. To generalize our results across broader finite source scenarios for the SSAF, we will develop an accurate reduced-order modeling technique, a scientific machine learning approach based on the interpolated proper orthogonal decomposition, to rapidly compute Green’s functions which could be used within CyberShake. Our analysis of physics-based site terms from ensembles of SSAF rupture scenarios highlights the benefits of high-resolution seismic hazard estimates and guides future developments combining high-performance computing and machine learning reduced-order modeling techniques for PSHA.
SHOW MORE

Citation: Rekoske, J., Gabriel, A., May, D. A., & Callaghan, S. (2025, 09). Increasing the spatial resolution in physics-based site term estimates: results from southern San Andreas ruptures. Poster Presentation at 2025 SCEC Annual Meeting.
Keywords: reduced-order model, cybershake, seismic hazard, machine learning
SCEC Award: 25259, 24103
Relevant SCEC Themes:
Advanced Modeling Frameworks
Research Computing & Cyberinfrastructure