Pushing the Limits of Regional-Scale Fully Deterministic Large Earthquake Ground Motion Simulations on High-Performance Computers with Three-Dimensional Earth Structure and Topography: Hayward Fault Scenarios and Generic Ruptures in Simple Models

Date: 4/25/2019

Time: 11:00 AM

Room: Elliott Bay

Numerical simulations of earthquake ground motions can account for region-, path- and site-specific three-dimensional (3D) earth structure and fault geometry and complement limited empirical data for large events. Advances in numerical methods, improvements in rupture models and 3D earth structure and the inexorable growth of computational power enable higher resolution earthquake ground motion simulations. We are modeling ground motions to frequencies of 5 Hz and higher from large earthquakes (moment magnitude M_W 6.5-7.2) on regional scales (~100 km). Simulations rely on world-class high-performance computing at DOE National Labs, including GPU-accelerated platforms. Simulations include generic faults with simple basin structures and stochastic heterogeneity and site-specific simulations of large ruptures on the Hayward Fault in the San Francisco Bay Area (SFBA). Simulations rely on the SW4 finite difference code with rupture models from Graves and Pitarka (2016) and a 3D geologic/seismic model from the United States Geological Survey (USGS) including topography. We have shown that simulated motions are consistent with ground motion models, such as those from the PEER NGA-West2 project (Bozorgnia, et al., 2014). In the SFBA, we demonstrate how path- and site-effects in the 3D model bias intensity values and propose a method to account for these epistemic effects in a non-ergodic ground motion model. Within the SFBA, we have shown how the assumed minimum shear wavespeed in the near-surface geotechnical layer can impact the response. For the Hayward Fault, 3D geometry of the fault plane plays a role in controlling near-fault motions, akin to hanging wall effects for normal and thrust ruptures. Generic ruptures in simple basin models allow us to study near-fault motions which are particularly sensitive to the distribution of slip on the fault, the depth to the top of the rupture and directivity. We show how velocity pulses and displacement steps result in large near-fault motions and intensities can vary by a factor 10 close to the fault.

Presenting Author: Arthur J. Rodgers

Authors