Events Calendar

Forecasting Earthquake Ground Motions Using Large-Scale Numerical Simulations

Speaker: Thomas H. Jordan, Professor, Department of Earth Sciences, USC, Director, Southern California Earthquake Center

Talk Title: Forecasting Earthquake Ground Motions Using Large-Scale Numerical Simulations

Abstract: The seismic radiation from complex fault ruptures in 3D crustal structures can now be numerically simulated for the largest earthquakes at frequencies of engineering interest. An interdisciplinary team of SCEC scientists recently achieved a milestone dubbed M8, the dynamic-rupture simulation of a magnitude-8, wall-to-wall earthquake on southern San Andreas fault up to seismic frequencies of 2-Hz. M8 was calculated on a computational grid of 436 billion elements; the production run sustained 220 teraflops for 24 hours on 223K cores of the NCCS Jaguar supercomputer. I will describe SCEC efforts to use such simulations in forecasting strong ground motions in Southern California, focusing on the CyberShake computational platform. In the CyberShake 1.0 hazard model for the Los Angeles region, about 440,000 earthquake simulations have been used to represent the probabilistic seismic hazard up to 0.3 Hz. The hazard maps are substantially different from those derived from empirical ground-motion prediction equations. At the probability levels appropriate for long-term forecasting, these differences are most significant (and worrisome) in sedimentary basins, where the regional seismic risk is concentrated. The higher basin amplifications obtained by CyberShake are due to the strong coupling between rupture directivity and basin-mode excitation. The simulations show that this coupling is enhanced by the tectonic branching structure of the San Andreas system. Large-scale simulations are being used in several other applications: (a) operational earthquake forecasting, which provides short-term earthquake probabilities using statistical models of seismic triggering and clustering; (b) earthquake early warning, which attempts to predict imminent shaking during an on-going event; and (c) post-earthquake information, including high-resolution maps of seismic shaking intensities needed for emergency response immediately following a large earthquake. These applications offer new (and urgent) computational challenges, including requirements for robust, on-demand supercomputing and access to very large data sets.



Host: Prof. Jean-Pierre Bardet

Location: Kaprielian Hall (KAP) - 209

Audiences: Everyone Is Invited

Contact: Evangeline Reyes