Abstract

CyberShake is a high-performance computational platform developed by the Southern California Earthquake Center (SCEC) to produce seismic hazard models from large suites of earthquake simulations. Code optimization and the development of workflow tools on the NCSA Blue Waters and OLCF Titan supercomputers have substantially reduced the computational costs. A recent workflow efficiently utilized both machines to generate 285 million two-component seismograms for the central California region from which 46 billion intensity measurements were extracted. We summarize the current suite of CyberShake models and apply averagingbased factorization (ABF) to decompose each model into a hierarchy of site, path, directivity, stress-drop, and source complexity effects. We summarize how the strength of the directivity effect decreases with increasing source complexity, and how the site and path effects vary with differences in crustal structure. Comparisons quantified by the ABF variances indicate that simulation-based hazard models can potentially reduce, the total unexplained variability in current ground-motion prediction equations, by as much as one-third.

Date

January 1, 1970

Authors

Thomas H Jordan, Scott Callaghan

Journal

Proceedings of the US national conference on earthquake engineering