PetaSHA3 Project

Geoinformatics: A Petascale Cyberfacility for Physics-Based Seismic Hazard Analysis (SCEC PetaSHA-3 Project) (NSF EAR - 0949443)

Project Dates: July 15, 2010 through June 30, 2010

SCEC PetaSHA-2 Research Groups

University of Southern California - Earth Sciences (Thomas H. Jordan)

University of Wyoming - Geophysics (Po Chen)

San Diego State University - Geological Sciences (Steve Day and Kim Olsen)

San Diego Supercomputer Center - (Yifeng Cui)

University of California San Diego - Geophysics (J. Bernard Minster)

Stanford University - Geophysics (Greg Beroza)

USC Information Sciences Institute - Computer Science (Ewa Deelman)

PetaSHA-3 Project Summary

Summary of a Proposal by the SCEC/CME Collaboration (T. H. Jordan, P.I.)

Earthquakes exemplify emergent phenomena that arise from nonlinear, multiscale interactions within complex natural systems. The proper use of system-level models to make valid scientific inferences about earthquakes in the real world requires an iterative process of model formulation and verification, simulation-based predictions, validation against observations, and, where the model is wanting, data assimilation to improve the model, re-initiating the inference cycle at a higher level.

We are developing a petascale cyberfacility for seismic hazard analysis, PetaSHA, that will enable scientists to work outward on this inference spiral, utilizing NSF’s open-science computing resources with an improved arsenal of data and model analysis tools. The PetaSHA computational platforms have been designed to vertically integrate the high-performance hardware, scalable software, and scientific expertise needed to execute the key computational pathways of seismic hazard analysis. These platforms are still evolving towards petascale capability, but they have already delivered major advances in earthquake system science: (1) The first time-dependent, uniform California earthquake rupture forecast (UCERF), released in April, 2008, was developed on, and is being delivered to, end-users via the OpenSHA platform. (2) A high-resolution simulation of a M7.8 earthquake on the San Andreas fault computed on the TeraShake platform was the scientific model used in November, 2008, for the Great Southern California ShakeOut, the largest earthquake disaster drill in U.S. history. (3) The first full-3D tomography was run for the LA area, and regional waveform data throughout most of Southern California are now being assimilated into the SCEC 3D community velocity models on the Tera3D platform. (4) Physics-based probabilistic seismic hazard analysis has been implemented on the CyberShake platform, and the first physics-based hazard maps, completed in June, 2009, are showing how source directivity, rupture complexity, and basin effects control strong ground motions.

The practical mission of PetaSHA is to provide society with better predictions of earthquake hazards. Southern California, the natural laboratory for the proposed project, comprises 23 million people and about half of the national earthquake risk. This project will provide the high performance computing required to achieve the objectives for earthquake source physics and ground motion prediction in the SCEC3 (2007-2012) research plan, which has engaged over 600 scientists at more than 60 research institutions. As demonstrated by UCERF and ShakeOut, its cyberinfrastructure and simulation results are being used by the SCEC community and its partners in earthquake engineering and disaster management. The project will benefit the USGS National Seismic Hazard Mapping Program. Experience with simulation-based research in a petascale environment will enhance geosystem science in general and the NSF EarthScope and NEES programs in particular. It will promote the vertical integration of cyberinfrastructure in parallel with petascale hardware development, thus supporting NSF’s plans to achieve petascale computing within several years. Several early-career scientists will be involved in project leadership, and a diverse set of undergraduate and graduate interns from the UseIT and ACCESS programs will participate, enhancing the career trajectories of women and minorities interested in HPC.

Project Plan. Our 2-yr research program will be organized into three project tracks, each with specific geoscience objectives measured by explicit milestones. (1) The Big Ten track will generate a hierarchy of simulations for ten of the most probable large (M > 7) ruptures in Southern California, with the objective of understanding how source directivity, rupture complexity, and basin effects control ground motions. (2) The Tera3D track will validate earthquake simulations using well-recorded regional events (M ≤ 6.7) and will use full-3D tomography to assimilate regional waveform data into the SCEC community velocity models. (3) The CyberShake track will validate the CyberShake hazard curves and extend the CyberShake maps to higher frequencies and more extensive geographic coverage, providing a rich new database that earthquake scientists and engineers can use for ground motion studies and performance based seismic design. We will coordinate across these project tracks to achieve a set of computer science/geoinformatics objectives and workforce-development objectives, which include equipping a diverse scientific workforce with petascale skills and capabilities. The project timetable is constructed around 14 simulation milestones that will provide metrics to evaluate how the PetaSHA platforms are being used and what impact they are having on the scientific community.