10 years of CyberShake: Where are we now and where are we going with physics-based PSHA?

Scott Callaghan, Robert W. Graves, Kim B. Olsen, Yifeng Cui, Kevin R. Milner, Christine A. Goulet, Philip J. Maechling, & Thomas H. Jordan

Submitted August 2, 2017, SCEC Contribution #7394, 2017 SCEC Annual Meeting Talk on Mon 11:00 (PDF)

2017 marks 10 years of development on SCEC's physics-based probabilistic seismic hazard analysis (PSHA) platform, CyberShake. CyberShake establishes an updated PSHA methodology that utilizes multiple SCEC community products, including UCERF to provide an earthquake rupture forecast (ERF), UCVM, to generate velocity meshes from SCEC community models, AWP-ODC code, to calculate Strain Green Tensors (SGTs), and Broadband Platform rupture generators to produce individual rupture realizations. These elements are combined via scientific workflows to enable computation of a complete set of PSHA data, including rupture realizations, seismograms, intensity measures, duration metrics, hazard curves, and hazard maps.

We will present results from the most recent CyberShake Studies, 15.4 and 17.3, produced on the largest open-science supercomputers. CyberShake Study 15.4 calculated a seismic hazard model for Southern California at 336 sites at 1 Hz using a tomographically-derived 3D velocity model. Using averaging-based factorization techniques, we can decompose CyberShake variability into source, site, path, and directivity effects, enabling direct comparison of this study with ground motion prediction equation (GMPE)-based PSHA results. We find that physics-based simulations, by capturing directivity and basin effects, have the potential to lower previously unexplained variability in the GMPEs, decreasing hazard estimates at high ground motions by orders of magnitude. In Study 17.3, CyberShake was migrated to the Central California region for the first time. PSHA results for two seismic hazard models, generated using a 3D tomographically produced velocity model and a regionally averaged 1D model, were calculated at 438 sites at 1 Hz. Of particular interest is the basin response in the southern San Joaquin Valley, lower than that for smaller basins as well as GMPE predictions.

We will discuss how CyberShake extends existing PSHA methods, and describe applications of the rich CyberShake dataset, including improving earthquake early warning algorithms (Böse et al., 2014), evaluating building response (Deierlein et al., 2016), and as input to seismic codes through the SCEC UGMS committee. Finally, we will outline future plans for CyberShake, including integrating new ERFs, integrating physics for high frequency ground motions, and migrating the methodology to new geographic regions, while remaining computationally tractable.

Key Words
seismic hazard, ground motion simulations

Citation
Callaghan, S., Graves, R. W., Olsen, K. B., Cui, Y., Milner, K. R., Goulet, C. A., Maechling, P. J., & Jordan, T. H. (2017, 08). 10 years of CyberShake: Where are we now and where are we going with physics-based PSHA?. Oral Presentation at 2017 SCEC Annual Meeting.


Related Projects & Working Groups
Community Modeling Environment (CME)