HR: 09:45h
AN: IN21D-08    [Abstracts]
TI: TeraShake-2: Next Steps
AU: * Minster, J
EM: jbminster@ucsd.edu
AF: Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92024 United States
AU: Olsen, K
IN21D-08 AF: San Diego State University, 5500 Campanile Drive, San Diego, CA 92182 United States
AU: Day, S
IN21D-08 AF: San Diego State University, 5500 Campanile Drive, San Diego, CA 92182 United States
AU: Cui, Y
IN21D-08 AF: San Diego SuperComputer Center, University of California, San Diego, La Jolla, CA 92093 United States
AU: Faerman, M
IN21D-08 AF: San Diego SuperComputer Center, University of California, San Diego, La Jolla, CA 92093 United States
AU: Moore, R
IN21D-08 AF: San Diego SuperComputer Center, University of California, San Diego, La Jolla, CA 92093 United States
AU: Okaya, D
IN21D-08 AF: University of Southern California, Southern California Earthquake Center, Los Angeles, CA 90089 United States
AU: Jordan, T
IN21D-08 AF: University of Southern California, Southern California Earthquake Center, Los Angeles, CA 90089 United States
AU: Archuleta, R
IN21D-08 AF: University of Southern California, Southern California Earthquake Center, Los Angeles, CA 90089 United States
AU: Hu, Y
IN21D-08 AF: San Diego SuperComputer Center, University of California, San Diego, La Jolla, CA 92093 United States
AU: Ely, G
IN21D-08 AF: Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92024 United States
AB: The SCEC ITR collaboration achieved a series of earthquake simulations on the southern segment of the San Andreas fault that brought to light an heretofore unsuspected pattern of seismic hazard, linked to fault geometry, rupture directivity, and wave propagation through the three-dimensional SCEC Community velocity model. This set of simulations, labeled TeraShake-1 relied on a kinematic fault rupture model scaled from inversion results for the 1999 Denali earthquake. That model is quite severely constrained, and questions remain about the level of uncertainty we should attach to these results. We report on the next stage of development that incorporates a spontaneous rupture model, using one or more of several codes being compared as part of the SCEC spontaneous rupture mode validation effort. There exist about a dozen such codes. We are initially restricting our focus on four such codes: (1) the 2nd-order Dynamic Fault Model (DFM) FD code that has passed numerous tests; (2) the 4th order FD Anelastic Wave Propagation Model (AWM) code, used in the SCEC TeraShake-I and Cybershake projects; (3) the UCSB Finite Element spontaneous rupture and wave propagation code; (4) the SIO 'mimetic' Support-Operator Rupture Dynamic (SORD) code. All these have quite different levels of maturity. A major challenge is that the range of scales involved in such simulations is enormous: the inner scale associated with rupture characteristics is in the range of 1-100 meters, while the outer scale associated with geology, wave propagation, and hazard assessment reaches 600 kilometers. Time scales cover an equally broad range of magnitudes. This combination raises a numerical 'grand challenge' that pushes the limits of available computing systems. Additional considerations include desired capabilities to model nonplanar faults, and to include surface topography in the calculation. Recent SCEC research results show that such complications are indeed potentially important for accurate ground motion predictions. The SCEC-ITR group has devised a consensus approach based on a 2-stage modeling strategy. In a first stage, a spontaneous rupture model is run, on a fine grid with limited geographical extent, using absorbing boundary conditions to control the effects of reflections on the rupture process. The output of this run, a complete rupture history along the fault, is then used as a kinematic source in an anelastic wave propagation code to propagate the seismic waves to regional distances, as was done in TeraShake-1. The simulations are being done on the TeraGrid collection of platforms. This approach represents admittedly an approximation that requires validation. This also raises very significant IT challenges, in terms of computational performance, storage, communications, and visualization. We report on progress achieved so far along all of these lines, and describe our plans to achieve the next level of simulations.
UR: http://www.scec.org
DE: 7212 Earthquake ground motions and engineering seismology
SC: Earth and Space Science Informatics [IN]
MN: Fall Meeting 2005