Your query was:

"NG11A-0181"

HR: 0830h
AN: NG11A-0181
TI: Earthquake Source Simulations: A Coupled Numerical Method and Large Scale Simulations
AU: * Ely, G P
EM: gely@ucsd.edu
AF: Institute of Geophysics and Planetary Physics, University of California, San Diego 9500 Gilman Drive, La Jolla, CA 92093-0225 United States
AU: Xin, Q
AF: San Diego Supercomputer Center, University of California, San Diego 9500 Gilman Drive, La Jolla, CA 92093-0505 United States
AU: Faerman, M
AF: San Diego Supercomputer Center, University of California, San Diego 9500 Gilman Drive, La Jolla, CA 92093-0505 United States
AU: Day, S
AF: Department of Geological Sciences, San Diego State University 5500 Campanile Drive, San Diego, CA 92182-1020 United States
AU: Minster, B
AF: Institute of Geophysics and Planetary Physics, University of California, San Diego 9500 Gilman Drive, La Jolla, CA 92093-0225 United States
AU: Kremenek, G
AF: San Diego Supercomputer Center, University of California, San Diego 9500 Gilman Drive, La Jolla, CA 92093-0505 United States
AU: Moore, R
AF: San Diego Supercomputer Center, University of California, San Diego 9500 Gilman Drive, La Jolla, CA 92093-0505 United States
AB: We investigate a scheme for interfacing Finite-Difference (FD) and Finite-Element (FE) models in order to simulate dynamic earthquake rupture. The more powerful but slower FE method allows for (1) unusual geometries (e.g. dipping and curved faults), (2) nonlinear physics, and (3) finite displacements. These capabilities are computationally expensive and limit the useful size of the problem that can be solved. Large efficiencies are gained by employing FE only where necessary in the near source region and coupling this with an efficient FD solution for the surrounding medium. Coupling is achieved through setting up and an overlapping buffer zone between the domains modeled by the two methods. The buffer zone is handled numerically as a set of mutual offset boundary conditions. This scheme eliminates the effect of the artificial boundaries at the interface and allows energy to propagate in both directions across the boundary. In general it is necessary to interpolate variables between the meshes and time discretizations used for each model, and this can create artifacts that must be controlled. A modular approach has been used in which either of the two component codes can be substituted with another code. We have successfully demonstrated coupling for a simulation between a second-order FD rupture dynamics code and fourth-order staggered-grid FD code. To be useful earthquake source models must capture a large range of length and time scales, which is very computationally demanding. This requires that (for current computer technology) codes must utilize parallel processing. Additionally, if larges quantities of output data are to be saved, a high performance data management system is desirable. We show results from a large scale rupture dynamics simulation designed to test these capabilities. We use second-order FD with dimensions of 400 x 800 x 800 nodes, run for 3000 time steps. Data were saved for the entire volume for three components of velocity at every time step and six components of stress at every tenth time step. This generated a 10 terabyte data set that was handled with the SDSC Storage Resource Broker (SRB). Run time was approximately 12 hours on SDSC Blue Horizon machine, and data archival to SRB took approximately 5 days.
DE: 7209 Earthquake dynamics and mechanics
SC: Nonlinear Geophysics [NG]
MN: 2003 Fall Meeting