HR: 08:15h
AN: SF31B-02 INVITED     [Abstracts]
TI: Cyberinfrastructure for the Unified Study of Earth Structure and Earthquake Sources in Complex Geologic Environments
AU: * Zhao, L
EM: zhaol@usc.edu
AF: Department of Earth Sciences, University of Southern California, Los Angeles, CA 90089-0740 United States
AU: Chen, P
EM: pochen@usc.edu
AF: Department of Earth Sciences, University of Southern California, Los Angeles, CA 90089-0740 United States
AU: Jordan, T H
EM: tjordan@usc.edu
AF: Department of Earth Sciences, University of Southern California, Los Angeles, CA 90089-0740 United States
AU: Olsen, K B
EM: kolsen@geology.sdsu.edu
AF: Department of Geological Sciences, San Diego State University, San Diego, CA 92182-1020 United States
AU: Maechling, P
EM: maechlin@usc.edu
AF: Southern California Earthquake Center, Univerisity of Southern California, Los Angeles, CA 90089-0742 United States
AU: Faerman, M
EM: mfaerman@cs.ucsd.edu
AF: Dept. of Computer Science and Engineering, University of California, San Diego, CA 92093-0114 United States
AB: The Southern California Earthquake Center (SCEC) is developing a Community Modeling Environment (CME) to facilitate the computational pathways of physics-based seismic hazard analysis (Maechling et al., this meeting). Major goals are to facilitate the forward modeling of seismic wavefields in complex geologic environments, including the strong ground motions that cause earthquake damage, and the inversion of observed waveform data for improved models of Earth structure and fault rupture. Here we report on a unified approach to these coupled inverse problems that is based on the ability to generate and manipulate wavefields in densely gridded 3D Earth models. A main element of this approach is a database of receiver Green tensors (RGT) for the seismic stations, which comprises all of the spatial-temporal displacement fields produced by the three orthogonal unit impulsive point forces acting at each of the station locations. Once the RGT database is established, synthetic seismograms for any earthquake can be simply calculated by extracting a small, source-centered volume of the RGT from the database and applying the reciprocity principle. The partial derivatives needed for point- and finite-source inversions can be generated in the same way. Moreover, the RGT database can be employed in full-wave tomographic inversions launched from a 3D starting model, because the sensitivity (Fr{\'{e}}chet) kernels for travel-time and amplitude anomalies observed at seismic stations in the database can be computed by convolving the earthquake-induced displacement field with the station RGTs. We illustrate all elements of this unified analysis with an RGT database for 33 stations of the California Integrated Seismic Network in and around the Los Angeles Basin, which we computed for the 3D SCEC Community Velocity Model (SCEC CVM3.0) using a fourth-order staggered-grid finite-difference code. For a spatial grid spacing of 200 m and a time resolution of 10 ms, the calculations took ~19,000 node-hours on the Linux cluster at USC's High-Performance Computing Center. The 33-station database with a volume of ~23.5 TB was archived in the SCEC digital library at the San Diego Supercomputer Center using the Storage Resource Broker (SRB). From a laptop, anyone with access to this SRB collection can compute synthetic seismograms for an arbitrary source in the CVM in a matter of minutes. Efficient approaches have been implemented to use this RGT database in the inversions of waveforms for centroid and finite moment tensors and tomographic inversions to improve the CVM. Our experience with these large problems suggests areas where the cyberinfrastructure currently available for geoscience computation needs to be improved.
DE: 7200 SEISMOLOGY
SC: Special Focus: Advances in Data Acquisition, Management, Analysis and Display [SF]
MN: 2004 AGU Fall Meeting