HR: 10:50h
AN: S11D-09
TI: Waveform Modeling of the Crust and Upper Mantle in China using S, Sp, SsPmP, and Shear-Coupled PL Waves
AU: * Pulliam, J
EM: jay@ig.utexas.edu
AF: Institute for Geophysics, University of Texas at Austin, 4412 Spicewood Springs Road, Bldg. 600, Austin, TX 78751 United States
AU: Sen, M
EM: mrinal@ig.utexas.edu
AF: Institute for Geophysics, University of Texas at Austin, 4412 Spicewood Springs Road, Bldg. 600, Austin, TX 78751 United States
AB: Teleseismic P waves are often used to produce "receiver functions", models of the crust and upper mantle beneath stations. However, these waves arrive at a station via steep propagation angles and therefore sample only a narrow cone beneath the station, resulting in poor depth resolution. A suite of phases that arrives around the direct S phase, including Sp (converted at the base of the Moho), SsPmP, and shear-coupled PL (SPL) waves, collectively sample the medium at more oblique angles and therefore have the potential to produce a better lateral average of structural properties than teleseismic P waves. SPL waves are sensitive to velocity structure, including gradients, Vp/Vs, and layer thicknesses in the crust and uppermost mantle. The relatively large-amplitude Sp and SsPmP phases can often be modeled simultaneously with SPL, which offers the potential to constrain the model more accurately. A combination of the high frequencies required over teleseismic distances, which make the computation of synthetic seismograms time-consuming, and the complex sampling of the Earth by SPL renders trial-and-error modeling infeasible. We have parallelized and optimized a synthetic seismogram code based on the reflectivity method for use with a variant of the simulated annealing global optimization method. We are thus able to compute complete seismograms up to 0.5 Hz in just over two minutes using eight 666 MHz Alpha processors, and the speed-up in computation time is nearly linear with the number of processors used. This approach also allows us to find not only the single best-fitting solution but also its uncertainty and uniqueness. Importance sampling can be used to estimate the posterior probability function (PPD), posterior mean, covariance and correlation matrices. We model observations of S, Sp, SsPmP, and SPL recorded at stations of the China Digital Seismographic Network (CDSN) and compare our results to receiver function models produced by others. Broadband waveforms are first compared to synthetic waveforms computed using receiver function models for five deep earthquakes located at distances of 31\deg -59\deg from CDSN stations. SV arrivals computed for receiver function models typically match the observations well, while Sp (and S-Sp time separation), SsPmP, and SPL phases match poorly. This result underscores the opportunity provided by these additional phases to add constraints to models of the crust and upper mantle. A global search of the model space, combined with analyses of sensitivity, resolution, and uncertainty, will evaluate tradeoffs between model parameters help build confidence in the final models.
DE: 7200 SEISMOLOGY
DE: 7218 Lithosphere and upper mantle
DE: 7219 Nuclear explosion seismology
DE: 7260 Theory and modeling
DE: 9320 Asia
SC: S
MN: 2001 AGU Fall Meeting