HR: 0830h
AN: S21D-0330    [PDF]
TI: Regional Travel-Time Uncertainty and Seismic Location Improvement Using 3-Dimensional {\it a priori} Velocity Models
AU: * Flanagan, M P
EM: flanagan5@llnl.gov
AF: Lawrence Livermore National Laboratory, Earth Sciences Division, L-205 PO Box 808, Livermore, CA 94551 United States
AU: Myers, S C
EM: myers30@llnl.gov
AF: Lawrence Livermore National Laboratory, Earth Sciences Division, L-205 PO Box 808, Livermore, CA 94551 United States
AB: We assess the ability of 3-D velocity models to better predict regional seismic travel times, relative to 1-D models, as well as quantify the travel time uncertainties. Accurate travel time prediction and uncertainty characterization is essential for properly identifying regional seismic phases and computing seismic event locations with representative error ellipses. To accomplish this we use an {\it a priori,} 3-D velocity model of Western Eurasia developed at LLNL [{\it Pasyanos et al.,} 2003] and a dataset of {\it P} and {\it Pn} phase picks from a catalog of calibration events which meet strict network criteria as investigated by {\it Bondar et al.,} [2003]. We evaluate the predictive power of the model by computing the median residuals between the observed arrivals and those predicted by 3-D finite difference computations. Statistical assessment of prediction accuracy is made in a non-stationary framework. We conclude that 1-D velocity models are not able to accurately characterize travel time uncertainties; not even radially symmetric models will work. Also, statistical estimates of uncertainty must be made in a region-specific framework ({\it i.e.,} 3-D). These results demonstrate quantitatively that 3-D velocity models accomplish two goals: improving travel-time prediction (by reducing overall variance of residuals), and reducing the dimensionality of the uncertainties by accounting for the non-stationary component. We also investigate improvement in seismic location using an adaptive station correction approach which combines both model predictions and empirical data. We find that no individual Earth model provides optimal travel-time prediction everywhere. We have, therefore, geographically merged travel-time predictions from a variety of different models and empirical observations to form a travel-time model (correction surface) for each station. Our approach combines the extrapolative advantages of model-based corrections and the interpolative/geostatistical advantages of kriging [{\it Schultz et al.,} 1998] to produce hybrid travel-time predictions and uncertainty models. For ease of use, model-based and empirical corrections are combined to produce one travel-time correction surface (per seismic station) and uncertainty model that is applied on top of the {\it iasp91} global Earth model in our location algorithm. Such 3-D models and uncertainty characterizations help to achieve location accuracy and computation of representative location error ellipses in a regional monitoring environment, particularly for small events not recorded teleseismically. {\it This work was performed under the auspices of the U.S. Department of Energy by the University of California Lawrence Livermore National Laboratory under contract No. W-7405-Eng-48, Contribution UCRL-JC-155079-ABS}.
DE: 7205 Continental crust (1242)
DE: 7218 Lithosphere and upper mantle
DE: 7219 Nuclear explosion seismology
DE: 9335 Europe
SC: Seismology [S]
MN: 2003 Fall Meeting