3-D a Priori Model Constraints and Uncertainties for Improving Seismic Event Location

We investigate our ability to improve seismic event location in the Middle East, North Africa, and Western Eurasia (MENAWE) by using an a priori 3-D velocity model for the crust and upper mantle in lieu of more commonly used 1-D Earth models. We develop our composite velocity model based on published literature: a global sediment thickness map, a regionalized crustal model based on geology and tectonics, and an upper mantle model [Pasyanos et al., 2001]. We test a variety of mantle velocity models developed either from teleseismic travel-times [Gudmundsson and Sambridge, 1998] or from geodynamic estimates of temperature, pressure, and composition [Nataf and Ricard, 1996]. Using a 3-D finite difference technique we compute travel times through the model and produce station-specific correction surfaces relative to iasp91. We then use these model-based correction surfaces as additional constraints in our event location algorithm to relocate a set of ground truth (GT) events (the 1991 Racha aftershock sequence). Myers and Schult [2000], using empirical kriged correction surfaces and a 1-D velocity model applied to a test network, reduced the average location bias of this aftershock sequence from 42 km (using a 1-D velocity model alone) to 13 km. With our model-based correction surfaces we reduce this bias from 42 to 26 km using the travel-time mantle model and from 42 to 10 km using the geodynamic mantle model. This is a significant result showing the power of a priori 3-D models to improve location to almost the same degree as empirical corrections, within the GT uncertainty bounds. This test relocation samples only one part of our MENAWE model, and we are continuing to test the model in other areas using geographically diverse GT events. A larger data set of GT0-GT10 events is being collected and will be used to further evaluate the effectiveness of the a priori model for improving event location accuracy. We also plan to explore a variety of validation techniques (e.g., cross-validation, sensitivity tests) and to model the uncertainty process for model-based corrections which will be required to compute representative error ellipses for the new locations. 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.