Investigating the use of Three-dimensional Travel-time Tables for Standard Regional Seismic Event Relocation

As computer processor speeds continually increase, we are routinely locating larger earthquake data sets with more complex algorithms. Despite the increase in processing speed and capacity, large global seismic catalogs are still produced primarily using one-dimensional (1-D) velocity models. Results using these models can be significantly improved through application of static or source-specific station corrections -- for example, the use of two-dimensional (2-D) correction surfaces developed with kriging methods has greatly improved relocation accuracies for regional events. Three-dimensional (3-D) velocity models, however, are improving in both resolution and reliability. Studies have demonstrated their superiority over 1-D models in terms of improved location and reduction in uncertainty. It therefore seems appropriate that routine seismic event location methods should migrate to the 3-D velocity model standard. We must consider, however, the increased processing time to produce 3-D relocations, even using faster computers. Further, it is important to assess the accuracy of 3-D travel-time (TT) tables and the ray tracing algorithms used to create them, to weigh -- in combination with the added computational burden -- the merits versus costs of such a transition. The objected-oriented relocation software (LocOO) developed by programmers at Sandia National Laboratories permits the use of 3-D TT tables to relocate events. The code is based on the "libloc" algorithms, accesses Oracle databases, and uses the Parametric Grid Library (PGL) for TT tables and station parameters. We plan to generate TT tables based on an a priori 3-D velocity model of the Central Asia region by selecting stations to maximize azimuthal coverage for the western China area. Using the LocOO software, we will test whether using 3-D TT tables improves the accuracy of regional seismic events that have corresponding ground-truth data for comparison. We will also compare these relocations with those produced using 1-D velocity models coupled with 2-D correction surfaces. Others have used 3-D TT tables for relocation but usually for specific event or event-cluster analyses. Our effort will focus on performing relocation tests on a catalog scale (or at least a large subset), and analyzing computational times to determine if using 3-D TT tables for regional location on a routine basis can be justified in an operational setting.