• journal_title：Bulletin of the Seismological Society of America
• Contributor：J. R. Murphy ; W. Rodi ; M. Johnson ; D. D. Sultanov ; T. J. Bennett ; M. N. Toksöz ; V. Ovtchinnikov ; B. W. Barker ; D. T. Reiter ; A. C. Rosca ; Y. Shchukin
• Publisher：Seismological Society of America
• Date：2005-
• Format：text/html
• Language：en
• Identifier：10.1785/0120040087
• journal_abbrev：Bulletin of the Seismological Society of America
• issn：0037-1106
• volume：95
• issue：4
• firstpage：1535
• section：Articles

It is well documented that lateral heterogeneities in the earth can introduce serious errors in seismic event locations derived from radially symmetric earth models, in particular, for small events recorded only by sparse networks of regional stations. It should be possible to significantly improve the seismic location accuracy for such events by applying travel-time calibration information specific to the regions where the sources and stations are located. This article describes a research investigation directed toward regional travel-time calibration of 30 International Monitoring System (ims) stations in eastern Asia. For this calibration an initial 3D velocity model covering the entire region was constructed from several submodels. Seismic travel times in this complex 3D velocity model have been computed by using a raytracing algorithm based on a finite-difference approximation to the eikonal equation, which is believed to be accurate to within 0.3–0.5 sec over the regional distance range of interest. A new regional tomography algorithm has also been implemented which solves the fully nonlinear problem by iterating over linear inversion steps in which event hypocenters, model velocities, and ray paths can all be updated. Extensive testing and validation of the final tomographically refined velocity model have been performed with data from numerous ground truth (gt) explosions and earthquakes throughout the region, and it has been demonstrated that this model predicts P-wave travel times with associated root-mean-square (rms) errors on the order of 1 sec across the study region. Finally, a completely new multistation kriging method, which satisfies seismic reciprocity constraints, has been formulated and applied to derive empirical corrections to account for remaining unmodeled error, further reducing the rms error to approximately 0.7 sec. These results support the conclusion that our new 3D velocity model for this region represents a significant improvement over the default IASP91 model and will provide improved seismic location capability, in particular, for events recorded by sparse networks of regional stations.

Online Material: Source-specific station correction and travel-time calibration files.