The software developed for the Comprehensive Nuclear-Test-Ban Treaty (CTBT) International Data Centre (IDC) in Vienna, Austria, uses the IASPEI91 seismic travel-time tables as default. However, in reality, travel times as a function of source location are different for each station, especially at regional distances. This is not modeled by a global one-dimensional model such as IASPEI91. The software therefore allows the application of a suite of travel-time corrections for improved travel-time prediction and consequently, more accurate hypocentral locations. To account for path dependency, we have developed gridded source-specific station corrections (SSSCs) for Pn, Sn, Pg, and Lg phases for 22 International Monitoring System (IMS) stations in North America. In deriving the SSSCs and associated modeling errors, we used a methodology that had been successfully applied previously for Fennoscandia. We first compiled regionalized one-dimensional travel times for North America. Then we assigned travel times as weighted sums of the travel times in each region to calculate SSSCs. Validation testing of the SSSCs was conducted using two separate sets of North American ground truth (GT) events with location accuracies within 10 km (GT0–GT10). These events were recorded at IMS stations between 1995 and 1999 and distributed throughout North America.
Test event location results with SSSCs applied show reductions in both mislocations and error ellipses. For GT events with location accuracies better than 2 km, the median change in location was a 5-km improvement, and the median area of the 90th percentile error ellipses was reduced from 2550 km2 to 910 km2. Using SSSCs for both data sets, the median improvement is less than 5 km. The testing emphasized that GT events with accuracies 5 km or better are crucial for validating SSSCs. Using SSSCs for all 51 GT0–GT10 events used in the validation testing, the median area of the error ellipses was reduced to less than 1000 km2, and the error ellipse reductions were achieved without significant degradation of coverage. The modeling errors used in this study may be somewhat underestimated, however, leaving the target goal of error ellipses less than 1000 km2 still somewhat in question. We expect the SSSCs, event locations, and uncertainties can be improved by future refinement of travel times and modeling errors. These corrections were implemented in the operational Prototype IDC system located in Arlington, Virginia, in March 2000 and have been delivered to the Vienna IDC in January 2001.