Nuclear explosion locations at the Balapan, Kazakhstan, nuclear test site: the effects of high-precision arrival times and three-dimensional structure

Abstract

We have investigated the potential contributions of improved arrival times (using waveform cross-correlation) and the use of three-dimensional (3-D) velocity models for seismic event location capability. Our analyses are applied to a dataset of nuclear explosions at Balapan, Kazakhstan, for which ground-truth locations and some absolute origin times are available. This ground-truth information allows us to determine excellent origin time estimates for the remaining explosions. The combination of excellent ground-truth location information and high-quality origin time estimates permits us to (1) carry out a detailed examination of the quality of ISC picks, (2) identify probable timing errors in the digital data, (3) evaluate relative and absolute location capability using data from a sparse network, (4) assess the influence of event signal-to-noise ratio (SNR) on relative location accuracy, (5) utilize the Balapan events as a source array for 3-D tomography beneath the test site, and (6) test the influence of 3-D structure (local and global) on relative location accuracy and precision in a “controlled” situation. Our principal finding is that improved arrival times are the primary contributor to improved locations. Joint and individual relocations of Balapan events using the full digital dataset result in average mislocations of less than 1 km and 95% confidence regions of a compatible size. To mimic a CTBT scenario more realistically, we also carry out relocations using very few stations (4–10 observations). Location accuracy degrades somewhat, but the high-quality picks generally result in mislocations less than 10 km, even for events with very large azimuthal gaps. Uncertainty is generally underestimated in these cases. Tests with artificially degraded SNR show that mislocation increases slowly as SNR decreases. 3-D velocity structure makes a smaller contribution to relative location accuracy than accurate time picks. Travel time variations due to global 3-D structure vary little across the source region, so that location scatter is not reduced when travel time corrections for global 3-D structure are applied. Travel time variations due to the local 3-D structure (estimated using source-region tomography) are also modest. Applying travel time corrections that account for the local structure does yield slight location improvement.