Abstract: | In the monitoring of earthquakes and nuclear explosions using a sparse worldwide network of seismic stations, it is frequently necessary to make reliable location estimates using a single seismic array. It is also desirable to screen out routine industrial explosions automatically in order that analyst resources are not wasted upon detections which can, with a high level of confidence, be associated with such a source. The Kovdor mine on the Kola Peninsula of NW Russia is the site of frequent industrial blasts which are well recorded by the ARCES regional seismic array at a distance of approximately 300 km. We describe here an automatic procedure for identifying signals which are likely to result from blasts at the Kovdor mine and, wherever possible, for obtaining single array locations for such events. Carefully calibrated processing parameters were chosen using measurements from confirmed events at the mine over a one-year period for which the operators supplied Ground Truth information. Phase arrival times are estimated using an autoregressive method and slowness and azimuth are estimated using broadband f{-}k analysis in fixed frequency bands and time-windows fixed relative to the initial P-onset time. We demonstrate the improvement to slowness estimates resulting from the use of fixed frequency bands. Events can be located using a single array if, in addition to the P-phase, at least one secondary phase is found with both an acceptable slowness estimate and valid onset-time estimate. We evaluate the on-line system over a twelve month period; every event known to have occured at the mine is detected by the process and 32 out of 53 confirmed events were located automatically. The remaining events were classified as “very likely” Kovdor events and were subsequently located by an analyst. The false alarm rate is low; only 84 very likely Kovdor events were identified during the whole of 2003 and none of these were subsequently located at a large distance from the mine. The location accuracy achieved automatically by the single-array process is remarkably good, and is comparable to that obtained interactively by an experienced analyst using two-array observations. The greatest problem encountered in the single array location procedure is the difficulty in determining arrival times for secondary phases, given the weak Sn phase and the complexity of the P-coda. The method described here could be applied to a wide range of locations and sources for which the monitoring of seismic activity is desirable. The effectiveness will depend upon the distance between source and receiver, the nature of the seismic sources and the level of regional seismicity. |