Developing regional discriminants at any given seismic station requires a ground truth database of waveforms from both earthquakes and explosions. Recently installed stations used for seismic monitoring have no single-charge explosions on which to base discriminants. We have developed a procedure to map regional seismic amplitudes from surrogate stations, which have a long recording history, to newly installed operational stations. This procedure uses an interstation amplitude correction factor (IACF), computed using a set of commonly recorded calibration earthquakes between the surrogate and operational station pair, to predict seismic amplitudes at the new station using the IACF and historical explosion amplitudes from the surrogate station.
We tested the method using an earthquake aftershock sequence in Pakistan recorded at the Kyrgyzstan broadband seismic network. The aftershocks had a range of depths and were clustered in a linear trend defining the fault rupture plane of the mainshock. We computed IACFs for all possible station combinations at KNET (45 combinations) for all events and then computed a mean IACF value for each station pair. Using the mean IACF values, we calculated predicted phase amplitudes at each station and compared the predictions to the actual values. The observed-predicted relationship performed well (r2-values>0.8) for all phases and frequency bands. At frequencies above 6 Hz, the observed-predicted relationship showed more scatter, a trend not due to interstation distance, source depth, or spatial location of the calibration earthquakes. We were able to reduce scatter in the observed-predicted relationship by applying path corrections to the data, suggesting that sensitivities in differing paths between certain surrogate-operational station combinations do play a role in the effectiveness of the method.
We used this methodology to predict explosion values at the Lop Nor nuclear test site in China using a suite of nearby calibration earthquakes to calculate IACF values. The results indicate good correlation between the observed and predicted amplitudes, suggesting that the method can be used to simulate explosion characteristics at stations that have never recorded explosions. Comparing the IACF values from the calibration dataset near Lop Nor to those calculated for the earthquakes in Pakistan, we see some similarity for certain station combinations; however, on the whole, there are differences that suggest that IACFs have a spatial variance and must be tuned to the region and path of interest for the method to be successful.