A waveform correlation method for identifying quarry explosions

Abstract

Abstract A waveform correlation method is presented for identifying quarry explosions by attributing them to known mines characterized by multiple master events. The objective is to provide a reliable automatic procedure for screening the large number of quarry explosions likely to be detected by networks of in-country stations monitoring compliance with test-ban treaties. The method generalizes existing correlation techniques to compare waveforms from an unlocated event recorded at an array of sensors with a linear combination of master event waveforms recorded at the same array. The use of a linear combination reduces the chance of a missed location caused by some variation in mechanism or spectral excitation between the events being compared. The weights in the linear combination are filters, offering some compensation for variations in source time functions and errors of waveform alignment. The use of array data reduces the likelihood of false attribution by reducing bias and variance in the correlation measurement. In a test conducted with P-wave data segments recorded at a 13-element array, the method successfully resolves two source regions separated by 4 km at a range of 150 km. Resolution with single-station waveform correlations is marginal due to the limited amount of data. The statistics of the sample waveform correlation coefficient are developed and demonstrate that single-station waveform correlations are unreliable unless estimated with large signal durations T or bandwidths B. A time-bandwidth TB product exceeding 100 (or smaller TB with more stations) is necessary for reliable event attribution. The related problem of separating superimposed waveforms from two events in different source regions may be solved by cancellation. The waveforms of one event are again approximated by a linear combination of waveforms from master events in the same mine. The residual signals, obtained by subtracting the approximation from the superimposed waveforms, estimate the waveforms from the second event. This method achieves significant separation of waveforms from events 6 km apart at a range of 150 km, using data from the 13-element array. Its resolution exceeds that of conventional beam-forming methods.