The Reduced Detection Rate of Signals That Are Hidden by Earthquakes: Case Studies with Spotlight Detectors That Operate at Seismic Arrays

Abstract

We directly test if seismic waveforms that are sourced by earthquakes and that interfere with signals sourced by underground explosions can significantly reduce the probability that such explosion signals are detected. We perform this test with multichannel correlation detectors (correlators) that use records of ground motion (templates) sourced by explosions to detect smaller signals from similar collocated sources. Our test applies these detectors against thousands of signals with a waveform injection method. This method amplitude scales a template waveform over a grid of amplitude values, adds these waveforms into a target data stream at times that create interference with background seismicity, and processes their sum with a correlator. We apply this method to explosion templates sourced in Nevada (United States) and recorded by the multichannel array NVAR. Our study compares correlator performance when we deliberately inject templates into earthquake signals relative to baseline operation that processes target waveforms injected into data that is absent of known seismicity. We find that a correlator that uses an explosion-sourced template and that can reliably detect a 1.7 ton shallowly buried explosion in background noise (a 0.97 detection rate) is unlikely to detect the same event in noisy earthquake interference (a 0.37 detection rate). This masking remains significant when explosion and earthquake origin times separate by as much as 100 s. We also find that the performance of correlators that use a template sourced by an earthquake is even more degraded and can fall from a 0.92 to a 0.16 detection rate during earthquake swarms. We conclude that earthquake seismicity can mask explosion signals with significant probability and that swarms can also mask significant repeating earthquake seismicity.