Waveform correlation detectors used in seismic monitoring scan multichannel data to test two competing hypotheses: that the data contain (1) a noisy, amplitude‐scaled version of a template waveform or (2) only noise. In reality, seismic wavefields include signals triggered by nontarget sources (background seismicity) and target signals that are only partially correlated with the waveform template. We reform the waveform correlation detector hypothesis test to accommodate deterministic uncertainty in template‐to‐target waveform similarity and thereby derive a new detector from convex set projections (the cone detector) for use in explosion monitoring. Our analyses give probability density functions that quantify the detectors’ degraded performance with decreasing waveform similarity. We then apply our results to three announced North Korean nuclear tests and use International Monitoring System (IMS) arrays to determine the probability that low magnitude, off‐site explosions can be reliably detected with a given waveform template. We demonstrate that cone detectors provide (1) an improved predictive capability over correlation detectors to identify such spatially separated explosive sources, (2) competitive detection rates, and (3) reduced false alarms on background seismicity.
Online Material: Description and illustration of several theoretical and practical aspects of implementing the correlation and cone detector on geophysical waveform data.