A new seismic discriminant based on spectral differences of regional phases from earthquakes and explosions recorded at a single station has been tested and found to work remarkably well. The test data consisted of a well-constrained set of 30 Nevada Test Site (NTS) explosions and 21 earthquakes located within about 100 km of NTS which were recorded on short-period seismographs at the Tonto Forest Observatory in central Arizona at an epicentral distance averaging 530 km. The events in the data set cover a magnitude range from 3.3 to 4.8 (mb) for which Pn, Pg, and Lg phases have been analyzed.
We found that, although Lg phases from earthquakes are typically more prominent than for explosions with comparable P-wave amplitude levels, simple time-domain Lg/P amplitude ratios do not result in a separation of the earthquake and explosion samples consistent enough to provide reliable discrimination. However, spectral analyses of the data over the frequency band from 0.5 to 5.0 Hz revealed significant differences in the spectra of certain regional phases which proved to be a quite reliable discriminant. In particular, both the Pg and Lg spectra from earthquakes have been found to be richer in high-frequency content than corresponding explosion spectra. A discriminant measure, defined as the ratio of average Lg spectral amplitude level in the 0.5- to 1.0-Hz passband to that in the 2.0- to 4.0-Hz passband, provides good separation of earthquake and explosion populations.