Spectral discrimination between Soviet explosions and earthquakes using short-period array data

SUMMARY Spectral ratio discriminants are applied to 61 presumed nuclear explosions, five large chemical explosions, and 44 earthquakes in the Soviet Union recorded at the four UK teleseismic arrays. The nuclear explosions were mainly from the Shagan test site (STS) and from peaceful nuclear explosions (PNE) ranging in magnitude from 4.2 to 6.2. The earthquakes were mostly located in the south-central Soviet Union and range in magnitude from 4.4 to 5.9. For each event/station pair, the P-wave spectrum is calculated out to 10 Hz using a 5 s window from the time-domain array beam and from a spectral stack. The spectral ratios from the array beam in the 0.5-1.0 Hz and 2.0-3.0 Hz frequency bands appear to give the best discrimination performance. At higher frequencies, reduced signal-to-noise ratios (particularly for the earthquakes) result in fewer measurements and increased scatter. In general, the explosions are characterized by the existence of more high-frequency energy than the shallow crustal earthquakes. A least-squares matrix factorization approach is used to separate out source and path effects. The separation of earthquakes and explosions is generally quite good down to magnitude 4.5 where the two populations merge because of signal-to-noise problems. The spectral values from the PNEs shows increased scatter relative to those from STS, presumably due to their wider range of emplacement conditions and geologic environment. Previous studies using regional phases indicate that NTS explosions are deficient in high frequencies relative to western US earthquakes. These contrasts in frequency content between NTS and STS explosions may be caused by differences in the near-source dynamic material response to the passage of the explosion shock wave or to depth-dependent effects of Q.