Spectral discrimination analysis of Eurasian nuclear tests and earthquakes recorded by the Israel Seismic Network and the NORESS array

The energy spectral ratio and the innovative spectral semblance discriminants, successfully performed previously on a local Israeli events, were verified on teleseismic short-period recordings. The events tested include 29 nuclear explosions and 41 earthquakes (m(b) = 5.2-6.5), mainly from China and Kazakhstan, recorded by the Israel Seismic Network (ISN) and the NORESS array. A 15-s window comprising P- and P-coda waves was selected for the analysis. The 'semblance' statistic commonly used in seismic prospecting for phase correlation in the time domain was modified and utilized as a measure of coherency of the smoothed spectra across the network/array. The semblance and the average spectral ratio of low-to-high frequency energy were evaluated, using a subset of 7-10 stations for a given event. Semblance and spectral ratio values, calculated from ISN seismograms, were found to be higher for earthquakes, where the analyzed waves are richer in low-frequency energy and have more coherent spectral shapes than explosions. These observations are contrary to those observed for local events. The best performance is provided in the frequency bands (0.6-1 Hz)/(1-3 Hz) (for the ratio) and (0.6-2 Hz) (for the semblance). Joint application of the two discriminants showed almost full separation (95%) between the two populations. Some explosions exhibited pronounced minima (nulls) near 1 Hz which could be interpreted in terms of interference of P- and pP-waves from a source at the depth of several hundreds of meters. Nevertheless, this feature could not be utilized as a discriminant: many explosions revealed strong variability of this minimum across the ISN network and some earthquakes also distinctly exhibited this feature. The ISN and NORESS discrimination performances were compared. The latter records showed the same (as the ISN) relation between spectral ratio values for earthquakes and explosions, whereas the character of semblance was reversed. The ratios in the frequency bands (0.6-0.8 Hz)/(0.6-10 Hz) yield full separation of the two populations. In general, the spectral ratio performance appears to be better for NORESS recordings, whereas the spectral semblance is better for the ISN data.