Time Domain Regional Discriminants

Abstract

The time and frequency domains are equivalent displays of seismic trace, information, though some qualities of the signal are more easily observed in one domain than the other. The relative frequency excitation of Lg, for instance, is most easily viewed in the frequency domain, but such waveform qualities as the sequence in which pulses arrive in the wave train or the sharpness of pulse onset are most easily studied in the time domain (Murphy and Bennett, 1982, Blandford, 1981). Because of the tremendous complexity of high frequency regional data, most attempts at using it for discrimination purposes have involved analysis of the frequency content of the various arrivals either through transforming selected windows or through multiple bandpass filtering. We report here on our initial attempts to explore the alternative and to discriminate events using those waveform characteristics most easily observed in the time domain. A second advantage of time domain analysis approaches is that they permit a deeper insight into the physical processes creating a seismic signal's character. For this reason, they can be more e3silv used to evaluate the transportabilty of a discriminant to varying geophysical and tectonic regimes. This is an especially important feature in the development of regional discriminants. The most prominent and successful spectral regional discriminants have been empirically developed. This means that they must be redeveloped and reverified in each new area. As we shall show in the following, through rigorous time domain analysis such features as regional depth phases can be identified and used to discriminate. Discriminants based on such simple physical features as source depth should be transportable anywhere. In work recently completed under the treaty verification program, we have proved that such time domain discriminants do exist. In analyzing a test discrimination data set from the western U. S., we have discovered that the onset of P_n is always very similar for explosions and that few earthquakes have this unique waveform character. This information can be constructed into a simple discrimination scheme by testing the correlation of observed P_n waveform onsets with average waveforms observed from explosions. High correlations indicate explosions and low correlations earthquakes. We have also discovered that the regional phase P_g is actually composed of a sequence of sub-arrivals which correspond to successively higher orders of reverberation in the crust. In realistic crust models, the depth phases play an important role in the waveshapes of these sub-arrivals. By selecting an appropriate frequency band to analyze, we have been able to accurately model this type of data from explosions in the western United States. Over the very relevant regional distance ranges of 200 to 600 km, it appears that a discrimination procedure very similar to the one which is known to work for P_n will also be effective for P_g. We are investigating whether similar discriminants can be constructed based on the phases S_n and S_g in areas where those phases are prominent arrivals.

Files

Additional details