A Large-Population LASA Discrimination Experiment

A computer program has been written and has been applied to LASA time series data from nearly 200 events in order to obtain data for discrimination studies. In general the previously estimated performance of M -m, , spectral ratio, and S D complexity has been corroborated. Some of the specific results are the following. Except for one anomalous earthquake, all shallow earthquakes with m,)4. 8 and explosions with m,)5.1 have been correctly identified using M -m, . No events with m, (5.1 could be unequivocally identified as explosions. However, the probability that an earthquake would be identified as such decreased to zero only gradually as m, was reduced to about 4.0. A modification of spectral ratio has been made which introduced the option to make no decision concerning the nature of an event when the signal-to-noise ratio is not sufficiently large. The probability that no decision will be made is zero for ni, )4. 8 and unity for m,(4. 0. The probability of correct identification is high for events which pass the signal-to-noise ratio test. The period of short period P-waves may be a valuable discriminant at low magnitudes. Many earthquakes with magnitudes below 4. 5 can be identified as such using the period data. Depth phase picks have been made for about 60% of the earthquakes in our population. About 70% to 95% of these picks correspond to valid depth phases. Unfortunately, depth phases were also picked for three presumed explosions. Accepted for the Air Force Franklin C. Hudson Chief, Lincoln Laboratory Office