AUTOMATIC PHASE IDENTIFICATION WITH ONE AND 2 LARGE

ection and the estimation of seismic signal parameters is done automatically and in real time by the signal processing systems at LASA and NORSAR. The operation and capabilities of these systems were described in an earlier paper (Shlien and Toks6z, 1973). Once a signal is detected, it can be one of the several short-period body phases observable at the arrays. Ignoring the depth phases, the more important phases are P, PKP, PeP, ScP, SKP, PP, PKKP, and P'P'. These phases must be properly identified before the epicenter and origin time of the earthquake can be determined. In this paper, the feasibility of automatically identifying earthquake phases in real time using either LASA, or LASA together with NORSAR. is investigated. The input to the automatic phase identifier is confined to the detection logs generated by the detection processors at LASA and NORSAR. The detection processors form 600 beams at LASA and 490 beams at NORSAR steered in specific signal directions. The signal-to-noise ratio (SNR) of each beam is monitored in real time: whenever this ratio exceeds a preset threshold a detection is declared and the beam number, detection time, and signal strength are recorded into the Detection Log. If several beams are triggered at the same time, only the beam with maximum signal strength is entered. Since a number of phases are generally observed from an earthquake, a single event may generate several contiguous detection entries on the Log. The group of detections shall be called a detection group. On an average day, 250 detection groups were recorded by LASA and 100 by NORSAR. It was estimated in the previous paper (Shlien and Toks/Sz, 1973) that about 80 per cent of these detection groups were false alarms having low signal strength. The problem of discrimination of false alarms from signals limits the capability of the arrays to reporting detections with high SNR. The underlying structure of the phase identifiers was based upon statistical pattern recognition theory. Phase identifications were made by computing the likelihood ratio function for the different interpretations and then choosing the interpretation with the highest likelihood ratio. The two-array phase identifier, although it is much more complicated, does not involve any more important theoretical concepts. For this reason, most of the emphasis will be placed on explaining the operation of the single array identifier and the two-array identifier will be described briefly.
Bulletin of the Seismological Society of America.
Vol. 64, No. 1, pp. 221-233. February 1974
AUTOMATIC PHASE IDENTIFICATION WITH ONE AND TWO LARGE APERTURE SEISMIC ARRAYS
BY SEYMOUR SHLIEN AND M . NAFI TOKSOZ
ABSTRACT
Schemes to automatically identify later phases using LASA and using LASA together with N O R S A R were designed and tested. With a single array, phase identification was accomplished by searching for a P or PKP phase that could be associated with a particular later phase such as PP, PcP, etc. On the basis of the time interval between the signals, their slowness, and relative amplitudes, the best phase interpretation was chosen by means of the Neyman-Person likelihood ratio test. With two arrays, detections from one array were checked against detections from the other array for signals originating from the same event. Testing the schemes on the LASA and N O R S A R detection logs, 9 later phases per day were detected and 11 events per day were found common to the LASA and N O R S A R detection logs. Confusion of later phases was very minor. The most serious problem was the identification of fictitious phases and events that occurred due to various coincidences.
TL, (TN)
Atz, (AtN)
var (.)
Azimuth (A) and distance (D) of LASA (NORSAR) beam with respect to LASA (NORSAR) Azimuth and distance of LASA (NORSAR) beam with respect to NORSAR (LASA), subscript refers to the coordinate system Azimuth and distance of beam NBM from array given hypothesis H Covariance of variables in parentheses Presumed epicenter for a given phase interpretation hypothesis False alarm Hypotheses for the single array and the two array phase identifier Maximum short-term averages of first detection, later detection, LASA detection, and NORSAR detection Beam numbers of first detection, later detection, LASA detection, and NORSAR detection Probability of observing X given Y Decision parameters for the two-array phase identifier Decision parameters for the two-array phase identifier with respect to LASA (NORSAR) Standard d69iation of variable in parentheses Time interval' between first and second detection at LASA or time interval between a LASA and a NORSAR detection Travel time of a seismic phase to LASA (NORSAR) Predicted time interval between the LASA and the NORSAR detection assuming the event location is given by the LASA (NORSAR) beam Variance of the variable in parenthesis