This paper presents the basic method and algorithm of a system for automatic on-line analysis of seismic data at three-component stations and gives preliminary experimental results. Data processing consists of four steps: (1) preliminary detection and estimation of the signal arrival time by using the energy-based γ-detector; (2) refinement of signal parameters (arrival time, amplitude, period, and coda duration); (3) estimation of the azimuth and incident angle of the seismic ray by using polarization analysis; and (4) approximate estimation of event parameters (epicenter coordinate, origin time, and magnitude) from teleseismic P waves for a constrained focal depth.
The software package, which is called “SEISMOSTANSIA,” has been tested in an on-line environment at an experimental seismic station in Eastern Kazakhstan. Testing results show that the system can handle an input flow of up to 100 signals in a 24-hr period. Comparison of the results of automatic detection with data of the station bulletin obtained by visual analysis of analog seismograms shows that: (1) with the present threshold setting, the automatic system detects practically all teleseismic signals seen by the analyst together with typically 10 to 15 signals per day not confirmed by visual analysis; (2) in 70 per cent of the cases, the residual between the arrival time obtained from the automatic and subjective procedures is 0.5 sec or less for SNR > 10 and 1 sec or less for SNR < 10; (3) in 70 per cent of the cases, the residual between the signal period obtained from the automatic and subjective procedures is 0.2 sec or less. Analysis of data indicates that an effective means of identifying false alarms with low SNR can be the level of polarization rectilinearity of P and S waves. By analyzing azimuth estimates for a suite of explosions at the Nevada Test Site, it is shown that a systematic bias forms the main contribution to the residual relative to true azimuth. In the general case, the azimuth residuals are due to the features of the medium both at the epicenter and at the station, and consequently depend on both azimuth and epicentral distance. An analysis of azimuth residuals points to a periodic dependence on azimuth. The value of the period is about 120°. Additional information (azimuth and angle of incidence) reported from three-component stations to data centers is shown to improve significantly the reliability of phase association in the processing of data from a global seismic network.