HR: 0830h
AN: S21D-0339    [PDF]
TI: Automatic picker of P \& S first arrivals and robust event locator
AU: * Pinsky, V
EM: vlad@gii.co.il
AF: Geophysical Inst. of Israel, PO Box 182, Lod, 71100 Israel
AU: Polozov, A
EM: vlad@gii.co.il
AF: Geophysical Inst. of Israel, PO Box 182, Lod, 71100 Israel
AU: Hofstetter, A
EM: ramk@gii.co.il
AF: Geophysical Inst. of Israel, PO Box 182, Lod, 71100 Israel
AB: We report on further development of automatic all distances location procedure designed for a regional network. The procedure generalizes the previous "loca l" (R $<$ 500 km) and "regional" (500 $<$ R $<$ 2000 km) routines and comprises: a) preliminary data processing (filtering and de-spiking), b) phase identificatio n, c) P, S first arrival picking, d) preliminary location and e) robust grid-search optimization procedure. Innovations concern phase identification, automa tic picking and teleseismic location. A platform free flexible Java interface was recently created, allowing easy parameter tuning and on/off switching to t he full-scale manual picking mode. Identification of the regional P and S phase is provided by choosing between the two largest peaks in the envelope curve. For automatic on-time estimation we utilize now ratio of two STAs, calculated in two consecutive and equal time windows (instead of previously used Akike Information Criterion). "Teleseismic " location is split in two stages: preliminary and final one. The preliminary part estimates azimuth and apparent velocity by fitting a plane wave to the P automatic pickings. The apparent velocity criterion is used to decide about strategy of the following computations: teleseismic or regional. The preliminary estimates of azimuth and apparent velocity provide starting value for the final teleseismic and regional location. Apparent velocity is used to get first a pproximation distance to the source on the basis of the P, Pn, Pg travel-timetables. The distance estimate together with the preliminary azimuth estimate provides first approximations of the source latitude and longitude via sine and cosine theorems formulated for the spherical triangle. Final location is based on robust grid-search optimization procedure, weighting the number of pickings that simultaneously fit the model travel times. The grid covers initial location and becomes finer while approaching true hypocenter. The target function is a sum of the bell-shaped characteristic functions, used to emphasize true pickings and eliminate outliers. The final solution is a grid point that provides maximum to the target function. The procedure was applied to a list of ML $>$ 4 earthquakes recorded by the Israel Seismic Network (ISN) in the 1999-2002 time period. Most of them are badly constrained relative the network. However, the results of location with average normalized error relative bulletin solutions e=dr/R of 5% were obtained, in each of the distance ranges. The first version of the procedure was incorporated in the national Early Warning System in 2001. Recently, we started to send automatic Early Warn ing reports, to the EMSC Real Time Bulletin. Initially reported some teleseismic location discrepancies have been eliminated by introduction of station corrections.
DE: 1734 Seismology
SC: Seismology [S]
MN: 2003 Fall Meeting