HR: 16:45h
AN: S22F-04    [PDF]
TI: Relative Earthquake Location Techniques: Tests using both Synthetic and Real Data
AU: * Lin, G
EM: gulin@ucsd.edu
AF: IGPP, SIO, UCSD, 9500 Gilman Dr., La Jolla, CA 92093-0225 United States
AU: Shearer, P
EM: shearer@igpp.ucsd.edu
AF: IGPP, SIO, UCSD, 9500 Gilman Dr., La Jolla, CA 92093-0225 United States
AB: Over the years, researchers have developed various techniques for reducing relative errors in earthquake location among nearby events. These techniques rely on the assumption that if the hypocentral separation between two earthquakes is small, travel time differences at a particular station are not biased by the effects of heterogeneity. Here, we compare three different earthquake location techniques: (1) the hypocentroidal decomposition method of Jordan and Sverdrup (1981), (2) the source-specific station term (SSST) method of Richards-Dinger and Shearer (2000), and (3) the modified double difference method (DD) of Waldhauser and Ellsworth (2001). In principle, all these methods should give approximately the same result for a single compact cluster of events, with significant differences appearing only for more distributed seismicity. We test these methods with both a synthetic data set and actual phase picks and waveform cross-correlation data from the M=5.4 Big Bear California aftershock sequence of February 2003. For the synthetic data, we generate a set of quake locations, station locations and arrival time picks in a simple half-space velocity model. We add random time noise to the data by including contributions of varying size from: (1) normally distributed random picking errors, (2) normally distributed station terms, which are constants for all events recorded by each station, and (3) spatially varying station terms computed by summing travel time anomalies along rays in random 3-D velocity models. In addition, each event is recorded by a random subset of the total set of stations. We quantify the performance of each method by characterizing both the absolute and relative errors in the computed locations as compared to the true locations, experimenting with both compact clusters and more distributed seismicity. The algorithms are also tested on real data, and their performances evaluated on a set of 785 events recorded by the Southern California Seismic Network (SCSN) from a region around the 2003 M=5.4 Big Bear earthquake, for which both phase pick and differential times resulting from waveform cross-correlation are available.
DE: 1734 Seismology
DE: 7200 SEISMOLOGY
SC: Seismology [S]
MN: 2003 Fall Meeting