HR: 16:15h
AN: S22F-02    [PDF]
TI: MULTIPLE-EVENT LOCATION WITH STATION-CORRELATED PATH CORRECTIONS
AU: Rodi, W
EM: rodi@erl.mit.edu
AF: Massachusetts Institute of Technology, 42 Carleton St., Cambridge, MA 02142 United States
AU: * Myers, S C
EM: smyers@llnl.gov
AF: Lawrence Livermore National Laboratory, Box 808 L-205, Livermore, CA 94551 United States
AB: Multiple-event location methods solve jointly for the locations of a set of seismic events and travel-time corrections associated with the paths from the events to the observing stations. Many of these methods assume the path corrections for a given station are the same for all events, which effects an important constraint on the event locations but which restricts the method to event clusters of small aperture. A notable exception is the double-differencing method, which allows for de-correlation of path corrections for well separated pairs of events. However, these methods do not typically incorporate a model of complete or partial correlation of path corrections between seismic stations, and thus omit potentially valuable information for many common network geometries. This paper investigates a new approach to multiple-event location that models the correlation of travel-time corrections for each pair of paths as a function of both event and station separation. The approach generates path corrections from a ``universal correction function'', which is sampled at the event location and station location to produce the correction for a given path. The universal correction function is estimated from observed travel-time residuals using a geo-statistical interpolation technique (kriging), making use of prior covariance information to impose correlations between paths. We have tested this approach on a data set from the Nevada Test Site (NTS) comprising Pn and Pg arrival times from 64 stations and 74 explosions for which precise location parameters are available. In this data set some stations are separated by less than 10 km while the median nearest-neighbor distance is only 70 km. Our tests compare event locations obtained using station-uncorrelated and station-correlated path corrections, each estimated from the observed travel-time residuals relative to the IASP91 earth model. In our initial tests, the path corrections are derived using the known locations of the events and then used in a single-event location algorithm applied to each event. The resulting event mislocations are consistently smaller when station-correlated corrections are used. More realistic tests entail estimating the path corrections jointly with the event locations in a multiple-event location process. Preliminary tests of this type on the NTS data indicate that multiple-event location methods incorporating station-correlated path corrections yield modestly smaller mislocations than the traditional methods.
DE: 7215 Earthquake parameters
DE: 7219 Nuclear explosion seismology
DE: 7299 General or miscellaneous
SC: Seismology [S]
MN: 2003 Fall Meeting