HR: 0830h
AN: S21D-0331 [PDF]
TI: ASSESSMENT OF REGIONAL-NETWORK SEISMIC LOCATION ACCURATY USING RELATIVE AND MASTER-EVENT TECHNIQUES AND
A NEVADA TEST SITE DATASET
AU: * Myers, S C
EM: smyers@llnl.gov
AF: Lawrence Livermore National Laboratory, Earth Sciences Division, Box 808
L-205, Livermore, CA 94551 United States
AU: Anderson, M L
EM: anderson@geo.arizona.edu
AF: University of Arizona, Geosciences
Gould-Simpson Building, Tucson, AZ 85721 United States
AU: Walter, W R
EM: bwalter@llnl.gov
AF: Lawrence Livermore National Laboratory, Earth Sciences Division, Box 808
L-205, Livermore, CA 94551 United States
AU: Ryall, F
EM: ryall@s160.llnl.gov
AF: Lawrence Livermore National Laboratory, Earth Sciences Division, Box 808
L-205, Livermore, CA 94551 United States
AU: Rodi, W
EM: rodi@erl.mit.edu
AF: Massachusetts Institute of Technology, 42 Carleton St., Cambridge, MA 02142 United States
AB:
We use a set of Nevada Test Site (NTS) nuclear explosions and the Grid Search Multiple Event Location (GMEL) algorithm to
test location accuracy achieved using master-event techniques. Arrivals from seventy-four nuclear tests recorded at 64
regional stations comprise the test dataset. Baseline relative location accuracy is established for 4 velocity models. We
find that location and known travel-time prediction accuracy (derived from known hypocenters) are highly correlated, but
root-mean-squared, a posteriori travel-time residuals (after relocation) are not a reliable indicator of either location or
travel-time prediction accuracy. Unlike most master-event algorithms, GMEL accommodates any number of master events, allowing
us to test the improvement achieved with more than one calibration event. For each of the 4 models, we test master-event
location accuracy by successively relocating events relative to 1, 5, 10, 37, and 67 fixed hypocenters. In the case where 1
event is fixed we successively relocate relative to each of the 74 events. For tests with more than one master event, we
randomly select the desired number of events in 74 instances. We find that, on average, one master event improves location by
58%. Using 5 and 10 master events improves location accuracy by 66% and 68%, respectively. Using 67 master events adds
a meager 4% improvement over 10 constrained events. From these tests we conclude that although calibration using one master
event is highly desirable, calibration from one event is susceptible to picking and other random errors that confound the
goal of estimated travel-time-prediction bias. Using more than one master event helps to further mitigate extraneous random
errors and results in more accurate determination of prediction bias. We find that between 5 and 10 calibration (master)
events within one spatial correlation length is an optimal number. More than 10 events are not needed and calibration with
fewer than 5 events may not provide sufficient averaging of random errors.
This work was performed under the auspices of the U.S. Department of Energy by the University of California, Lawrence
Livermore National Laboratory under Contract No. W-7405-Eng-48
DE: 7215 Earthquake parameters
DE: 7219 Nuclear explosion seismology
DE: 7299 General or miscellaneous
SC: Seismology [S]
MN: 2003 Fall Meeting