HR: 09:40h
AN: S31C-07 [Abstracts]
TI: Empirically Designed Filters for Matching Synthetic Seismograms to the Observed Seismograms
AU: Marshall, M
EM: margaret.e.marshall@saic.com
AF: SAIC/OSD, 1710 Saic Dr.
M/S 1-11-15, Mclean, VA 22102
United States
AU: * Salzberg, D H
EM: david.h.salzberg@saic.com
AF: SAIC/OSD, 1710 Saic Dr.
M/S 1-11-15, Mclean, VA 22102
United States
AB:
We are demonstrating a technique which designs empirical filters that, when applied to synthetic seismograms, results in much
better agreement between a synthetic waveform and the observed waveform. The filter is computed by deconvolving a synthetic
waveform for a specific event - station pair from the observed data. Applying the empirical filter to a
synthetic waveform for another event provides a more realistic synthetic waveform for the second event. Since the assumption
is that the misfit between the observation and the synthetic waveform is constant between events, the empirical filter from
multiple events can be combined using a weighted average. The weighing functions consist of two terms: the amplitude spectrum
of the synthetic waveforms divided by the amplitude spectrum of the pre-signal noise. The weighting function minimized both
the impact of small spectral values from the synthetic waveforms and the effects of noise. Applying the empirical filters to
synthetic waveforms for other events allows us to model the observed waveforms at frequencies up to 1 Hz. To demonstrate the
approach, empirical filters for the Parkfield, California region were computed for three stations based several larger
aftershocks. The empirical filters were then validated by using them to simulate the waveforms for other Parkfield
aftershocks. Comparing both the synthetic waveforms and the empirically filtered synthetic waveforms to the observed data
demonstrates the advantages of using the empirical filters. The empirically filtered synthetic waveform technique is useful
for a variety of applications. We have applied it to full waveform location, and have been able to obtain hypocenter
locations which are accurate to within 1 - 2 km from ground truth using only three regional (200 km <
range < 1000 km) stations. In addition, we have modeled the Parkfield mainshock by computing the synthetic seismograms at
each point of Chen Ji's finite fault model to simulate the mainshock waveforms. This work is Sponsored by
Air Force Research Laboratory Contract No. FA8718-05-C-0019
DE: 7215 Earthquake source observations (1240)
DE: 7290 Computational seismology
SC: Seismology [S]
MN: Fall Meeting 2005