HR: 1340h
AN: S33A-0303 [Abstracts]
TI: Two cross-correlation techniques applied to volcanic seismic waveforms
AU: * DeShon, H R
EM: hdeshon@geology.wisc.edu
AF: Department of Geology and Geophysics, University of Wisconsin-Madison,
1215 W. Dayton St., Madison, WI 53706
United States
AU: Rowe, C A
EM: char@lanl.gov
AF: Los Alamos National Laboratory, EES-11 M.S. D408, Los Alamos, NM 87545
United States
AU: Thurber, C H
EM: thurber@geology.wisc.edu
AF: Department of Geology and Geophysics, University of Wisconsin-Madison,
1215 W. Dayton St., Madison, WI 53706
United States
AB:
Two cross-correlation methods that employ very different mathematical approaches are compared: bispectrum-verified relative
lag estimates and cross-coherency-determined lags inverted for pick adjustments.
The bispectrum method is used to verify results from time-domain cross-correlation for cases in which correlation
coefficients fall below pre-determined quality thresholds. Bispectrum cross-correlation, or cross-correlation in the third
order spectral domain, identifies waveform similarity in cases where traditional time-domain methods fail due to correlated
noise contamination such as wind or site effects. For correlations passed through verification, the method incorporates
cross-spectral information for subsample precision in the output differential times. The algorithm produces differential
times, intended for use with the double-difference family of earthquake relocation and tomography methods.
The cross-coherency correlator calculates adaptive, cross-coherency-weighted correlation lags. Both integer and subsample
cross-correlation are employed. Following correlation, a dendrogram-based pair-group classification scheme identifies
clusters of similar waveforms. Correlation lag differentials are inverted for consistent pick adjustments within clusters,
using a 1-norm conjugate gradient solver, providing adjusted absolute phase picks for subsequent use in any location or
tomography algorithm.
We apply the methods to local seismicity recorded at Redoubt Volcano from 1989-1994 during and after the 1989/90 eruption.
Waveforms recorded at volcanoes are often noisy and contain distinct spectral components that arise from complex interactions
of tectonic and magmatic processes. Volcano seismic networks typically have few stations and often marginal coverage,
providing challenges for earthquake location in a complex, three-dimensional setting. We compare the performance of the two
methods, measured primarily by a direct comparison of correlation lags and earthquake relocations, but also by the success of
both techniques in robustly segregating families of repeating earthquakes.
DE: 7230 Seismicity and tectonics (1207, 1217, 1240, 1242)
DE: 7280 Volcano seismology (8419)
DE: 7290 Computational seismology
DE: 8419 Volcano monitoring (7280)
SC: Seismology [S]
MN: Fall Meeting 2005