An Automatic, Adaptive Algorithm for Refining Phase Picks in Large Seismic Data Sets

Abstract

We have developed an adaptive, automatic, correlation-and clusteringbased method for greatly reducing the degree of picking inconsistency in large, digital seismic catalogs and for quantifying similarity within, and discriminating among, clusters of disparate waveform families. Innovations in the technique include (1) the use of eigenspectral methods for cross-spectral phase estimation and for providing subsample pick lag error estimates in units of time, as opposed to dimensionless relative scaling of uncertainties; (2) adaptive, cross-coherency-based filtering; and (3) a hierarchical waveform stack correlation method for adjusting mean intercluster pick times without compromising tight intracluster relative pick estimates. To solve the systems of cross-correlation lags we apply an iterative, optimized conjugate gradient technique that minimizes an L 1-norm misfit. Our repicking technique not only provides robust similarity classification-event discrimination without making a priori assumptions regarding waveform similarity as a function of preliminary hypocenter estimates, but also facilitates high-resolution relocation of seismic sources. Although knowledgeable user input is needed initially to establish run-time parameters, significant improvement in pick consistency and waveform-based event classification may be obtained by then allowing the programs to operate automatically on the data. The process shows promise for enhancing catalog reliability while at the same time reducing analyst workload, although careful assessment of the automatic results is still important.

Improving Magnitude Detection Thresholds Using Multi-Station Multi-Event, and Multi-Phase Methods

“…Still most work and practical application has concentrated on the STA/LTA technique. The greatest advances of the application of correlation in recent years have largely been seen in location estimation, improving arrival time measurements of seismic events starting with pairs of similar events or doublets and currently working up to relocating hundreds of thousands of events (e. g. Poupinet et al 1984;Fréchet 1985;Ito 1985;Frémont & Malone 1987;Deichmann & Garcia-Fernandez 1992;Got et al 1994;Dodge et al 1995;Nadeau et al 1995;Shearer 1997;Lees 1998;Rubin et al 1999;Waldhauser et al 1999;Waldhauser & Ellsworth 2000;Phillips 2000;Rowe et al 2002;Schaff et al 2002;Moriya et al 2003;Shearer et al 2005;, Waldhauser & Schaff 2007. As remarked earlier, correlation similarity thresholds much lower than needed for location purposes may be adequate for detection.…”

Mount Hood Earthquake Activity: Volcanic or Tectonic Origins?

“…It has been used extensively to relocate earthquakes in volcanic regions. Examples include Rowe et al (2002) for Montserrat, West Indies, Rubin et al (1998) for Kilauea, and Musumeci et al (2002) and Fremont and Malone (1987) for Mount St. Helens. The method works because earthquakes with similar locations and source mechanisms produce similar seismograms at a given station.…”

Methods for Improving Seismic Event Location Processing

“…In the step 1 procedure, the entire seismogram was fihered and an eigenvector decomposition of the covariance matrix was appHed to the 3-component data (Rowe et al, 2002) prior to the calculation of Hilbert-envelopes for the principal direction. Test event data for each of the master stations were cross-correlated against 161 GT events with the position of the CC global maximum determining the step 1 location (Fig.…”

“…Average standard deviations of GT locations for the 19 test events indicate good control on latitude and longitude (±0.015° and ±0.02°) and moderate control on depth (±1.23 km). We chose GT solutions with a mean errors (absolute value) of 0.03°, 0.04°, and 1.26 km.In the step 1 procedure, the entire seismogram was fihered and an eigenvector decomposition of the covariance matrix was appHed to the 3-component data (Rowe et al, 2002) prior to the calculation of Hilbert-envelopes for the principal direction. Test event data for each of the master stations were cross-correlated against 161 GT events with the position of the CC global maximum determining the step 1 location (Fig.…”

“…Many catalog P picks required manual adjustment. We then prepared WCC data based on 3-and 5-second P windows (Rowe et al, 2002). A subset of non-test event data (P and S NCSN phase catalog times and WCC P-pick adjustments) augmented with test event data allows clustering of the test event with adjacent reference events.…”