Automatic measurement of shear wave splitting and applications to time varying anisotropy at Mount Ruapehu volcano, New Zealand

Abstract

[1] We present an automatic shear wave splitting measurement tool for local earthquakes, with the sole manual step of choosing an S arrival time. We apply the technique to three data sets recorded on Mount Ruapehu volcano in New Zealand that have previously been determined to have fast polarizations that vary in time and with earthquake depth. The technique uses an eigenvalue minimization technique, applied over multiple measurement windows. The dominant period of each waveform sets minimum and maximum window lengths. Cluster analysis determines the best solution among all the windows, and quality grading criteria assess the results automatically. When the same filters are used for events determined to be high quality from manual studies, the automatic technique returns virtually identical results. Scatter increases when the automatic technique chooses the best filter, but the average automatic results remain consistent with the manual results. When the automatic technique is used on sets that include data previously judged as poor quality, some stations yield distributions of fast polarizations that include peaks that were not present in previously published results. The difference may stem from two factors: automatic grading lets through some measurements that independent analysts consider poor quality, but also unconscious bias in the manual selection process may downgrade measurements that do not fit expectations. Nonetheless, the new objective analysis confirms changes in the average fast polarizations between 1994 and 2002 and between shallow and deep events. Therefore, this new technique is valuable for objective assessment of anisotropy and its variation in time.

The Erua earthquake cluster and seismic anisotropy in the Ruapehu region, New Zealand

“…Measurements that differed substantially across filters were removed, and at most one measurement is presented for each earthquake-station pair. Refer to Savage et al [2010b] for details on these quality control steps.…”

“…An automated program developed by Savage et al [2010b], and based on the algorithms of Silver and Chan [1991] and Teanby et al [2004], was used to perform all SWS measurements in this study. The program grades each measurement and marks any null measurements in which no splitting result is obtained.…”

Seismic detection of increased degassing before Kīlauea's 2008 summit explosion

“…The incidence angle of each ray at each station is calculated using the TauP Toolkit 28 with the one-dimensional velocity model 29 for the Kı %lauea region. In addition to the selection criteria enforced by the algorithm 27 , rays with incidence angles 435°from vertical were not included in the analyses as these lie outside the shear wave window, and S-P conversions at the surface could contaminate the waveforms. We observe that the steep velocity gradient in the top 0.5 km ensures that the majority of the local earthquakes that were recorded arrive at the stations with incidence angles o35°and so are included in the analysis.…”

“…It has been incorporated into code that conducts cluster analysis over a range of time windows to find the most stable result. The automated method 27 calculates the optimum three filters to apply to the data on the basis of signal-to-noise ratios and then determines the maximum and minimum time window around the S-wave arrival. These time window extremes are based on the dominant frequency of the first 3 s of the S-waveform.…”

Application of cluster analysis based on waveform cross-correlation coefficients to data recorded by ocean-bottom seismometers: results from off the Kii Peninsula

“…As an alternative approach, Schaff and Waldhauser (2005) calculated two different CCs per event pair using different time window lengths and accepted only results that produced consistent delay times for both. Similar methods have been introduced for measuring S-wave splitting (e.g., Savage et al 2010).…”