Characterization of global seismograms using an automatic-picking algorithm

Abstract

Abstract An automatic phase picker is useful for quickly identifying and timing phase arrivals in large seismic data bases. We have developed an automatic phase picker that is sensitive to small changes in amplitude and applied it to over 7 yr of global data distributed by the National Earthquake Information Center (NEIC). Our phase-picking algorithm is based on a short-term-average to long-term-average ratio (STA/LTA) taken along an envelope function generated from the seismogram. The algorithm returns arrival times and corresponding pick qualities. The procedure requires few input parameters and is easily adapted to various types of data. We produce global travel-time plots from both high-frequency (20- or 40-Hz sample rate) and low-frequency (1-Hz sample rate) data. These plots clearly image the predominant high- and low-frequency phases in the NEIC data base. Picks made from the long-period seismograms are less precise, but they reveal far more phase arrivals than the short-period picks. A number of phases resulting from reflections and phase conversions at upper mantle discontinuities can be identified in the low-frequency picks; however, a search of the short-period picks for upper mantle discontinuity phases, between P and PP and prior to P′P′, has so far been unsuccessful. In the long-period S and SS picks, we observe a discrepancy in SV and SH travel times, a possible result of upper mantle anisotropy. To check the accuracy and consistency of our algorithm, we present comparisons between hand-picked times and automatic-picked times for identical seismograms. Travel-time residuals from the short-period automatic picks and data reported to the International Seismological Centre (ISC) picks exhibit a comparable amount of scatter. Histograms of the ISC residuals and automatic-pick residuals are similar in shape and width for P and PcP. These observations suggest that human picking errors are not a major contributor to the scatter observed in ISC travel times, although direct comparisons between ISC reported picks and automatic picks on particular seismograms occasionally identify operator mispicks.