Optimal filter parameters for low SNR seismograms as a function of stationand event location

Global seismic monitoring requires deployment of seismic sensors worldwide, in many areas that have not been studied or have few useable recordings. U sing events with lower signal-to-noise ratios (SNR) would increase the amou nt of data from these regions. Lower SNR events can add significant numbers to data sets, but recordings of these events must be carefully filtered. F or a given region, conventional methods of filter selection can be quite su bjective and may require intensive analysis of many events. To reduce this laborious process, we have developed an automated method to provide optimal filters for low SNR regional or teleseismic events. As seismic signals an often localized in frequency and time with distinct time-frequency characte ristics, our method is based on the decomposition of a time series into a s et of subsignals, each representing a band with f/Delta f constant (constan t Q). The SNR is calculated on the pre-event noise and signal window. The b and pass signals with high SNR are used to indicate the cutoff filter limit s for the optimized filter. Results indicate a significant improvement in S NR, particularly for low SNR events. The method provides an optimum filter which can be immediately applied to unknown regions. The filtered signals a re used to map the seismic frequency response of a region and may provide i mprovements in travel-time picking, azimuth estimation, regional characteri zation, and event detection. For example, when an event is detected and a p reliminary location is determined, the computer could automatically select optimal filter bands for data from non-reporting stations. Results are show n for a set of low SNR events as well as 379 regional and teleseismic event s recorded at stations ABKT, KIV, and ANTO in the Middle East. (C) 1999 Els evier Science B.V. All rights reserved.