Automatic Processing of Multi-Array Long-Period Seismic Data

Abstract

Summary Long-period seismic data from the three large arrays ALPA, LASA, and NORSAR are telemetered to the Seismic Array Analysis Center in Alexandria, Virginia, where they are processed in real time by employing a fast frequency-wavenumber (f-k) detection scheme. The processor incor- porates a novel f-k filtering technique which allows the separation of interfering signals and the extraction of small signals from noise. Surface wave magnitudes (Ms) are derived directly from the processor output. The Seismic Array Analysis Center (SAAC) in Alexandria, Virginia, receives the long-period seismic data transmitted continuously from the Large Aperture Seismic Array (LASA) in Montana and other large long-period arrays located in Alaska (ALPA) and Norway (NORSAR). As part of the evaluation of this network these data are processed to detect and measure Rayleigh waves associated with seismic events listed in the LASAISAAC Daily Event Summary. The choice of data processing methods was influenced by the fact that approxi- mately 100 events per day have to be processed in real time. Time domain beam- forming and plotting is a slow process which suffers from the disadvantage that the propagation azimuth and velocity must be assumed. Any disturbance visible on a particular beam may be associated with adjacent beams or with a side lobe. For these and other reasons, a frequency-wavenumber Cf-k) approach has been used for routine analysis of the LP data (Mack 1972). The heart of the processing scheme is the fast signal detection algorithm FKCOMB (Smart 1972), which makes used of frequency-wavenumber transforms. This refer- ence also describes the principles of thef-k analysis of transient waveforms. FKCOMB was originally used for the automatic processing of microbarograph array data (Smart & Flinn 1971). In this paper we merely summarize the method of analysis and outline how it is applied in the SAAC to detect Rayleigh waves and to obtain seismic surface wave magnitudes on a real-time basis. Frequency wavenumber analysis Frequency-wavenumber Cf-k) spectral estimation is a powerful technique for signal detection and waveform analysis of digitally recorded array data. The f-k spectrum of a given segment of array output is the squared modulus of the multi- dimensional Fourier transform of the data in time and space. Thef-k spatial represen- tation of a propagating wave is shown schematically in Fig. 1. Using discrete Fourier t Present address: Exploration Services Center, Mobd Oil Corporation, Dallas, Texas 75221.