The application of multiply constrained minimum-variance adaptive beamforming to regional monitoring

Abstract

Abstract Broadband operation of small regional seismic arrays suggests the need for beams whose widths are nearly frequency independent and for an adaptive processor capable of suppressing nonstationary noise and interference. It is shown that multiply constrained minimum-variance (MCMV) adaptive beamforming when trained on signal-and-noise windows can achieve these objectives better than conventional beamforming. This paper presents the theory, implementation, characteristics, and performance of such a beamformer in the context of seismic monitoring for underground nuclear tests. The adaptive beamformer was applied to a large set of Pn arrivals observed on the NORESS array to estimate first- and second-order sample statistics of the array gain as a function of frequency and subarray size. The relative merit of adaptive beamforming was established by comparing the mean gain with that realized by conventional beamforming. These comparative results and other findings showed that the MCMV beamformer can (1) improve average signal-to-noise ratios from 0 to 8.5 dB, (2) better identify and suppress organized components in the noise field, and (3) better suppress interference in comparison to the conventional beamformer. The adaptive beamformer also produced peak narrowband gains on the larger subarrays in the neighborhood of 24 dB for specific evens. When implementational, seismic environmental, and practical aspects are added to the findings, important configurational and operational strategies are identified. The array gain statistics for both the adaptive and conventional beamformers are also useful for modeling the performance of monitoring networks.