The Influence of Spatial Filters on Infrasound Array Responses

• Published in: Pure and applied geophysics 2014-03, Vol.171 (3-5), p.575-585
• Main Authors: Brown, David J., Szuberla, Curt A. L., McCormack, David, Mialle, Pierrick
• Format: Article
• Language: English
• Subjects: Acoustics; Arrays; Atmospheric turbulence; Earth and Environmental Science; Earth Sciences; Environmental monitoring; Estimates; Filters; Fluid dynamics; Fluid flow; Geophysics/Geodesy; Infrasound; Microorganisms; Monitoring systems; Pipes; Plane waves; Turbulent flow
• ISSN: 0033-4553; 1420-9136
• DOI: 10.1007/s00024-012-0586-1

A spatial filter is often attached to a microphone or microbarometer in order to reduce the noise caused by atmospheric turbulence. This filtering technique is based on the assumption that the coherence length of turbulence is smaller than the spatial extent of the filter, and so contributions from turbulence recorded at widely separated ports will tend to cancel while those of the signal of interest, which will have coherence length larger than the spatial dimensions of the filter, will be reinforced. In this paper, the plane wave response for a spatial filter with an arbitrary arrangement of open ports is determined. It is found that propagation over different port-to-sensor distances causes out-of-phase sinusoids to be summed at the central manifold and can lead to significant amplitude decay and phase delays as a function of frequency. The determined spatial filter plane wave response is superimposed on an array response typical of infrasound arrays that constitute the International Monitoring System infrasound network used for nuclear monitoring purposes. It is found that signal detection capability in terms of the Fisher Statistic can be significantly degraded at certain frequencies. The least-squares estimate of signal slowness can change by up to 1.5° and up to 10 m/s if an asymmetric arrangement of low and high frequency spatial filters is used. However, if a symmetric arrangement of filters is used the least-squares estimate of signal slowness is found to be largely unaffected, except near the predicted null frequency.