Identifying suitable time periods for infrasound measurement system response estimation using across-array coherence

Abstract

SUMMARY Microbarometers deployed to measure atmospheric infrasound are often connected to, or housed within, a wind noise reduction system (WNRS). At infrasound arrays of the International Monitoring System (IMS), being deployed as part of Comprehensive Nuclear-Test-Ban Treaty verification measures, the WNRS typically comprises an 18 m diameter pipe array. Over the past decade an in situ method has been developed to estimate the measurement system (sensor + WNRS) response characteristics, by comparing its recordings with those made on a colocated reference sensor with known response and no WNRS. The method relies upon the identification of time periods for which the reference sensor and measurement system are subject to the same input pressure field. It has proven difficult to reliably identify such time periods at frequencies $\lt 0.1\,$ Hz using recordings at a single location, resulting in a negative bias in estimated measurement system gain values (the ‘dip artefact’) in the 0.02–0.1 Hz passband. The IMS is deploying arrays of microbarometers, and we show that a measure of across-array coherence can be used to identify time periods associated with acoustic signal propagation. Amplitude response estimates, using 1 yr of data from four IMS arrays, indicate that the dip artefact can be removed by retaining for analysis only those time periods that exhibit high across-array coherence. Moreover, our analysis confirms the hypothesis that the dip artefact is associated with time periods during which wind-generated pressure fluctuations dominate, leading to partial suppression of noise with length scales less than the extent of the WNRS. At two arrays within continental forests accurate amplitude responses are estimated across the 0.02–4 Hz passband, as acoustic signals at all frequencies can be identified. At two oceanic island arrays, the low numbers of time windows with above-noise acoustic signal in the 0.02–0.1 Hz passband make reliable response estimation at these frequencies difficult or impossible. It is recommended that the methodology for estimating the response of an infrasound measurement system at an array should incorporate a multichannel coherence measure; data centres may already routinely compute such measures in their signal detection algorithms.