Sound anomalies of Cornell Swift recorders affect ecoacoustic studies, and a workaround solution

Abstract

AbstractThe Swift Terrestrial Passive Acoustic Recording Unit from The Cornell Lab of Ornithology, running firmware v. STM32 0.18.6.3, produced an initial 4‐sec sound anomaly in each sound (wave file) recording, created by the power‐saving features of the unit as it switches from standby to record mode. The sound anomaly had a statistically significant impact on several soundscape indices calculated from the recordings. Here, as a case study of identifying and solving this problem, I dissected the nature of the anomaly and analyzed the variable effects it has on calculated ecoacoustic soundscape indices. I used a sample of 150, 10‐min sound files, recorded during my ecoacoustics study in central boreal Alaska during 2019 (June‐August) and 2020 (April‐September), stratified by several landscape conditions and by types of sounds representing anthrophony, biophony, and geophony conditions. The sound anomaly statistically significantly biased the calculations of 7 of 13 ecoacoustic indices analyzed from all of these landscape and soundscape conditions. There is no simple correction factor that can be applied to the calculated index values to account for the effects of the anomaly. I suggest several workarounds, notably to automate a procedure to delete a specified segment of each sound file to eliminate the anomaly prior to soundscape analysis, and in general to watch and correct for such anomalies when using Autonomous Recording Units recordings in ecoacoustic analyses.