Using microphone arrays to investigate microhabitat selection by declining breeding birds

Abstract

Understanding the microhabitat preferences of animals can help managers to develop better conservation and recovery strategies, but is challenging. Traditional methods are limited by cost, accuracy, and human resources. In this study, we investigated avian microhabitat preferences using microphone arrays that are capable of accurately localizing vocalizing birds. Our objective was to identify the microhabitat associations of two common species in steep population decline, the Boreal Chickadee Poecile hudsonicus and the Cape May Warbler Setophaga tigrina. We deployed 68 eight-channel arrays at random locations in Labrador, Canada during the 2016 avian breeding season. We returned in 2017 to the 18 array locations where the target species had been detected the previous year and characterized the microhabitat at the exact locations where they had been detected. We also characterized the microhabitat at randomly determined control locations. Results show that Boreal Chickadees select trees with greater diameter-at-breastheight that are surrounded by greater stem density. We did not find evidence that Cape May Warblers exhibit microhabitat selection during song production. The study shows that microphone arrays are an effective tool for identifying preferred microhabitat that could be incorporated into future conservation or recovery strategies.

Acoustic localization of terrestrial wildlife: Current practices and future opportunities

“…Six studies calculated animal abundance, either by direct calculation of number of individuals (Frommolt & Tauchert, 2014; Hedley, Huang, & Yao, 2017; Spillmann et al, Willems, van Noordwijk, Setia, & van Schaik, 2017; Wahlberg et al, 2003; Wilson & Bayne, 2018) or indirectly by calibration of acoustic indices, as described by Stevenson et al (2015) (Thompson et al, 2009). Five studies used localization to infer territory boundaries or habitat use, including assessing animals' relationships with anthropogenic or natural habitat features (Ethier & Wilson, 2019; Hennigar, Ethier, & Wilson, 2019; Kershenbaum et al, 2019; Spillmann et al, 2017; Wilson & Bayne, 2018). Three studies separated animal sounds from background noise to improve species classification (Kojima, Sugiyama, Hoshiba, Suzuki, & Nakadai, 2017; Kojima, Sugiyama, Suzuki, Nakadai, & Taylor, 2016; Suzuki, Matsubayashi, Nakadai, & Okuno, 2016).…”

“…Field deployment typically requires multiple people and can be time‐consuming, especially at larger scales and with many ARUs. Ethier and Wilson (2019) reported that two people required 1–2 hr to deploy four GPS‐synchronized ARUs that were separated by 40 m.…”

Singing behaviour of ruby‐crowned kinglets Regulus calendula in relation to time‐of‐day, time‐of‐year and social context

“…Their locations were selected at random, but with the constraints that they were within 1 km of road or trail access, at least 500 m apart from each other, and not in a swamp, bog or body of water. The microphone arrays were also used in other projects investigating the relationships among habitat characteristics, anthropogenic disturbance, avian community composition and avian behaviour (Ethier and Wilson , Hennigar et al ). One of those studies involved broadcasting experimental noise or light, but those manipulations were only conducted at a given location after the data for the current study had been collected (Hennigar et al ).…”

“…This second step was repeated 100 times, with each iteration using a finer‐resolution lattice and a smaller study area (equal to the spatial resolution of the previous iteration) centred on the estimated origin of the song from the previous iteration. (Wilson et al , Ethier and Wilson , Hennigar et al ).…”

“…Second, we retained only those songs with a localization error of 0.1 or less. We chose 0.1 because previous research showed that 95% of sounds with a localization error of 0.1 or less are localized to within 10 m of their true locations (Ethier ). That research involved broadcasting frequency upsweeps (200 ms in length, logarithmic increase from 0.5 to 2.5 kHz, broadcast at 90 dB sound pressure level, as measured 1 m from the speaker) from in and around eight of our arrays (but after our trials were complete), localizing them with our localization software, and comparing the location estimates to the true locations of the speaker, as determined with our GNSS.…”