Skip to main content
Dryad logo

Miniaturization eliminates detectable impacts of drones on bat activity

Citation

Kuhlmann, Kayla et al. (2022), Miniaturization eliminates detectable impacts of drones on bat activity, Dryad, Dataset, https://doi.org/10.5061/dryad.g4f4qrfs1

Abstract

A new way to survey wildlife populations may be possible with advancements in drones, or unmanned aerial vehicles (UAVs) that render aerial technology more accessible and promote surveying in inapproachable habitats. However, it remains unclear whether UAV disturbance deters animals, which would make this method inaccurate for data collection and hazardous to wildlife welfare. This study addresses the viability of UAV use for wildlife research by measuring the effects of UAV flight on acoustic bat detection and comparing bat activity in response to varying UAV models. Depending on the way UAVs effect bat detection rate, it may be possible to identify whether wildlife surveys should be done with UAVs and the drone models best suited for this purpose. The results reveal that larger and louder UAVs deterred significantly more bats, and the smallest and quietest model had no effect on bat detection. Indeed, drone noise was positively correlated with drone size, but drone size had little effect on the range of frequencies emitted. While detecting bats with small and quiet UAVs may be possible, complications still arise with acoustic detection and the species-specific effects of drone flight. The reliability of automatic identification with the acoustic detecting software is limited, as over a quarter of detections were triggered by non-bat noises yet still classified as bats (25.99%). Overall, using drones for wildlife detection should be approached with caution, as this study illustrates that some drones deter and disturb wild bats. If drones are used in wildlife habitat, consider flying smaller and quieter models, which are significantly less disturbing. Otherwise, large and loud drones will likely deter more animals and skew the results of the survey.

Methods

We performed trials starting 30 min after dusk from 21h00 to 0h00, following the procedure described by Ednie et al. (2021). Trials consisted of three phases where we acoustically recorded bat activity before, during, and after drone treatment. At each site, we placed a control detector 250-400m away from the treatment, enough distance to attenuate drone noise but remain in similar habitat. Upon arriving to one of our eight detection sites, we began the first phase of the trial by attaching the Echo Meter Touch 2 plugin acoustic bat detector (Wildlife Acoustics, Maynard, MA, USA) to a tablet and initiating recording with the corresponding Wildlife Acoustics app (Wildlife Acoustics, Maynard, MA, USA). Phase 1 consisted of audio recording from the ground with the Echo Meter Touch 2 for five minutes with no drone activity. After five minutes, we began Phase 2 by setting up one of three drones for the randomly chosen treatment. We either launched the drone to 15 m above the site of detection where the drone remained hovering; or we carried the drone to a launch site 50 m away, ascended it to 15 m above the launch site, and then flew the drone above the detection site. The drone hovered in place for five min, and afterwards was returned to its launch site. Once we powered off the drone, we continued to Phase 3, i.e. another five min of audio bat detection with no drone disturbance. This completed the trial, then we repeated the procedure at a new site, conducting one trial at each site every night of the experiment. Bat calls detected were automatically saved and identified to species using Kaleidoscope Pro Analysis Software (Wildlife Acoustics, Maynard, MA, USA) integrated in the Wildlife Acoustics app. 

Due to the difficulty differentiating between frequency spectrograms accurately, some species were combined: Myotis species: the eastern small-footed bat (Myotis leibii), the little brown bat (M. lucifugus), and the northern long-eared bat (M. septentrionalis) and Eptesicus complex: the big brown bat (Eptesicus fuscus) and the silver-haired bat (Lasionycteris noctivagans). Automatic identifications were then followed with blind manual classification to determine if all bat recordings were valid. We completed the manual classification using the spectrogram viewer in Kaleidoscope 5.4.1a (Wildlife Acoustics, Maynard, MA, USA), and the guide to acoustics for Québec bats (Fabianek, 2015); bats were identified by the shape and frequency range of their calls. False detections were excluded from and incorrect detections corrected for the new set of manually identified data.

Funding

Kenneth Molson Foundation

Natural Sciences and Engineering Research Council of Canada, Award: 241061