The response behaviour of bats to motion-controlled LED lighting of bicycle trails

Additional notes on methods used Recording of echolocation calls We used Batloggers A+ (Elekon AG, Luzern Switzerland) that were fixed with cable ties to the lamppost at about 2.5 m height above ground to record bat echolocation calls. The microphone was fixed about 20 cm above the Batlogger box, facing away from the bicycle trail. Batloggers were set to automatically record sound (mode: Crest Adv, Crest factor = 7, min frequency = 15 kHz, max frequency = 155 kHz) from up to 15 minutes before sunset and after sunrise using pre- and post-trigger times of 100 ms and 400 ms, respectively. Analysis of acoustic recordings We used the software BatExplorer (version 2.1.10.1, Elekon AG, Luzern Switzerland) to identify automatically the emitting species based on the recorded echolocation calls. We first optimized the species identification by adjusting the species library "European bats DE" of the BatExplorer to the local bat fauna. Then, we analysed the accuracy of the automatic identification by manually checking a subset of recordings. Based on the results of this validation, we derived rules about when automatic species identifications may prove acceptable without further manual checks. Whenever manual checks were conducted, we examined echolocation call parameters (start, peak, and end frequency, repetition rate, and shape of the call; spectrogram settings: Hamming window, FFT = 1024, overlap = 97 %), and compared them to values and characteristics in published literature (e.g., Barataud et al. 2015; Miller and Degn 1981; Obrist et al. 2004; Parsons and Jones 2000; Russo and Jones 2002; Skiba 2003). In general, we considered in our analysis only recordings for which BatExplorer detected at least three echolocation calls when using the following threshold criteria: Crest factor = 6, min. call length = 1 ms, lower frequency limit = 15 kHz, upper frequency limit = 150 kHz, hysteresis = 0.95, min. call intensity = 10 % of loudest call in the recording, intensity tolerance = 0.2 %, frequency tolerance = 7 kHz, species separation = 3 kHz. Alternatively, we used recordings with less than three calls only if recordings were at least 1.26 seconds long based on an accuracy test of automatic call detection. Classification of bat species into functional guilds Due to the strong overlap in echolocation call characteristics between species of the genera Nyctalus, Eptesicus and Vespertilio, we classified all recordings from the respective species into the group of "NEV". Recordings of these species were lumped into the open-space foraging bat guild. Similarly, we grouped all recordings from species of the genus Myotis into the group of "Myo". We defined this group together with recordings from the species Plecotus auritus as narrow-space foraging bats. The species Pipistrellus pipistrellus, P. nathusii and P. pygmaeus were identified to species level, but later lumped into the functional guild of edge-space foraging bats. In addition, we counted the number of terminal buzzes in a recording. A terminal buzz is a very characteristic sequence of echolocation calls with a short duration, a decreasing interval length between subsequent pulses and decreasing peak frequency, indicating the attempted capture of an insect (Donald R. Griffin, 1958). Light measurements During recording nights, we measured the relative illuminance of LEDs with the Pendant® MX Bluetooth Temperature+Light Data Logger (MX2202) (Datenlogger-Store, Eichstetten, Germany). Hereby, light from other sources, such as moonlight, was measured as well. At each recording site, we tied the sensor to the lamppost a few cm above the Batlogger. The resolution of the light sensor in faint light is 1 lux. The logger was set to record from sunset to sunrise at an interval of 1 s. Thermal imaging and 3D-tracks of bats We used two thermal imaging cameras (FLIR Tau 2 LWIR Thermal Imaging Camera Core) equipped with a ThermalCapture-Modul (TeAx Technology TC 2.0). Analyses and calculations were performed following published procedures (Abdel-Aziz, Karara 1971, Hedrick et al. 2014) and using the software Thermoviewer 3.0.7 (TeAx Technology), easyWand, DLTdv5 and Matlab (Mathworks, Version R2022b). The illustration of 3D-tracks is based on the R-packages plot3D (version 1.4, Soetaert 2021) and plot3Drgl (version 1.0.4, Soetaert 2023).