An Efficient Neural Network Design Incorporating Autoencoders for the Classification of Bat Echolocation Sounds.

Bats are widely distributed around the world, have adapted to many different environments and are highly sensitive to changes in their habitat, which makes them essential bioindicators of environmental changes. Passive acoustic monitoring over long durations, like months or years, accumulates large amounts of data, turning the manual identification process into a time-consuming task for human experts. Automated acoustic monitoring of bat activity is therefore an effective and necessary approach for bat conservation, especially in wind energy applications, where flying animals like bats and birds have high fatality rates. In this work, we provide a neural-network-based approach for bat echolocation pulse detection with subsequent genus classification and species classification under real-world conditions, including various types of noise. Our supervised model is supported by an unsupervised learning pipeline that uses autoencoders to compress linear spectrograms into latent feature vectors that are fed into a UMAP clustering algorithm. This pipeline offers additional insights into the data properties, aiding in model interpretation. We compare data collected from two locations over two consecutive years sampled at four heights (10 m, 35 m, 65 m and 95 m). With sufficient data for each labeled bat class, our model is able to comprehend the full echolocation soundscape of a species or genus while still being computationally efficient and simple by design. Measured classification F1 scores in a previously unknown test set range from 92.3% to 99.7% for species and from 94.6% to 99.4% for genera.

Collision risk of bats with small wind turbines: Worst-case scenarios near roosts, commuting and hunting structures.

Small wind turbines (SWTs) have become increasingly common within the last decade, but their impact on wildlife, especially bats, is largely unknown. We conducted an operational experiment by sequentially placing a mobile SWT with five different operational modes at six sites of high bat activity, including roosts, commuting structures, and highly frequented hunting areas. Bat flight trajectories around the SWT were documented at each site during five consecutive nights using a specifically designed high-spatial-resolution 3D camera. The recordings showed high bat activity levels close to the SWT (7,065 flight trajectories within a 10-m radius). The minimum distance to the rotor of each trajectory varied between 0 and 18 m, with a mean of 4.6 m across all sites. Linear mixed models created to account for site differences showed that, compared to a reference pole without a SWT, bats flew 0.4 m closer to the rotor (95% CI 0.3-0.6 m) if it was out of operation and 0.3 m closer (95% CI 0.1-0.4 m) if it was moving slowly. Exploratory behavior was frequently observed, with many bats deviating from their original flight trajectory to approach the rotor. Among 7,850 documented trajectories, 176 crossed the rotor, including 65 while it was in motion. The collision of one P. pygmaeus individual occurred during the experiment. These results demonstrate that, despite the generally strong ability of bats to evade moving rotor blades, bat casualties at SWTs placed at sites of high bat activity can reach or exceed the current threshold levels set for large wind turbines. As SWTs provide less energy than large turbines, their negative impact on bats should be minimized by avoidance measures such as a bat-friendly site selection or curtailment algorithms.