Data Sheet 1_Artificial clicks (Porpoise ALert) affect acoustic monitoring of harbour porpoises and their echolocation behaviour.pdf

IntroductionIn 2021, a continuous acoustic monitoring programme was initiated in the Baltic Sea of Schleswig-Holstein to determine the occurrence and seasonal distribution of harbour porpoises. At the same time, fishers in this area applied acoustic devices (Porpoise ALert, PAL) generating artificial porpoise clicks to reduce bycatches in set-net fisheries. The underlying hypothesis was that signals from porpoise acoustic warning devices (PALs) might be misinterpreted by the click loggers (C-POD) as genuine porpoise clicks, potentially leading to an increase in detections. The study aimed to determine whether PALs were being recorded at the monitoring stations, and to identify effective methods for filtering out artificial signals.
MethodsTherefore, we deployed an array of 11 C-PODs at distances between 50 and 350 m to a duty-cycled PAL in the middle over a period of 3 months. A sophisticated machine learning approach was employed that was able to discriminate natural porpoise signals from artificial PAL signals using the full click sequence parameters.
ResultsThe trained algorithm showed remarkable efficiency in discriminating between artificial PAL signals and natural harbour porpoise clicks, demonstrating good sensitivity (99.74%) and accuracy (97.12%) in the test dataset.
DiscussionThe consequences for compromised monitoring are significant, particularly in regions with low harbour porpoise densities, where artificial signals may influence the interpretation of diurnal and seasonal aspects of natural harbour porpoise behaviour, leading to misinterpretations. The effectiveness of management measures depends on the availability of reliable monitoring data, which is essential given the urgent need to improve the conservation of harbour porpoises, which are declining in the western Baltic Sea within the waters of Schleswig-Holstein. Finally, the study design was maximised to provide further information on PAL functionality and effectiveness as warning devices. The results revealed a reduction in the number of porpoise clicks during PAL operation, and changes in echolocation patterns characterised by increased minimum Inter-click-intervals (ICI), suggesting a shift from foraging or communication to orientation activity, and decreased maximum ICI, indicating enhanced long-range orientation. The function of these devices on echolocation behaviour remains therefore unclear, as it is not known whether they act solely as an alarm or rather as a deterrent.

引言
2021年，石勒苏益格-荷尔斯泰因州的波罗的海海域启动了一项持续性声学监测项目，旨在确定港湾鼠海豚（harbour porpoise）的出现情况与季节分布。与此同时，该区域的渔民正在使用人工模拟港湾鼠海豚点击声的声学装置（海豚警报装置，Porpoise ALert, PAL），以降低定置网渔业的兼捕率。本研究的核心假设为：点击记录器（C-POD）可能会将海豚警报装置（PAL）发出的信号误识别为真实的港湾鼠海豚点击声，进而可能导致检测数量虚增。本研究旨在确认监测站点是否记录到了PAL设备发出的信号，并筛选出可有效滤除人工信号的方法。
方法
为此，我们在占空比运行的PAL设备周边50至350米的范围内，部署了由11台C-POD组成的监测阵列，监测周期为3个月。本研究采用了一种先进的机器学习方法，可通过完整的点击序列参数区分天然港湾鼠海豚信号与人工PAL信号。
结果
经训练的算法在区分人工PAL信号与天然港湾鼠海豚点击声时表现出优异的识别效率，在测试数据集上展现出良好的灵敏度（99.74%）与准确率（97.12%）。
讨论
人工信号对监测结果的干扰影响显著，尤其在港湾鼠海豚种群密度较低的区域——此类区域中，人工信号可能会干扰对港湾鼠海豚自然行为的昼夜与季节特征解读，进而引发误判。渔业管理措施的有效性依赖于可靠的监测数据，而鉴于西部波罗的海石勒苏益格-荷尔斯泰因州海域的港湾鼠海豚种群数量持续下降，获取可靠监测数据的需求尤为迫切。本研究的设计旨在进一步探究PAL设备作为警报装置的功能与有效性。研究结果显示，PAL设备运行期间，检测到的港湾鼠海豚点击声数量有所减少；同时，回声定位模式出现变化：最小点击间隔（ICI）升高，提示其行为从觅食或交流转向定向活动，而最大点击间隔则有所降低，表明其长距离定向能力增强。不过，此类设备对港湾鼠海豚回声定位行为的具体影响仍不明确，目前尚不清楚其仅起到警报作用，还是兼具驱避效果。