Nocturnal soundscapes - updated version

Five non-consecutive nights were selected from each of the twelve sites (January 8th, 17th, 21st, 26th, and February 2nd of 2024) across two reserves in Australian Eucalyptus open-forest. Nights were selected based on manual inspections of the recording quality, aiming to minimize the effects of geophony (nights with calm weather conditions with minimum wind). From each night’s recording, a two-hour window was used for data extraction starting one hour after sunset to minimize contributions from dusk-active species (some of which are also diurnally active) while targeting the onset of nocturnal Orthopteran activity (crickets, katydids, grasshoppers, and locusts) during the reproductive (summer) season. Each two-hour window was segmented into 1-minute clips at 10-minute intervals, resulting in 12 1-minute clips per night, from which one 1-minute clip was randomly selected for analysis. With six sites per study area (and five nights per site), this resulted in 60 1-minute sample recordings (30 per study area) providing a comprehensive temporal and spatial coverage of the summer soundscape within the broader study region. Data quality was further improved by applying a band-pass filter to attenuate low-frequency energy (<300 Hz), reducing bias from anthropogenic noise (highway and railway). Sonotypes were manually annotated by listening to recordings and examining spectrograms within each 1-minute sample. Four taxon groups were identified: frogs, nocturnal birds, and mammals below 4,000 Hz (with only a bat sonotype annotated above 10,000 kHz); and Orthopterans (crickets, grasshoppers, katydids, and locusts) between 1,400 and 20,000 Hz (with the majority of these orthoptera calls between 3,500 to 10,000 Hz). A presence–absence sonotype matrix was compiled to assess richness (rows = site × night recordings; columns = sonotypes; 1 = detected, 0 = absent). All soundscapes present across the 60 1-minute samples were annotated and assigned to the four taxon groups (Orthopterans, birds, frogs, mammals) and to “other” (noise from anthropophony or geophony; presence– absence sonotype matrix in S1 Appendix).

本研究从澳大利亚桉树（Eucalyptus）开放林带的两个保护区内的12个采样点中，各选取5个非连续夜间（2024年1月8日、17日、21日、26日及2月2日）作为研究对象。夜间的选取基于录音质量的人工抽检，旨在最大程度降低环境声（geophony）的干扰，即优先选择风力微弱的晴朗无风夜晚。针对每个夜间的录音，我们选取日落1小时后开始的2小时窗口进行数据提取，以规避晨昏活动物种（部分物种同时兼具昼行性）的干扰，同时聚焦繁殖季（夏季）夜行性直翅目昆虫（Orthopteran，包括蟋蟀、螽斯、蚱蜢及蝗虫）的活动起始时段。每个2小时的录音窗口以10分钟为间隔被切分为1分钟片段，单夜间共得到12条1分钟片段，随后从中随机选取1条用于后续分析。每个研究区域包含6个采样点，每个采样点对应5个夜间，最终共得到60条1分钟样本录音（每个研究区域30条），实现了研究区域夏季声景的全面时空覆盖。为进一步提升数据质量，我们采用带通滤波器滤除300Hz以下的低频能量，以降低人为噪声（公路与铁路交通噪声）带来的分析偏差。我们通过收听录音并检视每条1分钟样本的声谱图，对声型（sonotype）进行人工标注。最终划定四类分类群：4000Hz以下的蛙类、夜行性鸟类与哺乳类（仅在10000kHz以上标注蝙蝠声型）；以及1400Hz至20000Hz频段内的直翅目昆虫（蟋蟀、蚱蜢、螽斯及蝗虫），其中绝大多数直翅目昆虫的鸣唱信号位于3500Hz至10000Hz之间。我们构建了声型有无矩阵以评估物种丰富度，其中行代表采样点×夜间录音样本，列代表声型，1表示检测到对应声型，0表示未检测到。60条1分钟样本中的全部声景均被标注并归类至上述四类分类群（直翅目昆虫、鸟类、蛙类、哺乳类）以及“其他”类别（人为声或环境声，声型有无矩阵详见附件S1）。