SM2BAT自动声学识别软件在新热带地区蝙蝠调查中的应用
Abstract
Bat populations are known to be affected by anthropogenic activities because bats are an extremely diverse group occupying almost all available niches in terrestrial environment. Hence, bats are considered bioindicators to monitor changes in the environment, but their value as such also depends on the ease to monitor and detect demographic trends in their populations. The long-term interest of researchers in the acoustic of bats results from the fact that it is a non- invasive, time-efficient method to monitor spatiotemporal patterns of bat diversity and activity.The analysis of sounds emitted by organisms has been considered useful to gain insight into species-specific biotic and abiotic interactions, which can further be applied to conservation. Besides manual identifications of bat calls, some automated species identification programs using regional call classifiers have been introduced into the market as an effective tool in the monitoring of bat populations. Most of these programs have not been validated using field data. This study evaluates the reliability of two automated software, SonoChiro, and Kaleidoscope Pro, in comparison to manual identifications of field data collected from the Neotropical region. There was low agreement between the two automated methods at the species level, fair agreement at the genus level and moderate agreement at the family level. There was also a significant difference between the proportions of correctly identified calls of the two-automated software at the species level identifications. Major challenges for using automated identification software include the need for comprehensive call libraries of the regions under scope; significant opportunities, on the other hand, include the widespread possibility to monitor spatiotemporal patterns of bat activity. Overall, there are serious gaps that preclude a widespread application of automated programs ecological and conservation studies of bats, but it has the potential to serve as a useful tool.
摘要：
众所周知，蝙蝠种群受到人为活动的影响，因为蝙蝠是一个极其多样化的群体，几乎占据了陆地环境中所有可用的生态位。因此，蝙蝠被认为是监测环境变化的生物指标，但它们的价值也取决于监测和检测其种群人口趋势的难易程度。研究人员对蝙蝠声学的长期兴趣源于这样一个事实，即这是一种非侵入性、省时的方法来监测蝙蝠多样性和活动的时空模式。分析生物体发出的声音被认为有助于深入了解物种特异性的生物和非生物相互作用，这可以进一步应用于保护。除了手动识别蝙蝠叫声外，一些使用区域叫声分类器的自动物种识别程序也被引入市场，作为监测蝙蝠种群的有效工具。这些程序中的大多数尚未使用现场数据进行验证。本研究评估了两种自动化软件SonoChiro和Kaleidoscope Pro的可靠性，并与从新热带地区收集的现场数据的手动识别进行了比较。两种自动化方法在物种水平上的一致性较低，在属水平上一致性较好，在科水平上一致度适中。在物种级别的识别中，两种自动化软件正确识别的呼叫比例也存在显著差异。使用自动识别软件的主要挑战包括需要范围内区域的综合呼叫库；另一方面，重要的机会包括监测蝙蝠活动的时空模式的广泛可能性。总体而言，存在严重的差距，阻碍了自动化程序在蝙蝠生态和保护研究中的广泛应用，但它有可能成为一种有用的工具。
关键词：SM2+声学记录器，Wildlife Acoustics，野生动物声学记录，动物被动声学监测，声景生态学、快速生物多样性评估、生态声学
