Calling in the wilderness: the use of Passive Acoustic Monitoring in biodiversity surveys

科学背景BTO目前正处于白俄罗斯和乌克兰为期5年的令人兴奋的景观恢复计划的第二个年头，该计划名为“无国界荒野：创造欧洲最大的自然景观之一”。该项目旨在指定新的保护区并对现有保护区进行升级，在覆盖约580万公顷的更广泛的普里皮亚特/波莱西亚区域内，创建一个120万公顷的跨界保护和相互关联的核心区。然而，对野生动物，特别是夜间野生动物的大规模监测仍然具有挑战性。研究方法本博士项目将研究被动声监测(PAM)作为为夜间野生动物提供大规模基线数据的工具的潜力。具体地说，学生将在Prypiat和Polesia荒野地区部署声学记录器与分析声学数据。由于呼叫库对于建立受监督的自动分类器至关重要，将确定物种覆盖率方面的差距，并将其列为2020年实地工作的优先事项。学生将评估BTO建立随机森林分类器的现有方法，与新的深度学习算法(卷积神经网络，CNN)相关，以开发用于自动物种识别的健壮框架和工具。通过四季(2019-2022年)的数据，学生将评估该方法在提供关于重点分类群体的分布、相对丰度和栖息地要求的可靠数据方面的潜力。培训成功的候选人将接受以下方面的培训：被动生物多样性监测方法；大型、长期监测和声学数据库的构建、管理和分析；包括CNN的机器学习，并有望在统计建模方面获得高水平的能力。此外，学生还将获得实地研究和设计技能，包括大规模样本设计、小型哺乳动物的捕获和处理以及多分类群鉴定。

Scientific backgroundThe BTO is currently in its second year of an exciting 5-year landscape restoration program in Belarus and Ukraine - 'Wilderness without borders: creating one of the largest natural landscapes in Europe'. The project aims to designate new, and upgrade existing conservation areas, to create a transboundary protected and interconnected core area of 1.2 million ha, within the wider Prypiat / Polesia area covering approximately 5.8 million ha.Underpinning this process, it is crucial for decisions to be made on robust and representative assessment of the biodiversity and ecological value of the region. However, large-scale monitoring of wildlife, and particularly nocturnal wildlife remains challenging. Research methodologyThis PhD project will examine the potential of passive acoustic monitoring (PAM) as a tool for providing large-scale baseline data for nocturnal wildlife. Specifically, the student will combine the deployment of acoustic recorders in the Prypiat and Polesia wilderness area with analysis of acoustic data. As call libraries are essential for building supervised automatic classifiers, gaps in species coverage will be identified and prioritised for fieldwork effort in 2020. The student will evaluate the BTO's existing approach for building random forest classifiers, in relation to new deep learning algorithms (Convoluted Neural Networks, CNNs), to develop a robust framework and tools for automated species identification. With four seasons of data (2019-2022), the student will evaluate the potential of the approach for providing robust data on the distribution, relative abundance and habitat requirements of the focal taxonomic groups.TrainingThe successful candidate will receive training in passive biodiversity monitoring approaches; the construction, management and analyses of large, long-term monitoring and acoustic databases; machine-learning including CNN's and is expected to achieve a high level of competency in statistical modelling. Furthermore, the student will obtain field research and design skills including in large-scale sample design, small mammal trapping and handling, and multi-taxa identification.