Using passive acoustic monitoring and BirdNET to identify drivers of bird populations: proposed analysis pipeline and application in an agricultural context

Abstract: The combination of Passive Acoustic Monitoring and automated bird vocalization
identification tools such as BirdNET offers great opportunities to document the factors
structuring bird communities and populations. However, the current lack of case studies using this approach raises questions regarding the collection and processing of acoustic recordings, as well as the use of BirdNET generated data while ensuring the reliability of inferred ecological relationships. We propose a general pipeline to identify and mitigate the effects of confounding factors influencing both bird detectability and BirdNET performance. This pipeline was applied in a case study investigating the effects of agricultural practices in apple orchards on Parus major and Erithacus rubecula, the two-dominant species in this system and potential pest regulators. Our results demonstrate that although Passive Acoustic Monitoring combined with BirdNET is not yet a turnkey method and requires careful requires careful data filtering and optimization of classification confidence from BirdNET to ensure robust ecological inference, it can be a powerful tool for studying the factors driving bird populations when such precautions are implemented.