National-scale acoustic monitoring of avian biodiversity and phenology

Abstract

AbstractBirds migrate over large spatial scales with complex dynamics which play out over extended time periods, making monitoring of phenology challenging with traditional biodiversity survey approaches. In this study, over a complete spring season, we collected 37,429 hours of audio from 28 networked sensors in forests across the latitudinal extent of Norway to demonstrate how acoustic monitoring can transform avian phenology monitoring. We used machine learning to automatically detect and identify bird vocalizations, and with expert validation found we were able to classify 55 species (14 full migrants) with over 80% precision. We compared audio data to existing avian biodiversity datasets and demonstrated that acoustic surveys could fill large data gaps and improve the temporal resolution at which metrics such as date of arrival for individual species could be estimated. Finally, we combined acoustic data with ecoclimatic variables from satellites and were able to map migratory waves of 10 species across the country at fine spatial resolutions (0.2 degrees). Our study demonstrates how acoustic monitoring can inexpensively and reliably complement existing national-scale biodiversity datasets, delivering high quality data which can support the design and implementation of effective policy and conservation measures.