Nighthawk: Acoustic monitoring of nocturnal bird migration in the Americas

Abstract

Abstract Animal migration is one of nature's most spectacular phenomena, but migratory animals and their journeys are imperilled across the globe. Migratory birds are among the most well‐studied animals on Earth, yet relatively little is known about in‐flight behaviour during nocturnal migration. Because many migrating bird species vocalize during flight, passive acoustic monitoring shows great promise for facilitating widespread monitoring of bird migration. Here, we present Nighthawk, a deep learning model designed to detect and identify the vocalizations of nocturnally migrating birds. We trained Nighthawk on the in‐flight vocalizations of migratory birds using a diverse dataset of recordings from across the Americas. Our results demonstrate that Nighthawk performs well as a nocturnal flight call detector and classifier for dozens of avian taxa, both at the species level and for broader taxonomic groups (e.g. orders and families). It achieves an average precision score above 0.80 for 50 species and a mean average precision of 0.96 across 4 orders. The model accurately quantified nightly nocturnal migration intensity (80% variation explained) and species phenology (78% variation explained) and performed well on data from across North America. Incorporating modest amounts of additional annotated audio (50–120 h) into model training yielded high performance on target datasets from both North and South America (average precision on order Passeriformes >0.99). By monitoring the vocalizations of actively migrating birds, Nighthawk provides a detailed window onto nocturnal bird migration that is not presently attainable by other means (e.g. radar or citizen science). Scientists, managers and practitioners could use acoustic monitoring with Nighthawk for a number of applications, including: monitoring migration passage at wind farms; studying airspace usage during migratory flights; monitoring the changing migrations of species susceptible to climate change; and revealing previously unknown migration routes and behaviours. Overall, this work will empower diverse stakeholders to efficiently monitor migrating birds across the Western Hemisphere and collect data in aid of science and conservation.

Keywords

• acoustic monitoring
• artificial intelligence
• bioacoustics
• bird migration
• machine learning
• machine listening