Search

Machine learning tools capable of identifying animals by sound have proliferated, making the challenge of interpreting their outputs much more prevalent. These tools, like their predecessors, quantify prediction uncertainty with scores that tend to resemble probabilities but are actually unitless scores that are (generally) positively related to prediction accuracy in species-specific ways. BirdNET is one such tool, a freely available animal sound identification algorithm capable of identifying > 6,000 species, most of them birds. We describe two ways in which BirdNET "confidence scores"—and the output scores of other detector tools—can be used appropriately to interpret BirdNET results (reviewing them down to a user-defined threshold or converting them to probabilities), and provide a step-by-step tutorial to follow these suggestions. These suggestions are complementary to common performance metrics like precision, recall, and receiver operating characteristic. BirdNET can be a powerful tool for acoustic-based biodiversity research, but its utility depends on the careful use and interpretation of its outputs. [ABSTRACT FROM AUTHOR]

Ecological disturbances are becoming more extensive and intensive globally, a trend exemplified by 'megafires' and industrial deforestation, which cause widespread losses of forest cover. Yet the hypothesis that contemporary environmental disturbances are affecting biodiversity has been difficult to test directly.The novel combination of landscape‐scale passive acoustic monitoring, a new machine learning algorithm, BirdNET and improved Bayesian model‐fitting engines enables cohesive, community‐level before‐after, control‐impact studies of disturbances. We conducted such a study of a 2020 megafire in the Sierra Nevada, USA. We used a bespoke dynamic multi‐species occupancy modelling approach, which enabled us to account for imperfect detection, misclassifications, and to share information among species.There was no community‐level difference in colonization between burned and unburned forest. In contrast, the probability of site extinction in burned forest, 0.36, was significantly higher than in unburned forest, 0.12. Of the 67 species in our study, 6 (9%) displayed a positive colonization response to the fire, while 28 (41%) displayed a significant extinction response.We observed a 12% decrease in avian biodiversity 1 year post‐fire, and a substantial shift in community composition. However, in this ecosystem, many species display time‐dependent responses to the fire that are unobservable after just 1 year.Synthesis and applications. We have shown that three emerging conservation technologies, passive acoustic monitoring, machine learning animal sound identification algorithms, and advances in Bayesian statistical tools, can provide previously unattainable information about biodiversity responses to ecological change. Critically, our approach is transferrable and scalable, as the workflow is agnostic to species or ecosystem and each component is either freely available (all relevant software) or relatively inexpensive (recording hardware). Environmental change is unfolding rapidly, but new analytical techniques may help our understanding and—thus interventions—keep pace. [ABSTRACT FROM AUTHOR]