Listening to the Night Sky: Characterizing Nocturnal Bird Migration in a Globally Important Flyway Using Bioacoustics and Machine Learning

This repository accompanies the manuscript: "Listening to the Night Sky: Characterizing Nocturnal Bird Migration in a Globally Important Flyway Using Bioacoustics and Machine Learning" The dataset provides the materials required to reproduce the workflow presented in the manuscript and to apply the retrained BirdNET classifier to new recordings. Contents This repository includes: A custom BirdNET classifier trained to identify nocturnal flight calls (NFCs) of 35 common migratory bird species and 6 background-noise classes. Training manifests listing all recordings used for model development, including their original source and metadata. Example labeled audio recordings collected during this study for testing the classifier. Processed datasets used in the statistical analyses. Python and R scripts for running the classifier, evaluating model performance, and reproducing the analyses and figures presented in the manuscript. Training data The classifier was trained using a dataset of 8,280 labeled recordings compiled from multiple sources: approximately 40% edited recordings from the Xeno-canto database; approximately 30% edited recordings from the Cornell Lab of Ornithology Macaulay Library; and approximately 30% validated recordings collected during this study. The original training recordings obtained from Xeno-canto and the Macaulay Library are not redistributed in this repository in order to comply with the licensing terms of the original contributors and repository policies. Instead, complete manifests of all training recordings are provided to ensure transparency and reproducibility. Citation If you use this dataset, please cite both this Zenodo record and the associated publication.