HARP North Atlantic beaked whales: Echolocation click collection for machine learning

This dataset package is publicly available to advance the detection of beaked whale populations through passive acoustic monitoring. This extensive dataset from the western North Atlantic has been instrumental in developing a deep neural network (DNN) to improve the detection of ephemeral events. 

The volume of data generated by passive acoustic monitoring can be overwhelming, complicating efforts to quantify species occurrence for effective conservation and management. Automation of data processing using machine learning algorithms enables efficient species identification using their sounds. Beaked whale acoustic events, often infrequent and ephemeral, can be missed when co-occurring with signals of more abundant, and acoustically active species that dominate acoustic recordings. Large-scale classification efforts using DNN, which included beaked whales as one of many classes along with other odontocete species and anthropogenic signals, often missed the ephemeral events in favor of more common and dominant classes. By training the DNN to focus on the taxonomic family of beaked whales, we demonstrate that ephemeral events can be correctly and efficiently identified to species, even with few echolocation clicks.

This classification method can support improved estimation of beaked whale occurrence in regions of high odontocete acoustic activity, and this dataset package can be used to develop classification methods to improve data availability for these rare species. We kindly ask that if this dataset is used, authors cite this repository and the accompanying article that provides detailed information on how the dataset has been collected and processed.