A Public Dataset of Annotated Orcinus orca Acoustic Signals for Detection and Ecotype Classification

Area of Science:

• Marine Biology
• Bioacoustics
• Conservation Science

Background:

• Killer whales (Orcinus orca) display significant ecological and genetic diversity across three Northeast Pacific populations: Resident, Bigg's (Transient), and Offshore.
• Distinct foraging, social structures, and vocal repertoires among populations complicate monitoring and conservation.
• Accurate identification of killer whale vocalizations is crucial for ecological studies and management.

Purpose of the Study:

• To create a comprehensive, annotated dataset of killer whale vocalizations for improved detection and classification.
• To facilitate the development of robust machine learning models for ecological monitoring.
• To support conservation initiatives and deepen the understanding of killer whale acoustic communication.

Main Methods:

• Compiled acoustic recordings from diverse sources over 11 years across Alaska, British Columbia, and Washington.
• Utilized multiple hydrophone systems for data collection.
• Annotated recordings with detailed metadata, including confounding signals, to aid classification.

Main Results:

• Developed a comprehensive dataset of killer whale vocalizations with extensive annotations.
• Dataset addresses challenges in distinguishing killer whale calls from other marine sounds and noise.
• Included specific examples of confounding signals to enhance model robustness.

Conclusions:

• The curated dataset significantly improves the potential accuracy of killer whale detection algorithms.
• This resource is vital for advancing conservation efforts for distinct killer whale populations.
• Enhances the study of killer whale acoustic communication and ecological diversity.