Automated association of fin whale calls for localization using distributed acoustic sensing

Area of Science:

• Marine bioacoustics
• Oceanographic technology
• Signal processing

Background:

• Submarine cables offer extensive spatial coverage for ocean monitoring.
• Distributed Acoustic Sensing (DAS) on these cables is an underutilized resource for marine mammal detection.
• High-resolution DAS data necessitates advanced automated processing methods.

Purpose of the Study:

• To develop and evaluate an automated method for associating and localizing fin whale calls using DAS data from the Ocean Observatories Initiative (OOI) Regional Cabled Array (RCA).
• To assess the performance of the developed method in terms of precision, recall, and F1-scores.

Main Methods:

• Utilized DAS data collected by the OOI RCA, including fin whale vocalizations, offshore of central Oregon.
• Employed a grid search approach for associating acoustic arrival times between two nearshore cable sections (65 and 95 km).
• Applied Gabor filtering and spatial windowing to enhance noisy call associations and used weighted least squares inversion for localization.

Main Results:

• The automated method achieved high recall values (0.92-0.94) when enhanced with a far spatial window.
• Precision was initially low (0.3) but can be improved through subsequent track formation.
• Manual annotation of five ten-minute data subsets was used for performance evaluation.

Conclusions:

• The developed automated method shows significant potential for associating and localizing fin whale calls using submarine cable DAS.
• DAS technology on submarine cables represents a promising tool for continuous, large-scale marine mammal monitoring.
• Further development in track formation can mitigate precision limitations and enhance the utility of DAS for bioacoustic surveys.