Machine learning-enabled retrieval of ocean current speeds from acoustic flow noise in the 2023 NESMA Pilot Experiment

The Naval Postgraduate School deployed a network of Moored Autonomous Noise Recorders (MANRs) in the Northwest Atlantic near the Atlantis II Seamounts during the 2023 NESMA Pilot Experiment. This study examines the retrieval of ocean current speeds from flow noise recorded over a 52-day deployment. A strong correlation was observed between acoustic noise intensity at infrasonic frequencies (<20 Hz) and current speeds. Distinct spectral properties enabled differentiation between flow noise and ambient sound, including shipping noise. A regression tree machine-learning model trained using data from MANR #1, equipped with both a hydrophone and a current meter, facilitated the inference of current speeds with 1-min resolution at MANR #2, which was equipped solely with a hydrophone. MANRs #1 and #2 were located on steep seamount flanks at depths of 2573 and 2994 m, respectively. The regression tree model estimated current speeds of up to 107.7 cm/s at MANR #2, which was validated by spectral comparisons of flow noise at the two moorings. These findings underscore the potential for hydrophones to function as effective tools for long-term current monitoring, offering critical insights into deep-sea currents and their impact on seafloor dynamics and sediment transport. [Work supported by ONR.]