Automatic Analysis of Hydroacoustic Signals Related to the Activity of the Fani Maoré Submarine Volcano

Abstract Due to the properties of sound propagation underwater, including the existence of the SOFAR channel, sounds in the ocean can propagate over large distances with little attenuation. This makes passive acoustics a relevant method for monitoring natural events such as ice calving, earthquakes, and underwater volcanic eruptions. However, the lack of automated techniques (e.g., such as beamforming for small‐aperture array) for large (50 km) hydrophone arrays makes the data analysis very time‐consuming; indeed, analysts must recognize signals such as earthquake T‐phases on several hydrophones, manually pick and associate their arrival times to build event catalogs. This lengthy iterative process by trial and error often leads to underutilization of the recorded data and exposes the resulting catalogs to the analyst's subjectivity. Here we propose a pipeline that fully automates the creation of event catalogs from continuous hydroacoustic data in two steps: (a) detection of events at each recording station and (b) association of detections at multiple stations to trilaterate the source. In order to evaluate this approach, we tested it on real data recorded by the Mayotte Hydrophone Network (MAHY), which monitors the activity of the new underwater volcano off Mayotte Island in the Mozambique Channel. Its seismicity and volcanic activity were analyzed with the catalogs obtained.