Using underwater acoustics from tagged elephant seals to evaluate variations in geophonic noise (wind/rain) and biophonic noise
Weather conditions have an important impact on different oceanographic processes including the concentration in nutrients in the surface layer necessary in primary production. However, current satellite data in the Austral Ocean is neither precise nor complete enough to characterize such fine-scale processes. My PhD looks at another way to acquire weather data in the Antarctic using acoustic data from tagged southern elephant seals (SES). Using clean underwater acoustic recordings and transfer learning techniques, a neural network is trained to predict weather conditions based on acoustic signatures and patterns in the recordings from the tagged SES. The acoustic data are coupled with wind direction and sea state data acquired through an accelerometer and magnetometer also on the tag. Besides their utility in meteorology, the acoustic recordings are used to study SES’ behavior by i) seeing if they follow biophonic gradients to find prey and ii) evaluate their success rate during their hunts. Furthermore, SES can be used as buoys for passive acoustic monitoring tasks such as detecting biophonic cues such as cetacean calls.
PhD student in the Marine Predators team
Supersivor: Christophe Guinet (CEBC)