Singing around the volcano: Detecting baleen whales in the Mozambique channel based on their song rhythms, from seismic and hydroacoustic data.

Abstract

Since March 2019 and October 2020, two long-term submarine networks-comprising ocean bottom seismometers and water-column hydrophones-have been deployed in the northern Mozambique Channel (Western Indian Ocean) to monitor the seismo-volcanic crisis that began offshore Mayotte in 2018. These deployments provide a valuable multi-year dataset for environmental monitoring in this remote region, including seasonal variations in the presence of baleen whales. To analyze these patterns, we developed an automated detector for stereotyped and regular signals, focusing on the characteristic inter-call intervals of each whale species, independently of the song structure. The detector's simplicity makes it computationally efficient and easily adaptable to species vocalizing in the 15-100 Hz frequency range over extended periods. Analysis of nearly six years of continuous seismo-acoustic data highlights the seasonal presence of Antarctic blue whales (May-August), Southwest Indian Ocean pygmy blue whales (bi-modal pattern from April-August and October-January), minke whales (June-December), and fin whales (July-November), and highlights the potential variability over years. Our results enhance and refine our understanding of the seasonal migration patterns of whale populations in this specific area of the Western Indian Ocean.