Speaker
Description
Fiber-optic submarine cables instrumented with distributed acoustic sensing (DAS) provide a cost-effective solution for large-scale, real-time monitoring of fin whales. We present an automatic workflow for the detection, characterization, and localization of fin whale songs, developed and tested using data recorded on a telecom submarine optical cable in the Alboran Sea (westernmost part of the Mediterranean, at the transition to the Atlantic Ocean) from late September 2023 to early February 2024. The workflow begins with a kurtosis-value picker (KVP) optimized for fin whale pulse detection. The resulting picks are then clustered using the DBSCAN algorithm to group coherent arrivals across DAS channels. This clustering step enables the characterization of individual notes by extracting key features related to energy, time, and frequency. Based on these features, signals can be classified into 20-Hz notes and backbeats, enabling a detailed characterization of songs. The identified clusters also provide high-quality candidates for localization by exploiting the multi-channel nature of DAS recordings. Time-of-arrival data are selected through a fitting consistent with the cable geometry, followed by grid-search localization. The proposed workflow supports near-real-time processing, while reducing data volumes for downstream analysis, and offers a scalable framework for fin whale acoustic monitoring applicable to other DAS datasets.