Identifying fish sounds underwater with spatial audio and 360° video

Despite the proliferation of passive acoustic monitoring (PAM) recordings in marine and freshwater habitats around the world, identifying which species produce which sounds has been a pervasive and fundamental challenge, especially for fishes. The overwhelming majority of sounds from these recordings remain unidentifiable to a taxonomic level beyond “fish” and are best categorized only by sound characteristics. Most techniques for sound I-D in nature have a high probability of misattribution or have not proven to be scalable. We have developed a scalable method for in situ fish sound identification in nature by overlaying visualizations of spatial audio onto 360 deg video recordings. In Curaçao, we have developed an extensive collection of identified sounds from reef-associated Atlantic fishes, and a starting point for an open-access library of fish sounds (fisheyecollaborative.org/library). We intend to expand this approach globally to enhance taxonomic resolution for PAM recordings, providing critical validated information for developing the detection models required for large scale analysis. Identified sounds can help enable PAM to become a transformational tool for fish biology.