Who said that? Deciphering Complex Coral Reef Soundscapes with Spatial Audio and 360 Degree Video

Abstract

Coral reef soundscapes hold an untapped wealth of biodiversity information. While they are easy to record, they are nearly impossible to decipher because we know very little about which sound is made by which species. With identified fish sounds, acoustic monitoring can directly measure biodiversity, detect key or invasive species, identify behavioral events and frequencies, and estimate abundance, all at temporal and spatial scales not possible with methods like eDNA or visual surveys. Using a novel approach to visualize underwater sound, we present the most extensive collection of identified natural sounds from reef-associated Atlantic fishes. These sounds were all ascribed to species in situ on a crowded Caribbean coral reef. The soundfield analysis technique combines visualizations of spatial audio with 360{degrees} video recordings, a method not previously accomplished underwater. We used these species identifications to decipher a representative section of a soundscape recording from a separate recording device. We have collected our identified recordings into a growing open-access resource to catalyze passive acoustic monitoring research, enabling a species-specific resolution of coral reef soundscape dynamics and providing critical validated information for developing machine learning models required to analyze an ever-expanding collection of long-term recordings.