Artificial Intelligence Is Mapping the Pacific Ocean’s Secret Soundscape
Dr. Hervé Glotin and his team have developed artificial intelligence models that aim to decipher 3,000 hours of underwater recordings from Seeker's trans-Pacific journey. If they are successful, they will build the most comprehensive dataset of its kind – an enormous "sound postcard of the ocean."
Each night at sea, the crew aboard Seeker deploys a "hydrophone" -- a specialized microphone designed for use underwater. While the boat stays at or navigates back to Ben's last GPS point, the hydrophone sits 25 meters below the surface and eavesdrops on the chatter of cetaceans and the hum of boat traffic.
Scientists like Dr. Hervé Glotin at the University of Toulon have built robust "machine listening" systems to analyze these some 3,000 hours of recordings - and actually recognize which species are present where, and which areas of the Pacific are most cluttered with man-made noise.
"A soundscape is a recording of a multiplicity of sounds," Hervé explains. "Like if you look to a mountain's landscape, you will see trees, rivers, roads, buildings, maybe snow, maybe clouds. This notion of objects is the same with sound." It is the job of his lab's AI models to parse and interpret this soundscape, and label these "objects" by species or sound type. "Biophony are the sounds from biology activities. Anthropophony are the sounds from human activities. And geophony is the sounds from wind, and the rain, and the tectonic," Hervé explains.
It is anthropophy, or anthropogenic noise, that presents a particular challenge to these marine mammals, who rely on clear communication channels to be able to find each other over long distances and navigate their environment.
"The more you have sounds in the sea... the less the animals can communicate between each other. So the less they meet, the less they mate, the less they reproduce. The less the species can hear, the less it can hunt, the less it's eating," Hervé explains. An ocean too noisy to hunt or call for a mate means species are in trouble.
However, if this project is successful, it will represent the largest dataset of its kind. If the evidence of the sounds of the Pacific Ocean, and the wildlife population densities they indicate, can reveal where the key routes and channels of migration and communication for cetaceans exist, policymakers can develop solutions such as "quiet zones" to help alleviate the stress of human noise on wildlife communication.
