At Florida Atlantic University's Harbor Branch Oceanographic Institute, researchers are training artificial intelligence to listen for the sound of predators crunching through shells—and in doing so, they're opening a window into one of the ocean's most invisible ecological dramas. Each crushed clam and shattered snail shell produces a distinct acoustic signature, brief but information-rich, and now a team led by Laurent Chérubin and Matt Ajemian has developed machine learning tools precise enough to detect and classify these feeding events from underwater recordings.
Interactions between hard-shelled marine mollusks like clams and snails and their predators play a foundational but largely unseen role in shaping coastal ecosystems. These organisms stabilize shorelines, filter water and support biodiversity—yet they face mounting pressure from ocean acidification and expanding populations of mobile shell-crushing predators. The problem is that many of these predators, including the whitespotted eagle ray, forage in subtidal environments where direct observation is nearly impossible. As a result, measuring how much mollusks are being consumed in natural systems has remained one of marine ecology's stubborn blind spots, despite decades of recognition that the process matters deeply.
The FAU team's breakthrough lies in passive acoustic monitoring—quite literally listening to predator-prey interactions as they happen. Rather than relying on a single detection method, their system uses a two-step approach. It first scans large datasets to flag potential shell-crushing sounds based on acoustic patterns, then applies a second layer of machine learning to separate real feeding events from background ocean noise. The researchers trained the system using controlled tank experiments with whitespotted eagle rays, highly mobile predators known for their ability to crack hard shells. Once validated, the system classifies prey type using both traditional and deep learning methods, including random forests, long short-term memory networks, and convolutional neural networks, each trained to recognize subtle patterns in acoustic structure.
A striking finding emerged from the work, published in Ecological Informatics: highly complex AI models were not always necessary for strong performance. Simpler methods using gammatone-based features proved nearly as effective as advanced deep learning systems at detecting shell-crushing sounds, while requiring far less computing power. That matters enormously for real-world application. It means these streamlined approaches could make long-term underwater monitoring more practical, scalable and cost-effective in the natural marine environments where they would be deployed.
Crucially, the system proved effective not only in controlled tank conditions but also in the field, using both animal-borne acoustic tags and fixed underwater recorders. "Shell-crushing sounds contain a surprising amount of ecological information about predator-prey interactions and feeding behavior," Chérubin said. "This work shows how passive acoustic monitoring can be used not only to detect these events, but also to better understand how marine predators interact with their environment in places that are otherwise difficult to observe."
For coastal conservation, the implications are transformative. Being able to remotely detect and classify feeding events means researchers can begin measuring predation pressure on mollusk populations at ecosystem scales, not just in isolated observations. As Ajemian put it, this represents a major step forward for understanding how human-altered ocean environments are reshaping the delicate balance between predator and prey. By training machines to listen, scientists have found a way to measure the unmeasurable.