Fort Lauderdale, United States – Researchers have developed an artificial intelligence system capable of identifying underwater predator-prey interactions by analyzing the sounds produced when marine animals crush shellfish.

The research focuses on hard-shelled mollusks such as clams and snails, which are essential to coastal ecosystems because they help stabilize shorelines, filter water and support biodiversity. These organisms are increasingly under pressure from ocean acidification and growing populations of shell-crushing predators.

These feeding interactions happen quickly and are difficult to observe directly in underwater environments, making them challenging to measure in natural conditions. However, each crushed shell creates a distinct acoustic signal that can be captured by underwater recording systems.

Scientists used machine learning tools to detect and classify these feeding sounds, training their system through controlled experiments involving whitespotted eagle rays, which are known to break open hard shells while feeding.

The system first scans large acoustic datasets to identify possible feeding events, then filters out background noise to reduce false detections. A second stage of analysis classifies prey types using a combination of traditional and deep learning techniques, including random forests, long short-term memory networks and convolutional neural networks.

Researchers found that simpler models based on gammatone-derived features performed almost as well as complex deep learning systems, while using significantly less computing power, making them more suitable for long-term monitoring in marine environments.

The acoustic data also revealed additional behavioral information, including differences in handling and processing of prey. The system performed effectively in both controlled tank environments and real-world deployments using underwater recorders and animal-borne tags.

Even when trained only on controlled experiments, the system was still able to detect feeding events and identify prey types in natural settings with strong reliability.

Scientists say this approach could enable remote measurement of predation pressure on mollusk populations across coastal ecosystems, improving understanding of marine food webs at large scales.

The method also offers potential for monitoring a wide range of habitats, from buried filter feeders to more mobile species, supporting future conservation and management efforts.

Podcast: Play in new window | Download