Researchers at the University of Gothenburg’s Tjärnö Marine Laboratory have used deep learning to analyse 26 years of underwater video from Kosterhavet National Park, Sweden’s first marine national park. Funded by the European Commission’s DTO-BioFlow project, the study processed over 4.4 million images to track seabed species and reveal ecological responses to protection measures and rising temperatures.

The results show recovery of sensitive communities, such as mussels, anemones, and soft corals, while some cold-water species retreat from shallower areas. This work demonstrates how AI can transform long-term video archives into actionable knowledge for Marine Protected Area management and ecosystem-based spatial planning, contributing to Europe’s Digital Twin of the Ocean initiative.

Read the press release: here.