Development and validation of artificial intelligence models for autonomous monitoring of Posidonia oceanica restoration projects.

Abstract

Posidonia oceanica meadows are a key marine ecosystem in the Mediterranean Sea, with approximately 2.2% of their cover lost annually. This seagrass generates unique habitats for multiple native marine species while retaining CO2 and producing oxygen. It also protects sandy beaches by attenuating wave energy, and dead Posidonia wrack deposits on the shore can provide natural protection against sand loss during extreme weather events. Given its ecological relevance and declining coverage, restoration efforts through active transplantation have gained increasing attention. However, evaluating the success of these transplantations remains challenging, as it requires sustained, long-term monitoring of sparse and heterogeneously distributed individual shoots, a scenario for which current tools are not well suited. This research focuses on the autonomous monitoring of Posidonia oceanica meadows and transplantations using marine robotics and artificial intelligence. Unlike existing approaches designed for continuous meadows, this research focuses on the autonomous detection and monitoring of sparse individual shoots in restoration sites, combining multi-scale observations from satellites, UAVs, surface robots, and fixed underwater observatories with AI-driven data fusion. The proposed system will enable continuous assessment of key indicators of P. oceanica health, with particular emphasis on early-stage, low-density restoration sites.