Semi-enclosed coastal systems are highly dynamic environments where benthic organisms are exposed to strong hydrographic gradients and increasing anthropogenic pressures. This study assessed the habitat suitability of the pearl oyster Pinctada radiata in two contrasting Mediterranean gulfs of Central Greece, the Maliakos and the South Evoikos, by integrating Copernicus Earth Observation (EO) products with an Artificial Intelligence (AI) modeling framework. Environmental variables, including sea surface temperature, salinity, chlorophyll-a concentration, current velocity, and dissolved oxygen, were derived from satellite and marine datasets and used to train a multi-algorithm ensemble combining Maximum Entropy (MaxEnt), Extreme Gradient Boosting (XGBoost), and a Convolutional Neural Network (CNN). The ensemble model showed strong predictive skill (AUC = 0.94; TSS = 0.80) and identified temperature, dissolved oxygen, and substrate type as the main drivers of habitat suitability. Spatial projections indicated that roughly two-thirds of the study area currently supports favorable conditions for P. radiata, particularly in shallow, low-energy, mesotrophic zones. Under a simulated +2 °C warming scenario, highly suitable habitats declined by about 20%, highlighting the species’ sensitivity to future thermal stress and subsequent oxygen depletion, demonstrating the value of EO-driven AI approaches for anticipating ecological change in vulnerable coastal systems.
Dimitris et al. (Thu,) studied this question.