Modeling the effects of climate change on the habitat suitability of Mediterranean gorgonians

Multiple stressors, including global warming, increasingly threaten the distribution and abundance of gorgonian forests. We built species distribution models (SDM) combined with machine learning algorithms to compare the ecological niche and distribution response to climate change under the worst IPCC scenario RCP8.5 for three Mediterranean gorgonian species (Paramuricea clavata,Eunicella cavolinii and Eunicella singularis. To obtain the potential habitat suitability and future distribution projections (2040-2050), we employed three Machine Learning models (XGBoost, Random Forest and the K-nearest neighbour) which considered 23 physicochemical and 4 geophysical environmental variables. The global sensitivity and uncertainty analysis was used to identify the most important environmental variables shaping habitat suitability for each species and to disentangle the interaction terms among environmental variables. For all species, bathymetry was the primary variable influencing habitat suitability, which had strong interactions with silicate concentration, salinity, and concavity. Under predicted future climatic conditions, P. clavata is predicted to shift its habitat suitability from lower to higher latitudes, mainly in the Adriatic Sea. For both E. cavolinii and E. singularis, a general habitat reduction was predicted. In particular, E. cavolinii is expected to reduce its occupancy area by 49%, suggesting that the sensitivity of symbiotic algae (zooxanthellae) may not be the principal cause of susceptibility of this species to thermal stresses and climate change.