MaxEnt modelling as a tool to inform discovery of deep-sea coral ecosystems

Coral reefs are now one of the most threatened marine ecosystems, due to local anthropogenic pressures and global changes in ocean conditions. Until recently, research has focused on the vulnerability of shallow coral ecosystems, however effects on deeper coral-dominated ecosystems become more severe with climate change. Deep-sea corals form large bioconstructions, including dense frameworks and mounds, which can remain even when the organisms are dead. The presence of these ecosystems alters the local seafloor morphology. To understand the changes that deep-sea coral ecosystems are undergoing, and the impacts these changes may have on local seafloor morphology, we must first know where they are located. Exploration of deep-sea habitats is costly, challenging and time-consuming, reducing the geographic extent at which it is possible. Methods are also often difficult in remote areas. Therefore, as such ecosystems are relatively sparse, extensive resources may be used attempting to locate them. Species distribution models can be used to identify key areas of interest with much less information, focusing surveys to a smaller extent, and making it much more cost-effective to study. MaxEnt is a habitat suitability modelling software which uses a unique maximum entropy algorithm to find the most dispersed distribution. It requires only presence data, allowing accurate results with less input data. Here we present the first habitat suitability models for deep coral habitats, as deep as 720 m, in the Red Sea. During the OceanX-NEOM ‘Deep Blue’ expedition in the Northern Saudi Arabian Red Sea and the Gulf of Aqaba in 2020, deep-sea coral ecosystems were identified and characterised by ROV and submersible dives. Geomorphometric variables were derived from bathymetry, including measures of aspect, curvature and local-scale rugosity. These variables were used as input data for MaxEnt, along with backscatter, which can provide information about the seafloor substrate, and environmental data from CTD casts and current models. The potential of such models to identify areas of interest is clear, and could become an important tool in order to focus limited conservation funding.