This dataset provides a raster layer (.img format) with the predicted distribution of the gorgonian Eunicella cavolini (Koch, 1887) on the continental shelf and submarine canyon off Cap de Creus (NW Mediterranean), an area included in a Site of Community Importance (SCI) of the Natura 2000 Network. The algorithm Random Forest (RF) was used to model areas suitable for the development of the species based on a combination of species density data extracted from 60 underwater video transects (80 to 400 m depth) and a set of environmental factors of a geomorphologic, oceanographic and human nature. The RF model predicted the presence of the species (densities of at least 1 col·m-2) in 1,061 ha along the whole modelled area, with 14 ha potentially hosting gorgonian densities above 5 col·m-2 and 4.9 ha with densities exceeding 10 col·m-2. The variability explained by the model was 70.1%, although a cross-validation assessment indicated that the variability explained could be lower, denoting some degree of spatial autocorrelation of the data within the underwater dives. Based on the percentage increase in mean squared error of the model used for predictions, the most important variables determining the likelihood of finding high-density patches of E. cavolini in the study area were average bottom current velocity, depth and rugosity of the seafloor (calculated as TRI, Terrain Ruggedness Index). This study provides relevant information about the spatial distribution of E. cavolini inside a recently declared marine protected area of NW Mediterranean, which could be used by policymakers when formulating specific management measures for the conservation of this species within the SCI.

This dataset is related to the article “Environmental niche, spatial distribution and size structure of shelf-dwelling populations of the gorgonian Eunicella cavolini (Koch, 1887) in NW Mediterranean” submitted to Estuarine, Coastal and Shelf Science by Dominguez-Carrió et al., which contains further information on how the data were obtained and processed., The development of the Random Forest (RF) model was funded by the European project LIFE+ INDEMARES (LIFE07/NAT/E/000732). The biological data used to feed the RF model was obtained through several surveys funded by the European project LIFE+ INDEMARES (LIFE07/NAT/E/000732) and the Spanish project Acciones Complementarias nº CTM2007-28758-E/MAR. The bathymetric data for the continental shelf was provided by the Spanish Fisheries General Secretary, and acquired in the frame of the ESPACE, DeepCoral and LIFE+ INDEMARES projects. Multibeam data for the submarine canyon was acquired by the University of Barcelona, AOA Geophysics and the Institute of Marine Sciences (ICM-CSIC). The bottom current model used to run the RF model was performed using HPC resources from the CALMIP platform (French region Occitanie) under grant P09115, with the SYMPHONIE model distributed by the SIROCCO group (https://sirocco.obs-mip.fr). VMS data was provided by the General Directorate of Fisheries Management of the Spanish Ministry of Agriculture, Food and Environment (MAGRAMA). The authors would like to thank the crew of the research vessel García del Cid and all the scientists and technicians who assisted during the different oceanographic cruises, as well as Gavin Newman (ROV pilot) and the members of the JAGO Team Jürgen Schauer and Karen Hissmann (Geomar, Kiel) for the acquisition of the video footage. This work acknowledges the AEI ‘Severo Ochoa Centre of Excellence’ accreditation (CEX2019-000928-S). CD-C acknowledges the financial support received by the EU funded projects Life+ Indemares (LIFE07/NAT/E/000732) and ENPI ECOSAFIMED (ENPI CBC MED 2007-2013) and by the FCT-IP Project UIDP/05634/2020., With funding form the Spanish government through the "Severa Ochoa Centre of Excellence" accreditation (CEX2019-000928-S)., Peer reviewed