Connectivity of Coral Ecosystems in the Northwestern Gulf of Mexico: Habitat suitability model predictions for corals, octocorals, sponges, black corals, and snappers in the northwestern Gulf of Mexico

This dataset contains continuous spatial predictions of habitat suitability developed for eight focal taxa across Flower Garden Banks National Marine Sanctuary and adjacent shelf-edge banks in the northwestern Gulf of America/Mexico. The modeled region encompasses banks within Flower Garden Banks National Marine Sanctuary and adjacent banks. Focal taxa include the scleractinian corals Montastraea cavernosa and Stephanocoenia intersepta, the octocorals Muricea pendula and Swiftia exserta, black corals (Antipathes spp.), the demosponge Xestospongia muta, Red Snapper (Lutjanus campechanus), and Vermilion Snapper (Rhomboplites aurorubens). Habitat suitability models were developed using a generalized additive model framework relating biological occurrence data to bathymetric and multiscale seafloor geomorphological predictors. Benthic-invertebrate occurrence data were compiled from remotely operated vehicle surveys, SCUBA surveys, and the NOAA Deep-Sea Coral & Sponge Database. ROV observations span 2003–2018 and SCUBA observations span 2006–2015; NOAA Deep-Sea Coral & Sponge Database records used in the analysis were predominantly from 2002–2018. Fish occurrence data were compiled from ROV observations and NOAA Fisheries Southeast Area Monitoring and Assessment Program drop-camera surveys. Models were fitted using a presence-versus-background design with 10,000 unique randomly located pseudo-absences. Pseudo-absence observations were weighted to constitute 50% of the fitted dataset. Bottom depth was included as the foundational environmental predictor, with additional predictors selected from 77 geomorphological metrics representing terrain characteristics across spatial scales from approximately 16 m to 32 km. These included bathymetric position index, distance to isobaths, rugosity, slope, aspect, curvature, and ruggedness metrics. Model selection and evaluation used spatial cross-validation based on five groups of banks. Bottom depth was entered first, with additional predictors selected iteratively while excluding variables highly correlated with depth or with one another, models that failed to converge, variables with low estimated significance, and models that did not improve predictive performance. Model selection was based primarily on cross-validated deviance explained. Additional evaluation metrics included area under the receiver operating characteristic curve, percent correctly classified, true positive rate, true negative rate, Cohen's kappa, and true skill statistic. The spatial predictions are distributed on a 16 m grid where high-resolution multibeam bathymetry is available. Continuous raster values range from 0 to 1, with larger values representing greater predicted habitat suitability under the fitted model. The products are intended for identifying areas of potentially suitable habitat, prioritizing future surveys, providing habitat context for connectivity analyses, and supporting spatial management and monitoring in and around Flower Garden Banks National Marine Sanctuary. Predictions represent modeled habitat suitability rather than observed abundance, population density, or occupancy and should be interpreted accordingly. This dataset is a result of research funded by NOAA’s National Center for Coastal Ocean Science’s Competitive Research Program and Office of Ocean Exploration and Research under award NA18NOS4780166 to Lehigh University.