Predictive distribution modelling of cold-water corals in the Newfoundland and Labrador region

Information on the distribution of cold-water corals in Newfoundland and Labrador (NL) waters largely comes from scientific multi-species trawl surveys and commercial fisheries observer programs. As a result, knowledge of coral distribution has been influenced by the type of gear used and by fishing effort distribution, leaving large knowledge gaps beyond the fishing footprint along the edge of the continental shelf. In support of international efforts to preserve marine biodiversity, maximum entropy (Maxent) species distribution models (SDMs) were generated for cold-water coral functional groups and individual species to predict their distributions throughout the NL region. Although functional group models have also been produced using a random forest (RF) approach in the past, the species level models provided here are thought to be the first of their kind in the NL region. Models generated for this study were found to be statistically robust, even for species with limited observation data, with average area under the receiver operating curve and true skill statistic values of 0.914 and 0.684, respectively. Findings indicated that models for functional groups typically overgeneralized habitat suitability and did not always reflect the distribution of the individual species they were based on. Furthermore, suitable habitats delineated by Maxent fit more closely to in situ observations than existing models generated for this region, providing more realistic estimates of habitat suitability at bathyal and abyssal depths. In general, the highest suitability was found along the continental shelf break and within canyons on the upper continental shelf, with Maxent models capturing variations in habitat suitability not previously described in the region.