Predictive modeling of coral distribution in the Central Aleutian Islands, USA

Logistic regression and generalized estimating equation (GEE) models were used to predict the probability of presence/absence for deep-sea corals in the central Aleutian Islands from 50 to 3000 in depth. Explanatory variables included in the models were depth, slope, and rugosity. Five of 15 logistic models were considered acceptable based on cross validation using a Kappa statistic threshold of 0.4. Models of occurrence north of the Aleutian chain were more successful than models for areas to the south of the chain, Based on the logistic regression predictive model, there are large swaths of sea floor between 100 and 1700 m depth with high rugosity and steep slope that have the potential to support coral gardens, particularly north of the Aleutian Islands arc and in Amchitka Pass. Improvements to the predictive models, including high resolution substrate and current data, as well as more in situ coral observations from a broader range of areas in the archipelago, would allow an improved evaluation of the adequacy of recent fishing gear restrictions in the Aleutian Islands to provide conservation benefits.