Modeling the spatial distribution of numbers of coral reef fish species and community types in the Western Indian Ocean faunal province

Predicting and mapping coral reef diversity at moderate scales can assist spatial planning and prioritizing conservation activities. We made coarse-scale (6.25 km(2)) predictive models for numbers of coral reef fish species and community composition starting with a spatially complete database of 70 environmental variables available for 7039 mapped reef cells in the Western Indian Ocean. An ensemble model was created from a process of variable elimination and selectivity to make the best predictions irrespective of human influences. This best model was compared to models using preselected variables commonly used to evaluate climate change and human fishing and water quality influences. Many variables (similar to 27) contributed to the best number of species and community composition models, but local variables of biomass, depth, and retention connectivity were dominant predictors. The key human-influenced variables included fish biomass and distance to human populations, with weaker associations with sediments and nutrients. Climate-influenced variables were generally weaker and included median sea surface temperature (SST) with contributions in declining order from SST kurtosis, bimodality, excess summer heat, SST skewness, SST rate of rise, and coral cover. Community composition variability was best explained by 2 dominant community richness axes of damselfishes-angelfishes and butterflyfishes-parrotfishes. Numbers of damselfish-angelfish species were ecologically separated by depth, and damselfishes declined with increasing depth, median temperature, cumulative excess heat, rate of temperature rise, and chronic temperature stresses. Species of butterflyfish-parrotfish separated by median temperature, and butterflyfish numbers declined with increasing temperature, chronic and acute temperature variability, and the rate of temperature rise. Several fish diversity hotspots were found in the East African Coastal Current Ecoregion centered in Tanzania, followed by Mayotte, southern Kenya, and northern Mozambique. If biomass can be maintained, the broad distributions of species combined with compensatory community responses should maintain high diversity and ecological resilience to climate change and other human stressors.