Environmental and spatial predictors of species richness and abundance in coral reef fishes

Abstract

Aim: We developed predictive models of coral reef fish species richness and abundance that account for both broad-scale environmental gradients and fine-scale biotic processes, such as dispersal, and we compared the importance of absolute geographical location (i.e. geographical coordinates) versus relative geographical location (i.e. distance to domain boundaries). Location Great Barrier Reef, Australia. Methods: Four annual surveys of coral reef fishes were combined with a 0.01°-resolution grid of environmental variables including depth, sea surface temperature, salinity and nutrient concentrations. A principal component-based method was developed to select candidate predictors from a large number of correlated variables. Generalized linear mixed-effects models (GLMMs) were used to gauge the respective importance of the different spatial and environmental predictors. An error covariance matrix was included in the models to account for spatial autocorrelation. Results (1) Relative geographical descriptors, represented by distances to the coast and to the barrier reef, provided the highest-ranked single model of species richness and explained up to 36.8% of its deviance. (2) Accounting for spatial autocorrelation doubled the deviance in abundance explained to 71.9%. Sea surface temperature, salinity and nitrate concentrations were also important predictors of abundance. Spatially explicit predictions of species richness and abundance were robust to variation in the spatial scale considered during model calibration. Main conclusions: This study demonstrates that distance-to-domain boundaries (i.e. relative geographical location) can offer an ecologically relevant alternative to geographical coordinates (i.e. absolute geographical location) when predicting biodiversity patterns, providing a proxy for multivariate and complex environmental processes that are often difficult or expensive to estimate.