Coral reefs, one of the world's most complex and vulnerable ecosystems, face an uncertain future in coming decades as they continue to respond to anthropogenic climate change, overfishing, pollution, and other human impacts [1, 2]. Traditionally, marine macroecology is based on presence/absence data from taxonomic checklists or geographic ranges, providing a qualitative overview of spatial shifts in species richness that treats rare and common species equally [3, 4]. As a consequence, regional and long-term shifts in relative abundances of individual taxa are poorly understood. Here we apply a more rigorous quantitative approach to examine large-scale spatial variation in the species composition and abundance of corals on midshelf reefs along the length of Australia's Great Barrier Reef, a biogeographic region where species richness is high and relatively homogeneous [5]. We demonstrate that important functional components of coral assemblages "sample" space differently at 132 sites separated by up to 1740 km, leading to complex latitudinal shifts in patterns of absolute and relative abundance. The flexibility in community composition that we document along latitudinal environmental gradients indicates that climate change is likely to result in a reassortment of coral reef taxa rather than wholesale loss of entire reef ecosystems.