Benthic assemblages are more predictable than fish assemblages at an island scale

Abstract Decades of research have revealed relationships between the abundance coral reef taxa and local conditions, especially at small scales. However, a rigorous test covariation requires robust dataset collected across wide environmental or experimental gradients. Here, we surveyed spatial variability in densities major functional groups 122 sites along 70 km expanse leeward, forereef habitat Curaçao southern Caribbean. These data were used to degree which community composition could be predicted based on assumed site-specific anthropogenic, physical, ecological conditions. In general, models less power describe fish biomass than cover builders (R 2 best-fit models: 0.25 [fish] 0.64 [reef builders]). The total benthic was best described by physical (wave exposure relief) (turf algal height recruit density) predictors. No metric anthropogenic pressure related variation builder cover. contrast, showed consistent (albeit weak) association with predictors (fishing diving pressure). As is typical most gradients, patterns both density spatially autocorrelated. Residuals from model for retained signature autocorrelation while removed autocorrelation, thus reinforcing our finding that better able biomass. seek understand communities scale management units (i.e., kilometer- island-scales), distinct scale-dependent perspectives will needed when considering different groups.