Among-habitat variation in herbivory on Sargassum spp. on a mid-shelf reef in the northern Great Barrier Reef

Herbivory is widely acknowledged as a key process determining the benthic community structure and resilience of coral reefs. Despite numerous studies that have examined herbivory across reef gradients in the Caribbean, few studies have directly quantified this process on Pacific reefs. Bioassays of two species of erect macroalgae (Sargassum swartzii and S. cristaefolium) were used to quantify variation in grazing intensity across seven habitats of varying depth and wave exposure on a mid-shelf reef in the northern Great Barrier Reef. Removal rates of Sargassum varied significantly among habitats, with both species displaying broadly similar patterns. The shallow habitats on the exposed aspect of the reef (i.e. reef crest, flat and back reef) experienced the highest reductions in mass (81.4-91.6% day(-1)) for both S. swartzii and S. cristaefolium, while the deeper exposed habitats (reef slope and base) displayed the lowest reductions (3.8-13.4% day(-1)) over a 24 h period. In contrast, the grazing intensity varied between the two species in the three habitats on the leeward aspect of the reef. Reductions in mass remained relatively high for S. swartzii on the patch reef and sheltered reef base and flat (62.7-76.5% day(-1)) but were considerably lower for S. cristaefolium (37.9-63.5% day(-1)) across the same habitats. Surprisingly, the rates of removal of Sargassum displayed no relationship with the density or biomass of roving herbivorous fishes or those species known to consume erect macroalgae, either collectively or independently. These results suggest that the relationship between browsing rates and herbivorous fish biomass is complex and may be driven by species that are underestimated in visual surveys. Direct quantification of browsing intensity using assays revealed a different pattern to inferences based on herbivore densities and highlights the potential difficulties of evaluating ecosystem processes based on visual census data alone.