Dataset: 9 Species Experiment (Months 3 and 7)

Methodology:

We conducted an experiment that incorporated a greater range of coral species richness to better assess the role of coral diversity per se, evaluate changes in the shape of this relationship with increasing species richness, and lessen the potentially confounding effects of species identity when evaluated across only three species. Forty-eight experimental plots were deployed within a back reef lagoon on Mo’orea (17°29′19″S 149°52′54″W). Treatments consisted of one, three, six, or nine common coral species drawn at random for each plot from a pool of nine coral species: P. rus, P. lobata, S. pistillata, P. damicornis, P. verrucosa, P. cactus, A. hyacinthus, A. pulchra, and A. cytherea (12 plots per treatment; 864 corals total). Individual corals were randomly embedded within each plot, and differences in growth and tissue mortality were assessed at three and seven months with permutation ANOVA and a post hoc permutation test for multiple comparisons using the R package predictmeans. Six and 27 coral replicates that became dislodged from their epoxy base were excluded from our analyses at three and seven months (0.7 and 3.1% of replicates), respectively. At both time points, we also excluded 18 corals from a nine-species plot that was colonized by damselfish (the only plot where this occurred). In instances where significant differences among treatments were detected (i.e., growth at three and seven months and tissue mortality at three months), we also conducted separate analyses comparing growth and tissue mortality among treatments for each of the nine species used in our manipulations. Macroalgal cover was absent among plots across all treatments.

Sampling and analytical procedures: 

We used permutation-based, LME models in the R package predictmeans to compare differences in the percentage mass change and tissue mortality of corals in each treatment. In each analysis, plot type (e.g., 3 species) was treated as a fixed factor, and individual replicate plots were treated as a random effect nested within plot type.