Table 3. Summary results of a distance-based permutational multiple regression analysis for the association of the prevalence of two coral diseases (Acropora and Porites growth anomalies) with 9 predictor variables across surveys (304 and 602, respectively) throughout the Indo-Pacific Ocean.

Disease	n	Predictor	BIC	Pseudo-F	P value	% variability	% total
Acropora GA	304	AcropCov	1925.5	21.18	0.0001	16.6	16.6
Porites GA	602	HumPop100	4349.2	36.88	0.0001	15.8
		PorDen	4335.9	19.98	0.0001	13.0
		UV	4325.8	16.57	0.0002	12.4	41.2

The optimal predictors of each disease and the proportion of variability (%) in the data set they explained are shown. Predictor variable codes and units are as per Table 2. Model development was based on step-wise selection and a Bayesian Information Criterion (BIC), with the total variation (r²) explained by each best-fit model shown (% total). Analyses based on 9999 permutations of the residuals under a reduced model.