Fig. 5: Reef fish dissimilarity correlates more strongly with shifts in coral composition than changes in coral cover (%).

Predicted values of reef fish dissimilarity (β diversity) of 92 reefs (three sites per reef per year) for a change in coral cover (%) and b shift in coral composition. Dots in a and b represent sites, and lines represent the fitted generalised linear mixed models (GLMM) (n reefs = 92). ‘Reef’ and ‘sShelf position’ were used as random effects in our models. Points in c represent means of the effect sizes of the GLMM and lines the 95% CIs (n reefs = 92). d Predicted values of reef fish dissimilarity with shifts in coral composition at latitudinal sector level: CL (n reefs = 8), CA (n reefs = 11), IN (n reefs = 7), TO (n reefs = 18), WH (n reefs = 9), PO (n reefs = 12), SW (n reefs = 17), CB (n reefs = 10). Shading bands in a, b and d represent the 95% CIs. In (c) and (e), slopes are significant if their 95% CIs do not overlap the vertical dashed line at zero. Fitted lines for CA, IN, and TO in (d) are omitted as their effects are non-significant. Points in e represent the mean effect size of shifts in coral composition on reef fish dissimilarity for each latitudinal sector from the GLMM; lines represent 95% CIs. Codes in d and e are latitudinal sectors: Cooktown/Lizard Island (CL), Cairns (CA), Innisfail (IN), Townsville (TO), Whitsunday (WH), Pompey (PO), Swain (SW) and Capricorn‐Bunker (CB) ordered in descending from lower to higher latitudes and colours from red to blue, respectively. Source data are provided as a Source Data file.