Boosted regression tree (BRT) analysis: Optimal parameter settings, predictive performance, and relative influence of environmental variables on total large-bodied reef fish biomass and presence/absence of key species.

Note: Total Biomass values were double-log+1 transformed to achieve pseudo-normality, species were analyzed based on presence/absence. Species selection was based on PRIMER BEST analysis, maximum data density, and management importance. CAAB = Carcharhinus amblyrhynchos, LUBO = Lutjanus bohar, MASP = Macolor spp., TROB = Triaenodon obesus, NAHE = Naso hexacanthus, SCRU = Scarus rubroviolaceus, CAME = Caranx melampygus, CHUD = Cheilinus undulatus.1Cross-validation (CV) deviance and standard error (SE) is shown as the measure of model performance (the lower the value the better the model performance).