ABSTRACT Aim Reliable biodiversity assessments using species distribution models (SDMs) are essential for effective conservation and management. Understanding factors influencing SDM performance is crucial for improving model reliability, yet such links remain underexplored in marine systems. To address this gap, we quantified the effects of geographical and species‐level factors on SDM performance in tropical reef fishes, aiming to provide practical guidelines on which species are more likely to yield reliable predictions. Location Global tropical reef ecosystems. Methods We built ensemble SDMs for 1941 tropical reef fish species using occurrence records. We evaluated model predictive performance using the continuous Boyce index, the most suitable performance metric given that we lacked quality absence data, and two other commonly applied metrics (the area under the receiving operating characteristic curve and the true skill statistic). We compiled 10 factors related to species' geographical and ecological characteristics and assessed their influence on model performance using phylogenetic generalised linear models. Results Ensemble SDMs for tropical reef fishes exhibited high predictive performance based on the three evaluation metrics. Phylogenetic generalised linear models relating evaluation metrics to species geographical and ecological characteristics showed modest explanatory power, with R 2 varying from 0.253 to 0.341. Across evaluation metrics, SDM performance was strongly associated with species' latitude, proximity to shore, and environmental similarity between training and evaluation datasets. For continuous Boyce index, there were additional significant effects for range size, parental care, range coverage, and species description year. Main Conclusions Our study provides a practical framework for identifying tropical reef fish species more likely to yield reliable SDM predictions. The identified factors offer guidance for researchers to anticipate model reliability before undertaking extensive modelling efforts. This large‐scale, multi‐species comparative approach is broadly applicable to other marine taxa and regions, advancing our ability to model and conserve marine biodiversity.
Liu et al. (Sun,) studied this question.