Causal inference and large‐scale expert validation shed light on the drivers of <scp>SDM</scp> accuracy and variance

Abstract Aim To develop a causal understanding of the drivers Species distribution model (SDM) performance. Location United Kingdom (UK). Methods We measured accuracy and variance SDMs fitted for 518 species invertebrate plant in UK. Our measure reflects variation among replicate fits, taxon experts assessed accuracy. Using directed acyclic graphs, we developed depicting plausible effects explanatory variables (e.g. species' prevalence, sample size) on SDM quantified those using multilevel piecewise path model. Results According to our model, size niche completeness (proportion covered by sampling) directly affect variance. Prevalence range have indirect mediated size. Challenging conventional wisdom, found that effect prevalence is positive. This facts has positive larger sizes are possible widespread species. It possible, however, omission an unobserved confounder biased this effect. Previous studies, which reported negative correlations between accuracy, conditioned Main conclusions explicates basis previously performance species/data characteristics. also suggests similarly large as Analysts should consider completeness, or proxies thereof, addition when deciding whether modelling worthwhile.