Quantitative understanding of extinction risk in freshwater fishes remains poor. Despite facing numerous threats, fewer than 40% of species have been assigned an IUCN threat level. However, ecological, life-history and biogeographical traits of species often correlate with extinction risk. These data may therefore be of use in predicting threat levels in taxonomic groups where detailed data on population and range size trends necessary for IUCN assessments are lacking.

Using information on ten representative traits, together with a composite phylogeny of over 300 brackish and freshwater species, we apply two statistical approaches (phylogenetic generalised least-squares regression and random forest modelling) to identify traits which contribute to the extinction risk of Australian freshwater fishes. These two approaches are complementary in their methods and assumptions, yet their conclusions are broadly consistent, and they display similar ability to correctly reclassify species already assigned a threatened status. Predictive models of extinction risk may therefore prove useful in identifying species needing formal (re)assessment and in informing future conservation strategies.

Although further trait data and model development are needed to refine our understanding of species' differing susceptibility to extinction, our results highlight the heterogeneity in threat type and severity across species of differing ecology and biogeography.