Determining the factors that drive the distributions of threatened species is often critical for informing effective conservation management actions. Species distribution models can be used to distinguish common habitat features shared by limited historical records and identify other areas where a species might persist. In this study, we built a species distribution model for the Endangered and cryptic Kangaroo Island dunnart (Sminthopsis fuliginosus aitkeni). We fitted generalised linear models using incidental records and presence-absence data from surveys between 1969 and 2018. In the models we included the variables rainfall, percentage native vegetation in the surrounding 2 km2, and post-fire vegetation age. The modelling suggested that rainfall and to a lesser extent post-fire vegetation age are good predictors of dunnart occurrence, with dunnart occurrence greatest in areas of high rainfall (>600 mm) and vegetation age classes <30 years post fire. Potentially suitable habitat for the KI dunnart was predicted to be on the central-western side of Kangaroo Island. These results suggest that careful fire management could benefit the dunnart, and that decreased rainfall (as projected by Australian climate models), will be a threat in the long term. Extensive recent fires on western Kangaroo Island suggest that climate-related threats are already being realised.