Predictively modelling the distribution of the threatened brush-tailed rock-wallaby (Petrogale penicillata) in Oxley Wild Rivers National Park, north-eastern New South Wales, Australia

Context. Species Distribution Models (SDM) can be used to investigate and understand relationships between species occurrence and environmental variables, so as to predict potential distribution. These predictions can facilitate conservation actions and management decisions. Oxley Wild Rivers National Park (OWRNP) is regarded as an important stronghold for the threatened brush-tailed rock-wallaby (Petrogale penicillata), on the basis of the presence of the largest known metapopulation of the species. Adequate knowledge of the species' ecology and distribution in OWRNP is a key objective in the national recovery plan for the species occurring in the Park. Aims. To model distribution using key GIS-derived environmental factors for the brush-tailed rock-wallaby in OWRNP and to ground-truth its presence through field surveys in areas of high habitat suitability. Methods. We used Maxent to model the distribution of the brush-tailed rock-wallaby within OWRNP on the basis of 282 occurrence records collected from an online database, elicitation of informal records from experts, helicopter surveys and historic records. Environmental variables used in the analysis were aspect, distance to water, elevation, geology type, slope and vegetation type. Key results. Vegetation type (37.9%) was the highest contributing predictor of suitable habitat, whereas aspect (4.8%) contributed the least. The model produced an area under the curve (AUC) of the receiver operating characteristic (ROC) of 0.780. The model was able to discriminate between suitable and non-suitable habitat for brush-tailed rock-wallabies. Areas identified in our model as being highly suitable yielded eight new occurrence records during subsequent ground-truthing field surveys. Conclusions. Brush-tailed rock-wallaby distribution in OWRNP is primarily associated with vegetation type, followed by distance to water, elevation, geology, slope and aspect. Field surveys indicated that the model was able to identify areas of high habitat suitability.