Context. Research on species’ habitat associations is strengthened if it combines coarse-grained landscape data with finer-scale parameters. However, due to the effort required to measure fine-scale parameters, studies on threatened species that unite these two scales remain relatively rare.

Aim. This study aimed to undertake a multi-scale analysis of the habitat association of the threatened Petrogale lateralis (MacDonnell Ranges race) in the Anangu Pitjantjatjara Yankunytjatjara (APY) Lands, South Australia.

Method. Analyses were conducted at four spatial scales: (1) across the Central Ranges IBRA Region (regional scale); (2) on hills in the APY Lands at which P. lateralis is extant and extinct (site scale); (3) at ‘core’ and ‘non-core’ areas within those hills (hillside scale); and (4) at rocky refuges. The maximum entropy approach through the software MaxEnt was used for the analysis at the regional scale. At the remaining scales, fieldwork was used to collect, and regression modelling to analyse, data.

Key results. At the regional scale, presence was associated with slope and geology. At the site scale, aspect, rock abundance and habitat type are likely to have facilitated animal persistence at extant sites. At the hillside scale, the aspect, vegetation type and rock complexity of core areas are likely to have contributed to their higher use. Size, exposure and accessibility were significant predictors of the use of rocky refuges.

Conclusions. All four spatial scales yielded novel information on the habitat associations of P. lateralis, supporting previous researchers’ suggestions that habitat modelling should be conducted at multiple spatial scales.

Implications. The study exemplifies the utility of combining MaxEnt modelling with fieldwork-derived data. The results may have conservation implications for this threatened race, and may also provide a model for other studies of faunal habitat associations.