• ecological scale;
  • mixed effects model;
  • resource selection;
  • semi-arid landscape;
  • spatial heterogeneity


The conservation of any species requires understanding and predicting the distribution of its habitat and resource use, including the effects of scale-dependent variation in habitat and resource quality. Consequently, testing for resource selection at the appropriate scales is critical. We investigated how the resource selection process varies across scales, using koalas in a semi-arid landscape of eastern Australia as a case study. We asked: at what scales does tree selection by koalas vary across regions? We tested the importance of the variation of our ecological predictors at the following scales: (i) the site-scale (a stand of trees representing an individual koala's perception of local habitat); (ii) the landscape-scale (10 × 10 km area representing a space within which a population of koalas exists); and (iii) a combination of these scales. We used a mixed-modelling approach to quantify variation in selection of individual trees by koalas among sites and landscapes within a 1600 km2 study area. We found that tree species, and tree height, were the most important factors influencing tree selection, and that their effect did not vary across scales. In contrast, preferences for trees of different condition, which is the state of tree canopy health, did vary across landscapes, indicating spatial variation in the selection of trees with respect to tree condition at the landscape-scale, but not at the site-scale. We conclude that resource selection processes can depend on the quality of those resources at different scales and their heterogeneous nature across landscapes, highlighting the consequence of scale-dependent ecological processes. Designing studies that capture the heterogeneity in habitat and resources used by species that have an extensive distribution is an important prerequisite for effective conservation planning and management.