An Integrated Multi‐Scale Approach to Habitat Modelling and Conservation for a Threatened Marsupial in a Dynamic and Fire‐Prone Landscape

We investigated the distribution and habitat associations of the southern brown bandicoot (<i>Isoodon obesulus obesulus</i>), a threatened marsupial, to understand how it occupies a dynamic, fire-prone landscape in south-east Australia. Our goal was to identify core habitat at landscape and local scales to inform conservation efforts. This study was conducted in Gariwerd (Grampians National Park), a fire-prone landscape in south-east Australia. We employed a multi-faceted approach. First, we built and field-validated Species Distribution Models (SDMs) to delineate landscape-scale core habitat. To rigorously evaluate model performance and minimise bias, we conducted broad-scale, systematic field surveys. Subsequently, we collected and analysed fine-scale habitat characteristics at survey sites, using occupancy modelling to enhance our understanding of the species' ecological niche at a local scale. Our SDMs demonstrated good discriminative ability between presence and absence when tested on field data (AUC = 0.75). However, a notable proportion of observed absences occurred in high-probability areas, leading to an overestimation of the species' distribution and negatively impacting overall evaluation metrics. This likely reflects the dynamic nature of bandicoot populations, introducing uncertainty into model predictions. Site-level occupancy modelling revealed specific habitat associations: bandicoot occupancy was positively influenced by greater coverage of spear grass, slender twine-rush, sedge, and the presence of diggings, while negatively associated with higher densities of <i>Callitris</i>. SDMs are valuable for demarcating core habitats for endangered species. However, to effectively inform conservation strategies for dynamic species like the southern brown bandicoot, model predictions must be rigorously validated with independent field surveys and complemented by fine-scale habitat analyses. This multi-scale approach is crucial for accurate conservation planning in dynamic landscapes.