Context The niche reduction hypothesis (NRH) predicts that the realised niche of declining species is reduced by threats that are mediated by environmental, biotic and evolutionary processes, explaining why species decline in some locations but not others. The critically endangered central rock-rat (CRR) survives only in rugged mountain range habitat in central Australia and is highly vulnerable to cat predation. We predicted that cat density and ranging behaviour, and, hence, predation risk, is mediated by habitat complexity, thus explaining the mechanism maintaining the CRR refuge.

Aims We sought to determine whether cat densities were lower in the rugged CRR refuge than in an adjacent valley dominated by less complex rocky habitats and no longer occupied by CRRs.

Methods We installed arrays of camera traps along two parallel mountain ranges in the refuge and in the intervening valley habitats. We identified uniquely patterned individual cats and compared spatially explicit capture–recapture (SECR) models to evaluate our hypothesis that cat density varies with topographic complexity.

Key results The dominant effect in all models was the significant negative relationship between cat detection probability and fine-scale topographic ruggedness. Two of the best three SECR models indicated lower cat densities and relative home-range sizes in the refuge than in the valley. In total, 17% of cats were detected in both habitat types.

Conclusions We found some evidence that cat density and home-range size were mediated by habitat complexity. Further, the negative relationship between cat detection probability and topographic complexity suggests that cats spend less time foraging in CRR refuge habitat.

Implications Cat management programs, aimed at reducing predation pressure on the CRR, must include the refuge and surrounding habitats to control cats that pose a threat to CRR subpopulations.