Determining correlates of density for large carnivores is important to understand their ecological requirements and develop conservation strategies. Of several earlier density studies conducted globally, relatively few addressed a scale (usually >1000 km2) that allows inference on correlates of density over heterogeneous landscapes. We deployed 164 camera trap stations covering ~2500 km2 across five areas characterized by broadly different vegetation cover in the Udzungwa Mountains, Tanzania, to investigate correlates of density for a widespread and adaptable carnivore, the leopard (Panthera pardus). We modelled data in a spatially explicit capture-recapture framework, with both biotic and abiotic covariates hypothesised to influence density. We found that leopard density increased with distance to protected area boundary (mean±SE estimated effect = 0.44±0.20), a proxy for both protected area extent and distance from surrounding human settlements. We estimated mean density at 4.22 leopards/100 km2 (85% CI = 3.33‒5.35/100 km2), with no variation across habitat types. Results indicate that protected area extent and anthropogenic disturbance limit leopard populations whereas no support was found for prey availability and trap array as drivers of leopard density. Such vulnerability is relevant to the conservation of the leopard, which is generally considered more resilient to human disturbance than other large cats. Our findings support the notion that protected areas are important to preserve viable population of leopards, increasingly so in times of unprecedented habitat fragmentation. Protection of buffer zones smoothing the abrupt impact of human activities at reserve edges also appears of critical conservation relevance.