Large protected areas are essential for the long-term conservation of wide-ranging or low-density large carnivore populations. However, small protected areas can also contribute to carnivore conservation if they have sufficient prey density, if wildlife crime is controlled and if they are connected to other protected areas and/or embedded within a larger matrix of at least partially suitable habitat. In the central highlands of Sri Lanka, the 31.6 km2 Horton Plains National Park (HPNP) is a mosaic of tropical montane cloud forest and patana grassland set within a mosaic landscape of forest and agriculture. We conducted a camera trapping study in 2017 and 2018 to estimate leopard densities in HPNP and predict their activity centers. A Bayesian multi-session spatially-explicit capture-recapture analysis of 249 independent captures across two 4-month camera trapping campaigns returned density estimates of 10.6 per 100 km2 (95% HDI = 8.0–16.0) in 2017 and 15.1 per 100 km2 (95% HDI = 13.3–18.7) in 2018. Female capture rates and activity centers remained stable over the two survey periods, indicating fidelity to the grasslands. In contrast, most male activity centers were predicted to fall outside of the park, which implied that a significant proportion of males’ activity occurred outside HPNP, likely with occasional returns to the prey-dense grassland area. As a result, HPNP is unlikely to be sufficient on its own to sustain the estimated leopard densities—particularly for males—without contribution from the surrounding landscape. Thus, although the Department of Wildlife Conservation maintains jurisdiction over wildlife anywhere in the country, interagency collaboration in research, management and conservation should be encouraged owing to multiple jurisdictions over land management that will affect habitat suitability for leopards. Finally, we highlight the potential benefits of maintaining an updated camera trap database for outreach and potential conflict management.