Complex landscapes interact with meteorological processes to generate climatically suitable habitat (refuges) in otherwise hostile environments. Locating these refuges has practical importance in tropical montane regions where a high diversity of climatically specialized species is threatened by climate change. Here, we use a combination of weather data and spatial modeling to quantify thermally buffered environments in a regional tropical rainforest. We do this by constructing a spatial surface of maximum air temperature that takes into account important climate-mediating processes. We find a strong attenuating effect of elevation, distance from coast and foliage cover on maximum temperature. The core habitat of a disproportionately high number of endemic species (45%) is encompassed within just 25% of the coolest identified rainforest. We demonstrate how this data can be used to (i) identify important areas of cool habitat for protection and (ii) efficiently guide restoration in degraded landscapes to expand extant networks of critical cool habitat.