Areas of suitable habitat for species and communities have arisen, shifted, and disappeared with Pleistocene climate cycles, and through this shifting landscape, current biodiversity has found paths to the present. Evolutionary refugia, areas of relative habitat stability in this shifting landscape, support persistence of lineages through time, and are thus crucial to the accumulation and maintenance of biodiversity. Areas of endemism are indicative of refugial areas where diversity has persisted, and endemism of intraspecific lineages in particular is strongly associated with late-Pleistocene habitat stability. However, it remains a challenge to consistently estimate the geographic ranges of intraspecific lineages and thus infer phylogeographic endemism, because spatial sampling for genetic analyses is typically sparse relative to species records. We present a novel technique to model the geographic distribution of intraspecific lineages, which is informed by the ecological niche of a species and known locations of its constituent lineages. Our approach allows for the effects of isolation by unsuitable habitat, and captures uncertainty in the extent of lineage ranges. Applying this method to the arc of rainforest areas spanning 3500 km in eastern Australia, we estimated lineage endemism for 53 species of rainforest dependent herpetofauna with available phylogeographic data. We related endemism to the stability of rainforest habitat over the past 120,000 years and identified distinct concentrations of lineage endemism that can be considered putative refugia. These areas of lineage endemism are strongly related to historical stability of rainforest habitat, after controlling for the effects of current environment. In fact, a dynamic stability model that allows movement to track suitable habitat over time was the most important factor in explaining current patterns of endemism. The techniques presented here provide an objective, practical method for estimating geographic ranges below the species level, and including them in spatial analyses of biodiversity.