Tiger Panthera tigris populations have declined dramatically in the Terai Arc Landscape (TAL; India and Nepal), and remaining populations are highly fragmented and endangered. As part of a research program to aid tiger management by identifying critical areas for conservation, we aimed to 1) identify the factors which affect the distribution of tigers in the TAL; 2) explore the role of spatial scale in habitat selection; 3) map potentially suitable habitats; and 4) assess the quality of potential corridors linking suitable habitats. We used an approach based on presence and pseudo-absence data, combining ecological niche factor analysis and generalized linear models. We used an information-theoretic approach to compare our data on tiger presence with different hypotheses on tiger habitat selection (i.e. protective habitat, prey species, human disturbance), and spatial scales. All hypotheses yielded models with high prediction accuracy (>79%). The most parsimonious model included variables characterizing habitat suitability of the 2 main prey species. More detailed assessment of potentially suitable areas using an extended source-sink approach suggested that most of the habitats outside the protected areas were attractive sink-like habitats (i.e. they suffered high levels of human disturbance in otherwise good habitats). Overall, 24% (ca 18 500 km2) of the study area was predicted as suitable (probability cut-off p>0.5), approximately 7% of which is under protection. Our models showed that protecting the remaining concentrations of tigers requires focusing management efforts on specific areas outside the currently protected areas. These are characterized by good natural suitability; however, they suffer from a high level of human disturbance. Our models underscore the importance of minimizing human disturbances in these areas to avoid that they act as attractive sinks but act as corridors between existing subpopulations.