Predictive models of habitat suitability for the Common Crane Grus grus in a wintering area of southern Portugal were derived using logistic multiple regression and Geographic Information Systems. The study area was characterized by landscape variables and surveyed uniformly for the presence of cranes. The most important variables were distance to roosts, to open Holm Oak woods and to villages, and the occurrence of unpaved roads, shrubby vegetation, slope and orchards. Two models were built, the second having one variable fewer than the first. The selection of the best model was based on statistical and biological criteria. Crane distribution was negatively related to: distance to open Holm Oak Quercus rotundifolia woods and roosts. Additionally, unsuitable vegetation and orchard areas are avoided. In these areas movement is difficult, food availability is reduced and the risk of predation increased. We also found that villages and roads were avoided; disturbance is a significant factor for this species. Some management guidelines are proposed for the area: (1) maintenance of open Holm Oak woodlands, (2) incentives to avoid the abandonment of traditional agriculture and pastoral use of the area, which would lead to an increase of shrubby vegetation areas, (3) preservation of suitable roosting places and (4) management of new patches of forest and orchards.