RESUMO
Species distribution models (SDMs), which statistically relate species occurrence to climatic variables, are widely used to identify areas suitable for species growth under future climates and to plan for assisted migration. When SDMs are projected across times or spaces, it is assumed that species climatic requirements remain constant. However, empirical evidence supporting this assumption is rare, and SDM predictions could be biased. Historical human-aided movements of tree species can shed light on the reliability of SDM predictions in planning for assisted migration. We used Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco), a North American conifer introduced into Europe during the mid-19th century, as a case-study to test niche conservatism. We combined transcontinental data sets of Douglas-fir occurrence and climatic predictors to compare the realized niches between native and introduced ranges. We calibrated a SDM in the native range and compared areas predicted to be climatically suitable with observed presences. The realized niches in the native and introduced ranges showed very limited overlap. The SDM calibrated in North America had very high predictive power in the native range, but failed to predict climatic suitability in Europe where Douglas-fir grows in climates that have no analogue in the native range. We review the ecological mechanisms and silvicultural practices that can trigger such shifts in realized niches. Retrospective analysis of tree species introduction revealed that the assumption of niche conservatism is erroneous. As a result, distributions predicted by SDM are importantly biased. There is a high risk that assisted migration programs may be misdirected and target inadequate species or introduction zones.