A bioclimate model predicting the presence or absence of aspen, Populus tremuloides, in western USA from climate variables was developed by using the Random Forests classification tree on Forest Inventory data from about 118,000 permanent sample plots. A reasonably parsimonious model used eight predictors to describe aspen's climate profile. Classification errors averaged 4.5%, most of which were errors of commission. Themodel was driven primarily by three variables: an annual dryness index, the ratio of summer to annual precipitation, and an interaction of growing season precipitation with the summer-winter temperature differential. Projecting the contemporary climate profile into the future climate provided by three General Circulation Models and two scenarios (SRES A2 and either B1 or B2) suggested that the area occupied by the profile should diminish rapidly over the course of the century, 6-41% by the decade surrounding 2030, 40-75% for that surrounding 2060, and 46-94% for 2090. The relevance of the climate profile to understanding climate-based responses is illustrated by relating trends in climate to the recent incidence of sudden aspen dieback that has plagued portions of the aspen distribution. Of the eight variables in the profile, four reached extreme values during 2000-2003, the period immediately preceding the appearance of damage in aerial surveys.