Background/Question/Methods  The climatic water balance is regarded as the most functionally significant control of the distribution of biomes and plant species, and is a fundamental process driving dynamic global vegetation models. However, the water balance has rarely been implemented in a wide array of statistical approaches of projecting climate change on species distribution. Likewise, the role of recurrent fire is a secondary process that affects the position of vegetation ecotones, but the spatial scale of the intermediating effect of fire on vegetation has rarely been quantified. We analyzed the spatial pattern of 34 vegetation classes in northwest North America at a resolution of 400 meters with respect to the climatic water balance and estimates of presettlement fire intervals from > 40 paleoecological studies. The “climatic signature” of each vegetation class was summarized using a Gaussian mixture model (GMM) on four alternative predictor variable sets. Each GMM was applied to the current climate to examine classification accuracy, and to future climate scenarios from output of eight Global Climate Models (2080-2099 AD). This analysis produced 1224 maps of projected distributions (34 vegetation classes, four predictor sets, and nine climate change scenarios).

Results/Conclusions  Our analysis shows that a set of three variables (actual evapotranspiration, AET; deficit, D; mean temperature of the coldest month, MTCO) distinguished among vegetation classes similarly or better than other common predictor variable sets. Similarly, these variables resulted in few areas of “no-analog” future climates (i.e., combinations of climate variables unlike any that exist today). In contrast, other commonly used variables resulted in a greater extent of no-analog climates, indicating the importance of using ecophysiologically meaningful variables when assessing analog conditions. In areas where analog climates existed, the projected shifts in vegetation classes were largest at the dry ends of vegetation gradients. Presettlement fire intervals were shortest in areas with a high AET (productivity) and D (summer drought), which occurred near the lower forest treeline. In a discriminant analysis where climate and presettlement fire intervals were predictors of vegetation type, fire provided no additional explanatory power with the exception of grasslands and the driest forest classes. These results show that a simple climatic water balance is a stronger predictor of major patterns of vegetation than previously shown, that historical fire explains forest boundaries in several regions, and that the presence of no-analog climatic conditions is highly sensitive to the set of variables employed.