Comparing future shifts in tree species distributions across Europe projected by statistical and dynamic process-based models

Abstract

Many tree species are predicted to shift their geographic ranges with changing climate, but the extents, timing, and magnitude of these shifts remain uncertain. Comparing various modeling strategies is crucial for reducing uncertainty related to these responses and for guiding the interpretation of model results. Here, we compared outputs of a dynamic vegetation model (DVM) and an ensemble of statistical bioclimatic envelope models (BEMs) in predicting range shifts of 14 representative tree species in continental Europe. Expanding the number of species and geographic extent compared to previous model comparisons, we found that the DVM produced more conservative range shift estimates, even in long-term equilibrium simulations. The differences in range shift projections were greatest for Mediterranean species, whose expansion northwards was inhibited in the DVM by more competitive prevailing temperate species. In contrast to our expectation, competitive traits of the species studied did not consistently affect the differences. The agreement between BEM and DVM results was highest in boreal species, suggesting that BEMs are an efficient method for modeling species under strong control of abiotic factors. BEMs produced substantially larger range contractions at the southern edge of distribution, in contrary to the DVM, where contractions were more modest. Despite these differences, both approaches also yielded consistent northwards shifts of forest types, which may have substantial negative impacts on forest economy, and alter species composition in natural forest stands.