If species responses to the environment can be predicted from a small number of measurable traits, then trait responses from isolated continents should be comparable. However, there are few tests of whether or not community weighted mean trait values respond to the environment in consistent ways across continents that are isolated from one another. Here, we synthesized eight national forest inventory datasets from two ecologically and climatologically comparable temperate regions: North America (north of Mexico) and Western Europe. Using generalized joint attribute modeling, we created two multivariate species distribution models based on the composition of basal area for 193 tree species in North America and 112 in Europe. To determine if traits jointly respond to the environment in the same ways across these two continents, we generated community weighted mean trait predictions and inference through a change-of-variable approach. We determined if traits on each continent could be accurately predicted with an model-based “artificial transplant” of the species from the opposing continent.
Results/Conclusions
High out-of-sample prediction accuracy within the continent in which models were fitted, confirmed strong within-continent controls on trait distributions. However, many traits were poorly predicted when floras were artificially transplanted to the alternative continent (Spearman’s rank correlations below 0.3). Despite a comparable number of inventory plots (135,2189 for North America and 169,225 for Europe) and comparable variation in environmental covariates (temperature, water balance index, and soil information), trait variation in North American is better explained by environment than in Europe. The relative importance of different covariates in explaining trait variation were similar between the continents (Spearman’s rank correlation 0.75), but North American community traits were overall more sensitive to every covariate than in Europe. Nonetheless, out-of-sample predictions of European plant traits using the North American model fit was unexpectedly poor. Taken together, these results suggest that community weighted traits may not have equivalent responses to environmental gradients even within ecologically similar biomes. Different continent-level responses may result from the very different glacial legacies, giving rise to divergent evolutionary histories that respond differently to climate as well as forest management.