Determinants of species abundance for eastern North American trees

Aim We aimed to quantify the relative contributions of trait-based selection and the flow of individuals across space created by differences in species abundance, i.e. mass effects, in determining local-scale species relative abundance (SRA) at 0.07 ha, 1 grid cell and subregion grains. Location The eastern USA. Methods We coupled a maximum entropy model (MaxEnt) with nine species-specific plant traits and a continental-scale forest inventory dataset to perform our analyses. Mass effects were estimated using the next grain size up as the metacommunity (or the eastern USA at the subregion grain, and also at the 1 degrees grid cell grain in a fourth analysis). Permutation tests were conducted to test the significance of trait constraints. The Kullback-Leibler divergence index was used to decompose the information content in each assemblage SRA. Results Trait constraints were significant in predicting community structures at coarse scales, but were largely not significant at fine or medium scales. At the 0.07 ha grain little of the variation in local SRA can be explained by either trait constraints or a regional SRA. At the 1 degrees grain c. 61% of the variation could be accounted for by a regional SRA represented by subregions. At the subregion grain c. 74% of the variation could be accounted for by trait constraints and the SRA of the entire study area. Main conclusions Our results suggest that the relative importance of community assembly processes is scale dependent. Estimated mass effects predominated at small spatial scales where environmental gradients were weak. Trait-based selection was hard to detect at small scales and was strongest at large scales where environmental gradients are stronger. The fact that, at all scales, mass effects accounted for more variation in assemblage SRA than trait-based selection in our analyses suggests that regional processes are important in determining community structures more locally.