Beta diversity partitioning
Spatial variation in plant community composition was estimated from two components, species replacement (turnover component, βturnover) and changes in species richness (nestedness component, βnestedness), which together contribute to the total β-diversity (βtotal). To test for effects on β-diversity that are independent of species richness differences, we first used the Simpson index of dissimilarity to examine how βturnover changed along the environmental gradient. We also calculated the βtotal with the Sørensen index to examine changes of the total β-diversity. We calculated the two β-diversity components via the betapart package in R (Baselga & Leprieur 2015). All these analyses were based on species presence/absence data.
Generalized Dissimilarity Modelling (GDM)
We used the GDM approach to analyse β-diversity patterns along environmental gradients, which is widely used to identify important environmental drivers for β-diversity and to test the independent significance of these drivers (using permutation tests). The advantage of GDMs is that they allow nonlinear relationships between dissimilarity and distance (Ferrier et al. 2007; Fitzpatrick et al.2013). The environmental matrix in our study included habitat variables (MAT, MAP, altitude, soil N, and soil P) and geographical distances (i.e. spatial distance from latitude and longitude) between sites.
We plotted the partial effect of each predictor against the level of a given predictor to visualize the results of each GDM (holding all other predictors constant). The maximum height of the line shows the relative importance of the studied predictor in explaining the variation of β-diversity in the model. The shape of the line shows how β-diversity varies along each environmental or spatial gradient, i.e. how the effect of a given predictor on β-diversity varies at a given level of that predictor. Furthermore, we also determined the proportion of deviance uniquely attributable to environment or distance, by comparing the deviance explained by a GDM containing all of the variables and a GDM with all variables except environment or distance, respectively. The unique deviance explained by environment or distance was calculated as the difference in deviance explained by these models. We then converted this to a percentage by dividing the deviance explained by the full GDM. These percentages can indicate the relative importance of geographic distance among sites (linked to dispersal limitation processes) and environmental heterogeneity (linked to niche differentiation processes) in determining variability in β-diversity (Chase & Myers 2011). GDMs were fitted to the β-diversity for turnover component (βturnover) and total (βtotal), separately, using the gdm function in the gdm library (Manion et al. 2017). The results were generally similar for βturnover and βtotal (Fig. 2-3 vs. Fig. S3-4) and therefore we only reported the results for βturnover in the man text for simplicity.