The decrease of beta diversity with increasing altitude in Central European forests

Patterns of beta diversity variation can provide insights into the assembly processes that shape plant communities. However beta diversity metrics are linked to variation in gamma and alpha diversity which make statistical comparisons difficult. Recently null models have been proposed to disentangle the multi-scale effects of processes operating either at the plot-scale (affecting alpha diversity) or at large biogeographical scales (influencing gamma diversity). Thus, the deviation between the beta diversity observed and that expected under a null model indicates the existence of ecological processes that generate non-random spatial patterns in vegetation assemblages along biogeographic or environmental gradients. We aim at testing whether a significant relationship exists between beta diversity and altitude after controlling for the confounding effect of alpha and gamma diversity with the use of null models. We used the forest egetation of the Czech Republic as a case study. More than 19,000 vegetation plots were extracted from the Czech National Phytosociological Database, and were divided into 10 altitudinal ransects, i.e. ecoregions comprising altitudinal gradients from lowland to mountain tops. We implemented three resampling schemes to alleviate possible bias due to the uneven distribution of plots across the study area, and tested the robustness of the inference by comparing the results obtained using the whole dataset and different resampling schemes. We used GLMMs to model the patterns of beta diversity and of its deviation from random expectations, along the altitudinal gradient while controlling for climatic and topographical variability, and the geographical spread of the plots. Model selection was performed using an Information-Theory approach that fully exploits the information contained across permutations. Beta diversity declined with increasing altitude because of a steeper decline in gamma than in alpha diversity. After controlling for the effects of alpha and gamma diversity using null models, this trend disappeared. This result was relatively robust both across different resampling schemes and across diversity metrics that give progressively less weight to rare species. Our study supports the hypothesis that the decreasing beta diversity (compositional heterogeneity) of forest communities with increasing altitude depends on the reduced species pool of high-altitude vegetation, and not on the variation of community assembly processes.