Variation in species composition among sites can provide significant insights into mechanisms underlying the generation and maintenance of biodiversity. Among biogeographic regions, beta-diversity typically decreases with increasing latitude and elevation. Geographic gradients in beta-diversity may result from parallel gradients in species characteristics (e.g., geographic range size or dispersal ability), characteristics of geographic regions (e.g., environmental heterogeneity or productivity), or the strength of local community assembly processes (e.g., habitat filtering or interspecific competition). In addition, large-scale gradients in beta-diversity could be shaped by random sampling effects when there are strong gradients in the size of regional species pools (i.e, gamma-diversity). In this study, we tested the extent to which elevational gradients in beta-diversity are driven by variation in gamma-diversity and non-random assembly processes at multiple spatial scales. Using a unique, large-scale dataset on the distribution of woody plants across a 4000-m elevational gradient in the Bolivian Andes, we compared the observed elevational gradient of beta-diversity to the gradient expected given a simulation model based on random assembly from observed regional species pools. Analyses were repeated at small and large spatial scales, which spanned an order of magnitude difference in area and geographic distance.
Results/Conclusions
We found strong elevational gradients in observed diversity. Irrespective of scale, gamma- and alpha-diversity decreased monotonically with elevation. In contrast, the elevational gradient in beta-diversity differed across spatial scales. At the small scale, beta-diversity decreased monotonically with elevation, while at the large scale, beta-diversity peaked towards mid-elevations. The simulation model of random assembly from observed species pools predicted linear decreases in beta-diversity with elevation at both scales. When controlling for differences in regional species pools across the elevational gradient, scale-dependent patterns of beta-diversity persisted. At the small scale, differences between observed and simulated beta-diversities (i.e., beta-deviations) peaked at mid-elevations, but the relationship was relatively weak. At the large scale, in contrast, there was a strong relationship between beta-deviations and elevation with a significant mid-elevation peak. In addition, beta-deviations were almost always greater than expected by the simulation model, and were consistently larger at the large scale relative to the small scale. We conclude that variation in gamma-diversity alone cannot account for elevational gradients in beta-diversity. Instead, our results suggest that the apparent importance of stochastic assembly in the production of beta-diversity gradients is strongly dependent on spatial scale, and that the relative importance of non-random assembly processes may peak towards intermediate elevations.