A key question at the interface of ecology, biogeography and evolution is whether community assembly mechanisms vary systematically across large-scale biodiversity gradients. Theory suggests that community assembly may shift from stronger stochastic assembly in high-diversity communities to stronger deterministic assembly in low-diversity communities. In high-diversity communities, rare species may interact infrequently with one another, resulting in more unpredictable outcomes of interspecific competition, more random changes in species and trait composition among local neighborhoods of interacting species, and more random co-occurrences among species pairs. In this study, we tested this hypothesis across a large-scale (4000 m) elevational-diversity gradient in the Bolivian Andes. Using 30 tree communities sampled in 1-ha forest plots (15-99 species per plot), we tested the degree to which species and trait composition of local tree neighborhoods is non-randomly assembled from local species pools. We utilized a null-model that breaks interspecific and intraspecific aggregation patterns while maintaining intraspecific aggregation patterns of the focal species. Comparing empirical neighborhood composition to null expectations allows us to disentangle the effects of stochastic sampling and local biotic interactions.
Results/Conclusions
Across forest plots, mean compositional dissimilarity among local tree neighborhoods increased with tree-species diversity. This relationship was strongest at the smallest neighborhood scale (5-m neighborhood radius: r2 = 0.81) and weakest at the largest neighborhood scale (20-m neighborhood radius: r2 = 0.55). After comparing the observed relationships to those expected from the null model, we found that local tree neighborhoods were more randomly assembled in higher-diversity communities. Standardized effect sizes in the three highest-diversity plots deviated little from the null expectation. In contrast, standardized effect sizes in the three lowest diversity-plots were generally positive, indicating that variation in species composition among local tree neighborhoods was higher than expected from random assembly. Across this biodiversity gradient, our results suggest that there is a geographic shift from more deterministic species interactions in low-diversity communities to more stochastic interactions in high-diversity communities. Our findings have implications for understanding how communities, especially those at high elevations, might respond to changing environmental conditions.