Inferring Ecological Processes from Taxonomic, Phylogenetic and Functional Trait β-Diversity

Understanding the influences of dispersal limitation and environmental filtering on the structure of ecological communities is a major challenge in ecology. Insight may be gained by combining phylogenetic, functional and taxonomic data to characterize spatial turnover in community structure (beta-diversity). We develop a framework that allows rigorous inference of the strengths of dispersal limitation and environmental filtering by combining these three types of beta-diversity. Our framework provides model-generated expectations for patterns of taxonomic, phylogenetic and functional b-diversity across biologically relevant combinations of dispersal limitation and environmental filtering. After developing the framework we compared the model-generated expectations to the commonly used "intuitive" expectation that the variance explained by the environment or by space will, respectively, increase monotonically with the strength of environmental filtering or dispersal limitation. The model-generated expectations strongly departed from these intuitive expectations: the variance explained by the environment or by space was often a unimodal function of the strength of environmental filtering or dispersal limitation, respectively. Therefore, although it is commonly done in the literature, one cannot assume that the strength of an underlying process is a monotonic function of explained variance. To infer the strength of underlying processes, one must instead compare explained variances to model-generated expectations. Our framework provides these expectations. We show that by combining the three types of b-diversity with model-generated expectations our framework is able to provide rigorous inferences of the relative and absolute strengths of dispersal limitation and environmental filtering. Phylogenetic, functional and taxonomic b-diversity can therefore be used simultaneously to infer processes by comparing their empirical patterns to the expectations generated by frameworks similar to the one developed here.