A novel multiple-site extension to pairwise partitioned taxonomic beta diversity

Our understanding of the causes of variation in taxonomic composition, or beta diversity, is progressing rapidly, thanks in part to recent methodological advances. For example, methods for partitioning beta diversity into its “replacement” and “richness” components have helped reveal patterns that had been undetected by traditional analyses. These partitioning methods are derived from pairwise dissimilarity measures, and are thus well suited to many conventional beta diversity analyses, including “distance decay” relationships. However, pairwise beta diversity measures have limitations, including their lack of information about taxa that are shared among three or more sites. Recently, a new suite of multiple-site counterparts to the pairwise partitioning measures of beta diversity was proposed, but the pairwise analogs upon which these were based were subsequently criticized, and compelling arguments were presented in favor of other partitioning approaches. Here, we introduce multiple-site partitioning measures that address these shortcomings, and illustrate their desirable properties using numerical simulations. We also provide an empirical example of their utility by analyzing the temporal beta diversity of breeding birds within the conterminous USA. We show that temporal beta diversity is predominantly driven by replacement rather than richness differences, and correspondingly, that correlations between temporal beta diversity and productivity and elevation are driven primarily by the replacement component. Furthermore, in contrast to existing multiple-site measures, we show that richness differences do play an important part in driving overall beta diversity patterns. Our new multiple-site measures therefore complement existing methods for analyzing beta diversity, and are especially suitable when compositional heterogeneity is the response of interest.