Assessing statistical tests of phylogenetic beta diversity

Background/Question/Methods

Synthesizing species' evolutionary relationships with spatial distribution patterns, phylogenetic beta-diversity (PBD) supplies the temporal dimension necessary for deciphering what processes govern species assemblage; where, assemblage processes are ones significantly associated with PBD. Given its simplicity and flexibility, the Mantel test and partial Mantel test are commonly employed to confer matrix association---these methods may however be inappropriate for accurately assessing matrix association.

Here, we present PROTEST and partial PROTEST as accessible and versatile matrix association alternatives to the Mantel tests that directly employ raw data. We framed our comparison in application to phylogenetic beta-diversity, a nascent avenue for examining ties between communities' structural and evolutionary relationships across space. For angiosperm communities in the Rocky Mountain National Park (Colorado, USA), we associate three taxonomic (Jaccard, Bray-Curtis and Gower) and six phylogenetic (PhyloSor, UniFrac, D_nn, D_pw, Rao's D, and Rao's H) beta-diversity indices to spatial and abiotic processes. First and foremost, we were interested in how the Mantel methods compared to the PROTEST methods; Second, if phylogenetic relationships provide novel insights uncaptured by taxonomic beta-diversity metrics; Third, if indices' sensitivity to evolutionary depth matter; and finally whether space and/or environment play a crucial role in species assemblage across an environmental gradient.

Results/Conclusions

We found the PROTEST superimposition methods to have more significant values than the Mantel and partial Mantel tests. Furthermore, PBD metrics were indeed distinct from TBD; but among the PBD metrics, some values were redundant (Rao's D & D_pw and D_nn & PhyloSor). Among the PBD metrics, indices with sensitivities to more recent phylogenetic shifts (t-PBD) were discernibly unique from indices sensitive to more historic events (b-PBD). Finally, while both space and environment affect species turnover, the underlying environmental gradient does so more strongly.