Multi-site generalised dissimilarity modelling: using zeta diversity to differentiate drivers of turnover in rare and widespread species

24 May 2017

1. Generalised dissimilarity modelling (GDM) applies pairwise beta diversity as a measure of species turnover with the purpose of explaining changes in species composition under changing environments or along environmental gradients. Beta diversity only captures turnover across pairs of sites and, therefore, disproportionately represents turnover in rare species across communities. By contrast, zeta diversity, the average number of shared species across multiple sites, captures the full spectrum of rare, intermediate and widespread species as they contribute differently to compositional turnover. 2. We show how integrating zeta diversity into GDMs (which we term multi-site generalised dissimilarity modelling, MS-GDM), provides a more information rich approach to modelling how communities respond to environmental variation and change. We demonstrate the value of including zeta diversity in biodiversity assessment and modelling using BirdLife Australia Atlas data. Zeta diversity values for different numbers of sites (the order of zeta) are regressed against environmental differences and distance using two kinds of regressions: shape constrained additive models and a combination of I-splines and generalised linear models. 3. Applying MS-GDM to different orders of zeta revealed shifts in the importance of environmental variables in explaining species turnover, varying with the order of zeta and thus with the level of co-occurrence of the species and, by extension, their commonness and rarity. In particular, precipitation gradients emerged as drivers in the turnover of rare species, whereas temperature gradients were more important drivers of turnover in widespread species. 4. Appreciation of the factors that drive compositional turnover across multiple sites is necessary for accommodating the full spectrum of compositional turnover across rare to common species. This extends beyond understanding drivers for pairwise beta diversity only. MS-GDM provides a valuable addition to the toolkit of GDM, with further potential for survey gap analysis and prediction of species composition in unsampled sites.