Abstract
A species set in a site comprises species that are present (realized diversity) and species that could inhabit this site but are absent (dark diversity; DD). DD can be both species-driven (a species' traits preclude its presence, independently of site features) and site-driven (site features preclude the species' presence, independently of its traits). DD affinity (DDA) is a measure of species' tendencies to be absent from sites that they could inhabit or of sites' tendencies to lack species that could be present. Decomposition of DDA into DDA for species (dda(sp)) and for sites (dda(site)) allows (a) disentangling these two mechanisms and (b) detecting species traits and site features contributing to their DDA. The species-site unified model is a Bayesian statistical model aimed at simultaneously estimating dda(sp) and dda(site). We applied it to flea and mite assemblages (a) within a host species across regions (component metacommunities; CtM; dda(site) = dda(region)) and (b) within a region across host species (compound metacommunities; CdM, dda(site) = dda(host)). In CtMs, dda(sp) and dda(region) equally contributed to DD, whereas the relative contributions of dda(sp) and dda(host) to DD in CdMs varied from the former being higher than the latter and vice versa. In CtM and CdM, dda(sp) increased in low-abundance ectoparasites exploiting a restricted number of hosts. In CtMs, dda(region) was associated with the regional environment, but we failed to find host traits affecting dda(host) in CdMs. We conclude that ectoparasite species and either regions in CtMs or host species in CdMs independently contribute to DD.