Modelling selection, drift, dispersal and their interactions in the community assembly of Amazonian soil mites.

Three processes can explain contemporary community assembly: natural selection, ecological drift and dispersal. However, quantifying their effects has been complicated by confusion between different processes and neglect of expected interactions among them. One possible solution is to simultaneously model the expected effects of each process within species, across communities and across species, thus providing more integrative tests of ecological theory. Here, we used generalized linear mixed models to assess the effects of selection, drift and dispersal on the occurrence probability of 135 soil oribatid mite species across 55 sites over an Amazonian rainforest landscape (64 km2). We tested for interactions between process-related factors and partitioned the explained variation among them. We found that occurrence probability (1) responded to soil P content and litter mass depending on body size and reproductive mode (sexual or parthenogenetic), respectively (selection); (2) increased with community size (drift); and (3) decreased with distance to the nearest source population, and more so in rare species (dispersal limitation). Processes did not interact significantly, and our best model explained 67% of the overall variation in species occurrence probability. However, most of the variation was attributable to dispersal limitation (55%). Our results challenge the seldom-tested theoretical prediction that ecological processes should interact. Rather, they suggest that dispersal limitation overrides the signatures of drift and selection at the landscape level, thus rendering soil microarthropod species ecologically equivalent and possibly contributing to the maintenance of metacommunity diversity.