Selection and dispersal are ecological processes that have contrasting roles in the assembly of communities. Variable selection diversifies and strong dispersal homogenizes them. However, we do not know whether dispersal homogenizes communities directly via immigration or indirectly via weakening selection across habitats due to physical transfer of material, e.g., water mixing in aquatic ecosystems. Here we examine how dispersal homogenizes a simplified synthetic bacterial metacommunity, using a sequencing-independent approach based on flow cytometry and mathematical modeling. We show that dispersal homogenizes the metacommunity via immigration, not via weakening selection, and even when immigration is four times slower than growth. This finding challenges the current view that dispersal homogenizes communities only at high rates and explains why communities are homogeneous at small spatial scales. It also offers a benchmark for sequence-based studies in natural microbial communities where immigration rates can be inferred solely by using neutral models.

In community ecology, drift refers to random births and deaths in a population. In microbial ecology, drift is estimated indirectly via community snapshots but in this way, it is almost impossible to distinguish the effect of drift from the effect of other ecological processes. Controlled experiments where drift is quantified in isolation from other processes are still missing. Here we isolate and quantify drift in a series of controlled experiments on simplified and tractable bacterial communities. We detect drift arising randomly in the populations within the communities and resulting in a 1.4-2% increase in their growth rate variability on average. We further use our experimental findings to simulate complex microbial communities under various conditions of selection and dispersal. We find that the importance of drift increases under high selection and low dispersal, where it can lead to~5% of species loss and to~15% increase in β-diversity. The species extinct by drift are mainly rare, but they become increasingly less rare when selection increases, and dispersal decreases. Our results provide quantitative insights regarding the properties of drift in bacterial communities and suggest that it accounts for a consistent fraction of the observed stochasticity in natural surveys.

In the original version of this article, the green and blue outlines in Figure 2b, top centre and right panels were inadvertently shifted left from the correct position. This has now been corrected in the PDF and HTML versions of the article.