Although dispersal has long been recognized as a key process in the assembly of plant and animal communities, its role in shaping microbial communities has only recently been established [1][2][3]. Directly tracking the movement of individual microorganisms in the field remains a technical challenge. As such, the fundamental properties of microbial dispersal, including the rate and distance that free-living microbes move in their environments, are largely unknown.

Dispersal rate, or the number of individuals immigrating into a defined area per unit time, is an important determinant of how strongly dispersal impacts community structure (species abundance and composition) relative to other processes, such as environmental selection [4,5]. Using microbial "traps" (e.g. filter paper, glass slides, or air samplers), previous studies have quantified the rate at which bacteria and fungi disperse in aquatic [6], aerial [7], and terrestrial [8,9] systems. However, these measurements remain rare for bacteria, and it is unclear the degree to which dispersal rates vary across space and thus may contribute to the biogeography of microbial communities.

To investigate the spatial variability of microbial dispersal, we quantified the rate of bacterial immigration along 30-m transects spanning two ecosystems, a grassland and shrubland (Supplementary methods, Fig. 1a). We placed two types of glass slides on the soil surface, sterile slides open to dispersal (accumulation rate samples, n = 96) and slides closed to dispersal (death rate samples, n = 35) and collected the slides over 2 months (Fig. 1b and c). Using f low cytometry, we measured the number of bacterial cells accumulating on the open glass slides and the number of cells declining on the death rate samples over time. The glass slides capture cells moving into the surface soil from different sources (e.g. soil, vegetation, or the atmosphere) and physical vectors (e.g. wind or rain), but provide no nutrients for growth. Therefore, the number of cells accumulating on the open slides (Fig. 2a) ref lects the immigration rate minus the death rate of cells (see Supplementary methods). To quantify cell death, death rate samples were inoculated with a microbial community extracted from either grassland or shrubland leaf litter before being placed in the field. Although death rates may vary between different source communities, leaf litter was previously demonstrated to be a major source of microbes immigrating into the surface soil at this site [9]. Additionally, ecosystem-specific litter was used, because microbial composition varies between the grassland and shrubland [10].

Death rate did not differ significantly between the grassland and shrubland (Table S1; ANCOVA: Time × Ecosystem interaction P value = .17) with 2.76% of the community dying on average per day across the landscape (Fig. 2a). Accumulation rate, or the number of intact cells (detected by f low cytometry) captured on the open slides over time, also did not differ between ecosystems (Fig. 2a; Table S2; P value = .61). Assuming a dynamic relationship between immigration and death rates (see Supplementary methods), the rate of bacterial dispersal into the soil surface averaged 1060 ± 90 cells/cm 2 /day (±SE). This rate represents 0.04% of the average bacterial abundance found in leaf litter at this research site (∼2.8 million cells/cm 2 ). The relatively small fraction of cells moving into the topsoil per day suggests that dispersal is likely not homogenizing the bacterial community on the soil surface (i.e. mass effects [4]).

Immigration rate captures how fast cells move into the surface soil, but it does not indicate how far cells are moving-i.e. the degree to which cells are dispersal limited. Dispersal limitation is critical for predicting the spread and gene f low of a specific microbial species, such as a pathogen [11], as well as the successional dynamics and biogeography of microbial communities as a whole [12]. In plants, dispersal limitation is commonly reported as a dispersal kernel: a probability density function describing the likelihood that seed deposition will occur at certain distances from a parent plant [13]. Dispersal kernels have been characterized for a few fungal species [8,14,15] and reveal that the degree of dispersal limitation of fungal spores ranges widely across species. These studies have focused on species that are host-associated, allowing for their abundance to be quantified at increasing distances from a known point source. In contrast, characterizing the dispersal kernel of free-living microorganisms adds another layer of complexity, because they may be dispersing simultaneously from many sources.

To address the challenge of tracking individual microbes, we used a community approach to estimate dispersal limitation in the field by measuring how the composition of dispersing microbes was impacted by surrounding vegetation, an important source of dispersal at our field site [16]. The composition of both bacteria and fungi on the accumulation rate slides were distinct between the grassland and shrubland (Fig. 2b; Table S3; PER-MANOVA: P value ≤.001; estimated variance explained by ecosystem: 3% and 18% for bacteria and fungi respectively). Within ecosystems, fungal, but not bacterial, composition on the slides was more variable in the shrubland than that in the grassland (Table S3; PERMDISP: P value <.001), potentially ref lecting greater heterogeneity of plant composition in the shrubland. These differences in the composition of microbes dispersing into the soil indicates that some bacterial and fungal cells are dispersal limited at this study's scale, within the 30 m transects.

Given that the grassland and shrubland are immediately adjacent with no major differences in slope, aspect, soil type, or climate [17], we hypothesized that shifts in plant community composition between the ecosystems underlie these distinct dispersal communities. To test the spatial scale at which plant composition most strongly contributes to the identity of immigrating taxa, we measured the correlation between microbial composition on the accumulation rate slides and plant composition within 11 increasingly large circles (Fig. 1d and e), with radii ranging from 0.1 to 4 m away from the slides. When controlling for variation explained by geographic distance among samples, the strength of the correlation between the microbial composition on the open slides and surrounding plant composition was highest at around 1 m for both bacteria and fungi (Fig. 2c; Table S4; Mantel's r = 0.13 and r = 0.59, respectively). This highly local signal of dispersal limitation corroborates the strong compositional differences in dispersing microbes between the grassland and shrubland (Fig. 2b). The correlation was also much stronger for fungi than bacteria at all radii. This result may ref lect stronger relationships between plants and fungi or alternatively, shorter average dispersal distances of fungi compared to bacteria [18]. Indeed, a SourceTracker analysis of potential dispersal sources (air, plant litter, and soil) suggests that dispersal from leaf litter contributed more to the fungal communities captured on the accumulation slides compared to the bacterial community (Fig. S1).

Together, our results quantify fundamental properties of microbial dispersal in a terrestrial ecosystem. Bacteria disperse into the soil surface at a similar, albeit relatively low, rate across a landscape. This rate seems to be determined by abiotic conditions (e.g. wind speed, precipitation, and landscape topography) that are shared between the grassland and shrubland [19,20]. Even though overall dispersal rates were not spatially variable, the composition of dispersing microbes was highly localized, with a signature of dispersal limitation detected at a meter scale for both bacteria and fungi. This result does not mean that individual microorganisms are not moving much longer distances, but that enough are restricted to shorter distances to generate these patterns. This local signature indicates that dispersal may interact with other eco-evolutionary processes, like genetic and ecological drift, to shape biodiversity and biogeography of microbial communities even at a meter scale [21].