Microorganisms have confounded biogeographers because of their high dispersal capability, small size, and vast diversity and abundance. Here we use pyrosequencing, bioinformatics tools, and geospatial modeling to reveal that the genetic relatedness of soil bacteria varies in a predictable pattern across a landscape. Microbial communities showed strong spatial autocorrelation to a distance of 240 meters and this pattern was driven by changes in the genetic relatedness and abundance of specific clades across the landscape. The spatial distribution of important clades across the landscape identified pH, plant abundance, and snow depth as major structuring variables. In turn, the abundance of these clades covaried with biogeochemical functions across the landscape, suggesting unexpected new niches for large groups such as the Rhodospirillales.