Abstract
Aim Organisms on our planet form spatially congruent and functionally distinct communities, which at large geographical scales are called "biomes". Understanding their pattern and function is vital for sustainable use and protection of biodiversity. Current global terrestrial biome classifications are based primarily on climate characteristics and functional aspects of plant community assembly. These and other existing biome schemes do not take account of soil organisms, including highly diverse and functionally important microbial groups. We aimed to define large-scale structure in the diversity of soil microbes (soil microbiomes), pinpoint the environmental drivers shaping it and identify resemblance and mismatch with existing terrestrial biome schemes. Location Global. Time period Current. Major taxa studied Soil eukaryotes and prokaryotes. Methods We collected soil samples from natural environments world-wide, incorporating most known terrestrial biomes. We used high-throughput sequencing to characterize soil biotic communities and k-means clustering to define soil microbiomes describing the diversity of microbial eukaryotic and prokaryotic groups. We used climatic data and soil variables measured in the field to identify the environmental variables shaping soil microbiome structure. Results We recorded strong correlations among fungal, bacterial, archaeal, plant and animal communities, defined a system of global soil microbiomes (producing seven biome types for microbial eukaryotes and six biome types for prokaryotes) and showed that these are typically structured by pH alongside temperature. None of the soil microbiomes are directly paralleled by any current terrestrial biome scheme, with mismatch most substantial for prokaryotes and for microbial eukaryotes in cold climates; nor do they consistently distinguish grassland and forest ecosystems. Main conclusions Existing terrestrial biome classifications represent a limited surrogate for the large-scale diversity patterns of microbial soil organisms. We show that empirically defined soil microbiomes are attainable using metabarcoding and statistical clustering approaches and suggest that they can have wide application in theoretical and applied biodiversity research.