Soil microbes provide multiple ecosystem functions such as nutrient cycling, decomposition and climate regulation. However, we lack a quantitative understanding of the relative importance of microbial richness and composition in controlling multifunctionality. This knowledge gap limits our capacity to understand the influence of biotic attributes in the provision of services and functions on which humans depend. We used two independent approaches (i.e. experimental and observational), and applied statistical modeling to identify the role and relative importance of bacterial richness and composition in driving multifunctionality (here defined as seven measures of respiration and enzyme activities). In the observational study we measured soil microbial communities and functions in both tree- and bare soil-dominated microsites at 22 locations across a 1200 km transect in southeastern Australia. In the experimental study we used soils from two of those locations and developed gradients of bacterial diversity and composition through inoculation of sterilized soils. Microbial richness and the relative abundance of γ-Proteobacteria, Actinobacteria and Bacteroidetes were positively related to multifunctionality in both the observational and experimental approaches; however, only Bacteroidetes was consistently selected as a key predictor of multifunctionality across all experimental approaches and statistical models used here. Moreover, our results, from two different approaches, provide evidence that microbial richness and composition are both important, yet independent, drivers of multiple ecosystem functions. Overall, our findings advance our understanding of the mechanisms underpinning relationships between microbial diversity and ecosystem functionality in terrestrial ecosystems, and further suggest that information on microbial richness and composition needs to be considered when formulating sustainable management and conservation policies, and when predicting the effects of global change on ecosystem functions.

Changes after Jun 9, 2018:

-Clarified that approximately 10% of the raw microcosm data was missing and subsequently imputed (KNN/Mean imputation). This has been clarified via corrigendum.

-Corrected Multifunctionality in microcosm B after detecting a small error when log-transforming glucose respiration data in Excel before determining averaging multifunctionality. Data is r ~ 1.00; P < 0.001 correlated with original multifunctionality. This has been clarified via corrigendum.

-Corrected Multifunctionality in field study after detection of an error in two values associated with CB and NAG in site 1 (open). This likely resulted from a transcription error to Excel. Data is r = 0.975; P < 0.001 correlated with original multifunctionality. This has been clarified via corrigendum.

-Added a link to the raw sequencing data used in the Field survey.

-Data has been rounded to three decimal places to reflect the resolution of our laboratory techniques.