Discussion

Biogeographic Patterns of Microbial Properties

We found significant biogeographic patterns of fungi, bacteria and their balance in topsoil along latitude, climate (MAP and MAT), plant (NPP and Croot), soil microclimate (SM and ST), and edaphic factors (SOC, TN, C:N ratio, soil pH, soil texture, and bulk density) (Fig. S2-6 ). Some of those have been reported in previous studies (Fierer et al. 2009; Waring et al. 2013; Chenet al. 2016; Bahram et al. 2018). For example, Bahramet al. (2018) also reported the inverse unimodal trend of BBC and positive linear trend of F:B ratio along latitude and significant positive linear trend of F:B ratio along MAP and MAT, Fierer et al. (2009) reported significant controls of plant NPP and microbial biomass, Waring et al. (2013) showed that F:B ratio decreased with low level C:N ratio, and increased at high C:N ratio, and de Vrieset al. (2012) found that finely textured soils tend to have higher fungal and bacterial biomass.
In addition, we did find different results compared with previous studies. Specifically, in contrast to the inverse unimodal trend of FBC along latitude, Bahram et al. (2018) found the significant positive linear relationships between FBC and latitude. Also, we observed the inverse unimodal relationship between F:B ratio and soil pH, with lowest F:B ratio at soil pH of 6.3, while Chen et al.(2015) reported a significant positive relationship between F:B ratio and soil pH in Mongolian Plateau and Eskelinen et al. (2009) found a negative relationship between F:B ratio and soil pH in the alpine tundra of northern Europe. These discrepancies may result from two reasons. First, the difference in sample size may lead to the variations in the relationships obtained among studies. Dataset in Bahram et al. (2018) was built based on the globally selective sampling plots (145 topsoil samples), while the dataset of this study is a comprehensive meta-analysis dataset with 1323 data points (Fig. 1 ). Second, the difference in spatial scales of research is responsible for the plausible distinction. Chen et al. (2015) covered soil pH>6.5, while Eskelinen et al. (2009) contained sampling sites of soil pH ranging from 4.7 to 7.0. Merging the negative and positive relationships between F:B ratio and soil pH found by Chen et al. (2015) and Eskelinen et al. (2009) reach the similar results as reported in this study.
FBC and BBC were largely distinct among biomes, but we observed generally similar patterns for FBC and BBC among biomes (Table 1 ). Consistent with our results, Xu et al. (2013) also found the highest soil microbial biomass in tundra among biomes, and soil microbial biomass was significantly higher in boreal forests than that in temperate forests and tropical/subtropical forests. Both Fiereret al. (2009) and Xu et al. (2013) reported lowest soil microbial biomass in deserts, the low SOC concentration may result in low FBC and BBC in deserts (Fig. S6 ). However, this study generated slightly different results from previous studies. Among forest biomes, Fierer et al. (2009) reported the higher soil microbial biomass in temperate and tropical forests than that in boreal forests, which exhibited opposite patterns with this study. In addition, soil microbial biomass in temperate forests was significantly higher than that in tropical/subtropical forests reported by Xu et al.(2013), while both FBC and BBC were significantly higher in tropical/subtropical forests than that in temperate forests in this study (Table 1 ). The seasonality of FBC and BBC could be a source for the inconsistency. Microbial biomass showed strong seasonal dynamics, samples taken in growing and non-growing seasons are expected to have distinct microbial biomass concentrations (Lipson et al.2002).
Our results also showed that F:B ratio was distinct among biomes, with the smallest F:B ratio in savanna and the highest in tundra (Table 1 ). Similar to our findings, Bahram et al. (2018) found significantly higher F:B ratio in boreal-arctic biomes (e.g., tundra and boreal forests) and temperate biomes (e.g., temperate forests and grassland) than that in tropical biomes (e.g., savanna and tropical/subtropical forests). Additionally, we found significantly higher F:B ratio in grasslands than that in pastures, which is consistent with de Vries et al. (2012), suggesting that management practices enhance the dominance of bacteria.
We estimated FBC and BBC storage in topsoil as 12.56 Pg C and 4.34 Pg C, respectively (Table 2 ). This result is consistent with overall terrestrial biomass estimates of FBC and BBC storage of 12 Pg C and 7 Pg C, respectively, in Bar-On et al. (2018). Differences in methods probably account for most of the differences between the results reported in these studies. Fungi are more sensitive to anoxic conditions, and bacteria and archaea are important components in deep soils such as subsurface environments (Bar-On et al. 2018). It is likely that the differences in the soil depths between this study (0-30 cm) and Bar-On et al. (2018) (entire soil profile) might underpin the discrepancy in estimated global budget of BBC.

Mechanisms for the Microbial Biogeography

We demonstrated that different factors underpin the biogeographic patterns of FBC, BBC, and F:B ratio in topsoil. These patterns can be related to the different nutrient stoichiometry of the microbial groups and their ecological tolerance. First, saprotrophic fungi have more efficient enzymatic machinery than bacteria to obtain C from complex organic material with high C:N ratio (de Vries et al. 2012; Chenet al. 2015). Second, highly carbon-rich soils usually display low soil pH that is relatively more difficult to cope with for bacteria compared with fungi (Eskelinen et al. 2009; Rousk et al.2010). These two interacting mechanisms may favor fungi-dominated ecosystem C and nutrient cycling in tundra and boreal forest ecosystems that exhibits particularly high F:B ratio. Third, fungi were more adapted to low‐temperature conditions and more heat-tolerant than bacteria (Pietikäinen et al. 2005). Meanwhile, fungi dominate early stages of litter decomposition that is more common in high-latitude than low-latitude (Steidinger et al. 2019). In concordance, we also found significantly higher F:B ratio in boreal forest and temperate forest than that in tropical/subtropical forest (Table 1 ). Grasslands feature significantly higher F:B ratio than pasture, indicating that management practices might enhance the dominance of bacteria as shown by de Vries et al. (2012). Natural wetland has F:B ratio comparable to unvegetated ground, desert, and shrub, which might be relevant to the low availability of oxygen that inhibits the growth of fungi and most soil bacteria (Lin et al.2012).
Edaphic properties rather than climatic variables determine much of the variation in FBC and BBC globally (Fig. 2a-b), which is consistent with Chen et al. (2016). This result indicates that FBC and BBC variations are driven by soil pH, SOC, nutrients (e.g., nitrogen and phosphorus), and soil texture (clay, silt, and sand) that control the availability of C energy, nutrients and oxygen – all determinants of fungal and bacterial growth (Brockett et al. 2012; de Vrieset al. 2012). Edaphic properties determine the nutrient and water availabilities, and even shelter from predation (Chapin et al.2011). Specifically, soil pH strongly influences abiotic factors, such as C availability (Andersson et al. 2000), nutrient availability (Pietri & Brookes 2008), and the solubility of metals (Firestoneet al. 1983).
In striking contrast to bacterial and fungal biomass, our results suggest that climate is the most important factor governing the F:B ratio (Fig. 2c ). Climate has a principal effect on soil properties and vegetation activities that control soil microbial composition (Classen et al. 2015; Bahram et al. 2018). Although the tight association between plant community and soil microbial community structure has been reported (Bardgett et al.1998), we observed negligible effects of plant productivity (NPP and Croot) on topsoil microbial community structure. It is likely due to two reasons: 1) we used plant productivity to represent the effect of plants in this study, while growing evidence points to plant functional traits as drivers of soil biological processes at a range of spatial scales (Bardgett & Wardle 2010); 2) vegetation distribution and functional traits are largely determined by climate (Aerts 1997), the inclusion of both climate and plants into the model may result in the dominant role of climate in deciding soil microbial community.

Implications for Global Carbon Cycle

We estimated the ratio of FBC and BBC to SOC as 1.8% and 0.6%, respectively, which agrees with the findings that microbial biomass C (MBC) generally comprises 0.5-5% of SOC (Insam 1990). Soil microbes, the living fraction of soil organic matter, have a much faster turnover rate than soil organic carbon (Xu et al. 2017); meanwhile, the BBC bas faster turnover rate than FBC (Bååth 1998; Rousk & Bååth 2007b). The changes in the MBC:SOC ratio indicates the integrated effects of soil organic matter input, soil microbial C use and C losses, and mineral protection of SOC; therefore, MBC:SOC ratio has been suggested as a useful and meaningful indicator of changes in soil organic matter status (Powlson & Jenkinson 1981; Sparling 1992).
However, MBC:SOC ratio was not constant. In addition to the natural variations due to the seasonal dynamics of MBC, MBC:SOC ratio is affected by climate, and land use change, soil texture, soil mineralogy, and SOC (Sparling 1992). For example, managed ecosystems such as croplands and pastures tend to have broad MBC:SOC ratio and high MBC, the increasing MBC will enhance the release of carbon dioxide from soil to atmosphere due to the facilitated microbial breakdown of soil organic matter. The acceleration of such processes by soil microbes could significantly exacerbate the soil C cycle; therefore, soil microbial community change is expected to have profound influence on global C cycle.

Limitations and Prospects

A few limitations need to be recognized when interpreting the results. First, we assumed that all samples were taken from surface soil representing 0-30 cm soil profile; while the sampling depth varies between 0 and 30 cm in this study, and 76% of soil samples were taken for topsoil of 0-15 cm. Considering the vertical distribution of microbial biomass C (Xu et al. 2013), this bias might lead to a trivial overestimate to the summarized BBC and FBC in our estimates. Second, the disproportion of the number of data points from each biome to its land area might lead to bias in spatial extrapolation. For example, the data points from forest, grassland, and cropland contribute approximately 80% of the dataset, while the land area of these biomes is approximately 50% of the global land area (Table 2 ). Third, the sampling date might be another reason for uncertainty; the data points were taken from various seasons and we assume the average across season represent the annual mean. In this aspect, future studies on seasonal variation of soil FBC and BBC would bring in improvements to our knowledge. Fourth, actinobacteria was categorized as bacteria in a portion of studies but not in others (Andersen et al. 2010; Royer-Tardif et al. 2010). This difference in classification may introduce minor uncertainties in simulating the relationships between FBC and BBC.