Diversity of dominant soil bacteria increases with warming velocity at the global scale

Understanding global soil bacterial diversity is important because of the key roles soil bacteria play in the global ecosystem. Given the effects of environmental changes (e.g., climate change and human effect) on the diversity of animals and plants, effects on soil bacterial diversity are expected; however, they have been poorly evaluated to date. Thus, in this study, we focused on the soil dominant bacteria because of their global importance and investigated the effects of warming velocity and human activities on their diversity. Using a global dataset of bacteria, we performed spatial analysis to evaluate the effects, while statistically controlling for the potential confounding effects of current climate and geographic parameters with global climate and geographic data. It was demonstrated that the diversity of the dominant soil bacteria was influenced globally by warming velocity (showing significant increases) in addition to aridity index (dryness) and pH. The effects of warming velocity were particularly significant in forests and grasslands. An effect from human activity was also observed, but it was secondary to warming velocity. These findings provide robust evidence, and advance our understanding of the effects of environmental changes (particularly global warming) on soil bacterial diversity at the global scale.

123 For the OLS regression, we constructed full models encompassing all explanatory 124 variables (AMT, Aridity index, HF, MDR, NPP, pH, PSEA, TS, UV, and WV), and 125 selected the best model to obtain the most simplified model and to simultaneously avoid 126 multicollinearity in the full model. The best model was selected based on the sample-127 size-corrected version of the Akaike information criterion (AICc) values using the R 128 package MuMIn (version 1.43.6). When focusing on the best model only, however, the 129 importance of certain variables may be overestimated; moreover, important variables 130 may be overlooked. To avoid such a model selection bias, we adopted a model-averaging 131 approach using MuMIn. We obtained the averaged model in the top 95% confidence set 132 of models. A global Moran's test was used to evaluate spatial autocorrelation in the 133 regression residuals using the function lm.morantest.exact in the R package spdep 134 (version 0.6.13).
135 As in [14,35,36], the richness and WV were log-transformed. Aridity index, PSAE, and 136 NPP were also log-transformed for normality. The variables were normalized to the same 137 scale, with a mean of 0 and standard deviation of 1, using the scale function in R before 138 the analysis. 139 We also considered a spatial eigenvector mapping (SEVM) modeling approach [22,37] to 140 remove spatial autocorrelation in the regression residuals. Specifically, we adopted the 141 Moran eigenvector approach using the function SpatialFiltering in the R package 142 spatialreg (version 1.1.5). As with the OLS regression analysis, we constructed full 143 models, and then selected the best model based on AICc values. The spatial filter was 144 fixed in the model-selection procedures [37]. We also obtained the averaged models.
145 The contribution (i.e., non-zero estimate) of each explanatory variable (i.e., 146 environmental parameters) to bacterial diversity was considered significant when the 147 associated p-value was less than 0.05. 148 We obtained the residuals of the explanatory variables and bacterial diversity according 149 to the SEVM modeling approach-based best models. 156 AMT, MDR, and TS indicate annual mean temperature, mean diurnal temperature range, 157 and temperature seasonality, respectively; PSEA represents precipitation seasonality; 158 NPP represents normalized difference vegetation index; UV represents UV radiation 159 index; HF and WV represent the human footprint score and warming velocity, 160 respectively. The estimates in the full, best, and averaged models based on the ordinary 161 least squared (OLS) regression and spatial eigenvector mapping (SEVM) modeling 162 approach are shown. R 2 denotes the coefficient of determination for the full and best 163 models based on the OLS regression and SEVM modeling. Values in brackets are the 164 associated p-values.
165 The Shannon index was associated with several environmental changes (Table 1). 166 Specifically, the Shannon index showed a positive correlation with warming velocity (Fig  167 1A) and HF score (human effects). The aridity index (i.e., dryness; Fig 1B) and pH (Fig  168 1C) were negatively and positively associated with the Shannon index, respectively. The 169 aridity index, pH, and warming velocity were also associated with the other diversity 170 indices (i.e., richness, Simpson index, and evenness); however, the contribution of human 171 effects to these diversity indices was not statistically significant concluded (Tables A-C 172 in S1 File). UV radiation and plant productivity were negatively and positively associated 173 with evenness, respectively (Table C in S1 File) but showed no correlation with the other 174 diversity indices (Table 1 and Tables A-C in S1 File). Annual mean temperature, mean 261 soil bacterial diversity. It has been shown that soil bacterial diversity is positively 262 associated with human population [52]. However, the direct effects of human effects are 263 currently ambiguous because they depend on the types of diversity indices and ecosystem 264 types investigated (see section 3). The effects of the other environmental factors (UV 265 radiation, plant productivity, temperature seasonality, and precipitation seasonality) at the 266 global scale may also still be uncertain because they were influenced by the types of 267 diversity indices and ecosystem types included in the analyses.
268 Dryness (based on the aridity index results) and pH appear to affect soil bacterial 269 diversity. Our findings that aridity index and pH had negative and positive correlations 270 with bacterial diversity, respectively, are consistent with a number of previous studies 271 [1,3,15,24], despite the limitations in terms of data and statistical analysis in these works 272 (see section 1). This suggests that the evidence for these effects are robust. 298 In conclusion, despite the limitations in our data analyses, our findings enhances our 299 understanding of the effects of environmental changes (particularly global warming) on 300 soil bacterial diversity.