A meta-analysis of the effects of grazing on plant diversity, biomass and 1 soil properties of alpine grassland on the Tibetan Plateau

: 20 Aim As one of the main human disturbance factors in the alpine grassland on the Tibetan Plateau, grazing 21 not only directly affects grassland plant diversity and biomass, but also indirectly changes soil carbon (C) 22 and nitrogen (N) of grassland. Despite of extensive field grazing experiments, the impacts of grazing on 23 grassland diversity, soil C and N remain uncertain due to different grazing management. 24 Methods We conducted a meta-analysis of 70 peer-reviewed publications to evaluate the general 25 response of 11 variables related to alpine grassland plant-soil ecosystems to grazing. 26 Results The results showed that grazing significantly increased species richness, Shannon-Wiener index 27 and Pielou evenness index by 9.8%, 7.3% and 3.7%, respectively. Aboveground biomass, belowground 28 biomass, soil organic carbon, soil total nitrogen, soil C: N ratio and soil moisture decreased by 41.9%, 29 17.7%, 13.1%, 12.6%, 3.3% and 20.8%, respectively. Soil bulk density and soil pH increased by 17.5% 30 and 2.2%, respectively. Specifically, moderate grazing, long-duration (>5 years) and winter grazing 31 contributed to the increase in the species richness, Shannon-Wiener index, and Pielou evenness index. 32 Aboveground biomass, belowground biomass, soil organic carbon, soil total nitrogen and soil C: N ratio 33 showed a decreasing trend with enhanced grazing intensity. Furthermore, grazing duration, grazing season, livestock type and grassland type also affected alpine grassland plant diversity, biomass, soil C 35 and N.


Introduction 43
As an important component of terrestrial ecosystems, grassland accounts for nearly 20% of the 44 global land surface (Scurlock and Hall 1998). Grassland ecosystem not only supports living and grazing 45 conditions, but also plays a key role in biodiversity conservation, carbon sequestration, sand fixation and 46 climate regulation (Ren et al. 2016;Zhang et al. 2015). There is 3.93 million km 2 grassland in China, 47 response ratio (RR++) was used to improve the statistical accuracy, and the weight factor (w) of effect 151 value (RR) of each study was the inverse of the variance (w=1/v). The means of response ratio (RR++) 152 was calculated from each pair of control and grazing treatment of individual RR (Zhou et al. 2017a). The 153 formula for calculating the weighted response ratio is shown in Eq (3). 154 where wij is the weight factor for each group. The m and k are the number of datasets and data points 156 in the category group, respectively. The effect of grazing was considered significant if the 95% 157 confidence interval (CIs) value of RR++ for a variable did not overlap zero with CIs given in Eq (4)  The overall standard error (S) was calculated using the following formula by Eq (5). We applied the random-effects models to calculate the mean effect size for each study, which used 162 bootstrapping method to obtain the lowest and highest values to derive the bootstrap 95% confidence 163 intervals (95% CIs) based on 5000 iterations (Janssens et al. 2010; Zhou et al. 2018). When the 95% CIs 164 of RR++ did not overlap with zero, it indicated that grazing had a significant impact on selected variables. 165 On the contrary, the amount of effect overlapping with zero indicated that there was no obvious difference 166 under various grazing conditions. The percentage change of the variable was calculated with the 167 following formula by Eq (6). To further examine the effects of categorical classes, the total heterogeneity (QT) was composed of 170 within-group heterogeneity (QW) and between-group heterogeneity (QB) (Ren et al. 2018). To clarify whether there was a distinct difference among different treatments within the same group. If the 172 probability value of QB was lower than 0.05, the response rates were significantly different among various 173 subgroups (Li et al. 2016b). We used Rosenthal's fail-safe to check for publication bias (Table S1) 174 (Rosenberg 2005). If the fail-safe number is more than 5n+10 (n is the number of observations used in 175 the analysis), then the result is considered as a reliable estimate of the true effect (Ren et  Overall, our meta-analysis showed that grazing significantly increased all the grassland diversity 184 indices, including species richness (+9.8%), Shannon-Wiener index (+7.3%) and Pielou evenness index 185 (+3.7%) (Fig. 2). Moderate and free grazing significantly increased species richness by 18.8% and 13.1%, 186 respectively, but light and heavy grazing had no effect on species richness (Fig. 2a). The Shannon-187 Wiener index and Pielou evenness index significantly increased except for heavy grazing ( Fig. 2b and  188 2c). For the experimental duration, short and medium grazing duration did not significantly increase 189 species compared to non-grazing, but long duration grazing increased species richness by 13.8%. In 190 contrast, short-duration grazing significantly reduced Shannon-Wiener index (-13.0%) and Pielou 191 evenness index (-9.2%). In terms of the grazing season, winter grazing contributed to increasing species 192 richness (+34.8%), Shannon-Wiener index (+26.6%) and Pielou evenness index (+6.3%). With respect 193 to livestock type, Tibetan sheep grazing significantly increased Shannon-Wiener index (+10.2%) and 194 Pielou evenness index (+8.1%), but did not change species richness. For grassland type, grazing 195 significantly increased Shannon-Wiener index (+18.4%), Pielou evenness index (+11.7%) of alpine 196 steppe, and species richness (+11.0%) of alpine meadow (Fig. 2). 197 [Here insert Fig. 2

] 198
Effects on grassland biomass 199 On average, the results showed that grazing significantly decreased AGB (-41.9%) and BGB (-200 17.7%) (Fig. 3). With respect to different grazing intensity, all grazing intensities had significant negative 201 effects on AGB as the intensity of grazing increased, but only heavy and free grazing decreased BGB. 202 When the short duration grazing had the greatest impact on AGB (-58.2%), BGB did not change. 203 Moreover, medium and long duration grazing had significant negative effects on AGB and BGB. For 204 different grazing seasons, winter grazing had less impact on AGB than summer and annual grazing, while 205 it had no significant effect on BGB. Different livestock type grazing showed different magnitudes of 206 changes for biomass, and mixed grazing had the greatest reduction in AGB compared with yak and 207 Tibetan sheep grazing. Furthermore, grazing had significantly reduced AGB and BGB in different alpine 208 grassland (Fig. 3). 209 [Here insert Fig. 3] 210 Effects on grassland soil C, N, and related variables 211 Overall, grazing significantly decreased SOC (-13.1%), TN (-12.6%), C: N ratio (-3.3%) and SM (-212 20.8%), but enhanced BD (+17.5%) and soil pH (+2.2%) (Fig. 4). Specifically, with the increase of 213 grazing intensity, grazing had a negative impact on SOC, TN, C: N ratio and SM, while it had a positive 214 effect on BD and pH. Regarding grazing duration, long duration grazing had the greatest impact on SOC 215 increased soil BD and pH, while it was decreased SM. With reference to grazing season, all grazing 217 season significantly reduced SOC, TN and SM, but enhanced soil BD. When grouped by livestock type, 218 mixed grazing had the greatest impact on SOC, TN and C: N ratio. Tibetan sheep grazing had the most 219 significant impact on soil BD, SM and pH. Additionally, based on the limited number of observations 220 (<20), grazing had no significant effect on soil C, N and SM in steppe and desert steppe grassland. 221 However, the opposite result was discovered for meadow grassland (Fig. 4). 222 [Here insert Fig. 4] 223

Relationship between response variables and climate factors under grazing 224
We used linear regression to investigate the relationship between the response ratio (RR) and climate. 225 Specifically, our analysis showed that there was no significant correlation between the response ratio of 226 plant diversity (e.g., SR, H, and E) and MAT and MAP (Table 1). No significant relationships were 227 observed between the response ratio of biomass, SOC and TN and climate. Moreover, the response ratio 228 of soil C: N ratio, SM and soil pH were significantly negatively correlated with MAT (p<0.05), but it 229 was not significantly correlated with MAP. The response ratio of BD declined with the increase of MAP 230 (p<0.05), but it was not significantly correlated with MAT. 231 [Here insert Table 1] 232 Taken together, the results illustrated the effect of grazing on plant diversity, biomass and soil C, N 233 and their possible mechanisms in alpine grassland (Fig. 5). In short, grazing directly reduced grassland 234 productivity and SM, enhanced soil BD, and affected soil chemical properties. Moreover, grassland 235 vegetation and soil properties were mutually influential (Fig. 5). Response of grassland plant diversity to grazing 239 Our meta-analysis suggests that grazing had significantly increased species richness, Shannon-240 Wiener index and Pielou evenness index in alpine grassland on the Tibetan Plateau (Fig. 5). These results 241 were in line with a recent meta-analysis by Lu et al. (2017), which indicated that grazing could increase 242 the spatial heterogeneity, enhance the niche of grassland communities, promote the coexistence of species 243 and improve plant species diversity in alpine grassland. With regard to grazing intensity, we found that 244 moderate and free grazing remarkably increased species richness, but the effect was no significant in the 245 light and heavy grazing. This finding was in accordance with the previous study of alpine grassland 246 Notably, long-term (>5 years) grazing increased species diversity compared to non-grazing, but the 255 effects of short and medium duration grazing on species diversity were not significant, which might be 256

Response of grassland biomass to grazing 276
Livestock grazing directly and indirectly affected the normal growth and biomass of grassland. In 277 our meta-analysis, the results consistently demonstrated that grazing significantly decreased AGB and 278 BGB (Fig. 5). The finding was supported by many previous studies that showed alpine grassland biomass Response of soil C, N and related variables to grazing 298 Soil C and N are materials that store energy and limit plant productivity in terrestrial biomass (Song 299 et al. 2017). Overall, our meta-analysis indicated that grazing significantly reduced SOC, TN, and C: N 300 ratio in alpine grasslands (Fig. 5) perspective. All grazing seasons had a significant reduction in SOC and TN, but winter and annual 312 grazing did not change C: N ratio (Fig. 4). This was consistent with Wang et al. Meanwhile, long-duration grazing significantly increased soil pH, probably because grazing had an 330 additive effect on soil trampling. Regarding grazing season, summer grazing most significantly increased 331 soil BD, but had little effects on SM, which is likely due to the abundant precipitation in summer. 332 Furthermore, our results showed that BD was negatively correlated with MAP (Table 1). The effect of 333 Tibetan sheep grazing on soil BD and pH was more than that of mixed grazing (Fig. 5), which was in 334 agreement with a previous study (Xiao et al. 2018). Grazing significantly increased the soil BD and pH 335 of alpine grassland, based on a number of studies in alpine steppe and desert steppe. Grazing significantly 336 reduced soil BD in the alpine desert steppe, but did no change soil BD in the alpine steppe, possibly 337 because different grassland types had different soil properties. 338

Regulating mechanisms of climate factors 339
The Tibet Plateau is the highest geographic in the world with the harshest and most sensitive 340 environment, and climate change will affect grassland vegetation growth (Wu et al. 2014a). In this study, 341 we found that each single plant diversity indices index was not significantly relevant to MAT and MAP 342 under grazing. This result may indicate that grazing disturbance was the main cause of changes in 343 grassland species diversity (Collins and Barber 1985). There was no evidence for variation of responses 344 of plant biomass to grazing as a function of climate (Table 1), which might be attributed to the reduction

Conclusions 374
This study explored the effects of grazing on plant diversity, biomass and soil properties in alpine 375 grasslands on the Tibetan Plateau. Our results revealed that moderate grazing showed the most significant 376 increase species richness, Shannon-Wiener index and Pielou evenness index, indicating that moderate 377 grazing intensity might be an effective management approach for improving the species diversity of 378 alpine grassland on the Tibetan Plateau. With increasing grazing intensity, there was a greater decrease 379 in the loss of biomass, SOC and TN. In addition, long duration (>5 years), winter and mixed grazing can 380 enhance grassland diversity. Given these results, this study indicates that grazing should be chosen 381 according to local environmental conditions, in order to realize the sustainable utilization, biodiversity 382 and environment protection of alpine grassland on the Tibetan Plateau. 383

Acknowledgments 384
We are thankful to all the scientists whose research work was used to perform this meta-analysis.