Published papers were collected using keywords “grazing” and “nitrous oxide or N₂O” and “Tibet*” in Web of Science from January 1990 to April 2022. Furthermore, 49 articles were selected and conserved in the Endnote library (Endnote X9). Most full-version papers and doi numbers were automatically added by the function “find full text.” Others were captured through doi by using Google Scholar.

Then, complete articles were screened based on the following criteria: 1) all studies were conducted, including control and its grazing activity (light, moderate and high grazing). 2) N₂O concentrations were analyzed using the static chamber method and chromatographic concentration analysis. Some data were extracted from published paper figures by using WebPlotDigitizer software. Here, we collated 21 published studies, including 40 field experiment results (Figure 1). Soil and plant characteristics were also recorded.

We calculated the log response ratios (RR, hereafter response ratios) as a measure of effect size. The 95% confidence intervals (CIs) were calculated. A random-effect-model meta-analysis was performed, and the data were analyzed with R statistical software by using the Meta package. ln ⁡ R = L n x e x c = ln ( x e ) −   ln ( x c ) where X_e and X_c are the mean values of each individual trait measure in the treatment and control groups, respectively. In addition, lnR <0 indicated a decrease in trait response to grazing activity; otherwise, it indicated an increasing effect. The variance in lnR was calculated as follows: V ln R =   S e 2 N e x e 2   +   S c 2 N c x c 2   where S_e, N_e, x_e, S_c, N_c, and x_c are standard deviations, sample sizes, and mean values of the grazing treatments and control, respectively.

The effect size of grazing on N₂O emission rates, soil and plant characteristics, and CI were calculated based on the random-effect model: Weight   of   an   individual   study   w i ∗ = 1 / ( v i + τ 2 ) where v _i represents the intra-study variance, and τ 2 represents the inter-study variance. Average   effect   size   y ¯ = ∑ i = 1 k w i ∗ y i ∑ i = 1 k w i ∗ Standard   error   S E = 1 ∑ i = 1 k w i ∗ 95% CI of average effect value: CI = y ¯ ± 1.96 SE, and y _i refers to the single study effect value.

Meta-statistical analyses were performed using R 3.6.2, and a random-effect model of the meta-analysis was run in metafor1.9-8 (Benítez-López et al., 2017). The random-effect models were used to analyze the estimated values and standard errors (rma). Then, mixed-effect models (mods) were used to explain significant residual heterogeneity with different moderators, including categorical and continuous variables. The explained moderator heterogeneity statistic (Qm) was also calculated to test for significance in single covariate meta-regressions.

Drive factors of climate factors and soil characteristics for effect sizes were analyzed with the structural equation model using “piecewiseSEM” package. All figures were drawn using SigmaPlot 10.0 and R statistics.

Biases against publishing negative results may exist in research fields. In this study, a regression test for funnel plot asymmetry of publication bias was performed using a mixed-effect meta-regression model (funnel and Egger’s test, rma).

Grazing activities had a robust and positive effect on N₂O emission rates (p < 0.0001, Table 1), with an estimated value of 0.31 ± 0.08 effect size (95% CI: 0.15–0.47). This indicated that grazing disturbance increased alpine meadow N₂O emission rates by 36.27% on the Tibetan Plateau. Furthermore, high grazing activity significantly increased N₂O emission rates by approximately 62.16% (p < 0.0001), and it was significantly higher than that of moderate grazing (p < 0.05, Table 1). Light and moderate grazing increased by 34.62 and 19.48%, respectively. Thus, moderate grazing is optimal for mitigating N₂O emissions on the Tibetan Plateau.

Grazing activity increased grassland soil ammonia and nitrate levels, pH, temperature, dissolved organic carbon, and moisture by 12.24, 8.60, 6.51, 5.33, 3.56, and 1.16%, respectively (Figure 2). The effect size of grazing on pH was significant (p < 0.05). However, grazing decreased soil total carbon, total phosphorus, available phosphorus, soil organic carbon, available potassium, total nitrogen, and bulk soil by 14.4, 10.25, 10.15, 6.57, 5.85, 4.3, and 1.97%, respectively. In addition, the effect size of available potassium was significant (p < 0.01, Figure 2). Grazing significantly decreased plant diversity (p < 0.05), richness, and aboveground biomass (p < 0.01) by 15.16, 23.7, and 30.7%, respectively (Figure 2).

Aboveground biomass significantly influenced the effect size based on test for moderators (p < 0.05, Table 2). The mixed-effect model results indicated that aboveground biomass could explain 13.36% of the variations in effect size, and the effect size decreased as the aboveground biomass increased. Furthermore, nitrate content and precipitation were the main factors that affected the effect sizes by approximately 4.49 and 3.18%, respectively (Table 2).

The effect size of grazing activity on N₂O emission rates was mainly controlled by grassland aboveground biomass with a structural equation model. The direct coefficient of aboveground biomass on effect size was −0.631 (p < 0.05, Figure 3). The direct coefficients of temperature and precipitation on the effect sizes were 0.579 and −0.456, respectively. The indirect coefficients of precipitation, air temperature, and nitrate were −0.365, −0.311, and 0.153, respectively.

Egger’s regression test for funnel plot asymmetry indicated that this result was not significantly affected by publication bias by using meta-analysis models (z = 0.0862, p = 0.9313, Figure 4).

The average global temperature in 2021 was approximately 1.11°C above pre-industrial levels, and atmospheric N₂O was 333.2 ppb, approximately 123% of pre-industrial levels (WMO, 2021). Grassland N₂O emission rates increased by 5.05% because of nitrogen excreted by grazing animals in Europe (Oenema et al., 1997). In South Brazil, grazing increased grassland N₂O emission rates from 5.54 to 15.83 μg m⁻² h⁻¹ (Schirmann et al., 2019). Across the Tibetan Plateau, grazing treatments increased N₂O emissions from 5.42 to 266.71% (Lin et al., 2009; Ma S. et al., 2021). Light grazing increased N₂O emission rates by 5.40–63.91% (Guo et al., 2019; Luo et al., 2020), and extreme grazing decreased N₂O emission rates by 28.57% on the central Tibetan Plateau (Wei et al., 2012). The opposite effect and strong heterogeneity across different studies were because of different aboveground biomass, soil organic carbon content, air temperature and precipitation (Yao et al., 2019; Feyissa et al., 2021). Grazing increased soil N₂O emissions by regulating nirK and nosZ denitrifiers in alpine meadows (Zhang et al., 2021).

In the present study, we found that grazing increased alpine meadow emission rates by 36.27% on the Tibetan Plateau. As the grazing intensity increased, the influence on extent of emission first decreased from 34.62% (light) to 19.48% (moderate) and then increased by approximately 62.16% (high grazing). The effect was the lowest for moderate grazing, which was revealed to significantly increase grassland biomass and diversity on the northeast and southeast Tibetan Plateau (Zhu et al., 2015; Hu et al., 2017; Ma L. et al., 2021).

There are three reasons for this increased effect of grazing on N₂O emission rates on the Tibetan Plateau. First, grazing activities increased available nitrate and pH in grassland soils, providing adequate substrate and suitable environments for N₂O production. In this study, we also found that grazing increased soil ammonia and nitrate by 12.24 and 8.60%, respectively. Second, grazing livestock provide a lot of dung and urine to grassland soils. Ruminants may excrete 75–95% of the consumed nitrogen to grasslands (Feyissa et al., 2021). The feeding and trampling behaviors of grazing animals and their excreta can directly or indirectly increase soil emissions of N₂O (Ma S. et al., 2021). Cumulative N₂O emissions for both urine and dung patches were 1.8–3.7 times greater than those of control plots in alpine meadows (Lin et al., 2009). Third, grazing significantly decreased grassland biomass and richness by 30.70 and 23.70%, respectively, on the Tibetan Plateau. This reduction in aboveground biomass and richness increased the competition priority of microorganisms. Subsequently, N₂O emission rates significantly increased on the Tibetan Plateau.

Limiting warming to around 1.5°C in the Paris Agreement requires global greenhouse gas emissions to be reduced by 43% by 2030 (WMO, 2021), and N₂O should be reduced by 20% (IPCC, 2022).

Grazing activity has significant effects on plant communities and soil properties (Feyissa et al., 2021). Heavy grazing intensity accelerated N cycling rates and increased plant-soil system N in the alpine meadow (Zhong et al., 2017). In this study, we revealed that the effect size of grazing on N₂O fluxes was driven by aboveground biomass, with a direct coefficient of −0.631 based on the structural equation model. Furthermore, aboveground biomass could explain 13.36% of the variations in effect size by using the mixed-effect model. Thus, an increase in grassland aboveground biomass would help to significantly mitigate N₂O emission rates on the Tibetan Plateau.

Different vegetation types affected grassland N₂O emission rates; emission rates in Kobresia humilis and Potentilla fruticosa meadows were 47.8 and 60.6 μg m⁻² h⁻¹, respectively (Du et al., 2008). N₂O is mainly driven by the simultaneous effects of grassland biomass and soil temperature on the Tibetan Plateau (Lin et al., 2009). N₂O fluxes were significantly positively correlated to soil nitrate content, organic carbon, and biomass (Yao et al., 2019). Soil temperature and biomass are the major contributors to alpine meadow N₂O fluxes on the northwest Tibetan Plateau (Yin et al., 2020). Grassland denitrification genes increased when biomass was higher, which was concomitant with increased N₂O emissions (Wang et al., 2016; Tang L. et al., 2019).