We collected published papers using the keywords “grazing,” “soil available nutrient,” and “Tibet*” in the Web of Science from January, 1990 to June, 2024. Furthermore, 165 articles were selected and conserved to the endnote library (Endnote X9). Full-version papers were automatically added by its function of “find full text” and Google Scholar.

Full articles were subsequently screened based on the following criteria: (1) all studies were conducted, including controls and grazing activity. Grazing intensity is categorized into light, moderate, and heavy grazing according to the classification standards set forth in the paper. (2) Soil AN, AP, and AK were analyzed using the alkali hydrolysis nitrogen method, molybdenum antimony colorimetry, and ammonium acetate extraction flame photometric method, respectively. Some data were extracted from the table, whereas the others were sampled from published figures using WebPlotDigitizer. Herin, we collated the available nitrogen (26 groups), available phosphorus (35 groups), available potassium (13 groups, Figure 1 ), and 74 sets of field experimental results ( Figure 2 ). Additionally, soil and plant characteristics and climatic factors were collected.

Log response ratios (RR, hereafter response ratios) were calculated as the measures of the effect size. A 95% confidence interval (CI) was calculated. When the 95% confidence interval crosses zero, the result is considered not statistically significant. Meta-analysis using a random-effects model was performed, and the data were analyzed using R.

Xe and Xc are the mean values of each trait measured in the grazed and control grassland respectively. An ln R < 0 reveals a decrease in the trait response to grazing activity; otherwise, it indicates an increasing effect. The variance in ln R is calculated as follows:

Se, Ne, Xe, Sc, Nc, and Xc represent the standard deviations, sample sizes, and mean values in the grazing treatments and control, respectively.

The effect size of grazing on soil available nutrients and confidence interval based on the random effects model were calculated using the following equation:

Individual study weight w i * = 1 / ( v i + τ 2 )

Both v_i and τ² represent the intra- and inter-study variance, respectively. The total variance of a single case (vi*) = within-case variance (vi) + between-case variance (τ²).

Average effect size y ¯ = ∑ i = 1 k w i * y i ∑ i = 1 k w i *

y_i refers to the single study effect value

Standard error S E = 1 ∑ i = 1 k w i *

95% confidence interval of average effect value: CI = y ¯ ± 1.96 SE

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

Meta-statistical analyses were performed using R 3.6.2, and a random-effects model of the meta-analysis was run in metafor1.9-8 (22). The estimated values and standard errors (rma) were analyzed using the random effects models. A statistically significant difference is indicated when the 95% confidence interval does not include zero. In meta-regressions, total heterogeneity can be partitioned into variance explained by moderators (Qm) and residual error variance (Qe). The Qm statistic, a Q-test, assesses whether moderators account for significant heterogeneity within the data. Specifically, Qm functions as a Wald-type test for model coefficients, where a significant Qm value indicates that moderators contribute meaningfully to variations in effect sizes. A model selection approach was applied using the glmulti package, which evaluates multiple models based on Akaike Information Criterion corrected (AICc) values to identify the best-fitting model. The process involved ranking models by their AICc scores, visualizing these rankings, and examining the weights and summary of the model with the lowest AICc, as well as any additional models with similar AICc values (within a difference of less than 2).

Grazing significantly increased the AN content in the alpine grassland soils, with an effect size of 0.112 (P < 0.05, increase of 11.9%, Figure 3 ). Conversely, the effect of grazing on AP was not statistically significant, with an effect size of 0.023, corresponding to an approximate increase of 2.4%. Grazing significantly reduced the AK content, with an effect size of -0.157 (P < 0.05, decrease of 14.5%).

This meta-analysis systematically evaluated the effects of varying grazing intensities on the availability of nitrogen, phosphorus, and potassium in alpine grasslands. The results indicated that both light grazing (LG) and moderate grazing (MG) significantly enhanced soil available nitrogen (AN), with increases of 6.88% and 17.90%, respectively (P < 0.01, Figure 4 ). Conversely, heavy grazing (HG) exhibited a detrimental effect on AN, with a decrease of 1.52%. Similarly, both LG and MG were found to increase soil available phosphorus (AP), while HG resulted in a substantial decrease in soil AP, approximately 11.04%. Furthermore, LG contributed to a 2.05% increase in available potassium (AK), whereas both MG and HG led to significant reductions in soil AK, with decreases of 11.01% and 10.35%, respectively.

Grazing impacts soil nutrient dynamics in alpine grasslands differently depending on the nutrient and environmental factors. Temperature significantly influences reducing the effect of grazing on available nitrogen (P = 0.01, Figure 5 ) and increasing the effect size of available potassium (P < 0.0001, Figure 6 ), making it a key driver under climate change scenarios. In contrast, total nitrogen, total phosphorus, total potassium, precipitation, and soil organic carbon showed no significant effects on nutrient dynamics, highlighting their limited role in regulating available nitrogen, phosphorus, and potassium under grazing conditions ( Figure 7 ). These results emphasize the critical role of temperature in nutrient cycling, especially in the context of climate warming.

The mixed-effects model shows that soil organic carbon and temperature significantly affect the effect size of grazing on AN. The model explains 39.57% of the variance in effect size, and the high I² value of 92.54% indicates substantial heterogeneity ( Table 1 ), reflecting considerable variability among studies. Precipitation explained only 6.51% of AP variance. The high I² value of 96.22% again indicates substantial heterogeneity. temperature significantly affects the effect size of AK, explaining 67.28% of the variance, with a lower but still high I² value of 83.35%.

This study demonstrated that grazing activity significantly increased available nitrogen (AN) in alpine grassland soils by approximately 11.9%, while the effect on available phosphorus (AP) was not statistically significant, resulting in a minor increase of 2.4%. Conversely, grazing led to a marked reduction in available potassium (AK) by 14.5% across the Tibetan Plateau. This study found that light grazing improves soil nutrient levels, while moderate grazing has mixed effects. In contrast, heavy grazing generally reduces key soil nutrients. Therefore, light grazing should be recommended as the main grazing strategy.

These findings highlight the differential impacts of grazing on soil nutrient availability, with significant implications for nutrient cycling and ecosystem management in fragile alpine environments. Similarly, soil nitrogen content was the highest under light grazing on the Eastern Tibetan Plateau (16). Grazing substantially increased grassland AN from 117.39 to 129.35 mg/kg and AP from 3.75 to 4.50 mg/kg (23). Grazing enhanced alpine meadow AN from 23.53 to 23.83 mg/kg on the Northeast of the Tibetan Plateau (24). The alpine meadow soil available nitrogen content was 26.07, 26.62, and 24.61 mg/kg in light grazing, moderate grazing, and heavy grazing, respectively (25). Similarly, a previous study indicated that heavy grazing decreased soil nitrogen in the Eastern Tibetan Plateau (26). This was because, under light and moderate grazing conditions, the nitrogen mineralization process in the soil was enhanced by stimulated microbial activity, which accelerated the decomposition of soil organic matter and increased the release of available nitrogen (25). However, heavy grazing inhibited microbial activity by reducing the input of plant and root residues, leading to a decrease in nitrogen mineralization rates (27). Furthermore, grazing decreased AP and exchanged potassium by 33.89% and 41.02%, respectively, in the central Ethiopian highlands (28). Additionally, grazing reduced the soil AP of grasslands from 6.96 to 5.76 mg/kg on the Tibetan Plateau (12). The increase in soil potassium under light grazing is particularly significant, mainly due to the greater amount of potassium left in plant residues, but in long-term high grazing treatments, the availability of potassium tends to decrease, primarily due to the reduced input of biomass potassium (29). Thus, our study offers a more comprehensive understanding than previous research, providing a clearer basis for the sustainable management of alpine grasslands on the Tibetan Plateau.

Grazing is a traditional grassland management practice that significantly alters nutrient cycling processes within the ecosystem (23, 30). Concentrations of plant nutrients are mainly derived from soil nutrients (30, 31). Soil available nutrients are derived from microbial mineralization and decomposition and are driven by multiple factors, such as the environment, climate, and physiochemical properties.

Furthermore, soil nutrients appeared to play crucial roles in root decomposition, which increased at lower altitudes in the Bromus inermis grasslands of Northwest China (32). In this study, the mixed-effects model highlights temperature as a critical determinant of grazing impacts on soil nutrient dynamics, particularly by reducing available nitrogen and increasing available potassium, underscoring its pivotal role under ongoing climate change scenarios. These findings highlight the importance of incorporating temperature variability into grazing management strategies to predict and mitigate grazing impacts on nutrient cycling in alpine grasslands, contributing to greater ecosystem resilience under climate change. Temperature was identified as a key driver, reducing available nitrogen (AN) and increasing available potassium (AK) by accelerating organic matter decomposition and nutrient mineralization (6). Given its significant influence, addressing temperature variability is crucial for effective nutrient management and the sustainability of fragile alpine ecosystems amid climate change (14).

The significant heterogeneity observed in the mixed-effects model, indicated by high I² values (e.g., 92.54% for AN and 83.35% for AK), reflects the complexity of the underlying research. This heterogeneity likely arises from variations in geographical regions, soil types, grazing intensities, and diverse environmental contexts across the studies, leading to substantial variability in the outcomes (2). Such variability underscores the complex interactions between grazing and soil nutrient dynamics, influenced by multiple interrelated factors that differ depending on the specific study environment (19).