1 Introduction

Nitrogen (N) is one of the essential and limiting nutrients for plant growth and development in terrestrial ecosystems (Elser et al. 2007; Yuan and Chen 2012; Moreau et al. 2019). Thus, moderate N enrichment may have a positive impact on plant growth and terrestrial ecosystem productivity (LeBauer and Treseder 2008; Bobbink et al. 2010; Tian et al. 2020). However, extensive atmospheric N deposition, owing to human activities (e.g., fertilization and urbanization), has occurred in current terrestrial ecosystems and continues to increase on a global scale (Lamarque et al. 2005; Elser and Bennett 2011; Dietrich et al. 2017). These actions stimulate plant growth and affect plant nutrient cycling (Kallenbach et al. 2017; Bellenger et al. 2020; Iqbal et al. 2020; Tognetti et al. 2021). Therefore, it is crucial to fully understand the changes in terrestrial ecosystem responses to future N deposition (Zaehle and Dalmonech 2011; Greaver et al. 2016; Soong et al. 2020).

Plant nutrient concentration reflects plant growth, nutrient uptake, and use strategies in terrestrial ecosystems, which are collectively affected by N enrichment (Yuan and Chen 2009; Lihavainen et al. 2016). However, this issue is still controversial, although most previous studies have focused on the effect of N enrichment on changes in plant nutrient concentration. For example, field experiments reported varied effects of N enrichment on plant P levels, either increasing (Lu et al. 2013; Carate-Tandalla et al. 2018; Gonzales and Yanai 2019), decreasing (Sardans et al. 2016; Tiruvaimozhi et al. 2018; Liu et al. 2021a), or not affecting (Zong et al. 2018; Tian et al. 2020) P concentration. Nitrogen addition also had positive, negative, and neutral effects on the plant leaf N and P concentrations in the same ecosystem (Lu et al. 2013; Su et al. 2021). These differences may result from the comprehensive effects of multiple factors such as species (Van Heerwaarden et al. 2003), soil nutrient conditions (Gusewell 2005), N addition rates, experimental duration (Hao et al. 2018), and environmental factors. Our study complements the limited knowledge about the complex responses of plant nutrient concentration and nutrient resorption to N addition in varying grassland types and among plant functional groups.

Plants reduce their dependence on external nutrients and enhance their survival and adaptability by resorbing nutrients from senesced tissues (Yuan et al. 2006; Gerdol et al. 2019). This process can be quantified by computing the nutrient resorption efficiency (Milla et al. 2005). Previous studies reported that N addition reduced the nitrogen resorption efficiency (NRE) owing to the increase in soil N availability (Soudzilovskaia et al. 2007; Ren et al. 2015; Li et al. 2016). However, the effect of N addition on phosphorus resorption efficiency (PRE) has not been determined in grassland ecosystems (Lu et al. 2013; Yuan and Chen 2015a, b; Zheng et al. 2018). Nitrogen addition to an ecosystem was thought to transform N limitation to P limitation or co-limitation of N and P (Dong et al. 2019). It is unclear how N addition affects nutrient resorption efficiency in grassland ecosystems, which are usually limited by P (Van Dobben et al. 2017). Therefore, knowledge of nutrient resorption responses to N enrichment has great significance for managing plant nutrient conservation and plant productivity under future global environmental changes.

Grasslands account for approximately 26% of the global terrestrial area (Obermeier et al. 2016). Extensive atmospheric N deposition inevitably affected plant nutrient concentrations and nutrient uptake strategies in grassland ecosystems (Sattari et al. 2016; Shi et al. 2021; Hu et al. 2022). Many studies reported that different N addition rates produced varying responses of leaf nutrient concentrations and nutrient resorption (Cerasoli et al. 2018; Bai et al. 2019; Graff et al. 2020). However, the patterns and mechanisms of leaf nutrient concentrations and nutrient resorption responses to N addition in grassland ecosystems have not been defined, particularly the complex plant responses to changes in multiple biotic and abiotic factors. Therefore, we conducted a meta-analysis to integrate the available data from different field experiments in global grassland ecosystems and define the responses of grassland leaf nutrients and nutrient resorption to N addition.

Here, by using a weighted meta-analysis of a global data set of 1935 observations at 98 sites from 127 publications based on N addition, we examined how leaf nutrient concentration and nutrient resorption respond to N addition in grassland ecosystems. We tested three hypotheses. (1) Nitrogen addition enhances the N concentration in green and senesced leaves due to an N-induced increase in soil N availability (Soudzilovskaia et al. 2007; Li et al. 2016). The corresponding P concentration in green and senesced leaves will also increase to maintain the balance of N and P (You et al. 2018b). (2) Plants will reduce N resorption from senesced tissues due to increased soil N availability (Ren et al. 2018), thereby reducing NRE but not PRE (to balance N and P). (3) Numerous experimental and environmental factors alter leaf nutrient and resorption responses to N addition by modulating plant nutrient cycle patterns.

2 Material and Methods

2.1 Data Collection

We collected peer-reviewed publications that reported changes in plant nutrient concentrations and resorption efficiency in global grassland ecosystems under the addition of N to fields. We performed Boolean searches of the Web of Science (http://apps.webofknowledge.com) and the China National Knowledge Infrastructure (CNKI, https://www.cnki.net) databases using the following keywords: (a) ‘N addition’ or ‘nitrogen addition’ or ‘nitrogen amendment’ or ‘nitrogen deposition’, (b) ‘leaf N’ or ‘leaf nutrient’, (c) ‘leaf P’ or ‘leaf nutrient’, (d) ‘NRE’ or ‘nutrient resorption’, (e) ‘PRE’ or ‘nutrient resorption’, (f) ‘AGB’ or ‘aboveground biomass’, (g) ‘BGB’ or ‘belowground biomass’, and (h) ‘grassland’ or ‘grassland ecosystem’. Seven criteria were used to select suitable publications: (1) Experiments were conducted on global grassland ecosystems; (2) experiments were conducted in the field; (3) studies reported comparisons between controls (i.e., without nitrogen addition) and treatments (i.e., nitrogen addition); (4) studies reported means, standard deviations (SD) or standard errors (SE), and sample sizes of the selected variables; (5) studies reported the treatment method, magnitude, and duration; (6) the publications were peer-reviewed journal articles, conference collections, theses, or dissertations; and (7) where data were published in different papers for studies at the same site, we only reserved one recent publication to ensure the relative independence of data. These search and selection criteria yielded 1935 experimental observations from 127 papers at 98 sites across the globe (Fig. 1; Supplementary information, data sources list).

Fig. 1
figure 1

Map of sites conducting field studies of nitrogen addition in grasslands that were included in the meta-analysis. Points with different colors represent different N application rates (g N m−2 year−1) in each study site

Full size image

We performed a stratified analysis to evaluate whether grassland type, plant group, fertilizer type, and experimental duration affected plant nutrient concentration and resorption efficiency responses to N addition. The data classifications included the following categories: Grassland types were categorized into temperate grassland and alpine grassland; plant functional group was categorized into grass and forb; experimental duration was categorized into short-, intermediate- and long-term (i.e., < 3 years, 3–6 years, and > 6 years, respectively), and fertilizer type was categorized into urea and NH4NO3. We extracted site information related to the experiments to construct an integrated database of geographic variables (longitude, latitude, and altitude) and climatic factors [mean annual temperature (MAT) and mean annual precipitation (MAP)]. Data presented in the figures were extracted using GetData version 2.20 (http://www.getdata-graph-digitizer.com). We investigated how the climatic factors affected the plant nutrient concentration and resorption efficiency responses to N addition using the De Martonne aridity index to compute the humidity combined with MAT and MAP data at each experimental site [aridity index = MAP/(MAT + 10)] as described in previous studies (Song et al. 2019; Su et al. 2021).

2.2 Data Analysis

We conducted a meta-analysis to determine the effects of N addition on plant nutrient concentrations and resorption efficiency in global grasslands. The natural logarithm of the response ratio (lnRR) was calculated to indicate the effect size of each treatment (Hedges et al. 1999) [Eq. (1)]:

$$\mathrm{lnRR}=\ln\left(\frac{{\overline{\mathrm X}}_{\mathrm T}}{{\overline{\mathrm X}}_{\mathrm C}}\right)=\ln\;{\overline{\mathrm X}}_{\mathrm T}-\ln\;{\overline{\mathrm X}}_{\mathrm C}$$
(1)

where \({\overline{X} }_{T}\) and \({\overline{X} }_{C}\) are the mean treatment and control values, respectively. The variance (v) of each lnRR was calculated using Eq. (2):

$$\mathrm v=\frac{\mathrm S_{\mathrm T}^2}{{\mathrm n}_{\mathrm T}\overline{\mathrm X}_{\mathrm T}^2}+\frac{\mathrm S_{\mathrm C}^2}{{\mathrm n}_{\mathrm C}\overline{\mathrm X}_{\mathrm C}^2}$$
(2)

where \({n}_{T}\) and \({n}_{C}\) are the sample sizes, and \({S}_{T}\) and \({S}_{C}\) are the SD of means for each treatment and control. Most of the selected studies reported SE, which was transformed to SD according to Eq. (3):

$$\mathrm{SD}=\mathrm{SE}\times\sqrt{\mathrm n}$$
(3)

where \(n\) was the sample size. Based on previous studies, the lnRR was calculated separately for each control-treatment pair and treated as independent data when data were extracted from multifactor experiments with multiple single-factor treatments and a single control (Lajeunesse 2011; Song et al. 2019).

The weighted response ratio (lnRR++) and bias-corrected 95% bootstrap-confidence interval (CI) were calculated using inverse-variance weighted regressions and random-effects models with the rma function in the “metafor” package version 3.0–2 of R version 4.1.2 (The R Project for Statistical Computing, https://www.r-project.org/) (Hedges et al. 1999). The effects of treatments on selected variables were considered statistically significant if the 95% CI did not overlap zero, whereas the effects between groups or under different conditions differed if their 95% CIs did not overlap. To clarify data interpretation, the percentage change (%) was calculated based on weighted response ratios using the equation [exp (lnRR++) – 1] × 100 (Yan et al. 2020). We computed multiple comparisons to examine differences in treatment effects on different groups or under different conditions. Statistical results were reported as differences among group cumulative effect size (QB) and residual error (QE). Regression analysis (including univariate covariance analysis) was conducted to examine the effects of N addition rate and environmental factors (e.g., aridity index) on the response ratio of plant nutrient concentration and resorption efficiency under N addition and evaluate the relationships of response ratios of objective variables (e.g., green and senesced leaf N and P). Statistical differences were considered significant when P < 0.05. All statistical analyses were performed in R version 4.1.2.

3 Results

3.1 Effects of N Addition on Leaf Nutrient Concentration and Resorption

Nitrogen addition significantly increased the concentrations of leaf N, green leaf N, senesced leaf N, senesced leaf P, AGB, and BGB by 29%, 32%, 50%, 7%, 74%, and 19%, respectively (P < 0.001, Fig. 2). By contrast, N addition reduced NRE and PRE by 9% and 5%, respectively (P < 0.001, Fig. 2). Nitrogen addition did not significantly affect the concentrations of total leaf P and green leaf P (P > 0.05, Fig. 2). Regression analysis identified significant and positive relationships between the concentrations of green leaf N and P, senesced leaf N and P, and NRE and PRE. Regression analysis also identified significant differences in the relationships between NRE and PRE, and green leaf N and P in the control and N enrichment plots, but the relationship between N and P in senesced leaves in the control and N enrichment plots did not significantly differ (Fig. 3a–d, Supplementary Table S1).

Fig. 2
figure 2

Responses of leaf nutrient concentrations, nutrient resorption, and biomass to experimental nitrogen addition in global grassland ecosystems. N, leaf nitrogen concentration; P, leaf phosphorus concentration; Ng, green leaf nitrogen concentration; Pg, green leaf phosphorus concentration; Ns, senesced leaf nitrogen concentration; Ps, senesced leaf phosphorus concentration; AGB, aboveground biomass; BGB, belowground biomass; NRE, nitrogen resorption efficiency; PRE, phosphorus resorption efficiency. Error bars represent 95% confidence intervals (CI). The vertical dashed line represents the response ratio = 0. Treatment effects were statistically significant (denoted by *) if 95% CI did not overlap zero. *, **, and *** indicate significant correlations at p < 0.05, p < 0.01, and p < 0.001, respectively. The sample size for each variable is given in parentheses

Full size image
Fig. 3
figure 3

Relationships between (a) leaf nitrogen (N) and phosphorus (P), (b) green leaf N and P, (c) senesced leaf N and P, and (d) nitrogen resorption efficiency (NRE) and phosphorus resorption efficiency (PRE) under control (N0) and N addition (N) conditions

Full size image

3.2 Effects of N Enrichment on Leaf Nutrient Concentration and Resorption Among Varying Subgroups

Subgroup analysis indicated that the responses of leaf nutrient concentration and nutrient resorption to N addition differed in subgroups (Fig. 4, Supplementary Table S2). Nitrogen addition increased the green and senesced leaf N concentration in temperate and alpine grasslands (P < 0.001, Fig. 4a,b), but the responses were more sensitive in alpine than in temperate grassland. The N addition impacts on leaf P concentration in different grassland types varied among green and senesced leaves (Fig. 4d,e). Nitrogen addition decreased NRE and PRE in temperate grassland, but did not significantly affect NRE and PRE in alpine grassland (Fig. 4c,f, Supplementary Table S2).

Fig. 4
figure 4

Subgroup analysis of the response of leaf nutrient concentrations and nutrient resorption to experimental nitrogen addition in global grassland ecosystems. Ng, green leaf nitrogen concentration; Ns, senesced leaf nitrogen concentration; Pg, green leaf phosphorus concentration; Ps, senesced leaf phosphorus concentration; NRE, nitrogen resorption efficiency; PRE, phosphorus resorption efficiency. Error bars represent 95% confidence intervals (CI). The vertical dashed line represents the response ratio = 0. Treatment effects were statistically significant (denoted by *) if 95% CI did not overlap zero. *, **, and *** indicate significant correlations at p < 0.05, p < 0.01, and p < 0.001, respectively. The number represents the sample size of each variable

Full size image

Nitrogen addition enhanced green and senesced leaf N concentration in both grass and forb (P < 0.001, Fig. 4a,b), but grass had a higher response ratio than forb. The effect of N addition on green leaf P concentration was opposite in grass and forb (i.e., decreased in grass and increased in forb; P < 0.05, Fig. 4d), while N addition impact on senesced leaf P concentration was not significantly different between grass and forb (P > 0.05, Fig. 4e). Nitrogen addition significantly reduced NRE and PRE in both grass and forb, while forb had greater responsiveness than grass (P < 0.001, Fig. 4c,f). Ammonium nitrate and urea increased the green and senesced leaf N concentration, but the NH4NO3 effect on plant N concentration was stronger than that of urea (Fig. 4a,b, Supplementary Table S2). Ammonium nitrate increased the senesced leaf P concentration (P < 0.001), whereas urea did not (P > 0.05, Fig. 4e). The effect of NH4NO3 and urea on green leaf P concentration was not significantly different (P > 0.05, Fig. 4d). Ammonium nitrate and urea both reduced NRE (P < 0.001, Fig. 4c), but the effect of urea on NRE was greater than that of NH4NO3. Phosphorus resorption efficiency was only reduced by NH4NO3 (P < 0.001, Fig. 4f). Long-term experiments yielded larger increases in senesced leaf N concentration than short- and intermediate-term experiments. Long-term experiments increased senesced leaf P and reduced green leaf P, whereas short-term and intermediate-term experiments did not affect green leaf P. Larger increases in green leaf N and decreases in NRE were observed in intermediate-term and short-term experiments, respectively (Fig. 4a,c).

3.3 Effects of N Application Rates and Humidity Conditions on Leaf Nutrient Concentration and Nutrient Resorption Responses to N Addition

The response ratios exhibited significant differences under different N application rates (Fig. 5). The aboveground biomass and green and senesced leaf N concentrations enhanced with increasing N addition rates (Fig. 5a,b,g), and the maximum occurred at the N application rate of ~ 40 g N m−2 year−1. Phosphorus resorption efficiency decreased with the increasing N addition rates (Fig. 5f), and NRE also decreased to a minimum at the N application rate of ~ 40 g N m−2 year−1 (Fig. 5e). Our results indicated that the effects of N addition on green and senesced leaf P concentrations were shifted from decrease to increase when the N application rate was ~ 10 g N m−2 year−1 (Fig. 5c,d). The green and senesced leaf N concentration response to N addition was enhanced with increasing aridity index (Fig. 6a,b), whereas the green and senesced leaf P concentration decreased with increasing aridity index (Fig. 6c,d). The aridity index did not significantly affect NRE and PRE responses to N addition (Fig. 6e,f).

Fig. 5
figure 5

Relationships of the natural logarithm of the response ratio (lnRR) of leaf nutrient concentrations, nutrient resorption, and biomass with changes in the nitrogen application rates. Ng, green leaf nitrogen concentration; Ns, senesced leaf nitrogen concentration; Pg, green leaf phosphorus concentration; Ps, senesced leaf phosphorus concentration; AGB, aboveground biomass; BGB, belowground biomass; NRE, nitrogen resorption efficiency; PRE, phosphorus resorption efficiency

Full size image
Fig. 6
figure 6

Relationships of the natural logarithm of the response ratio (lnRR) of leaf nutrient concentration, nutrient resorption, and biomass with changes in the aridity index [aridity index = MAP/(MAT + 10)]. MAP, mean annual precipitation; MAT, mean annual temperature; Ng, green leaf nitrogen concentration; Ns, senesced leaf nitrogen concentration; Pg, green leaf phosphorus concentration; Ps, senesced leaf phosphorus concentration; AGB, aboveground biomass; BGB, belowground biomass; NRE, nitrogen resorption efficiency; PRE, phosphorus resorption efficiency

Full size image

4 Discussion

Our global data analyses of N addition experiments indicate that N addition alters leaf nutrient uptake strategies and patterns. The relationship between green leaf N and P and between NRE and PRE changed under N addition, but the relationship between senesced leaf N and P was not altered. Green leaves in grassland ecosystems tend to have higher N:P ratios. These results verify that nutrient resorption may be an important pathway regulating the relationships between leaf N and P concentrations. The multiple responses of varying grassland types and plant groups to N addition also highlight the complex impacts of future N deposition on global grasslands.

4.1 Nitrogen Addition Alters Leaf Nutrient Concentrations and Resorption in Grasslands

Consistent with our first hypothesis, N addition significantly enhanced the green and senesced leaf N concentration. This was primarily attributed to enhanced soil N availability (Yuan and Chen 2015a; Taylor et al. 2021). These results are also consistent with previous global-scale synthesis, which demonstrated that N addition significantly increased foliar N concentration under all nutrient-limited conditions and increased the foliar P concentration under P-limited conditions (You et al. 2018a). However, the results in our study demonstrated that N addition only increased P concentration in senesced leaves across global grasslands universally limited by P (Van Dobben et al. 2017). The nonsignificant change of P concentration in green leaves under N addition may be due to the dilution of green leaf P concentration caused by N-induced enhancement of aboveground biomass (Sardans et al. 2016). Overall, these findings suggested that plants alter the nutrient uptake strategies and tend to have higher N:P ratios in green leaves across grassland ecosystems under future N enrichment.

Our analysis also revealed that N addition reduced NRE in grassland ecosystems, which confirmed our second hypothesis. Nitrogen input enhanced soil N availability and reduced N resorption from senesced tissues (Soudzilovskaia et al. 2007; Li et al. 2016). Previous studies also reported the consistent results that N addition significantly reducing the NRE for all plant growth types by about 12–13% on a global scale (Yuan and Chen 2015b; You et al. 2018b). However, this is slightly higher than the 9% reduction of NRE under N addition we found in grassland ecosystems. This may be due to the limitation of soil nutrient availability across global grassland (Craine and Jackson 2010). Plants tend to have high nutrient resorption efficiency to maintain their growth under limited soil nutrient availability (Wright and Cannon 2001; Zong et al. 2018). The results of our study also indicated that N addition reduced PRE, which was in agreement with previous reports of N addition reducing PRE in forest, grassland, and shrubland ecosystems (Yuan and Chen 2015b; Su et al. 2021). This suggested that soil was the primary P source for plants under N addition, even in P-limited grassland ecosystems. Previous studies also showed that N addition leads to an imbalance in N and P cycles in soil and plants during biogeochemical processes (Penuelas et al. 2012) and can decouple N and P cycles in plants (Lü et al. 2016; Yan et al. 2018). However, the results in our analysis showed that plant N and P cycles remained coupled under N enrichment in global grassland ecosystems, whereas the relationships between NRE and PRE, and between green leaf N and P concentrations, changed significantly under N enrichment. Early studies reported that nutrient resorption had a key role in maintaining the coupled balance between N and P cycles in senesced leaves (Lü et al. 2016). Plants uptake critical nutrients from senesced leaves, which reduces their dependence on soil nutrient availability (Killingbeck 1996). This can explain why PRE was reduced under N enrichment. Nutrient resorption from senesced leaves may be an important pathway regulating the coupled relationship between N and P cycles.

4.2 Nitrogen Addition Differentially Affected Leaf Nutrient Concentrations and Nutrient Resorption in Different Subgroups

Our results indicated that N addition differentially affected leaf nutrient concentrations and nutrient resorption in temperate and alpine grasslands. Alpine grasslands displayed more sensitivity in leaf N concentration changes under N addition than temperate grasslands. Plants in alpine environments tend to have efficient nutrient uptake and utilization strategies to adapt to nutrient limitations under extreme environments (Zong et al. 2017; Zimmer et al. 2018). Combined with the decrease of green leaf P and unchanged senesced leaf P under N addition in alpine grasslands, it suggests that alpine plants will tend to have a higher N:P ratio than temperate plants under N addition. Generally, a higher N:P ratio means that N is abundant while P is relatively deficient, which suggests that plant growth is limited by P availability (Zong et al. 2018). Alpine grasslands will probably show a more severe limitation of P availability than temperate grasslands under future N enrichment. By contrast, temperate grasslands displayed greater sensitivity in nutrient resorption efficiency under N addition than alpine grasslands. Nutrient resorption efficiency reflects soil nutrient conditions and the plant adaptation strategies to soil nutrient availability in different environments (Kobe et al. 2005). The nonsignificant response of nutrient resorption efficiency to N addition in alpine grasslands may be due to the low soil availability of N and P, which was likely induced by the inhibition of enzyme and microbial activities caused by low temperature under high altitude (Liu et al. 2021b). The increase of plant biomass in alpine grassland under N enrichment means that plants demand to acquire extensive nutrients, which is obviously difficult for alpine grassland soil to satisfy. Therefore, alpine plants still need to uptake substantial nutrients from senesced tissues to meet their own growth needs even under N addition. Plants in alpine grasslands tended to display stronger dependence on nutrient resorption from senesced tissues than those in temperate grasslands under N addition.

The green leaf N concentration displayed a greater increase under N addition in grasses than in forbs, but nutrient resorption efficiency displayed a greater decrease under N addition in forbs than in grasses. As one of the dominant functional groups in grassland ecosystems, grasses always have higher nutrient resorption efficiency than other species (Zong et al. 2018), which is an effective strategy for the successful competition of grasses with other species. The greater decrease in green leaf P concentration of grass than that of forb under N addition may be due to higher P utilization efficiency in grasses than in forbs (Hayes et al. 2018; Pereira et al. 2018). These results demonstrated that grasses possess a higher N:P ratio under N addition than forbs. This suggests that P limitation in grasses under future N deposition may be more pronounced than in forbs in global grasslands. Long-term N addition and NH4NO3 fertilizer were more likely to shift the leaf nutrient in grassland ecosystems, as NH4NO3 is more soluble in water and more easily acquired and utilized by plants. In summary, our analyses indicated that the responses of leaf nutrient concentrations and nutrient resorption to N addition differed among grassland types, plant groups, and experimental conditions. These results will deepen our understanding of the effects of N enrichment in global grassland ecosystems.

4.3 Factors that Shift the Impact of N Addition on Leaf Nutrient Concentrations and Nutrient Resorption

Consistent with our third hypothesis, the responses of leaf nutrient concentration and nutrient resorption to N addition were modulated by factors such as N application rates and humidity. Our analysis showed that N application rates of ~ 40 g N m−2 year−1 had greater impacts on leaf N concentration and NRE than other N application rates. The plant N uptake gradually increased due to enhanced soil N availability when the N addition rate was below the critical threshold, whereas plant N uptake was limited by the availability of other nutrients such as carbon and phosphorus when the N addition rate was above the critical threshold (Yuan and Chen 2015a, b). This similar critical threshold was also reflected in our analysis of the response of aboveground biomass to N addition. Our results also exhibited that the turning point from decrease to increase in leaf P concentration to N addition was when the N application rate was ~ 10 g N m−2 year−1, which was consistent with the results of a previous study of grasslands across northern China (Su et al. 2021). Low N addition (< 10 g N m−2 year1) tends to inhibit plant uptake of P, but high N addition (> 10 g N m−2 year−1) stimulates plant P uptake. This also may be why soil available P differed due to varying responses of soil phosphatase activity to N application rates (Gong et al. 2020; Widdig et al. 2020). Moreover, previous studies reported that N addition increased aboveground biomass by 20–35% and the effects of N addition on aboveground biomass had a unimodal distribution (Liu and Greaver 2010; Jiang et al. 2019). These responses are also reflected in our analysis. Dilution effects resulting from the increase in plant biomass may also partly explain the complex responses of leaf P concentration to N addition due to mismatched changes in plant biomass and soil phosphatase activity (Yuan and Chen 2015a). However, the mechanism of a critical threshold for the changes in plant P concentration to N addition still needs to be further studied.

Humidity (aridity index) also shifted N addition impacts on leaf nutrient concentrations. Our results indicated that the response of green and senesced leaf N to N addition increased with greater aridity index, whereas the response of green and senesced leaf P to N addition decreased with rising aridity index. This may result from a synergy between N enrichment and humidity, which was reported in previous studies (Grunzweig and Korner 2003; Copeland et al. 2012). Grassland ecosystems can be particularly sensitive to changes in nutrient and water availability (Su et al. 2021). An increase in soil moisture usually led to a higher N mineralization rate and improved the availability of soil inorganic N, which further promoted plant aboveground biomass and nitrogen absorption (Schuster and Dukes 2017; Dijkstra et al. 2018). However, this synergy may also lead to an inhibition of plant growth induced by excessive nitrogen and water. The aridity index did not affect plant biomass and nutrient resorption efficiency responses to N addition, which may be due to the multiple responses of different grassland types and plant groups to the effects of N addition and water availability. For example, N addition and water availability impacts on leaf nutrient concentration and nutrient resorption of grass and forb plants were different in semi-arid grassland, and the distinction also was reflected in varying grassland ecosystems (Lu and Han 2010; Zhang et al. 2019). Overall, our combined results indicated that temperature and precipitation jointly modulate the responses of plant nutrient concentrations and nutrient resorption to N addition in grassland ecosystems.

5 Conclusions

Our meta-analysis synthesized data on changes in leaf nutrient concentration and resorption across global grassland ecosystems under experimental N addition. The results suggested that N enrichment altered leaf nutrient concentration and nutrient resorption in grassland ecosystems, which was modulated by N application rates, temperature, and precipitation. The sensitivity of leaf nutrient concentration and resorption to N enrichment varied greatly across grassland types and plant groups. These findings provide evidence that nutrient resorption plays a significant role in regulating the plant nutrient strategies to respond to future N deposition. Our results contribute to better predictions of changes in the plant nutrient cycle under N enrichment in global grassland. Future changes in temperature and precipitation may affect the soil–plant nutrient cycles of global grassland ecosystems by mediating plant nutrient responses to N addition.