Introduction

Grasslands, which constitute approximately 41% of the earth's ice-free land (~ 52.5 million km2), play a crucial role in providing various ecosystem services (White et al. 2000). Ungulate grazing is a widespread land use practice in grasslands. It is a significant driver of plant community structure, which is fundamental to ecosystem function, including aboveground net primary productivity (ANPP) (Li et al. 2017; Piipponen et al. 2022). Grazing is not just a physical disturbance to plant productivity, but also has indirect effects mediated by diversity and community structure (Zhang et al. 2018). Grazing is a complex biotic disturbance involving multiple components, including defoliation, trampling, and dung and urine return, which collectively influence grassland ecosystems, shaping plant community composition and productivity (Liu et al. 2015). However, there remains a knowledge gap in understanding the mechanistic links between community functional composition and productivity under grazing disturbance, particularly in exploring how these multiple disturbances operate, which could advance our understanding of the biodiversity-productivity relationship in managed grasslands.

Plant functional traits can serve as indicators of plant response to disturbance at both species and community levels (Dı́az et al. 2007; Jäschke et al. 2019), which help us to understand the mechanisms underlying the effects of grazing on plant communities. Generally, the community functional composition is described by both community-weighted mean (CWM) and functional diversity (FD). CWM reflects the dominant trait value within the community, while FD quantifies the trait dissimilarity among coexisting species. Previous studies examining the response of community functional composition to grazing components have primarily concentrated on interspecific trait variation (caused by species turnover), assuming that such variation captures the majority of the underlying processes (Violle et al. 2012). For instance, defoliation diminishes the dominance of taller species pursuing a resource-conservative strategy and mitigates community light limitation (Borer et al. 2014), creating opportunities for shorter species with larger SLA to thrive and shifting CWM towards resource-acquisitive strategies. Nutrient return favors tall species as they have an advantage in acquiring light, thereby driving the change of CWM through species turnover. As competition for light intensifies, shorter-statured species may be excluded from the community gradually (Wang et al. 2022). Defoliation and nutrient return can potentially reduce functional diversity by selecting species with specific traits (Zhang et al. 2023; Zheng et al. 2021). Additionally, few studies have found that trampling affects community-level functional traits and FD by causing species turnover (Li et al. 2023).

However, plant traits are not fixed and can be shaped by biotic or abiotic factors via intraspecific trait variation (ITV). A global meta-analysis reported that ITV explained ~ 25% of the trait variation within communities and ~ 32% among communities (Siefert et al. 2015). Consequently, the effects of ITV on community functional composition, community assembly, and ecosystem functions have gained increasing attention in recent years (Wright et al. 2016). In response to ungulates, ITV reflects different species' adaptive strategies to mammalian herbivory (Amaral et al. 2021). The species with grazing-avoidance strategies tend to reduce their palatability, such as short stature, tough leaves with high leaf dry matter content (LDMC), and small specific leaf area (SLA) (Archibald et al. 2019). Conversely, grazing-tolerant plants are associated with high SLA or leaf N concentration (LNC), enabling fast compensatory growth, which trade-offs exist with grazing avoidance strategies (Li et al. 2017; Wright et al. 2004). These response strategies are primarily related to defoliation, but the contributions of dung and urine return and trampling to the ITV response in relation to ungulate grazing have not been adequately documented. Similar to nutrient addition, dung and urine return may lead to changes in community functional composition by intraspecific variation (Wang et al. 2022). Specifically, increased soil nutrient availability increased individual plant height and SLA (Zhang et al. 2019). Moreover, trampling with moderate intensity can induce compensatory growth in specific species, such as Medicago ruthenica in the Tibet Plateau (Xiao et al. 2018). However, understanding the contribution of these species-level changes to community-level functional composition under trampling is still limited (Li et al. 2021b).

Plant functional traits can also provide insights into the relationship between plant communities and ANPP (Hanisch et al. 2020). According to the classical theoretical framework of biodiversity-ecosystem functioning, changes in the species composition (species turnover) are an essential factor influencing productivity. The relationship of CWMs with ANPP suggests that species with a high (or low) trait value will promote productivity (Zuo et al. 2022), which may emphasize the role of dominant species in driving ecosystem function. Traits, such as SLA and leaf nitrogen content, which are thought to be positively correlated with photosynthetic rate, can ultimately affect ANPP (Ali et al. 2013). On the other hand, communities with higher FD may exhibit higher niche differentiation, resource utilization, and thus higher productivity, an idea derived from diversity theory (Bongers et al. 2021). However, the maintenance of grassland productivity is not only dependent on species turnover, intraspecific trait variation may also play an important role. A study in a semi-arid grassland found that ITV of plant stem/leaf ratios, influenced by changes in grazing density, was a crucial factor mediating the relationship between interspecific trait variation and ANPP (Zheng et al. 2021). Additionally, intraspecific trait variation due to nutrient addition (e.g., increased SLA) is an important pathway through which nutrient inputs increase ANPP. The impact of species turnover and intraspecific trait variation on ANPP under different grazing components is still unknown, limiting our understanding of the relationship between plant communities and ecosystem functioning in disturbed grassland.

In this study, we conducted a long-term field experiment in a semiarid grassland in Inner Mongolia, China, simulating the disturbances from ungulate herbivores through various treatments, including defoliation, dung and urine return, trampling, and their combinations. We measured four traits (plant height, SLA, LDMC, and leaf C/N) linked to resource uptake strategies, the competitive ability for light, and adaptation to grazing (Dı́az et al. 2001). We asked the following questions: (a) How do plant diversity and community functional composition (CWM and FD) respond to simulated grazing? (b) To what extent is the response of functional composition to simulated grazing due to species turnover vs. ITV? (c) How does the simulated grazing affect ANPP through intra- and interspecific trait variations, and FD? Specifically, we tested the following hypothesis: (a) Defoliation is a dominant factor in the effects of grazing on plant diversity and functional structure, leading to community convergence (i.e., a community composed of species with similar functional traits or having a low FD) toward resource-acquisitive strategies (higher SLA, lower leaf C/N, and LDMC). Dung and urine return increase plant height and specific leaf area, while trampling decreases plant height. (b) Variations in CWM and FD under defoliation will be mainly explained by species turnover, but the effects of dung and urine return and trampling are more interpreted through ITV due to the change of soil properties. (c) CWM and FD together drive the productivity response to simulated grazing through species turnover.

Material and methods

Study site and experimental design

This experiment was carried out in a temperate steppe at the Duolun Restoration Ecology Station of the Institute of Botany of the Chinese Academy of Sciences (42.02°N, 116.17°E, 1324 m a.s.l.), Inner Mongolia, China. The study area has a semiarid continental monsoon climate with a mean annual temperature of 2.1 °C and annual precipitation of 378 mm (Xu et al. 2022). The soils belong to Haplic Calcisols according to ISSS Working Group RB (Deckers and Nachtergaele 1998). Dominant plant species in this area are perennial grasses, forbs and subshrub, including Stipa krylovii Roshev. (Poaceae), Cleistogenes squarrosa Keng (Poaceae), Leymus chinensis Tzvel. (Poaceae), Potentilla acaulis L. (Rosaceae), and Artemisia frigida Willd. (Asteraceae). Our study site was fenced since 2003 to exclude large herbivores and for hay-making at the end of the growing season until 2014.

The simulated, moderate-intensity cattle grazing was established in a randomised block design with eight blocks in 2015 (64 plots in total, 4 × 4 m per plot). We choose mowing to a stubble height of 5–6 cm three times per year (in the middle of June, July, and August) to simulate defoliation (Baoyin et al. 2014). In this study, our defoliation has no selectivity. The former results showed that the diet selectivity of cattle was not as strong as that of sheep, and it usually happened under light grazing intensity with higher plant coverage and height (Chen et al. 2015; Fraser et al. 2009). Trampling was performed by a 75-kg man carrying a weight of 25 kg, who wore shoes made from the hind pair of a cow's hooves, and applied three times per year after mowing. It was equivalent to 65% of the pressure of an adult cow. The trampling induced soil compaction after five years of treatments (Fig. S1a), and the bulk density is equivalent to that under moderate grazing intensity (Ren et al. 2018). The dose of fresh dung and urine mixture applied to the plot was determined by the mean herbage production per hectare in the pasture and the mean digestibility of the grazed herbage (Mikola et al. 2009). A total of 1.5 L urine per m2 mixed with dung was applied to the soil, split into two applications per year, right after each of the first two mowing treatments and trampling was finished. In total, the two applications of the mixture had 1,152 g of dry matter, 495 g of C, 87 g of N, and 12 g of P per plot per year. Consequently, the experiment included factorial combinations of mowing (M; no mowing vs. mowing), a liquid mixture of dung and urine addition (D; no addition vs. addition), and trampling (T; no trampling vs. trampling), resulting in eight different treatment combinations (CK: control, D: dung and urine return, M: mowing, T: trampling, D + M: dung and urine return + mowing, M + T: mowing + trampling, D + T: dung and urine return + trampling, D + M + T: dung and urine return + mowing + trampling).

Plant community and trait sampling and measuring

In the middle of August 2019, after five years (2015–2019) of treatment, we surveyed plant communities at their peak standing biomass before the third time treatment implementation. Aboveground vegetation was sampled by clipping all plants using a 0.5 × 0.5 m quadrat randomly placed in the plot. All living vascular plants were sorted into species, oven-dried at 65 °C for 48 h and weighed. All plants were classified into five functional groups (rhizome grass, bunchgrass, legume, forb, and subshrub). The dry weight of all live plants in mid-August was used to estimate aboveground net primary productivity (ANPP) for the unmowed plots. In the defoliated treatment plots, we recorded the biomass removed after each mowing by drying and weighing it. The ANPP for these plots was calculated as the sum of the biomass removed each time and the standing biomass in mid-August.

The effects of our treatments on plant community structure gradually increased from 2016 to 2019 (Fig. S2). To explore the cumulative effects of simulated grazing treatments, we sampled and measured functional traits in 2019 after plant community survey. We measured traits from 3 to 5 individuals for each species in each plot. The functional traits were quantified using methods proposed by Perez-Harguindeguy et al. (2013). Plant height was measured by averaging the height of the three tallest individuals of each species. We randomly sampled at least five healthy, fully expanded leaves from 3 to 5 individuals of each species to measure leaf traits. After sampling, the fresh leaves were immediately scanned (PER-V700P, EPSON, Japan) to measure leaf area and weighed. Then, the leaves were oven-dried (65 °C) to a constant mass and weighed. LDMC was calculated as dry leaf mass per unit of fresh leaf mass, and SLA was the leaf area per unit of dry leaf mass. These dried leaves were ground in a ball mill (MM400, Retsch, Germany) and used to determine carbon (C) and nitrogen (N) contents (Elementar vario MICRO cube elemental analyzer, Elementar, Germany).

Soil sampling and physicochemical properties determination

In the middle of August 2019, we sampled soil. A stainless steel cutting ring (5 cm diameter, 100 cm3) was used to evaluate soil bulk density and soil water content. Three soil cores (5 cm diameter, 0–10 cm depth) were randomly collected from the quadrats for biomass evaluation in each plot and combined to gain one composite soil sample. After homogenization and removal of roots gently, the soil was sieved through a 2-mm mesh and stored at 4 °C.

Soil pH was measured in soil suspension with a 1:5 ratio of soil to deionized water with a pH/Conductivity Benchtop Meter (ORION STAR A125, Thermo Scientific, USA). Soil ammonium nitrogen (NH4+–N) and nitrate nitrogen (NO3−–N) were extracted with a 1 M KCl solution and measured with a continuous flow analyzer (AA3, SEAL Analytical, Germany). A subsample of 2.5 g of soil was extracted with 0.5 M NaHCO3 (pH 8.5) to measure extractable P concentration using the molybdenum blue method to get the content of AP (Olsen 1954).

Quantification of inter- and intraspecific functional variance

The changes in functional traits at the community level are typically described by community weight means (CWM), reflecting the dominant trait value based on the mass-ratio hypothesis (Grime 1998). Functional diversity (FD) indicates community assembly (Laliberte and Legendre 2010) and describes how dissimilar the coexisting plants are in a community. All traits were weighted by the relative biomass of each species to calculate CWM for each plot using the formula:

$${\text{CWM}}=\sum_{{\text{i}}=1}^{{\text{n}}}{{\text{p}}}_{{\text{i}}}\times {{\text{t}}}_{{\text{i}}}$$

where pi is the relative biomass of species i in the plot, and ti is the mean trait value of species i in the plot (specific trait values).

We used a method presented by Lepš et al. (2011) based on the sum of squares (SS) decomposition analyses to assess the relative importance of species turnover and ITV in driving CWM. In this method, we first calculated three CWM parameters for each functional trait: (1) 'Specific average', i.e., CWM as mentioned above, whose response to total change factors can be caused by both species turnover and ITV. (2) 'Fixed average' (CWM_Fixed), calculated by mean trait values of all plots, and their response variation caused by species turnover. (3) 'Intraspecific variability effect' (CWM_Intra), i.e., the differences between 'fixed' and 'specific' averages. Next, three-way ANOVA for 'specific', 'intraspecific' and 'fixed' were performed separately with the three grazing components and their interactions as predictors. We extracted the SS for each trait value that was explained by the three grazing components and their interactions, so that SS_specific, SS_fixed and SS_intraspecific represented the treatment-induced total variation, effect of species turnover, and effect of ITV, respectively. Finally, the SS_cov component, which is the effect of covariation between inter- and intraspecific trait variability, was calculated by subtracting SS_fixed and SS_intraspecific from SS_specific (SS_total). Positive covariance effects between species turnover and ITV indicate that they change in the same direction, potentially amplifying the response of CWM (or FD) to treatments. In contrast, negative covariance effects between species turnover and ITV suggest that they change in opposite directions. Ultimately, SS can be expressed like this: SS_total = SS_fixed + SS_intraspecifc + SS_cov.

FD with Rao index for each plot was calculated using relative biomass and standardized specific trait value dissimilarity (Euclidean distance) between species (de Bello et al. 2011).

$${\text{FD}}=\sum_{{\text{i}}}^{{\text{n}}}\sum_{{\text{j}}}^{{\text{n}}}{{\text{p}}}_{{\text{i}}}\times {{\text{p}}}_{{\text{j}}}\times {{\text{d}}}_{{\text{ij}}}$$

where pi and pj are the relative biomass of species i and j. dij is the specific trait dissimilarity (Euclidian distance) between each pair of species i and j.

To quantify the ITV and species turnover in changes of FD, we also calculate the FD parameters based on the methods of Niu et al. (2020) and de Bello et al. (2011). 'Inter_FD' was calculated using species-level dissimilarity, the distance of mean trait values. We calculated 'Intra_FD' as the dissimilarity of deviation of trait values from the grand mean value for each species. And then, we calculated rITV as the ratio of Intra_FD to Inter_FD. Finally, using a sum of squares decomposition analysis (de Bello et al. 2011), similar to that of Lepš et al. (2011) based on ANOVA, we quantify the contribution of species turnover and ITV to functional diversity. Similarly, SS is expressed by the following formula: SSFD = SSInter_FD + SSIntra_FD + SScov_FD.

In order to investigate the impact of treatments on the trait values of various functional groups, we computed the CWMs of different functional groups using the methodology mentioned above (Fig. S3). Notably, the parameter pi in the formula uses the relative biomass of species i in the functional group to which it belongs.

Data analysis

All calculations and statistical analyses were performed in R 4.0.3. We used species richness and Shannon index as indicators of species diversity based on the number and relative biomass of the species present. To visualize the variation of plant composition under simulated grazing, we applied non-metric multidimensional scaling (NMDS) ordination via the metaMDS function in the R package vegan. Our ordination was based on the Bray–Curtis dissimilarity matrix of the vegetation datasets with 64 plots. The effects of defoliation, dung and urine return, trampling and their interactions on vegetation composition were tested with a permutational multivariate analysis of variance (PERMANOVA, adonis in vegan) on the same dissimilarity matrix (999 permutations).

Linear mixed-effects models were employed to evaluate the effects of defoliation, dung and urine return, trampling, and their interactions on species diversity, community functional composition, functional groups' CWM and soil properties. The lme4 package was utilized to implement the models, with the block included as a random effect to account for potential spatial variability. Prior to analysis, normality of the data was assessed using the Shapiro–Wilk test. For the data that did not meet the assumptions of normality, we used logarithmic or aligned rank transform (a preparation for nonparametric analyses of variance; ARTool package).

Finally, a Bayesian linear mixed effect model (brms package) was performed to predict the effect of soil physicochemical properties on community functional properties, and to explore the effect of plant richness, intra- and interspecific changes of CWM and FD on ANPP. Before the Bayesian linear mixed effect models of functional properties, all soil factors were standardized, and the descdist function (fitdistrplus package) was used to examine the distribution of response variables. Before the model procedure of ANPP, we conducted principal component analyses (PCAs) to extract the first principal component (PC1) of intra- and interspecific changes of CWM (CWM_Intra and CWM_Fixed) and FD (FD_Intra and FD_Inter) based on the four functional traits. PC1 explained 37.06% and 69.03% of the total variance in CWM (Fig. S4 a, b) and 44.41% and 65.28% of the total variance in FD (Fig. S4 c, d) for intraspecific and interspecific trait variability. Beta brms-model was employed for ANPP divided by its maximum value, and we used block as a random term. Markov Chain Monte Carlo (MCMC) simulations were performed to estimate the fit of the models (median and 95% interval of R2) and range of effect size (95% credibility intervals) of predictors (intra- and interspecific changes of CWM and FD).

Results

Plant species diversity and community structure

After five years of simulated grazing treatments, we found that all three grazing components did not significantly affect species richness (Fig. 1a). However, defoliation reduced the Shannon diversity index (Fig. 1b) and significantly changed the community structure (Fig. 1d, Table 1). Specifically, defoliation increased the relative abundance of subshrub Artemisia frigida (P < 0.001), while sharply decreasing the dominance of perennial rhizome grass (i.e., Leymus chinensis, Agropyron cristatum) (P < 0.001; Fig. 1c, Table S1). PERMANOVA revealed that dung and urine return also caused a significant change in community composition (P < 0.05) by reducing the proportion of forbs (P < 0.05) and increasing subshrub (P < 0.05). Dung and urine addition only promoted the growth of rhizome grasses in un-defoliated communities (M × D, P < 0.05; Fig. 1c, Table S1). Trampling had no significant effect on plant community structure.

Fig. 1
figure 1

Effects of defoliation (mowing, M), dung and urine return (D) and trampling (T) on species richness (a), and Shannon diversity index (b). Relative biomass of each functional group in community (rhizome grass, bunchgrass, legume, forb and subshrub) in all eight treatments (c). Ordination of community structure under different treatments by non-metric multidimensional scaling (NMDS) (d). Box-plot shows quartile values (n = 8) for each treatment. × denotes the interaction between grazing components, while the + symbol represents a treatment that combines several grazing components. Significance levels of the factors are as follows: *P < 0.05, **P < 0.01, ***P < 0.001

Full size image
Table 1 The effect of soil physicochemical properties on plant community properties for four traits
Full size table

Response of community-weighted mean trait values

The defoliation significantly changed CWM values of all plant traits, and overrode other grazing components. Specifically, defoliation decreased plant height, LDMC, and leaf C/N ratio, while increasing the SLA (Fig. 2a–d). Dung and urine addition led to taller stature, higher SLA, and lower leaf C/N ratio (Fig. 2a, b, d, P < 0.05). Trampling increased the LDMC at the community level (Fig. 2c, P < 0.05). Dung and urine return in combination with defoliation increased LDMC compared to control (CK), but decreased it when mowing was absent (D × M, P < 0.01; Fig. 2c).

Fig. 2
figure 2

Community-weighted mean (CWM) of plant height a, specific leaf area (SLA), b, leaf dry matter content (LDMC), c and leaf C/N ratio d under control (CK), defoliation (mowing, M), dung and urine return (D), trampling (T), and their combinations. Sum of squares decomposition analyses showing the relative importance of intraspecific trait variability (ITV) vs. species turnover in explaining CWM (e–h). In plots a–d, box-plot shows quartile values (n = 8) for each treatment. The + symbol represents a treatment that combines several grazing components. Significance levels of the factors are as follows: *P < 0.05, **P < 0.01, ***P < 0.001. In plots e–h, light and dark green bars represent species turnover and ITV effects, respectively, and black lines indicate the total CWM variation (including species turnover, ITV, and their covariation effects) explained by grazing components. Covariation effects are represented by the space between the top of the bar and the black line. The line above the bar denotes positive covariation, and the line across the bar denotes negative covariation. The × symbol denotes the interaction between grazing components

Full size image

Overall, species turnover (23.91–68.67%) and ITV (40.61–54.01%) contributed almost equally to the total variation of the four CWM values in this study (Fig. S5). The effect of defoliation on SLA, LDMC, and leaf C/N was mainly achieved by affecting species turnover (Fig. 2). However, the influence of defoliation on plant height was primarily driven by ITV. Furthermore, there was a positive covariation between species turnover and ITV for height and leaf C/N (Fig. 2e, h), but a negative covariation for SLA and LDMC (Fig. 2f, g). Under dung and urine return treatment, ITV explained more variance in CWM values than species turnover, particularly for SLA and leaf C/N. Similarly, the variation in CWM of LDMC under trampling was mainly driven by ITV. Considering the interactive effect between defoliation and nutrient addition, the contribution of species turnover (2.17%) and ITV (2.33%) were almost equal for CWM of LDMC.

Response of functional trait diversity

Defoliation significantly decreased the functional diversity of the four traits (Fig. 3a–d). Dung and urine return reduced the functional diversity of leaf C/N (Fig. 3d). However, trampling had no significant effect on FD. Compared to the CWM, the variations of FDs were driven more by ITV (Fig. S5). Specifically, for plant height and leaf C/N, defoliation primarily affected FD by reducing ITV (Fig. 3e, h). The relative contributions of species turnover and ITV to LDMC variation were almost equal under defoliation, while the change in SLA was mainly contributed by species turnover (Fig. 3f, g). The variation in leaf C/N under dung and urine return was attributed to species turnover and a positive covariation (Fig. 3g, h).

Fig. 3
figure 3

Community functional diversity (FD) of plant height a, specific leaf area (SLA), b, leaf dry matter content (LDMC), b and leaf C/N ratio d under control (CK), defoliation (mowing, M), dung and urine return (D), trampling (T), and their combinations. Sum of squares decomposition analyses showing the relative importance of ITV vs. species turnover in explaining the FDs (e–h). In plots a–d, box-plot shows quartile values (n = 8) for each treatment. The + symbol represents a treatment that combines several grazing components. Significance levels of the factors are as follows: *P < 0.05, **P < 0.01, ***P < 0.001. In plots e–h, light and dark orange bars represent species turnover and ITV effects, respectively, and black lines indicate the total FD variation (including species turnover, ITV, and their covariation effects) explained by grazing components. Covariation effects are represented by the space between the top of the bar and the black line. The line above the bar denotes positive covariation, and the line across the bar denotes negative covariation. The × symbol denotes the interaction between grazing components

Full size image

Predicting functional properties from soil physicochemical properties

Soil available nutrients (soil inorganic nitrogen and available phosphorus) have a greater impact on functional properties than bulk density and pH (Table 1). The CWM, CWM_Fixed (calculated by mean trait values of all plots), intraspecific variation, and FD of plant height increased with increasing soil available N. Moreover, both CWM_Fixed and FD of LDMC exhibited positive correlations with soil available N. Additionally, FD_intra (FD variation associated with ITV) and rITV (relative importance of ITV for FD) of LDMC were positively influenced by phosphorus availability. The functional properties (CWM, CWM_Intra, FD, FD_Intra, and rITV) of leaf C/N ratio showed positive correlations with soil available N and P, while there was a negative and positive effect of bulk density on CWM_Intra and rITV, respectively. For SLA, there was only a significant negative effect of phosphorus availability on CWM.

Responses of aboveground net primary productivity

The ANPP increased significantly under dung and urine return treatment (P < 0.05, Fig. 4a), while other grazing components did not have significant effects. Bayesian linear mixed effect model demonstrated that functional composition explained 32.05% of the variation in ANPP (Fig. 4b). The intraspecies variation of CWM (CWM_Intra) had a significant effect on ANPP, and between-population ITV (FD_Intra) had a large but non-significant effect. However, plant richness and interspecific trait variation did not contribute significantly to ANPP.

Fig. 4
figure 4

ANPP a under control (CK), defoliation (mowing, M), dung and urine return (D), trampling (T), and their combinations. *Indicates the significant effect (P < 0.05). Error bars represent ± SE. The effect of plant richness, the intraspecific variability of the CWM (CWM_Intra), the CWM based on fixed trait value (CWM_Fixed), between-population ITV (Intra_FD) and interspecific variation (Inter_FD) on the ANPP (b). The slope in Bayesian regression models indicates the effects with 95% credibility intervals, and the fit of the models is indicated by the median and 95% interval of R2. Positive versus negative effects are highlighted in blue and red, respectively, and the significant effects occur when the 95% credibility intervals don't overlap zero

Full size image

Discussion

Our study first disentangled the effects of individual or combined three grazing components on community functional composition and ANPP. We found that defoliation and dung and urine return significantly altered the community composition and community-level height, SLA, LDMC and leaf C/N. On the other hand, the effect of trampling was comparatively weaker. The variability of CWM induced by defoliation was mainly driven by species turnover, while that induced by dung-urine return and trampling was driven primarily by ITV after 5-year simulated grazing treatments. Consequently, species turnover and ITV were equally important in explaining the simulated grazing effects on CWM trait values. Furthermore, defoliation significantly reduced FD, with ITV explaining most of the variation (Fig. 5). Dung and urine return increased ANPP mainly mediated by changes in intraspecific trait values. These findings highlight that grazing may through multiple pathways affect functional composition and their variation to regulate productivity in grassland ecosystems, with ITV playing an important role in the process (Fig. 5).

Fig. 5
figure 5

A conceptual framework illustrating the mechanisms of grazing mechanisms (defoliation, trampling, and dung and urine return) changes on plant community functional composition and productivity. Red, blue, and yellow arrows represent the effects of defoliation, dung and urine return and trampling respectively, with the thickness of the line indicating the magnitude of the effect. ITV intraspecific trait variation, Intra_FD between-population ITV, SLA specific leaf area, LDMC leaf dry matter content, C/N leaf C/N ratio, ANPP aboveground net primary productivity

Full size image

Species turnover and ITV together drive the effects of grazing components on CWMs

As hypothesized, defoliation resulted in a shift of traits to the strategies of rapid growth and resource-acquisitive (higher SLA, lower LDMC and leaf C/N) at the community level, and the main driver of changes caused by grazing was species turnover. Defoliation reduced the dominance of species with conservative strategy (L. chinensis and S. krylovii, Table S2) and decreased litter accumulation (Fig. S6), and consequently altered irradiance and temperature within the community (Jessen et al. 2020; Liu et al. 2019), which optimized the habitat for some forbs and subshrubs colonization. Additionally, high SLA and low LDMC values positively correlate with fast growth rate and grazing tolerance (Laliberté et al. 2012; Zheng et al. 2010), such as the species A. frigida in our study.

The results of the variance decomposition analysis revealed that species turnover and ITV have different covariance effects on different CWM values under defoliation, suggesting that they can reinforce or counteract each other (Fig. 2e–h). The positive covariations were observed for plant height and leaf C/N, suggesting that defoliation promotes the dominance of plants with short-statured and low leaf C/N at both the community and species levels (Fig. S3). In contrast, although the variation of CWM for SLA and LDMC resulted from species turnover under defoliation, there were negative covariations with ITV. Specifically, defoliation led to higher SLA and lower LDMC at the community level, but some individuals or species tended to decrease SLA (e.g., legumes and forbs) and increase LDMC (e.g., subshrub, Table S3, Fig. S3). These species-level changes may be considered avoidance strategies against grazing, resulting in decreased visibility and palatability to livestock, which differs from the grazing tolerance strategy at the community level (Dı́az et al. 2007; Jessen et al. 2020). The differences in grazing response at community and species level may reduce the trade-off between the traits associated with these two strategies, i.e., the community eventually exhibits a simultaneous increase in compensatory growth rate and grazing resistance (Fig. 2a–d). This finding suggests that ITV may cause community-level traits to behave differently from the leaf economic spectrum, complementing the theory of Wright et al. (2004).

In comparison, dung and urine return increased plant community height and SLA, and decreased leaf C/N, mainly explained by ITV (Fig. 2, Fig. S7). These results agreed with observations that dung and urine addition can alleviate nutrient limitation, stimulate plant growth and increase light demand (Eskelinen et al. 2012; Zhang et al. 2019). Plants tend to develop leaves with higher SLA in nutrient-enriched soil to maximize light capture efficiency (Rozendaal et al. 2006). This plasticity in plant traits can enhance their competitiveness under nutrient enrichment, even when exposed to defoliation (Jessen et al. 2020). Trampling increased the CWM of LDMC, which may be an adaptive response to mechanical stress through trait plasticity. Mechanical stress caused by trampling stimulates plants to produce more mechanical tissue (e.g., collenchyma and fibers) with more structural carbohydrates and lower water content (Leroux 2012). We found significant interactions between defoliation and dung-urine return on the CWM values of LDMC (Fig. 2c), but there were no significant interactions among the three grazing components in the CWM values of the other traits. While previous studies reported interactive or additive effects of the three components (Liu et al. 2015; Zhang et al. 2022), the role of defoliation was dominant in the changes of plant community structure in our simulated experiment. Despite the different pathways of grazing components influencing CWM, the similar contributions of species turnover and ITV to CWM variation under the simulated grazing (Fig. 2, Fig. S5) suggest that defoliation didn't ultimately counteract the effects of dung and urine return and trampling.

ITV is the main driver of the effects of grazing components on FDs

Grazing is known to reduce FD compared to no grazing in arid or semiarid grassland (Chillo et al. 2017). In our study, defoliation's effects on Rao's FD override those of trampling and dung and urine return (Fig. 3), indicating that defoliation as a habitat filter affects the range of trait values (Fischer et al. 2019). Defoliation leads to a convergence of community traits towards rapid growth and resource-acquisitive strategies. The contribution of deterministic processes to community assembly may increase under the habitat filters, reflecting the mechanism of α-diversity (Shannon index, Fig. 1b) reduction under grazing (Mammola et al. 2021). In addition, defoliation decreased niche differentiation due to light competition, allowing plants with similar traits to coexist in the community (Keddy 1992).

The changes in FD also indicate that both biotic and abiotic stress tend to alter plant phenotypic plasticity (Niu et al. 2020; Valladares et al. 2007). Our study found that defoliation significantly reduces Intra_FD and rITV (Fig. S8), and that ITV contributed the most to the variation in FD (Fig. 5, Fig. S5). Defoliation directly damages plants and induces compensatory growth or defense against herbivores (Valladares et al. 2007). The compensatory growth might lead to deficiencies of available soil nutrients (Fig. S1), indirectly affecting plant functional diversity. Moreover, defoliation removed aboveground biomass and decreased litter inputs (Fig. S6), aggravating these negative effects on soil nutrients and FD (Niu et al. 2016). We found a significant positive effect of soil available nutrients on FD of LDMC and leaf C/N (Table 1), suggesting that nutrient limitation caused by defoliation leads to the convergence of traits related to resource acquisition (Niu et al. 2020). Plants tend to enhance their ability to access resources to increase their inter- and intraspecific competition. Contrary to our initial prediction, dung and urine return did not significantly change soil available nutrient content (P > 0.05, Fig. S1c, d) in this study, which indicated that nutrient return may not fully compensate for the soil nutrient loss caused by defoliation.

Our study demonstrates that ITV plays a crucial role in shaping changes in plant functional composition (CWM and FD) in response to grazing components (Fig. 5, Fig. S5), suggesting that local plants have likely evolved adaptations to herbivore grazing over a long history of grazing (Adler et al. 2004; Li et al. 2021a). The plasticity of plant traits allows for changes in functional composition under disturbance without relying solely on species turnover, which can serve as a buffer against changes in plant composition and protect species from extinction (Lepik et al. 2005). Hence, this mechanism ensures the relative stability of plant richness in local communities under grazing (Fig. 1a) (Volf et al. 2016).

ITV-induced changes in CWM mediate the effects of grazing components on ANPP

Our results revealed that the simulated grazing positively affected ANPP, especially under dung and urine return (Fig. 4a). Contrary to our third hypothesis, intraspecific variation-induced changes in CWM played a key role in the changes in productivity, while the effects of both intra- and interspecific FD on ANPP were not significant (Figs. 4b, 5), suggesting that mass-ratio theory mainly explained the increase of ANPP (Mokany et al. 2008). This could be attributed to the fact that nutrient inputs increased SLA and leaf nitrogen content within species (Fig. S7, Table S1), which consequently increased the photosynthetic rate at the community level (Ali et al. 2013). Although the absence of significant effects of defoliation and trampling on ANPP, the positive effect of dung and urine return, along with the compensatory effect of defoliation, explained the mechanism of ANPP enhancement under moderate grazing. We found that A. frigida compensated for the biomass loss of perennial rhizome grass caused by defoliation (Fig. 1c, Tabel S3). But this compensation may have a negative impact on animal production performance (Wang et al. 2007). It is worth noting that dung and urine return will not be distributed as evenly for real livestock grazing as in our simulated grazing (Moir et al. 2011). Therefore, the significant contribution of grazing to productivity may not be observed in some livestock grazing experiments due to the sampling effect.

It is important to acknowledge that our simulated grazing treatments do not fully replicate real grazing by ungulates. Our treatments did not account for the selectivity and persistence of defoliation (Onatibia et al. 2023), and the spatial heterogeneity of defoliation, trampling, and dung and urine return (Liu et al. 2018). Furthermore, it is possible that our anthropogenic disturbance had different effects on other herbivores (such as arthropods) compared with real grazing (Wang et al. 2020), which also has a significant effect on vegetation. Despite these limitations, we believe that our study provides valuable insights into the mechanisms through which management practices, such as grazing and mowing, impact grassland.

Conclusion

This study provides novel insight into how large herbivores affect plant community functional composition and ANPP, and is crucial to understanding plant–herbivore interactions. Based on a simulated grazing experiment, we demonstrate that defoliation had the greatest influence on CWM and FD. Under defoliation, species turnover accounted for most of the variance of CWM, while intraspecific trait variability played a more critical role in the decrease of FD. Dung and urine return and trampling changed CWM more through ITV than species turnover, and these two components had almost no significant effect on FD. Furthermore, ANPP was increased by dung and urine return and mediated by changes in intraspecific trait values. These results highlighted the key role of ITV in shaping grassland plant community functional composition and productivity under ungulate grazing. Our study also pointed out that even at moderate intensities, continuous removal of aboveground biomass (defoliation) had potential negative effects on individual plants, resulting in soil nutrient deficiencies and a decrease of α diversity. Therefore, grazing regimes and intensities should be adjusted in grassland management according to key species' phenotypic plasticity and grazing adaptability, and nutrient addition should be implemented for long-term grazing grasslands.