Understanding how plant nutrient acquisition strategies respond to grazing at the community level is critical to understanding ecosystem structure and functioning in grasslands. However, few studies have simultaneously compared the difference in aboveground (leaf) and belowground (root) nutrient-acquisition strategies in response to long-term grazing, especially at the regional scale. Here, we measured a set of leaf and fine-root traits that correspond to the fast-slow economic spectrum at the community level in 10 experimental sites from paired grazed and ungrazed grasslands across a soil nutrient gradient covering three major types of grasslands in northern China. We found that patterns of variations of leaf and fine-root traits were consistent with both a leaf and root economic spectrum at the community level for both grazed and non-grazed plots. Grazing had a minor effect on community-level leaf nutrient-acquisition strategies but strongly influenced community-level root nutrient-acquisition strategies. Specifically, root nutrient-acquisition strategies were shifted to more exploitative resource use in grazed communities. Moreover, soil nutrients contributed to the changes in both leaf and root nutrient-acquisition strategies, which tended towards a more resource-acquisition strategy with increasing soil nutrient levels. Grazing significantly interacted with soil nutrients to affect root nutrient-acquisition strategies, and grazing contributed more to root nutrient-acquisition strategies than soil nutrients. Our results demonstrated completely inconsistent responses of community-level above- and below-ground resource acquisition strategies to long-term grazing, and below-ground acquisition strategies were more sensitive to long-term grazing. Our findings also suggest that high-intensity anthropogenic activities such as grazing may strongly modify below-ground resource acquisition strategies.

Study area and experimental design

The study area was located in temperate grasslands of the Inner Mongolian Plateau in Northern China (111.23 E–120.03 E, 41.25 N–49.52 N), where the climate is predominantly arid and semi-arid continental; Mean annual precipitation ranged from 225 to 433 mm and mean annual air temperature ranged from −2.3°C to 5.4°C. We selected three different types of grasslands along this transect from east to west: meadow steppe, typical steppe, and desert steppe. A total of ten sites with livestock long-term grazing and with various dominating plant species were selected along this transect, including five meadow steppes, three typical steppes, and two desert steppes (Zhang et al., 2023; See Table S1; Figure S1, S2 for details of the study sites, their locations, and their climatic conditions). Within each site, a long-term grazing exclosure was established in an area with a history of long-term heavy grazing. Livestock were excluded via the exclosure for over 10 years at each of the ten sites, and we observed many important changes in vegetation structure (Zhang et al., 2023; Figure S1).

At each site, a 50 m × 50 m plot was selected randomly, and five 1 m × 1 m quadrats were set at the four corners and the centre of the plot. The adjacent quadrats were separated by about 20 m. In total, there were 100 quadrats (10 sites × 2 grazing treatments × 5 quadrats per plot).

Field sampling and measurements 

We measured leaf and root traits on several species of plants that were among the more common species at each site (those comprising at least 90% of the total plant abundance in each site) in late July 2020. In each quadrat, we randomly selected at least 5 mature plants from each species. For each sampled individual, we carefully excavated the plant and surface soil (0–30 cm), and removed all soils that adhered to the roots. We then placed samples in vacuum bags and put them in a car refrigerator for transport to the laboratory, where they were frozen at −20°C for subsequent morphological and chemical analyses.

For leaves, we determined the leaf projected area using an area meter (YMJ-CH, Top, China). We then weighed leaves for wet weight and dry weight after oven-drying at 65 °C for 48 h. We estimated specific leaf area (SLA) as the leaf area per unit of dry leaf mass and leaf dry matter content (LDMC) as dry mass per unit of water-saturated fresh mass. We then ground leaves to fine powder and measured leaf N concentration (LNC) using a Discrete Auto Analyzer (Smartchem 600, AMS Alliance, Italy).

For roots, we conducted measurements on fine roots (root diameter < 2mm). Prior to measurements, we washed fine roots in deionized water and then arranged and scanned roots from each quadrat on an Epson scanner (Epson Expression 12000XL, Epson, Japan). Thereafter, we oven-dried roots from each sample at 65°C for 48 hr and weighed them. From these, we estimated root diameter (RD) as the total root length and volume using the scanned images with the WinRHIZO software, specific root length (SRL) as the total root length divided by its dry mass, and root tissue density (RTD) as the ratio of root dry mass to its volume. After grinding root tissues to a powder, we measured root N concentration (RNC) using a Discrete Auto Analyzer (Smartchem 600, AMS Alliance, Italy).

To measure soil nutrient availability, we collected three soil cores (2.5 cm diameter) at 10 cm depth in each of the five quadrats per plot. We then mixed these soil cores to form one composite sample representing each quadrat. After removing the rocks and roots, we passed the soil through a 2-mm-mesh sieve and separated it into two parts. We kept the first part at -20 °C until analyses could be made and then dried samples at 105 °C for 24 h to measure soil water content (SWC). We air-dried the second part at room temperature and further sieved it to 0.15 mm to determine soil total nitrogen (STN) and total phosphorus (STP) with a Discrete Auto Analyzer (Smartchem 600, AMS Alliance, Italy).

Statistical analyses

For each species in the non-grazed and grazed plots at each site, we calculated the mean values of leaf traits (SLA, LDMC and LNC) and root traits (SRL, RD, RTD and RNC) and then calculated the community-weighted means (CWM) of each trait for each quadrant. To evaluate the existence of a fast-slow economic spectrum of leaf and root (LES and RES), we performed principal component analysis (PCA) on the standardized CWM of leaf and root traits from non-grazed and grazed grasslands. The first PCA axis captured a high proportion of the CWM variation (>50%). Consequently, its scores can be used as a proxy of the economics spectrum given that they represent gradients of trait variation across sites (PC1). For the PCA analysis, we used the rda function within the vegan package (Oksanen, 2023). We used linear mixed-effects models to analyze the interactive effects of grazing and soil nutrients on the leaf and root economic spectrum (LES and RES; PC1) and leaf and root traits. To do so, grazing, soil nutrients and their interaction were fitted as fixed factors, and quadrat and site (plot IDs) nested within grassland types were fitted as random factors to control for pseudo-replication. To reduce the number of variables characterizing the soil nutrient and their collinearity, the first PCA axis represented the majority of the variation used in the above analysis. The function lmer within the lme4 package was used to fit linear mixed effects models (Bates, Maechler, Bolker, & Walker, 2014). The percentage of variance explained by each variable and its interaction (soil nutrient and grazing and their interaction) was determined using the pamer.fnc function in the LMERConvenienceFunCtions package (Tremblay & Ransijn, 2011). We chose the lower.p.val. Also, we used FDR method to adjust the p-value. To distinguish the relative contribution of ITV and species turnover on changes in the CWM for each leaf and root trait, we used a sum of squares decomposition following Lepˇs et al. (2011). We computed three parameters for each functional trait at the community level: (1) CWMspecific were calculated by the relative organ abundance and trait values of each species in a given soil core, where the total variation included both species turnover and ITV effects. (2) CWMfixed were calculated by the relative organ abundance of each species in a given soil core and the trait values were averaged over the total number of plots, which only included the effects of species turnover. (3) ITV was calculated as the difference between the CWMspecific and CWMfixed. General linear models for specific average, fixed average, and intraspecific variability were performed separately with grazing as the predictor. We extracted the sum of squares (SS) for each trait value that was explained by the grazing factor, so that SStotal, SSturnover, and SSintra, represented the grazing-induced total variation, effect of species turnover, and effect of ITV, respectively. The effects of covariation (SScov) were then calculated by subtracting SSturnover and SSintra from SStotal.