The study area is located in the Siziwang base of the Comprehensive Experimental Demonstration Center of the Inner Mongolia Academy of Agricultural and Animal Husbandry Sciences (41°47′17″N, 111°53′46″E, elevation 1450 m), which belongs to the short-flowered coniferous desert steppe zone, and the climate is a medium-temperate continental type. The average annual precipitation is 280 mm, evapotranspiration is 2300 mm, the average annual temperature is 3.4°C, ≥10°C cumulative temperature is 2200~2500°C, the frost-free period is 90~120 days, and the soil is dominated by light chestnut-calcium soil, which provides a typical environment for exploring the spatial response of the vegetation to the rate of livestock carrying.

In this study, a grazing control experiment was established in June 2023 in a meadow in northern China, with a total test area of 8.4 ha, and two gradient treatments were set up: a no-grazing control (CK) and heavy grazing (HG) ( Figure 1 ) (Zhang et al., 2023). Among them, the HG treatment unit was configured at a loading rate of 0.45 sheep hectare^-1 month^-1 and 12 healthy adult sheep were placed in each independent sample plot (about 26.7 ha). CK1 and HG1 were selected as the core observation plots, and the grazing intervention period covered the complete plant growing season (June-November), with 12 hours (06:00-18:00) of continuous grazing per day. All experimental sheep were uniformly bred as two-year-old de-stemmed rams to ensure the consistency of individual physiological status and feeding behavior.

The study analyzed the vegetation distribution pattern through nested multi-scale sampling: a 50 m × 50 m main sample plot (vertex as the coordinate origin) was set up in the loading rate experimental area, and 25 m × 25 m (5 m × 5 m grid) and 2.5 m × 2.5 m sample plots were nested sequentially, the latter matching plant coordinates through a virtual grid (0.25 m to 0.05 m scale), combining field measurements with computer analysis to collect 36 grid vertex data (species, height, biomass). The data integration covered both macro (50 m/25 m) and micro (below 2.5 m) scales to analyze the variation thresholds of interspecific competition intensity with spatial scales, and to reveal the mechanism of the influence of stocking rate on the spatial pattern of vegetation ( Figure 2 ).

The plant community of the Stipa breviflora desert grassland is divided into four functional groups based on life type and ecological function, and each functional group maintains the stability of the ecosystem through synergistic effects (Wang et al., 2020). The perennial grasses are the established functional groups in this grassland, with Stipa breviflora as the dominant species. The functional groups of Stipa breviflora desert grassland are categorized as follows ( Table 1 ).

To quantitatively verify the sensitivity of the sampling scale and reveal the spatial pattern changes caused by grazing, we adopted geostatistical semi-variogram analysis. Based on the importance values of plant functional groups and their exact coordinates of each grid vertex within splines, the empirical semi-variation function was calculated.

Significant differences in functional group diversity indices were observed across grazing treatments. At all spatial scales, heavy grazing exhibited a reduced Shannon-Wiener diversity index (H’), Simpson’s dominance index (D), and Pielou’s evenness index (J’) compared to control groups. However, at the 25m×25m scale, heavy grazing showed an elevated Margalef richness index relative to controls.

The effects of spatial scale on functional group diversity were non-significant. At the finer 2.5m×25m scale, species diversity indices were lower than those at broader scales (50m×50m and 25m×25m). Nevertheless, minimal divergence occurred between 50m×50m and 25m×25m scales, indicating relatively weak scale-dependent impacts on diversity within larger spatial extents ( Table 2 ).

At different spatial scales, perennial grasses had the highest importance values when there was no grazing, and under heavy grazing conditions, perennial grasses continued to have higher importance values, but the importance values of Annual and biennial plants increased significantly, and those of perennial forbs and shrubs and semi-scrubs decreased significantly ( Table 3 ).

The Across spatial scales, the overall interspecific association (variance ratio, VR) in heavily grazed areas consistently exceeded that in ungrazed areas ( Table 4 ).

At a scale of 50m×50m, under non-grazing conditions, there was an overall insignificant negative correlation among functional groups (VR< 1), while under heavy grazing, there was no correlation (VR = 1), and it was also insignificant. This indicates that on a larger scale, there is interspecific competition for environmental resources among functional groups in ungrazed areas, and grazing may alleviate this competitive relationship to a certain extent. At the 25m×25m scale, the functional groups under both grazing treatments showed an insignificant positive correlation (VR > 1), and the VR value under heavy grazing was higher than that under non-grazing conditions. At a fine scale of 2.5m×2.5m, ungrazing still showed an insignificant negative correlation (VR< 1); However, heavy grazing triggered a significant positive association (VR > 1, P< 0.05). This result indicates that on a smaller spatial scale, grazing may have promoted facilitating interactions among functional groups or significantly altered the original competitive landscape.

At the 50m×50m spatial scale ( Figure 6 ), Pearson correlation coefficients revealed positive correlations between shrub/semi-shrub groups and perennial forbs, as well as between shrub/semi-shrub groups and annual/biennial plants in the ungrazed area, whereas negative correlations were observed among -all other functional group pairs. In the heavily grazed area, positive correlations occurred between annual/biennial plants and perennial forbs, with negative correlations prevailing among all remaining functional groups.

At the 25m×25m spatial scale ( Figure 6 ), Pearson correlation coefficients indicated positive correlations between annual/biennial plants and both perennial forbs and shrub/semi-shrub groups in the ungrazed area, while negative correlations characterized other functional group pairs. In the heavily grazed area, only shrub/semi-shrub groups and perennial forbs exhibited a positive correlation. Spearman’s rank correlation coefficients yielded congruent results with Pearson coefficients across all analyses.

At the 2.5 m × 2.5 m spatial scale ( Figure 6 ), Pearson and Spearman correlation coefficients show that only perennial forbs and annual and biennial plants are positively correlated in both ungrazed and heavily grazed areas. In contrast, the remaining functional groups are negatively correlated.

Stability analysis of functional groups using M. Godron’s stability method revealed differential stability patterns across spatial scales and grazing treatments. The importance values show that perennial grasses and perennial forbs are the dominant functional groups, and the number of species of perennial grasses and perennial forbs is high in the experimental area. In contrast, Annual and biennial plants, shrubs, and semi-shrubs show functional group deficiency at smaller spatial scales and under heavy grazing conditions, so their stability is not investigated. Perennial grasses exhibited higher stability in ungrazed areas compared to heavily grazed areas at all spatial scales, though remaining in an unstable state ( Table 5 ). Perennial forbs demonstrated greater stability at larger spatial scales (50m×50m), with higher stability in ungrazed areas than in grazed conditions ( Table 6 ). The maximum stability for perennial forbs occurred in ungrazed 50m×50m plots, evidenced by an intersection point at (19.97, 80.03), closest to the theoretical stable equilibrium (Davison et al., 2020; Tran et al., 2021). Annual/biennial plants and shrub/semi-shrub groups consistently remained unstable across all spatial scales and grazing treatments. According to the Godron stability method, the functional groups under all treatments did not reach the theoretical stable state (Euclidean distance > 0). However, perennial forbs show a tendency to be closer to the theoretical equilibrium point in no-grazing areas than in heavily grazing areas, and their Euclidean distance (50m×50m scale: 17.9) is significantly smaller than the corresponding value under heavily grazing conditions (50m×50m scale: 37.8). This indicates that although the system as a whole is unstable, the absence of grazing interference enhances the relative stability of this functional group.

In the absence of grazing, the semivariance of perennial grasses increases significantly with distance, indicating increasing spatial heterogeneity. Under heavy grazing, the semivariance remains low and flat, suggesting that grazing pressure destroys patch structure, leading to either uniform distribution or fragmentation, and suppressing spatial heterogeneity (Bisigato et al., 2005). For perennial forbs, in ungrazed areas of 25m, the semivariance reaches a peak at medium to short distances before declining, implying the presence of local patches at small scales; in 50m ungrazed areas, the semivariance increases slowly, reflecting weak spatial autocorrelation at large scales; under heavy grazing, the extremely low semivariance indicates that grazing pressure eliminates local heterogeneity, resulting in a random distribution (Bisigato et al., 2005) (Figure 7). For annual and biennial plants, under heavy grazing, the semivariance increases significantly at medium distances, reflecting how grazing pressure creates microhabitats that promote the establishment of annual plants and the formation of medium-scale patches. The fluctuating and disorderly curves in ungrazed areas indicate weak spatial autocorrelation (Figure 7). For shrubs and semi-shrubs, in 50m ungrazed areas, the semivariance continues to increase with distance, demonstrating strong spatial autocorrelation at large scales; in 25m ungrazed areas, the semivariance increases at short distances before declining, suggesting more concentrated shrub patches at small scales; under heavy grazing, the extremely low semivariance shows that grazing pressure suppresses shrub growth and destroys their spatial heterogeneity (Komac et al., 2011) (Figure 7).

Grazing intensity significantly alters plant functional group diversity in Stipa breviflora desert steppe, exhibiting pronounced spatial scale dependency. Overall, heavy grazing markedly reduces diversity indices (e.g., Shannon-Wiener, Simpson, Pielou) due to decreased species evenness and simplified functional group structure (Tran et al., 2021). This simplification manifests as overconsumption of dominant species (e.g., Stipa breviflora) causing community homogenization, while grazing-tolerant species (e.g., annual weeds) rapidly colonize released niches to establish dominance (Sachura et al., 2024). Consequently, high-nutritional-value grass functional groups decline, whereas disturbance-resistant, high-fiber plants increase. Diversity responses diverge across spatial scales: At finer scales (2.5m×2.5m), diversity indices are generally lower, constrained by limited environmental heterogeneity and sampling bias—Stipa breviflora root competition inhibits coexistence of sensitive species, and sparsely distributed forbs may be undersampled (Ravenek, 2015). Conversely, at intermediate scales (25m×25m), the Margalef richness index shows a distinct increase, likely driven by localized outbreaks of grazing-tolerant species (e.g., Chenopodium aristatum) and patch dynamics. Heavy grazing suppresses dominant species expansion, enabling annual plants to rapidly colonize vacated niches and transiently boost species richness (Bajwa et al., 2021). Grazing-induced vegetation patchiness allows this scale to capture more ephemeral species in residual grazing areas, while control plots exhibit lower species counts due to uniform dominance—aligning with the scale-dependent intermediate disturbance hypothesis, where medium scales show greater sensitivity to short-term disturbance responses than broader scales (50m×50m), where spatial averaging masks local variations (Stone, 1995; Mayor et al., 2015; McConaghy, 2016). At broad scales, non-significant diversity differences between grazing treatments likely stem from compensatory effects among functional groups (e.g., spatial complementarity between deep-rooted shrubs and shallow-rooted grasses buffering individual group impacts) (Bakker, 2018; Gamadaerji et al., 2020; Wang et al., 2022). Collectively, grazing regulates diversity dynamics through multi-scale mechanisms: local competition and morphological adaptations [e.g., cuticle thickening in Cleistogenes songorica (Sachura et al., 2024)] dominate species turnover at finer scales, while functional complementarity and resource partitioning sustain system stability at broader scales (Guo et al., 2020; Li et al., 2022a; Montoya et al., 2015).

Grazing intensity and spatial scale interactively reconfigure interspecific associations among plant functional groups in Stipa breviflora desert steppe. Variance Ratio (VR) analysis reveals scale-dependent patterns: at finer scales (2.5m×2.5m), heavy grazing generates significantly positive associations (VR > 1) as it suppresses dominant groups (e.g., perennial grasses), releasing niche space for complementary distributions between grazing-tolerant species (e.g., annual/biennial plants) and residual dominants (Tang et al., 2019; Liu et al., 2022). Conversely, at larger scales (50m×50m), ungrazed areas exhibit non-significant negative VR values indicating inherent resource competition, plants achieve resource division through spatial division of their root systems (root domain territorialization), forming symmetrical competition (resource allocation in proportion to biomass), and reducing intraspecific consumption (de Kroon et al., 2003), while heavy grazing homogenizes resource consumption, driving VR toward neutrality (VR = 1) (Le Bagousse-Pinguet et al., 2012). This supports the Competitive Release Hypothesis where disturbances reduce niche overlap (Segre et al., 2016). The fine-scale positive associations under grazing arise from synergistic environmental filtering, habitat heterogeneity, and interspecific facilitation - reflecting adaptive coordination among stress-tolerant species rather than classical competitive antagonism (Magura et al., 2018; Dai et al., 2020; Jiang et al., 2022).

Functional group pairwise associations show distinct scale-grazing interactions: Ungrazed conditions feature non-significant associations at larger/intermediate scales (50m×50m,25m×25m) due to stable coexistence via resource partitioning (Yang et al., 2018; Pescador et al., 2021; Beck et al., 2022; Homulle et al., 2022; Yang et al., 2015; Sinclair et al., 2020), but significant positive perennial grass-forb correlations at fine scales (2.5m×2.5m) via microhabitat root complementarity (Lencinas et al., 2007; Nagel et al., 2015; Liu et al., 2023). Heavy grazing fundamentally disrupts these patterns: at large scales, grazing-tolerant species form positive associations (Cole, 2003) while grass-forb competition intensifies (Fu et al., 2019); at fine scales, amplified competition generates negative associations among most groups (Pescador et al., 2021), though transient grass-annual plant cooperation persists through shared resource exploitation (Nagel et al., 2015; Zhang et al., 2021). Shrub-related associations further demonstrate scale effects: negative grass-shrub correlations at large ungrazed scales [resource competition (Gamadaerji et al., 2020)] shift to positive shrub-forb associations under grazing [collaborative stress tolerance (Alvarez et al., 2011; Puigdefabregas et al., 1999; Thorpe et al., 1998; Saldo and Bartolome Filella, 2021)]. Collectively, grazing reshapes association networks by altering resource distribution and competitive hierarchies (Saiz and Alados, 2012; Zhang et al., 2020; Newman, 2006; Olesen et al., 2007).

Co-occurrence (PC) and association strength (OI/DI) indices corroborate these dynamics: Ungrazed large scales show high PC (0.6–1) and distributed OI values ([0.3,1]), (Dai et al., 2019; Shan et al., 2018) indicating multi-tiered networks sustained by environmental heterogeneity (Zainelabdeen et al., 2021; Rodríguez et al., 2023; Zheng et al., 2024; Zainelabdeen et al., 2020; Ren et al., 2018). Heavy grazing universally reduces PC values and concentrates OI in higher ranges ([0.6,1]) but with fewer functional groups, forming simplified high-co-occurrence/low-biodiversity communities (Saiz and Alados, 2012; Zainelabdeen et al., 2021; Hao et al., 2022; Zainelabdeen et al., 2020; Ren et al., 2018). Scale reduction intensifies these trends at fine scales (2.5m×2.5m), ungrazed communities maintain elevated OI through microhabitat symbiosis (Gamadaerji et al., 2020; Hu et al., 2022; Hu et al., 2020), while grazing collapses associations to near-zero PC/OI values for most pairs, leaving only rudimentary grass-annual plant networks (Saiz and Alados, 2012; Jiménez et al., 2014). Grazing consistently weakens interspecific linkages by reducing species occurrence probabilities, with scale diminution exacerbating this effect (Tsutsumi et al., 2002; Saiz and Alados, 2012; Zhang et al., 2013; Liang et al., 2018).

Pearson correlations integrate these patterns: Positive shrub-forb/annual correlations in ungrazed large scales reflect microhabitat amelioration (Archer et al., 2017; Davison et al., 2020), contrasting with grazing-induced positive annual-forb correlations through complementary resource-use (Herlocker et al., 1999; Mbembok et al., 2017). Intermediate scales strengthen mutualistic networks ungrazed (Setiawan et al., 2023) but only retain shrub-forb synergies under grazing (Bascompte et al., 2003; Bascompte and Jordano, 2007). Critically, large-scale associations are primarily environmentally mediated [heterogeneity-driven resource allocation (Wang et al., 2014; Hai et al., 2021)], while fine-scale linkages are governed by direct biotic interactions (Losapio et al., 2018). Grazing drives transitions from multispecies synergies to simplified tolerance-dominated networks via resource redistribution and physiological stress (Abu Bakar et al., 2023), offering actionable insights: restoring degraded grasslands requires optimizing resource heterogeneity at broad scales while enhancing key species interactions at fine scales (Peters et al., 2006; Wang et al., 2018; Zhao et al., 2023).

Plant functional groups exhibited distinct stability responses to grazing across spatial scales in Stipa breviflora desert steppe. Perennial grasses showed higher stability in ungrazed areas than heavily grazed plots at all spatial scales yet maintained overall instability. Their long life cycles and regenerative capacity supported greater biomass and structural stability without grazing (Zhang et al., 2020), but grazing directly damaged aboveground tissues (reducing photosynthesis and reproduction), while soil compaction and nutrient depletion in heavily grazed areas further impaired recovery (Hamza and Anderson, 2005). Notably, even ungrazed perennial grasses displayed relatively low stability, likely due to natural disturbances (e.g., climatic fluctuations) and interspecific competition (Rayburn, 2011; Siebert et al., 2020). Perennial forbs reached peak stability at broader scales (50m×50m) under ungrazed conditions [intersection point approaching (Davison et al., 2020; Tran et al., 2021)], where increased environmental heterogeneity enabled adaptive niche strategies (e.g., divergent root architectures and nutrient uptake) to buffer environmental fluctuations (Liang et al., 2022; Ning et al., 2022; Meng et al., 2023). Heavy grazing reduced forb stability by damaging biomass and disrupting soil-nutrient cycles (Giuliani et al., 2024; Hao et al., 2024). Conversely, annual/biennial plants and shrubs/semi-shrubs remained unstable across all scales and grazing intensities: short-lived species exhibited high vulnerability to environmental variability due to brief life cycles and seed-dependent regeneration (Zhang et al., 2020), further compromised by grazing consumption of seedlings/seeds; woody plants suffered from slow growth, browsing/trampling sensitivity, patchy distributions, and weak interspecific interactions (El-Mahi, 1990). These differential responses arise from scale-grazing interactions modulating growth traits, ecological strategies, and disturbance resilience—perennial groups show grazing-sensitive stability (especially at broader scales), while short-lived and woody groups persist unstable throughout the grazing gradient.