This study was conducted at Eco-animal Husbandry Experimental Zone, Chinese Academy of Sciences, located in the Xeltala cattle breeding farm in Hulunbuir, Inner Mongolia (120°0′E, 49°17′N, 680 m a.s.l.). The primary soil classification prevalent in this area is kastanozems, categorized in accordance with the FAO/Unesco System of Soil Classification. This area is situated within a semi-arid continental climate, characterized by an average annual temperature of −3°C and an average annual precipitation of 375 mm. The vegetation is characterized as meadow steppe, including Leymus chinensis, Carex duriuscula, Carex pediformis, Poa pratensis, Thalictrum squarrosum, Pulsatilla turczaninovii, Cleistogenes squarosa, Artemisia tanacetifolia, and Potentilla bifurca.

Within the enclosed natural steppe, the Natural Grassland Restoration Multi-factor Experiment Platform (207 m × 111 m) took place in 2021. The experiment was a split-plot design involving seed addition and fertilizer application. We selected the N and P addition treatments on this platform, resulting in a full-factorial block-designed experiment. It consisted of 40 plots divided into ten blocks, each accommodating four plots (10 m²), including Control (ambient, no nutrient addition), P addition (5 g P m⁻² yr⁻¹, calcium superphosphate), N addition (10 g N m⁻² yr⁻¹, urea), and the combined treatment of N and P addition (10 g N m⁻² yr⁻¹ and 5 g P m⁻² yr⁻¹). During mid-May of each year, the solid particles of nutrients were manually scattered to experimental plots by hand in an even distribution manner.

In early September of 2022 (at the end of growing season), we conducted an assessment of the belowground bud density and aboveground shoot density within each experimental plot’s 0.3 m² square quadrat. Recognizing that the majority of belowground buds are concentrated within the 0–30 cm soil profile of this steppe, we excavated all belowground components to this depth [31,34]. Throughout the sampling process, the integrity of the connection between aboveground shoots and belowground organs was maintained to facilitate subsequent species identification for belowground buds. Soil was meticulously cleaned, and the samples (comprising rhizomes, roots, and aboveground plant parts) were placed in plastic bags and transported to the laboratory. The samples stored in an incubator (4°C) before bud identification and counting. According to the procedure of previous studies [31], belowground buds were categorized into four types: tiller buds (axillary buds growing at the base of shoot in caespitose species and rhizomatous grasses), rhizome buds (axillary buds and apical buds on hypogeous rhizomes), bulb buds (meristems wrapped in the swollen leaf base or scale leaf of bulbous species) and dicot buds (buds growing on belowground parts of dicotyledonous herbs) (Table A1). Direct counting was feasible for buds on rhizomes and roots, whereas shoot bases needed to be dissected for tiller bud and bulb bud counting. During this process, only turgid buds were considered for counting, excluding any necrotic or obviously dead tissues. All tasks should be completed within one week to prevent any deterioration of the samples. Meanwhile, aboveground shoots were counted for each species, and then sorted into perennial rhizomatous grasses (featuring tiller and rhizome buds), perennial bunchgrasses (featuring tiller buds), and perennial forbs (featuring bulb and dicot buds).

The ANPP in each experimental plot was also estimated by harvesting all aboveground biomass at its growth peak in the middle of August 2022 by mowing all plant species to soil surface within a 1 m² square quadrat each plot. The samples were put in a drying oven at 60°C for 48 h and weighted (0.1 g in accuracy).

Prior to conducting statistical analysis, the quantity of belowground buds and aboveground shoots within per quadrat were used to calculate the densities of buds and shoots, respectively, expressed on a per square meter basis. In this study, annual species (their relative abundance in most quadrats <10%, and they constituted less than 5% of total aboveground biomass) were excluded from the analyses. We mainly focused on perennial species, which were classified into three functional groups: perennial rhizomatous grasses, perennial bunchgrasses, and perennial forb. We also calculated the ratio of bud to shoot density (i.e., meristem limitation index-MLI) to estimate the degree to which aboveground population regeneration was constrained by belowground bud bank [38].

Levene’s tests for heteroscedasticity and Shapiro–Wilk tests for normality were performed on all data before statistical analyses. The confirmation of variance homogeneity and normal distribution allowed for the use of untransformed data in our statistical analyses. To examine the interactive effects of nutrient addition treatments on belowground bud density, aboveground shoot density, and MLI for the whole community and three plant functional groups, we used Generalized Linear Mixed Effects Models with N addition and P addition as fixed factors and block as a random factor. We compared the bud density, shoot density, and MLI across four treatment using analysis of variance (ANOVA), followed by Tukey Test for subsequent multiple comparisons within different functional groups and the whole plant community. We further performed regression analysis to explore the relationship between the belowground bud bank and aboveground vegetation. All analyses and figures were carried out using R 4.2.1 software (R Core Team, 2022).

For the whole plant community, there were no N × P interaction on total bud density, shoot density and MLI (Table 1). Both N addition and P addition and their combination had no significant influences on the bud density, shoot density and MLI of the whole plant community (Fig. 1).

As regard to belowground bud bank composition, we found specific responses in bud density, shoot density and MLI of three plant functional groups to soil nutrient addition. N addition significantly increased the bud density and shoot density of PR (bud density: p < 0.001, shoot density: p < 0.01, Table 1, Figs. 2A and 2B), especially with combination of P addition, even though P addition had no significant influences (p > 0.05, Table 1). N addition, P addition and their combination had no significant effects on the MLI of PR (p > 0.05, Table 1, Fig. 2C). Conversely, N addition significantly reduced shoot density of PF (p < 0.01, Table 1, Fig. 2H), particularly combined with P addition, although P addition alone did not have any significant effects (p > 0.05, Table 1), while the soil fertilization treatment had no significant influences on their bud bank and MLI (p > 0.05, Table 1, Figs. 2G and 2I). In addition, N addition, P addition, and their interactions had no significant influences on the bud density, shoot density, and MLI of PB (p > 0.05, Table 1, Figs. 2D–2F).

Considering all nurtient treatments together, we found the significantly linear relationship between belowground bud density and aboveground shoot density (R² = 0.39, p < 0.01, Fig. 3A). However, no clear relationships were observed between aboveground biomass and either bud density or shoot density. (Figs. 3B and 3C).

Under soil nutrient addition, the enhanced grassland productivity could arise from the increased tiller production, as observed in two bunchgrasses [30], or from larger tiller size, as documented in the rhizomatous grass Panicum virgatum [39]. In this study, although a positive correlation was observed between belowground bud density and aboveground shoot density (Fig. 3A), the increased aboveground biomass following nutrient addition might be attributed to the enlargement of individual aboveground shoot size (Table A2, Fig. A1). This could be inferred by the lack of significant relationship between aboveground biomass and shoot density (Fig. 3B), along with no significant influences of nutrient addition on aboveground shoot density (Table 1). Previous studies showed that nutrient addition may modify biomass allocation pattern of plant species [40], resulting in a higher proportion of biomass being allocated to aboveground organs relative to belowground parts [41]. These results suggest that the enhancement of aboveground biomass under short-term nutrient addition may be driven more by enlarging individual size than by increasing shoot abundance.

In the temperate semiarid steppe with both N and P limitation [42], soil nutrient addition to some extent could alleviate soil nutrient limitation. Under these conditions, plant species tend to reduce the biomass allocation to belowground for nutrient competition, while increasing resource allocation to aboveground components to compete for light and space [43]. The reduced biomass allocation to belowground constrains the resource supply for bud production, thus nutrient addition has negative influences on belowground bud density [44]. However, our result showed that belowground bud bank and its relationship with aboveground vegetation have no significant responses to short-term soil nutrient addition. This contrasts with our previous findings in temperate typical steppe, where a decrease in belowground bud density was observed under long-term (7 years) N addition [31]. This might be attributed to the fact that the resources stored in the belowground bud-bearing organs [24,25] could alleviate the instantaneous pressure of reducing belowground input under short-term soil nutrient addition, and thus the aboveground population regeneration/recruitment originated from belowground bud bank remains unaffected. Thus, we infer that a response threshold may exist for the belowground bud bank to nutrient addition or other environmental changes (e.g., drought) [34,45,46]. Short-term nutrient addition does not exceed the buffering capacity of bud banks, resulting in no response in total bud density. Besides, P addition was also found to have no significant impact on aboveground vegetation (i.e., ANPP [24]) and consequently may not change the distribution of biomass between aboveground and belowground, resulting in the constraint imposed by the belowground bud bank on aboveground population regeneration remains unaffected. Meanwhile, the response of belowground bud density may be attributed to the characteristic of P element, with low diffusion and high fixation in soil [47]. The added P might initially be adsorbed onto the surface of clay minerals and Fe/Al oxides through the formation of various complexes, and thereby become unavailable to plants [48]. Therefore, belowground bud bank had no significant response to short-term P addition. In addition, the lack of interaction between N addition and P addition on bud density indicates that N and P might independently influence the belowground bud bank and its potential regenerative contribution in semi-arid meadow steppe.

Nutrient addition did not change MLI for the whole plant community (Table 1), and the MLI kept constant under both N addition and P addition (Fig. 1). This implies that short-term soil nutrient addition also does not modify the relationship between belowground bud bank and aboveground shoot population. In simpler terms, the constraint imposed by the belowground bud bank on aboveground population regeneration/recruitment does not seem to be influenced by short-term nutrient addition.

The responses of various bud types to environmental changes would significantly influence the relative dominance of plant functional groups, thereby reshaping the composition and structure of plant communities under global changes [29,31,34]. Our results demonstrate that PF and PR showed contrasting responses to the short-term soil nutrient addition, i.e., N addition (itself as well as combined with P addition) increased the bud density and shoot density of PR but decreased the shoot density of PF. The varying responses between PR and PF might be ascribed to their differing physiological characteristics. Under nutrient addition, the increasing NH⁴⁺ could to a greater extent suppress the enzymes activity and reduce the photosynthetic rate of PF in compared with PR [37], resulting in the reduced resource allocation to belowground for bud production and thus affecting the population regeneration of PF. Compared with PF, PR usually elongated rhizomes and increased rhizome bud density to occupy more growth space [40]. Previous studies have confirmed that under N addition, rhizomatous grasses could allocate more photoassimilates to belowground organs (i.e., rhizomes) to produce more rhizome buds for space occupation and population colonization [36,37]. These changes may significantly increase the belowground bud bank and facilitate the population regeneration/recruitment of PR. Therefore, our study suggests that nutrient addition has positive effects on rhizomatous grasses both for belowground bud bank and aboveground vegetation, while its negative impacts on forbs mainly act on their aboveground shoot population. The addition of N, P, and their interaction all have no significantly influences on the bud density, shoot density, and MLI of PB, suggesting that both the belowground bud bank and its regenerative capacity of this functional group were insensitive to short-term nutrient addition. This insensitivity of both whole plant community and PB might be attributed to grassland soil nutrient context and fertilization experimental duration [36], thus different grassland types and experimental durations should be incorporated in future studies to explore the potential responses of belowground bud density and composition as well as its regenerative contribution to nutrient addition.