This experiment was conducted at Duolun Restoration Ecology Station, which is located in southern Inner Mongolia Autonomous Region (42°02'N, 116°17'E, 1,324 m a.s.l). The long-term (1954–2013) mean annual precipitation is 385 mm, and the mean annual temperature is 2.1°C. Ninety percent of the precipitation occurs between May and October. Monthly mean temperature ranges from −17.6°C in January to 19.2°C in July. The soil is classified as chestnut according to the Chinese classification. Dominant plant species in the temperature steppe include the perennial herbs Stipa krylovii, Cleistogenes squarrosa, and Agropyron cristatum (Yang et al., 2012).

In this study, the two most common grasses, S. krylovii and C. squarrosa, were selected as the research objects in the grasslands of Inner Mongolia. Stipa krylovii is a grass that tends to grow in clusters with large individual (high: 30–80 cm), which is advantageous under nutrient enrichment (Zhao et al., 2016). By contrast, C. squarrosa is a lower cluster grass with small individual (high: 10–30 cm), which makes it more prone to loss under nutrient enrichment. Moreover, with a fibrous root system, C. squarrosa is considered as a key species for sustainable grassland development (Liang et al., 2002).

Our experiment was nested within an existing long-term mowing and N addition experiment that began in 2012 (Wang et al., 2020, 2021b). Five 24 × 4 m blocks were arranged into one row and five columns. Each block was randomly assigned to four plots, each 4 × 4 m, with four treatments: (1) control (C), (2) mowing (M), (3) N addition (N, ambient plus 10 g N m⁻² year⁻¹, NH₄NO₃), and (4) combined M with N addition (MN, Wang et al., 2020). The subset of 15 plots of control, mowing, and N addition treatments were used in this experiment.

We collected the seeds of our two study species from a natural community in 2016. These seeds were sown in a seedbed in the field on May 10, 2017, and were carefully nursed for 20 d. Previous observations indicated that the roots of these two species were distributed in the top 15 cm of soil (Yang et al., 2011). Cylindrical root ingrowth containers were made from rigid plastic mesh (diameter = 7 cm, length = 15 cm, mesh = 4 × 4 mm square) (Chen et al., 2017). These root ingrowth containers contain two specifications, one is made of dense mesh (15 cm long, 5 cm width, 50 μm mesh) and the other is made of sparse mesh (15 cm long, 5 cm width, 2 mm mesh). Dense mesh can isolation root competition of surrounding plants for target plants, but sparse mesh cannot. On June 1, 2017, 20 root ingrowth containers were installed in each subset plot, including 10 dense mesh and 10 sparse mesh, and filled with in situ soil. Ten S. krylovii seedlings were placed in five root ingrowth containers with dense mesh and five root ingrowth containers with sparse mesh in each subset plot. Ten C. squarrosa seedlings were placed in the other half of root ingrowth containers in each subset plot. The in each subset plot were watered with 200 ml every day and dead seedlings were replaced for the first week. The seedlings were left to grow naturally until the end of September when they were sampled. The results under the C, M and N treatment were analyzed in this experiment.

Photosynthetic active radiation (PAR) on the ground was measured three times per month near the seedling within each plot using a Li-Cor Quantum Sensor (Li-Cor, Lincoln, NE, USA) on clear days. Two PAR values of the upper part of the canopy (PARu) and the surface (PARs) were measured at each site. Intercept photosynthetic active radiation (PARi) was calculated using the following formula: PARi = (PARu – PARs)/PARu.

On September 30, 2017, we recorded the number of surviving individuals and measured the maximum height of each plant. All seedling in each plot were then taken out from the ingrowth containers. Because of the short time of the experiment, all the roots were located inside the ingrowth containers. Each seedling was separated into shoots and roots. Roots were gently washed from the soil. All samples were oven-dried at 65°C for 48 h and weighed.

In the middle of August 2017, we harvested the biomass of surrounding vegetation at the peak aboveground plant biomass according to species in a 1 × 1 m square in each plot. Aboveground net primary productivity was estimated using standing biomass. Two 50-cm-deep holes were excavated with a soil auger (5-cm internal diameter) in each plot. Soil was sieved through a 2-mm screen, and roots were washed to measure the belowground net primary productivity. All samples were oven-dried at 65°C for 48 h and weighed. Net primary productivity (NPP) is equal to aboveground net primary productivity plus belowground net primary productivity.

We used three-way ANOVAs to assess the effects of species, root isolation, and management strategy on biomass production, height, survival, shoot, root, and shoot/root ratio. Duncan's multiple range test was used to compare differences between treatments. Regression analyses were used to assess the contributions of NPP and PARi to seedling characteristics of the two species. All statistical analyses were performed in R 3.5.0 (Team, 2018).

The belowground root isolation (RI) treatment significantly increased mean individual biomass and height by 207 and 46% (Table 1; Figure 1) across both S. krylovii and C. squarrosa, respectively. N addition significantly decreased mean individual biomass and survival by 60 and 18%, respectively (Supplementary Table S1; Figure 1). There was a significant interaction effect between RI and N on individual biomass (Supplementary Table S1). RI significantly increased the individual biomass of S. krylovii and C. squarrosa by 226 and 189% and their height by 47 and 50%, respectively (Table 2; Figure 1). N addition significantly decreased the individual biomass of S. krylovii and C. squarrosa by 57 and 63% and their survival by 20 and 15%, respectively (Supplementary Table S2; Figure 1). Mowing did not affect the individual biomass of S. krylovii and C. squarrosa. However, the survival of S. krylovii was reduced by 14% under the mowing treatment (Supplementary Table S2; Figure 1). There was a significant interaction between RI and N addition on the biomass of S. krylovii (Supplementary Table S2).

Across the two species, RI treatment significantly increased mean shoot and root mass by 231 and 168%, respectively, and reduced the root/shoot ratio by 0.12 (Table 1; Figures 2, 3). N addition significantly decreased mean shoot and root mass by 61 and 58%, respectively (Supplementary Table S1; Figure 2). There was a significant interaction effect between RI and N on shoot and root mass (Supplementary Table S1). RI significantly enhanced the shoot mass of S. krylovii and C. squarrosa by 216 and 224% and root mass by 249 and 98%, respectively (Supplementary Table S2; Figure 2). N addition significantly decreased the shoot mass of S. krylovii and C. squarrosa by 54 and 68% and root mass by 63 and 52%, respectively (Supplementary Table S2; Figure 2). RI significantly decreased the root/shoot ratio of C. squarrosa by 0.25 (Table 2; Figure 3). N addition significantly increased the root/shoot ratio of C. squarrosa by 0.15 (Supplementary Table S2; Figure 3). There were significant interaction effects between RI and N on the shoot and root mass of S. krylovii and between RI and M on the shoot/root ratio of C. squarrosa (Supplementary Table S2).

Simple linear regression analyses showed that the biomass of S. krylovii was negatively correlated with NPP under the no-isolation treatments (R² = 0.29, P = 0.040, Figure 4A). The biomass of S. krylovii was negatively correlated with PARi under the isolation treatments (R² = 0.35, P = 0.021, Figure 4B). The biomass of C. squarrosa was negatively correlated with NPP (R² = 0.29, P = 0.039, Figure 4C) and PARi (R² = 0.27, P = 0.045, Figure 4D) under the no-isolation treatments but was only negatively correlated with PARi under the isolation treatments (R² = 0.31, P = 0.029, Figure 4D).

Nitrogen addition significantly decreased the individual seedling biomass of the two species, whereas mowing did not affect the biomass of these species. The negative responses of seedling biomass of the two species can be explained by indirect factors. Planting experiments showed that N addition increased biomass of seedling because of higher soil N availability (Ceulemans et al., 2017; Luo et al., 2020). But, these studies did not competition from surrounding plants. We found that N addition increased NPP, which was consistent with other studies conducted in situ ecosystems (DeMalach et al., 2017; Wang et al., 2017; Zhao et al., 2018). Therefore, N addition inhibits the growth of seedlings by increasing competition from surrounding vegetation (Jensen and Löf, 2017). On the one hand, N addition increase the height of the surrounding vegetation, increasing light competition (DeMalach et al., 2017). On the other hand, N addition can also increase the belowground root competition (Wang et al., 2019). Therefore, both aboveground light and belowground root competition combines to reduce biomass of seedling. Mowing can have a positive effect on seedling establishment (Bissels et al., 2006; Gibson et al., 2011). In some cases, mowing can decrease vegetation cover or NPP and increase ground light intensity (Collins et al., 1998; Gibson et al., 2011). However, we did not find a significant effect of mowing on the seedling biomass of the two species. This may stem from that the frequency of mowing was only once a year, which thus did not affect NPP and PARi.

Nitrogen addition and mowing decreased seedling survival in this experiment. Our results were inconsistent with previous studies on herbs (Jutila and Grace, 2002; Bissels et al., 2006; Zhang et al., 2018) or woody plants (Walters and Reich, 2000). This inconsistency might be explained by the different approaches used. Many previous studies have conducted planting experiments in greenhouses or fields in which the surrounding vegetation was absent (Walters and Reich, 2000; Zhang et al., 2018). However, the plots in our study were nested within a long-term N addition and mowing experiment. Both root and light competition are important factors that affect seedling survival (Gunaratne et al., 2011; Tomlinson et al., 2018; Hu and Wan, 2019). For example, nutrient enrichment can decrease seedling establishment in grassland by enhancing light asymmetry and interspecific competition (Xia and Wan, 2013; DeMalach et al., 2017). N enrichment can also increase the availability of toxic metals, which decreases seedling survival (Bobbink et al., 2010). Mowing can increase seedling establishment by removing the most productive plants and decreasing light competition (Collins et al., 1998; Gibson et al., 2011). But the decrease in seedling establishment due to mowing observed in this study may be caused by the lower soil nutrient content and soil quality after long-term clipping (Wang et al., 2020).

In our study, the performance of seedlings significantly increased in the RI treatment. These findings are consistent with previous studies showing that a low level of belowground root competition can maximize the success of seedling recruitment (Haugland and Tawfiq, 2001; Liu et al., 2013a). McConnaughay and Bazzaz (1991) suggested that root competition in the soil not only depletes water and nutrient but also creates physical barriers to root growth. The isolation of neighboring roots may, therefore, increase the physical space available for the growth of target seedling roots as well as reduce competition for other resources (Liu et al., 2013a). However, the effects of neighboring interactions on community structure differ at different phases of population growth. For example, competition associated with neighbors can accelerate seedling emergence (Dyer et al., 2008) but decrease seedling survival and biomass (Fayolle et al., 2009).

We used a simple correlation analysis to assess the relationship between seedling biomass and environmental factors. The negative relationships between seedling biomass and NPP are consistent with the results of many theoretical and empirical studies under the no-isolation treatments (Liu et al., 2007, 2013a). However, interspecific interactions can be complex (Martorell et al., 2014). Negative interspecific competition can occur when one species occupies the space required for another species to establish (e.g., mats of vegetation), and positive interactions can occur when, for example, adult plants create an optimal microclimate that facilitates the recruitment of small seeds and seedlings (Martorell et al., 2014). In our study, most species share similar ecological niches, so the relationship between species is more competitive than mutually reinforcing. Root and light competition are considered two important aspects of interspecific competition. Previous studies show that light competition is one of the main factors affecting seedling growth (Liu and Han, 2007; Fayolle et al., 2009; Liu et al., 2013a). However, other studies show that belowground root competition has been found to be more important than light competition in grasslands (Cook and Ratcliff, 1984; Haugland and Tawfiq, 2001). In our study, light competition becomes an important factor affecting seedling biomass when belowground root competition is isolation (Figure 4). Further analysis of previous studies found that the experiments that considered light competition as the main factor were mostly greenhouse experiments or planting experiments (Hautier et al., 2009), while the experiments that considered belowground root competition to be the main factor were mostly in situ experiments (Cook and Ratcliff, 1984; Haugland and Tawfiq, 2001; Wang et al., 2021c). Therefore, our study suggests that belowground root competition alters the relationship between light and seedling establishment (Graphical Abstract).