The study area is located in the Xundian County, Southwest China (25°36′42′′N, 103°13′38′′E; 2040 m a.s.l.). The study area features a low latitude plateau monsoon climate with annual mean temperature of 14°C. The dry season runs from November to April, with a clear division between dry and rainy seasons. The average precipitation during the dry season is 160 mm. The annual evapotranspiration is 2034 mm. Ci and Wu (1997) classified the study area as the sub-humid area prone to drought according to the United Nations Convention to Combat Desertification. The soil is red loam with a high clay content. The study area has previously been used for forage production and grazing without nutrient addition for approximate 30 a, which has resulted in severe degradation in grassland. The dominant native species in this area were Imperata cylindrica (L.) P. Beauv. (Gramineae), Artemisia dubia Wall. ex Besser subf. intermedia Pamp. (Compositae), Vicia sativa L. (Leguminous), Bidens pilosa L. (Compositae), Dichondra micrantha Urb. (Convolvulaceae), and Hypochaeris ciliata (Thunb.) Makino (Compositae).

In this study, we selected three 100 m×100 m sites, representing fenced grassland (FG), once tilled grassland (OTG), and twice tilled grassland (TTG). For the OTG, grass-legume mixtures were established with one tillage event at seeding, with Dactylis glomerata L., Lolium perenne L., and Trifolium repens L. at a ratio of 2:1:1 in June 2015. D. glomerata was seeded with the density of 15.0 kg/hm², whereas L. perenne and T. repens were seeded with the density of 7.5 kg/hm². For TTG, the fields were tilled at seeding and two times when used as a pea/silage corn rotation system for 2 a (2013 and 2014) prior to the establishment of grass-legume mixtures in June 2015. This pre-seeding practice primarily ensures weed control. For tillage in both OTG and TTG, the soil was ploughed to a depth of 20 cm using a moldboard plough, followed by seedbed preparation with a disc harrow and a flexible harrow. We applied fertilizers in OTG and TTG (N:P (phosphorous):K (potassium)=2:5:1) according to local agronomic recommendations.

When selecting different study sites, we ensured the homogeneity of these sites in elevation and slope. At each site, we established three sampling plots (10 m×10 m), with each plot containing three random quadrats (50 cm×50 cm) in November 2020. Quadrat was positioned at least 20 m away from the edges to mitigate edge effects. In each quadrat, plant species were determined and counted, and their aerial cover was visually assessed (Jiang et al., 2017). Shannon-Wiener diversity index was calculated as:

where P_i is the relative coverage of species i; and S is the number of species (Sun et al., 2021). The plants were categorized into three functional groups based on their characteristics: grasses, legumes, and non-N₂-fixing forbs (hereafter forbs). Table 1 shows information on the main plant species in these grassland sites.

At each grassland site, we collected 3 soil cores with a diameter of 4.5 cm from each of the three sampling plots, totaling 9 soil cores per site. These soil samples were combined for reducing errors from soil heterogeneity (Hoshino et al., 2009; Veres et al., 2015; Yang et al., 2019). The soil samples were processed by passing them through a 2-mm mesh sieve, which removed unwanted debris, litter, and stones. Next, the samples were separated into two sub-samples. One of the sub-samples was air-dried at a temperature of 25°C (±5°C) for conducting physical-chemical analysis of the soil, while a portion of the other sub-sample was immediately tested for soil MBC and MBN, and another part was stored at a temperature of -20°C for measuring dissolved nutrients.

Soil pH was determined using a glass electrode in a soil-to-water ratio of 1.0:2.5 (Dong et al., 2022b). Ammonium (NH₄⁺-N) and nitrate (NO₃^--N) were extracted from fresh soil using 2 M KCl and analyzed by an ultraviolet visible spectrophotometer (UV-1800PC, MAPADA Instrument Co., LTD., Shanghai, China) (Sorrenti et al., 2016). The Walkley and Black potassium dichromate oxidation method was employed to measure SOC (Walkley and Black, 1934), and the Dumas combustion method was employed for the analysis of soil total nitrogen (TN) concentration. An elemental analyzer (Elementar Vario EL III, Elementar, Hanau, Germany) was used to perform this analysis (Jones, 2001). Soil MBC and MBN contents were obtained by chloroform fumigation extraction (Brookes et al., 1985). Total 20 g fresh soil samples were fumigated in the dark with ethanol-free chloroform in a vacuum dryer for 24 h. Then, the fumigated and non-fumigated samples were extracted using 0.5 M K₂SO₄ for 1 h (Jones and Willett, 2006) and passed through a 0.45-μm membrane. The total dissolved nitrogen (TDN) and dissolved organic carbon (DOC) in both non-fumigated and fumigated soils were determined using the Multi N/C 3100^® analyzer (Analytik Jena, Jena, Germany) (Brookes et al., 1985). The amounts of soil MBC and MBN were counted as the gap between organic C and N from fumigated and unfumigated soil. Microbial biomass recovery of C and N was further adjusted by the factors of 0.45 and 0.54, respectively (Wu et al., 1990).

The methods described by Gamalero et al. (2003) and Dick et al. (2013) were used to estimate the activities of β-glucosidase (βG) and N-acetyl-β-D-glucosaminidase (NAG), respectively. Briefly, air-dried soil was incubated at 37°C for 1 h, and absorbance at 400 nm was determined via an enzyme labeling instrument (SpectraMax 190, Molecular Devices Co., Ltd., San Francisco, USA). Air-dried soil was utilized for measuring soil enzyme activity. This method reduces the influence of soil moisture on enzyme activity calculations compared with fresh soil, thereby enabling more precise quantification. Previous research also indicates that the air-drying pretreatment of soil has an insignificant effect on enzyme activity, compared with fresh soil (Zornoza et al., 2006). Both βG (C cycling) and NAG (N cycling) enzymes are frequently used to indicate the microbial nutrient demand, and their potential activities are associated with microbial metabolism and biogeochemical processes (Peng and Wang, 2016; Zhao et al., 2018).

For all response variables, the Shapiro-Wilks test was used to verify normally distributed residuals.

And if necessary, square root transformation was implemented. The collected data from FG and grass-legume mixture site was subjected to statistical analysis using one-way analysis of variance (ANOVA), followed by Duncan's multiple range tests with a significance level of P<0.05. Paired t tests were used to compare differences between OTG and TTG in SPSS v.27.0. To gain a better understanding of the relationship between redundant information within the dataset and the characteristics of plants and soil, we applied redundancy analysis (RDA) in Canoco v.5.0. Correlations were determined using Pearson's correlations with heatmap. All figures were generated via Origin v.2023.

Establishing grass-legume mixtures increased plant community diversity and coverage. Plant diversity increased 63.3% and 70.9% in OTG and TTG, respectively, compared with FG). In addition, the total plant coverage was significantly increased in grass-legume mixtures compared with FG. Compared with OTG, TTG increased the relative coverage of legumes by 380.3%, but significantly decreased the relative coverage of forbs by 61.2% in plant communities (Fig. 1).

The establishment of grass-legume mixtures promotes the activity of soil microbes in soil C cycling. Soil MBC concentrations in OTG and TTG increased by 16.0% and 16.4%, respectively, compared with FG. The ratio of MBC:MBN was greater in tilled grass-legume mixtures than in FG, whereas the soil MBN concentrations were higher in FG than in grass-legume mixtures (Fig. 2). The tillage treatments showed a significantly increased ratio of MBC:SOC but a decreased ratio of MBN:TN, compared with FG (Fig. 3). Compared with FG, TTG significantly decreased the activity of NAG by 72.3%, whereas βG in OTG and TTG increased by 55.9% and 27.3%, respectively, compared with FG (Fig. 4).

Figure 5 shows the N contents under different grassland treatments. Compared with FG, NO₃^--N availability was significantly greater in both OTG and TTG (Fig. 5b). However, compared with FG, soil NH₄⁺-N availability significantly decreased in OTG and TTG (Fig. 5a). Compared with FG, tillage treatments significantly decreased soil pH (Fig. 5c). Compared with FG, SOC in OTG and TTG decreased by 18.7% and 25.8%, respectively, and soil DOC decreased by 27.2% and 31.0%, respectively (Fig. 6). Whereas the greatest TDN storage level was recorded in TTG (Fig. 5d). However, no significant differences in the concentration of TN were detected between tilled grass-legume mixtures and FG (Fig. 3a).

In the RDA, two axes explained 64.6% of the total variance (Fig. 7). Tilled grass-legume mixtures (OTG and TTG) were significantly different from FG, with a greater coverage of legumes, along with higher levels of MBC, NO₃^--N, plant diversity, and βG activity, a lower grasses coverage, and decreased levels of soil pH, MBN, SOC, DOC, NH₄⁺-N, and NAG activity. MBC concentrations were positively correlated with plant diversity and βG activity, but negatively correlated with NH₄⁺-N availability, SOC, and DOC (Fig. 8). SOC were negatively correlated with MBC concentrations and βG activity. MBN concentrations were positively correlated with NH₄⁺-N, SOC, DOC, and NAG activity.

Our findings support our hypothesis that compared with FG, the establishment of grass-legume mixtures promotes plant coverage and plant community diversity, consequently improving degraded grasslands. In TTG, we observed a higher proportion of legumes and a decreased proportion of forbs, compared with OTG. This is mainly attributed to the disruptive effects of intensive tillage on the asexual reproductive structures and seed bank of native weedy grasses present in the soil (Haring and Flessner, 2018). Our results showed that establishing grass-legume mixtures increases soil N availability, creating favorable conditions for neighboring plant species and thereby enhancing plant diversity. A study conducted in arid areas on grass-legume mixtures also confirmed this finding (Sun et al., 2023). Tillage effectively creates ecological niches for seeded plant species, reducing competitive exclusion from dominant species and resulting in greater species diversity in tilled grasslands. Moreover, we noted a notable decline in the proportion of grasses within grass-legume mixtures, compared with FG. This decline was mainly attributed to the dominance of the native grass species, I. cylindrica, characterized by spiky and irritating leaves and having poor nutritional content, making it unpalatable for livestock (Katoch, 2022). As an alternative to N fertilization, establishing grass-legume mixtures in degraded grasslands, coupled with appropriate management strategies, can be an effective and practical restoration method that enhances grassland productivity while preserving plant species diversity. Our findings provide a foundation for restoration management in degraded grasslands. Implementing grass-legume mixtures can rehabilitate degraded grasslands by increasing the coverage of highly palatable and high-quality forage.

Although the use of tillage during the establishment of grass-legume mixtures facilitated legume stability and weed management, tilled grasslands had significant lower SOC and DOC compared with non-tilled areas. Contrary to the findings of two studies conducted in semi-arid areas (Li et al., 2016; Singh et al., 2023), our research indicates that the establishment of grass-legume mixtures does not benefit SOC storage. We believe this negative result is due to the increased βG enzyme activity and MBC. The decomposition of legume residues adds N to the soil, which can boost microbial growth and the creation of extracellular enzymes. This process may activate a priming effect, causing increased C and N losses in specific areas (Ramirez et al., 2010). RDA results indicate that in grass-legume mixtures, the rise in soil N availability increases the activity of βG. The positive correlation between βG and MBC suggests that greater enzyme activity accelerates soil C mineralization, resulting in the loss of SOC. Soil microbial organisms play a crucial role in maintaining soil ecosystem services. The activity of soil microbial communities regulates nutrient turnover, transport, and the rate of decomposition of SOC (Tarafdar, 2022). Our results showed that tillage breaks apart SOC protection and aerates soil, increasing the availability of substrates that boost microbial activity and raise the proportion of MBC in SOC. The close association between the increase in MBC and the rapid decline in SOC suggests that MBC is a key driver of C storage potential variation in soil (Dong et al., 2022a). Another study in southwest China discovered a link between total soil MBC and SOC decomposition (Piao et al., 2001). Soil microorganisms can efficiently use lower C stocks and promote soil C mineralization in grasslands that were established with moderate tillage. In addition, tillage often results in a decreased nutrient availability and a rapid decomposition of SOC (Zhang et al., 2018; Parajuli et al., 2021). Tillage disrupts soil aggregates, exposing organic matter that was previously protected, and enhances microbial respiration, thus further accelerating the loss of soil organic matter (Thapa et al., 2023). Previous research has indicated that soil nutrients can influence soil microbial biomass and activity (Sistla et al., 2012). Higher soil TDN concentrations in grass-legume mixtures with more frequent historical tillage suggest that leguminous plants within the mixture increase soil TDN by fixing N in their root nodules (Becana and Sprent, 1987). Higher levels of soil N can enhance microbial activity and accelerate the decomposition of organic matter, leading to lower levels of SOC (Sistla et al., 2012). Additionally, we found that grass-legume mixtures significantly reduced soil pH and increased the concentration of NH₄⁺-N in the soil. And there was a significant negative correlation between MBC and soil NH₄⁺-N, SOC, and soil DOC. N fixation by legumes induces soil acidification, shifting the plant's preference for absorbing inorganic N forms towards favoring NH₄⁺ (Li et al., 2013b). This disturbs the NH₄⁺ and NO₃^- balance in the soil, causing a decline in soil NH₄⁺. Previous research has shown that an increase in soil NH₄⁺-N availability has been shown to inhibit soil microorganism activity and reduce soil MBC (Li et al., 2013a). Hence, reduced NH₄⁺ levels in grass-legume mixtures foster MBC accumulation. This could also be a contributing factor to the loss of SOC.

Our results also revealed significant correlations between changes in plant diversity with MBC, βG enzyme activity, and soil MBN. Indeed, a greater microbial biomass is associated with species-rich plant communities (Chen et al., 2019). Our findings support the proposed positive relationships between plant communities and MBC (Eskelinen et al., 2009; Grigulis et al., 2013; Chen et al., 2019; Song et al., 2019; Moreno et al., 2021). Nutrient uptake, litter inputs, root turnover, and root exudates from diverse plant species can alter soil food web community structure, microbial abundances, and soil decomposition processes (Bing et al., 2016). According to the results of a meta-analysis across global terrestrial ecosystems, plant species mixtures typically increase the fungal to bacterial biomass ratio, soil respiration, and soil microbial biomass levels (Chen et al., 2019). We propose that the positive impacts of increased plant diversity on MBC and βG enzyme activity are primarily due to the diverse composition of plant communities in grass-legume mixtures (Tabel 1), which offers a favorable habitat for microorganism proliferation (Jangid et al., 2011; Li et al., 2012; Zhou et al., 2019b; Lama et al., 2020).

MBN content and NAG activity in tilled areas were lower than those in non-tillage grassland. According to Wutzler et al. (2017), resource heterogeneity results in variations in extracellular enzyme activity of the soil, influencing the patterns of C and N distribution in different grasslands. Our findings indicate positive correlations among MBN, SOC and DOC contens, and NAG activity. Soil decomposers may produce more N-acquiring enzymes to access available resources when soil C is abundant (Tiemann et al., 2015). The changes in βG and NAG enzyme activities caused by this resource heterogeneity can be explained by resource allocation theory (Xiao et al., 2018). Compared with grass-legume mixtures, low N availability in fenced grasslands prompts soil microbes to secrete NAG and to decompose organic matter, thus obtaining N (Craine et al., 2007). The significant increase in MBN in fenced grasslands also supports this.

Conversion of native prairie land for forage production has led to significant loss and redistribution of SOC, resulting in declining soil fertility in low-productivity semi-arid agroecosystems (Zhang et al., 2018; Parajuli et al., 2021). Our study indicates that moderate tillage can support legume establishment and increase soil nitrogen, although the associated soil disturbance enhances microbial activity and organic matter decomposition. As Hurisso et al. (2013) reported, grass-legume mixtures have substantial SOC storage potential compared with no-till wheat-fallow practices in semi-arid soils. Therefore, it is essential to continuously investigate the impact of tillage-established legume-grass mixtures on SOC storage and the mechanisms behind carbonate formation in these low-input semi-arid grasslands.