The research area is located in the central and western part of Zoige County on the northeast edge of the Qinghai-Tibet Plateau (33°43′27″–33°51′43″ N, 102°25′40″–102°33′34″ E), China, which is characterized by arid and semi-arid transitions with a cold alpine climate. Annual precipitation fluctuates between 600 and 750 mm (average 656.8 mm), 90% of which occurs from April to October. The annual mean temperature varies in the range of 0.6°C–1.2°C, with monthly averages of −10.8°C in January and 10.9°C in July. The geological landform is a hilly plateau formed by the intense uplift of the Himalayan tectonic movement and neotectonic movement. The elevation varies from 3,400 to 3,450 m above sea level, and the main water systems are the White River and the Black River (tributaries of the Yellow River). It is interspersed with a subalpine region, river valley plains, and several lake depressions. Subalpine meadows, wet meadows, and marshes dominate the main landscape. The soils in this study area were classified as Cambic Coarsic Leptosols in the World Reference Base for Soil Resources (IUSS Working Group WRB, 2022).

Subalpine meadow grasslands have degraded universally by natural and human factors over 20 years ago. The desertified grasslands were widely distributed in Xiaman Town (33°43′27.3″–33°46′13.6″ N, 102°25′35.3″–102°32′45.1″ E), Maixi Town (33°51′35.9″–33°51′43.8″ N, 102°32′56.7″–102°33′34.7″ E), and Axi Town (33°40′59.7″–33°41′4.6″ N, 102°55′57.8″–102°56′4.2″ E) in Zoige County. The degradation characteristics of grasslands were mainly embodied in decrease of vegetation coverage, biomass, and herbage edibleness. To restore degraded grasslands, a series of restoration measures was carried out by Zoige Forestry and Grassland Administration, China, such as planting Avena sativa, Poa pratensis, Elymus nutans, and Tamarix ramosissima artificially. Three primary ecological restoration patterns were carried out in each town respectively, including enclosure plus grass pattern (EG), enclosure intercropping shrub-grass pattern (ESG), and enclosure plus sand-barrier and shrub-grass pattern (ESSG). Each pattern has three restoration periods of ≤5 years, 7–9 years, and ≥15 years. We have not adopted anthropogenic management measurements such as mowing return and applying fertilization during grassland restoration.

We set 10 sampling sites, including three restoration patterns with three restoration years, and a control site with unrestored desertified grassland (CK) in each subarea ( Figure 1 ). Then, we selected three sampling quadrats (each 2 m × 2 m) in each sampling site based on the theory of biological replicates. The plant community characteristics (e.g., functional groups, density, and coverage) and biomass were investigated, and we obtained plant samples following the method described in Zhang and Zhao (2015) ( Table 1 ). The soils (0–10, 10–20, and 20–40 cm) were sampled from five random locations at each quadrat, and 90 samples were collected totally. After removing litter and root material, all soil samples from the same plot were mixed and diminished to 800 g approximately. A part of the soil samples was stored in ice bags and reserved at 4°C for measurement of microbial indicators, and the rest of the soil samples were air-dried to determine the physicochemical indicators. Soil bulk density was measured by the cutting ring method simultaneously.

Soil particle composition was determined by the simple specific gravity method using Stokes’ law (Lu, 2000). The total organic carbon (TOC) content was determined by the Walkley–Black potassium dichromate sulfuric acid oxidation procedure (Nelson and Sommers, 1983). The permanganate-oxidizable carbon content was determined following the protocol proposed by Culman et al. (2012). The dissolved organic carbon content was determined by using Vario TOC analyzer (Elementar, Germany) after extracting with deionized water (Chen et al., 2003). The TN content was determined by the Kjeldahl method. After 2 mol L⁻¹ KCl extraction, ammonium nitrogen (NH₄ ⁺) was determined by indophenol blue colorimetry, and nitrate nitrogen (NO₃ ^-) was determined by UV spectrophotometry colorimetry (Lu, 2000). The dissolved organic nitrogen (DON) content was calculated as the difference between TN and TIN (the sum of NO₃ ^–N and NH₄ ⁺–N) (Mariano et al., 2016). The total P and available P contents were measured using the colorimetric molybdenum blue method via NaOH fusion and 0.5 mol L⁻¹ NaHCO₃ (pH = 8.5) extraction, respectively (Lu, 2000). The total potassium (TK) and available potassium (AK) contents were determined using the flame photometer method after digesting in a nickel crucible with NaOH at 750°C and extracted by 1 mol L⁻¹ CH₃COONH₄ (pH = 7.0) respectively (Lu, 2000). The number of cultivable bacteria, fungi, and actinomycetes was determined using the dilution plate method (Hou et al., 2014), which were expressed as colony-forming units per gram of soil. The bacteria were incubated in beef protein medium, and the fungi and actinomycetes were cultured in Martin medium and Gauze’s synthetic no. 1 medium (Shamiyeh and Johnson, 1973). After soil sampling was performed using the chloroform fumigation–extraction method, Vario TOC cube (Elementar, Germany) analyzer was used to determine the MBC and microbial biomass nitrogen, and MBP was determined by the colorimetric molybdenum blue method (Brookes et al., 1985; Vance et al., 1987). The urease activity was measured by the phenol-sodium hypochlorite method using urea as substrate, which was expressed as milligrams of NH₄ ⁺ per gram of dry soil per hour after incubation at 37°C for 24 h (Samborska et al., 2004). The proteinase activity was determined by the Folin-Ciocalteau reagent and expressed in milligrams of tyrosine per gram of dry soil per hour produced by the incubation of sodium caseinate at 50°C for 2 h (Weber and Tihanyi, 1994). The sucrase activity was measured by the 3,5-dinitrosalicylic acid colorimetric method, expressed as the amount of glucose released per gram per hour of soil sample in 24 h (Frankeberger and Johanson, 1983). The amylase activity was determined by the dinitrosalicylic acid colorimetric method and expressed as the milligram of maltose in 1 g soil after 24 h (Zhang et al., 2011). Both neutral and alkaline phosphatase activities were determined by the disodium 4-nitrophenylphosphate colorimetric method, which were expressed by the release of 4-nitrophenol per gram of soil samples after incubation at 37°C for 1 h from disodium 4-nitrophenylphosphate (Guan, 1986; Tabatabai, 1994).

PCA was also used to calculate the weight values of indicators, including vegetation coverage, density, and biomass as well as silt, TN, carbon–nitrogen ratio (C/N), TK, MBC, MBP, and fungi in soils. The calculation method of indicator weights was similar to that of soil property weights in soil quality assessment using Eqs. (1) and (2). The degrees of membership of the above-mentioned indicators were calculated using Eq. (3) because these indicators were in accordance with the function of “more is better” (Mamehpour et al., 2021). The ecological restoration effect index (EREI) of restoration measurements was calculated according to Eq. (4).

The experimental data analyses were performed using SPSS 19.0 (IBM Corp., US) (i.e., maximum, minimum, mean, and standard error). One-way analysis of variance, followed by the Duncan test, was applied to test differences of soil quality indicators among depths, restoration patterns, and years (p< 0.05). The correlation coefficients among soil microbial and physicochemical indicators were analyzed via Pearson correlation analysis. The figures were drawn by using Origin 2022b (Origin Lab Corp., Northampton, MA, USA).

The soil quality indicators for unrestored grassland (CK) had no significant difference in three subareas (p > 0.05) ( Supplementary Table S1 ), and they had no significant differences except bacteria and alkaline phosphatase among depths in the alpine grassland (p > 0.05) ( Supplementary Table S2 ). Accordingly, the weighted average of soil quality indicators from different layers was performed for the same quadrat in this research.

As shown in Table 2 , the values of soil quality indicators, including physical and chemical properties in restored grasslands, were preferable to those under CK treatment at each restoration year. The soil bulk density and sand content among different ecological restoration patterns were lower than that in unrestored grasslands, and they showed a decreasing trend with increasing restoration years. However, the contents of soil silt, clay, TOC, TN, DON, NH₄ ⁺–N, and NO₃ ^–N in restored grasslands were significantly higher than those in unrestored grasslands (p< 0.05), and they generally significantly increased with increasing restoration years (p< 0.05). Moreover, the contents of soil silt and clay in ESSG were significantly higher than in EG and ESG. Meanwhile, our previous results showed that the soil microbial properties, including microbial abundance, microbial biomass, and enzyme activity, in the three patterns of restored grasslands significantly improved (p< 0.05) compared with unrestored grasslands ( Supplementary Table S4 ) (Hu et al., 2018a). The improvement effects of the microbial properties enhanced with extended restoration years at each pattern, and those in ESSG were generally better than EG and ESG ( Supplementary Table S4 ) (Hu et al., 2018a).

Table 3 indicates that different principal components (PCs) were selected with eigenvalues ≥1 for soil physical, chemical, and microbial properties. PC1 explained 82.39% of the total variance in soil physical properties including four indicators, and soil silt was selected as the appropriate indicator representing physical properties ( Table 3 ). In terms of chemical properties encompassing 12 indicators, three PCs explained 79.23% of the total variance, and PC1, PC2, and PC3 explained 59.48%, 11.18%, and 8.57% respectively. Combined with the correlation coefficients of soil chemical indicators, we chose TN, C/N, and TK in the MDS, representing chemical properties on account of their higher loadings and norm values ( Table 3 ; Supplementary Table S3 ). Concerning microbial properties involving 12 indicators, two PCs explained 79.42% of the total variance, and PC1 and PC2 explained 71.06% and 8.36% of the variation, respectively. MBC, MBP, and fungi were contained in the MDS because of higher loadings and norm values ( Table 3 ; Supplementary Table S4 ). As a result, sensitive and important indicators comprising soil silt, TN, C/N, TK, MBC, MBP, and fungi were selected to establish the MDS of alpine grassland soil quality.

As shown in Table 4 , the weights of soil physical, chemical, and microbial properties were almost equal. The scores of soil properties increased with increasing restoration years. Figure 2 shows that the value of SQI varied from 0.102 to 0.671. The SQI in ESSG was higher than EG and ESG overall ( Figure 2 ). The SQI in each restoration pattern increased significantly with increasing restoration years. The SQI of ≤5 years, 7–9 years, and ≥15 years increased by 144.83–249.14%, 319.63–463.10%, and 506.28–561.24%, respectively, compared with CK (p< 0.05) ( Figure 2 ). Moreover, the increasing rate of SQI began to reduce after 5 years in EG and ESG ( Figure 2 ), whereas the increasing rate of SQI between ≤5 years and 7–9 years improved by 120.13% compared with ≤5 years in ESSG, which started to decrease after 7–9 years (p< 0.05) ( Figure 2 ).

The value of EREI ranged from 0.104 to 0.547 ( Figure 3 ). There was no significant difference among the three restoration patterns (p > 0.05), but EREI in ESSG was generally higher than that in EG and ESG, especially before 7–9 years ( Figure 3 ). EREI increased constantly with increasing restoration years, of which 7–9 years and ≥15 years significantly increased by 6.93%–107.91% and 93.23%–178.89% respectively, compared with 0–5 years (p< 0.05) ( Figure 3 ). In addition, the increasing rate of EREI in EG was lower than ESSG. Moreover, the increasing rate of EREI in ESG was lower than ESSG from 5 to 7–9 years, while that in ESG was higher than ESSG after 7–9 years ( Figure 3 ).

In this research, soil silt, TN, C/N, TK, MBC, and MBP as well as fungi have been selected as soil quality indicators, and they were associated with vegetation growth during grassland restoration. The changes of soil silt content under the application of restoration measurements were the most significant compared with that of sand and clay, similar to the finding of He et al. (2021), which was ascribed to the fact that vegetation restoration could weaken wind erosion, thereby reducing the loss of fine particles (Zhang et al., 2022). The decomposition of plant litterfall increased the soil organic matter under the conditions of planting herbs and/or shrubs plus grazing prohibition in desertified grassland (Yu et al., 2018). Thus, litterfall was enhanced with an increase of aboveground biomass in restored grasslands ( Table 1 ), which induced the changes of C/N and TN contents in soils. The potassium (K) requirement for plant growth greatly increased in the middle and late stage of restoration, which could promote the conversion of soil mineral K to AK via the mobilization of organic acids and enzymes from K-solubilizing bacteria, and mineral K was the main form of TK in soils (Zhang et al., 2022). The increase of litter input led to the improvement in soil water storage capacity, which was conducive to the reproduction of microorganisms together and further increased the fungi abundance as well as the contents of MBC and MBP in soils (Liu et al., 2017; Zhang et al., 2021). In recent years, some studies have also established the MDS of soil quality assessment based on physicochemical properties in temperate grassland ecosystems (Wang et al., 2021b; Li et al., 2023b). Nevertheless, our results found that the weights of soil microbial properties were almost equal to the soil physicochemical properties and that the soil microbial properties had positive and significant effects on TN ( Supplementary Figure S1 ), indicating that microbial properties had non-negligible significance to soil quality in this study area. Therefore, microbial properties should be considered to assess the soil quality of alpine grassland under restoration measurements.

Different restoration patterns and years had various improvement effects on soil assessment indicators. As far as soil physical properties were concerned, silt content increased, which was similar to the findings of Hu et al. (2022) and can be attributed to the fact that vegetation restoration effectively resisted soil sand outside and promoted the optimization of soil texture (Zhang and Zhao, 2015). Soil structure and texture improved by applying restoration measurements, which may further lead to an increase in soil nutrients (Hu et al., 2022). In this study, the soil nutrient contents of the restored grassland were also significantly higher than those of the unrestored grassland, particularly the significant increase of TN content, similar to the implementation effects of restoration measurements in Shenza County in the Tibetan Plateau as reported by Zhang et al. (2022). This was mainly because the humification of litterfall could directly increase the soil nutrient content, and the improvement of physical properties (e.g., texture, structure, and water retention) might reduce the mineralization of organic matter (Kooch et al., 2022). Previous studies revealed that the vegetation biomass had positive relationships with soil microbial biomass generally (Zhang and Zhao, 2015; Yang et al., 2022). The increase of carbon and nitrogen sources of microorganisms further enhanced the soil microbial biomass (Yu et al., 2018). Therefore, the contents of MBC and MBP as well as the quantity of fungi improved significantly, which was caused by the increase of vegetation biomass together with soil carbon and nitrogen contents in each restoration pattern ( Tables 1 , 2 ).

In general, soil quality indicators are gradually optimized due to increased vegetation coverage, density, and biomass (Zhang et al., 2022; Wan et al., 2023). In this research, the vegetation coverage, density, and biomass of restored grassland increased ( Table 1 ), thus inducing soil quality indicators gradually ameliorated with the increasing restoration years, which was similar to the findings of Li et al. (2023c). Furthermore, the contents of slit and MBC in ESSG were significantly higher than EG ( Table 2 , Supplementary Table S4 ). These results can be explained by the combined effects of sand barriers and vegetation that were beneficial to sand fixation and water conservation, further effectively improving the soil quality (Wang et al., 2021a; Hu et al., 2022).

Vegetation biomass had significant positive relationships with SQI and EREI (R ² > 0.6, p< 0.05), indicating that soil quality and ecological benefits improved with the increase of biomass in alpine grasslands. We found that the restoration measurements improved the soil quality and ecological benefits of desertified grassland ( Figures 2 , 3 ), which was consistent with the findings of positive effects of the fenced enclosure on the soil quality of sandy grassland in the Tibetan Plateau as reported by Zhang and Zhao (2015); Hu et al. (2022), and Zhang et al. (2022). The main reason that was attributed to is that restored vegetation and/or sand barriers could be able to separate the sand sources and resist wind erosion effectively in areas with strong winds (Wan et al., 2023), which was beneficial for soil and water conservation as well as nutrient retention in desertified grassland. Moreover, plants grow preferably based on superior conditions of soil water and nutrients. On the one hand, vigorous plants could retain water and return organic matter via litterfall (Yu et al., 2018). On the other hand, the decomposition and humification of dead root and root exudates could promote soil quality (Wu et al., 2023). Therefore, the soil–vegetation system could form positive mutual feedback in restored grassland, further improving soil quality and restoring ecological service functions.

Soil quality and ecological benefits in ESSG were optimal among three restoration patterns because setting sand barriers was of great significance to prevent wind and fix sand as well as reserve moisture in sandy soils (Liu et al., 2019). Particularly, shrubs with deep roots promoted the accumulation of organic matter and nutrients in shallow soils combined with the roles of sand barriers (Kidron and Gutschick, 2013; Wang et al., 2021a), which could strengthen the positive mutual feedback between vegetation and soil. Similarly, Scotton (2019) also reported that combining sand fixation barriers and phytoremediation was a good way to restore the ecology of desertified grasslands.

Vegetation is an important and positive driving force for restoring desertified grasslands, as Wu et al. (2023) have reported. Furthermore, the significant improvement of vegetation (e.g., biomass) with the increase of restoration years was beneficial to restore ecological function directly and indirectly in desertified grasslands (Wang et al., 2023). As the restoration years of degraded grasslands in our study increased, the aboveground and root biomass of vegetation also gradually improved ( Table 1 ). This was conductive to enhance the abilities of retaining water and nutrients and increase litter return and root exudates, which can also further strengthen the positive feedback between soil and vegetation. Thus, soil quality and ecological benefits increased significantly, corresponding to the increasing restoration years in each restoration pattern.

The increasing rate of SQI in EG and ESG as well as EREI in EG began to decrease after 5 years of the application of restoration measurement, indicating that short-term fencing was more beneficial than long-term fencing. A probable explanation for this outcome was that the increasing requirement of plant nutrients slowed down the amount of nutrient accumulation and optimization of microbial characteristics in soils under the circumstances of long-term grazing exclusion (Li et al., 2023c). A similar result has been shown in the study of Sun et al. (2020) and Zhang et al. (2022), respectively, such that longer-term grazing exclusion had a little effect on vegetation growth without the construction of sand prevention belts. Interestingly, the increasing rate of SQI and EREI in ESSG was reduced after 7–9 years. This may be ascribed to the fact that the decay of sand barriers combined with a large amount of requirement of plant nutrients could also slow down the amelioration of soil features after 7–9 years based on the result of Liang et al. (2023). Thus, it was hard to ensure that the soil–vegetation system was constantly a virtuous cycle with the increase of restoration years.

In addition, plant species diversity appeared to decrease along with increased restoration years ( Table 1 ) because the growth of some plants may have been inhibited during grassland restoration (Li et al., 2023c). This would cause a weakening in the stability of ecosystems, according to a research reported by Zhou et al. (2017). As a result, combined with the increasing rate of SQI, EREI, and plant diversity, we suggest that the appropriate time of anthropogenic monitoring and management measurements, such as applying organic fertilizer in soils and mowing return, should be reasonably considered at the beginning of 7 years in ESSG and 5 years in the other two patterns to maintain sustainable ecological benefits.