The experimental area is located in Jun Ranch of Sanjiangyuan Alpine Meadow Research Observatory (Jun Ranch, Machin County, Golog Tibetan Autonomous Prefecture, Qinghai Province, 34°22′-34°20′N, 100°30′-100°29′E), at an altitude of 3,783.5 m. It is a continental semihumid climate type on the Qinghai-Tibet Plateau, with an average annual temperature of -5.6–3.8°C, with an average temperature of −12.6°C in January and 9.7°C in July. The average annual precipitation is 500–600 mm, and the annual evaporation is relatively large, generally 730–1700 mm. There is no absolute frost-free period, and the forage grows for approximately 150 days. The soil type is alpine meadow soil, and the grass type is alpine meadow. The vegetation types are Kobresia pygmaea C. B. Clarke and K. humilis Sergievskaya. and K. capillifolia C. B. Clarke as the dominant species and Poa crymophila Keng, Carex atrofusca Schkuhr subsp, and Polygonum viviparum L as companion species in alpine meadows. Due to climate change, human activities and rodent feeding activities, the alpine meadow has been degraded to alpine meadow with Aconitum pendulum busch, Morina chinensis Diels, Pedicularis kansuensis Maxim and Ligularia virgaurea Mattf, and Artemisia hedinii Ostenf. et Pauls, were the dominant severely degraded meadows.

In August 2020, an experimental plot was selected in the Military Pasture of Sanjiangyuan Alpine Meadow Research and Observation Station. According to the ‘Grading Standard of Natural Grassland Degradation, Desertification and Salinization’ (GB19377-2003) and the classification standard of Ma et al. (2002) (Table 1), the method and principle of temporal succession were replaced by spatial sequence. The meadows with basically the same climate and soil conditions in the study area were divided into slightly degraded alpine meadow (D1), moderately degraded alpine meadow (D2), highly degraded alpine meadow (D3) and severely degraded alpine meadow (D4) (She et al., 2021) (Table 2). The plots with different recovery years were compared with the severely degraded meadow, and the native vegetation was sparse in the severely degraded stage. The vegetation was low or dead, the soil surface was exposed in patches, poisonous weeds were overgrown and spread, and the grazing value basically disappeared. Before the establishment of artificial grassland, it was a typical severely degraded grassland. The dominant species of Cyperaceae were replaced by poisonous weeds, and the coverage of native vegetation was less than 10%. The vegetation is composed of A. pendulum, Potentilla chinensis, Ligularia virgaurea, P. resupinata, and Ajania tenuifolia. In 2019, 2015 and 2010, perennial artificial grasslands dominated by Elymus nutans, Festuca sinensis and P. crymophila were replanted on severely degraded grassland. The seeds of Elymus nutans, Festuca chinensis and P. crymophila used in the artificial grassland were produced by the local grass seed breeding farm with a sowing rate of 37.5 kg/hm² and a sowing rate of 3:1:1, and the amount of fertilizer applied was 45 kg/hm² (diammonium hydrogen phosphate compound fertilizer). The sowing distance was 20 cm, and the agronomic measures were as follows: deep turning-raking-fertilizing-sowing-covering-soil-compacting-fencing. Each sample plot was 20 hm², and each treatment had 5 replicates. The management measures after planting were fencing and culturing, without livestock grazing disturbance factors. In August 2020, severely degraded alpine meadow (D4), artificial grassland established in 2019 (1-year-old artificial grassland, 1AG, R1), artificial grassland established in 2015 (5-year-old artificial grassland, 5AG, R5), and artificial grassland established in 2010 (10-year-old artificial grassland, 10AG, R10) were tested (Figure 1; Table 2).

Five randomly set 50 cm × 50 cm samples were set in each sample plot for the vegetation community survey. To avoid edge effects, the interval between each sample square should not be less than 10 m. The vegetation sample survey mainly included the survey of overall community characteristics, including meadow health, dominant species, number of species, total cover, average plant height (multiple replicates to take the average value), and the survey of species characteristics, including the subcoverage of each species, number of plants (multiple degrees), plant height and so on. The average height of the community in the sample square was recorded with a tape measure, and then the total cover of the community in the sample square was measured by visual inspection, followed by measuring the height of a single species in the sample square with a tape measure (5 replicates, less than 5 plants were all measured), measured the cover of a single species by visual estimation, and calculating the multiplicity of species. Afterwards, the plants in the sample plots were cut in flush ground and put into envelopes, brought back to the laboratory and dried at 65°C for constant weight, then weighed and recorded.

After taking the plant samples, the underground parts of the plants were collected in the sample cubes using the soil column method (Φ = 50 mm). Three augers were taken for each sample cube, and each auger was divided into 3 layers of 10 cm soil columns each (0–10, 10–20, 20–30 cm), and the soil samples were sieved with a mesh of 5 mesh. After rinsing the soil with adequate water, the live roots were picked out according to the flexibility of the root system and the color of the cross-section. Finally, washed live root systems were put into large envelopes, dried in an oven at 65°C for constant weight and then weighed. Soil samples from the 0–30 cm soil layers were passed through a 2 mm sieves to remove debris and used for soil property determination. Soil microorganisms were obtained by the X-type five-point method, and soil samples 0–10 cm were taken. All samples were collected in sterilized self-sealing bags, stored frozen and shipped back to the laboratory in time. Ten grams of fresh soil was stored at −80°C in the refrigerator for DNA extraction and high-throughput sequencing. Sample sequencing was performed at Majorbio Cloud platform (http://www.majorbio.com/).

The alpha diversity of bacterial and fungal communities showed opposite trends between degradation and recovery processes (Table 4). The Shannon and Chao indices of bacteria and fungi increased significantly with increasing degradation, the Chao indices of fungi were in the order of D4 (697.1) > D3 (660.5) > D2 (643.3) > D1 (443.7). In contrast, the fungal Shannon and Chao indices of the recovery process were significantly lower (p < 0.05) from R5 to R10. The bacterial Chao index showed a similar trend, while the bacterial Shannon index did not differ significantly between the different recovery years. The Simpson diversity index of bacteria was significantly different (p < 0.05) between the highly and severely degraded stages of degradation, while there was no significant difference between the recovery years, and the Simpson diversity index of fungi was not significantly different between degradation stages and recovery years. The ACE diversity index of bacterial and fungal communities showed no significant patterns during either degradation or recovery.

Indicators incorporated in ecosystem multifunctionality are all quantitative indicators of soil nutrient cycling, nutrient pool accumulation and good above- and belowground productivity, where nitrogen and phosphorus are key factors for ecosystem production and biomass are proxies for good grassland productivity. In light and moderate degradation stages, alpine meadow degradation has a positive plant‒soil feedback effect on ecosystem multifunctionality and reaches its maximum at moderate degradation; in the high and severe degradation stages, alpine meadow degradation has a negative plant‒soil feedback effect on ecosystem multifunctionality and reaches its maximum by severe degradation. In different stages of alpine meadow degradation, the plant‒soil feedback on ecosystem multifunctionality changed to negative feedback after reaching the maximum in moderate degradation. Negative plant‒soil feedback of alpine meadow degradation on biological factors was observed in mild degradation, and positive plant‒soil feedback was observed from moderate degradation to severe degradation, which showed a tendency to increase with the degradation degree. In the early stage of the degradation process, alpine meadow degradation has a negative plant‒soil feedback effect on abiotic factors; the heavy and severe degradation stages, alpine meadow degradation has a positive plant‒soil feedback effect on abiotic factors, and it shows a trend of increasing gradually. It can be seen that at the stage of severe degradation, planted artificial meadow has a negative plant‒soil feedback effect on ecosystem multifunctionality, and at 1 and 5 years of recovery, planted artificial meadow has a positive plant‒soil feedback effect on ecosystem multifunctionality, and it reaches the maximum at 1 year of recovery. During the recovery process, plant‒soil feedback on ecosystem multifunctionality changed to negative feedback after reaching a maximum at 1 year of recovery. For severe degradation, the constructed artificial grassland had a positive plant‒soil feedback effect on biotic factors, negative feedback at 1 year of recovery, positive feedback at 5 years of recovery, and negative feedback at 10 years of recovery, indicating that aboveground community diversity and belowground microbial community diversity were richer at 5 years of artificial grassland than at 1 and 10 years of recovery, and the change pattern of the feedback effect of abiotic factors was consistent (Figure 10).

The species diversity index showed an overall decreasing trend from D1 to D4, but a maximum occurred at D3, which is essentially consistent with the findings of this study (Chen et al., 2016; Zhang et al., 2019). The reason for the maximum species diversity in D3 is mainly because the D1 and D2 meadow ecosystems have strong stability, while with the accelerated process of degradation, edible forage grasses decrease, and species such as Stellera chamaejasme, Aconitum pendulum and L. virgaurea begin to occupy the dominant species ecological niche in the community. The increase in the number of species within the community will cause changes in plant community structure and function and will influence the high level of species diversity within the community (Zhou et al., 2012). Disturbances from human activities during the process from D1 to D4 can also have an impact, such as grazing and the survival of rodents in the meadow, cattle, sheep and rodents with foraging preferences, trampling of the meadow during foraging and acts such as seed dispersal as a vector during foraging, which can result in a higher number of species in D3 than in D2 (Wang et al., 2014; Zhang et al., 2021). The species richness and Shannon‒Wiener diversity index during meadow degradation showed a unimodal change, with the maximum value being present at D2 (Hao et al., 2020). This study showed no significant difference in species evenness and Simpson’s diversity index from D1 to D4, suggesting that anthropogenic disturbance inhibited the dominant species in the community, which provided opportunities for the invasion and expansion of toxic weeds and enhanced the competitiveness of these species to dominate the community, resulting in an increase in the species diversity index, thus indicating that moderate disturbance would improve the species diversity within the community (Zhou et al., 2005; Zhou et al., 2012). The community composition is gradually homogenized, and the species richness and Shannon Weiner diversity index decreased significantly (Zhou, 2001; Jianli et al., 2004; Yuan et al., 2004).

The establishment of artificial grasslands at the stage of severe degradation has a positive effect on long-term recovery of artificial grassland ecosystems (Shang et al., 2018). The establishment of artificial grassland not only improves the productivity of the ecosystem but also corresponds to the introduction of new vegetation and the acquisition of organic matter contributed to the soil system by plant residues and dead leaves, thus improving the structure and function of the soil (Zi et al., 2017; Xing et al., 2020). In this study, plant aboveground biomass showed a decreasing trend with the extension of recovery years, and in the early stage of recovery, competition between species was low, and plants of the grass family were larger and soon occupied a higher ecological position and dominated the whole community, so aboveground biomass was higher (Li et al., 2021). After 5 years of recovery, sedge family plants with low resource requirements gradually increased, and a small number of sedge family plants appeared during the gradual succession of artificial grassland, indicating that the artificial grassland entered the natural recovery process with low weed-like biomass (Zhang et al., 2003). After 10 years of recovery, soil fertility declined, and when nutrients in the soil were insufficient, plants tended to transfer resources to the underground root system to improve the uptake and utilization of limited soil resources, the dominant species status of Gramineae declined, and species dominated by Asteraceae began to be in subdominant species status (Zhang et al., 2012). With the increase in recovery years, all four species diversity indices showed a “V” shape change, which was due to the strong competition of planted species of artificial grassland at the early stage of recovery and their rapid dominance, making the species diversity index decrease; however, with the extension of recovery years, the native species gradually invaded and gained dominance, and the species diversity increased significantly (Wu et al., 2019).

Soil available potassium, total nitrogen and total potassium showed a decreasing trend as the degradation level increased. There was no significant change in the distribution of soil total potassium content among soil layers in the meadow samples with different degradation levels (Gao et al., 2006). In the same soil layer of different alpine meadow degradation stages, the soil total potassium content varied significantly, and in different soil layers of the same alpine meadow degradation stages, the total potassium content in the same soil layer did not vary significantly (Chen and Zhang, 2013). Soil available nitrogen, available phosphorus and total phosphorus showed an overall decreasing trend but differed in that a small increase was observed at a specific stage. Soil available nitrogen showed a small increase at the stage of severe degradation, and soil available phosphorus showed a small increase at the stage of mild degradation, while soil total phosphorus showed a small increase at the stage of severe degradation. With the intensification of soil degradation, the total phosphorus content of the soil showed a decreasing trend and then an increasing trend, while the available phosphorus content showed a decreasing trend (Zhao et al., 2009). In alpine meadows of the Tibetan Plateau, mineralization of organic nitrogen is the main source of available nitrogen in the soil (Gao, 2007). Plant uptake and natural factors reduce the content of available nitrogen, and effective nitrogen has a small rebound in the process of severe degradation, which may be because the level of severe degradation plant activity is not high, so the amount of uptake is not high, resulting in a small rebound phenomenon. The soil water content showed a decreasing trend with the deepening of degradation (Zhou et al., 2005; Li et al., 2020). The response of alpine meadows to soil at different stages of degradation focuses on soil bulk, and the specific response is expressed as an increase in bulk weight (Zi et al., 2015). Both of these processes reduce the water holding capacity of the soil, leading to a decrease in soil water content (Zhou et al., 2005). At the same time, enhanced evapotranspiration and infiltration also led to the loss of soil water. The loss of soil water in the surface layer of the meadow may be due to evaporation and infiltration from the ground surface, and in the later stages of degradation, the decrease in soil water content decreased (Fan et al., 2019), probably due to low vegetation cover, low plant activity level and reduced transpiration. With the deepening of degradation, the soil capacitance and soil pH showed an increasing trend. Soil pH was higher in the severely degraded stage than in the other degraded stage and showed a decreasing trend from severely degraded to moderately degraded, which was consistent with the integrated results of this study (Gao et al., 2006). Soil organic carbon showed an overall decreasing trend, but a small increase was observed at the stage of severe degradation, and an overall decreasing trend was observed with increasing soil depth, which was consistent with the integration results of this study (Han et al., 2011).

Soil is the most important carrier of nutrient cycling, water balance and microbial activity in terrestrial ecosystems and plays a key role in plant growth, and its physicochemical properties and nutrient status are the most important indicators of ecosystem stability (Liu and Tan, 1994; Yang et al., 2007). Soil nutrients, water content and pH are key indicators for restoring and maintaining the function of damaged ecosystems and play an important role in plant community succession (Li and Ma, 2002; Cheng et al., 2021). In this study, we found that the soil nutrient content in the 0–10 cm soil layer was higher than that in the 10–20 cm and 20–30 cm soil layers, which is due to the short growing season of alpine plants, which are mostly shallow-rooted plants, resulting in shallow root distribution and decomposition of dead roots into organic matter and other nutrients in the soil, which accumulate in the soil surface layer (Yue et al., 2015). Total nitrogen, phosphorus and potassium in artificial grassland showed an increasing trend with the increase in recovery years, but available N, P and K showed the opposite trend, and some studies also showed that the content of total nitrogen, available nitrogen and available potassium in artificial grassland soil increased with the increase in recovery years (Sun et al., 2019), which was not consistent with the results of this study, which was due to the heterogeneity of artificial grassland in different recovery years. Compared with severe degradation, the soil organic matter and organic carbon of artificial grassland increased (p < 0.05), but with the extension of the recovery period, the trend of the “V" shape was observed in 1-year, 5-year,10-year recovery, the lowest in 5-year recovery, and the overall trend was “high-low-high”. The overall trend of “high-low-high” is consistent with the findings of others (Yang et al., 2015). Compared with severe degradation, the soil pH, water content and bulk density of artificially established grassland will decrease, while high vegetation cover by the time the vegetation community steps into recovery succession will not only reduce evaporation of soil water content but also provide a rich source of organic matter for the soil, further indicating that the improvement of the soil environment is beneficial to the settlement of exotic species, seed germination, root expansion and their nutrient uptake (Wang et al., 2009).

Overall, alpha diversity of bacteria and fungi increased with increasing degradation but decreased with increasing recovery years during recovery. The alpha diversity of bacteria was much higher in artificial grasslands restored for 1–4 years than in those restored by 10–18 years, which are similar to other studies (Gao et al., 2021). Bacterial diversity was higher in the early successional stages of vegetation (4 years) than in the later stages of vegetation succession (8 and 12 years) (Hu et al., 2019). For the degradation process, patch formation significantly increased fungal alpha diversity by approximately 40% (Che et al., 2019). Overall, it is plausible that the significant response of soil bacterial and fungal community alpha diversity to degradation and recovery processes may be influenced in multiple ways. First, soil pH changes frequently during degradation and recovery (Lennon and Houlton, 2017), and an important environmental factor affecting microbial communities is soil pH (Fierer, 2017). Many studies have found that soil pH is negatively correlated with microbial α-diversity during degradation and recovery (Tripathi et al., 2018; Guo et al., 2019). Second, in severely degraded meadows, reduced vegetation cover can accelerate topsoil erosion, increase roughness, and increase light, thus improving the competitiveness of aerobic and photosynthetic autotrophic microorganisms (Lin et al., 2015; Zhang et al., 2017). Bacterial and fungal community structure varies significantly across degradation and recovery processes, with significant differences in microbial communities during the degradation or recovery of alpine grasslands on the Tibetan Plateau (Li et al., 2016; Guo et al., 2019; Zhou et al., 2019). These differences are mainly due to the different vegetation characteristics and soil properties of different degraded and artificial grasslands. Degraded meadows limit plant growth, leading to physical damage to soil structure and nutrient status, which further accelerates soil degradation (Zhou et al., 2019), and alpine meadow degradation will significantly alter bacterial community structure due to reduced availability of microbial substrates as a result of reduced apoplactic inputs (Zhang et al., 2014).

Different soil conditions can have soil-specific effects on plant species (van de Voorde et al., 2011), and changes in soil properties directly affect the direction of plant community succession, providing a theoretical basis for plant‒soil feedback (Bever et al., 1997; Bardgett and Van Der Putten, 2014). When studied at the individual plant level, a positive feedback effect value indicates that the plant grows better in the original or the same plant-influenced soil; a negative feedback effect value indicates that the plant has a negative feedback effect in the soil (Baxendale et al., 2014). Similarly, when the plant‒soil feedback is positive at a certain stage of plant succession. It means that the plant is growing better at that stage and the soil is in a depleted situation, and vice versa. A meta-analysis at the individual species level showed positive plant‒soil feedback effects accounting for approximately 28% of the statistics (Kulmatiski et al., 2008). Plant‒soil positive feedback can increase the asymmetry of plant growth, leading to the dominance of some plants in a given habitat and the formation of communities dominated by a single plant species, influenced by (1) an increase in available soil nutrients due to plant growth, for example, through plant nitrogen fixation (Chapman et al., 2006), and (2) the accumulation of plant symbionts among the root system (Klironomos, 2002). Most studies have shown that plants in late successional stages are more dependent on interroot symbionts, exhibit greater environmental tolerance, and tend to have a positive plant‒soil feedback (Kardol et al., 2006), becomes the dominant species in the community, and gradually become the dominant species in the community. The results of this study showed that in ecosystem multifunctionality, positive feedback was observed in the predegradation period, positive feedback was observed in the moderate to severe degradation for biotic factors, and positive feedback was observed in the moderate degradation for abiotic factors. In the prerecovery stage, plant‒soil feedback in ecosystem multifunctionality showed positive feedback, and in the recovery year of 5 years, all three aspects of ecosystem multifunctionality, biotic factors and abiotic factors were positive feedback effects. In the early stage of succession, the preferentially planted plants promote their own growth by decomposing the dead material produced by aboveground plants and the mineralization of nutrients by soil microorganisms while reducing the competitiveness of other plant species in the community, such as the establishment of Molinia caerulea in the abandoned land where the distribution density of Calluna vulgaris is high. The vegetation type gradually changed to grassland (Van der Putten et al., 1993; Berendse, 1994). The positive plant‒soil feedback effect is a key factor in early plant changes (Reynolds et al., 2003), and contemporaneous changes in soil properties can have potential long-term effects on plant distribution and productivity (Grman and Suding, 2010; Kulmatiski and Beard, 2011). Plant-soil feedback studies are important for the study of plant community succession, ecosystem multifunctionality and productivity formation and maintenance mechanisms, as well as for theoretical guidance in understanding ecosystem responses to global ecological events, such as climate change and biological invasions.

Negative plant‒soil feedback occurs more frequently at the individual species level, accounting for approximately 70% of the included statistics (Kulmatiski et al., 2008). It promotes a balance between the survival states of two plants in a community and, to some extent, species coexistence, with the main influences being (1) nutrient fixation or depletion in the soil, resulting in inadequate supply (Berendse, 1994; de Kroon et al., 2012), (2) accumulation of predators or pathogens under the roots (Van der Putten, 2003) or phytochemical effects of chemicals released by plants (Bais et al., 2003), and (3) some degree of soil carbon aggregation (Ehrenfeld et al., 2005). The results of this study showed that in ecosystem multifunctionality, the late stage of degradation had a negative feedback effect, and biotic factors had a negative feedback effect in mild degradation. The abiotic factors had a negative feedback effect in moderate to severe degradation, and the feedback intensity changed from large to small with the increase in recovery years and the trend from positive to negative feedback. At the recovery year of 10 years in artificial grasslands, the ecosystem multifunctionality, biotic factors and at 10 years of artificial grassland recovery, negative feedback was observed in three aspects: Ecosystem multifunctionality, biotic and abiotic factors. Studies have shown that negative plant‒soil feedback effects are generally present in the middle and late stages of semiarid grassland evolution but have less impact on the dominant species (Reinhart, 2012). Some studies have also found that in early successional stages, when an individual plant exhibits a negative feedback effect on itself but a positive feedback effect on other plants growing together within the community, the dominance of this plant is replaced by the coexist plant if other plants exhibit a negative feedback effect on this plant (van de Voorde et al., 2011). During plant community succession, early successional plants exhibit strong negative plant-soil feedback effects on themselves, and if the co-existing species in the community exhibit negative feedback effects on them, this plant species will be replaced by the co-existing species. Thus, negative plant‒soil feedback effects can affect plant community structure by suppressing dominant species, reducing the number of most species and decreasing the competitiveness of all species within the community.