Abstract
The restoration of "Heitutan" degraded grasslands on the Qinghai-Tibetan Plateau was hindered by suboptimal grass species mixtures, leading to low vegetation productivity, impaired soil nutrient cycling, and microbial functional degradation. Based on a 22-year controlled field experiment, this study systematically elucidated the regulatory mechanisms of different artificial grass mixtures on vegetation community characteristics, soil physicochemical properties, and bacterial community structure and function. The results demonstrated that mixed-sowing treatments significantly improved soil conditions and enhanced aboveground biomass. The HC treatment (Elymus nutans Griseb. + Poa crymophila Keng ex L. Liu cv. 'Qinghai' + Festuca sinensis Keng ex S. L. Lu cv. 'Qinghai') achieved aboveground biomass of 1580.0 and 1645.0 g·m(-2), representing 66.14% and 60.91% increases, respectively, compared to the HA monoculture (E. nutans). Concurrently, this treatment increased soil organic matter content by 52.3% and 48.4%, total nitrogen by 59.4% and 69.2%, while reducing electrical conductivity by 48.99% and 51.72%, with optimal pH stabilization (7.34-7.38). These findings confirmed that optimized grass mixtures effectively enhance soil physicochemical properties and carbon-nitrogen retention. Microbiome analysis revealed that the HE treatment (E. nutans + P. crymophila + F. sinensis + Poa poophagorum Bor. + Festuca kryloviana Reverd. cv. 'Huanhu') exhibited superior α-diversity indices (OTU, Shannon, Ace, Chao1, Pielou) with increases of 9.36%, 4.20%, 15.0%, 1.76%, and 13.4%, respectively, over HA, accompanied by optimal community evenness (lowest Simpson index). Core bacterial phyla included Pseudomonadota (22.7-29.9%), Acidobacteriota (21.5-23.6%), and Actinomycetota (13.6-16.0%), with significant suppression of pathogenic bacteria. Co-occurrence network analysis identified specialized functional modules, with HC and HD treatments (E. nutans + P. crymophila + F. sinensis + P. poophagorum) forming a "nitrogen transformation-antibiotic secretion" network (57.3% positive connections). Structural equation modeling (SEM) revealed that mixed sowing had the strongest direct effect on bacterial diversity (β = 0.76), surpassing indirect effects via soil (β = 0.37) and vegetation (β = 0.11). Redundancy analysis (RDA) identified vegetation cover (24.7% explained variance) and soil pH (20.0%) as key drivers of bacterial community assembly. Principal component analysis (PCA) confirmed HC and HD treatments as the most effective restoration strategies. This study elucidated a tripartite "vegetation-soil-microorganism" restoration mechanism, demonstrating that intermediate-diversity mixtures (3-4 species) optimize ecosystem recovery through niche complementarity, pathogen suppression, and enhanced nutrient cycling. These findings provided a scientific basis for species selection in alpine grassland restoration.