Species Combinations Shape Vegetation and Soil Nutrient Responses to Grass–Legume Reseeding in a Degraded Semi-Arid Grassland

Reseeding is widely used to restore degraded grasslands, but the extent to which restoration outcomes depend on grass–legume species combinations remains unclear. We conducted a two-year field experiment in a degraded semi-arid grassland in Inner Mongolia, China, using two grasses (Elytrigia elongata and Bromus inermis) and two legumes (Caragana microphylla and Lespedeza davurica) reseeded as monocultures or grass–legume mixtures. Reseeding resulted in higher vegetation cover than the unseeded control across all treatments, whereas biomass responses depended on species combination, with B. inermis + L. davurica showing the highest mean biomass and cover in both years. Most reseeding treatments reduced Shannon–Wiener and Simpson diversity, whereas E. elongata + C. microphylla maintained both indices at levels comparable to the unseeded control. Legume-containing treatments consistently increased soil total N, ammonium nitrogen, nitrate N, and microbial biomass N, while available P and microbial biomass C increased across all reseeding treatments. Vegetation productivity was positively associated with soil nutrient availability and microbial biomass but negatively associated with plant diversity. Overall, reseeding promoted early vegetation and soil recovery, but grass–legume mixtures were not universally superior to monocultures. These findings suggest that B. inermis + L. davurica may be preferred for rapid vegetation recovery, whereas E. elongata + C. microphylla may be considered when maintaining plant diversity is a priority.