We selected five grassland sites with different grazing exclusion times (0, 5, 9, 22, and 30 years) on Yunwu Mountain as the study sites, where the roots of Stipa plants in the 0-30 cm soil layer and soil of the 0-20 cm soil layer were collected. Roots were washed and identified to the species level based on attached aboveground parts and root color, texture, and branching. Root samples of the same species were scanned using an Epson Scanner to obtain images for analysis of root morphological traits using WinRhizoPro software. Root length, surface area, and volume were analyzed at 0.1 mm intervals in root diameter, the measurements of which could be used for calculations of specific root length (SRL), specific root surface area (SRS), and root tissue density (RTD). Soil samples were analyzed to determine soil moisture, soil bulk density, soil pH, soil organic carbon, total soil nitrogen, total soil phosphorous, available soil nitrogen, available soil phosphorus, microbial biomass carbon, and microbial biomass nitrogen. One-way analysis of variance was conducted to determine the effect of grazing exclusion time on Stipa root traits, and to determine differences in the roots of different Stipa species. Redundancy analysis was conducted to determine soil variables that best explain the variations in Stipa root traits. The main results were as follows. (1) The root biomass, root length density, root surface area, and root volume of Stipa plants showed a slight decrease at the beginning of grazing exclusion, but thereafter gradually increased and showed significant increases in grassland with grazing exclusion for 30 years. (2) All of the root trait indexes showed similar species compositional changes with grazing exclusion time. Specifically, the proportion of Stipa bungeana was the highest in grazed grassland, then gradually declined with grazing exclusion time, and disappeared after 30 years' grazing exclusion; the proportion of Stipa grandis initially increased, peaked in grasslands after 20 years' grazing exclusion, and then sharply decreased to the initial status; and Stipa przewalskyi appeared only in grassland with 30 years' grazing exclusion and became the dominant Stipa species. (3) Compared with the roots of S. grandis, the roots of S. bungeana and S. przewalskyi were thinner and had higher SRL and SRS, mainly resulting from a larger proportion of root traits in the 0-0.6 mm root diameter class. Besides, root tissue density of S. bungeana was higher than that of S. grandis and S. przewalskyi.(4) Long-term grazing exclusion significantly increased soil water content, total soil nitrogen, total soil phosphorus, soil nitrate nitrogen, and available soil phosphorus, and showed limited impacts on soil bulk density, soil pH, soil ammonium nitrogen, soil microbial biomass carbon, and soil microbial biomass nitrogen. With significantly increased total soil nitrogen, long-term grazing exclusion significantly decreased the ratio of soil carbon to nitrogen and significantly increased the ratio of soil nitrogen to phosphorus. (5) Redundancy analysis showed positive correlations among root biomass, root length density, root surface area, and root volume, and total soil nitrogen and ammonium nitrogen were significant factors explaining variations in the root traits of Stipa plants after grazing exclusion. Furthermore, root traits were positively affected by soil water content and available soil phosphorus, and were negatively affected by soil pH. In conclusion, inherent differences in the root traits of three Stipa species and changes in soil properties collectively drove the root trait patterns of Stipa plants in semiarid grassland after long-term grazing exclusion in this area.