Southern Xinjiang located in the far northwest of China is experiencing serious particulate matter (PM) pollution. The wind erosion has been recognized as a great contributor of PM₁₀ and PM_2.5 concentrations in southern Xinjiang. In this study we developed a method that used the Wind Erosion Prediction System (WEPS) to establish the PM₁₀ and PM_2.5 emission inventories from the wind erosion source in southern Xinjiang with a 10 km × 10 km spatial resolution and a temporal resolution of one month. The PM₁₀ and PM_2.5 emission inventories from wind erosion sources were provided as input to the Models3/SMOKE (Sparse Matrix Operator Kernel Emission). The Community Multi-scale Air Quality (CMAQ) model was employed to simulate PM₁₀ and PM_2.5 concentrations from wind erosion source and the results were compared with the monitoring data to verify the emission inventories. The total PM₁₀ and PM_2.5 emissions were 12 × 10⁶ t and 4 × 10⁶ t, respectively, and the emission per unit area were 14.6 t/km² and 4.9 t/km², respectively, in southern Xinjiang in 2016. The PM₁₀ and PM_2.5 emissions per unit area were highest in Kashi being 19.1 t/km² and 9.1 t/km², respectively. The total emissions peaked at 5.4 × 10⁶ t/km² in PM₁₀ and 1.4 × 10⁶ t/km² in PM_2.5 in Bazhou. The PM₁₀ and PM_2.5 emissions were highest in spring, followed by summer, winter and autumn. The Normalized Mean Error (NME) between the predicted monthly average PM₁₀ and PM_2.5 concentrations from wind erosion source and the results from multiplying the monitoring data by the source apportionment results of PM₁₀ and PM_2.5 were 25.3% and 26.1%, respectively, which indicated that the accuracies of the PM₁₀ and PM_2.5 emission inventories from wind erosion source in southern Xinjiang were satisfactory.