Abstract: Soil bulk density (SBD) is an important parameter to reflect soil quality and soil productivity. Due to the time-consuming, laborious and on-site sampling required by the method of determining soil bulk density (ring tool method), the number of bulk density samples was very limited, which affected the estimation accuracy and precise control of soil carbon and nitrogen storage. In this study, based on soil bulk density and organic carbon data obtained from 437 grassland sampling sites in Xinjiang, the statistical characteristics of bulk density and organic carbon were analyzed to build a bulk density prediction model for the whole Xinjiang and 6 regions, respectively. The results showed that: (1) The ranges, mean values and coefficients of variation of bulk density and organic carbon in the main grassland soils in Xinjiang were 0.24 - 1.99 g cm⁻³ and 0.01% - 24.51%, 1.24 g cm⁻³ and 1.86%, 23.15% and 129.96%, respectively; (2) The optimal prediction model was SBD = 1.3861 e^{−0.0704 SOC}, (R² = 0.35, P < 0.0001); (3) Compared to the optimal prediction model at the Xinjiang scale, the regional optimal models built for the Altai Mountains and the Ili river valley and other areas had an improvement in accuracy. The results could provide important parameters for the accurate assessment of soil carbon storage in Xinjiang grasslands.