The soil measured data in this dataset mainly comes from the soil organic carbon density data of 263 sampling points at the slope and watershed scales obtained by the National Natural Science Foundation of China projects (41771252, 31270482, 91025002).

Based on the "Scorpion" (Soil, Climate, Organisms, Relief, Parent material, Age, Geographic position) framework, digital soil mapping (DSM) method and tile structure calculation were used to simulate the spatial distribution of soil organic carbon density at a resolution of 2m in the 0-100 cm soil layer of the Dayekou watershed in the Qilian Mountains. This was achieved by integrating 2-meter DEM data from the Dayekou watershed (Zhang Yanli, 2020) and Quickbird 2.5m resolution multispectral images (Guo Jianwen, 2019). The prediction method is mainly based on the Extreme Gradient Boosting algorithm (XGBoost) in machine learning, which uses grid data such as climate, precipitation, radiation, terrain, vegetation index, spectral information, and spatial position as input variables for spatial mapping. Through repeated modeling of bootstrap, spatial modeling is performed on each bootstrap sample to obtain the frequency distribution of the modeling results. The uncertainty of the modeling is represented by standard deviation (SD). After 30 rounds of 10 fold cross validation, the RMSE and R2 of the model were 5.34 kg/m2 and 0.84, respectively. In the final data product, mean and SD represent the mean and standard deviation of 30 repeated modeling, respectively, in kg/m2, indicating the mass of soil organic carbon in the 0-100cm soil layer per unit area.