The data set was prepared using the measured data of 973 soil sampling points in the Qilian Mountains and its surrounding areas. The main sources of data are as follows:

(1) The research team undertook the projects of the National Natural Science Foundation of China, "Research on the law and mechanism of soil organic carbon change along the gradient in Qilian Mountains" (41771252), "Investigation of ecological and hydrological transects in the Heihe River Basin" (91025002), and "Research on soil carbon simulation of typical ecosystems in the arid inland river basin" (31270482), The data of the major project of Gansu Province, "Research on the interaction between Qilian Mountain water conservation ecosystem and hydrological process and its adaptation to climate change" (18JR4RA002), the scientific and technological innovation project of Gansu Provincial Forestry and Grass Administration, "Research on the effectiveness of ecological governance in Qilian Mountain National Park (Gansu Area)" (GYCX [2020] 01), and the project funded by Gansu Provincial Science and Technology Plan, "Gansu Qilian Mountain Ecological Environment Research Center" (18JR2RA026). (2) The original soil profile data, document integration data (Xu et al., 2018 DOI: 10.1038/s41598-018-20764-9), and the World Soil Reference and Information Center (ISRIC) WoSIS soil profile database published by the National Glacier Frozen Desert Scientific Data Center and the Qinghai-Tibet Plateau Scientific Data Center（ https://www.isric.org/explore/wosis ）Etc. After the quality control of the original data, the sampling points with the accuracy of longitude and latitude positioning less than 0.001 and the abnormal values of soil organic carbon content significantly deviating from the environmental background were removed

According to the framework of "Scorpan" (Soils, Climate, Organisms, Relief, Parent material, Age, Geographic position), the spatial distribution of soil organic carbon density at 10m resolution of 0~100 cm soil layer in Qilian Mountains was simulated using the digital soil mapping (DSM) method and the calculation based on tile structure. The prediction method is mainly based on the extreme gradient boosting algorithm (XGBoost, eXtreme Gradient Boosting) in machine learning. The environmental covariate data include: Sentinel-2 multispectral image, Sentinel-1 radar image, temperature, precipitation, radiation, terrain, vegetation index, location and other grid data.

The digital soil mapping framework is constructed in R language. In order to reduce the memory consumption and improve the calculation speed, the whole Qilian Mountain area is divided into 33 150 km × 150km sub-region, each sub-region is built based on tiles (tile size: 15km × 15km) and parallel computing. The bootstrap method is used to repeat the modeling, and each bootstrap sample is spatially modeled to obtain the frequency distribution of the modeling results. The uncertainty of modeling is expressed by standard deviation (sd)