The wind erosion factor is a key parameter in determining the area of sand formation and the amount of sand and dust emissions in sandstorm patterns. However, the wind erosion data currently used in the model is usually assumed to be static, which fails to fully reflect the highly heterogeneous and dynamically changing characteristics of the dust source area, resulting in significant errors and uncertainties in the dust simulation results. To address this issue, we propose a new method that integrates multi-source data such as soil moisture, vegetation cover, soil texture, and land use to construct a physical process based wind erosion dataset. This dataset covers the whole world with a resolution of 1 kilometer, which can more finely depict the spatial characteristics of dust source areas. The results of sand and dust simulation based on WRF Chem show that the new dataset significantly improves the overall performance of sand and dust process simulation. The root mean square error (RMSE) of PM10 simulated using the new erosive data decreased by 32.4%, and the correlation coefficient (R) increased by 82.4% compared to the default data. In addition, the spatial distribution of simulated dust aerosol optical thickness (AOD) is closer to satellite AOD products.