Abstract: Real-time dynamic high-resolution soil moisture products can provide important support for regional agricultural production safety and security, and the current commonly used soil moisture remote sensing products have problems of spatial resolution and time series continuity. In order to generate spatio-temporally continuous high-resolution soil moisture results, we introduced the random forest (RF) and gradient boosting machine (GBM) algorithms in ensemble learning, and innovatively constructed a high-resolution soil moisture simulation (HRSMS) model integrating multi-source data. Input data comprised microwave soil moisture observations from the soil moisture active passive (SMAP) mission (2017–2022), vegetation indices, land surface temperature, and in-situ soil moisture measurements. Missing values in vegetation indices and temperature datasets were reconstructed using Savitzky-Golay filtering and multivariate regression. The model downscaled SMAP-derived surface soil moisture (0～5 cm depth) from 9 km to 1 km resolution. Validation was conducted in Jilin Province, a major agricultural region in northeastern China, and the results of the study indicated that: 1) The accuracy of HRSMS model was significantly improved compared to the commonly used polynomial regression fitting method. The root mean square error (RMSE), mean absolute error (MAE), and r square (R²) improved by 22.2%, 44.0%, and 0.27 compared to the polynomial regression fitting method, and the error was reduced by 33.2% in the main grain producing areas of northwestern Jilin Province; 2) In the HRSMS model, RF and GBM algorithms have similar computational efficiency, and RF algorithm R², RMSE, and MAE are higher than GBM algorithm by 4.9%, 0.7%, and 2.3%, respectively, and therefore either algorithm can be selected for model construction when carrying out related research in Jilin Province. The HRSMS framework effectively improved both spatial resolution and accuracy of soil moisture products while maintaining temporal consistency. This advancement supports precision agriculture through enhanced soil moisture monitoring capabilities. The operational flexibility between RF and GBM algorithms provides adaptability for diverse regional data conditions. The methodology holds significant potential for scaling to other agricultural regions requiring high-resolution soil moisture assessments.