 This paper proposes a novel framework combining the machine-learning downscaling algorithm and post-processing procedures for generating high-resolution precipitation data. For ML-based models, including support vector regression, random forest, spatial random forest, and extreme gradient boosting, were tested for downscaling and compared with conventional downscaling methods. The results showed that the ML-based methods outperformed the conventional regression and interpolation approaches. Additionally, the geographical difference analysis, GDA, calibration process significantly improved the downscaled results. Finally, the combination of the spatial random forest, SRF, downscaling algorithm and the GDA calibration method was found to be the most effective approach for generating high-resolution precipitation data. This framework could be used to generate high resolution precipitation data, especially in areas where data are scarce, which would benefit regional water resource management and hydrological disaster prevention. This article was authored by Hung Lin Zhu, Huazeng Lu, Cheeming Zhou, and others. We are article.tv, links in the description below.