<p>Understanding the hydrological processes of mountainous watersheds is a highly significant and challenging process. Precipitation exhibits high spatial and temporal variability in these regions, making the selection of appropriate precipitation datasets crucial for accurate hydrological modelling. Precipitation data closely aligned with local climatic conditions will help simulate more accurate and reliable results. Even after merging datasets to improve the accuracy of precipitation datasets, uncertainties often persist. These uncertainties can significantly impact hydrological applications. Therefore, it is crucial to assess not only the statistical consistency of precipitation datasets but also their ability to capture the realistic hydrological responses of complex mountainous regions. This study proposes a comprehensive framework for evaluating and improving the precipitation dataset for hydrological applications in mountainous regions. It includes multi-matrix statistical and hydrological evaluation, the development of a composite pixel-wise ranking score, the merging of selected datasets using various blending techniques, and the validation of the merged datasets through streamflow simulations using a fully distributed hydrological model. Based on the composite performance score most promising datasets for merging are ranked. Two merging methods were applied, and the resulting products were validated through streamflow simulations using the fully distributed WATFLOOD model. The two-stage blending approach demonstrated superior performance, producing the most accurate streamflow simulations and outperforming individual datasets. These findings highlight the value of incorporating both statistical and hydrological evaluations when selecting and improving precipitation datasets for regional-scale hydrological modelling in complex mountains.</p>