<p>Precipitation estimation is of vital importance for hydrological simulation and water resources management. However, precipitation datasets usually have large uncertainties in high-alpine regions due to the scarce gauged observations and complex terrains. Data fusion technologies are commonly used to integrate the advantages of multi-source precipitation datasets, but the spatial information of precipitation is usually ignored. To address this limitation, we developed a two-step machine learning framework for merging multi-source precipitation datasets based on the 2D convolutional neural network (CNN), incorporating neighboring spatial information, hereafter referred to as N-CNN. The framework combines classification and regression models to merge three gridded precipitation products (i.e., ERA5-Land, TPReanalysis, and GPM) and gauged observations over a high alpine watershed in China during the period 2001–2019. Two merged precipitation datasets were generated by CNN and the proposed N-CNN framework, respectively. Additionally, two regional reanalysis datasets with different spatial resolutions were merged to assess the effect of resolution on the merged datasets. The results show that the proposed framework effectively integrates the advantages of multiple datasets. The CNN and N-CNN merged precipitation datasets have a similar spatial distribution to the original products but differ in precipitation amounts. Precipitation amounts of merged data are much closer to gauged observations than original precipitation products. Both merged datasets outperform the original products in terms of statistical and categorical indices evaluated based on 25 independently meteorological stations with a complete time period (covering 2001–2019). However, the N-CNN merged dataset performs better than the CNN merged dataset in capturing precipitation amounts and detecting precipitation events, especially for moderate (5–10 mm d⁻¹) and heavy precipitation (&gt;10 mm d⁻¹). Compared with the CNN merged result, the N-CNN framework reduces the station-averaged root mean square error (RMSE) from 4.25 to 3.74 mm d⁻¹ for moderate precipitation and from 9.43 to 8.57 mm d⁻¹ for heavy precipitation, while increasing the station-averaged critical success index (CSI) by 0.03 and 0.04, respectively. Resolution analysis further reveals that the effect of regional reanalysis resolution mainly concentrates on precipitation amounts, while it has little effect on event detection. In general, this study highlights the importance of incorporating spatial information in precipitation merging, especially for high-alpine regions.</p>