Abstract: The correction of multi-source precipitation data is critical for analyzing and modeling hydrological patterns in data-scarce regions. However, existing correction methods often overlook the influence of climatic factors. To address this, a precipitation bias correction model was developed based on BiLSTM, integrating multi-source precipitation datasets (ERA5, ERA5-Land, MSWEP-V2, and PERSIANN-CDR) with teleconnection factors. The XGBoost-SHAP model was employed for feature selection and causality analysis, while a Bayesian Optimization (BO) strategy was applied to identify the optimal combination of model hyperparameters to further enhance correction accuracy. Using the Upper Hanjiang River basin as the study area, multi-source precipitation data bias correction was conducted. The results indicate: ①Atmospheric circulation factors are the primary influences on precipitation formation in the Upper Hanjiang River basin, with the position index of the Northern Hemisphere Subtropical High Ridge having the most significant impact. ②Compared with traditional statistical methods, the BO-BiLSTM model, while slightly less effective than the optimal parameter transformation method, provides greater flexibility in incorporating multiple influencing factors. ③After considering teleconnection factors, the Nash-Sutcliffe efficiency coefficient of the corrected multi-source precipitation data during the test period improved by an average of 5.4%, the mean squared error decreased by 24.6% on average, and the Kling-Gupta efficiency coefficient increased by 10.5% on average. These findings offer a practical technical solution for high-precision monthly precipitation estimation and extension in data-scarce regions.