Abstract: The intra-annual precipitation of the headwater catchment of the Yellow River basin (HCYRB) occurs in a concentrated manner, which leads to a high flood risk. It is of great scientific significance and application value to reconstruct the rainy season precipitation and flood season streamflow of the HCYRB for improving the runoff prediction accuracy and for flood and disaster prevention. A nested principal component hierarchical Bayesian regression model, which estimates the posterior distribution of parameters instead of fixed values to consider the uncertainty, was used to reconstruct the rainy season precipitation in the past 1 160 a using 16 tree-ring chronologies in and near HCYRB; categorical proportion regression model based on annual streamflow was proposed to reconstruct the flood season streamflow of the HCYRB up to C.E. 159 year. The results showed that: ① Reduction of error (E_R) and coefficient of efficiency (E_C) values obtained with the nested principal component hierarchical Bayesian model were significantly higher than 0, and those obtained with the categorical proportion regression model were up to 0.90 and 0.88, respectively, which indicates that the reconstructions of the rainy season precipitation and flood season streamflow of HCYRB are reliable; ② Even in the millennium time scale, 1979—1985 represented an unusual period of high runoff during the flood season.