A significant decrease in annual runoff (AR) in the headwater region of Yellow River, China, has been observed during the past decades, which produces a recognized impact on long-term water resources management and planning. In this paper, the Pettitt method and Sequential Clustering method are used to detect the change point of AR and annual precipitation (AP) series. On the basis of the change point, the whole observed period is divided into two periods, before-point (Period I) and after-point period (Period II). The distribution characteristic of occurrence number of given extreme AP and AR and the relationship between the AP and AR in the two periods are analyzed. Then, the impact of AP change on the design AR is quantified. The results indicate that the AP and AR has a change point at the location of 1989.The distribution of occurrence number of given AP with more than 500 mm and AR with more than 150 mm in Period I is clearly different than that in Period II as well as the AP-AR relationship. The quantiles of AP and AR with the given non-exceedance probabilities of 0.8, 0.9, 0.95, 0.98, 0.99, 0.995, 0.998 and 0.999 in Period I decrease significantly compared to those in Period II. The AP decrease accounts for 53, 53, 54, 57, 60, 62, 66 and 68% of the total decrease in AR at the level of quantile with respect to the given eight non-exceedance probabilities from 0.8 to 0.999. Overall, the decrease in the AR quantile is mainly caused by the decrease in AP change.