The coincidence risk of the non-stationary flood peak between the Yellow River and Qingshui River was analyzed. The variability of flood peak time series was detected using the comprehensive diagnosis method. The mixed Von Mises distribution and mixed P-Ⅲ distribution were used to fit the flood occurrence dates and peak magnitudes, respectively. The asymmetric partially nested Archimedean Copula function was selected to establish the joint distribution of flood peaks of the two rivers, and the occurrence dates and corresponding magnitudes between the two rivers were studied. Diagrams of coincidence risk of floods with different combinations of frequencies of the two rivers were provided. The results indicate that the flood time series in both rivers had changed; the coincidence risk of the flood peak occurrence dates of the two rivers peaked on July 31 and August 31, with values of 3. 36×10^-4 and 1. 15×10^-4, respectively; and the peak coincidence risks of flood peaks with difference frequencies were obtained in two periods, from July 26 to August 2 and from August 31 to September 5.