When predicting monthly runoff with set pair analysis (SPA) method, it will become much difficult to determine the difference degree coefficients as the number of runoff levels increased. To solve this problem, the SCEM-UA algorithm was employedfor optimizing the difference degree coefficients of SPA. This method was applied to forecast the monthly runoff. The results showthat the proposed method can effectively distinguish the effects of difference degree among set pairs, and the difference degree co -efficient optimized by the method is effective and reasonable, the runoff prediction accuracy is improved greatly. Also, the parame -ter uncertainty occupies the largest proportion in the model uncertainty.