Based on the basic principle of hydrologic cycle,the basic concept and classification of mid-to long-term runoff forecast are defined from the perspective of impact factors of water vapor sources of river basin runoff,the meteorological factors (midterm), climate factors (long-term),and astronomical factors (very long-term)are selected as the predictors,and the ten-day, monthly,seasonally (flood)or annual runoff targeting to the outlet section of river basin (or typical section)and drought and flood trend are forecasted.Through the comparison of the difference in the physical movement mechanism of water vapor system under different time scales,the rationality of short-term climate predictionas the theoretical basis for mid-to long-term runoff forecast is demonstrated.The impact factors for the runoff sources in river basin and their performance are summarized as three laws:the periodic law of astronomical factors is the main rule,which reflects the basic state of hydrological and climatic process; the random law of atmospheric circulation has interference on the basic state of hydrology and climate,which leads to fluctuation; and the characteristic law of river basin reflects the comprehensive effects from various factors,which has particularity.Finally, the mid-to long-term runoff forecast can be divided into two categories:runoff process forecast in normal year and drought and flood forecast in abnormal calamity year.The former takes the hydrological and climatic elements as predictors whereas the latter takes the astronomical factors as predictors.In terms of service objects,the former serves for the regular operation and scheduling of water projects whereas the latter provides disaster forecast for disaster prevention and reduction in flood control and drought resistance departments.In terms of publication of results,the former provides the quantitative forecast of runoff process whereas the latter provides the qualitative forecast of level 3 (or level 5)and similar years.In terms of forecast methods,the former adopts mathematical statistics method or physical cause correlation analysis whereas the latter adopts comprehensive forecasting identification method of periodicity,randomness,and watershed particularity.