Abstract: Copula function is used to combine the correlation degree and relevance pattern of independent variable and dependent variable. A joint distribution function can be constructed, and its interrelated structure with the two variables is analysed. Then, expressions for the relation of the marginal distribution of dependent variable versus the joint distribution are formulated, and the unknown value of dependent variable is predicted through a combination with the known sample value of probability distribution of independent variable. Case study of annual runoff prediction shows that this Copula prediction method has a clear theoretical basis and it is stable and accurate. This method not just provides new ideas for hydrological prediction, but offers new insights into Copula function.