Abstract:The random simulation of flood hydrographs is a new way for hydrological safety design. The Gumbel Hougaard Copula was employed to construct a bivariate joint probability distribution for describing flood peak and flood volume, whose marginal distributions are both the Pearson type III. From the derived joint probability distribution of flood peak and flood volume, the flood peak series sample and the flood volume series sample were randomly drawn simultaneously and used to construct the synthetic flood hydrographs. The joint probability distribution of flood peak and flood volume could reflect the correlation between flood peak and flood volume much better than the two separate one-variable probability distributions, one only for flood peak and the other only for flood volume. Compared with other models used for flood random simulation, such as the seasonal first-order autoregressive model and the disaggregation model, the method proposed here performed much better, in terms of the bias and the root mean square error.