Abstract: Annual maximum flood occurrence dates and magnitudes both can provide important information for the hydraulic engineering design and the reservoir operation.The existing literatures only consider the distribution of flood magnitudes,but ignors the flood occurrence dates.In this paper,Von Mises distribution and Pearson Type Ⅲ distribution are used to describe the occurrence dates and magnitudes of annual maximum flood respectively.A bivariate joint distribution with Von Mises distribution and Pearson Type Ⅲ distribution margins is developed based on the Gumbel Archimedean Copula and used to describe the annual maximum flood series.The approaches for calculating conditional probability,joint probability and bivariate return period are presented.Case study shows that the bivariate joint distribution can fit both occurrence dates and magnitudes of annual maximum flood series well.It can mine more flood information and provide a new way for flood frequency analysis.