LAUSR

Global analyses show that climate change has already increased the frequency and strength of extreme weather events. The observed and remote-sensed precipitation productions have been deeply studied considering the stationary and non-stationary characteristics of extreme series. However, as these studies largely addressed the evolution of precipitation extremes at a single station, the consistency risk of extremes at multiple stations within close days has been little researched. In this paper, a multivariate frequency analysis method is proposed to reveal the exceeding and coincidence probabilities of the most dependent extreme precipitation series. First, a generalized extreme distribution, which has proved better at describing the tail behavior of extreme series, is developed to simulate the marginal distribution of rainfall extremes. Then, a mixed von Mises distribution is introduced to fit the occurrence time interval of the extreme events. Taking advantage of its flexibility in choosing arbitrary marginals, a copula is constructed to implement a joint probability analysis of extreme precipitation events at multiple stations. The upper reaches of the Yangtze River, where frequent heavy rains are the main source of the regional floods, were selected as a typical research area. Five hundred twenty-eight copula models were finally constructed by arbitrarily choosing two precipitation extreme series among the total 33 meteorological stations. The results revealed the risk of extreme rainfall occurrence once in 2, 5, 10, 20, 50, or 100 years, within a time interval of 1–90 days. The research has great significance for understanding complicated extreme events and is helpful for the disaster emergency response and management.