Spatial Extreme Value Model Of Regional Precipitation Event In The Middle And Lower Reaches Of The Yangtze River

Abstract Climate change and water circulation is the forefront of the interna-tional global change and water science research, and uncertainty is one of the difficulties of the climate change impacts on water resources; for example, due to the low frequency of extreme events and the large time span, traditional statis-tical methods can not be used directly to the analysis of extreme precipitation events. In addition, since the regional characteristics of the environmental statis-tics are becoming more and more important, the statistical modeling of extreme precipitation event must consider these regional impacts and correlation. In theory, there are methods based on Extreme Value Theory to predict extreme precipitation events. Although univariate extreme value model which is applied to single-site data have already been developed, but the modeling of regional precipitation events is not simply to extend univariate methods to multivariate ones, the methods of modeling and diagnosis are of great difficulties.In the present paper, composite likelihood method are used to fit annual max-imum daily precipitation of44years of the15stations around the middle and lower reaches of the Yangtze River, which is a spatial extreme model based on max-stable process. The fitting procedure is taking into account both the s-patial extreme correlation structure and marginal distributions, and also using the fitted model to predict extreme precipitation events in the middle and lower reaches of the Yangtze River Basin.