Flood statistics with seasonal differentiated mixed distributions

点击翻译按钮获取中文摘要

In the last decades the estimation of flood probabilities by extreme value statistics for annual maximum flows was largely standardized in Germany by rules such as the DWA M -552. However the basic conditions for the application of extreme value statistics to the annual maximum values which were already emphasized by Gumbel in 1941 (the sample is homogeneous, i.e. each value of the sample is subject to the same distribution, and the causes of the events are the same) are not satisfied. The causes of the observed flood events are complex and varied. E.g. floods, caused by rain and snowmelt are mixed within series of annual maximum flows. It can be shown that winter and summer floods have different frequencies, which can be described by different distribution functions. The seasonal differences in the flood genesis, which is reflected in the seasonal variability of the occurrence of flood events of different sizes, are an important feature of the flood regime of a catchment. An inclusion of this feature in flood statistical analyzes can reduce their uncertainty and improves the estimation of probabilities for extremely rare flood events. As part of the planned work the seasonality of floods in different German regions will be analyzed. The results will be taken into account in the probabilistic assessment of extremely large events by mixed distributions. It is assumed that distribution functions of flood events, which differ in their origins can be determined with the help of seasonal differentiated analyzes. The different seasons will be separated by hydrological indicators which are derived from analyses of observed flood events. The seasonal flood probabilities can be described by different statistical approaches and finally combined to distributions of annual maximum values . Several mathematical options for calculating seasonal differentiated flood statistics will be compared. By considering a larger number of catchments in Bavaria, Thuringia and Saxony the generality of the newly developed approaches for generation of mixed distributions will be demonstrated. A significant improvement of flood statistical analyzes is expected as a result of the planned research work. The results of this basic research will be widely applicable for practice.