Interdependence of Extreme Floods

本研究方案涉及大尺度洪水的空间方面。本文提出了一种从观测到的放电的时间序列特性中推导出常见事件的方法，而不是传统的使用一组极端事件进行事件调查。基本性质可以用傅立叶变换和哈尔变换得到，这两种变换允许描述对应于不同时间尺度的相关性。利用这些技术可以确定关键的时间尺度，并可以量化它们对洪水的贡献。复杂的高维依赖结构不能用两两调查来检测。基于光谱特性和拓扑结构的新技术将用于检测这种依赖性并量化它们的作用。请注意，这些依赖关系的作用可能不同，在某些情况下，它们的后果是更高的，而在另一些情况下，它们的后果是更低的大规模洪水风险。将比较气象模式的观测降雨量、序列和输出，以发现空间依赖性和空间结构的差异。这样就有可能对同时发生的洪水作出气象学上合理的解释。为了评估不同地点同时发生极端放电的概率，将开发随机放电和降雨发生器。将开发两种不同的方法-基于成对依赖的传统生成器和反映高阶依赖的生成器。对于降雨发生器的空间特性，气象模式的输出也会得到反映。将比较这两种模型的实际性能。降雨发生器的输出将作为水文模型的输入，以获得模拟的流量序列。然后将这些数据与观测序列和从排放产生模式得到的序列进行比较，这样就可以更好地了解模式的可信性以及气象因子和地表性质的作用。

This research proposal deals with the spatial aspects of large scale floods. Instead of the traditional investigation of events using a selected set of extreme events, a methodology to derive common occurrences from time series properties of observed discharges is suggested. The essential properties can be derived using Fourier and Haar transformations which allow a description of the dependence corresponding to different time scales. Using these techniques critical time scales can be identified and their contribution to floods can be quantified. Complex high dimensional dependence structures cannot be detected using pairwise investigations. New techniques based on spectral properties and topological structures will be used to detect such dependences and to quantify their role. Note that these dependences may act differently, in some cases their consequence is a higher in others a lower large scale flood risk. Observed rainfall, series and output of meteorological models will be compared to detect differences in spatial dependence and spatial structure. This way a meteorologically plausible explanation of simultaneous flood occurrences will be possible. In order to assess probabilities of simultaneously occurring extreme discharges at different locations stochastic discharge and rainfall generators will be developed. Two different approaches – a traditional generator based on pairwise dependences and one reflecting high order dependence will be developed. For the rainfall generator spatial properties of the meteorological model output will also be reflected. The two types of models will be compared with respect to their areal behavior. Outputs of the rainfall generator will be used as input for hydrological models to obtain simulated discharge series. These will then be compared to the observed series and to the series obtained from the discharge generating model and both the plausibility of the models and the role of the meteorological factors and surface properties can be better understood.