Characterisation of seasonal flood types according to timescales in mixed probability distributions
Characterisation of seasonal flood types according to timescales in mixed probability distributions
复制标题
DOI：
10.1016/j.jhydrol.2016.05.005
复制
发表时间：
2016-08
影响因子：
6.4
通讯作者：
S. Fischer;A. Schumann;M. Schulte
中科院分区：
文献类型：
--
作者：
S. Fischer;A. Schumann;M. Schulte
当洪水统计是基于年最大系列（AMS），样本往往包含洪峰，其成因不同。如果事件类型之间的比率在观测范围内变化，则概率分布函数（pdf）的外推可以由属于某一洪水类型的大多数事件主导。如果这种类型对于非常大的极值不是典型的，那么pdf的这种外推是误导的。为了避免这种违反同质性假设的情况，开发了冬季和夏季洪水之间不同的季节模型。我们表明，夏季和冬季洪水之间的区别并不总是足够的，如果季节性系列包括不同成因的事件。在这里，我们区分洪水的时间尺度成组的长期和短期的事件。提出了一种区分事件的统计方法。为了证明其适用性，德国河流流域的冬季和夏季洪水的时间尺度进行了估计。结果表明，夏季洪水可以分为两大类，但在我们的研究区域，冬季洪水样本至少包括三个不同的洪水类型。通过一个新的混合模型，将两组夏季洪水的PDF合并。该模型认为，有关并行事件的信息，只使用其最大值是不完整的，因为一些实现是重叠的。提出了一种修正统计参数的统计方法。在德国的案例研究中的应用表明了新模型的优势，特别强调洪水类型。
When flood statistics are based on annual maximum series (AMS), the sample often contains flood peaks, which differ in their genesis. If the ratios among event types change over the range of observations, the extrapolation of a probability distribution function (pdf) can be dominated by a majority of events that belong to a certain flood type. If this type is not typical for extraordinarily large extremes, such an extrapolation of the pdf is misleading. To avoid this breach of the assumption of homogeneity, seasonal models were developed that differ between winter and summer floods. We show that a distinction between summer and winter floods is not always sufficient if seasonal series include events with different geneses. Here, we differentiate floods by their timescales into groups of long and short events. A statistical method for such a distinction of events is presented. To demonstrate their applicability, timescales for winter and summer floods in a German river basin were estimated. It is shown that summer floods can be separated into two main groups, but in our study region, the sample of winter floods consists of at least three different flood types. The pdfs of the two groups of summer floods are combined via a new mixing model. This model considers that information about parallel events that uses their maximum values only is incomplete because some of the realisations are overlaid. A statistical method resulting in an amendment of statistical parameters is proposed. The application in a German case study demonstrates the advantages of the new model, with specific emphasis on flood types.