Catchments as non-linear filters: understanding catchment similarities for the regionalisation of rainfall-runoff transformation processes using wavelets

在没有流量测量或只有少量流量测量的集水区（所谓的无流量测量集水区）进行流量预测正在成为一项重大的科学和社会经济挑战。目前可用的水文模型很难从有资料的集水区转移到无资料的集水区，由于在流域尺度上的排放形成的水文过程的了解不足。这种水文模型的空间转移是可能的假设下，类似的景观单元产生类似的水文过程。本研究的目的是调查这一假设，并提出一种新的方法来量化水文景观相似性。这种相似性度量将基于这样的假设，即不同的景观在植被径流转化过程中具有不同的非线性过滤特性，并且这些特性可以量化。滤波效果将通过信号处理技术，即小波分析来确定。Wavelet能够定量分析系统输入时间序列的某些频率特征如何在系统输出时间序列中被过滤。在本研究中，这一主要属性将被利用的不同景观单元的过滤特性的调查。基于这种动态特征的相似性度量的主要优点是将集水区或景观单元的水文响应直接与制约这种响应的过程联系起来。因此，它不仅有助于水文响应的建模，而且有助于对其的理解。新方法的效率和获得的见解将通过一个流量区域化研究与数据集从德国部分的莱茵河集水区进行验证。

Discharge prediction in catchments with no or only a few discharge measurements - so-called ungauged catchments - is emerging as a major scientific and socioeconomic challenge. Currently available hydrological models are difficult to transfer from gauged to ungauged catchments due to an insufficient understanding of the hydrological processes underlying the discharge formation at the catchment scale. Such a spatial transfer of hydrological models is possible under the assumption that similar landscape units produce similar hydrological processes. The objective of the present research is to investigate this assumption and to propose a new approach for the quantification of hydrological landscape similarity. This similarity measure will be based on the assumption that different landscapes have different non-linear filtering characteristics during the rainfall-runoff transformation process and that these characteristics can be quantified. The filtering effects will be identified through signal processing techniques, namely wavelet analysis. Wavelets enable a quantitative analysis of how certain frequency features of a system input time series are filtered in the system output time series. In the present research, this main property will be exploited for the investigation of the filtering characteristics of different landscape units. A similarity measure based on such dynamic characteristics has the main advantage of connecting directly the hydrological response of a catchment or landscape unit to the processes that condition this response. It contributes therefore not only to the modelling of the hydrological response but to its understanding. The efficiency of the new approach and the gained insights will be validated through a discharge regionalisation study with data sets from the German part of the Rhine catchment.