Development of Design Hyetographs for Ungauged Sites Using an Approach Combining PCA, SOM and Kriging Methods

A design hyetograph represents the temporal distribution of rainfall intensity associated with a return period. The choice of the design hyetograph will have a significant influence on the shape and peak value of the hydrograph. Hence, the determination of design hyetographs is an important task in the hydrologic designs. In this paper, an approach is proposed to develop design hyetographs for ungauged sites. The proposed approach is composed of four steps: principal component analysis (PCA), self-organizing map (SOM)-based clustering, region delineation, and kriging-based construction. Firstly, PCA is applied to obtain the principal components of the design hyetographs. Then the transformed data resulting from PCA and the three geographic characters of the gauges are used as input data to the SOM, which is applied to group the rain gauges into specific clusters. Thirdly, the regions for these clusters are delineated and then the regions map is made. Finally, the design hyetographs for ungauged sites is constructed by using the kriging method. The proposed approach is applied to estimate the design hyetographs of ungauged sites in Taiwan. For comparison with the proposed approach, three other approaches are executed. Four gauges are treated as ungauged and the three approaches are used to construct the design hyetographs. The results show that accurate estimated design hyetographs can be obtained by the proposed approach. Cross-validation tests further have been performed to examine the stability and the accuracy of these approaches. Again, the results indicate that the proposed approach is more accurate and stable than the other approaches. Overall, the results demonstrate that the proposed approach is useful to develop design hyetographs for ungauged sites.