Author(s): Kazuhiro Matsumoto; Mamoru Miyamoto
Linked Author(s):
Keywords: Clustering; Distributed hydrological model; Flood forecasting; Hydrograph; IFAS; Mathematical optimization
Abstract: A mathematical optimization procedure is presented to group multiple hydrographs into a small number of clusters for the purpose of helping to understand various runoff behaviors observed in flood events in a basin. In grouping, the hydrographs belonging to each cluster can be estimated with certain accuracy by the corresponding parameter set. The effectiveness is demonstrated using twenty-seven hydrographs observed in nine flood events and at three water level stations in the Abe River basin in Japan. The optimization results illustrate that eight sets of parameters are necessary to estimate such hydrographs with sufficient accuracy. One parameter set commonly estimates as many as seven out of twenty-seven hydrographs while some other parameter sets estimate the other hydrographs with different characteristics specific to flood events or water level stations. Most of the previous research is based on continuous optimization; however, a presenting procedure such as clustering is based on combinatorial optimization. Thus, new insight into understanding the runoff behaviors is brought by combinatorial optimization which is not often used in conventional research.
Year: 2018