Author(s): Hideyuki Yamaji; Tetsuya Takeshita
Linked Author(s): Hideyuki Yamaji, Tetsuya Takeshita
Keywords: Climate Change d4PDF Self-Organizing Map RRI model
Abstract: In light of concerns that climate change will increase rainfall and cause severe flood damage, the National Institute for Land and Infrastructure Management (NILIM) has used long-term ensemble climate projection data to calculate the rainfall change multiplier [the ratio of rainfall in the present and future climate], which is used to develop the basic river management policy based on climate change in Japan. The rainfall change multiplier is an index that focuses on the total rainfall in one rainfall event, but even if the total amounts of rainfall are the same in several rainfall events, there is concern that differences in the spatiotemporal distribution of rainfall may result in greater flood damage. On the other hand, considering all rainfall events is a heavy analytical burden. Therefore, in this study, a cluster analysis of rainfall spatiotemporal distribution by Self-Organizing Map was conducted in the Kuzuryu River basin using long-term ensemble climate projection data from the 2°C rise experiment in order to efficiently extract rainfall spatiotemporal distribution to be considered for the basic river management policy from a vast amount of rainfall data. Among the rainfall events in each cluster of rainfall spatiotemporal distribution, two representative rainfall events with total rainfall equivalent to approximately 1/150 annual exceedance probability of occurring were extracted and rainfall-runoff calculations were conducted. The cluster with the most rainfall events was the “Upper and middle Kuzuryu River, Peak in the Late” type, and the results of rainfall-runoff calculations using eight representative rainfall events indicated that the cluster with the largest flow was the “Upper Kuzuryu River, Peak in the Middle” type.
Year: 2025