Investigation of Changes in Hydrologically Homogeneous Regions for Regional Frequency Analysis under Climate Change

 Regional frequency analysis (RFA) is a more reliable method for estimating hydrological quantities than at-site frequency analysis, particularly in countries like South Korea where the observation period for hydrological data is relatively short. The results of RFA vary depending on the classification of hydrologically homogeneous regions. With the increasing occurrence of extreme climate events due to climate change not only in South Korea but also globally, the validity of existing hydrologically homogeneous regions defined solely by historical rainfall data is now in question. Currently, South Korea’s flood estimation guidelines classify the country into 26 homogeneous regions based on hydrological data collected up to 2017, without considering the impacts of climate change. Therefore, it is necessary to evaluate whether the currently used Generalized Extreme Value (GEV) distribution remains appropriate by conducting a goodness-of-fit test after redefining hydrologically homogeneous regions. This study aims to reclassify South Korea's hydrologically homogeneous regions for rainfall regional frequency analysis using the up-to-date rainfall data and clustering analysis techniques. After collecting recent rainfall data, the data will be corrected using the Inverse Distance Weighting (IDW) method, followed by the reclassification of homogeneous regions through three clustering methods. The clustering methods to be applied include k-means, Self-Organizing Maps (SOM) based on artificial neural networks, and t-Distributed Stochastic Neighbor Embedding (t-SNE), a dimensionality reduction technique for high-dimensional data. The results of the homogeneous region classifications derived from each clustering method will be compared using measures of discordance(H) and heterogeneity(D_i). This study is expected to provide insights into how climate change affects the classification of homogeneous regions in regional frequency analysis.