Potential Flood Risk Analysis Considering Climate Change Impact

In the present study, we used multiple machine learning methods to assess flood risks caused by the effects of climate change and predict future flood risks based on climate change scenarios. In particular, we aimed to verify the usefulness of machine learning algorithms as a tool to predict future flood risks. Through this analysis, we sought to provide fundamental data for preparing plans to improve flood risk predictions and establishing climate change adaptation measures. To this end, we used three machine learning algorithms (naïve Bayes, k-nearest neighbor, and random forest) to predict flood risks in certain areas of Cheongju, Chungcheongbuk Province, South Korea, where heavy rain occur frequently, and assessed future flood risk for each regional climate model and climate change scenario (Representative Concentration Pathway [RCP] 4.5 and 8.5). RF (AUC: 0.616), among all machine learning algorithms tested, achieved the best flood risk predictive performance. Furthermore, for each climate change scenario, the models confirmed a gradual increase in flood risk over time, even when the uncertainty of the future is considered. In particular, if the greenhouse gas mitigation policy fails (RCP8.5), the flood risk will increase significantly.