On the 12th, President Lee Jae-myung visited the Seoul Han River Flood Control Center for his first livelihood initiative after taking office. He chaired a meeting to directly assess the flood preparedness in response to an early start to the rainy season compared to previous years.
Floods and disasters have become increasingly severe as climate change accelerates, to the extent that the president prioritized them. Their form is also becoming more complex. Experts have relied on past experiences and intuition to assess regional flood risks, but now it is difficult to foresee even a moment ahead.
Professor Gam Jong-hoon of Pohang University of Science and Technology POSTECH expects that artificial intelligence (AI) will play a significant role in predicting floods that cause localized damage. Professor Gam's research team developed a 'flood risk map' that predicts and visualizes regional damage risks based on 189 flood incidents across 236 cities and counties in the country from 2002 to 2021, and published it in an international academic journal.
During a meeting at Pohang University of Science and Technology POSTECH on the 11th, Professor Gam noted, "In the past, we gathered experts and held surveys to determine risk levels, which was time-consuming and inevitably subject to subjective judgment," adding, "AI quantifies this process, creating judgments that are repeatable and comparable."
AI combined various data points, such as the area of impermeable surfaces, river ratios, population density, and past damage history, to quantitatively assess risk without relying on subjective human judgment. The AI analysis classified large cities like Seoul and Incheon as high-risk flood areas.
Professor Gam said, "However, we should not interpret this as a simple conclusion that 'cities are more dangerous.'" He explained, "In large cities, there is less regional variance, so risk signals are clearly evident, whereas in rural areas there is greater variability in damage, making risk less distinct, but that does not mean they are safe."
The research team also suggested policy directions. Since cities expose a large population and facilities to risks, policies are needed to reduce 'exposure,' while rural areas should improve their preparedness against disasters, that is, increase 'response capacity.'
Professor Gam explained, "This study is a macro analysis based on nationwide data," adding, "Local governments need to rely on much more detailed local data, so more precise data is required to link it to practical policies."
In particular, Professor Gam emphasized that AI is not omnipotent and remains a supportive tool. The factors considered significant may vary depending on the AI model used in this study. He stressed, "AI can provide draft policy judgments, but expert interpretation and judgment must also be conducted."
The research team is also considering the establishment of a real-time flood risk monitoring website based on AI. By analyzing citizens' responses during disasters through social media (SNS) or app data, they are developing a system to help government and local authorities respond more agilely to complaints and demands.
Professor Gam stated that for AI to advance in analyzing and preparing for disasters, research infrastructure is essential. He noted, "Pohang University of Science and Technology POSTECH has its own server cluster enabling this, but most local universities and small research organizations struggle to even start their research," urging that "the government should invest in equipment and infrastructure in the long term."
Pohang University of Science and Technology POSTECH is currently the only university in the country operating a school-level cluster center. It has a structure where servers are installed for research labs in a space equipped with temperature and humidity control and emergency power systems, currently operating three clusters and preparing for a fourth.
Professor Gam said, "Floods are not frequent disasters, but they cause significant damage each time they occur," and noted, "AI can be a powerful tool in managing such uncertainties. The government and society need to actively support the application of AI in environmental fields."
References
Journal of Environmental Management (2025), DOI: https://doi.org/10.1016/j.jenvman.2025.125640