AI for Flood Damage Prevention: Standing Strong Against Natural Disasters
▲ Senior Research Fellow Yoon Kwang-seok, Department of Hydro Science and Engineering Research (AI Flood Forecasting Research Team), KICT
Last summer, the Korean Peninsula was hit hard by severe torrential rains. The monsoon front, which began on June 25, persisted until July 26. As the water level of the Seomjin River Dam neared its flood control limit of 194 meters, the dam began releasing water at a rate of up to 300 tons per second. This severe flooding led to significant loss of life and property damage. This flood, and other floods of 2000, 2022, and 2023 have underscored the growing threat of natural disasters driven by climate change.
AI-based Flood Forecasting System Enabling Rapid Decision-Making
The Department of Hydro Science and Engineering Research at the Korea Institute of Civil Engineering and Building Technology (KICT) identified urbanization and the high population density in developed cities as the primary causes of flooding, attributing it to a reduction in areas for rainfall infiltration. The department also predicted that the flooding issue is likely to worsen in the future. To address this, an "AI-based Flood Forecasting System" was proposed as a new solution to thoroughly prepare for the extreme floods that can occur at any time.
‘If flood forecasting were conducted solely by human resources, predictions and warnings would rely on manual analysis, leading to slower decision-making and delayed crisis responses. Starting this year, the Ministry of Environment and the Flood Control Office have decided to adopt the KICT's AI-based flood prediction model to enable more efficient forecasting and warning systems. This marks the world's first implementation of an AI-driven flood prediction model.
The AI-based Flood Forecasting System consists of four stages: observation and investigation, transmission and prediction, prediction, and delivery. It automatically analyzes national flood forecasting points at 10-minute intervals and autonomously learns from big data on weather and hydrological conditions in the Han River basin. Flood forecasters verify AI-based prediction results, make a judgment on the situation, and issue flood warnings.
Enhanced Accuracy, Speed, and Stability in Flood Prediction
The AI-driven Long Short-Term Memory (LSTM) model applied to the system automatically predicts river water levels by learning statistical correlations from observational data, such as rainfall, water levels, and dam discharge volumes. This is a physical model that combines hydrological and hydraulic models, calculating river water levels using flow rates determined through the storage function method. Warnings are issued at points where water levels are predicted to exceed the warning threshold.
The prediction scope will soon be significantly expanded. Until 2023, predictions were limited to 75 flood warning points focused on major rivers, leaving tributaries and streams more vulnerable. Starting this year, the number of flood warning points will be increased to 223, covering tributaries and smaller streams. Currently, the AI-based flood forecasting model is used in four flood control offices, with plans for gradual expansion. Forecasters can quickly predict disasters, allowing for countermeasures to be taken promptly by using the system’s dam-river digital twin technology to simulate water level changes and pinpoint areas at risk of flooding.
Notably, the upstream points of rivers added as flood information provision points from this year have faster runoff speeds, making prediction difficult with conventional physics-based models alone. The AI-based flood prediction model assists in predicting and decision-making for such points. As this is the first application of an AI-based flood prediction model, the research team is continuing its research and development to achieve accuracy, speed, and stability. Senior research fellow Yoon Kwang-seok, the principal researcher, expects the AI-based flood prediction system to spread not only domestically but also globally.
"As this is the first time an AI-based flood prediction model is being applied in practice, the Department of Hydro Science and Engineering Research is focusing on research to advance the technology and improve its accuracy. In particular, we expect to increase the efficiency of flood prediction by linking with conventional physics-based models and establishing an improved decision-making system. Our goal is for the AI-based flood prediction system we developed to become the world's best system."
"Our goal is for the AI-based flood prediction system
we developed to become the world's foremost system."
KICT's Technology Expanding Globally
The research team’s focus extends beyond flood-related issues. Last year’s torrential rains caused severe damage and casualties in areas like Gangnam Station and Sillim-dong in Seoul. To respond, the team is analyzing past damage caused by urban inundation and actively conducting research on monitoring and predicting urban inundation damage, with plans to continue this work through 2025.
The goal is to develop flood monitoring equipment capable of measuring inundation depths in urban areas. Furthermore, the team plans to create a model that predicts inundation based on the monitoring results. These developments will be tested in countries such as the Philippines, Indonesia, and Laos to verify their adaptability. The researchers are committed to their work, believing that these advancements will pave the way for domestic technologies to reach international markets
Since its establishment, the KICT's Department of Hydro Science and Engineering Research has worked consistently to address national water-related issues, such as floods, droughts, climate change, and coastal disasters, while preserving the value of the national territory. The department believes that the AI-based Flood Forecasting System will improve citizens' quality of life and lead to more effective water management. Driven by a sincere desire for a better world, the team’s research will continue to bring about meaningful changes.
