Science1 publisher2 min readPublished
Radar detection of stalled rainbands goes live in Seoul's flood forecasting pilot
KICT has put radar-based detection of quasi-stationary rainbands and sudden downpours into the Han River Flood Control Office's platform, now piloted in two Seoul districts. The two-hour forecast comes later.
The Scientist · Science desk
What happened
- KICT said it will apply its early detection technology for quasi-stationary linear rainbands and sudden localized torrential rainfall to the Urban Flood Forecasting Platform run by the Han River Flood Control Office.
- The detection has gone into the platform's observation and monitoring functions, and the platform is being piloted in areas including Gangnam and Kwanak in Seoul.
- KICT researchers have sat on the Ministry of Climate, Energy and Environment's Urban Flood Forecasting Task Force since November 2025, reviewing platform technologies and helping establish operational procedures.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- capability Drainage and emergency staff in the pilot areas get a display that flags a stalled band or a growing downpour while it is forming, so crews can be moved before stream levels rise.
- constraint Procedures written today have to work on detection, so usable response time is bounded by how long a band takes to reach the ground once radar can see it; the two-hour number is not yet available to plan against.
- decision District offices have to settle what a detection flag obliges them to do, and at what threshold, while the same task force that vetted the technology is still writing the operating rules.
- exposure Every action keyed to a flag inherits that flag's error rate, and the districts absorb the cost of each evacuation or crew deployment that turns out to have been unnecessary.
A quasi-stationary linear rainband is defined partly by its shape: a narrow, elongated band of precipitation that stays over one area and produces large amounts of rain [5]. Shape and persistence are things a radar rainfall field shows directly, and the paper behind the platform work is titled "Development of a Hazardous Weather Classification Algorithm Based on Radar Rainfall Data Using Morphological Features" [7]. The detector tracks a band's formation range and propagation path, and separately the initiation and development of sudden localized downpours, from radar observations alone [3].
What runs in the pilot today sits in the platform's observation and monitoring functions [2]. It reports hazardous rainfall as radar sees it form, and passes real-time detection results to operators for assessment and decision-making [10]. The forecast layer is still being built. Yoon Seong-Sim of KICT said, "Once the AI-based hazardous rainfall prediction technology is fully developed, it is expected to enable flood prediction up to two hours in advance, contributing to faster response and securing critical response time" [9]. The two-hour figure belongs to that system, and the source states no lead time for the detection now in the pilot [12].
The detection reports rain, and flooding depends on drainage. KICT's framing of the hazard runs through drainage capacity: a linear rainband can produce rainfall beyond what an urban drainage system carries, which is how the flooding happens [5], and a sudden localized downpour can raise water levels in urban streams fast enough to cause flash floods, rapid currents and incidents in which people are stranded [6]. Whether a particular catchment goes under depends on the state of its network as well as on the band overhead. The announcement does not report detection accuracy, false-alarm rates or results from the pilot [13].
The procedure work started before the deployment. KICT researchers joined the Urban Flood Forecasting Task Force of the Ministry of Climate, Energy and Environment in November 2025, reviewing the technologies the platform needs and helping establish operational procedures [4]. Operational procedure is where a research prototype turns into an operating layer, because someone has to write down what a detection flag obliges a district office to do, and at what threshold. Yoon said, "This achievement represents a notable example of putting into practice the hydrometeorological technology developed by KICT for urban water disaster response by applying it to urban flood forecasting" [8].
What to watch
- Whether the Han River Flood Control Office publishes hit and false-alarm rates for detection flags from the Gangnam and Kwanak pilot.
- Completion of the AI-based hazardous rainfall prediction, and whether the two-hour lead time holds under operational testing.
- Whether the task force's written procedures specify the action a detection flag triggers, and whether the pilot expands past two districts.