Research Information

Accurately and Timely Flash Flood Forecasting
  • Date2020-07-30
  • Hit212

Accurately and Timely Flash Flood Forecasting


Seok Hwan Hwang (Research Fellow, Flash Flood Research Center)


Korea and the rest of East Asia have suffered massive damage this year after experiencing the longest rainy season on record as well as repeated torrential downpours. These challenging experiences have highlighted the importance of being able to accurately predict sudden flash floods. The Flash Flood Research Center at the KICT is currently developing a system for predicting localized flash floods and providing early warning notifications for urban inundation. Once implemented, the system will be able to quickly and accurately identify flash floods at least one hour before they actually occur, giving residents a chance to evacuate or take other precautions. The system is expected to help drastically reduce the loss of life and property due to water damage during the summer.


Flood Forecasting in Populated Areas

Numerous water-related disasters have occurred in Korea in the summer of 2020. Rivers flowing through the city overflowed, landslides buried human dwellings in mud and soil, and apartment complexes were inundated with flood water. This type of water damage happens every year, resulting in the loss of both lives and property. These accidents occur suddenly, without warning, making them even more damaging.

Dr. Seok Hwan Hwang(Ph.D.), who is the head of the Flash Flood Research Center, has met with numerous public servants working for related authorities, and they have unanimously stated that even less than an hour’s warning would be hugely effective in preventing the loss of human life.

“Public servants working with related authorities have consistently called for threat prediction information for disaster response operations. Predicting threats may be difficult, but it is also essential. At this point in time, flood forecasts are provided only for the country’s major rivers, but flood damage reaches far beyond these major rivers. That is why there is a need for flash flood warnings to cover the entire country. How safe Koreans feel should not be dependent on where they live.”


▲ [{hoto 1] Head of the Flash Flood Research Center, Dr. Seok Hwan Hwang(Ph.D.)

The Flash Flood Research Center at the KICT is currently developing a system to predict localized flash floods and provide early warnings for urban inundation. Effective forecasting of localized flash flooding across vast areas involves the use of rainfall data gathered remotely using radar technology. Accurate flash flood forecasting is only possible using highly accurate rainfall measurements obtained remotely by radar. The Flash Flood Research Center is currently developing technology to enhance the accuracy of such technologies.

“The Flash Flood Forecasting System uses meteorological radar data to predict rainfall. However, the accuracy ofradar-based predictions decreases significantly when making forecasts more than hour into the future. In order to overcome these limitations, we are developing a technology that combines short-term radar prediction data with numerical weather prediction data so that we can ensure accurate rainfall predictions up to three hours in advance.”

The localized flash flooding prediction system under development is currently undergoing field testing at the Ministry of Environment’s Flood Control Office with the aim of putting the system into field operations by 2021. Dr. Hwang introduces the localized flash flooding prediction system as a technology that brings flood forecasting into people’s“everyday living sphere.” Once implemented, the technology will be able to provide a one-hour warning for flash floods anywhere in the country.

“Another characteristic of the system is that it analyzes open information on the web and social media to self-diagnose flood threat prediction errors. This will allow the accuracy of the system’s flood predictions to increase as the system matures. We are also developing techniques to track torrential rainfall to predict the spread and proliferation of flood risk. We are also working to develop techniques to calculate, early on, based on cloud formations, whether there is a possibility that the resulting rainfall will develop into torrential downpours that are likely to cause flash flooding.”

Dual-polarization Technology for Enhanced Accuracy
Flash floods are caused by convective precipitation, which develops vertically, or by precipitation fronts, which develop in one area and then move to another. The flooding that happened in Busan and Daejeon this summer developed in this same way. This past summer, many people wondered why flooding happened in the heart of the city, away from the rivers. Dr. Hwang explains that flood characteristics, damage, and risk levels vary substantially depending on the topology.

“In low-lying urban areas, such as the Gangnam district of Seoul, the amount of water that gathers in one place and the speed at which it flows will be completely different than in a village on Mt. Jirisan. That is why current special weather reports based on precipitation data often fail to predict the flash flood conditions that people actually experience in real life. This type of data is insufficient in providing early warning for sudden flash floods.”

Localized flash flood forecasting isa process that is divided into three steps: 1) First,  rainfall is observed using meteorological radars, and rainfall predictions are made up to three hours in advance using observed data; 2) Second, any flood (inundation) risk is calculated using predicted rainfall and the hydrologic characteristics of each region, including topography; 3) Third, the predicted rainfall and flood (inundation) risk data is used to assess flash flood risk levels and provide flash flood warnings in realtime.

The system under development by the Flash Flood Research Center uses existing flood damage data and detailed regional flood characteristics to extract factors that determine flood depth and degree of flood damage in various regions under varied conditions. The formula derived using this process is used to predict flash flooding risk for areas with no prior flooding data. A dual-polarized radar system is employed to boost the accuracy of the observation data.

“Dual polarization is a method of emitting radar beams to observe airborne precipitation particles. In the past, a single-polarized beam that oscillates only in a horizontal direction was emitted to observe precipitation. Even though single-polarized radar is able to measure the intensity of precipitation based on the strength of the reflected radar signals, this technology is unable to distinguish between different types of precipitation, such as rain, snow, or sleet. However, a dual-polarized radar simultaneously emits horizontally and vertically oscillating beams, which allows for the accurate determination of the type of precipitation, including rainfall.”

Dr. Hwang also points out that flood risk is manifested through a dynamic process of precipitation, flooding, surging, and landslides. This process was also seen in the flood damage caused by extended periods of torrential rainfall this past summer. Dr. Hwang also notes that the special rainfall alerts, flooding risk alerts, river overflow warnings, and landslide risk forecasts that are currently provided to the public are given individually, and fail to take into consideration all the factors of a given region. With such poor warning systems, people who live in areas with underdeveloped disaster response infrastructure cannot be expected to effectively cope with flood-related disasters. Dr. Hwang also stresses that the state needs to provide general flood risk forecast data to local governments and municipalities, so that they can also engage in disaster prevention and response efforts for small river overflows and landslide risks in upstream areas as well as for flood risks in downstream urban areas.

“Local governments need to be able to utilize this information and provide feedback on its effectiveness, as well as to give suggestions for improvement. An organically-linked system allows for the continual enhancement of prediction accuracy. By implementing an accurate prediction system, we will be able to prevent overlapping government investments in disaster prevention systems, and ensure that all Koreans are guaranteed the same standards of safety, regardless of their economic standing.”


▲ [Photo 2] Datafrom meteorological radars used to predict precipitation

Efforts Toward a Stronger Forecasting System

The Flash Flood Research Center is working on developing original flash flood forecasting technologies, based on quaternary convergence technologies, and implementing a national flash flood forecasting hub under the motto “Confidence in Technology Toward a Better Life.”

Once implemented, the flash flood forecasting system currently under development will be capable of generating applied forecasting data that considers regional and user characteristics. This will enable optimized flood forecasting tailored to each user type (i.e. industrial or agricultural). It will also allow the central and local governments to provide accurate and detailed flash flood forecasts one to three hours in advance at the local level. These advancements are expected to greatly improve the general safety of the Korean public in the face of flood risks.

“We assessed the system using urban flooding cases from 2020 and found that the system was capable of accurately predicting flood risks in numerous cases. However, our assessments also helped us identify a number of areas for improvement. After we address these areas, the system will be able to more accurately and reliably predict flash flood risks. Differences in urban and mountain village environments mean that calculated risk levels may be manifested entirely differently in real life depending on the area. Differences between areas are caused by artificial structures and varying degrees of urbanization. The flash flood forecasting system we are currently developing will be even more accurate next year as we make further improvements to address area differences.”


▲ [Photo 3] Radar meteorology station onSeodaesan Mountain