HR: 0830h
AN: H51D-02    [Abstracts]
TI: Ensemble Streamflow Prediction in Korea: Past and Future 5 Years
AU: * Jeong, D
EM: jung922@snu.ac.kr
AF: School of Civil, Urban and Geosystems Engineering, Seoul National University, San 56-1, Shillim-Dong, Gwanak-gu, Seoul, 151-742 Korea, Republic of
AU: Kim, Y
EM: yokim05@snu.ac.kr
AF: School of Civil, Urban and Geosystems Engineering, Seoul National University, San 56-1, Shillim-Dong, Gwanak-gu, Seoul, 151-742 Korea, Republic of
AU: Lee, J
EM: myroom1@snu.ac.kr
AF: School of Civil, Urban and Geosystems Engineering, Seoul National University, San 56-1, Shillim-Dong, Gwanak-gu, Seoul, 151-742 Korea, Republic of
AB: The Ensemble Streamflow Prediction (ESP) approach was first introduced in 2000 by the Hydrology Research Group (HRG) at Seoul National University as an alternative probabilistic forecasting technique for improving the 'Water Supply Outlook' That is issued every month by the Ministry of Construction and Transportation in Korea. That study motivated the Korea Water Resources Corporation (KOWACO) to establish their seasonal probabilistic forecasting system for the 5 major river basins using the ESP approach. In cooperation with the HRG, the KOWACO developed monthly optimal multi-reservoir operating systems for the Geum river basin in 2004, which coupled the ESP forecasts with an optimization model using sampling stochastic dynamic programming. The user interfaces for both ESP and SSDP have also been designed for the developed computer systems to become more practical. More projects for developing ESP systems to the other 3 major river basins (i.e. the Nakdong, Han and Seomjin river basins) was also completed by the HRG and KOWACO at the end of December 2004. Therefore, the ESP system has become the most important mid- and long-term streamflow forecast technique in Korea. In addition to the practical aspects, resent research experience on ESP has raised some concerns into ways of improving the accuracy of ESP in Korea. Jeong and Kim (2002) performed an error analysis on its resulting probabilistic forecasts and found that the modeling error is dominant in the dry season, while the meteorological error is dominant in the flood season. To address the first issue, Kim et al. (2004) tested various combinations and/or combining techniques and showed that the ESP probabilistic accuracy could be improved considerably during the dry season when the hydrologic models were combined and/or corrected. In addition, an attempt was also made to improve the ESP accuracy for the flood season using climate forecast information. This ongoing project handles three types of climate forecast information: (1) the Monthly Industrial Meteorology Information Magazine (MIMIM) of the Korea Meteorological Administration (2) the Global Data Assimilation Prediction System (GDAPS), and (3) the US National Centers for Environmental Prediction (NCEP). Each of these forecasts is issued in a unique format: (1) MIMIM is a most-probable-event forecast, (2) GDAPS is a single series of deterministic forecasts, and (3) NCEP is an ensemble of deterministic forecasts. Other minor issues include how long the initial conditions influences the ESP accuracy, and how many ESP scenarios are needed to obtain the best accuracy. This presentation also addresses some future research that is needed for ESP in Korea.
DE: 1800 HYDROLOGY
DE: 1833 Hydroclimatology
DE: 1860 Runoff and streamflow
DE: 2447 Modeling and forecasting
SC: Hydrology [H]
MN: 2005 Joint Assembly