The Ensemble Streamflow Prediction (ESP) technique, most widely known as a way to express the uncertainty of hydrological prediction, began to be studied in Korea at the start of the 21st century. Since then, the flood control office and K-water, which officially make hydrological predictions in Korea, have been conducting flow probability predictions with ESP. Unfortunately, however, hydrological probability prediction has not yet led to dam operation. So, even after creating several ensemble scenarios, they have been presenting a single scenario-based dam operation plan using ESP’s average or median values. This study aims to suggest how optimal release changes according to the different spreads of the ESP. First, the 21 multi-purpose dams in South Korea were grouped according to the ratio(r) of average annual inflow to the total storage capacity, and Dam Jangheung (JH,r=0.8), Dam Imha (IH,r=1.1), and Dam Chungju (CJ,r=1.8) were selected from each group. Then, the monthly inflow scenarios were calculated using ESP, and spread values were then evaluated. Subsequently, using the predicted monthly inflow ensemble scenarios, statistical analysis, and probability prediction were performed to generate the final inflow scenarios for simulation. The optimal solution for water utilization from October to June was computed using quadratic programming according to the generated scenarios. The optimal solutions were finally compared with the historical release rules as well as the solutions derived from the ESP scenarios with both large and small spreads. The result from this study suggested that a novel reservoir operation strategy is more effective in case of future inflow with larger spreads.
