Journal of Korean Society of Agricultural Engineers ISSN:1738-3692(Print) 2093-7709(Online) 한국농공학회논문집

As climate change intensifies extreme rainfall events, the risk of dam breaches in aging agricultural reservoirs is increasing. In Korea, Emergency Action Plan (EAP) is typically initiated using one-dimensional models such as DAMBRK, followed by high-resolution simulations for detailed analysis. This study proposes a method for predicting reservoir outflow hydrographs by integrating a Dynamic Time Warping–K-Nearest Neighbors (DTW-KNN) algorithm with reservoir outflow hydrographs generated by DAMBRK and FLOW-3D. After applying DTW-KNN, peak flow errors were generally reduced to within approximately 10%, and substantial improvements in peak timing accuracy were achieved, while the average Nash-Sutcliffe efficiency (NSE) increased from 0.34 to 0.93. Acceptable performance was also maintained at an extrapolated elevation of EL. 12.43 m, indicating that the trained model preserves reasonable generalization capability beyond the training range. These results demonstrate that the proposed DTW-KNN approach provides a practical, data-driven linkage between DAMBRK and FLOW-3D hydrographs, offering a reliable framework for rapid dam breach assessments under limited data conditions.