Author(s): Donghee Lee; Illwon Jung; Jaeyoung Yoon
Linked Author(s):
Keywords: NN; Climate indices; Dam inflow forecast; MLR; SVM; Teleconnection
Abstract: Water supply of South Korea is highly dependent on dam operation because dams provide approximately 65% of annual water demands. Therefore, reliable long-range dam inflow forecast is essential to successfully manage water supply from dams. This study developed dam inflow forecast models with three months lead time applied to Boryeong Dam watershed based on hydroclimatic teleconnection between monthly dam inflow and climatic variables. For this purpose, multiple lineal regression (MLR), support vector machines (SVM) and artificial neural networks (ANN) are used to develop dam inflow forecast models except wet season (June to October). Models have been applied on a leave-one-out cross-validation mode. To evaluate the ability of above model’s forecasts skill, expected error criteria S was used. The S values of the MLR model ranged from 0.21 to 0.55, the ANN model ranged from 0.20 to 0.52, and the SVM from 0.26 to 0.56. These are considered to be the satisfactory results. Our results indicated that teleconnection-based models using climate information have the potential to be used to predict dam inflow from a small basin with 3-month lead time for sustainable water resources management and could serve as a good supplementary tool to make a decision for water release in preparing for the droughts. However, further study deems necessary to incorporate local predictors such as flows and temperature in addition to teleconnection indices for improved performance.
Year: 2018