Streamflow Estimation in The Chi and Ping Basins Using an Artificial Neural Network Model

Abstract

Efficiently evaluating runoff is crucial for water management. Among the data-driven models applied to the rainfall-runoff relationship, the Artificial Neural Network (ANN) model stands out for its ability to estimate runoff based solely on meteorological data. However, its results have been deemed unsatisfactory, prompting several studies to improve its accuracy. Some of these have used antecedent streamflow or separating components (direct flow and base flow) to enhance accuracy. In this study, we used antecedent and cumulative rainfall predictors representing streamflow components to estimate runoff in the Upper Chi River basin and Upper Ping River basin. The results demonstrate that the ANN model has the potential for accurately assessing runoff, particularly in the upper Chi River basin with the average NSE of 0.87, while the upper Ping River basin has an average NSE of 0.75. Furthermore, the ANN model outperformed the NAM model in terms of NSE R2 RMSE, except for KGE.