Flood forecasting for the upper reach of the Red River Basin, North Vietnam

Flood forecasting remains a very important task. Good forecast values with sufficient lead times can help reduce flood damages significantly. This paper proposes two types of black-box model obtained by using multiple regression analysis and backpropagation neural networks in forecasting 6-h water levels at three important stations on the upstream section of the Red River basin, North Vietnam. The results obtained show that highly accurate forecast values can be obtained with lead times of up to 18 h by using two most recent past values of the water level at the station considered or two most recent past values at this station and two most recent values of an upstream station.