Prediction of water level in the Surma River of Bangladesh by Artificial Neural Networks

Abstract

Prediction of water level in a river system is of great interest for water management and flood control. In this study, using past rainfall and water level information, artificial neural networks (ANNs) with feed forward multi-layer perception have been designed and applied to predict the water level in one of the principal river (Surma) at agriculturally and ecologically important flash flood prone northeast hydrological region of Bangladesh. The network is trained by supervised learning and error back propagation algorithm. The obtained results have been used to calculate statistical indices for assessing the accuracy of prediction. The results indicate that ANNs are capable of mapping the nonlinear correlation between rainfall and water level.