CS(AR)-16/99-2000 : Rainfall-runoff modelling using artificial neural network technique

Abstract

The artificial neural network (ANN) methodology has been reported to provide reasonably good solutions for circumstances where there are complex systems that may be poorly defined or understood using mathematical equations, problems that deal with noise involve pattern recognition, and situations where input data are incomplete or ambiguous by nature. Because of these characteristics, it was believed that ANN could be applied to model the daily rainfall runoff relationship. Accordingly, a research study was conducted by employing ANN computing approach to forecast daily runoff as a function of daily precipitation and previous values of runoff. The model was trained and tested for the data of the Baitarani River Basin, Orissa. Two ANN algorithms were considered while developing the model, namely back error propagation network (BPN) and radial basis function network (RBF). The sensitivity of the prediction accuracy to the number of hidden layer neurons in a back error propagation algorithm was investigated. Based on this analysis, two BPN models were selected to represent the rainfall-runoff transformation. These two BPN models and the RBF model were compared for their performance using various statistical indices. The performance ANN model for Baitarani river basin was compared with that of existing models. The study demonstrates the applicability of ANN approach in developing effective non-linear models of Rainfall Runoff process without the need to explicitly represent the internal hydrologic structure of the watershed. The developed ANN model was found performing to a good degree of accuracy compared to other models in use.