Artificial neural network based predictionmodel forestimating the water quality of the river Ganga

Abstract

The Ganga River is the national river of India and recently got the right of living entity. It is necessary to monitor its water quality frequently. But monitoring and assessment of a river WQ is a provocative task in ecological management. In this study, soft computing based one of the computation technique called Artificial Neural Network (ANN) is used to developpredictionor forecasting modelfor estimating the WQ of the river Ganga in the stretch from Devprayag to Roorkee, Uttarakhand, India. Five sampling stations along this river stretch areselected. The monthly data sets of five water quality parameters temperature, pH, dissolved oxygen, biochemical oxygen demand and total coliform is usedfrom year 2001 to 2015. In developing the ANN model for the WQ prediction, the feed forward error back propagation method is used to develop predictionmodel for conducting various experimental investigationsfollowing many training parameters by adopting the neural network configurationof 5-10-1. The prediction performance of the developed model is evaluated using one of the statistical means named mean square error. The predicted WQ model resulted in a higher accuracy by producing best MSEvalue of 0.041 when using trainlm(Levenberg-Marquardt backpropagation)as a training function. The present study concluded that the ANNs are capable of predicting WQ of the river Ganga with acceptable accuracy of 95.9%.