Artificial neural network: An alternative approach for assessment of biochemical oxygen demand of the Damodar River, West Bengal, India

Abstract Biochemical Oxygen Demand (BOD), representing the biodegradable organic load in water, is a prime parameter to assess water quality. Estimation of biochemical oxygen demand requires prolonged incubation of water samples. Thus, it is a time-consuming as well as energy-consuming process. Therefore, it is not possible to respond rapidly for mitigation if the BOD level goes beyond the permissible limit. In this study on River Damodar, the BOD value was predicted from electrical conductivity (EC), turbidity and chloride, using Artificial Neural Network (ANN)-based empirical model. Since predictor parameters of these models are measurable rapidly, the BOD value can also be predicted quickly. As all the predictor parameters are highly correlated (correlation coefficient: 0.9 or more) with BOD, the models are valid for prediction of BOD. Additionally, for further validation of the models, a portion of field data (20%) was used for model testing. As the model predictions are close to the actual values (r = 0.98, MAE = 0.43, RMSE = 0.57), the model, developed in this study, can be considered as successful in BOD prediction.