Applications of back propagation neural networks in predicting nutrient effluent concentrations from permeable pavements

Artificial intelligence techniques such as neural networks are modelling tools that can be applied to analyse urban runoff water quality issues. Artificial neural networks are frequently used to model various highly variable and non-linear physical phenomena in the water and environmental engineering fields. The application of neural networks for analysing the performance of permeable pavements is timely and novel. Artificial neural networks are a promising tool for environmental process assessment and modelling. Feed-forward neural networks are the most widely adopted methodology for the prediction and forecasting of water quality and quantity variables. This paper presents the application of back-propagation neural networks and the testing of the Levenberg-Marquardt, Quasi-Newton and Bayesian Regularization algorithms. The back-propagation neural network models incorporating these algorithms performed classification and regression tasks without knowledge of the underlying physical processes occurring throughout the pavement system. The neural networks were statistically assessed for their goodness of prediction of outflow water quality with respect to ammonia-nitrogen, nitrate-nitrogen, and ortho-phosphate-phosphorus by numerical computation of the mean absolute error, root square mean error, mean absolute relative error and the coefficient of correlation for the prediction versus measured dataset. These performance indices compared the measured and estimated water quality parameters. The neural network models were functions of the readily available water quality parameters. Three neural network models were assessed for their efficiency in accurately simulating effluent water quality parameters from various experimental pavement systems. The models predicted all key parameters with high correlation coefficients and low minimum statistical errors. The back-propagation and feed-forward neural network models performed optimally as pollutant removal predictors with regard to these two sustainable technologies. Using an input layer consisting of pH, temperature, electric conductivity, turbidity, total dissolved solids, dissolved oxygen, and redox potential, the back-propagation feed forward neural network models performed optimally as nutrient predictors with regard to sustainable (urban) drainage systems such as permeable pavements. The results showed that for varied effluent concentrations of nutrients, applications of artificial neural networks computed a convincing outcome.