Author(s): Sundarambal Palani; Pavel Tkalich; Rajasekhar Balasubramanian; Jegathambal Palanichamy
Keywords: Tmospheric deposition; Nutrients; Eutrophication; Water quality management; Neural network
Abstract: In this paper a recurrent neural network model for the prediction of atmospheric deposition of nutrients such as total nitrogen (TN) and organic nitrogen (ON) concentrations onto coastal waters of Singapore is proposed. ANN models were developed using a combination of meteorological and time-scale input variables. A statistical procedure for the selection of the input variables has been evaluated. Experimental trials show that the developed and tuned model is appropriate, giving small values of root mean square error (RMSE), mean absolute error (MAE) and mean square error (MSE). The results of the neural network models were rather satisfactory; with values of the Nash–Sutcliffe coefficient of efficiency (E) ranging 0.9to 0.98 and 0.7 to 0.8 for independent over-fitting test and validation sets respectively. Their performance was also found adequate in the case of high-concentration events, with acceptable minimum error. In addition, the related correlation coefficient ranges from 0.91 to0.96 for both predicted N species, underlying a small difference between the predicted and the measured values. The developed ANN models could be used as a convenient forecasting tool which complements current TN and ON analysis within the atmospheric deposition monitoring program carried out in the region. The successful and on-time predictions of episodic atmospheric nitrogen deposition concentration are of particular importance for the pollution loading estimation onto the surface water and to forecast the possible impact on aquatic ecosystem.