Author/Authors :
Mohd Nasir، Mohd Fahmi نويسنده 1Department of Environmental Sciences, Faculty of Environmental Studies, UPM Serdang, Selangor, Malaysia , , Abdul Zali، Munirah نويسنده 1Department of Environmental Sciences, Faculty of Environmental Studies, UPM Serdang, Selangor, Malaysia , , Juahir، Hafizan نويسنده 1Department of Environmental Sciences, Faculty of Environmental Studies, UPM Serdang, Selangor, Malaysia , , Hussain، Hashimah نويسنده Department of Environment, Federal Government Administrative Centre, Environment Institute of Malaysia, Putrajaya, Malaysia , , M Zain، Sharifuddin نويسنده Department of Chemistry, Faculty of Science, Universiti Malaya, Kuala Lumpur, Malaysia , , Ramli، Norlafifah نويسنده Surface Water Monitoring Unit, Water and Marine Division, Department of Environment Malaysia, Federal Government Administrative Centre, Putrajaya, Malaysia ,
Abstract :
Recent techniques in the management of surface river water have been expanding the demand on the method
that can provide more representative of multivariate data set. A proper technique of the architecture of artificial
neural network (ANN) model and multiple linear regression (MLR) provides an advance tool for surface water
modeling and forecasting. The development of receptor model was applied in order to determine the major
sources of pollutants at Kuantan River Basin, Malaysia. Thirteen water quality parameters were used in principal
component analysis (PCA) and new variables of fertilizer waste, surface runoff, anthropogenic input, chemical and
mineral changes and erosion are successfully developed for modeling purposes. Two models were compared in
terms of efficiency and goodness-of-fit for water quality index (WQI) prediction. The results show that APCS-ANN
model gives better performance with high R2 value (0.9680) and small root mean square error (RMSE) value (2.6409)
compared to APCS-MLR model. Meanwhile from the sensitivity analysis, fertilizer waste acts as the dominant
pollutant contributor (59.82%) to the basin studied followed by anthropogenic input (22.48%), surface runoff
(13.42%), erosion (2.33%) and lastly chemical and mineral changes (1.95%). Thus, this study concluded that receptor
modeling of APCS-ANN can be used to solve various constraints in environmental problem that exist between
water distribution variables toward appropriate water quality management.
