Modelling nitrogen composition in streams on the Boreal Plain using genetic adaptive general regression neural networks. (5th November 2008)

• Record Type: Journal Article
• Authors: Li, Xiangfei; Nour, Mohamed H.; Smith, Daniel W.; Prepas, Ellie E.
• Is Part Of: Journal of environmental engineering and science. Volume 7(2008)Supplement 1
• Page Start: 109
• Page End: 125
• Publication Date: 2008-11-05
• Subjects: environment -- genetic adaptive general regression neural network -- nitrate -- ammonium -- total dissolved nitrogen -- surface water quality modelling -- watershed; réseaux de neurones à régression générale adaptatifs génétiques (GA-GRNN) -- nitrate -- ammonium -- azote dissous total -- modélisation de la qualité des eaux de surface -- bassin récepteur; Environmental engineering -- Periodicals; Sanitary engineering -- Periodicals; Environmental engineering -- Canada -- Periodicals; Environnement, Technique de l' -- Périodiques; Technique sanitaire -- Périodiques; Environnement, Technique de l' -- Canada -- Périodiques; Environmental engineering; Sanitary engineering; Canada; Electronic journals; Computer network resources; Periodicals; 628.05
• DOI: 10.1139/S08-041
• Languages: English
• ISSNs: 1496-2551

Abstract

Abstract : Increased release of nitrogen to hydrological networks due to watershed disturbance may cause aquatic problems and affect water uses. Therefore, effective nitrogen modelling is an important element of total watershed management. The objective of this study was to develop an artificial neural network modelling tool to predict nitrogen concentrations in streams using easily accessible data as model inputs. Genetic adaptive general regression neural network (GA-GRNN) models were applied to predict nitrate, ammonium, and total dissolved nitrogen concentrations in three forested watersheds in Alberta, Canada. The performance and generality of the developed models for dry and wet weather conditions in the studied watersheds were verified by the coefficient of multiple determination, the root mean squared error, swapping the testing and validation data sets, and plotting measured and predicted values over time. The successful application of GA-GRNN models to predict nitrogen compositions in the watersheds by using five major input variables and relevant time-lagged inputs, fully demonstrated the models' generality. It implies the high potential of applying GA-GRNN models for predicting other surface water quality parameters on other watersheds with similar or different characteristics.