Methods to Identify Non-Point Contaminant Flux to Groundwater : Comparison of Properties of ANN, LGA and IGA methods

概要

論文の詳細を見る
Contamination of groundwater due to non-point source, e. g., excessive fertilizer use in farming, is becoming a serious problem in many parts of the world. Quality of groundwater is often monitored at wells for its appropriate management. However, methodologies to identify the incoming flux values inversely from the monitored data are not completely established yet and still developing. In this research, three types of the identification methods, i. e., artificial neural network (ANN), lumped genetic algorithm (LGA) and iterative genetic algorithm (IGA) methods are presented and their applicabilities, convergencies and accuracies are discussed. Through the demonstrative applications, it is found that all the models possess quite satisfactory abilities of the identification. The IGA method shows better performance especially in the spatiotemporal problems, the LGA method requires less computational burden and the ANN method is useful in the problems without initial conditions. As a result, all these methods are expected to contribute to the management of groundwater quality.
日本雨水資源化システム学会の論文