Groundwater management in the presence of uncertainty using artificial neural networks and genetic algorithms

The problem addressed in this study is how to maximize the utilization of groundwater aquifers subject to quality constraints with subsurface heterogeneity and uncertainty. The quality constraint is imposed because of the presence of contamination sources near the aquifer. The problem is complicated by the uncertainty stemming from the heterogeneity of subsurface environments and that of the exact location of leaks from contaminant sources. The presence of conflicting objectives and uncertainty complicates the management process. We capitalize on the capabilities of evolutionary computational techniques, including artificial neural networks and genetic algorithms, for creating a comprehensive analytical package that can be used to manage and evaluate groundwater resources near potential contamination sources. The developed framework is found to be very efficient in evaluating millions of development scenarios that would not otherwise have been evaluated with traditional techniques. The main conclusion is that ANN and GA are robust techniques that can lead to improved management plans that maximize the benefits from groundwater aquifers while minimizing adverse impacts on the environment.