Filtered Stepwise Clustering Method for Predicting Fate of Contaminants in Groundwater Remediation Systems: A Case Study in Western Canada

This paper presents the development of a filtered stepwise clustering (FSC) method for facilitating the pump-and-treat (PAT) designs in groundwater remediation programs. To investigate the performances of different remediation strategies, a subsurface model is employed to simulate contaminant transport. Multivariate relationships between decision variables and selected modeling outputs are developed through the FSC method. Based on the developed statistical relationships, a set of possible outcomes for the remediation design can be presented; the solution space has been confined to a narrowed range. The proposed method can aid the PAT design by (a) quickly providing predicted outcomes given different remediation strategies and (b) directly locating the optimum remediation strategy for any outcome. The FSC method is examined through its application to a real-world aquifer remediation case in western Canada. The prediction results can help decision makers to evaluate the remediation design in an explicit way.