Optimal groundwater remediation design subject to uncertainty in hydraulic conductivity with regional geologic variations

Abstract

Groundwater pump-and-treat remediation designs where contaminant containment is the objective are developed using predictive groundwater flow models. The uncertainty of the hydraulic conductivity of an aquifer is included in the optimal cost remediation design by using a multi-scenario approach called robust optimization. Multiple regions within the optimization model are examined in this research. Each region represents a different geologic formation and as such is modeled with different hydraulic-conductivity fields. The variability of these fields is modeled using statistically distinctive distribution curves. In each scenario individual regions are represented with one hydraulic-conductivity value that is determined through sampling the associated distribution. Initially all combinations of all sampled hydraulic-conductivity values, say n values, for all regions, say r regions, are examined such that n r scenarios are considered in the robust-optimization model. This approach is then compared with a more simple approach whereby each distribution field is sampled n times and the hydraulic-conductivity values for each region in the k th scenario is assigned the k th lowest value sampled so that only n scenarios are examined in the robust-optimization model. The more simple approach to determining the hydrologic features of each scenario is an attempt to create a optimization model that is both computationally reasonable while representative of the uncertainty in the groundwater model.