SOLUTE TRANSPORT MODELING UNCERTAINTY: THE IMPORTANCE OF ALTERNATIVE CONCEPTUAL GEOLOGIC MODELS FOR CHARACTERIZING HETEROGENEITY

Insufficient subsurface data often contribute uncertainty to groundwater flow and contaminant transport modeling efforts. Fundamental uncertainty in the form of alternative conceptual models can be difficult and costly to evaluate if it requires the development and calibration of parallel models constructed with significantly different assumptions about the distribution of aquifer properties or boundary conditions. Yet such a level of effort may be necessary to ensure that remedial actions are robust under the complete range of possible conditions. In this investigation, numerical groundwater flow and contaminant transport modeling were conducted to evaluate remedial alternatives and predict contaminant concentrations of cis-DCE, benzene, and MTBE in an unconfined glacial aquifer located in the Village of Milford, Michigan, USA. Limited subsurface control contributed geologic uncertainty related to the geometry and continuity of a regional aquitard unit underlying the aquifer at the site. Consequently, two alternative conceptual models were constructed and independently calibrated to evaluate this uncertainty: a regional aquitard and an extended aquifer model.

Advective transport modeling demonstrated significant differences in predicted solute transport pathways between the regional aquitard and extended aquifer conceptualizations. However, subsequent simulation of advective-dispersive-reactive transport using these models predicted similar order-of-magnitude breakthrough curve responses in the remedial pumping wells and municipal water supply wells for each contaminant species at the site. Screened intervals for these wells ranged from approximately 30 to 100 feet. Thus, while the choice of conceptual model demonstrably affected predicted contaminant pathways, it had only a limited effect upon transport model predictions that were integrated over pumping wells screened across multiple model layers. In contrast, measurable changes in contaminant concentrations have been observed in monitoring wells screened over four to five foot intervals following the start of remedial pumping at the site. Therefore, the choice of conceptual model has important practical implications for evaluating the placement and performance of monitoring wells with smaller screened intervals.