INTEGRATION OF DETERMINISTIC AND STOCHASTIC METHODS TO MODEL THE HYDROSTRATIGRAPHIC ARCHITECTURE OF GLACIAL SEDIMENTS IN ANN ARBOR, MICHIGAN, USA

Glacial aquifer systems are notoriously complex, exhibiting a high degree of variability in lithologies and hydrogeologic properties over relatively small distances. Consequently, efforts to characterize hydrostratigraphic architectures in remedial investigations are often impeded by widely spaced subsurface control points. Hybrid approaches to hydrogeologic modeling seek to overcome data limitations by incorporating stochastic variability to address uncertainty within a deterministic stratigraphic framework. Geostatistical modeling relies upon the assumption of stationarity; hence a prerequisite to stochastic infill is the interpretation of a larger-scale sedimentary architecture. But to what degree does this primary, deterministic hydrostratigraphic architecture influence the outcome of groundwater flow and contaminant transport model predictions after stochastic variability is introduced?

This presentation describes the use of hybrid models to investigate the migration of 1,4-dioxane within an 80m thick Quaternary glacial aquifer system beneath Ann Arbor, Michigan, as it advances toward the Huron River through a groundwater Prohibition Zone established in 2005. Despite the installation of monitoring wells at more than 130 locations, subsurface control remains sparse with an average density of 7 wells per km² across an impacted area of approximately 20 km². An allostratigraphic approach, constrained by hydraulic head and contaminant concentration data, was used to interpret a three-dimensional distribution of aquifer and aquitard units using available well control. Sequential Gaussian simulation was subsequently used to populate the hydraulic conductivity field of aquifer and aquitard units within 30x30x3m cells in the central 14 km² area of a regional groundwater model. 100 stochastic realizations were ranked using harmonic mean K values for flow paths along the primary migration direction between the source area and the Huron River. We propose to compare relevant transport metrics (e.g., first arrivals and breakthrough times at the river) among realizations to evaluate the degree to which stochastic variability influences transport metrics and to test whether a priori rankings can be used to identify realizations with representative and extreme behaviors.