Matching Solute Breakthrough with Deterministic and Stochastic Aquifer Models

Two different deterministic and two alternative stochastic (i.e., geostatistical) approaches to modeling the distribution of hydraulic conductivity (K) in a nonuniform (sigma2ln(K)) = 0.29) glacial sand aquifer were used to explore the influence of conceptual model selection on simulations of three-dimensional tracer movement. The deterministic K models employed included a homogeneous effective K and a perfectly stratified 14 layer model. Stochastic K models were constructed using sequential Gaussian simulation and sequential i ndicator simulation conditioned to available K values estimated from measured grain size distributions. Standard simulation software packages MODFLOW, MT3DMS, and MODPATH were used to model three-dimensional ground water flow and transport in a field tracer test, where a pulse of bromide was injected through an array of three fully screened wells and extracted through a single fully screened well approximately 8 m away. Agreement between observed and simulated transport behavior was assessed through direct comparison of breakthrough curves (BTCs) and selected breakthrough metrics at the extraction well and at 26 individual multilevel sample ports distributed irregularly between the injection and extraction wells. Results indicate that conceptual models incorporating formation variability are better able to capture observed breakthrough behavior. Root mean square (RMS) error of the deterministic models bracketed the ensemble mean RMS error of stochastic models for simulated concentration vs. time series, but not for individual BTC characteristic metrics. The spatial variability models evaluated here may be better suited to simulating breakthrough behavior measured in wells screened over large intervals than at arbitrarily distributed observation points within a nonuniform aquifer domain.