Effects of Model Formulation and Calibration Data on Uncertainty in Dense Nonaqueous Phase Liquids Source Dissolution Predictions
Water Resources Research
Simplified models have been proposed to describe dense nonaqueous phase liquids (DNAPL) source mass discharge versus time assuming that aqueous mass transfer is proportional to the DNAPL mass remaining raised to the power of a "source architecture" coefficient. Recent studies suggest that source discharge predictions may be improved using multiple functions for different architectures. However, multiple-source functions require estimation of additional parameters, which can be difficult to do from available field data. A hypothetical problem was considered that involves a DNAPL source zone consisting of seven discrete trichloroethylene (TCE) sources within a roughly 120,000 m2 area, each having different release dates, masses, and rate parameters. A transport model was used to simulate dissolved TCE concentrations in 22 monitoring wells over a 23 year period. "Noise" representing conceptual model errors and sampling or measurement error was applied to produce synthetic monitoring data. Source mass and dissolution rates were computed for the total source and for various depth intervals for a 5 year period and were also overlaid with suitable "noise." Data sets consisting of various subsets of the noisy monitoring and source data were employed to calibrate various model formulations. Results indicate that calibration using only near-source discharge rates yield highly uncertain estimates of future mass discharge rates. Prediction accuracy is improved and uncertainty is reduced by calibrating to whole plume data in conjunction with source discharge estimates, especially if the latter are available for multiple depths. More complex models reduce prediction uncertainty up to a point beyond which added complexity was a liability unless additional or more accurate data are available. Copyright 2010 by the American Geophysical Union.
Parker, Jack C.; Kim, Ungtae; Widdowson, Mark; Kitanidis, Peter; and Gentry, Randy, "Effects of Model Formulation and Calibration Data on Uncertainty in Dense Nonaqueous Phase Liquids Source Dissolution Predictions" (2010). Civil and Environmental Engineering Faculty Publications. 215.