A numerical solute transport model was history matched to a high-resolution monitoring dataset to characterize a multicomponent source of nonaqueous phase liquids (NAPLs) and evaluate the uncertainty of estimated parameters. The dissolution of NAPL mass was simulated using the SEAM3D solute transport model with spatially-varying NAPL saturations and mass transfer rate coefficients, representing the heterogenous architecture of the source zone. Source zone parameters were simultaneously estimated using PEST from aqueous-phase concentrations measured in a multilevel monitoring transect and from mass recovery rates measured at extraction wells during a controlled field experiment. Data-worth analyses, facilitated by PEST ancillary software, linked maximum aqueous-phase concentrations of all compounds to reductions in prior uncertainty of mass transfer coefficients. In turn, transient concentrations of the most soluble NAPL fraction constrained the source mass estimation. Accurately estimating the source mass and reducing prior uncertainties was possible by removing concentrations measured during early NAPL dissolution stages, identified as prior-data conflicts using the iterative ensemble smoother PESTPP-iES. Prior-based Monte Carlo analyses highlighted model limitations for representing sub-grid-scale heterogeneity of source zone architecture and NAPL dissolution, yet history-matching of final dissolution stages measured at multilevel ports eliminated parameter bias and produced long-term projections of source depletion with multistage behavior. Including mass discharge constraints further improved the accuracy of source mass estimation, complementing multilevel monitoring constraints on the source architecture and mass transfer coefficients

Estimating dissipation timeframes and contaminant mass discharge rates of dense non-aqueous phase liquids (DNAPLs) source zones is of key interest for environmental-management support. Upscaled mathematical modeling of DNAPL dissolution provides a practical approach for assimilating site characterization and downgradient monitoring data to constrain future system behavior. Yet significant uncertainties on predictions of source zone dissipation rates may arise from inadequate or inaccurate conceptual assumptions in parameterization designs. These implications were investigated through upscaled modeling, sensitivity, and uncertainty analyses of high-resolution flow-cell experiments. Sensitivity results emphasized the role of local groundwater velocity and source dimensions in mass transfer scaling by strongly influencing error with respect to DNAPL persistence and dissolution rates. Linear uncertainty analyses, facilitated by PEST ancillary software, demonstrated the worth of monitoring profiles for constraining DNAPL saturations and dispersive mass transfer rates, responsible for source zone longevity. Nonlinear analyses performed with the iterative ensemble smoother PESTPP-iES, facilitated the quantification of unbiased source dissipation uncertainties from DNAPL delineation data. Conversely, monitoring data assimilation without consideration of flow-field heterogeneity and saturation distribution along the flow path biased model predictions. Our analyses provided practical recommendations on upscaled model design to assimilate available site data and support remedial-decision making.