A Computationally Cost Effective Approach to Modeling Transport of Multicomponent VOC mixtures in the Vadose Zone. Validation With Field Data

A major difficulty that is typically encountered when real contamination problems are investigated is the typically large number of constituents in the actual VOC mixture compared to the number of species that can be handled numerically. An approach to dealing with complex mixtures of individual hydrocarbons while reducing the computational cost is to define a small number of effective composite constituents to simulate the average behavior and fate of groups of individual hydrocarbons in the subsurface. However, there are two additional sources of uncertainty associated with this technique that may influence the reliability of model predictions, namely, the time dependence of the effective properties of a composite constituent due to temporal composition changes, and the approximation in determining their effective property values. An approach to limit these sources of uncertainty was developed recently by the authors based on experimental data from a lysimeter study that involved artificially mixed kerosene. The minimization of the time dependence of the effective thermodynamic properties of the composite constituents is achieved through a methodology for optimizing component grouping, based on the minimization of an objective function. Simple arithmetic averages are found to provide poor approximations to the effective group properties and alternative algorithms are proposed to express the effective property values of composite constituents in terms of property values of the member components. The applicability of this approach to large-scale contamination problems is demonstrated here using data from field studies. Compared to the case of simulating all organics as individual compounds, our approach offers the advantage of a greatly reduced computational cost, reduced data requirements, as well as a smaller impact of the uncertainty in the compound properties on modeling results because effective values are used for each composite constituent, which follow a pdf with reduced variance compared to that of the individual values.