WITHDRAWN: Influence of Porous Media Heterogeneity on NAPL Dissolution Fingering and Upscaled Mass Transfer

Abstract

AbstractThe utility of existing models for describing upscaled mass transfer from non‐aqueous phase liquid (NAPL) were examined when preferential dissolution pathways form in NAPL‐contaminated zones. Such models are needed because the coarse discretizations often used in simulators cannot capture the details of local‐scale preferential dissolution pathways.Laboratory experiments were conducted in two well‐characterized, heterogeneous packings that differed only in the correlation lengths of the permeability field. Experimental results were used to validate a numerical simulator for capturing the growth of centimeter‐scale preferential NAPL dissolution patterns. Using data from these experiments and simulations, three methods for upscaling the mass transfer rate coefficient for NAPL dissolution [Imhoff et al., 2003; Saenton and Illangasekare, 2007; Christ et al., 2006] and an equilibrium stream tube model for predicting contaminant flux [Basu et al., 2008] were evaluated. These models account for the influence of dissolution fingering [Imhoff et al., 2003] or NAPL architecture [Saenton and Illangasekare, 2007; Christ et al., 2006; Basu et al., 2008] on contaminant flux from NAPL source zones, and were applied here to local‐scale preferential NAPL dissolution. When the correlation length of permeability perpendicular to the mean water flow direction was 6.0 cm, greater than the scale of the dissolution fingers, only 4.2% of the NAPL resided in pools. Dissolution fingers formed in this experiment and all models predicted effluent concentrations with similar accuracy, with root mean square errors (RMSE's) between 0.03 and 0.05. When the correlation scale was smaller (1.0 cm), 31.3% of the NAPL was in pools and preferential dissolution pathways were dominated by channeling, preferential dissolution caused by spatial variations in aqueous phase velocity and NAPL dissolution rates. For this experiment all models performed poorly, with RMSEs between 0.12 and 0.35. Dissolution fingering was important when the permeability correlation length was sufficiently large that dissolution finger formation was not disrupted and NAPL pools were not dominant.