Nonequilibrium Sorption of Dimethylphthalate—Compatibility of Batch and Column Techniques

1986). In contrast, an apparent disagreement may be produced by applying an equilibrium model to data This study reconciles an apparent inconsistency between the dialthough sorption equilibrium was not reached within methylphthalate (DMP) batch and column data reported in the literature. Some researchers found that retardation coefficients obtained the time scale of the batch or column experiments. Failby fitting a linear two-stage model to column data were about 50% ure to reach sorption equilibrium is likely to occur since smaller than those calculated from the distribution coefficient of a sorption of nonionic organic compounds in soil is very 14-d isotherm. In this study, sorption parameters were derived indeslow (Streck et al., 1995). If sorption equilibrium is aspendently by fitting a linear and a nonlinear two-stage model simultasumed erroneously, incorrect equilibrium constants that neously to the 3and 14-d isotherms as well as to a sorption rate study are lower than the true equilibrium values will be estireported by others. With the sorption parameters and the dispersion mated (Brusseau et al., 1991). A direct estimation of coefficient obtained from a tritium tracer experiment, the column the equilibrium constant is difficult as equilibration may data were adequately predicted. A major part of the differences in take several weeks, months, or even years (Ball and the retardation coefficients observed in other research could be related Roberts, 1991). To overcome this problem inverse modto experimental difficulties in detecting the tailing of nonequilibrium breakthrough curves. eling can be used to estimate sorption rate and equilibrium parameters from column or batch data. In batch experiments, slow sorption rates result in continuously decreasing solute concentrations even O contaminants are introduced into the subafter the typical time span of the experiment. In column surface through diffuse sources, such as regionally experiments, slow sorption or desorption rates lead to applied agrochemicals, as well as point sources, includpronounced tailing of the breakthrough curves. For the ing accidental spills, hazardous waste sites, or landfills. case of a pulse input, the column experiment is termiDevelopment of remedial actions to deal with these nated when the quantification limit of the target comcontaminants requires understanding the physical, pound is reached. Breakthrough curves may lack much chemical, and biological processes affecting their mobilof the tail data, although most of the solute has actually ity. Among these processes, sorption is believed to be been recovered. Sorption parameters estimated by inan important mechanism that dramatically influences verse modeling of batch and column data are therefore their environmental fate. At the laboratory scale, batch subject not only to random errors but also to errors or packed column experimental techniques are comcaused by the necessity to use a truncated set of data monly employed to gain insight into sorption behavior specific to the experimental technique employed. of organic compounds. Results obtained from the laboMost studies that found deviating sorption parameter ratory are then utilized to predict actual transport in values between batch and column experiments are limthe field. ited to comparisons of parameter sets obtained from The sorption behavior in batch and column studies each experimental technique (MacIntyre and Stauffer, has been found to be in agreement in many studies (e.g., 1988; Lion et al., 1990; Piatt et al., 1996). However, Lee et al., 1988; MacIntyre et al., 1991; Gaber et al., as pointed out by Addiscott et al. (1995), this popular 1992). However, results disagreed considerably in others strategy (e.g., obtaining parameters by fitting to the data (e.g., Bilkert and Rao, 1985; MacIntyre and Stauffer, to be simulated) is the least desirable when attempting 1988; Piatt et al., 1996). The disagreement was attributed to identify shortcomings of either the model or the fit to different causes, such as loss of sorbent from the procedure. A more stringent model validation is the column, variations in column flow, immobile water in determination of all parameters independent from the the column (MacIntyre et al., 1991), mixing differences data to be simulated (Addiscott et al., 1995; Brusseau, between the two methods (Schweich et al., 1983), reduc1998). tion in soil particle spacing in the column compared For aquifer materials low in organic C such a strategy with batch systems (Celorie et al., 1989), different soil/ was applied by Larsen et al. (1992). The authors used water ratios (Karickhoff and Morris, 1985), and the posa two-compartment model to estimate kinetic sorption sibility of soil abrasion in batch experiments (Boesten, parameters from batch data and found good agreement S. Altfelder, Federal Institute of Geosciences and Natural Resources, between measured and predicted transport data. For Stilleweg 2, 30655 Hannover, Germany; T. Streck, Department of aquifer material restricted to a narrow size fraction, Geoecology, Technical University Carolo-Wilhelmina, Langer Kamp Young and Ball (1994, 1999) applied the same strategy 19 c, 38106 Braunschweig, Germany; M.A. Maraqa, Department of using a diffusion model. While Young and Ball (1994) Civil Engineering, UAE University, Al-Ain, United Arab Emirates; T.C. Voice, Department of Civil and Environmental Engineering, were able to predict transport data with good agreeMichigan State University, East Lansing, MI 48824. Received 13 Apr. 1999. *Corresponding author ([email protected]). Abbreviations: DMP, dimethylphthalate; LEA, local equilibrium assumption. Published in Soil Sci. Soc. Am. J. 65:102–111 (2001).