Granular Activated Carbon (GAC) filters are frequently designed based on the results of the expensive and time-consuming pilot plant experiments. This makes it attractive to explore the mathematical modeling possibilities for GAC filter performance. If a target compound is not biodegraded in GAC filters, there is no essential difference between the models that can be applied to predict its removal in filters with and without enhanced bioactivity. In both cases, model parameters are determined either from the breakthrough of the target compound in pilot plant experiments or from the lab- and bench-scale experiments with preloaded GAC. Thus, in principle, currently applied adsorption models indirectly account for the biodegradation of Background Organic Matter (BOM) that takes place in biologically active filters.

Two generally available models for the adsorption in GAC filters, the simple Adams-Bohart (AB) model and the Plug Flow Homogenous Surface Diffusion (PFHSD) model, were applied to predict the breakthrough of atrazine in pilot plant GAC filters operated. The parameters of the AB model, removal capacity and removal rate, were determined from the breakthrough of atrazine observed during the initial six months of the pilot plant operation. The parameters of the PFHSD model were determined from the experiments with GAC preloaded for seven months. GAC’s adsorption capacity was determined from adsorption isotherms for (pulverized) preloaded GAC, while atrazine’s mass transfer coefficients were determined from the Short Fixed Bed tests with GAC particles.

The accurate prediction of the GAC filter performance proved not yet possible. The simple Adams-Bohart model allowed a close description of the initial atrazine breakthrough at the EBCT of 7 minutes but, when extrapolated, resulted in much faster breakthrough than observed during the remaining 1.5 year of the filter operation. The model was not applied for the EBCT of 20 minutes, because the effluent atrazine concentrations were too low to allow for model calibration. The PFHSD model provided a less inaccurate prediction of atrazine breakthrough at the EBCT of 7 minutes than the Adams-Bohart model, but predicted much lower breakthrough than observed at the EBCT of 20 minutes. The inaccuracy of the models applied can be explained because, due to the complexity of the processes that simultaneously take place during GAC filtration, the prediction of its performance involves many inevitable assumptions and simplifications.