Modeling equilibrium adsorption of organic micropollutants onto activated carbon D.J. de Ridder a, *, L. Villacorte b , A.R.D. Verliefde a,c,d , J.Q.J.C. Verberk a , S.G.J. Heijman a , G.L. Amy a,b,e , J.C. van Dijk a a Delft University of Technology, P.O. Box 5048, 2600 GA Delft, The Netherlands b Unesco-IHE, P.O. Box 3015, 2601 DA Delft, The Netherlands c University of New South Wales, Sydney, Australia d KWR Watercycle Research Institute, P.O. Box 1072, 3430 BB, Nieuwegein, The Netherlands e King Abdullah University of Science and Technology, Thuwal, Saudi Arabia article info Article history: Received 8 October 2009 Received in revised form 5 January 2010 Accepted 19 February 2010 Available online 1 March 2010 Keywords: Activated carbon QSAR Binning abstract Solute hydrophobicity, polarizability, aromaticity and the presence of H-bond donor/ acceptor groups have been identified as important solute properties that affect the adsorption on activated carbon. However, the adsorption mechanisms related to these properties occur in parallel, and their respective dominance depends on the solute prop- erties as well as carbon characteristics. In this paper, a model based on multivariate linear regression is described that was developed to predict equilibrium carbon loading on a specific activated carbon (F400) for solutes reflecting a wide range of solute properties. In order to improve prediction accuracy, groups (bins) of solutes with similar solute properties were defined and solute removals were predicted for each bin separately. With these individual linear models, coefficients of determination (R 2 ) values ranging from 0.61 to 0.84 were obtained. With the mechanistic approach used in developing this predictive model, a strong relation with adsorption mechanisms is established, improving the interpretation and, ultimately, acceptance of the model. ª 2010 Elsevier Ltd. All rights reserved. 1. Introduction Since the presence of low concentrations of pesticides, phar- maceuticals, industrial waste constituents and personal care products has been confirmed in water sources, their occur- rence levels, effects on (human) health and efficacy of treat- ment processes for their removal from drinking water have been of primary concern to water utilities and environmental agencies (Schwarzenbach et al., 2006). It is a time-consuming and expensive process to experimentally determine all these different aspects for every individual micropollutant. The number of organic micropollutants present in water sources not only is vast, but also variable as new products are continuously introduced. In order to minimize experimental work in drug design, the pharmaceutical industry applies quantitative structure activity relationship (QSAR) models, which can predict drug metabolic activity and toxicity a priori, based only on chemical structure (Kruhlak et al., 2007). Environmental protection agencies, such as the U.S. EPA and the Danish EPA, also apply QSAR models to predict micropollutant toxicity. QSAR models to predict micropollutant removal in water treatment processes, however, have rarely been used, although some models have been proposed for membrane * Corresponding author. E-mail address: d.j.deridder@tudelft.nl (D.J. de Ridder). Available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/watres water research 44 (2010) 3077–3086 0043-1354/$ – see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.02.034