Modeling NOM Breakthrough in GAC Adsorbers Using Nonparametric Regression Techniques

Abstract

A nonparametric, nonlinear regression technique—which is introduced to the environmental engineering community in this article—was used to develop models to predict total organic carbon (TOC) breakthrough time for a wide assortment of granular activated carbon (GAC) adsorbers. The models were developed using breakthrough data extracted from the U.S. EPA's Information Collection Rule (ICR) treatment studies database and included breakthrough results from 221 small-scale tests performed using 35 different source waters. Model development included an evaluation of twelve independent (and possibly correlated) variables, of which only four proved to be of statistical importance: influent TOC concentration, pH, empty bed contact time (EBCT), and field-scale GAC size. The predictive capability and overall robustness of the models were evaluated using three methods. First, the models were externally validated using smalland field-scale breakthrough data. Second, a sensitivity analysis was performed to ensure that the models effectively capture expected breakthrough trends. Finally, the models were directly compared to other empirical models reported in the literature, which were developed using traditional linear regression techniques.