Main content area

A general-applicable model for estimating the binding coefficient of organic pollutants with dissolved organic matter

Li, Yi-Long, He, Wei, Wu, Rui-Lin, Xing, Baoshan, Xu, Fu-Liu
The Science of the total environment 2019 v.670 pp. 226-235
aquatic environment, dissolved organic matter, models, phenols, pollutants, regression analysis
The binding constant (Kdoc) of organic pollutants (OPs) with dissolved organic matter (DOM) is an important parameter in determining the partitioning of OPs in the aquatic environment. Most estimation models have focused on calculating the Kdoc of a specific group of OPs but failed to obtain Kdoc values of different OPs effectively over the last three decades. In this study, we attempted to build a general-applicable Kdoc model based on various organic compounds' Kdoc values from the literature since 1973. Two multiple linear regression models, a DOM nonspecific model and an Aldrich HA model, were developed based on two solid and easy to access parameters—molecular connectivity indices (MCI) and polarity correction factors (PCF). In addition, the models' corresponding Kow-Kdoc models, which were mostly used in previous model studies, were developed for comparison. The adjusted determining coefficient (adj-R2) and standard error of the estimate (SEE) of the DOM nonspecific MCI-PCF-Kdoc model were 0.815 and 0.579, respectively, whereas the adj-R2 and SEE for the MCI-PCF-Kdoc model of Aldrich HA reached 0.907 and 0.438, respectively. The Aldrich HA model showed higher pertinence to the nonspecific model. Furthermore, both models exhibited better fit than the Kow-Kdoc models. The dipole moment modification attempts did not significantly improve either MCI-PCF-Kdoc models; hence, the two models were not altered with the dipole moment. The robustness tests by a Jackknifed method showed that the two MCI-PCF-Kdoc models exhibited higher robustness than the Kow-Kdoc. Of all of the OPs, the phenols contributed the most to their robustness. Furthermore, a sensitivity analysis showed that the two MCI-PCF-Kdoc models were sensitive to the robust parameters.