Quantum chemical modelling of soil sorption coefficients: Multiple linear regression models

Soil sorption coefficients for nonpolar compounds easily can be modelled with log POW or with connectivity indices. This is not the case for polar compounds: for polar substances either compound class dependent models or models using correction factors based on substructures are described in the literature. In the present study a systematic search on multiple linear regression models using log POW and several quantum chemical descriptors was performed to model the log KOC for a set of 66 substances from different compound classes (pesticides and other nonpolar and polar chemicals. It is shown that the soil sorption coefficient KOC can be modelled with one equation for the whole data set.