Quantitative relationships between the structure of phenylurea herbicides a
nd their transformation in different matrices have been developed. Experime
ntal data on microbial transformation by pure and mixed cultures of soil mi
cro-organisms in inoculated and native soil, as well as on chemical transfo
rmation by hydrolysis in sterile soil and water, were available from previo
us studies. Around 60 experimental or calculated descriptors were used. Qua
ntum chemical calculations were performed with three different semi-empiric
al methods. Models developed with multiple linear regression were generally
easier to interpret than those derived with partial least-squares projecti
on to latent structures. Quite simple and interpretable models could be fou
nd to predict transformation rates by pure cultures from lipophilicity, by
mixed cultures from adsorption distribution coefficients, and by chemical o
r enzymatic hydrolysis from electronic properties. Transformation in inocul
ated soil could not be predicted, but for native soil the use of a quantum
chemical descriptor for reactivity (energy of LUMO) together with molar ref
raction resulted in a general model. (C) 2001 Society of Chemical Industry.