At least three important contributions to modeling K_OC values have been reported since Chemical Property Estimation. Theory and Practice was published. Great variability is noted in the reported K_OC values of chemicals. There are several reasons for this. There are no standard, agreed-upon test materials for the measurements, and the variable and changing nature of organic matter in soils used lead to great variability in results. Both fast and slow adsorption processes take place in test media as well as in the environmental soil/sediment column, but no standard equilibration time is used. Most reported values are based on measurements made before equilibrium was attained and represent the fast adsorption processes primarily. Chiou et al (1998) point out that the variable and changing nature of natural soils and ambient conditions ensure that the K_OC value of a chemical will vary in the environment, too.

A chemical's K_OC value depends on medium and structure. Sediment K_OC values can be as much as twice those of the soil K_OC values. PAHs have higher soil/sediment K_OCs than other non-polar chemicals. PCBs have comparatively long equilibration times.

Gawlick et al (1997) reviewed 175 different correlations between K_OC and various descriptors such as K_OW, aqueous solubility, and topological indices. They found no general correlation to be superior to any other. Indeed, an accurate and reliable general correlation covering many chemical types is not likely to be found. Seth et al (1999) suggest using K_OC = 0.33 K_OW with limits of 0.89 K_OW >K_OC > 0.14 K_OW. More accurate class-specific correlations are available, however.

A great deal of information is available on K_OC-K_OW correlations, and they are widely used. But there is a serious flaw inherent in any K_OC - K_OW correlation. Seth et al. (1999) show that K_OC and K_OW values are both greatly determined by the chemical's activity coefficient in water. Thus, the two are auto-correlated. While simple statistics show correlations between the two to be highly reliable, the observed extreme variability of K_OC/K_OW ratios is not accounted for and predictive accuracy is very low. Seth et al. (1999) show that by correlating the K_OC/K_OW ratio with other descriptors, the auto-correlation problem is revealed. For instance, they report that:

log (K_OC/K_OW) = 0.03 log K_OW - 0.61

n = 117, r² = 0.02, s = 0.38.

Since the coefficient on log K_OW is not significantly different from zero, Seth et al. (1999) suggest using log (K_OC/K_OW) = -0.48 with limits of -0.05 >log(K_OC/K_OW)> -0.86.

Chiou, S.T., McGroddy, S.E., and Kile, D.E., Environ. Sci. Technol., 1998, 32, 264.

Gawlik, B. M., Sotiriou, N., Feicht, E.A, Schulte-Hostede, S., and Kettrup, A. Chemosphere, 1997, 34, 2525.

Seth, R., Mackay, D., and Muncke, J.. Environ. Sci. Technol. 1999, 33, 2390-2394.