PREDICTING SOIL SORPTION COEFFICIENTS OF ORGANIC-CHEMICALS USING A NEURAL-NETWORK MODEL

The soil/sediment adsorption partition coefficient normalized to organ ic carbon (K-oc) is extensively used to assess the fate of organic che micals in hazardous waste sites. Several attempts have been made to es timate the value of K-oc from chemical structure or its parameters. Th e primary purpose of this study was to develop a nonlinear model for e stimating K-oc applicable to polar and nonpolar organics based on arti ficial neural networks using the octanol/water partition coefficient ( K-ow) and water solubility (S). An analytic equation was obtained by s tarting with a neural network, converging the bias and weight values u sing the available data on water solubility, octanol/water partition c oefficient, and the normalized soil/sediment adsorption partition coef ficient, and then combining the equations for each node in the final n eural network. For the 119 chemicals in the training set, estimates us ing the neural network equation lie outside the 2 sigma region (the st andard deviation for the training set, sigma = 0.52) for only five che micals, while all the chemicals in the test set lie within the 2 sigma region. it was concluded that the neural network equation outperforms the linear models in fitting the K-oc values for the training set and predicting them for the test set.