9304481 Rubin Predictions of contaminant transport by groundwater, as well as monitoring and management, are based on quantitative models of solute motion. Present methods for prediction of contaminant transport are inadequate because they fail to account explicitly for measured data at its locations. They also do not account realistically for the effects of heterogeneity in the conductivity and temporal variability in pressure (or only under some simplifying assumptions), and they do not offer an economic means of designing a monitoring scheme where the observed data are used to update the predictive models. Furthermore, little attempt has been directed so far to obtain a better understanding of transport in highly heterogeneous geologic formations and the effects of chemical reactions on transport at the field scale. Lastly, current predictive methods fail to link the predictions with any criterion of goodness of fit. In this study, we propose to develop a modeling strategy which considers the aquifer hydraulic properties, the velocity field, and the solute concentration as random space functions, expressed in terms of initial/boundary conditions and various physical parameters, and which allow, both spatial variability and prediction uncertainty to be analyzed with probabilistic, quantitative tools. Three-dimensional random function models will be obtained for steady and unsteady nonuniform velocity fields and will be used for predicting the various moments of the concentration of reactive and non-reactive contaminants for anisotropic conductivity fields. Bayesian analysis will make the predictions locally conditioned. The model will be applied to two field studies. We plan to achieve a better understanding of transport phenomena at the fundamental level and obtain risk- qualified, locally-conditioned estimates of solute spreading in applications. We plan to generalize existing stochastic transport theories to account for higher-order variability i n the flow field, to investigate the influence of flow unsteadiness and nonuniformity, to evaluate the relevance of unsteadiness to the prediction of solute transport under typical conditions, and finally to suggest simplified methods to incorporate flow unsteadiness, flow nonuniformity and chemical reactions into predictions. ***
