PROBABILISTIC MODELING OF AQUIFER HETEROGENEITY USING RELIABILITY METHODS

A probabilistic model of groundwater contaminant transport is presente d. The model is based on coupling first- and second-order reliability methods (FORM and SORM) with a two-dimensional finite element solution of groundwater transport equations. Uncertainty in aquifer media is c onsidered by modeling hydraulic conductivity as a spatial random held with a prescribed correlation structure. FORM and SORM provide the pro bability that a contaminant exceeds a target level at a well, termed t he probability of failure. Sensitivity of the probability of failure t o basic variabilities in grid block conductivity is also obtained, at no additional computational effort. The effect of the choice of the pr edetermined target level at the observation well is provided, along wi th its effect on the relevant sensitivity information. Considerable sa ving in computational time was achieved by superimposing a coarse rand om variables mesh on a finer numerical mesh. The presence of regions o f lower conductivity on the probabilistic event is analyzed, and the r egions in which conductivity most affects the results are identified. Copyright (C) 1996 Elsevier Science Limited