Numerical modeling of parametric uncertainties in flow through porous media: Development and initial testing of PORSTAT

The advantages of a stochastic approach for repository analysis are outlined. They are: (1) the likelihood of model predictions can be quantified, and (2) information can be gained about how to reduce the uncertainty in these predictions. The PORSTAT solves the stochastic ground water flow equation coupled with the deterministic heat transfer and mass transport equations. An integrated finite difference numerical scheme is used in PORSTAT to solve the governing equations. The stochastic groundwater flow equation is approximated by a second order uncertainty analysis technique. Stochastic variables input to PORSTAT may be hydraulic conductivity, specific storage, boundary conditions, and initial conditions. The output from PORSTAT consists of the expected values and covariances of hydraulic heads and Darcian velocities. PORSTAT will be used to stochastically model groundwater flow in the thermally influenced zone around the repository. The results from two test cases run by PORSTAT and BWIP's Monte Carlo groundwater flow computer code (MAGNUM-MC) are compared. It is shown that PORSTAT tends to overestimate the uncertainty in hydraulic head predictions.