Uncertainty Analysis with Site-Specific Groundwater Models: Experiences and Observations

Groundwater flow and transport predictions are a major component of remedial action evaluations for contaminated groundwater at the Savannah River Site. Because all groundwater modeling results are subject to uncertainty from various causes; quantification of the level of uncertainty in the modeling predictions is beneficial to project decision makers. Complex site-specific models present formidable challenges for implementing an uncertainty analysis. We present an overview of our experiences in conducting uncertainty analyses using the Monte Carlo method on two MODFLOW/MODPATH/MT3DMS groundwater flow and transport models in the reactor areas of the Savannah River Site. Each analysis required extensive computational resources, which were managed using pseudo-distributed computing with an Excel-based application on desktop computers. The Monte Carlo results were filtered using flow model calibration targets. The C-Area Groundwater Operable Unit uncertainty analysis focused on TCE discharge to the model boundary streams and used a nine layer (56,682 cell) finite difference grid with 5,000 eight-parameterrealization sets – 926 of which met calibration filtering requirements. The uncertainty analysis for the RReactor Seepage Basin Operable Unit focused on strontium-90 transport and used a 20 layer (214,200 cell) finite difference grid with 18,000 nineteen-parameter-realization sets – 1729 of which met calibration filtering requirements.