Data Decision Analysis: Project Shoal

Abstract

The purpose of this study was to determine the most appropriate field activities in terms of reducing the uncertainty in the groundwater flow and transport model at the Project Shoal area. The data decision analysis relied on well-known tools of statistics and uncertainty analysis. This procedure identified nine parameters that were deemed uncertain. These included effective porosity, hydraulic head, surface recharge, hydraulic conductivity, fracture correlation scale, fracture orientation, dip angle, dissolution rate of radionuclides from the puddle glass, and the retardation coefficient, which describes the sorption characteristics. The parameter uncertainty was described by assigning prior distributions for each of these parameters. Next, the various field activities were identified that would provide additional information on these parameters. Each of the field activities was evaluated by an expert panel to estimate posterior distribution of the parameters assuming a field activity was performed. The posterior distributions describe the ability of the field activity to estimate the true value of the nine parameters. Monte Carlo techniques were used to determine the current uncertainty, the reduction of uncertainty if a single parameter was known with certainty, and the reduction of uncertainty expected from each field activity on the model predictions. The mean breakthrough time to the downgradient land withdrawal boundary and the peak concentration at the control boundary were used to evaluate the uncertainty reduction. The radionuclide 137Cs was used as the reference solute, as its migration is dependent on all of the parameters. The results indicate that the current uncertainty of the model yields a 95 percent confidence interval between 42 and 1,412 years for the mean breakthrough time and an 18 order-of-magnitude range in peak concentration. The uncertainty in effective porosity and recharge dominates the uncertainty in the model predictions, while the other parameters are less important. A two-stage process was used to evaluate the field activities. For all of the field activities combined there were five activities that were found to be optimal in terms of uncertainty reduction per unit cost: two-well, natural-gradient, energy budget, and single-well tracer tests, and the vadose zone modeling. A subset of the field activities was chosen such that there would be no duplication in parameter characterization. Of this subset, the vadose zone model, barometric test, energy budget, and the two-well tracer test were found to be for the peak breakthrough time metric, while the single-well tracer test and the hydraulic head measurements are also considered for the peak concentration metric. The environmental tracer activity was not found to be optimal, yet this activity may provide additional information on the transport system. Care must be taken in using this analysis to design a field characterization plan, as many assumptions were required in the analysis. First, many subjective assumptions were required to assess the reliability of the field activities in terms of their ability to reduce the uncertainty in the mean parameters. Actual field characterization may not result in the same reduction in model output uncertainty as estimated by this analysis. Second, this analysis focused on the reduction in model uncertainty due to the reduction in the uncertainty in the mean parameters. If the uncertainty in the mean parameters is reduced to zero, there still exists uncertainty in the natural heterogeneity that can never be reduced to zero. Therefore, this analysis should be used in combination with expert judgement when designing a field characterization strategy.