Historical Reconstruction of PCE-Contaminated Drinking Water Using Probabilistic Analysis at U.S. Marine Corps Base Camp Lejeune, North Carolina

The historical reconstruction process is used to derive contaminant concentrations and exposure levels needed as input by health scientists conducting retrospective epidemiological studies—models are an integral part of this process. However, the models and associated calibrated parameters are inherently uncertain because they are based on limited historical data and information. This gives rise to the question, what is the reliability of the historically reconstructed estimates of contaminant concentrations determined using calibrated models? To answer this question and address the overarching issues of model and parameter variability and uncertainty, a probabilistic analysis is used to generate uncertainties of model inputs (e.g., hydraulic conductivity or contaminant source mass loading rate) so that estimates of uncertainties in model outputs (e.g., water level or contaminant concentrations in groundwater) can be made. In this paper, the authors describe the application of a probabilistic analysis using Monte Carlo simulations to assess model uncertainty and parameter variability. The probabilistic analysis is applied to groundwater-flow and contaminant fate and transport models (MODLFOW and MT3DMS, respectively) as part of an ongoing health study—exposure to volatile organic compounds in drinking water and specific birth defects and childhood cancers—being conducted by the Agency for Toxic Substances and Disease Registry at U.S. Marine Corps Base Camp Lejeune, North Carolina (Figure 1). Results of probabilistic analysis are presented in terms of the probability of occurrence tetrachloroethylene (PCE) contamination in groundwater, the probability of exceeding the current maximum contaminant level for PCE in finished drinking water (5 micrograms per liter), and the range of expected PCE concentrations in finished drinking water during the historical period of interest (1968–1985) for the health study.