Assessing the Value of Including and Reducing Uncertainty Using Bayesian Monte Carlo Analysis. M. E. Dakins, SUNY-College of EnvironmentaI Science and Forestry, Syracuse, NY 13210; J. E. Tolland, Enserch Environmental, 10900 NE 8th St. Bellevue, WA 98004; and M.J. Small, Carnegie Mellon University, Pittsburg, PA 15213
A methodology for incorporating uncertainty in model predictions into a risk-based decision for environmental remediation is illustrated considering polychlorinated biphenyl (PCB) sediment contamination and uptake by winter flounder in New Bedford Harbor, Massachusetts. Sensitivity and uncertainty analyses are conducted for a model which predicts the sediment remediation volume required to meet a biota tissue concentration criterion. These evaluations help to identify the variables which most significantly contribute to uncertainty in the model prediction and allow for calculations of the expected value of including uncertainty (EVIU) and the expected value of perfect information (EVPI) for the remediation decision. Once uncertainty is estimated, Bayesian Monte Carlo analysis provides a formal method of incorporating new data into the model to obtain an updated information state with reduced uncertainty. Bayesian Monte Carlo analysis is a sequential procedure in which probability mass functions on model outputs generated by Monte Carlo methods are updated using Bayes' rule to reflect information contained in experimental observations of the modeled system. Further analyses include a calculation of the expected value of sample information (EVSI), the difference between the expected loss of the optimal management decision based on the results of a full uncertainty analysis and the expected loss of the optimal management decision from an updated information or posterior state with reduced uncertainty, for three different sampling plans.