A Bayesian Synthesis Approach to Assessing Exposure to PCBs in New Bedford Harbor. S. C. Bates, Department of Statistics, University of Washington, Seattle, WA; A. C. Cullen, Graduate School of Public Affairs, University of Washington, Seattle, WA; and A. E. Raftery, Department of Statistics, University of Washington, Seattle, WA
A Superfund clean-up is currently underway at the New Bedford Harbor site (NBH) in MA, where marine sediments are contaminated with poly-chlorinated biphenyls (PCBs). Harbor dredging may result in exposure of inhabitants of the surrounding region to PCB contaminated air, soil and food. A probabilistic exposure assessment underway at this site requires distributions to represent PCB concentration in soil and in root, leafy and vine plants. A simple compartmental model for PCB concentration in soil was adapted by Cullen (1992). We apply the Bayesian synthesis method of analysis which incorporates both prior knowledge about the inputs to the model, and available data on PCB air and soil concentration in the NBH area. The method uses Monte Carlo simulation from the prior distributions of the inputs and resampling of these simulations with weights determined by the observed data. This method allows observed data on both model inputs and outputs to update the (prior) distribution of model inputs. It yields posterior distributions of PCB concentration in soil from which distributions for soil exposure quantiles are derived. The method also allows the separation of the contributions of variability and uncertainty on the posterior distribution of soil concentration.
Go to . . .