Bayesian Statistics in Crayfish Bioaccumulation of Polycyclic Aromatic Hydrocarbons. H. Lin, D. Berzins, J. Bollinger, K. H. Watanabe, Tulane University
Uncertainty evaluation is one of the important components of risk characterization, the final phase of ecological risk assessment. Currently, the U.S. EPA guidelines for ecological risk assessment combine uncertainty (e.g. measurement error, model structure error, etc) and variability (e.g. population heterogeneity in body weight, lipid content and spatial variability in the environment, etc). Including population heterogeneity into models can increase the credibility of model prediction and risk estimation. Bayesian statistical analysis, which treats model parameters as random variables, provides a mechanism to quantify the variability in model predictions based upon the variability in model parameters. The posterior distribution of parameters can be determined from the prior knowledge of parameters (historical data or information from the literature) and empirical data. In this study, we developed a hierarchical population model to account for the inter-individual variability of lipid content in a crayfish bioaccumulation model. This model predicts the concentration in organism based upon the sediment concentration, sediment organic carbon fraction, and the organism’s lipid fraction. Field-exposed and field-captured crayfish at control and contaminated sites in the LaBranche Wetlands, Louisiana were analyzed for polycyclic aromatic hydrocarbon (PAH) concentration and lipid fraction. The PAH concentrations in water and sediments at the sites were also analyzed. The posterior distributions of the model parameters were derived from the joint posterior parameter distribution using a Markov chain Monte Carlo approach and experimental data. The results were used to predict the distribution of contaminant concentration in the crayfish population. This approach can also be applied to more complex food web models to provide a quantitative expression of the variability in predicted concentrations.
Go to . . .
2002 SRA Annual Meeting Table of Contents
2002 SRA Annual Meeting Author Index
Main Abstracts Menu Page
RiskWorld Home Page