(Led by B. Williams, Ecological Planning and Toxicology, Inc., and J. Gagne, American Cyanamid Company)
Dr. Williams noted that each step in ecological risk assessment is more complex and less understood than the corresponding step in human health risk assessment. Although hazard can be assumed when a toxic chemical is released, the species and populations at risk must first be defined. The appropriate selection of surrogate species for testing in the laboratory is usually unclear. Measurement of environmental concentrations is only the first step in exposure characterization. Exposure assessment also requires consideration of foraging behavior, avoidance, and food-web considerations, as well as spatial and temporal variability. Risk characterization involves comparison of exposure estimates with measures of hazard; this process might result in compounding of errors. Ecological risk assessments do not track individuals over time and so do not accurately reflect population changes.
The activities presented in the case study have a large research component, which is focused on dose-response assessment and exposure assessment. One discussant characterized risk assessment, as presented in the case study, as a retrospective exercise based on focused characterization of hazard and exposure in wildlife. Given the difficulties in conducting environmental risk assessments, the four-part paradigm might not be applicable at levels of organization above that of the population.
CASE STUDY 3A:Models of Toxic Chemicals in the Great Lakes: Structure, Applications, and Uncertainty Analysis
D. M.DiToro, Hydroqual, Inc.
This paper reviewed and summarized efforts to model the distribution and dynamics of toxic chemicals in the Great Lakes, with applications to PCBs, TCDD, and other persistent, bioaccumulated compounds. The models were based on the principle of conservation of mass (Thomann
and Di Toro, 1983). Analysis proceeded through five steps: water transport, dynamics of solids, dynamics of a tracer, dynamics of the toxicant, and bioaccumulation in aquatic organisms. Mechanisms considered include settling, resuspension, sedimentation, partitioning, photolysis, volatilization, biodegradation, growth, respiration, predation, assimilation, excretion, and metabolism. The model of toxicant dynamics considered three phases (sorbed, bound, and dissolved) in each of two media (water column and sediments) and 21 pathways into, out of, or between these phases. The model of bioaccumulation included 25 compartments (four trophic levels with one to 13 age classes at each level) with five pathways into or out of each compartment. Because of the large number of coefficients (rate constants), sparseness of knowledge of inputs, and little opportunity for field calibration, uncertainty analysis was important in all the modeling exercises.
The first example modeled the dynamics of total PCBs in Lake Michigan (Thomann and Connolly, 1983). Plutonium-239 was used as a tracer to analyze sediment dynamics, and the model suggested that resuspension is an important mechanism. Calculation of PCB concentrations was limited by an order-of-magnitude uncertainty in the mass loading. Predictions of PCB concentrations and their rate of decline were sensitive to the value assumed for the mass-transfer coefficient for volatilization.
The second example modeled TCDD in Lake Ontario and attempted to predict the relationship between one source of input and the resulting incremental concentrations of TCDD (Endicott et al., 1989). In the absence of knowledge of other inputs, field data could not be used to calibrate the model. Hence, a formal uncertainty analysis was performed with Monte Carlo methods and assumed probability distributions of the rate coefficients. The 95% confidence limits of predicted TCDD concentrations in water and sediment differed by a factor of 10-100. Uncertainties in rate constants for photolysis and volatilization were the most important sources of uncertainty in predicted TCDD concentrations.
The third example extended the Lake Ontario TCDD model to eight other hydrophobic chemicals and incorporated a food-chain model to predict concentrations in lake trout (Endicott et al., 1990). The model predicted wide differences in toxicant concentrations, depending primarily on the degree of hydrophobicity as indexed by the octanol-water