• Bioaccumulation models;
  • Great Lakes food web;
  • Uncertainty analysis


The steady-state models of Gobas and Thomann for predicting chemical residues in aquatic food webs were compared. Bioaccumulation factors (BAFs) predicted on the basis of freely dissolved chemical in the water were not significantly different for piscivorous fish, based upon the 10th and 90th percentile predictions, except for chemicals with log n-octanol/water partition coefficients (Kows) ranging from 6.5 to 6.9, the region of maximum differences for chemicals with log Kows less than 8. For chemicals with log Kows greater than 8, the BAFs predicted by the models diverged. The BAFs predicted using the Gobas model were in slightly better agreement with measured BAFs (determined from Lake Ontario data) than those predicted using the Thomann model. Sensitivities of the input parameters used by both models were very similar. The most sensitive input parameters were lipid contents, Kow, and sediment–water column chemical concentration quotient (Πsocw) for both models, and the feeding preferences for benthic invertebrates (Diporeia) for the Thomann model. Simulations performed using uncertainties for the input parameters demonstrated that the Kow and Πsocw were the dominant sources of uncertainties for predicted BAFs by both models for the Great Lakes food web. For piscivorous fish, overall uncertainties in the predicted BAFs ranged from a factor of 3.3 to 5.5 (Gobas model) and from a factor of 3.3 to 8.7 (Thomann model) for chemicals with log Kows less than 7.6 (based upon the 10th and 90th percentile predictions).