An Error Budget for the Determination of the Atmospheric Mass Loading of Toxic Chemicals in the Great Lakes

Abstract

Using propagation of error analysis and a simple loading model, an estimate is made of the budget of errors involved in making an atmospheric loading prediction for toxic chemicals to the Great Lakes. The model is simplified in order to use the limited data available at present from monitoring near the Great Lakes. The paper focusses on the major sources of error and on processes which need further work before loading estimates can ultimately be made more accurate. The intent of the paper is to identify those processes that are poorly quantified rather than to provide a precise estimate of either the loadings or the exact error in those loadings. The results of the study are intended to encourage the improvement of the estimate of errors in calculating loadings to the lakes. It is suggested that certain parameters may be deemphasized with considerable savings in cost and effort in monitoring programmes designed to determine these loadings. For Pb, the model predicts that the fractional coefficient of variation in the loading (ɛ) is 0.80 and the error can be reduced by first improving the knowledge of the paniculate deposition velocity and then the precipitation concentration. For γ–HCH, the estimate based on the parameters chosen gives ɛ = 0.57. This estimate is controlled primarily by the accuracy of the Henry's law constant and air concentration errors. For B(a)P, the initial estimate predicts precision of ɛ = 1.5. The major source of that error is the poor knowledge of the particle deposition velocity. Improvement in the knowledge of v d to a factor of two reduces ɛ to 0.38. For PCB 28, a trichlorobiphenyl congener, the model estimates ɛ to be 8.9. The controlling factor in that error is the poor knowledge of the Henry's Law constant.