Testing the conservativeness of a screening model in a model validation exercise.

Abstract

Although models for calculating derived emission limits have been around for many years, the opportunity to test them against independent data sets did not arise until the Biospheric Model Validation Study (BIOMOVS) was organized in 1985. Within BIOMOVS, two scenarios tested predictions of the movement of 131I and 137Cs from air to soil, pasture, milk, and beef. One of these scenarios was a model intercomparison of a chronic release over 30 y. The second scenario, using data gathered world-wide after the Chernobyl accident, allowed predictions to be compared directly with observations. The Canadian Standards Association's Guidelines for Calculating Derived Release Limits for Radioactive Material in Airborne and Liquid Effluents for Normal Operation of Nuclear Facilities was tested in both these scenarios to see whether its predictions were suitably conservative, as they should be for a screening model. A comparison was made between results of the Canadian Standards Association's model and those of other screening models on the one hand and results of models attempting to predict best estimates on the other hand. This analysis shows that often screening models are not conservative, and thus there should be much more effort to test the models against observations.