Probabilistic Accident Consequence Uncertainty Analysis of the Whole Program Package COSYMA

The overall uncertainty analysis of the program package COSYMA for assessing the radiological consequences of nuclear accidents builds on the results of a series of individual uncertainty and sensitivity analyses of its submodules. A set of 186 model parameters was identified which contribute most to the uncertainties of endpoints. Probabilistic accident consequence assessments with 144 weather sequences were performed for each of 300 sample sets derived from the uncertainty distributions of these parameters by Latin hypercube sampling. The evaluation of the results provided confidence bounds for the complementary cumulative frequency distributions of endpoints for three different source terms covering a wide range of release fractions. Concluding sensitivity analyses identified the most important model parameters responsible for the uncertainties of endpoints.