ROBUSTNESS OF AN UNCERTAINTY AND SENSITIVITY ANALYSIS OF EARLY EXPOSURE RESULTS WITH THE MACCS REACTOR ACCIDENT CONSEQUENCE MODEL

Uncertainty and sensitivity analysis techniques based on Latin hypercu be sampling, partial correlation analysis and stepwise regression anal ysis were used in an investigation with the MACCS model of the early h ealth effects associated with a severe accident at a nuclear power sta tion. The following results were obtained in tests to check the robust ness of the analysis techniques: two independent Latin hypercube sampl es produced similar uncertainty and sensitivity analysis results; sett ing important variables to best-estimate values produced substantial r eductions in uncertainty, while setting the less important variables t o best-estimate values had little effect on uncertainty; similar sensi tivity analysis results were obtained when the original uniform and lo guniform distributions assigned to the 34 imprecisely known input vari ables were changed to left-triangular distributions and then right-tri angular distributions; and analyses with rank-transformed logarithmica lly-transformed data produced similar results and substantially outper formed analyses with raw (i.e., untransformed) data.