Sensitivity analysis via ‘derived parameters’

Abstract ‘Derived parameters’ are simple arithmetic combinations of uncertain input parameters, that are expected to have a strong effect on an output variable of the model. Knowledge acquired by experienced modelers from the analysis of model equations and from other sources (running codes) help defining this sort of input parameters. This work summarizes the way to find derived parameters in a well-known model (Level E) and shows the benefits of introducing these parameters in the sensitivity analysis: strong dependence of outputs on them and direct physical interpretation of sensitivity analysis results.