Deterministic vs. probabilistic analyses to identify sensitive parameters in dose assessment using RESRAD

The dose assessments for sites containing residual radioactivity usually involve tire use of computer models that employ input parameters describing the physical conditions of the contaminated and surrounding media and the living and consumption patterns of the receptors in analyzing potential doses to the receptors. The precision of the dose results depends on the precision of the input parameter values. The identification of sensitive parameters that have great influence on the dose results would help set priorities in research and information gathering for parameter values so that a more precise dose assessment can be conducted. Two methods of identifying site-specific sensitive parameters, deterministic and probabilistic, were compared by applying them to the RESRAD computer code for analyzing radiation exposure for a residential former scenario. The deterministic method has difficulty in evaluating the effect of simultaneous changes in a large number or input parameters on the model output results. The probabilistic method easily identified the most sensitive parameters, but the sensitivity measure of other parameters was obscured. The choice of sensitivity analysis method would depend on the availability of site-specific data. Generally speaking, the deterministic method would identify the same set of sensitive parameters as the probabilistic method when 1) the baseline values used in the deterministic method were selected near the mean or median value of each parameter and 2) the selected range of parameter values used in the deterministic method was wide enough to cover the 5th to 95th percentile values from the distribution of that parameter.