Comparing Two Methods for Addressing Uncertainty in Risk Assessments

Abstract

The Monte Carlo method is a popular method for incorporating uncertainty relative to parameter values in risk assessment modeling. But risk assessment models are often used as screening tools in situations where information is typically sparse and imprecise. In this case, it is questionable whether true probabilities can be assigned to parameter estimates, or whether these estimates should be considered as simply possible. This paper examines the possibilistic approach of accounting for parameter value uncertainty, and provides a comparison with the Monte Carlo probabilistic approach. The comparison illustrates the conservative nature of the possibilistic approach, which considers all possible combinations of parameter values, but does not transmit (through multiplication) the uncertainty of the parameter values onto that of the calculated result. In the Monte Carlo calculation, on the other hand, scenarios that combine low probability parameter values have all the less chance of being randomly selected. If probabilities are arbitrarily assigned to parameter estimates, without being substantiated by site-specific field data, possible combinations of parameter values (scenarios) will be eliminated from the analysis as a result of Monte Carlo averaging. This could have a detrimental impact in an environmental context, when the mere possibility that a scenario may occur can be an important element in the decision-making process.