Including uncertainty in Quantitative Risk Assessment (QRA) is recommended in most published Best Practice Recommendations for QRA, including those published by TNO, CCPS, and Norwegian authorities. Unfortunately typical commercial QRA techniques and models used in Process Risk Analysis do not include the capabilities of uncertainty within the model formulations. Effects on uncertainty may be based on qualitative estimates or ignored, or at most accounted for by sensitivity analysis by “swinging” variables to estimate individual variable sensitivity. By not presenting the uncertainty, the risk assessment results may be misrepresented as possessing an accuracy greater than is warranted.
Failure to account for uncertainty also tends to support centrist thinking and poor quality decision making. The natural tendency of users of risk analysis is to believe in their results, especially if much effort has been put into developing the results. If the result is a single number, that number may be assumed by the analyzer and users as “right”. In addition, there is a tendency in people to assume that by picking the most likely values for the inputs into a model, the end result is also the most likely result. This may not be true if the model is complex and non-linear. The most likely results must be determined by including variable distributions and uncertainty modeling in the analysis.
Decision analysis is also deceptively simple when single point criteria or simple curves are used for decision making purposes. A challenge in risk assessment is utilizing results and presenting them in a straight forward way that capture the uncertainty and complexity while allowing clear, high quality decisions by end-users.
This paper presents sources of uncertainty, analysis methods, examples of potential errors introduced by ignoring uncertainty, and illustrates how one can include uncertainties in your quantitative risk assessment (QRA) and benefit from it in your risk assessment decisions.