One of limitations of modeling uncertainty on decisions is the use of so-called "conservative" assumptions and parameterizations widely used in model development because the data needed to support and analyze model requirements may be difficult to obtain. Assumptions made for modeling efficiency or to compensate for a pervasive lack of data, providing “worse or worst” results, can distort quantification of uncertainty and compromise the use of model for making useful decisions.
Depending on setting and the decision parameter uncertainty can be of significant consequence, and can be addressed through relatively well-known methods. Consideration of more complex forms of uncertainty – conceptual model uncertainty – has created a new problem in that it is not always obvious what uncertainty permissible with the data actually has an impact on the decision. Application of techniques for expressing the consequences of this uncertainty to decision makers are still in their infancy.
Uncertainty analysis, whether parameter or conceptual, that focuses on changes to model outcomes that are not informed by measured data may be more useful to decision makers than a general “uncertainty analysis”. These types of approaches begin to address the impact of uncertainty on decision making, which is a very real value of modeling. Clearly communicating these uncertainties, including both the strengths and weakness of these modeling efforts, to decision makers is a continuing challenge.
See more of: Topical Sessions