1.3 Limitations in Performing Uncertainty Analysis

The main limitation in performing comprehensive uncertainty analyses of transport-transformation models is the associated cost and effort. The computer resources required for uncertainty propagation using conventional methods (to be discussed in the following paragraph) can sometimes be prohibitively expensive. Further, the incorporation of uncertainty associated with structural and formulation aspects of the models requires significant effort. Finally, the data needs for characterizing input uncertainties are often substantial.

Conventional methods for uncertainty propagation typically require several model runs that sample various combinations of input values. The number of model runs can sometimes be very large, i.e., of the order of many thousands, resulting in substantial computational demands. On the other hand, in order to estimate the uncertainties associated with model formulation, several different models, each corresponding to a different formulation of the mathematical problem corresponding to the original physical system, have to be developed. The model results corresponding to all the possible combinations give an estimate of the range of the associated model uncertainty. Development and application of several alternative computational models can require substantial time and effort. Thus, the costs associated with uncertainty analysis may sometimes be prohibitively high, necessitating a large number of model simulations and/or the development of several alternative models.