Sensitivity and uncertainty Sample Clauses

Sensitivity and uncertainty. ‌ In the verification and validation process the user needs to consider the effects of errors and the level of uncertainty when using models. Although it might be possible to minimize uncertainties by following best practice it might not be possible to eliminate them. This section reviews ways of treating uncertainty. In general, validation is about uncertainty, though possibly not always quantification of uncertainty. To validate simulations of physical systems observations and predictions of quantities are compared to determine whether the model, described by the mathematical equations, "replicates reality to an acceptable degree" (Ivings et al, 2013). Validation therefore involves assessing the difference between observed and predicted quantities (as set out in Section 3) both of which will have associated uncertainties to decide whether the model is acceptable given the difference and uncertainty. The need to consider uncertainty was recognised when the Model Evaluation Protocol (MEP) (MEG, 1994) for models used in hazard assessment was being developed. One of the items listed in the documentation describing MEPs by the Model Evaluation Group is "uncertainty estimation (of model and data)". If a model has not been verified then there can be no confidence that the solutions it produces are correct, therefore the complete process of validation can be described as Verification, Validation and Uncertainty Quantification (Xxx and Xxxxxxxxx, 2010, Xxxxxx et al, 2015). In the following subsections, sources of uncertainty in model use are reviewed, followed by a brief description of different types of uncertainty. The approach to treating uncertainty used in MEPs is described, followed by other approaches that are used to treat uncertainty. 4.1.1 Sources of uncertainty‌ There are a number of different sources of uncertainty and identifying these provides a basis for understanding the uncertainty in model validation and prediction. Model setup uncertainty relates to propagation of uncertainty in simulations, from uncertainty about model input and boundary conditions and model parameters. In Section 3.5, two practical examples are provided on the effect of location of computational domain boundaries for release and dispersion and deflagration simulations. In Section 3.6, further examples are provided on the effects of applying different boundary conditions.