Uncertainty analysis in risk assessment for environmental applications

Environmental risk assessment is an essential part of any decision-making process because it allows to evaluate all potential risks associated with human activities that may cause environmental damage. However, environmental risk assessment studies are affected by significant uncertainties: randomness due to inherent variability in the system behavior (aleatory uncertainty) and imprecision due to lack of knowledge and information on the system (epistemic uncertainty). Traditionally, probabilistic distributions have been used to characterize both types of uncertainty. However, resorting to a probabilistic representation of epistemic uncertainty may not be possible when sufficient data is not available for statistical analysis or information is of qualitative nature. In such cases, epistemic uncertainty may be best accounted for by possibilistic distributions. In this thesis, the uncertainties characterizing the inputs of an hydraulic model for the risk-based design of a flood protection dike have been analyzed. Different methods of joint propagation of aleatory and epistemic uncertainties have been embraced depending on the different frameworks adopted for uncertainty modeling and epistemic uncertainty representation. Two uncertainty model frameworks have been analyzed: in the first framework a mixture of purely aleatory and purely epistemic uncertainties is considered (“level-1” setting); in the second framework, aleatory and epistemic uncertainties are separated into two hierarchical levels (“level-2” setting). In addition, two frameworks for epistemic uncertainty representation have been adopted, i.e., probability and possibility theories. Within these frameworks, the efficiency of purely probabilistic and “hybrid” (i.e., mixed probabilistic and possibilistic) approaches has been compared in the task of jointly propagating aleatory and epistemic uncertainties, in both “level-1” and “level-2” settings.