The role of uncertainty in dam failure frequency estimates: a conceptual case study

The return period of combinations of events that are relevant to dam risk analysis is very high. It falls in a range of hundreds to millions of years, but remains hard to quantify with precision. Deterministic approaches to dam safety assessment typically rely on the detailed study of selected scenarios to gain knowledge about risk, necessarily leaving out a vast number of possibilities. Most current probabilistic approaches aim to analyze more systematically the range of frequencies that dangerous events might display, generally doing so at the expense of a less detailed evaluation but, despite that, still focus on a reduced number of hazards or structural elements and, therefore, offer space for improvement. This work employs the Generic Multi-Risk (GenMR) probabilistic frame- work to dynamically model a conceptual dam-reservoir system that is exposed to multiple hazards. GenMR is capable of coping with the dynamical nature of the system and, going beyond the aleatoric uncertainty that is normally required for hazard intensity characterization, it can model epistemic uncertainty and reproduce element responses based on fragility. The importance of considering epistemic uncertainty and responses based on fragility is illustrated with the analysis of 40 million years of simulations. For the conceptual case study, an increase of dam failure rates of 40% resulted from considering epistemic uncertainty. Accounting for element responses based on fragility contributed with an additional 50% augmentation. Overall, both types of uncertainty led to a 70% decrease in the return period of failures which, despite resulting from a conceptual case, illustrates well the need to explore the impacts of uncertainty on dam safety assessments.