Dynamic Bayesian networks for modelling chemical degradation in concrete structures of nuclear waste repositories exposed to climate change

Concrete is the ultimate engineering barrier that prevents the release of radioactive contaminants from Nuclear Waste Repositories (NWRs). Concrete suffers from two main chemical degradation processes that are expedited by Climate Change (CC): carbonation-induced corrosion and chloride ingress. Climatic variables, like temperature, relative humidity, and CO2 concentration, have an impact on the process of chemical degradation of concrete; this impact is uncertain, as it depends on the evolution of CC. In this work, we use Dynamic Bayesian Networks (DBNs) to model concrete chemical degradation, within the Performance Assessment (PA) of NWRs, and considering the effects of CC. The DBN models are applied to assess aquifer contamination, due to CC-induced concrete degradation, and dose intake from a realistic near-surface NWR. The results obtained show that the severe CC scenario (SSP-5 8.5) leads to the largest levels of concrete degradation and dose intake, as CC-induced degradation processes exceed the critical thresholds about a decade sooner than for moderate and low-emission scenarios. Moreover, as expected, deep uncertainty in CC variables significantly broadens the tolerance intervals of dose-intake violation probabilities, exposing long-term risk-informed PA to larger uncertainty, that are to be duly considered.