Methods for artificial intelligence-based forecasting of natural external impacts on final repositories during the operational phase

During the operational phase of a final repository, both underground facilities and aboveground structures with safety-related significance must be protected against external influences (EVA) (BGE 2021). A distinction is made between natural and anthropogenic external events (EVA), although this project considers only natural external events. These can influence the stability and safety of the repository and are assessed on a site-specific basis. Protection requirements are generally based on events with a defined recurrence frequency that are expected according to the current state of science and technology. To date, the characterization of relevant events (frequency, intensity, sequence) has been based primarily on historical data and extreme value methods, assuming a stationary climate regime. However, in light of climate change, it is questionable whether these approaches remain suitable (GAO 2024), making an adjustment to the modeling approach necessary. Within this context, the project investigates the use of AI methods to forecast distribution functions for safety-related design as well as to predict the temporal and spatial progression of EVA events, in order to derive appropriate measures in the event of an incident.