How to find the right site for a final repository? A science-based procedure is currently being employed to find the site in Germany that guarantees the best possible safety for the final disposal of high-level radioactive waste. In the future, this process of geoscientific exploration will generate extremely large amounts of geodata. Can artificial intelligence (AI) help in dealing with these data sets and thus in the search for a repository site? A study commissioned by the Federal Office for the Safety of Nuclear Waste Management (BASE) investigates this question.

Big Data

The geodata generated during the geoscientific exploration of potential sites are analysed and evaluated together with other information. As a result, a huge database is created in the site selection process. This database is not uniformly structured, and it is subject to various types of uncertainties. Finally, there is also a connection to the data already available today.

AI and machine learning: more than just buzzwords

In recent years, significant progress has been made in dealing with such large and inconsistent amounts of data. Under certain conditions, AI methods and tools are able to recognise patterns in data sets that are too large or too complex for human evaluation. One example is "machine learning" (ML) - a process that specialises in pattern recognition. It could be used in the search for a final repository. Patterns discovered in this way can then be used for more precise models of a repository and possibly also to predict repository behaviour over long periods of time.

Relevance for the site selection process

How can artificial intelligence be used in the search and selection of a final repository site? Artificial intelligence is already being used more and more in the geosciences. For this reason, BASE has commissioned a study to better assess the possibilities and limitations of these methods. The goal is to demonstrate the possibilities of artificial intelligence and to evaluate its application for geoscientific questions. At the same time, the study aims to present the limitations, the necessary prerequisites and the risks of AI application, and to classify and evaluate them with regard to the site selection procedure.