Internal Research Project I-11

One part of research project D03 is the development of methods for sensitivity analysis using model surrogates. In this internal project, we will apply these methods to the field of geosciences, particularly to the modelling of aqueous transport processes with low-level radioactive nuclides. By this application, we aim to improve our understanding of the governing processes and the influence of the various parameters, such as dispersivity, soil permeability, half-life period, sorption magnitude and rainfall intensity.

First, a suitable scenario for the sensitivity analysis will be specified. This includes setting up the geometry, initial and boundary conditions of a landfill containing radioactive nuclides, the surrounding soil and the aquifer beneath it. The parameters that are to be investigated, together with their assumed probability distributions, will also be specified. The second step is realising the coupling of the computational forward model implemented in DuMux with BayesValidRox, the central tool developed in D03-Flemisch/Oladyshkin. BayesValidRox is an open-source Python package that provides methods for surrogate modeling, Bayesian inference and model comparison. In the third step, the quality of the calculated model surrogates will be investigated and optimized. Finally, the sensitivity analysis will be performed, yielding the Sobol indices of the input parameters. The project will be concluded by an interpretation of the results.