Radionuclide sorption in the far field: Geostatistical simulation of crystalline rock to assess uncertainties due to heterogeneities

In the event of radionuclides leaking from a deep geological repository for radioactive waste, they can reach the ecosphere through fluid migration pathways in the rock and aquifers. Retention mechanisms such as the sorption of radionuclides on the minerals along such pathways influence the migration patterns and are thus an essential part of the safety requirements. Consequently, determining the mineral composition and its spatial distribution of a crystalline host rock is an important task in the safety assessment for potential repositories.

In the SANGUR project (Systematic sensitivity analysis for mechanistic geochemical models using field data from crystalline rock) we aim to determine which parameters and their uncertainties are essential for developing models for the simulation of radionuclide retention in crystalline rock. Radionuclide retention is substantially affected by sorption processes on the mineral surfaces, described by distribution coefficients (Kd values). A subsequent sensitivity analysis will help to identify the most influential parameters.

In addition to the groundwater composition and the thermodynamic sorption data, the mineralogy and its heterogeneity of the host rock play an important role in establishing the model. For the sensitivity analysis, in turn, it is vital to be able to describe the uncertainties of the individual parameters in the model.

To quantify the uncertainties, we simulate crystalline rock based on MLA (Mineral Liberation Analyzer image) data using Multinary Random Fields geostatistics. The focus is not only on the mineral composition of the bulk rock as a function of number of mineral phases and variability in grain sizes, but above all on the determination of the mineral composition of the exposed surfaces with which the aqueous phase comes into contact and on which sorption processes will thus preferentially take place.

Besides the question of how detailed the rock must be modeled in order to adequately capture the heterogeneities, the question of the model scale or the size of the representative volume element is also addressed.

In addition to the discussion of the methodology and the results of the host rock simulations, we show the results of an initial study that enables us to determine what size the representative volume element should have in order to best describe the heterogeneities of the host rock for the subsequent calculation of the Kd values and their uncertainties.