Efficient Modelling of Radionuclide Transport in Highly Heterogeneous Media and Under Variable Hydrochemical Conditions Using an "Intelligent Kd” Approach

Abstract:

Due to the high heterogeneity of fractured media and the ubiquitous lack of a complete site characterization, deterministic simulations of radionuclide transport in fractured rocks are notoriously highly uncertain. This epistemic uncertainty is typically addressed using stochastic methods; e.g. the connectivity structure of the medium is described using one or multiple realizations of Discrete Fracture Networks (DFN), which are then combined to Time Domain Random Walk (TDRW) simulations (e.g. Painter and Cvetkovic, 2005). In these formulations, many complex geochemical retention processes are usually lumped into a single parameter: the distribution coefficient (K_d). Although this approach is mathematically robust and numerically efficient, it relies on an important assumption: the K_d value of each radionuclide is constant in time. This assumption could be critical under long-term geochemical changes as it is demonstrated that the distribution coefficient depends on the pH, redox conditions and major chemistry of the system. In this work, we present a novel methodology that combines the robustness of stochastic methods with a sound and explicit description of water-solute-rock interaction processes. The reconciliation of all these is achieved by using an “intelligent K_d” approach. The hydrogeochemical evolution of the site of study is first computed using long-term and large-scale mechanistic reactive transport simulations. The simulated hydrochemical conditions are then used to generate a complete database of K_d values, which represent the hydrochemical conditions in every position and time of the model domain. Then, TDRW simulations, based on one or multiple DFN realizations, are fed with these data and the results (e.g. radionuclide breakthrough curves) implicitly bring the signature of the underlying changes in the background geochemistry.