A new methodology for utilizing multidimensional smart Kd-matrices in transport programs for long-term safety assessment

Sorption on mineral surfaces is one important retardation process to be considered in long-term safety assessments for radioactive waste repositories. So far a conservative concept with temporally and spatially constant distribution coefficients (Kd values) for each model unit is applied in the respective transport simulations.
In this work we describe a new methodology for integrating temporally and spatially variable distribution coefficients, so-called smart Kd values, into the transport code r³t [1]. These smart Kd values are pre-calculated with a bottom-up approach from mechanistic sorption models (namely surface complexation and ion exchange) and stored in a multidimensional matrix by coupling of three computer codes: PHREEQC [2], UCODE [3] and SimLab [4]. This strategy has numerous advantages over reactive transport codes: (1) One can calculate smart Kd values for a reasonable number of environmental parameter combinations; (2) It is possible to perform uncertainty analysis based on such smart Kd matrices; (3) In contrast to UCODE, SimLab also provides methods for global sensitivity analysis; (4) The overall methodology is much more efficient in computing time than a direct coupling of the geochemical speciation code with the transport code r3t.
The capability of this new methodology is demonstrated exemplarily for the sorption of uranium UVI on a natural sandy aquifer [5] and is able to describe the sorption behavior in dependence of changing geochemical conditions quite well. Complex proof-of-concept scenarios were simulated for the sedimentary overburden of the Gorleben salt formation. Here, results for the upper aquifer are presented and discussed. As a prerequisite, the data processing of the field data [6, 7] is explained, including approximate correlations of environmental input parameters. Eventually, five environmental factors span the multidimensional parameter space for the Kd matrices.
Smart Kd distributions based on Latin-Hypercube samplings are given for different radionuclides covering the elements Am, Cs, Ni, NpV, Ra, SeVI, Th and UVI; with Am and Th also used as chemical analogues for Cm and Pu, respectively. They are then compared to constant values, determined in laboratory experiments [8], which have been used so far in transport calculations for long-term safety assessment. In a last step, ranked sensitivity indices are provided for all the investigated radionuclides to identify the most relevant environmental factors, which should be investigated in more detail in the future to improve both the data base and process understanding.