Epistemic and aleatory uncertainty in recommended, generic rock Kd values used in performance assessment studies

Over the past decade or so there has been an explosion in the number of sorption modelling approaches and applications of sorption modelling for understanding and predicting solute transport in natural systems. The most widely used and simplest of all models, however, is that employing a constant distribution coefficient (K-d) relating the sorbed concentration of a solute on a mineral surface and its aqueous concentration. There are a number of reasons why a constant partitioning coefficient is attractive to environmental modellers for predicting radionuclide retardation, and in spite of all the shortcomings and pitfalls associated with such an approach, it remains the leitmotif of most performance assessment transport modelling. This paper examines the scientific basis underpinning the K-d-approach and its broad defensibility in a performance assessment framework. It also examines sources of epistemic and aleatory uncertainty that undermine confidence in K-d-values reported in the open literature. The paper focuses particularly upon the use of so-called "generic" data for generalised rock types that may not necessarily capture the full material property characteristics of site-specific materials. From the examination of recent literature data, it appears that there are still a number of outstanding issues concerning interpretation of experimental laboratory data that need to be considered in greater detail before concluding that the recommended values used in performance assessments are indeed conservative.