Deriving probabilistic soil distribution coefficients (K_d). Part 1: General approach to decreasing and describing variability and example using uranium K_d values

A general approach is presented to derive probabilistic radionuclide distribution coefficients (K_d) in soils from a K_d dataset. The main aim was to derive informed estimates with a low inherent uncertainty by restricting the K_d value data to subsets based on key soil factors and the experimental approach used to calculate the K_d value (e.g., sorption and desorption tests). As an example, the general approach was applied to uranium (U) K_d values that are part of a critically reviewed dataset containing more than 5000 soil K_d entries for 83 elements and an additional 2000 entries of K_d data for 75 elements gathered from a selection of other, non-soil, geological materials. The overall soil U K_d dataset included 196 values spanning a range of four orders of magnitude (1-67,000 L kg^-1), with additional 50 entries for other geological materials. Whereas the effect of the experimental approach could be disregarded, major factors in decreasing U K_d variability were pH and organic matter content (OM). Limitation in the number of entries made it difficult to use texture information (sand, silt, clay) to further decrease U K_d variability. The integrated combination of pH + OM permitted some soil groups to have U K_d confidence intervals as narrow as two orders of magnitude. Specifically for U K_d, data in the Mineral (< 20% OM) and Organic (≥ 20% OM) partial datasets were significantly different. Analogue data from geological materials other than soils, such as subsoil, till and gyttja (a lacustrine mud having elevated organic matter (OM) contents), were also statistically evaluated to determine whether they could be used to fill U K_d data gaps. It was shown that U K_d from subsoils and tills, but not gyttjas, could be used to enhance soil U K_d datasets. Selection of probabilistic K_d values for risk modelling can be made more reliably and with less uncertainty by using appropriate geochemical data representative of the study site to narrow the wide range of potential K_d values.