Deriving probabilistic soil distribution coefficients (Kd). Part 3: Reducing variability of americium Kd best estimates using soil properties and chemical and geological material analogues

Abstract

The solid-liquid distribution coefficient (Kd) is a key input parameter in radioecological risk models. However, its large variability hampers its usefulness in modelling transport processes as well as its accuracy in representing soil-radionuclide interactions. To assist in the selection of Kd values and their cumulative distribution functions for study areas without site specific information, a critically reviewed dataset was developed, containing more than 5000 soil Kd entries for 83 elements and an additional 2000 entries of Kd data for 75 elements gathered from a selection of other geological materials. For the specific case of americium (Am), the dataset contained 109 entries for soils and 33 additional entries for sediment and subsoils. The analysis of the Am Kd soil dataset showed that values varied 4-orders of magnitude, and consequently the resulting Am Kd best estimate (geometric mean (GM): 4.6 × 103 L kg−1) lacked sufficient reliability. The objective of this study was to calculate cumulative distribution functions and statistically evaluate this dataset to determine if the Am Kd variability in soils could be reduced by considering various factors, including: 1) measurement methods, 2) key soil properties, 3) the use of chemical analogue data, and 4) the use of analogue data. Accounting for Am Kd experimental method (i.e., sorption vs. desorption; long-vs. short-term experiments) had little effect on reducing variability. However, accounting for key soil factors (i.e., organic matter content (OM), pH, soil texture) succeeded in reducing variability of Am Kd, especially when combining pH and OM. While previous data sets have used 20% OM content as a critical value to distinguish between mineral and organic soils, this study shows that this critical value should be reduced to 10% OM to minimize Am Kd variability. The inclusion in the dataset of Am Kd from other geological materials (e.g., gyttjas, tills, and subsoils) and Kd values from trivalent lanthanides (Ln (III)) and actinides (An (III)) (172 additional entries) did not statistically affect the Am Kd geometric means of the various pH and OM partial datasets. The larger composite dataset (> 310 entries), with both chemical analogues and geological material analogues to address data gaps, increased the statistical power for calculating Kd best estimates with lower variability, thereby enhancing their usefulness for radionuclide risk calculations.