Deriving probabilistic soil distribution coefficients (K). Part 2: Reducing caesium K uncertainty by accounting for experimental approach and soil properties.

The solid-liquid distribution coefficient (K) is a key input parameter in radioecological models. However, its large variability hampers its usefulness in modelling transport processes as well as its accuracy in representing soil-radionuclide interactions. For the specific case of radiocaesium, the analyses of a Cs K soil dataset (769 entries) showed that values varied over a five order of magnitude range, and the resulting Cs K best estimate (calculated as a geometric mean = 2.5 × 10 L kg) lacked reliability and representativity. Grouping data and creation of partial datasets based on the experimental approach (short-term (< ~1 yr) vs. long-term experiments (> ~1 yr)) and soil factors affecting Cs interaction (i.e., the ratio of the radiocaesium interception potential (RIP) to the potassium content in soil solution (K); organic matter content (OM) and soil texture) succeeded in reducing variability a few orders of magnitude, with Cs K best estimates also differing by one-two orders of magnitude depending on the type of soil and experimental approach. The statistical comparison of the Cs K best estimates and related cumulative distribution functions of the partial datasets revealed a relevant effect of the sorption dynamics on Cs K values (with long-term values systematically higher than short-term ones), and that the RIP/K ratio was an excellent predictor of Cs K for short-term scenarios, whereas the RIP parameter could be predicted on the basis of texture information. The OM threshold to distinguish between OM threshold to distinguish between Mineral and Organic soils subclasses, regarding Cs interaction was determined to be 50% and 90% OM for short- and long-term scenarios, respectively. It was then recommended to select the Cs K input data depending on the soils and scenarios to be assessed (e.g., short- vs. long-term; OM %) to improve the reliability and decrease the uncertainty of the radioecological models.