Evaluation of semi-mechanistic models to predict soil to grass transfer factor of 137Cs based on long term observations in French pastures

Highlights: • Assess the ability of existing models to predict ¹³⁷Cs pasture concentrations. • Better characterizing and modeling of ¹³⁷Cs bioavailability in mid and long terms. • Comparison observed TF among monitoring sites and IAEA data are discussed. • Predicted TF tend to be unconvinced at mid and long terms compared to observed data. • Modeling of Kd using ¹³⁷Cs labile soil fraction led to more reasonable predictions. The aim of this study was to evaluate and improve the accuracy of the semi-mechanistic models used in regulatory exposure assessment tools, to describe the transfer factors of ¹³⁷Cs from pasture soils to grass observed in different grazing areas of France between 2004 and 2017. This involved a preliminary parameterization step of the dynamic factor describing the ageing of radiocesium in the root zone using a Bayesian approach. A data set with mid-term (10 years about) and long term (more than 20 years) field and literature data from 4 European countries was used. A double kinetics of the bioavailability decay was evidenced with two half-life periods equal to 0.46 ± 0.11 yr and 9.57 ± 1.12 yr for the fast and slow declining rates respectively. We, then, tested a few existing alternative models proposed in literature. The comparison with field data showed that these models always underestimated the observations by one to two orders of magnitude, suggesting that the solid-liquid partition coefficient (Kd) was overestimated by models. The results suggest that semi mechanistic models might fail in the long-term prediction of the radionuclide transfer from soil-to-plant in the food chain. They highlight the need to calculate Kd using easily exchangeable ¹³⁷Cs (i.e. labile fraction) rather than total soil ¹³⁷Cs.