Construction and validation of parametric models to predict radium sorption in soils

Elucidating the factors affecting the transfer of naturally occurring radionuclides (NOR) between environmental compartments is a key part of the assessment of ecosystem's exposure to naturally occurring radionuclide materials (NORM). For that, the sorption and desorption solid-liquid distribution coefficients (K_d) of radium (Ra) were quantified in a collection of 31 soil samples with contrasting edaphic properties under controlled conditions in laboratory batch experiments. Ra sorption was demonstrated to be moderate to high, with K_d (Ra) values ranging from 10² to 10³ L kg^-1. Ra sorbed was mostly irreversible, as evidenced by desorption percentages lower than 2 %. An exploratory analysis with partial least squares (PLS) regression identified the soil properties that correlated with K_d (Ra) and discarded those that were not relevant for describing K_d variability. A dataset of the sorption K_d (Ra) values and associated soil properties was built from our own data and from the literature after performing an in-depth review of similar Ra sorption studies. For the first time, K_d (Ra) parametric prediction models were constructed using univariate linear regression (ULR) and multivariate linear regression (MLR). Ra sorption in soils was mostly explained by the soil properties directly or indirectly related to the available exchange sites, such as the levels of water-soluble and exchangeable Ca and Mg as well as the pH of the contact solution. The most promising models explained around 80 % of the K_d (Ra) data variance, only needing K_d (Ca + Mg) or additional soil descriptors such as pH, Mn content, and the specific surface area. The validation of the proposed models confirmed that K_d (Ra) can be predicted with only a few soil properties that can be characterised in routine analysis. Thus, the proposed models could be used to estimate the interaction of Ra in soils in risk assessment.