A successful data modeling could only be achieved if considering (i) cation exchange competition with aqueous Ca 2+ and Mg 2+ , originating from the dissolution of the clay mineral itself or accessory minerals and (ii) precipitation, at high pH, of phases incorporating the divalent metal of interest. However, in many studies, the concentration of aqueous Ca 2+ and Mg 2+ and, more generally, a full characterization of solution composition (e. g., Si, Al, dissolved organic carbon, alkalinity) is not provided, thus impeding accurate data modeling. Further difficulties stem from uncertainties related to the nature of the solids precipitating at high pH and missing or inaccurate thermodynamic data related to mineral solubility.

Log K D values cannot be predicted from available literature data with an accuracy better than ±1 log unit, without sorption data obtained on a material representative of in situ conditions. In addition, the most reactive sites, commonly referred to as strong sites, which have been characterized only on some, but not all, reference clay materials, should not be considered in blind predictions based on the mineralogical composition of clay materials.