Generating knowledge on the availability of metals in soils is facilitated by accessible mechanistic multisurface models that predict the distribution of metal over the solid and solution phases (solid-solution partitioning) and between the reactive surfaces to which the metal is bound within either phase (speciation). An essential soil property used as a model input is the reactive metal pool. This pool can be measured through single extraction methods, the suitability of which is metal- and soil-dependent. Using 34 diverse Colombian soils with mostly naturally elevated Cd contents, we aimed to (i) define the suitability of multiple single extraction methods (HNO3, EDTA, DTPA or Mehlich-3), and (ii) evaluate model performance with each method as input and across levels of model complexity. Applying the isotopic dilution method (E-value) as a benchmark, we found that no single extraction consistently outperformed the others for particular soil types. Relatively strong extractions (HNO3, EDTA) tended to overextract reactive Cd and overpredict dissolved Cd, and vice versa for relatively weak extractions (DTPA, Mehlich-3). Using the E-value led to a more robust model accuracy across levels of model complexity. Residual overprediction can be attributed to uncertainty in the model parameters and other input values. To adequately model cadmium speciation, it proved necessary to include measured concentrations of competing cations. Accurate predictions of solid-solution partitioning could be made with a simplified model that only included pH, reactive cadmium, soil organic matter, and clay and excluded metal-oxides and competing cations. Our results strengthen multisurface models as accessible tools for environmental risk assessment.
Wiersma et al. (Thu,) studied this question.