Modelling the effects of soil properties on the concentration of Cd extracted by 10 mM CaCl₂ from soils of the Sydney Basin

Undisturbed surface soils on the peri-urban fringe of the Sydney Basin are mostly acidic, with low concentrations of total cadmium (Cd-t) and a wide range of other properties. In taxonomically similar soils on vegetable farms, Cd-t is typically about 20-fold greater and undesirably large concentrations of Cd occur in some of the vegetables. We used a sequence of multiple regression models to evaluate how soil properties, taxonomy, and farming affected the behaviour of Cd in surface composites (0-15 cm) from 29 farmed and 12 unfarmed sites in the Basin. The dependent variable was the log(10) transformation of the concentration of Cd extracted in 1 h by 10 mM CaCl2 solution (log(10) Cd-Ca). The range of values of Cd-Ca was 0.2-60 mug/L. The independent variables in the base model were pH(Ca) and log(10) Cd-t (R-2=0.885, r.s.d.=0.245). The final model contained 2 additional log(10)-transformed soil properties: effective cation exchange capacity (ECEC) and oxalate-extractable Fe (Fe-ox) (R-2=0.974, r.s.d.=0.121). The effect of log(10) Fe-ox was significant (P<0.05), but only when pH(Ca) was >5.6, the approximate sorption edge for Cd on goethite. The effects of other soil properties could not be adequately tested because the residual variation was too small. The measured values of Cd-Ca had a median error of +/-17% and a maximum error of +/-58% relative to the back-transformed fitted values from the final model. The coefficient of log(10) Cd-t in the final model was 1.33+/-0.05 for the farmed soils and 1.01+/-0.04 for the unfarmed soils, i.e. log(10) Cd-Ca was more sensitive to changes in log(10) Cd-t in the farmed than the unfarmed soils (P<0.01). This difference is consistent with the effects of the greater load and briefer duration of contact for Cd in the farmed soils. The coefficients of pH(Ca) and of log(10) ECEC in our final model had values of 0.49 (+/- 0.03) and 0.69 (+/- 0.08). These values are remarkably similar to those obtained when we fitted our data, using a model that had been used to describe the effects of soil properties on a compilation of Cd-sorption data, and to those in a partitioning model based on the desorption of Cd from contaminated soils. That is, the behaviour of Cd may be influenced by the duration of contact between Cd and the soil, and perhaps the load; however, neither factor appears to change the effects of pH(Ca) and log10 ECEC.