Predicting the solubility of Cd, Cu, Pb and Zn in uncontaminated Croatian soils under different land uses by applying established regression models

Soil pH, soil organic matter (SOM), dissolved organic carbon (DOC) and total trace metal concentration (M(tot)) control the solubility of metals in the soil. Several regression models have included these soil chemical variables for the prediction of metal solubility and free metal ion (FMI) concentrations in contaminated soils. We hypothesize that models developed on contaminated soils (after optimization of the coefficients) can be used on samples from uncontaminated sites. Soil samples were collected from unpolluted agricultural and forest soils located in Eastern Croatia and extracted with water to determine the concentrations of Cd, Cu, Pb and Zn. We used these data to test the applicability of three regression models on existing conditions under different land uses. The same predictors issued in the three models and the same regression coefficients were utilized in the present study. The results showed a good correlation between the observed and predicted values of metal solubility. However, the models overestimate the total solution concentration (M(sol)) and the concentrations of free metal ions (FMI) in solution, and therefore the same regression coefficients were optimized to fit our own observations. This was found to be very successful. The results showed that pH and DOC played a very important role in controlling metal solubility, while SOM and CEC were somewhat less significant. The impact of total soil concentration of metals (M(tot)) was rather minor. However, we feel that to carry out good predictions of M(sol) and FMI, the M(tot) is needed in such regression models.

► Regression models predicting metal solubility and free ion activity in soil.
► Testing most commonly used models from polluted soils on unpolluted soils.
► The same predictors are significant for both soils.
► Tested models have a good correlation but they overestimate the observed data.
► Optimization of regression equation is required to obtain the best-fit model.