Influence of soil properties on the sorption and retention of cadmium, copper and lead, separately and together, by 20 soil horizons: comparison of linear regression and tree regression analyses.

In this paper we compare linear regression with tree regression for analysis of the influence of soil properties on the sorption and retention of added Cd, Cu and Pb by 20 soil horizons typical of cropped soils in Galicia (N.W. Spain); our measure of sorption/retention capacity was K(r), a recently introduced adimensional parameter. Sorption and retention of Cd was depressed by the presence of Cu and Pb. The soil parameters that were most associated, overall, with differences in Cd, Cu and Pb sorption and retention were cation exchange capacity (CEC), pH and Mn oxides' content. Tree regression, which can take into account variation on both global and local scales, afforded better-fitting models than linear regression, which only reflects global tendencies; but for coherent interpretability of tree regression results it is just as important to avoid overfitting as in the case of linear regression.