MODELING SULFUR TRANSFORMATIONS IN SOILS

Abstract

A multiple regression model was developed to predict the mineralization/immobilization of soil organic S from soil chemical, biological, and environmental factors using data from a 2-year field study on S transformations in soil that included control, elemental S, and CaSO4 (gypsum) treatments. A monthly S mineralization/immobilization index (MII) was created from the monthly differences in total organic S over the 2-year period. Using data from the control plots, stepwise variable selection for multiple regression analysis along with variance inflation and collinearity diagnostics were used to arrive at a final model. The monthly differences or change in the residual (non-reducible) S fraction were found to be the single best predictor of MII, with R2 = 0.72. The monthly differences in the ester-SO4 and C-bonded (Raney-Ni reducible) S fractions were found not to be useful predictors of the mineralization index. The optimal model developed (R2 = 0.75) had four variables: monthly differences in residual S, arylsulfatase activity, microbial biomass C levels, and a soil temperature variable. The model was validated with data obtained from plots receiving CaSO4 which resulted in a R2 value of 0.73. However, the model was less useful in predicting MII when soils were treated with elemental S (R2 = 0.53) because the elemental S content of the soil could not be monitored, and thus all S pools could not be accurately specified for this soil treatment.