Prediction of primary salinity, sodicity and alkalinity in South African soils

Abstract

Several attempts have in the past been made to map the primary salinity, sodicity and alkalinity of South African soils. These have typically been based on personal experience and/or limited data and also failed to quantify the accuracy of prediction. This paper therefore attempted to map these variables from soil profile data, and 1:1 000 000 topographical, geological and mineral maps. In the end 58% of the variability in soil electrical conductivity in the <550 mm rainfall class and 39% of the variability in the >550 mm rainfall class could be explained. Coefficients of determination for exchangeable sodium percentage were 85% in the <550 mm rainfall class and 52% in the >550 mm rainfall class. Multiple linear regression relationships were unable to predict pH_water, indicating that log-transformation of pH_water values are needed. Of the 122 × 10⁶ ha land in South Africa, only 23% were calculated to be slightly saline, 5.1% saline, 1.4% moderately saline, 0.4% strongly saline; 3.8% non-alkaline saline-sodic, 6.3% alkaline saline-sodic and only 0.4% sodic.