In this study, the random forest (RF) algorithm and statistical model based on environmental covariates (ECOVs) representing climate, biota, topography, soil types, texture class, and selected properties including saturation percentage (SP), particle size fractions (clay, silt, sand), pH, soil organic matter (SOM), carbonate content (CaCO3), and cation exchangeable capacity (CEC) were utilized to develop pedotransfer functions (PTFs) examining the relationships between water-soluble potassium (WSK) and exchangeable potassium (ExchK) as responses to these variables in various soils in drylands of Sudan.