Surface soil moisture (SSM) is a crucial variable for water and energy balance in land-atmosphere system as well as a key information for agricultural, hydrological and climatological research. The SSM aids in deciding time of farm operations like sowing of crops, irrigation, and can be a component for better agro-advisory and drought monitoring system. The high spatial and temporal variability however, makes it impractical to assess through field observations. On the other hand, satellite remote sensing can update the SSM on a regular basis over a large area. Based on this background, the present study combines optical, thermal and microwave remote sensing data to map the SSM by using cubist, random forest (RF; bagging), gradient boosting machine (GBM; boosting) and stacking (combination of all) machine learning algorithms. The RF model was the best in the estimation (r = 0.71; RMSE = 5.17%) during independent validation, whereas cubist model recorded the lowest bias (MBE = 0.21%). The accuracy of SSM mapping were further improved by stacking cubist, GBM, and RF with Elastic Net as meta-learner. The primary factors influencing the SSM were radar backscatter, modified normalised difference water index, and land surface temperature. The study reveals the advantage of synergies of remote sensing data and recommends stacking multiple machine learning models for improving accuracy of digital SSM mapping.