DSM techniques were evaluated in a semiarid environment using remote sensed data.
Random forest and linear regression can be used to map soil texture.
The most significant covariates were b3/b7 for sand and clay and GSI for silt.
RFM had higher R2 and lower RMSE than MLR for sand and clay.
For silt, the MLR was better than RFM with higher R2 and lower RMSE.