Improving the kriging of a soil variable using slope gradient as external drift

Universal kriging with a topographic variable used as external drift was compared with ordinary kriging, ordinary kriging with external drift, and universal kriging for predicting the thickness of a superficial deposit on part of the Beauce of central France. Thickness was recorded at 62 sites and at another 25 sites to obtain a sample for validation. A Digital Elevation Model (DEM) was computed from 484 sites where relative elevation was measured. The slope gradient which is significantly correlated to thickness was used as an external drift variable for kriging thickness. The sample for validation was used for testing the four prediction methods. In testing each of the methods, two indices were computed from the validation sample (observed data) and predicted values. These two indices are the mean error (ME) and the root mean square error (RMSE). The results show that universal kriging with the slope gradient used as external drift was more precise than the other kriging methods. Further analysis based on the mean ranks of performance confirmed these results.