Optimising interpolation as a tool for use in soil property mapping

[摘要] Inverse distance weighting (IDW) and kriging are robust and widely used estimation techniquesin earth sciences (soil science). Variance of Kriging is often proposed as a statistical techniquewith superior mathematical properties such as a minimum error variance. However, therobustness and simplicity of IDW motivate its continued use. This research aims to compare thetwo interpolation techniques (Inverse Distance Weighting and Kriging), as well as to evaluatethe effect of sampling density on mapping accuracy of soil properties with diverse spatialstructure and diverse variability in a quest to improve interpolation quality for soil chemicalproperty mapping.The comparison of these interpolation methods is achieved using the total error of crossvalidationand validation statistics. Mean Prediction Error and Root Mean Square Error arecalculated and combined to determine which interpolator produced the lowest total error. Theinterpolator that produced the lowest total error portrays the most accurate soil propertypredictions of the study area.The finding of this study strongly suggests that the accuracy achieved in mapping soil propertiesstrongly depends on the spatial structure of the data. This was clearly visible, in that, when thesubset training data set was decreased, the total error increased. The results also confirmedthat systematic sampling pattern provides more accurate results than random sampling pattern.The overall results obtained from the comparison of the two applied interpolation methodsindicated that Kriging was the most suitable method for prediction and mapping the spatialdistribution of soil chemical properties in this study area.

[发布机构] University of the Free State