SPATIAL PREDICTION OF SOIL PROPERTIES FROM LANDFORM ATTRIBUTES DERIVED FROM A DIGITAL ELEVATION MODEL

Digital elevation models (DEMs) provide a good way of deriving landfor m attributes that may be used for soil prediction. The geostatistical techniques of kriging and cokriging are increasingly being applied to predicting soil properties. Whereas ordinary kriging (and universal kr iging) utilise spatial correlation to determine the coefficients of th e linear predictor, cokriging involves both inter-variable correlation and spatial covariation among variables. Multi-linear regression mode lling also offers an alternative to predicting a soil variable by mean s of covariation. The performance of predicting four soil variables by these methods and two regression-kriging models are compared. The pre cision and bias of prediction of the six methods were dependent on the soil variable predicted. The mean error of prediction indicates reaso nably small bias of prediction for all the soil variables by almost al l of the methods. With the exception of topsoil gravel, for which mult i-linear regression performed best, the root mean square error showed the two regression-kriging procedures to be best. Further analysis bas ed on the mean ranks of performance by the methods confirmed this. All the kriging methods involving covariables (landform attributes) have a more smoothing effect on the predicted values, thus minimising the i nfluence of outliers on prediction performance. Both the methods of re gression-kriging show promise for predicting sparsely located soil pro perties from dense observations of landform attributes derived from th e DEM. Histograms of subsoil clay residuals show outliers in the data set. These outliers are more evident in multi-linear regression, ordin ary kriging and universal kriging than regression-kriging. There was a clear advantage in using the regression-kriging methods on those vari ables which had a small correlation with the landform attributes: root mean square errors for all the soil variables are much smaller than t hose resulting from any of the multi-linear regression, ordinary krigi ng, universal kriging or cokriging methods.