Comparison of Linear and Nonlinear Kriging Methods for Characterization and Interpolation of Soil Data

Abstract

Characterization and analysis of large quantities of existing soil data represent highly complicated tasks because of the spatial correlation, uncertainty, and complexity of the processes underlying soil formation. In this work, three linear kriging (simple kriging, ordinary kriging, and universal kriging) and three nonlinear kriging (indicator kriging, probability kriging, and disjunctive kriging) algorithms are compared to determine which is best suited for the characterization and interpolation of soil data for applications in transportation projects. A spherical model is employed as the experimental variogram to aid the spatial interpolation and cross-validation. The kriged data are subjected to leave-one-out cross-validation. The data used are in both vector and raster format. Statistical measures of correctness (mean prediction error, root-mean-square error, standardized root-mean-square error, average standard error) from the cross-validation are used to compare the kriging algorithms. Using indicator and probability kriging with the vector data set yielded the best results.