Spatial Prediction of Topsoil Texture in a Mountain-plain Transition Zone Using Unvariate and Multivariate Methods Based on Symmetry Logratio Transformation

Abstract

High-resolution soil texture maps are essential for land-use planning and other activities related to forestry, agriculture and environment protection. The objective of the article was to find suitable methods for predicting soil texture through comprehensive comparison of different prediction methods (e.g., univariate and multivariate methods) by completely taking account of its characteristics as composition data with the same auxiliary information. This article, taking elevation as auxiliary variable, predicted the soil texture using univariate [ordinary kriging (OK)] and multivariate [i.e. regression kriging (RK), simple kriging with locally varying means (SKlm), and cokriging (COK)] methods. Soil texture was transformed by symmetry log ratio (SLR) to meet the requirements of the spatial interpolation for the compositional data. The root mean squared errors (RMSE), the relative improvement (RI) values of RMSE and Aitchison's distance (DA) were utilized to assess the accuracies of different prediction ...