Rapid estimation of soil variability from the convex hull biplot area of topsoil ultra-violet, visible and near-infrared diffuse reflectance spectra

Abstract

Rapid prediction of properties to describe soil variability is essential for site-specific crop management. Accurate predictions require the collection and analysis of a large number of soil samples, which is laborious and costly, and sometimes impossible. On the other hand, the diffuse reflectance spectrum of a soil sample provides multivariate data that are often related to various physical and chemical properties. One way to calculate and plot the spectral variation of different soil types is with the principal component biplot. For the soil that we studied, the first two principal components represent more than 90% of the variation among soil spectra. Our objective was to evaluate the hypothesis that the convex hull biplot area of a geographical region is proportional to the soil variation found in that region. An initial experiment that includes two different geographical regions suggested that the region, which was more variable in relation to pH, has a larger biplot area than the less variable one. A further analysis of the biplots of five fields and the associated variances of pH, organic carbon and clay showed that there was no direct relationship between the convex hull biplot area and the variation in the above soil properties. In this case, the convex hull biplot area might be a combined result of many soil variables, which we have not measured individually. The question of whether the spectral biplot area is a good and quick method of measuring multi-property soil variation is still open.