Vol. 11, No. 1, 25-32, 2012

Abstract
Laboratory imaging spectroscopy as a tool for studying the three-dimensional properties of soils is introduced. Hyperspectral images of parallel slices of a soil core sampled in a custom-made steel box were made. We used image spectra from chemically analysed samples to train Support Vector Machine and Partial Least Squares regression models of chemical soil properties. Special attention has to be paid on the correct number of estimator variables. The models were applied on the images to create sub-millimetre scale maps of carbon, nitrogen, iron, aluminium and manganese content in the soil profiles.