Soil discrimination using diffuse reflectance Vis–NIR spectroscopy in a local toposequence

Vis–NIR spectroscopy is nowadays presented as a possible routine method for soil sample analysis. However, there is still no consensus on which is the best multivariate statistical method to use. We propose to use principal component analysis to complete the spectral data treatment. The soil samples came from a pedological cover made up of red–yellow Latosols: 88 samples of 11 soil profiles on four toposequences were collected; clay, organic matter, silica, iron, aluminum and titanium total contents were determined; the contents of goethite, hematite, gibbsite, and kaolinite were calculated. Diffuse reflectance Vis–NIR spectroscopy at wavelengths from 400 to 2400 nm combined with principal component analysis (PCA) was sufficiently sensitive to discriminate different Latosols. Wavelengths of 700 nm and 2200 to 2300 nm were influenced by content ratios of organic matter and iron oxides (700 nm), and kaolinite and gibbsite absorption (2200 and 2300 nm). The spectral responses were affected not only by the content of these constituents, but also by the composition of the minerals, so that the same class of Latosol may have different or similar spectral responses. The role of microaggregation is discussed.