Spectroscopy is widely recognized as an effective tool for the analysis of soil constituents. It is an efficient alternative to conventional laboratory analysis that can be cumbersome, time consuming and expensive. The majority of studies on the use of spectroscopy focus on the spectroscopic modeling to predict soil properties. However, information derived from spectra can also be used to describe the soil and how it varies across landscapes. The reason is that spectra contain information on the fundamental composition of soil: its organic matter, and iron oxide, clay and carbonate minerals as well as on water and particle size. In this study we use visible near infrared (vis–NIR) spectra to describe topsoils across Denmark. We used 693 agricultural topsoil samples (0–30cm) from the Danish soil collection and measured them with a vis–NIR spectrometer. Spectra were collected in the range between 350–2500nm. We interpreted the soils by gleaning the organic and mineralogical information from the spectra. To summarize the information content in the spectra we performed a principal component analysis (PCA). The first three PC’s explained 94% of the variability in the spectra. The scores from the PCA were clustered using k-means to help with interpretation. Soil properties of the clusters were described using the mean spectrum of each class. We mapped the scores of the first three principal components using ordinary kriging. These maps and a cluster map derived with k-means clustering were used in the final discussion. The spectroscopic cluster map showed clearly soils with large clay contents, soils that are predominantly sandy, those that are silty and those with large amounts of organic matter, respectively. Their distribution reflects the general pattern of soil variability in Denmark.
Soil Science Society of America. Journal, 2013, Vol 77, Issue 2, p. 568-579