Predicting carbon and nitrogen by visible near-infrared (Vis-NIR) and mid-infrared (MIR) spectroscopy in soils of Northeast Brazil

Determinations of soil carbon and nitrogen stocks are important to evaluate land fertility and agricultural potential and because of their influence on the global climate. Spectroscopic determinations are faster, cheaper and less pollutant than traditional methods. The potential use of spectroscopy in the visible (Vis), near infrared (NIR) and mid-infrared (MIR) regions and their combination to estimate total C and N concentrations were evaluated using seven different pre-processing and two regression models and comparing to the concentrations determined by dry combustion of 701 soil samples from different soil classes and land uses in Northeast Brazil. Better C and N concentration predictions were obtained with the MIR region than with the Vis-NIR region and no significant improvement occurred when the two spectra were combined. The support vector machine (SVM) and the partial least squares (PLSR) models had similar performances both for C and N. The multiplicative scatter correction pre-processing is recommended for C and the standard normal transformation technique for N. Equations to estimate soil C and N concentrations of the predominant soil classes in the region and of the set of all classes are provided. Their high accuracy confirm the potential of reflectance spectroscopy as a useful and rapid tool to quantify C and N concentrations in different tropical soils and under different land uses.