Abstract
A group of 91 soils from arid-zone regions were used to develop a nondestructive near-infrared analysis (NIRA) procedure for estimating carbonate concentration in soils. Carbonate concentration was measured gasometrically and reflectance in the NIR was measured by a high-resolution FT-NIR instrument. The reflectance data were converted to absorption and derivatized, and statistical procedures were applied to obtain the multivariate regression equation between the various manipulated optical data and carbonate concentration. Reflectance from preheated (600°C) and nonheated soils was compared to test the ability to remove spectral interferences caused by hydroxyl groups. Forty-three soils were used for calibration, and subsets were selected by different strategies which either maximized variation of soil properties (as measured by their specific surface area) in the selected set, or minimized it. A validation set of 48 soils was used to test the prediction-ability of the various calibration equations obtained. Most of the wavelengths entered into the calibration equations were spectroscopically explainable as belonging to carbonate absorption features. Satisfactory prediction was obtained, however, in only one combination of conditions (heated soils; 21 samples in calibration set; second derivative of absorption). Standard error of prediction (SEP) was similar to standard error of calibration (SEC) in this case (7.9% and 5.9%, respectively) if 7 samples belonging to one soil type (Xerochrept) were omitted from the validation set. We conclude that carbonate in soils in the concentration range of 10 to 75% can be satisfactorily estimated for pedological and soil classification purposes by the rapid and nondestructive NIRA procedures.