SPECTRAL BAND SELECTION FOR QUANTIFYING SELECTED PROPERTIES IN HIGHLY WEATHERED SOILS

Accurate estimates of soil properties pertinent to separation of different soils in the field would enhance soil survey efforts. Spectral data obtained from eight soils from northern Alabama using a field multiband radiometer were used to evaluate the effectiveness of spectral reflectance in differentiating among soils of similar properties and to identify spectral band combinations useful in estimating selected soil properties. The wavelength bands used were 0.45 to 0.52, 0.52 to 0.60, 0.63 to 0.69, 0.76 to 0.90, 1.15 to 1.30, 1.55 to 1.75, 2.08 to 2.35, and 10.40 to 12.50-mu-m. Correlation, regression, and discriminant analyses were used in analyzing the data. High correlations were found among the spectral data and the soil variables studied. The soil types were sufficiently differentiated based on either the spectral properties or the physical and chemical properties of the soils. The results indicate that selective bands of the electromagnetic spectrum are important in the development of models that will significantly predict soil properties from spectral data. A set of important spectral bands was identified in the regression equations developed for the prediction of the soil properties. The thermal spectral band was found to be significant in contributing to the separability of soil types which were pedogenically and characteristically related, hence significant in the prediction of clay, organic matter, and iron oxide content of the soils.