Relationships between the mineralogical and chemical composition of tropical soils and topography from hyperspectral remote sensing data

Airborne Visible/InfraRed Imaging Spectrometer (AVIRIS) data acquired in 1995 in the vicinity of Campo Verde, central Brazil, were analyzed to investigate the relationships between the mineralogical and chemical composition of different soil types and topography. Band depth analysis following the continuum removal separated areas of exposed soils from the other scene components. Principal component analysis (PCA) was applied to this subset of pixel spectra of exposed soils. The Spectral Feature Fitting (SFF) technique was used for mineral identification in the scene. Regression relationships between the silica/aluminum ratio and the absorption band depth values at 2210 nm (kaolinite) and 2260 nm (gibbsite), calculated from laboratory spectra after the continuum removal, were used to estimate the Ki index (1.7SiO2/Al2O3) in AVIRIS spectra, an indicator of the degree of soil weathering in soil surveys. Results were plotted over an Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)-derived digital elevation model showing that several soil properties varied with surface elevation. Low reflectance soils (e.g., Rhodic Acrustox with negative first component (PC1) scores, clay texture and higher organic carbon, SiO2, Al2O3, Fe2O3 and TiO2 content) predominated at higher elevations and high reflectance soils (e.g., Ustic Quartzipsamments with positive PC1 scores, sand texture and lower content of these constituents) at lower elevations. In some portions of the study area, soil composition changed gradually from Rhodic Acrustox at higher elevations to Xanthic Acrustox at lower elevations, or from lower average soil reflectance (negative PC1 scores) and deeper kaolinite absorption band to higher average soil reflectance (positive PC1 scores) and deeper gibbsite absorption band. When applied to AVIRIS data, the laboratory-derived relationship between the Ki index and the depth of the 2210 and 2260 nm absorption bands showed a good correspondence with spectral fits for kaolinite and gibbsite. Kaolinitic areas were associated with high Ki values due to a high SiO2 and low Al2O3 content, whereas gibbsitic areas corresponded to low Ki values and highly weathered soil surfaces due to a low SiO2 and high Al2O3 content.