The objective of this research was to determine a methodology for the use of spectral data obtained in the laboratory and from orbital images for the recognition and discrimination of three soil classes in the State of São Paulo, Brazil. Reflectance data were obtained by a spectroradiometer/by Landsat 5 images in the laboratory as standards from soil samples of Oxisol, Ultisol, and Entisol profiles. Orbital data were converted to reflectance and classified by the Spectral Angle Mapper. The spectral curves obtained by a spectroradiometer in the laboratory showed three different patterns, based on the differences in the shape of the curves, soil mineral absorption features, and the reflectance intensity. The Oxisol presented smaller reflectance due to the clayey texture and greater iron content, in relation to the sandier Ultisols and Psament. The characteristic absorption bands of kaolinite (2,200 nm), OH-, and water (1,400 and 1,900 nm) occurred in all soils. Sandy soils such as the Psament presented a spectral curve with positive tendency. The soils were discriminated by the images (bands 5 and 7), being the terrestrial sensor data an important support in this evaluation. The soil line showed spectral distinction for both orbital and terrestrial data, indicating an individual tendency for each soil. The spectral standards obtained in the laboratory were important for the efficiency of the methodology to identify the same in the satellite images. The methodology was efficient to detect areas with exposed soil in the image. The methodology showed that the soil classes used in this work can be recognized and discriminated by orbital data. The obtained results of the digital classification indicated that this technique can be used to support soil survey at a semi-detailed level of high intensity.
soil classification; remote sensing; digital classification; satellite images; soil survey; spectroradiometry
