Abstract: The objective of this study was to evaluate the use of a laboratory and an orbital sensor on the classification of soils in a complete 180 ha bare soil area located in Brazil. The 180 points were located as a regular grid of 100x100 m (one point per ha). In each point, soil samples were georreferenced (GPS) and collected (0-20 and 80-100 cm depths). Based on the traditional soil analysis and field work (with 18 profile evaluation), a detailed soil map was developed as to be our “real pattern”. This soil map determined 18 soil classes and 53 polygons. Other soil maps were developed based on the following described methods. The first method was based on the orbital image (landsat) interpretation. It was used a color composition 5R, 4G, 3B. Based on the visual interpretation it was determined 16 classes of soils and 35 polygons. A statistical method was used to compare the traditional soil map with the color composition soil map. The traditional soil map was more accurate although the color composition had several. The second method was performed with laboratory sensor information. Spectral data (400-2500 nm) was acquired from soil samples (both depths of each point). Data was modelized and cluster analysis grouped the spectral curves. The third method consisted on the evaluation of the surface soil information (obtained in laboratory but convoluted for the landsat spectral bands). With this method 9 groups were discriminated. The fourth method was determined by quantitative analyses of each pixel information extracted from a processed and reflectance transformed landsat image. The number of groups determined were nine. The main conclusion was: Any sensor method reached the detailed soil map information.