Spectral Mixture Analysis (SMA) has been used to extract qualitative and quantitative information about common features in an imagery at the subpixel level. At the Walnut Gulch Experimental Watershed (Arizona) this technique was used to produce the soil spectral map and quantify the fractions of each soil endmember on an image from the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) sensor. SMA was applied on the base of image and reference endmembers in a geometrically correced AVIRIS data. Atmospheric correction was performed by using the Atmospheric Removal Program (ATREM). SMA assumes that few features (endmembers) on the ground are responsible for the integrated radiance received by a sensor for a single píxel. SMA also assumes that the at the sensor spectral radiance represents a linear combination of the spectral response o each endmember multiplyied by its individual fraction within a single pixel. Four image endmembers (McAllister and Stronghold soils, green vegetation, and shade) were needed to model AVIRIS data on the base of image endmembers. There was a general agreement between field observation and the delineation of the boundaries of McAllister and Stronghold soil series through the spectral map generated by SMA. Dark soils (Graham and Epitaph soil series) and shade fractions was separated only in the spatial context. A technique called target test was used to detect for the presence of reference endmembers on AVIRIS data. On the base six detected reference endmembers SMA was successfully applied for subset of image pixels.