Sub-pixel estimation of tree cover and exposed surface density using Landsat-TM and IKONOS images.

Abstract. The sub-pixel analyses are capable to generate continuous fields representing the variability of a certain thematic objective. This work had general objective develops numeric models to represent the variability of the tree cover and expose surfaces within each forested formation. This research was led in the riparian buffer inside sub-watershed medium of São Francisco to the North of Minas Gerais, Brazil. IKONOS and Landsat TM image were used with GUIDE algorithm to constructions models. The generated results were: two numeric models of regression tree and the index image of tree cover and expose surface extrapolated for the image TM. The use of the model non parametric and non linear by regression tree presented satisfactory result for depict Wetland, Deciduous and Savanna patterns of forest formation.

Key Word: Regression Tree, Tree Cover, Expose Surface, Riparian Buffer, Geoprocessing