Principal vegetation types in a natural area close to the city of Rome as observed by ERS-1 SAR and Landsat TM

The identification of plant communities in natural areas is fundamental to many applications. Among these are conservation and management of natural resources, production of thematic maps, and monitoring of relevant phenomena such as expansion of built-up areas and loss of grazing and cultivated lands. Multi-temporal radiometric characteristics as derived from ERS- 1 SAR and Landsat TM images have been analyzed for an area, located 30 km NW of Rome, which was studied by extensive ground surveys. The purpose of this study was to compare the information content of SAR images and Landsat TM bands analyzed together with topographical data, such as elevation, slope, and aspect, in order to develop an efficient classification methodology. Prior to classification, SAR images were corrected for the effect of scanning angle. Homogeneously vegetated areas found at different angle in the images allowed evaluation of the efficiency of this method. Canonical variable analysis was used to compare SAR and TM data in combination with topographic variables. After that, all the variables available were selected by a stepwise discriminant analysis for classification purposes. The variables considered in the model were used for the classification of test areas with an algorithm of linear discriminant analysis.