The deciduous forests (SDTF) are a set of remaining isolated, composed of species of seasonal phenology, scattered throughout the globe and with high levels of endemism. The agricultural and urban expansion has considerably reduced the footprint of this vegetation is found in South America and the largest nuclei in better condition. Because of this distribution in patches, these forests have close contact with several other faces ecotones resulting in poorly known to science. This study investigated the species-environment relationships in ecotones of Cerrado SDTF-and-SDTF semideciduous forests in southeastern Brazil, to understand the degree of reciprocal influence in the communities of trees, deciduous areas under ecological stress. For this, we sampled the tree species with CAP (circumference at breast height) 10 cm in 180 plots of 400 m2, equally distributed in six ecotonais remaining in the São Francisco, three with transition between SDTF- cerrado and three-SDTF semideciduous forests. Besides the biotic sampling were conducted in each plot topographic data collection and composite samples of soil for analysis of micro-nutrients and fertility. In addition, climatic data were obtained from each area of the World Bank Clim. Classification of vegetation and species distribution patterns were evaluated by techniques nonmetric multidimensional scaling (NMDS) and the type piecewise regressions. The NMDSs were made for the amount of coverage (VC) and data preseça-absence (binary matrix). The realação species- environment was assessed retrospectively using multiple linear regressions that related predictor variables (variables soil, topography and climate) with the response variables (scores of axes 1 and 2 and VC-NMDSs of binary, wealth species, basal area and density). The presence of spatial was observed in the construction of all types and corrective measures are taken when needed, in order to avoid inflation of the error type I. We found a clear relationship between species composition and environmental gradient in terms of soil fertility and climatic variations. The species composition showed high spatial autocorrelation, especially within the remnant, as well as for non-stationary data resulting from geographic isolation between areas. Multiple linear regression models showed that climatic variables influence the distribution of species between areas, as well as edaphic variables conditioned the distribution of species on a local scale. The analyzes showed that ordering many species are shared by the faces and the similarities are greater between SDTFs and semideciduous seasonal forests than among SDTFs and Cerrado. These patterns were observed in both local and regional scale
