We investigated the landscapes and vegetation cover of the Ubur-Dzokoi hollow of the Selenga middle mountains. It was found that according to the physical-geographical zoning, the study area is a hummocky terrain of dry steppe landscape. The geosystems of the hollow include steppe and forest-steppe communities. Barchan relief formation is due to human activities and the dynamics of the wind regime. An analysis of key morphometric parameters of hollow relief is carried out based on the SRTM digital elevation model. As a result of the automated classification of Landsat TM multispectral images with a further identification of landscape types, multitemporal (for the years 1990, 2000 and 2010) landscape maps are created. These maps are validated by field landscape and geo-botanical investigations. The algorithm of the ISODATA unsupervised classification method was used, which allows the outlines with non-contrast structure of spectral brightness to be identified. The maps show clear changes in vegetation cover, indicating a decrease in the forest area, a gradual replacement of sedge-forb steppe communities with free-standing trees as well as an increase of sand deposits on unused arable and fallow lands. The spatial structure of vegetation on aeolian landforms was studied. It was concluded that their main component is psammophyte vegetation. We have made 25 geobotanical descriptions. The floristic composition of the coenocomplex includes 52 species of higher vascular plants. The ecological structure of the florocoenocomplex was identified on the basis of analyzing groups of plants in relation to the factor of soil moisture content. The method of ordination was used to determine the ecological differentiation of vegetation. To identify the spatial and structural organization of vegetation used geobotanical profiles and geobotanical descriptions of the sites, and satellite imagery. A large-scale (1:10 000) vegetation map for aeolian landforms was compiled. A correlation of the species in environmental groups showed a predominance of xerophytes (48 %).Keywords: vegetation, psammophytes, Landsat, automated classification, spatio-temporal dynamics, mapping.