The article raises questions about landscapes mapping using remote sensing data. The method of land classification on the basis of Ensemble Learning concept based on effective ensemble systems was described in details. The developed method is based on a combination of different monoclassifiers in a single system via a metaclassifier component. In the construction of the ensemble systems it is necessary to carry out the selection of monoclassifiers, determine the metaclassifier algorithm, train mono- and metaclassifier and evaluate the quality of the ensemble based classification system. Metaclassifier learning algorithm based on the method of monoclassifiers weighted voting was described. Weight coefficients are calculated on the basis of the confusion matrix metrics. In the presented solution, in contrast to the already proposed decision-making methods, metaclassifier allows to consider the individual classifiers accuracy to determine specific classes of geoobjects. In addition to this the authors introduces the concept of a lower threshold of accuracy, which determines the possibility of the classifier system to make the hypothesis of class accessory. The approbation of method was carried on the base of test polygons «Cheberchinka», «Inerka», «Smolny», located at the edge of the Volga Upland. Landsat-7 satellite imagery was selected as input. The approbation of ensemble systems for geophysical data analysis made it possible to identify the advantages of using the proposed classification method over existing approaches (begging, busting, stack generalization, voting). The experiments confirmed that the use of ensemble systems can improve the final accuracy, objectivity and reliability of the analysis.