Автоматизированное распознавание генетических типов рельефа Кольского полуострова по морфометрическим данным

Abstract

The automatic geomorphological mapping based on the Earth’s surface remote sensing data has been developed in recent years. The aim of the research is an attempt of automatic creation of the Kola Peninsula geomorphological map at the morphogenetic legend’s principle. It was gained based on the random forest classification technique. As input data a several geomorphometric variables were used only (the basic ones – elevation, slope angle, curvatures etc., and many those of surface texture). Developing such algorithms for different territories and hierarchical levels of landforms analysis contributes to propagate the techniques of the fast geomorphological mapping. On the training data covering only 1,3 % of the study area with known labels for one of thirteen probable landform types, the reconstruction of geomorphological boundaries and the automatic creation of the geomorphological map were carried out. The accuracy of the resulting map was 81 %. Elevation has the greatest discrimination power according to Kola Peninsula landform types. In addition, some area-based local geomorphometric variables characterizing terrain pattern has big discrimination power. The lowest power relates with the “classical” local-based geomorphometric variables. The results of the work can be used in the development of automated landforms mapping systems at the level of morphogenetic types.