Monitoring of Arctic landscape changes under the influence of global warming become an important goal for nowadays geosciences. New methods of machine learning have achieved impressive results in landscape's recognition accuracy. We have analyzed Sentinel imagery (31/07/2020) and obtained index map images in Google Earth Engine. We assessed spatial distribution of identified land classes, which concern to different steps of thermokarst activity. Our work have showed possibility of using random forest method for mapping periglacial landforms.