Applying the Random Forest Algorithm to Analyze the Dynamics of Taiga-Tundra Forest Ecosystems

The article presents the results of studying the dynamics of taiga-tundra forest ecosystems in the Arkhangelsk Region based on the Earth remote sensing data and the application of image classification using the Random Forest algorithm. The change in the proportion of forested, non-forested and unforested areas in the study area is noted in the forest register. The results of the study show a significant increase in forest area between 2016 and 2023 (by 10.28 %) due to a reduction in non-forested areas and unstocked forest lands. This dynamics is due to the processes of successful natural restoration of non-forest covered areas, as well as their advancement northward taking-up lands due to climate change. The evaluation of the accuracy of automated classification of satellite images using the Random Forest algorithm by comparing them with reference data using criteria such as overall accuracy and the Kappa coefficient (the degree of correspondence between the model estimates and the actual data) has confirmed the reliability of the results obtained. The benchmarks have been taken from inventory databases, stationary sample plot data and the state forest inventory data. Before starting field work, cartographic databases have been studied and sample plots have been selected. Based on experimental data for the study area, a large number of polygons have been created, reflecting the diversity of forest stands and non-forest areas, to train the algorithm for classifying satellite images. Image processing, including corrections, mosaics, geoprojection and return, has been performed using SNAP (Sentinel Application Platform), an open source program. 100 points in various forest vegetation conditions in the study area have been analyzed. Studying the dynamics of forest ecosystems based on the Earth remote sensing data and the application of image classification using the Random Forest algorithm will improve the accuracy of assessing the resource and environmental potential of northern taiga and tundra forests of the Arkhangelsk Region.