MAPPING VEGETATION TYPES ON DIFFERENT SLOPES AND ASSESSING THE DYNAMICS OF THEIR CHANGE OVER A LONG PERIOD OF TIME

Abstract

Vegetation mapping is one of the most important tasks of the remote sensing in the field of environmental protection. The use of modern remote sensing technologies has many advantages, including significant time savings, coverage of large areas and provision of effective long-term observation. These methods can significantly speed up the execution of tasks compared to traditional ground-based methods. The article considers a method for land cover classification based on the analysis of Planet Scope satellite images of time series for the period from 2019 to 2023. The purpose of this study is to analyze the influence of various input data on the classification results and the dynamics of vegetation changes over the selected period. A study was conducted in the north-eastern part of East Kazakhstan using the Random Forest (RF) algorithm to select the most suitable parameters and analyze changes in vegetation types. The results of land cover classification using five different combinations of classification features and an assessment of the overall accuracy are presented. According to the results of the study, it was found that using a combination of such input parameters as spectral channels, topographic indices and tree crown height layer gives the best results in land cover mapping. The vegetation cover was classified into three types: trees, shrubs, and grass cover, and the areas of water resources and bare soil were also determined. The analysis showed that between 2019 and 2023, the total area of ​​trees and water bodies decreased, while the areas of grass cover, shrubs, and bare areas increased. Therefore, the results of the study can contribute to further analysis of ecosystem changes and identify important changes in vegetation structure and ecosystem processes.