Remote clustering of pastures

Abstract

This article discusses the problem of pasture clustering, where the objective is to separate the territory into groups with similar characteristics of vegetation cover and conditions of use. The authors propose to perform multi-time clustering of pastures using the k-means technique in SNAP software. The latter was applied to the remote sensing data of the Normalized Difference Vegetation Index (NDVI) for March-October 2023 based on the images obtained from the Sentinel-2 satellite group. The number of clusters that were discovered changed from three to five as a consequence of data processing. The authors decided to investigate the option with three clusters after comparing the collected data with the results of ground study. Planning resource management, vegetation assessment, yield forecasting, and controlling soil degradation can all be optimized with the help of clustering results. This important ecosystem can be managed and protected more effectively with the help of the suggested clustering technique, which divides the region into homogeneous pasture sections in a stable and informative manner. The suggested approach may be applied to different climate zones, and the authors intend to investigate this potential further in their future research.