Mapping of the Khabarovsk Region arable lands by machine learning using Sentinel-2 images

Automated classification of arable lands using machine learning methods is one of the most important tasks in the transition to digital agriculture. The classification of arable lands in the Khabarovsk Region was carried out using random forest (RF), minimum distance (MD), and K-means clustering methods based on Sentinel-2 images for July, August, September, and October 2022. The values of spectral bands, EVI (Enhanced Vegetation Index), and NDVI (Normalized Difference Vegetation Index) were considered as input data. Based on the results of statistical processing, it was found that the RF method demonstrated the greatest stability when changing the date of shooting and the type of input data. The accuracy of recognition of arable lands in the Khabarovsk Region in 2022 was 92.5% when using NDVI values calculated from the September Sentinel-2 image in the classifier. The proposed approach can be used for automated classification and subsequent mapping with expert correction of arable lands in the southern part of the Far East.