Quantitative comparative analysis of cropland masks of the Bryansk region

Abstract

A comparative quantitative analysis of croplands masks of the Bryansk region was conducted by an expert method using the archive of remote sensing data from Google Earth. The first mask, with a spatial resolution of 250 m per pixel, was exported as part of the FAO GSNmap project. This mask is a product of the Land Cover global map provided by the Copernicus Global Land Service and is based on PROBA-V satellite images. The second mask was obtained from the Land Use/Land Cover (LULC) global map which is derived from Sentinel-2 satellite data. The mask was exported at a spatial resolution of 250 m per pixel. The overall accuracy of the Agrochemical Service, FAO, and LULC masks was calculated based on the visual interpretation of 900 randomly selected land plots in the Bryansk region. The accuracy values were 88.11% for the FAO and 88.00% for the LULC mask, and the kappa indices were 0.68 for FAO, and 0.67 for LULC mask. A comparative analysis of the FAO and LULC masks revealed a similarity of 81% between them. The conducted research indicates a high level of reliability of the cropland masks of the Bryansk region, based on satellite data from PROBA-V and Sentinel-2.