Accuracy Assessment of ESA WorldCover 2020 and ESRI 2020 Land Cover Maps for a Region in Syria

Obtaining high-accuracy land use/land cover (LULC) maps is very costly; therefore, the use of the global LULC maps may prove to be very effective in several domains, including for geographic information system (GIS) applications. However, it is essential to assess the accuracy of LULC maps before using them. The objective of this study was to assess the accuracy of the two latest free and high-resolution (10 m) global land cover maps, namely, European Space Agency (ESA) WorldCover2020 and Environmental Systems Research Institute (ESRI) 2020 Land Cover in Tartous governorate (Syria), using geospatial techniques. Two sampling methods are used to assess the accuracy of the two LULC maps: stratified random sampling (SRS) and equalized stratified random (ESR). The overall accuracy (OA) and kappa coefficient were calculated, in addition to other accuracy assessment indicators. In the first method, the OA of the ESA WorldCover2020 map increased from 74.4 (global accuracy) to 78.3% (local accuracy). On the other hand, the OA of the ESRI 2020 Land Cover map decreased by 16%, from 86 (global accuracy) to 74.1% (local accuracy). The kappa coefficient for the ESA WorldCover2020 map was 0.72 against 0.62 for ESRI 2020 Land Cover map. In the second method, the OA reached 70% and 73% for the ESA WorldCover2020 map and the ESRI 2020 Land Cover map, respectively. Furthermore, the kappa coefficient for the ESA WorldCover2020 map was 0.66 against 0.69 for ESRI 2020 Land Cover map. A comparison was also carried out between the eight classes of two LULC maps. In addition, the map of agreement and disagreement between ESA and ESRI land cover 2020 datasets was created, which highlighted that the disagreement was not uniformly distributed in the study area.