A comparative analysis of the applicability of typical land cover datasets in the karst strongly heterogeneous terrain in Southern China

The international scientific community pays close attention to global land surface remote sensing mapping. Accurate LULC datasets are crucial for sustainable development assessments at various scales. Taking the southern China's karst as an example, this study selected three sets CRLC-10m (CRLC Land Cover), ESRI-10m (ESRI Land Cover), and ESA-10m (ESA World Cover) were select for 2020 to conduct correlation analysis, spatial consistency analysis and accuracy evaluation. The results indicate that: (1) There is good correlation between ESRI-10m and ESA-10m; (2) Only 3.67% of the study area shows full consistency across all three datasets, with good spatial consistency in crop, forest, grassland types; (3) The overall accuracy of the land cover data sets in the study area is CRLC-10m (55.40%), ESRI-10m (55.26%), and ESA-10m (49.40%), and the accuracy of CRLC-10m is less affected by slope changes. This study provides theoretical guidance for selecting LULC data and enhancing remote sensing classification accuracy in karst regions.