Finer-Resolution Mapping of Global Land Cover: Recent Developments, Consistency Analysis, and Prospects

摘要: Land-cover mapping is one of the foundations Earth science. As a result combined efforts many scientists, numerous global land-cover (GLC) products with resolution 30 m have so far been generated. However, increasing number fine-resolution GLC datasets imposing additional workloads as it necessary to confirm quality these and check their suitability for user applications. To provide guidelines users, in this study, recent developments currently available (including three thematic four different types, i.e., impervious surface, forest, cropland, inland water) were first reviewed. Despite great toward improving accuracy that there decades, current still suffer from having relatively low accuracies between 46.0% 88.9% GlobeLand30-2010, 57.71% 80.36% FROM_GLC-2015, 65.59% 84.33% GLC_FCS30-2015. The reported maps vary 67.86% 95.1% eight surface reviewed, 56.72% 97.36% seven forest products, 32.73% 98.3% six cropland 15.67% 99.7% water products. consistency was then examined. showed good overall agreement terms spatial patterns but limited some vegetation classes (such shrub, tree, grassland) specific areas such transition zones. Finally, prospects also considered. With rapid development cloud computing platforms big data, Google Engine (GEE) greatly facilitates production by integrating multisource remote sensing advanced image processing classification algorithms powerful capability. synergy spectral, spatial, temporal features derived satellite stored will definitely improve spatiotemporal In general, up now, most not able achieve maximum (per class or overall) error 5%–15% required Therefore, more are needed especially which has wetland, tundra, maps.