HYBMAP

Land-cover products provide important data support for monitoring and analyzing land surface changes. In this context, many products have been developed for use by researchers. However, comparisons between land cover products reveal disagreement between them, suggesting that individual products remain highly uncertain. Here, we mixed four sets of land cover products (MODIS-IGBP, ESA CCI, GLC_FCS30 and GlobeLand30) to produce hybrid maps covering the globe from 2000 to 2020 with a resolution of approximately (0.0083333°, 30''). The results showed that HYBMAP effectively reduced the disagreement between data sets and improved the agreement by 8.7%. The results of validation samples showed that HYBMAP improved the classification accuracy of land-cover products with relatively high accuracy. The overall accuracy of HYBMAP for level-0 classification system (nine coarse land-cover types), level-1 classification system (eleven land-cover types), and level-2 classification system (17 land-cover types) was 86.2% (kappa: 0.822), 80.5% (kappa: 0.765) and 75.9% (kappa: 0.733), respectively. From the perspective of time series, the area change of each type of land cover in the world is relatively stable with good time continuity