Harmonized global land cover maps at 10 m (2020) and 30 m (annual, 2020 to 2024) resolutions were generated using a deterministic, rule-based ensemble framework that integrates primary spatial backbones with specialized thematic and contextual layers. Using ESA WorldCover, GLAD GLCLUC, and GLC_FCS30D as baseline inputs, the workflow sequentially applies prioritized conditional rules to refine generic categories into up to 44 detailed hierarchical classes. The classification incorporates specialized crop, grassland, forest, and coastal data sets alongside topographic, spectral, and coastal-distance variables to account for landscape contexts. Finally, both resolution streams were independently validated at Level 1 against reference samples from the Global Ensemble Land Cover compilation.