The new GHSL description of global built-up (GHS-BUILT) areas relies on novel classification methods applied for joint assessment of Sentinel Multi Spectral Instrument (MSI), Landsat Enhanced Thematic Mapper (ETM), Thematic Mapper (TM), and Multi Spectral System (MSS) remotely sensed imagery data organized in five collections: 2018, 2014, 2000, 1990, and 1975.

Radiometric, textural and morphological image features extracted from the 10-m resolution Sentinel-2 data composite support the new classification of the residential (RES) vs. the non-residential (NRES) or industrial built-up surfaces and the new land vs. water surface assessment at 10m of spatial resolution. RES and NRES built-up Markers are salient points in the data labelling built-up structures with high confidence. They are generalized to the Morphological Settlement Zone delimiting the perimeter of the human settlements with 10-m resolution spatial detail.

Multiscale morphological decomposition of the Sentinel-2 image data supports the inferential model predicting the new continuous sub-pixel built-up surface fraction information incorporated in the GHS-BUILT, improving the previous assessments based on Boolean classification.

A new Composite of the Maximal Cumulative Divergent Confidence process applied to the historical satellite image data supports the automatic multiple-sensor, multiple-date change detection applied to the built-up surfaces. A new image segmentation and classification schema improves the consistency of the change detection across different sensor resolutions available in the different epochs, going back to the origin of the civilian remote sensing program.

Previous classification of Built-up surfaces obtained from Landsat and Sentinel 1-2 satellite data are augmented by the injection of semantic labels. These were extracted from sub-meter resolution imagery by automatic processes (Facebook, Microsoft building footprints), from available land use land cover data, and from volunteered geographical information (Open Street Map data): they are used as prior knowledge and learning set in the GHS-BUILTclassification process.

At present date, the new GHS-BUILT production process is at an advanced state but not finalized yet. The data showed here should be considered preliminary and for demonstrative purposes only. They may include gaps and inconsistencies that will be fixed in the final quality control process. The validation will be based on direct visual inspection of the image data, using a random stratified sampling approach. Digital cartography data reporting the building outlines in 500 test areas around the globe are used to assess the error in the built-up surface fraction estimates.