Table 2.

Approximate accuracy trends of the GHS-BUILT built-up surface datasets across different versions, gathered from evaluation studies based on the MTBF-33 reference dataset. Note that the reported F-1 scores are approximate averages across all counties, and that GHS-BUILT-R2018A was only evaluated in the state of Massachusetts, encompassing 14 out of 33 counties covered by MTBF-33.

GHSL release	Source data	Spatial resolution	Epoch	F1 (most rural)	F1 (most urban)	Data reference or related publication	Validation study
GHS_LDSMT_2015	Landsat	38 m	2014	0.11	0.85	https://doi.org/10.2788/253582	Leyk et al., 2018
GHS_LDSMT_2017	Landsat + Sentinel-1	30 m	2014	0.05	0.67	https://doi.org/10.1080/20964471.2019.1625528	Uhl et al., 2018
GHS-BUILT-S1	Sentinel-1	20 m	2016	0.05	0.60	https://doi.org/10.1080/20964471.2017.1397899	Uhl et al., 2018
GHS-BUILT-R2018A	Landsat + Sentinel-2	30 m	2014	0.45	0.80	https://doi.org/10.2905/jrc-ghsl-10007	Uhl & Leyk, 2022a
GHS-BUILT-S2	Sentinel-2	10 m	2018	0.52	0.63	https://doi.org/10.2905/016D1A34-B184-42DC-B586-E10B915DD863	This study