Thematic accuracy assessment of land cover/use products requires reliable reference data that enable their qualitative and quantitative evaluation. Such dataset with up-to-date information on a predefined class composition and spatial distribution is rarely available and its preparation requires an appropriate methodological approach adjusted to a specific product.Development of a new pan-European land cover/use map, generated from Copernicus Sentinel-2 data 2017 within the Sentinel-2 Global Land Cover (S2GLC) project carried out under a programme of and funded by the European Space Agency, provided an opportunity to design and develop an unique dataset dedicated to validation of this product. The dataset was prepared by twofold stratified random sampling. The first selection designated validation sites represented by Sentinel-2 image tiles and was performed on a country level with county borders used as a stratum. In the second selection validation samples were chosen randomly within the validation sites with stratification based on classes of the CORINE Land Cover database.The final dataset composed of samples visually checked by experienced image interpreters consists of a total number of 52,024 samples spread over the European countries. The samples represent 13 land cover/use classes including artificial surfaces, natural material surfaces (consolidated and un-consolidated), broadleaf tree cover, coniferous tree cover, herbaceous vegetation, moors and heathland, sclerophyllous vegetation, cultivated areas, vineyards, marshes, peatbogs, water bodies and permanent snow cover. Each sample provides information about the occurrence of one of the predefined land cover or land use classes within an area of 100 m² represented by a single pixel (10 m size) of Sentinel-2 imagery for the year 2017. The described dataset was used for the accuracy assessment process of the product Land Cover Map of Europe 2017 resulting from the S2GLC project and provided an estimate of the overall accuracy at the level of 86.1%.

S2GLC - Land Cover Map of Europe 2017 reference dataCBK PAN, http://s2glc.cbk.waw.pl/extensionAttribute table fields:'S2GLC' – a land cover/use class symbol according to the S2GLC classification system'TILE' – a symbol of the Sentinel-2 granule (a tile of the Military Grid Reference System)'NAME_ENG' – a country English name (valid for inland and coastal areas). Data source of country names and administrative boundaries: 'Countries, 2020 - Administrative Units - Dataset' of European Commission, Eurostat (ESTAT), GISCO, https://ec.europa.eu/eurostat/web/gisco/geodata/reference-data/administrative-units-statistical-units/countriesClassification system:111 - Artificial surfaces211 - Cultivated areas221 - Vineyards231 - Herbaceous vegetation311 - Broadleaf tree cover312 - Coniferous tree cover322 - Moors and heathland323 - Sclerophyllous vegetation331 - Natural material surfaces335 - Permanent snow cover411 - Marshes412 - Peatbogs511 - Water bodiesData projection: Lambert Azimuthal Equal Area (LAEA)EPSG: 3035For more technical information on this dataset please refer to Malinowski et al. (2020).