Reference point samples used in the production of the global maps of annual grassland class and extent for 2000—2022 within the scope of the Global Pasture Wath initiative. The reference samples (estabilished by Feature Space Coverage Sampling-FSCS) comprises 2.3M points visually classified (using Very High Resolution imagery) in: Cultivated grassland, Natural/semi-natural grassland Other land cover The file gpw_grassland_fscs.vi.vhr_tile.samples_20000101_20221231_go_epsg.4326_v1.gpkg aggregates the samples by visual interpretation units ( 1x1 km) and includes the follow collumns: cluster_id: Cluster id defined by k-means (FSCS), cluster_distance: Distance from the sample tile to center of the cluster (FSCS), cluster_size: Size of cluster (strata) defined by the FSCS, priority: Priority used by the visual interpretation, tile_id: Sample tile id, imagery: VHR reference images used by the visual interpretation, min_year: Minimum of year covered by the reference samples, max_year: Maximum of year covered by the reference samples, n_years: Number of years covered by the reference samples, n_samples_c1: Number of reference samples for "Seeded grass" (1), n_samples_c2: Number of reference samples for "Natural / Semi-natural grass" (2), n_samples_c3: Number of reference samples for "Not grass" (3), n_samples_all: Total number of reference samples, The file gpw_grassland_fscs.vi.vhr_point.samples_20000101_20221231_go_epsg.4326_v1.gpkg provides individual points (with 60-m spatial support) and include the follow collumns: sample_id: Sample id deribed by MD5 Hash of columns x, y, imagery and year, x: Longitude in WGS84 (EPSG:4326), y: Latitude in WGS84 (EPSG:4326), vi_tile_id: 1-km tile id, imagery: VHR Reference image used by the visual interpretation (Google; Bing; Interpolated), ref_date: Reference date of VHR image, year: Reference year of VHR image, class: Class id (1: Cultivated grassland; 2: Natural/semi-natural grassland; 3: Other land cover) , class_label: Class labels (Cultivated grassland; Natural/semi-natural grassland; Other land cover) , esa_worldcover_2020: Land cover class labels extracted from ESA WorldCover 2020, glad_glcluc_2000: Land cover class labels extracted from UMD GLAD GLCLUC 2000, glad_glcluc_2005: Land cover class labels extracted from UMD GLAD GLCLUC 2005, glad_glcluc_2010: Land cover class labels extracted from UMD GLAD GLCLUC 2010, glad_glcluc_2015: Land cover class labels extracted from UMD GLAD GLCLUC 2015, glad_glcluc_2020: Land cover class labels extracted from UMD GLAD GLCLUC 2020, glc_fcs30d_2000: Land cover class labels extracted from GLC_FCS30D 2000, glc_fcs30d_2005: Land cover class labels extracted from GLC_FCS30D 2005, glc_fcs30d_2010: Land cover class labels extracted from GLC_FCS30D 2010, glc_fcs30d_2015: Land cover class labels extracted from GLC_FCS30D 2015, glc_fcs30d_2020: Land cover class labels extracted from GLC_FCS30D 2020, ml_cv_group: spatial block CV group (based on vi_tile_id), ml_type: expecify if the sample was used for (1) training or (2) calibration. The file gpw_grassland_fscs.vi.vhr_grid.samples_20000101_20221231_go_epsg.4326_v1.gpkg provides the grid samples (with 10-m spatial support) and include the follow collumns: tile_id: 1-km tile id, bing_class: Class labels (Cultivated grassland; Natural/semi-natural grassland; Other land cover) defined using as reference Bing Maps Images, bing_image_start_date: Start date of the Bing Maps Images used in the visual interpretation, bing_image_end_date: End date of the Bing Maps Images used in the visual interpretation, google_class: Class labels (Cultivated grassland; Natural/semi-natural grassland; Other land cover) defined using as reference Google Maps Images, google_image_start_date: Start date of the Google Maps Images used in the visual interpretation, google_image_end_date: End date of the Google Maps Images used in the visual interpretation, missing_image_date: No images available, same_image_bing_google: Images from the same date available in Google and Bing Maps. The dataset was produced through the QGIS plugin Fast Grid Inspection. Related resources Maps of dominant grassland:2000-2002 2003-2005 2006-2008 2009-2011 2012-2014 2015-2017 2018-2020 2021-2022 Probability maps of cultivated grassland:2000-2022 (All URLs) Probability maps of natural/semi-natural grassland:2000-2022 (All URLs) Grassland reference samples based on VHR imagery (2000–2022):GeoPackage files Global machine learning models (Random Forest):Parquet and joblib python files Reference sampling design derived by FSCV:GeoPackage and raster files Harmonized reference samples based on existing LULC dataset:GeoPackage and raster files Source code for reproducibility:GitHub release Mapping feedback tool:GeoWiki Data catalogues:OpenLandMap STAC Google Earth Engine Support For questions of bugs/inconsistencies related to the dataset raise a GitHub issue in https://github.com/wri/global-pasture-watch