# Source: ESA WorldCover v200 — Global 10m Land Cover 2021
# Type: cog
# Catalog: Manual
# Created for: LULC benchmark pipeline (lulc-benchmark-pipeline ticket)
#
# ESA WorldCover is the first global 10m land cover map, produced from
# Sentinel-1 and Sentinel-2 data. v200 (2021) improves on v100 (2020)
# with better urban detection and updated methodology. 11 land cover
# classes following the UN-FAO Land Cover Classification System (LCCS).
#
# Reference: Zanaga et al. 2022 — doi:10.5281/zenodo.7254221
# License: CC-BY-4.0

source:
  id: esa-worldcover-v200
  name: ESA WorldCover 2021 v200
  type: cog
  url: https://esa-worldcover.org/en
  description: >
    Global land cover map at 10m resolution for 2021, produced by the
    European Space Agency from Sentinel-1 SAR and Sentinel-2 optical
    data. 11 land cover classes following the UN-FAO LCCS. Distributed
    as Cloud-Optimized GeoTIFFs on a 3-degree MGRS-based tile grid.
    WorldCover v200 achieved 76.7% overall accuracy in independent
    validation (Venter/Xu et al. 2024).
  provider: European Space Agency (ESA)
  license: CC-BY-4.0
  metadata:
    connector_id: null
    tags:
    - Land Cover
    - Global
    - Sentinel-1
    - Sentinel-2
    - 10m
    - ESA
    - WorldCover
    - LCCS
    dataset:
      tier: 1
      thematic_group: Vegetation & Agriculture
      format: cog
      resolution: 10m
      crs: EPSG:4326
      classes: 11
      bands:
      - id: map
        description: >
          Land cover class code: 10=Tree cover, 20=Shrubland, 30=Grassland,
          40=Cropland, 50=Built-up, 60=Bare/sparse vegetation, 70=Snow/ice,
          80=Permanent water, 90=Herbaceous wetland, 95=Mangroves, 100=Moss/lichen
        dtype: uint8
      temporal:
        range: 2021-01-01/2021-12-31
        frequency: static
        type: snapshot
        default_range: "2021"
      quality:
        methodology: >
          Machine learning classification of Sentinel-1 (VV/VH) and Sentinel-2
          (10 bands) annual composites. CatBoost classifier trained on 2.3M
          globally distributed reference samples. Post-processing with ancillary
          data (JRC GSW, OSM urban, GEDI canopy height).
        accuracy: >
          Overall accuracy 76.7% (independent validation, Venter/Xu 2024).
          Best at Tree cover (UA 88%, PA 82%) and Water (UA 95%, PA 92%).
          Weakest at Shrubland (UA 47%, PA 52%) and Herbaceous wetland.
        coverage: global
      layout:
        type: tiled
        spatial:
          scheme: mgrs_3deg
          description: >
            Tiles follow a 3-degree grid based on MGRS. Tile naming uses
            compass direction format: e.g., S06W048 = 6 degrees South,
            48 degrees West. Each tile covers a 3x3 degree area.
        uri_template: "https://esa-worldcover.s3.eu-central-1.amazonaws.com/v200/2021/map/ESA_WorldCover_10m_2021_v200_{tile}_Map.tif"
        parameters:
        - var: tile
          description: >
            MGRS 3-degree tile ID in compass format (e.g., N45E006, S06W048).
            Format: {N|S}{lat}{E|W}{lon} where lat/lon are zero-padded to
            2/3 digits respectively.
      original_source_url: https://esa-worldcover.s3.eu-central-1.amazonaws.com/v200/2021/map/
      homepage: https://esa-worldcover.org/en
      gee_asset_id: ESA/WorldCover/v200