What you get

Two outputs for any area and year you choose:

Training data

Every point carries a confidence tier, an independent-verification flag and spectral features, delivered as GeoJSON and CSV with a README and metadata.

Quality by class

ClassDefinitionLabel purity

Label purity is the share of labels confirmed by independent reference data: 1,595 expert-checked points in six African cities, and EU field-survey points for tree cover and cropland.

Water subtypes

Every Water point also has a water_type, based on how the water looks in the imagery. These are spectral subtypes, not measurements of water clarity or quality: use Water as one class unless you specifically need them.

Known limitations

Land-cover map beta

Every 10 m pixel classed as Water, Tree cover, Other vegetation, Built-up or Bare ground, with a confidence layer showing where the map is unsure. Delivered as a GeoTIFF with a QGIS style, class areas and a README. Your first 3 maps are free, up to 100 km² each.

How far to trust each class

Maps are new: their accuracy has not yet been measured on independent reference points, so check them against the imagery before relying on them.

Help us improve Xadov

Two minutes, all optional except a rating or a comment.

What did you try?
How useful is Xadov for you?
1 = not useful, 5 = very useful
What would make it more useful? (pick any)
Which would you pay for?