Global Land Cover

What this dataset is

A harmonised, planet-wide land-cover map at 10-metre resolution. Each pixel is assigned to one of eleven surface classes — water, trees, grass, crops, shrub & scrub, flooded vegetation, built area, bare ground, snow & ice, clouds, and no-data.

The classification model is re-trained every year on the most recent Sentinel-2 imagery and validated against an independent set of ground-truth points distributed across all biomes.

What you get

• A single global mosaic for each year, delivered as a Cloud-Optimized GeoTIFF, ready to read with rasterio, gdal, or any modern GIS.
• Pre-built vector tiles at zoom levels 0–14 for fast web visualisation.
• A confidence map for every classification, indicating how certain the model was at each pixel.

Typical uses

• Tracking deforestation, urban expansion, and cropland change.
• Generating inputs for hydrological, climate and ecological models.
• Building base maps for journalism, advocacy and education.

Notes

For year-over-year change analysis, prefer the land cover change dataset, which uses a consistent classifier across years rather than the per-year model.

Suggested attribution