Earth Index Imagery V7

A global, cloud-free yearly mosaic of Sentinel-2 L2A imagery for 16 August 2025 to 16 August 2026, produced by Earth Genome for Earth Index. It covers 17,849 MGRS tiles between roughly 60.5°S and 70.3°N. All bands except B10 (not present in L2A) are available at their original native resolution.

This is the V7 generation and supersedes earthindeximagery. The V7 embeddings were computed from this imagery.

This data is licensed under the Creative Commons Attribution 4.0 international license, which, in summary, only requires attribution. It is derived from Copernicus Sentinel-2 data.

Resources

• Earth Index: explore and search this imagery in the browser
• Earth Genome
• STAC API: https://stac.earthgenome.org, collection sentinel2-yearly-mosaics-v7. Search by place, MGRS tile or date.
• STAC Browser, a human-friendly interface to the same catalog
• Embeddings computed from this imagery: earthindex-embeddings-v7

The href of each STAC asset points to Earth Genome's own S3 bucket (s3://ei-imagery-sentinel2-prd/v7_global_2025-08-16_2026-08-16/…). The identical files are in this repository under the same relative path.

How it is made

Each tile is built with the s2mosaic library from Sentinel-2 L2A Collection 1 scenes (Earth Search STAC):

• Cloud masking with OmniCloudMask (OCM).
• A medoid composite, so each pixel is a real observation and spectral relationships between bands are preserved (not a per-band median).
• Seasonal-first scene selection with vegetation weighting (veg_weight 0.25). Geographic overlays adjust the selection for snow, rainforest and corn-belt regions. Up to 25 scenes are used per tile.

The exact parameters, the scenes that contributed, and the coverage are recorded on every STAC item (composite:* properties) and in each tile's scene_metadata.json.

Naming

Each tile is a directory named by MGRS tile, then start and end date: 31TDF_2025-08-16_2026-08-16/ holds the bands for the Barcelona tile (31TDF). The directory name equals the STAC item id.

Files in each tile directory (15)

All rasters are cloud-optimized GeoTIFFs (512 × 512 blocks), on the native 10 m grid of the tile (10,980 × 10,980 px, one UTM zone per tile, e.g. EPSG:32631 for 31TDF). Lower-resolution bands (20 m, 60 m) are resampled to 10 m, so every band in a tile shares the same grid. The CRS of each tile is in its STAC item as proj:code.

Unlike the original earthindeximagery, V7 is not reprojected to Web Mercator.

STAC properties worth knowing

• composite:scene_ids, composite:scene_dates, generated_from: the Sentinel-2 L2A scenes used
• composite:scene_count, composite:valid_pixels, composite:coverage_pct: how much of the tile has data
• composite:cloud_mask, composite:library, composite:veg_weight, composite:snow_weight: how it was built
• processing:software, processing:git_commit: the code that produced the tile

Example

The same objects are readable over S3: s3://us-west-2.opendata.source.coop/earthgenome/earthindeximagery-v7/.

Contact

Using this? Want to work together? Ping us at info@earthgenome.org or info@earthindex.ai