# A year of satellite observation, made lighter to carry.

TESSERA provides 10-meter, pixel-wise annual embeddings learned from Copernicus Sentinel-1 and Sentinel-2. ESA says the model first launched in 2025; its repository and a peer-reviewed CVPR 2026 paper now document the work.

Canonical: https://brightaifuture.com/discoveries/tessera-earth-embeddings
Format: discovery
Source publication: 2026-06-04
Bright publication: 2026-09-19
Substantive update: None recorded
Evidence and review: Demonstrated; confidence: high; approved; ai-assisted. AI-assisted editorial comparison with the cited primary sources, explicit evidence limits, and held alternatives. Publication authorized by the site owner on 2026-09-19; no human source review or independent replication is claimed.

## The human problem

Small research teams may need patterns from large satellite archives without the storage and compute to process every raw scene.

## The prior constraint

Earth-observation archives are large, multi-sensor, and costly to preprocess for each downstream task.

## AI’s actual role

The model compresses a year of radar and optical observations into reusable per-pixel representations.

## The documented result

ESA announced the peer-reviewed CVPR work on 4 June 2026; the paper reports 10-meter annual embeddings, and the linked repository provides implementation material. The available sources do not establish that every artifact was first released on that date.

## Why it may matter

Reusable representations can lower the starting cost of a local mapping project while leaving validation with the user.

## Limitations

An embedding discards information and is not a universal map.

Local crop, fire, or land-use claims require labels and independent validation.

Repository and data licenses must be checked separately.

## Unresolved questions

Which regions and seasons show weaker representations?

How much compute and expertise does local fine-tuning require?

What tasks fail compared with raw imagery?

## Provenance and history

{
  "dates": {
    "eventDate": "2026-06-04",
    "publicationDate": "2026-06-04",
    "captureDate": "2026-09-19",
    "lastReviewedDate": "2026-09-19"
  },
  "provenance": {
    "origin": "editorial",
    "externalId": "https://github.com/ucam-eo/tessera"
  },
  "revisions": [
    {
      "id": "revision:sept26-tessera-01",
      "recordedAt": "2026-09-19",
      "summary": "Initial open Earth-embedding draft.",
      "sourceIds": [
        "source-tessera-esa",
        "source-tessera-papers",
        "source-tessera-code"
      ]
    }
  ],
  "corrections": []
}

## Original sources

- [Tessera AI model offers accessible way to view Earth](https://www.esa.int/Applications/Observing_the_Earth/Copernicus/Tessera_AI_model_offers_accessible_way_to_view_Earth)
- [TESSERA — A pixel-wise earth observation foundation model](https://geotessera.org/papers/)
- [ucam-eo/tessera: TESSERA is a foundation model that can process time-series satellite imagery](https://github.com/ucam-eo/tessera)

## Continue exploring

- [Open intelligence](https://brightaifuture.com/worlds/open)
- [Frontier](https://brightaifuture.com/worlds/frontier)
- [What changes when computing leaves Earth?](https://brightaifuture.com/threads/space)
- [What changes when powerful models become open-weight?](https://brightaifuture.com/threads/open)