Tessera AI model offers accessible way to view Earth

A foundation model trained on Earth observation data from Copernicus Sentinel-1 and Sentinel-2 has been made widely available to researchers, it was announced at a computer industry conference this week in Denver, US. The institutional report frames the development in practical terms and ties it to the broader mission or observing effort.

It matters because Earth science becomes stronger when local observations can be placed inside a broader physical pattern that spans time and geography. The planet operates as a coupled system in which atmospheric, oceanic, cryospheric and solid-Earth processes interact across timescales from days to millions of years. A measurement that captures one variable at one location and one moment has limited interpretive value until it is embedded in the longer series and wider spatial coverage that allow natural variability to be separated from forced change. Crucially, the encoded datasets, called 'embeddings', use far less data than the pixellated images that are transmitted to Earth from satellites. A paper on Tessera was published at the 2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), held 3-7 June.

The model itself was first launched in 2025 and the paper marks the first fully peer-reviewed announcement of Tessera to the scientific community. The foundation model, Temporal Embeddings of Surface Spectra for Earth Representation and Analysis, or Tessera for short, was developed by researchers at the University of.

According to Nuno Miranda, Mission Manager for Sentinel-1 at the European Space Agency (ESA), this is an innovative and exciting step in the development and use of AI in the field. Our embeddings make the data more accessible to users from traditionally unserved communities, especially those from ecology, conservation, plant science and zoology.

We’ve also made these available without requiring registration and at no cost, opening the door to many new classes of critical problems. Optical data from Sentinel-2, and advanced radar data, known as synthetic aperture radar (SAR) data, from Sentinel-1.

The broader interest lies in linking the observation to climatic, geophysical or environmental dynamics that extend well beyond the immediate event or location. Earth science is unusual in that its most important questions operate on timescales that no single research career can observe directly, making the archival record, whether in ice, sediment, rock or satellite data, as important as any new measurement. Results that can be embedded in that record, and that either confirm or challenge the patterns it reveals, carry disproportionate scientific weight.

So, rather than the data-heavy and pixellated imagery from satellites, Tessera compresses data heavy, cloudy satellite imagery to create an embedding layer of Earth data. This model was funded and supported through ESA’s Foundation Models for Climate and Society (FM4CS) project.

Because the account originates with ESA Space News, it functions best as a primary institutional report that is close to the data and operations, not as independent scientific validation. Institutional communications are produced by organizations with legitimate interests in presenting their work in a favorable light, which does not make them unreliable but does make them partial. Details that complicate the narrative, including instrument limitations, unexpected failures and results below projections, tend to be minimized relative to progress messages. Technical documentation and peer-reviewed publications, where they exist, provide the complementary layer that institutional releases cannot substitute.

The next step is to place the result inside longer time series and to compare it with independent instruments and independent sites. Earth system observations gain most of their interpretive power from network density and temporal depth, not from any single measurement however precise. Model simulations that assimilate the new data will help clarify whether the observation fits comfortably within known natural variability or represents a shift that existing models do not reproduce.

Source