Tessera Foundation Model Revolutionizes Earth Observation Data Access for Researchers

A new foundation model designed for Earth observation, named Tessera, has been made publicly accessible to researchers, as announced at a computer industry conference in Denver, Colorado. Developed by a collaboration involving the University of Cambridge and various European and global partners, Tessera utilizes data from Copernicus Sentinel-1 and Sentinel-2 satellites, providing high-accuracy datasets that represent what the satellites capture of the Earth’s surface throughout the year.

Tessera generates compressed datasets, known as ’embeddings,’ which significantly lower the amount of data required compared to traditional pixelated images from satellites. This innovative approach permits users to create detailed maps for diverse applications, such as monitoring agricultural health, assessing burnt areas, and analyzing forest coverage. The model was initially launched in 2025, and the recent presentation at the 2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition marked its first formal peer-reviewed introduction.

Researchers highlight numerous advantages of the Tessera model for the Earth observation field. Its pretrained embeddings allow for the rapid identification of patterns and temporal changes, enabling even those without AI expertise to tackle remote sensing challenges on a global level with minimal labeled data. This makes the technology accessible via personal devices, eliminating the need for extensive computational resources. Moreover, as an open-source project, Tessera can be adapted for various applications, offering significant potential for scientific exploration.

Nuno Miranda, Mission Manager for Sentinel-1 at the European Space Agency (ESA), described the introduction of Tessera as a pioneering advancement in the application of AI to remote sensing data. He emphasized that the model effectively demonstrates real-world use of Sentinel-1 and Sentinel-2 data, enhancing analysis capabilities for Earth’s systems. Co-leader of the project, Professor Srinivasan Keshav from the University of Cambridge, noted that Tessera is designed to transform the handling of extensive data generated by the Copernicus program, particularly benefiting users in ecological, conservation, and biological sciences.

Tessera processes large volumes of remote sensing data by integrating optical data from Sentinel-2 with synthetic aperture radar data from Sentinel-1. The result is a time-series representation of Earth features that permits researchers to visualize changes over the year rather than static snapshots of landscapes. This capability is further enhanced by tools that allow users to search for similar geographic features, monitor landscape alterations, and predict trends in vegetation health and urban development.

One ongoing project using Tessera involves assessing the effectiveness of the UK’s nature protection initiatives using satellite data to monitor habitat changes in areas designated for conservation in northern England’s Cumbria region. This collaboration, engaging Tessera with the Endangered Landscapes and Seascape Programme and other national partners, seeks to provide empirical insights into government investment in agriculture and conservation.

Tessera is part of a broader movement to employ foundation models in Earth observation, with ESA also developing several other models through its open innovation laboratory, Φ-lab. These models, including Thor and TerraMind, focus on varied approaches to Earth observation data integration, further pushing the boundaries of geospatial intelligence. While Tessera emphasizes embedding long-term datasets, these newer models aim to leverage individual observations while maintaining detailed spatial context.

Overall, Tessera signifies a paradigm shift in the accessibility and usability of Earth observation data, fostering transparency and encouraging innovative approaches to ecological monitoring and analysis within the research community.