At a computer vision conference in Denver this June, researchers unveiled Tessera—a powerful, freely available artificial intelligence model that transforms how scientists can study Earth from space. Rather than wrestling with massive, cloudy satellite images, the global scientific community can now tap into compressed Earth observation data that works on a laptop or even a mobile phone, opening unprecedented possibilities for monitoring everything from crop health to forest fires.
Tessera, which stands for Temporal Embeddings of Surface Spectra for Earth Representation and Analysis, represents a breakthrough in making satellite science accessible. Developed by researchers at the University of Cambridge alongside partners at Aalto University in Finland and global collaborators, the model was first launched in 2025 and has now received its first full peer-reviewed validation from the scientific community.
The model works by fusing two types of satellite data from the European Space Agency's Copernicus program: optical imagery from Sentinel-2 and advanced radar data from Sentinel-1. Rather than delivering the data-heavy, pixelated images that satellites traditionally transmit to Earth, Tessera compresses these datasets into what researchers call "embeddings"—streamlined layers of information that capture what happened at each point on Earth's surface throughout an entire year. Working at a 10-meter resolution, the same as Sentinel-2's highest capability, Tessera has processed data spanning from 2017 to 2025 across the globe.
What makes Tessera transformative is not just what it does, but who can now use it. The pretrained embeddings capture patterns and changes over time that other methods must laboriously learn from scratch. This means non-AI experts can solve complex remote sensing problems at a global scale using only a fraction of the labeled data previously required. Because the model is open-source, freely available without registration, and lightweight enough to run on modest computing hardware, it opens doors for researchers from traditionally underserved communities—ecologists, conservation scientists, plant scientists, and zoologists who previously lacked the computational resources or technical expertise to tap into satellite datasets at scale.
The applications are already evident. Users can search for geographic regions similar to each other, track landscape changes over time, and make predictions about vegetation health and urban growth. One U.K.-based project is already using Tessera to evaluate the government's nature protection schemes by tracking habitat change through satellite data.
Nuno Miranda, Mission Manager for Sentinel-1 at the European Space Agency, called the release "an innovative and exciting step" in applying AI to Earth observation. Srinivasan Keshav, the University of Cambridge professor who co-leads the Tessera project, emphasized the democratizing impact: the team has addressed the fundamental challenge of making sense of Copernicus's vast data streams, creating tools that let researchers across disciplines understand Earth's systems more efficiently and without expensive infrastructure.
As climate change accelerates and biodiversity faces unprecedented pressure, the ability for scientists worldwide to rapidly analyze satellite data at no cost and without technical barriers could prove decisive. Tessera demonstrates how open-source AI and public satellite data can combine to give the global research community the tools they need to monitor, understand, and ultimately protect our planet.