Researchers at the University of Cambridge and international partners have expanded access to the Tessera AI model, a foundation model built using Earth observation data from the European Copernicus programme’s Sentinel-1 and Sentinel-2 satellites.

The system has now been formally introduced to the scientific community through a peer-reviewed paper presented at the 2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

The Tessera AI model transforms massive volumes of satellite imagery into compact datasets known as embeddings, allowing scientists to analyse environmental change without processing raw imagery.

By combining radar and optical satellite observations, the model creates detailed annual representations of Earth’s surface at a 10-metre resolution.

The result is a significantly more accessible way to work with Earth observation data. Researchers can generate maps, track environmental change and develop predictive models using far less computing power and labelled training data than traditional approaches require.

The Tessera AI model explained

Tessera, short for Temporal Embeddings of Surface Spectra for Earth Representation and Analysis, is an AI foundation model designed specifically for Earth observation applications.

Developed through a collaboration involving the University of Cambridge, Finland’s Aalto University and several European partners, the model processes satellite data collected by the Copernicus Sentinel missions.

Unlike conventional satellite imagery, which consists of large pixel-based image files, Tessera compresses years of observations into searchable embeddings.

These embeddings capture patterns and changes over time, enabling users to study how landscapes evolve rather than simply viewing individual snapshots.

The model combines optical imagery from Sentinel-2 with synthetic aperture radar (SAR) data from Sentinel-1. This fusion enables Tessera to overcome common limitations, such as cloud cover, while providing a more complete picture of environmental conditions.

How Tessera improves Earth observation analysis

One of the most significant advantages of the Tessera AI model is its ability to simplify access to complex satellite datasets.

Traditional Earth observation workflows often require extensive computing resources, specialised expertise and large quantities of labelled training data.

Tessera’s pretrained embeddings already contain learned representations of environmental patterns, reducing the effort required to build analytical models.

Because the datasets are highly compressed, researchers can access and analyse them on standard laptops and even mobile devices. This lowers barriers for scientists working in fields such as ecology, conservation, agriculture and biodiversity monitoring, where advanced computing infrastructure may not be readily available.

The platform also follows FAIR data principles, ensuring that datasets remain findable, accessible, interoperable and reusable across the research community.

Applications across agriculture, climate and conservation

The Earth observation capabilities enabled by Tessera support a broad range of scientific and environmental applications.

Researchers can use the model to monitor crop development, assess vegetation health, measure wildfire damage, analyse forest canopies and identify landscape changes over time. The system also allows users to search for geographically similar regions and detect patterns across large areas.

A UK-based conservation project is already using Tessera embeddings to evaluate habitat changes in protected areas across Cumbria.

The initiative aims to help policymakers assess the effectiveness of environmental protection measures and agricultural subsidy programmes using satellite-derived evidence.

As governments and organisations increasingly rely on Earth observation data to guide climate and conservation strategies, tools that reduce analytical complexity could play an important role in improving decision-making.

Open-source alternative to proprietary AI systems

A notable feature of the Tessera AI model is its open-source architecture.

While several commercial organisations are developing AI systems for satellite data analysis, many operate as closed platforms that limit transparency and independent validation.

Tessera’s developers have chosen to make both the model and its datasets freely available without registration requirements.

This approach allows researchers to modify, adapt and improve the system for specialised use cases while supporting greater reproducibility in scientific research.

The model also provides an alternative to proprietary Earth observation foundation models that generate similar compressed representations of satellite data but do not offer public access to their underlying methods.

Europe’s growing leadership in Earth observation AI

Tessera forms part of a broader European effort to advance artificial intelligence for Earth observation.

The European Space Agency (ESA), through its Φ-lab innovation programme, has supported the development of several foundation models designed to extract insights from satellite data.

Among these are Thor, a multimodal model trained on Sentinel-1, Sentinel-2 and Sentinel-3 data, and TerraMind, a geospatial AI system developed with IBM Research Europe.

Unlike Tessera, which creates annual embeddings that summarise long-term environmental change, both Thor and TerraMind focus on learning from individual observations while preserving detailed spatial information.

TerraMind also introduces cross-modal reasoning capabilities by combining satellite imagery with land-use data, elevation information and geographic context. This enables more sophisticated interactions with Earth system datasets and expands the potential applications of geospatial AI.

Expanding access to global environmental intelligence

The release of the Tessera AI model marks an important development in the evolution of Earth observation technologies.

By converting vast quantities of satellite imagery into compact, information-rich embeddings, the platform makes advanced environmental analysis accessible to a wider research community.

As demand grows for better tools to monitor ecosystems, agriculture, climate impacts and biodiversity, foundation models such as Tessera are helping bridge the gap between raw satellite data and actionable insights.

Its open-source framework, combined with global Earth observation coverage, positions the model as a significant new resource for scientists seeking to understand environmental change at scale.