Global Quant Geospatial Model “TESSERA” Unlocks New Earth Analytics for 2026
Artificial Intelligence & Machine Learning
One of the most promising machine-learning tools emerging in satellite analytics and remote-sensing research is TESSERA, a foundation model created to generate unified, information-rich representations of Earth from space. The model introduces scalable time-series processing for satellite imagery, advancing everything from ecosystem monitoring to carbon and biomass assessments.
TESSERA, short for “Temporal Embeddings of Surface Spectra for Earth Representation and Analysis,” is an open-source AI framework designed to extract meaning from the continual stream of satellite data. Unlike task-specific models, TESSERA functions as a general-purpose representation system, producing compact latent embeddings of Earth’s surface at 128 dimensions and at a granular 10-meter spatial resolution. Its capabilities draw from both Sentinel-1 and Sentinel-2 data pipelines, enabling both optical and radar modalities to be incorporated into one unified feature space.
TESSERA captures long-term temporal signatures rather than scraping individual static frames, giving analysts the ability to study seasonal variations, land-use transitions, vegetative health shifts, and surface changes that standard snapshot-based methods often overlook. The high resolution, combined with global coverage, positions the model as a highly scalable tool for ecological science, agricultural monitoring, urban infrastructure tracking, and climate-based analytics.
Because it is open-source and self-supervised, TESSERA becomes particularly accessible in research and industry contexts. Users can deploy its representations without re-training domain-specific variations, accelerating prototyping cycles and shortening experimentation times for geospatial data scientists. Its documentation and pipelines allow for seamless integration into established GIS workflows.
Where does TESSERA have application?
The model offers immediate relevance to multiple quantitative geospatial domains:
- Crop-Type Prediction
Time-series spectral characteristics allow agricultural analysts to label crop varieties, monitor growth stages, and anticipate yield estimates. - Carbon and Biomass Assessment
Forest height and canopy density can be estimated with improved fidelity, supporting carbon-credit accounting and deforestation monitoring programs. - Wildfire Surveillance
Temporal embeddings spotlight vegetation stress, recovery cycles, and burn scars, helping quantify wildfire extent and impact in near real-time. - Large-Scale Environmental Monitoring
The global resolution enables characterization of habitat loss, water-body changes, and land degradation patterns across continents.
How Can Researchers Access the Model?
TESSERA’s embeddings can be leveraged directly via GeoTessera, the associated platform offering programmatic access and documentation. Analysts, academics, and practitioners can incorporate the precomputed embeddings within their own geospatial pipelines, including Python-based tooling and conventional GIS platforms.
By introducing a unified temporal embedding framework for Earth observation, TESSERA may pave the way toward the next wave of remote-sensing analytics — one in which quant models incorporate geospatial context, spectral patterns, and temporal signatures. For climate economics, insurance risk, agricultural commodity forecasting, carbon-credit validation, infrastructure planning, and natural-resource monitoring, foundation models like TESSERA represent an important new class of quantitative tools.
Join us: QuantVision 2026: Fordham’s Quantitative Conference
This image of Earth’s city lights was created with data from the Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS). Originally designed to view clouds by moonlight, the OLS is also used to map the locations of permanent lights on the Earth’s surface. The brightest areas of the Earth are the most urbanized, but not necessarily the most populated. (Compare western Europe with China and India.) Cities tend to grow along coastlines and transportation networks. Even without the underlying map, the outlines of many continents would still be visible. The United States interstate highway system appears as a lattice connecting the brighter dots of city centers. In Russia, the Trans-Siberian railroad is a thin line stretching from Moscow through the center of Asia to Vladivostok. The Nile River, from the Aswan Dam to the Mediterranean Sea, is another bright thread through an otherwise dark region. Even more than 100 years after the invention of the electric light, some regions remain thinly populated and unlit. Antarctica is entirely dark. The interior jungles of Africa and South America are mostly dark, but lights are beginning to appear there. Deserts in Africa, Arabia, Australia, Mongolia, and the United States are poorly lit as well (except along the coast), along with the boreal forests of Canada and Russia, and the great mountains of the Himalaya. Data: Marc Imhoff/NASA GSFC, Christopher Elvidge/NOAA NGDC; Image: Craig Mayhew and Robert Simmon/NASA GSFC – https://visibleearth.nasa.gov/view.php?id=55167 (image link)
High Value Source Links
https://geotessera.readthedocs.io
https://github.com/earth-ai/tessera
https://sentinel.esa.int/web/sentinel/missions/sentinel-1
https://sentinel.esa.int/web/sentinel/missions/sentinel-2
