A "virtual satellite" to map Earth in unprecedented detail
New AI model integrates petabytes of Earth observation data that stitches together data from actual satellite images, radar, climate simulations, and more to map Earth’s land and coastal waters.
AlphaEarth combines multi-source Earth observation data to create consistent, on-demand maps of land use and surface properties. MapBiomas uses the dataset to understand agricultural and environmental changes across Brazil, enabling faster and more accurate conservation strategy development.
The Satellite Embedding dataset powered by AlphaEarth enables organizations like Global Ecosystems Atlas to classify ecosystems into categories (coastal shrublands, hyper-arid deserts) at scale. Compact embeddings reduce storage by 16x, making planetary-scale ecosystem mapping feasible for the first time.
AlphaEarth Foundations integrates optical satellite imagery, radar, and climate data into unified 64-dimensional embeddings that reveal crop stages and variations invisible to raw satellite data. Users can track agricultural changes with 24% lower error rate than competing systems and see through cloud cover in regions like Ecuador.
The model processes 10x10 meter resolution data across terrestrial land and coastal waters, allowing detection of new construction and urban growth. Unified embeddings enable consistent comparison across regions and time periods without relying on single satellite passes.
AlphaEarth integrates 3D laser mapping, radar, and optical imagery to estimate surface properties and track water-related changes. The 64-dimensional embedding representation captures complex environmental signals that inform water resource management decisions.