Google DeepMind has released AlphaEarth Foundations, a new AI model that stitches together vast, diverse Earth-observation data into a unified, efficient representation of our planet, says IEEE Spectrum. The model integrates inputs such as satellite imagery (optical, radar), topographic maps, lidar, climate models, and even geotagged text (e.g., from Wikipedia) to produce “embeddings” for every 10 × 10 meter square of land (and coastal areas) on Earth. These embeddings cover each year from 2017 through 2024.
An embedding is a compact numeric summary of many features of a place (vegetation, land cover, climate, etc) distilled into a 64-dimensional vector. DeepMind claims this approach preserves fine spatial and temporal details while reducing storage requirements by about 16× compared to prior methods.
Because many ground truth labels (e.g., field measurements, ecological surveys) are sparse or geographically uneven, AlphaEarth was tested in “data-scarce” scenarios. It still outperformed previous approaches on tasks such as land cover change detection, crop classification, and ecosystem mapping—even with limited labeled data.
Among its use cases are urban expansion monitoring, tracking melting ice, wildfire risk assessment, agricultural cycle detection, and conservation. It helps scientists see patterns that were difficult to detect before, like in areas with frequent cloud cover (e.g., in Ecuador) or remote terrain (e.g., Antarctica).
DeepMind is making the annual embedding dataset (“Satellite Embedding dataset”) available via Google Earth Engine, free for non-commercial use. Over 50 organizations are already using it, from ecosystem monitoring in the Amazon to classification of previously unmapped ecosystems.
AlphaEarth Foundations offers a scalable, accurate, and compact way to monitor Earth’s changes over time. It doesn’t generate new imagery but makes diverse existing data much more usable for science, policy, and environmental action. The model represents a major step forward in how we observe and understand our changing planet.