Google DeepMind’s AI maps Earth with 24% better accuracy

Google DeepMind announced AlphaEarth Foundations, a breakthrough AI system that can map the entire planet’s surface with unprecedented accuracy by integrating massive amounts of satellite data into unified digital representations. The system reduces error rates by 23.9% compared to existing approaches while requiring 16 times less storage space, potentially transforming environmental monitoring and resource management for governments, conservation groups, and businesses worldwide.

How it works: AlphaEarth Foundations processes information by creating “embedding fields” — highly compressed digital summaries that capture essential characteristics of Earth’s surface in 10-meter squares.

• The system maintains sharp 10×10 meter precision while tracking changes over time, allowing organizations to monitor individual city blocks, small agricultural fields, or forest patches.
• Rather than treating each satellite image separately, it weaves together data from optical satellites, radar, 3D laser mapping, climate simulations, and more into a coherent picture.
• The model architecture, dubbed “Space Time Precision” (STP), can create accurate maps for any specific date range, even interpolating between observations or extrapolating into periods with no direct satellite coverage.

Real-world applications: More than 50 organizations have tested the system over the past year, with transformative results across multiple sectors.

• In Brazil, MapBiomas, a collaborative network tracking land use changes, uses the technology to track agricultural and environmental changes across the Amazon rainforest in near real-time.
• The Global Ecosystems Atlas initiative employs the system to create the first comprehensive resource for mapping the world’s ecosystems, helping countries classify unmapped regions like coastal shrublands and hyper-arid deserts.
• The system excels in “sparse data regimes” where ground-truth information is limited, potentially reducing the cost of creating detailed maps for large areas.

What they’re saying: Early adopters emphasize the system’s transformative potential for environmental monitoring.

• “The Satellite Embedding dataset can transform the way our team works,” said Tasso Azevedo, founder of MapBiomas. “We now have new options to make maps that are more accurate, precise and fast to produce — something we would have never been able to do before.”
• “The Satellite Embedding dataset is revolutionizing our work by helping countries map uncharted ecosystems — this is crucial for pinpointing where to focus their conservation efforts,” said Nick Murray, Director of the James Cook University Global Ecology Lab.

Technical breakthrough: The system solves satellite imagery’s biggest challenge by handling missing data and cloud cover that often obscures tropical regions.

• “To the best of our knowledge, AEF is the first EO featurization approach to support continuous time,” the researchers note, meaning it can maintain accuracy even with incomplete satellite coverage.
• In testing evapotranspiration rates — how water transfers from land to atmosphere — AlphaEarth Foundations achieved an R² value of 0.58 while all other tested methods produced negative values, indicating they performed worse than simply guessing the average.

Privacy and accessibility: Google designed the system with environmental monitoring in mind rather than individual tracking.

• The 10-meter resolution cannot capture individual objects, people, or faces, balancing utility with privacy protection.
• The Satellite Embedding dataset, described as “one of the largest of its kind with over 1.4 trillion embedding footprints per year,” is available through Google Earth Engine with annual snapshots from 2017 through 2024.

The big picture: This positions Google at the forefront of “Google Earth AI” — a collection of geospatial models that already power weather predictions, flood forecasting, and wildfire detection systems used by millions in Google Search and Maps.

• The technology democratizes access to sophisticated Earth observation capabilities that previously required significant computational resources and expertise.
• For enterprises involved in supply chain monitoring, agricultural production, or environmental compliance, the system offers new possibilities for verifying sustainable sourcing claims and optimizing operations at planetary scale.