Introduction to AlphaEarth Foundations
On July 30, 2025, the AlphaEarth Foundations team introduced a new AI model that integrates petabytes of Earth observation data to generate a unified representation, revolutionizing global mapping and monitoring.
Satellites capture rich images and measurements daily, providing scientists with a near real-time view of our planet. However, the complexity, multimodality, and refresh rate of this data present challenges in effectively connecting disparate datasets. To address this, we launched AlphaEarth Foundations, an AI model that functions like a virtual satellite, accurately characterizing terrestrial land and coastal waters by integrating vast amounts of Earth observation data into a unified digital representation.
Main Features and Applications
AlphaEarth Foundations enables scientists to gain a more complete and consistent picture of Earth's evolution, aiding informed decisions on critical issues like food security, deforestation, urban expansion, and water resources. To accelerate research and unlock use cases, we released a collection of annual embeddings from AlphaEarth Foundations as the Satellite Embedding dataset in Google Earth Engine. Over the past year, we collaborated with over 50 organizations to test this dataset in real-world applications, yielding significant benefits.
Technical Highlights
AlphaEarth Foundations addresses two major challenges: data overload and inconsistent information. It aggregates information from various public sources, including optical satellite images, radar, 3D laser mapping, and climate simulations, analyzing the world's land and coastal waters in sharp 10x10 meter squares. Additionally, it creates highly compact summaries for each area, requiring 16 times less storage space than other AI systems tested, significantly reducing the cost of planetary-scale analysis.
Performance Testing and Feedback
In comparisons with traditional methods and other AI mapping systems, AlphaEarth Foundations consistently demonstrated superior accuracy across various tasks over different time periods, achieving a 24% lower error rate than the models tested, especially in scenarios with scarce label data.
Future Outlook
AlphaEarth Foundations marks a significant advancement in understanding the dynamics of our changing planet. We believe that when combined with general reasoning LLM agents like Gemini, its annual embeddings could be even more useful. We will continue to explore the application of our model's time-based capabilities as part of Google Earth AI to address the planet’s most critical needs.
Blogger's Review:
The launch of AlphaEarth Foundations not only enhances the precision and efficiency of global mapping but also provides robust data support for environmental monitoring and resource management. As AI technology advances, its application potential will expand significantly, making it worth watching.