Google DeepMind and the World Resources Institute have launched ForTy, a global map that categorizes land into six classes at a 10-meter resolution based on 2020 data. The project aims to support compliance with the EU Regulation on Deforestation-free Products, which restricts goods linked to forest degradation after December 31, 2020.
The system utilizes the AlphaEarth Foundations model, introduced in July 2025, to process satellite imagery, radar, and climate data. Developers refined the model using 43,000 expert-verified examples and the Gemini language model to align classifications with EU definitions.
Validation tests showed a 90.2 percent accuracy rate, though performance varies by category. The map correctly identified 89 percent of primary forests, while accuracy for tree crops and plantations was lower. The tool outperforms existing European Commission datasets in forest versus non-forest classification.
Global estimates from the project indicate approximately 4.1 billion hectares of forest, including 2 billion hectares of primary forest. The dataset, code, and classification layers are available under a CC BY 4.0 license via Google Earth Engine and the GeeFlow and JEO repositories.
