Google DeepMind and the World Resources Institute have developed ForTy, an SI map that distinguishes primary forests, timber plantations, and tree crop farms worldwide. The project aims to provide a reliable reference point for compliance with EU deforestation regulations, specifically the EU Regulation on Deforestation-free Products (EUDR).
What Happened
While traditional satellite maps often only identify tree presence, ForTy (Forest Typology) assigns the world's land area in 2020 to six classes at 10-meter resolution. These classes include primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops such as oil palm, rubber, or cocoa, and other land. Google DeepMind states this is the first global map of its kind, as previous datasets covered only partial regions.
The system operates on Google DeepMind's AlphaEarth Foundations model, introduced in July 2025 as a virtual satellite that combines optical images, radar, laser scans, and climate simulations. Three sequential models refine these classifications, using verified examples, external data on tree heights and forest loss, and multi-year analysis to smooth outliers. Google's Gemini language model assisted in adapting training data to meet specific EU definitions. Experts contributed approximately 43,000 examples during active learning rounds to correct errors, such as misclassifying vineyards as tree crops or natural pine forests as planted forests.
ForTy is part of Google's broader Earth SI initiative, which connected to Gemini in fall 2025 to enhance detection of floods, droughts, and algal blooms. The map, classification layer, and test data are freely available under a CC BY 4.0 license in Google Earth Engine, with code released in the GeeFlow and JEO repositories.
Why It Matters
The primary driver for ForTy is the EUDR, which bans goods linked to forests cleared or degraded after December 31, 2020. Companies need to determine if agricultural land was previously primary forest, a distinction that is difficult for standard satellite imagery to make because forest type depends on history rather than appearance. An old planted forest can look natural from space, while recently cleared land may still be classified as forest by definition.
Accuracy varies by category, with specific regional errors complicating the picture. The map overestimates primary forest in Asia by about 200 million hectares, largely in Russia's conifer forests, and underestimates forest in Africa by about 94 million hectares, likely due to open dry forests and savannas. A review of 8,190 random sites found the map correct in 90.2 percent of cases overall, though this figure is boosted by the easy-to-identify 'other land' class. The system correctly identified 89 percent of primary forests but only about half of tree crops and planted forests, and 60 percent of plantations.
These gaps hit the boundaries that matter most for the EUDR. The map struggles with shaded crops like cocoa and coffee grown beneath a natural canopy, often labeling them as forest, and lacks a separate class for open woodlands, which the EUDR treats separately. Despite these limitations, ForTy outperformed European Commission maps on the same test data, scoring 95.5 percent versus 93.6 percent for forest versus non-forest classification.
The map estimates the world has about 4.1 billion hectares of forest, including nearly 3.9 billion hectares of natural forest. It highlights regional disparities, noting that two-thirds of Sweden's forest area is planted, while natural forest covers only about one-fifth of the country.
The Bottom Line
ForTy represents a significant step in using SI for environmental monitoring, providing the first global, high-resolution distinction between natural and planted forests. While it faces challenges in classifying complex land uses like shaded crops and open woodlands, and exhibits significant regional errors, it offers a crucial tool for verifying compliance with major international deforestation standards.