What ForTy does
Google DeepMind and the World Resources Institute have published ForTy, short for Forest Typology, a global map that assigns the world’s land surface in 2020 to six classes at 10-meter resolution: primary forest, naturally regenerating forest, planted forest, forest plantation, tree crops such as oil palm, rubber or cocoa, and other land. According to the authors, it is the first global map of its kind; previous datasets, including the European Commission’s, cover only parts of the world.
Traditional satellite maps indicate whether trees grow in an area, not what type of forest or crop it is. ForTy is based on DeepMind’s AlphaEarth Foundations model, which compresses data from multiple satellites into a compact description of each pixel; image-recognition models are then layered on top, with a final step that analyzes multiple years to smooth out anomalous values. The Gemini language model helped adapt older training data to European definitions.
Why it matters for European rules
The project’s driving force is the European regulation on deforestation-free products (EUDR), which bans commodities linked to forests felled or degraded after December 31, 2020, including those produced after the conversion of primary forest into plantation. Because oil palm and cocoa plantations are considered agricultural land, companies assessing supply-chain risks must know how a site was used in 2020. ForTy provides that reference point and, according to the authors, could feed other initiatives based on the same data, including the Tropical Forest Forever Facility and the UN goal to protect 30% of land by 2030.
The limits of accuracy
The verification involved review by two experts of 8,190 randomly selected sites, with a correctness rate of 90.2%. The authors warn, however, that the overall score is inflated by the “other land” class, which is easier to identify. Results vary widely: the map identified 89% of primary forests confirmed on the ground, but only about half of tree crops and planted forests and 60% of plantations. About a quarter or a third of areas mapped as plantations or tree crops were found to be misclassified, and the map was wrong for a third of areas labeled as planted forest.
These gaps hit precisely the boundaries most relevant to the EUDR, and uncertainty in the assessment could make the results look slightly better than they are: experts were often uncertain about managed areas, and doubtful cases were excluded. Despite these limitations, ForTy outperforms the European Commission’s maps created for the EUDR on the same test data: 95.5% versus 93.6% in distinguishing forest from non-forest, and 90.9% versus 87.8% in classifying forest types.
Regional errors and open data
The map also inherits a weakness in the training data, whose annotators rarely classified forests near roads or settlements as primary, even when they appeared intact. Regional errors complicate the picture: the map overestimates primary forest in Asia by about 200 million hectares, especially in Russian coniferous forests, while in Africa it misses about 94 million hectares of forest, probably because of open dry woodlands and savannas. In Sweden, about two-thirds of forest area is mapped as planted and only about one-fifth as natural forest; according to the authors, overall tree cover may be adequate, but the share of natural forest, which matters for biodiversity, may be critically low. The Democratic Republic of the Congo, Indonesia and Brazil have the highest shares of natural forest.
Some land uses remain difficult: in certain regions, cocoa or coffee shaded under a natural canopy is still labeled as forest, and selectively logged or burned forests pose problems. ForTy does not yet have a separate class for open woodlands, which the EUDR treats separately, but the researchers plan to add it in future versions. Users can set their own thresholds using the per-class confidence scores. The map, classification layer and test data are available free of charge under a CC BY 4.0 license, including in Google Earth Engine; the code is in the GeeFlow and JEO repositories. The researchers estimate about 4.1 billion hectares of forest worldwide, of which nearly 3.9 billion are natural forest and about 2 billion are primary forest, plus about 250 million hectares of planted forests and nearly 200 million hectares of tree crops, largely in Asia.
