Google DeepMind maps the world's forests at 10-meter resolution

Google DeepMind and the World Resources Institute have built the first global map that tells primary forest from timber plantation and tree crop farm, at 10-meter resolution, for the year 2020. The system, called ForTy, short for Forest Typology, scored 90.2 percent accuracy when two experts reviewed 8,190 randomly selected sites. It beat European Commission maps built specifically for EU deforestation rules on the same test data.

That comparison matters because the project exists largely to serve one regulation. The EU Regulation on Deforestation-free Products (EUDR) prohibits goods linked to forests cleared or degraded after December 31, 2020, including land where primary forest was converted to a plantation. Oil palm and cocoa farms count as agricultural land, so companies assessing supply chain risks need to know exactly how a site was used in 2020. ForTy supplies that reference point.

It could also feed other initiatives that depend on the same kind of data, the authors note, including the Tropical Forest Forever Facility (TFFF) and the UN goal of protecting 30 percent of the world's land by 2030.

Why forest type can't be read from a photo

Satellite maps usually show whether trees grow in a given area. They stop there. A forest's type depends on its history, not just how it looks, DeepMind says. An old planted mixed forest can look like a natural one from space, and a recently cleared plantation still counts as forest by definition even though satellite images show bare ground.

ForTy assigns the world's land area in 2020 to six classes:

• Primary forest
• Naturally regenerating forest
• Planted forest
• Plantation forest
• Tree crops, such as oil palm, rubber, or cocoa
• Other land

Earlier datasets, including one from the European Commission, cover only parts of the world. The authors say ForTy is the first global map of its kind.

How the three-stage system works

ForTy runs on Google DeepMind's AlphaEarth Foundations model, introduced in July 2025 as a kind of virtual satellite. AlphaEarth combines optical satellite images, radar, laser scans, and climate simulations into a compact description of each 10-by-10-meter area, with annual embeddings covering every year since 2017, freely available in Google Earth Engine.

Several image recognition models build on those descriptions. The first learns from verified examples. Its predictions, combined with outside data on tree heights, buildings, and forest loss, become expanded training data for a second model. A final step analyzes several years of data together to smooth outliers. Google's Gemini language model helped adapt older training data to the EU's definitions.

During correction rounds, experts added about 43,000 examples to fix errors such as vineyards in Europe classified as tree crops, natural pine forests in Central America labeled as planted forests, and oil palms missed in Africa and Indonesia.

Some distinctions remain out of reach for satellites alone. Satellite images rarely show how land is managed, so the researchers use a 40-year rotation period to separate plantations from other planted forests. That cutoff does not fit every region.

Where the map falls short

The 90.2 percent headline number gets a large boost from the easy-to-identify "other land" class, the authors warn. Results vary sharply by forest type. The map correctly identified 89 percent of primary forests that experts confirmed on the ground. It found only about half of tree crops and planted forests, and 60 percent of plantations.

Flipping the lens, roughly a quarter to a third of areas mapped as plantations or tree crops turned out to be misclassified. The map was wrong for a third of areas labeled as planted forest. These gaps hit exactly the boundaries that matter most for the EUDR, and the evaluation itself may flatter the results: experts were most often unsure about managed areas, and the team excluded those uncertain cases from its assessment.

The model also inherited a bias from its training data, whose annotators rarely classified forests near roads or settlements as primary forest, even when the forests appeared untouched.

Still, on the same test data, ForTy scored 95.5 percent versus 93.6 percent for the Commission maps on forest versus nonforest classification, and 90.9 percent versus 87.8 percent for forest types.

What the map says about the world's forests

The researchers estimate the world holds about 4.1 billion hectares of forest. Nearly 3.9 billion hectares of that is natural forest, including about 2 billion hectares of primary forest. Planted forests cover about 250 million hectares, and tree crops cover nearly 200 million hectares, mostly in Asia.

Regional errors complicate the picture. The map overestimates primary forest in Asia by about 200 million hectares, largely in Russia's conifer forests. In Africa, it misses about 94 million hectares of forest, likely because of open dry forests and savannas.

Sweden offers a sharp example of why forest type matters. About two-thirds of the country's forest area is planted, and natural forest covers only around one-fifth of Sweden. The authors say overall tree cover there may be adequate, but the share of natural forest, which matters for biodiversity, could be critically low. The Democratic Republic of the Congo, Indonesia, and Brazil hold the highest shares of natural forest.

What's still missing, and what comes next

Some land uses remain hard to classify. In some regions, the map still labels shaded cocoa or coffee grown beneath a natural canopy as forest. Selectively logged or burned forests also pose problems. ForTy does not yet have a separate class for open woodlands, which the EUDR treats separately, though the researchers say future versions should add one. Users can make their own classification choices using the dataset's confidence scores for each class, which let them set their own thresholds.

The map, classification layer, and test data are freely available under a CC BY 4.0 license, including in Google Earth Engine. Google DeepMind has released the code in the GeeFlow and JEO repositories.

ForTy is one of the first large mapping projects built on AlphaEarth embeddings, and Google says it is part of Earth AI. That connection places it inside a broader effort: in fall 2025, Google connected Earth AI to Gemini to help detect floods, droughts, and algal blooms faster. With plantation and tree-crop detection still stuck near 50 to 60 percent, the usefulness of ForTy as an EUDR compliance tool will depend on how quickly the next iterations close those gaps.