Many MindedA 10-meter map that tells natural forest from plantation
Google DeepMind and Google Research, working with the World Resources Institute and IIASA, released Forest Typology (ForTy) 2020 on October 22, 2026. It is a global map that assigns every 10-by-10-meter patch of land to one of six classes based on FAO definitions: primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops and agroforestry, and other land. An early-release time series covers each year from 2017 to 2025 and shows conversions such as natural forest turning into plantations or farms.
The model was trained on about 1.7 million labeled locations from expert surveys, national and regional datasets, and targeted reviews of places where it struggled. A key input is AlphaEarth Foundations embeddings, compact summaries of years of satellite observations for each 10-meter pixel. ForTy also gives a probability score per class, so users can set their own thresholds, for example a strict one for EU deforestation regulation (EUDR) risk assessment. It follows last year's Natural Forests of the World 2020, which separated natural forest from all other tree cover.
This matters because EUDR rules turn on forest type. Coffee, cocoa and palm oil must not come from land that was forest in 2020 and has since been converted to agriculture, and wood products must not come from primary or naturally regenerating forest converted to plantation. A shared, open-resolution baseline that separates these classes makes such rules cheaper to check, and gives conservation and carbon accounting a clearer picture of what is real forest.
Caveats: the accuracy figures come from the developers' own validation with WRI and IIASA. Over 9,000 random sites, each reviewed by two trained experts, ForTy named the right forest type 90% of the time and separated forest from non-forest 95% of the time, and the developers say it outperforms other global forest maps. The 2017–2025 series is an early release. Some classes depend on land use rather than appearance, which made training hard. The source text we worked from was truncated, so further detail sits with the original.
Source: Google DeepMind