Google DeepMind and WRI release ForTy, a global 10-meter forest typology map for 2020 built on AlphaEarth Foundations

Google DeepMind and the World Resources Institute released ForTy, a free global map of forest types for 2020 at 10 metre resolution. It sorts land into six classes: primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops and other land. It is built on AlphaEarth Foundations embeddings, and Gemini was used to adapt older training data to EU definitions. The map and test data are CC BY 4.0, including in Google Earth Engine, and the code is in the GeeFlow and JEO repositories. The authors report 90.2% overall accuracy on 8,190 expert-reviewed sites, and say it beats European Commission EUDR maps on the same test data (95.5% vs 93.6% for forest/non-forest, 90.9% vs 87.8% for forest types).

Companies and regulators dealing with the EU Deforestation Regulation get an openly licensed, forest-type reference map for 2020 that is a bit more accurate than the Commission's own maps on the authors' test. Because it separates primary forest, planted forest and plantations, it is more useful for supply-chain checks than a plain tree-cover map. The gains over the Commission maps are a few percentage points, and the weak classes matter most for compliance. It finds only about half of tree crops and planted forests and 60% of plantations, so it is a screening aid, not proof for individual shipments. It is also a single snapshot of 2020, not a monitoring tool.

The "first global map of its kind" claim comes from the authors and is unverified. The links to the Tropical Forest Forever Facility and the UN 30-by-2030 goal are speculation about possible uses, not announced adoption. The accuracy and benchmark figures are self-reported, and the coverage we scored is a news report, not the paper, so independent validation is missing. The 200 million hectare overestimate of Asian primary forest is a large error that headlines about 90% accuracy tend to bury.

1. 01Independent validation or critiques of ForTy's accuracy, especially for tree crops, plantations and Asian primary forest, from academic groups or NGOs within the next few months
2. 02Whether the European Commission, EUDR-compliance vendors or major commodity traders say they will use ForTy, or whether the Commission updates its own reference maps
3. 03Whether a post-2020 version or yearly update follows, and whether the weak classes (planted forest, tree crops, plantations) improve in it