A new global map distinguishes primary forests, regrowing forests, plantations and tree-crop farms at 10-meter resolution, giving regulators and conservation teams information that ordinary tree-cover maps miss.
Google DeepMind and Google Research developed Forest Typology 2020, or ForTy, with the World Resources Institute and the International Institute for Applied Systems Analysis. Its six categories also cover planted forests and non-forest land. An early time series applies a similar method annually from 2017 through 2025, showing whether a cleared natural forest became a timber plantation, farm or something else.
The model uses compact summaries of satellite observations from AlphaEarth Foundations and about 1.7 million labeled locations. In validation against more than 9,000 randomly selected sites reviewed by two experts, the developers report 90% accuracy for forest type and 95% accuracy when separating forest from non-forest land. Probability scores let users choose stricter thresholds where false classifications carry regulatory risk.
Both the 2020 baseline and early time series are freely available through Google Earth Engine under an open license. They are designed in part to support the EU deforestation regulation, but the time series is still an early release that the team plans to refine.