ForTy: Google maps the world's forest types at 10 m

What changed

Google DeepMind and Google Research, with the World Resources Institute (WRI) and IIASA, released the Forest Typology 2020 (ForTy) map (Google DeepMind blog). It assigns every 10x10 meter patch of land on Earth to one of six categories based on FAO definitions.

An early time series comes with it, tracking change every year from 2017 to 2025. It shows not only where forest was lost but what replaced it: a timber plantation, a tree-crop farm or other land.

The six categories:

• primary forest, with no visible signs of human disturbance,
• naturally regenerating forest, for example after logging or fire,
• planted forest, managed over long cycles,
• plantation forest: fast-growing, intensively managed, often single-species tree farms,
• tree crops and agroforestry, such as oil palm, rubber, cocoa or fruit,
• other land, from grassland and cropland to cities and water.

Under the hood

The model uses a custom architecture designed for remote sensing data. Its key input is AlphaEarth Foundations embeddings, compact per-pixel summaries of years of satellite observations. From these it picks up signals like planting patterns, growth cycles and harvest rhythms.

Training data covers about 1.7 million labeled locations, drawn from expert surveys, national and regional datasets, and targeted reviews of places where the model struggled.

Instead of a yes/no label per pixel, ForTy gives a probability for each class. Users set their own threshold: a company checking EUDR compliance can go strict, while a researcher estimating forest extent can use the full probability picture.

How accurate is it?

With WRI and IIASA it was validated at more than 9,000 randomly chosen sites, each reviewed by two trained experts on high-resolution imagery and years of satellite records. ForTy picked the right forest type 90% of the time and separated forest from non-forest 95% of the time. Google says this outperforms other global forest maps. The 2020 map is described in Scientific Data.

Why it matters

The EU deforestation regulation (EUDR) requires that products like coffee, cocoa or palm oil don't come from land that was forest in 2020. A plain tree-cover map isn't enough for that: a natural forest that was logged and replanted looks green again within a few years. ForTy makes that difference visible.

For developers, the interesting part is that the model doesn't work from raw satellite imagery but builds on ready-made AlphaEarth Foundations embeddings. A task-specific model sits on top of the output of a general geospatial foundation model. ForTy is the latest addition to Google Earth AI, Google's collection of geospatial models and datasets.

Try it

ForTy 2020 and the time series are freely available on Google Earth Engine under an open license, and the map can be explored at 10 meter detail in Google's Earth Engine app. The time series is an early release that Google plans to improve with partners over the coming months, so check it against your own region before making compliance decisions.