Beyond tree cover: mapping the world’s forest types

Google DeepMind, Google Research, WRI and IIASA have released ForTy 2020, a global forest map that classifies every 10-by-10-meter patch of land into six types. The pitch is simple, but the point is big: a green pixel is not always a forest, and not every forest stores the same carbon or supports the same wildlife.

The map splits land into primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops and agroforestry, and everything else. That matters for rules like the EU’s deforestation regulation, which cares not just about where forest once stood, but what it became afterward. Coffee, cocoa and palm oil are one issue; wood products are another, because degradation counts too.

ForTy goes beyond a single snapshot. Alongside the 2020 baseline, the team is also publishing an early time series from 2017 through 2025 that shows what changed each year, including cases where natural forest was turned into plantation or farm land. That is the bit standard tree-cover maps miss, especially when a logged area quickly turns green again.

Training the model took 1.7 million labeled locations gathered from expert surveys, regional and national datasets, and targeted checks on hard cases. The system also uses AlphaEarth Foundations embeddings, which package years of satellite observations for each 10-meter pixel so it can catch clues like planting patterns, harvest rhythms and growth cycles.

To see if it held up, the researchers tested ForTy against more than 9,000 randomly chosen sites worldwide. Two trained experts reviewed each site using high-resolution imagery and satellite history. The result: the model got forest type right 90% of the time and forest versus non-forest right 95% of the time, which Google says beats other global forest maps. The data are now free on Google Earth Engine under an open license.