Beyond tree cover: mapping the world’s forest types

Google DeepMind and Google Research, working with the World Resources Institute and the International Institute for Applied Systems Analysis, have released Forest Typology (ForTy) 2020, a global map that assigns every 10-by-10-meter patch of land to one of six categories. Based on UN Food and Agriculture Organization definitions, these are primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops and agroforestry, and other land. An early time series covering 2017 to 2025 accompanies it, showing year by year what replaced lost forest.

DeepMind argues that tree-cover maps fall short because a logged forest replanted as a plantation soon looks green again from space, despite holding far less carbon and biodiversity. The distinction has regulatory weight: the EU's deforestation regulation bars products such as coffee, cocoa and palm oil grown on land that was forest in 2020 and later converted to agriculture, and it imposes forest degradation limits on wood products.

The model was trained on about 1.7 million labeled locations and draws on AlphaEarth Foundations embeddings, which condense years of satellite observations for each pixel. In validation against more than 9,000 randomly selected sites assessed by two trained experts each, the company reports ForTy identified the correct forest type 90% of the time and separated forest from non-forest 95% of the time, beating other global forest maps. It outputs probability scores for each class, so users can set stricter or looser thresholds.

The map, described in Scientific Data, and the time series are freely available on Google Earth Engine under an open license, and the team plans further refinements to the time series in the coming months.