Beyond tree cover: mapping the world’s forest types
deepmind ·
Google DeepMind, in collaboration with WRI and IIASA, has released Forest Typology (ForTy) 2020, a global map that categorizes every 10-by-10-meter patch of land into six distinct forest types. This initiative addresses the limitations of traditional tree cover maps by distinguishing between natural forests, plantations, and agricultural lands, which is crucial for implementing regulations like the EU Deforestation Regulation. The model utilizes AlphaEarth Foundations embeddings to identify subtle structural and seasonal patterns that differentiate primary forests from timber plantations or tree crops. Validation against over 9,000 expert-reviewed sites demonstrated a 90% accuracy rate in classifying forest types and 95% in distinguishing forest from non-forest areas. Alongside the static 2020 baseline, an early time series tracks changes from 2017 to 2025, revealing hidden conversions such as natural forests being replaced by plantations. By providing probability scores rather than binary labels, the tool allows users to adapt classifications to specific regional contexts and regulatory needs. The data is now freely available on Google Earth Engine to support governments, companies, and scientists in building deforestation-free supply chains. This release represents a significant advancement in using AI for nature conservation and environmental monitoring.
deepmind