Beyond tree cover: mapping the world’s forest types
Google DeepMind, Google Research, WRI, and IIASA have released Forest Typology (ForTy) 2020, a global map categorizing land into six forest and land-use types at 10-meter resolution. It tracks changes from 2017-2025, showing what replaces lost forests.
Intelligence analysis by Gemini 2.5 Flash
Traditional forest maps only indicate tree presence, making it difficult to differentiate between natural forests, plantations, and farms, which have vastly different ecological values. The new ForTy map uses AI to classify forest types and track land-use changes over time, providing critical data for conservation and regulatory compliance.
Imagine trying to tell if a patch of green land is a wild jungle, a fruit farm, or a tree farm just by looking at its green leaves from far away. It's hard! But now, a super-smart computer program, like a super-detective, can look at satellite pictures and tell exactly what kind of plants are growing, even if they look similar. This helps grown-ups protect the real wild forests and make sure our food doesn't come from places where those forests were chopped down.
Analysis
ForTy 2020
The Forest Typology (ForTy) 2020 map represents a significant leap in global forest monitoring, moving beyond simple 'tree cover' detection to a nuanced classification of land use. Developed by Google DeepMind, Google Research, the World Resources Institute (WRI), and the International Institute for Applied Systems Analysis (IIASA), this map categorizes every 10-by-10-meter patch of land into six distinct types: primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops and agroforestry, and other land. This granular detail, coupled with an early time series tracking annual changes from 2017 to 2025, allows users to identify not just where forest was lost, but precisely what replaced it, such as timber plantations or tree-crop farms.
The ability to differentiate between these forest types is paramount because their ecological value, carbon storage capacity, and biodiversity levels vary dramatically. A primary forest, for instance, holds significantly more carbon and supports richer biodiversity than a single-species timber plantation. By providing this detailed typology, ForTy addresses a critical gap in previous mapping efforts, which often treated all tree cover as ecologically equivalent. This enhanced understanding is vital for effective conservation strategies and for accurately assessing the impact of human activities on forest ecosystems globally.
EUDR
A key driver and beneficiary of the ForTy map is the European Union's regulation on deforestation-free products (EUDR). This regulation mandates that products like coffee, cocoa, palm oil, and wood do not originate from land that was forest in 2020 and has since been converted to agriculture or from primary forests converted into plantations. The EUDR requires companies and regulators to possess precise knowledge of forest types and land-use changes, a need that conventional 'tree cover' maps could not adequately fulfill.
ForTy directly supports the implementation of the EUDR by providing the necessary detailed baseline and ongoing monitoring data. It allows companies to verify their supply chains and demonstrate compliance, ensuring their products are free from deforestation and forest degradation. This level of transparency is crucial for both regulatory enforcement and for commodity producers committed to avoiding natural forest conversion, fostering a more accountable global trade system that aligns with environmental protection goals.
AlphaEarth Foundations
The technical backbone of the ForTy model relies on a custom architecture designed to handle the complexities of remote sensing data, leveraging advanced AI capabilities. A critical input for the model is AlphaEarth Foundations embeddings, which are data-compact summaries derived from years of satellite observations for each 10-meter pixel. These embeddings enable the model to discern subtle yet crucial clues that differentiate forest types, such as specific planting patterns, distinct growth cycles, and characteristic harvest rhythms.
Training the ForTy model was a significant undertaking, involving a large-scale collection of approximately 1.7 million labeled forest-type examples. These examples were meticulously gathered from expert surveys, national and regional datasets, and targeted reviews of areas where initial model performance was challenging. This extensive training dataset, combined with the sophisticated AI architecture and AlphaEarth Foundations embeddings, allows ForTy to accurately categorize forest and land-use types, even when they appear visually similar from above, achieving a 90% accuracy rate in identifying the correct forest type during validation against over 9,000 randomly chosen sites worldwide.
Key points
- Google DeepMind and partners released Forest Typology (ForTy) 2020, a 10-meter resolution global map of forest and land-use types.
- ForTy categorizes land into six types, including primary forest, plantations, and tree crops, and tracks annual changes from 2017-2025.
- The map addresses the limitations of 'tree cover' data by differentiating forest types with varying ecological values and carbon storage.
- It provides crucial data for implementing the EU's deforestation-free products regulation (EUDR), helping companies and regulators ensure compliance.
- The AI model uses AlphaEarth Foundations embeddings and was trained on 1.7 million labeled locations, achieving 90% accuracy in validation.
The new Forest Typology map offers unprecedented detail, enabling more effective global conservation efforts and supporting stricter regulatory compliance. This enhanced transparency can empower companies and governments to make informed decisions, fostering sustainable land use and protecting vital ecosystems.
Despite its accuracy, the map's effectiveness relies on widespread adoption and consistent enforcement by regulators and companies, which may face logistical or political hurdles. Furthermore, the dynamic nature of land use means continuous updates are crucial, posing a challenge for maintaining real-time relevance and preventing new deforestation.
