Google’s Forest Map Separates Natural Forests From Plantations—At 10-Meter Detail

From above, a mature forest and a commercial tree plantation can both look like green canopy. Their ecological roles, histories and legal relevance may differ sharply. Google’s new ForTy dataset tries to make that difference visible at a global scale, turning satellite and geospatial inputs into forest-type classes. The resolution is striking; the map’s limits are just as important for anyone using it to guide conservation or supply-chain decisions.

Why a tree-cover map can miss the point

Google DeepMind’s October 22 announcement describes Forest Typology (ForTy) 2020 as a global classification at 10-by-10-meter grid spacing. It distinguishes six categories: primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops and agroforestry, and other land. The project partners include Google Research, the World Resources Institute and the International Institute for Applied Systems Analysis.

That distinction matters because canopy presence alone cannot tell whether an area is an intact natural ecosystem, a planted timber stand or an agricultural tree crop. The map also comes with an early annual series from 2017 through 2025, intended to show what may have replaced forest over time. The open dataset repository provides code and data access details, including the benchmark data used to train and evaluate classification models.

Ten-meter spacing is not ten-meter certainty

A 10-meter grid tells readers the spatial sampling scale, not that every cell is correct. The map is a model output assembled from remote-sensing inputs and labels; class boundaries can be uncertain where landscapes mix, cloud cover limits observations, or the reference data are sparse. The project’s ForTy benchmark paper addresses the challenge of distinguishing forest types and describes its global dataset. Even a well-performing model remains dependent on definitions, training labels, geography and the period being mapped.

That is why users should inspect metadata, confidence information and validation before treating a change as fact on the ground. A map is useful for directing attention: where should analysts investigate, compare imagery or request a field check? It is weaker as a standalone answer to who owns a parcel, what a product contains, or whether a particular shipment meets a legal requirement.

A baseline for monitoring, not a compliance certificate

The European Commission’s EUDR guidance sets due-diligence obligations for covered commodities and products. Geolocation and supply-chain information matter to those obligations. A public global map may help organizations prioritize risk reviews and understand land-cover context, but it cannot replace product-specific traceability records or determine legal compliance on its own.

Google frames ForTy as a potential baseline for the EU deforestation regulation, not as a substitute for the regulation’s due diligence. The two can complement each other: a standardized map can help locate possible land-use change, while operators still need to connect a product to its origin and verify the relevant facts. That is especially important when a pixel falls near the boundary between natural forest, plantation and agriculture.

The best near-term use is therefore analytical. Conservation groups can compare regional patterns; researchers can test methods against an openly described benchmark; and supply-chain teams can use map signals to prioritize checks. Each should document the map version, date and confidence, then verify consequential findings with more specific evidence. Our coverage of Earth AI geospatial agents discusses a related but distinct question: how tools can help analysts investigate geographic information without turning an answer into a substitute for source data.