Global forest map ForTy: AI classifies land, but coverage has gaps

What happened

Google DeepMind and the World Resources Institute (WRI) have released ForTy, a global map that uses satellite imagery and AI to classify forests and other land into six categories, including primary forest, plantations and tree crops. The Decoder reports that the map could help organisations track deforestation against EU rules.

The map has clear limits. According to the reporting, it identifies only about half of planted forests and tree-crop areas, and about 60% of plantations. Those gaps matter most for the land types that compliance teams are most likely to need to verify.

Why it matters

ForTy is a practical example of AI turning raw satellite data into a classified, usable dataset at global scale. Work that once depended on patchy national inventories or manual surveys can now be drawn from a single model-driven map. For organisations with supply chains in agriculture, forestry or commodities, that changes what can be checked, how often, and by whom.

The more important lesson for leaders is about how such outputs are used. A map that is strong on one category and weaker on another can easily be read as complete. Teams building AI-powered monitoring tools will need to show clearly where a model's confidence is high and where it is not, so that decisions made on top of it stay defensible.

By the numbers

• Six land categories classified by the map, including primary forest, plantations and tree crops.
• About half of planted forests and tree-crop areas identified by the map, according to the reporting.
• About 60% of plantations identified by the map, according to the reporting.

The Renascence take

Most readers will focus on the headline achievement: a global forest map built with AI. The more useful signal is the coverage gap. When a model's accuracy varies sharply by category, the people relying on its output are the ones who need the most design care, because they are rarely the ones who built it.

The headline is about what AI can now map; the gap is about what users will assume it has mapped. Organisations that treat a dataset like ForTy as a finished answer will take on risk they cannot see. The behavioural principle at work is that people trust a single clean output and rarely ask what sits beneath it. A customer-obsessed operator should publish the coverage limits alongside the result and design workflows that route the weaker categories to human review.