Companies sourcing commodities like timber, coffee, or palm oil increasingly face regulations requiring them to prove their products did not come from recently deforested land. A key challenge is that satellite data often shows tree cover, but not whether that cover is a natural forest or a plantation. Google Research's new AI model aims to close that gap by classifying tree cover with greater nuance.
The model is trained to separate natural forests from other tree-covered areas, such as commercial plantations or mixed landscapes. This distinction matters because many deforestation-free frameworks define deforestation as the loss of natural forest, not the harvest of planted trees. Without this separation, companies can be falsely flagged or, worse, inadvertently source from recently cleared natural forest.
Google Research positions the tool as a practical step for supply chain monitoring. By making the classification more accurate and scalable, it gives businesses and regulators a clearer basis for verifying claims. The approach does not replace on-the-ground verification but provides a stronger first-pass screening layer from orbit.
The work highlights how AI can turn raw satellite imagery into decision-ready information. As deforestation regulations tighten, tools that can reliably tell a primary forest from a palm oil plantation will become a core part of responsible sourcing infrastructure.