Google Deepmind creates a global map of forests with satellite data and AI
decoder ·
Google DeepMind and the World Resources Institute have developed ForTy, an AI-powered global map that classifies land cover into six forest-related categories at a 10-meter resolution. This system utilizes AlphaEarth Foundations to distinguish between primary forests, plantations, and tree crops by analyzing growth cycles and harvest patterns rather than relying solely on visual appearance. The project aims to provide a reliable 2020 reference point for the EU Regulation on Deforestation-free Products, helping companies assess supply chain risks regarding deforestation. While the map achieved an overall accuracy of 90.2%, it struggled with specific classes, correctly identifying only about half of tree crops and planted forests. Despite these limitations, ForTy outperformed existing European Commission maps in comparative tests on the same dataset. Researchers noted significant regional errors, such as overestimating primary forest in Asia and missing large areas in Africa due to open woodlands. The dataset and code are freely available under a CC BY 4.0 license to support broader environmental monitoring initiatives.
decoder