Satellite AI Maps Cashew Orchards Across Guinea-Bissau to Combat Deforestation

Researchers have developed a novel, scalable, and cost-effective method utilizing satellite imagery and machine learning to map cashew orchards throughout Guinea-Bissau. This initiative addresses a critical gap, as the country currently lacks a comprehensive nationwide database for these economically vital but environmentally impactful agricultural areas. The uncontrolled expansion of cashew production is a significant driver of deforestation and biodiversity loss in the region, underscoring the urgent need for better monitoring.

The methodology employs Sentinel-2 satellite data combined with margin-based Active Learning techniques. This approach allowed for the creation of an optimized training dataset, enabling the automatic detection of cashew orchards with a high balanced accuracy of 94.0%. Crucially, the entire process was conducted remotely, demonstrating the efficiency and practicality of the solution for large-scale application.

The project has generated two open-access datasets and a high-resolution 2021 cashew map (10m spatial resolution), available on GitHub. These resources provide a foundational tool for environmental management and policy-making, offering a new pathway for monitoring and potentially regulating cashew cultivation to mitigate its negative ecological effects across Guinea-Bissau and potentially other West African nations facing similar challenges.