Detection of degraded forests in Guinea, West Africa, using convolutional neural networks and Sentinel-2 time series

A convolutional neural network (CNN) model, specifically U-Net, demonstrates superior accuracy and robustness in detecting forest degradation across diverse Guinean forest regions using Sentinel-2 satellite imagery. With a 94% agreement with photo-interpreted maps in the Ziama massif and over 91% accuracy in Mount Nimba Strict Nature Reserve, the U-Net model outperformed other algorithms like Random Forest (RF), SegNet, and ResNet-UNet. This study highlights the potential of CNN-based methods to replace traditional photo-interpretation techniques for land cover mapping, offering a more efficient and reliable approach for monitoring forest degradation in West Africa.

Vo Quang, A., Delbart, N., Jaffrain, G., & Pinet, C. (2025). Detection of degraded forests in Guinea, West Africa, using convolutional neural networks and Sentinel-2 time series. Frontiers in Remote Sensing, 6. https://doi.org/10.3389/frsen.2025.1538808