Mapping of the forest extent in the Congo Basin at 10m-resolution with Sentinel-2

The tropical forests of Central Africa, storing approximately 10% of the world’s terrestrial carbon play a vital role in the global carbon cycle. These forests also constitute a hotspot of biodiversity and a source of livelihood and ecosystem services for local communities. The need to monitor forest loss and gain in the region has led to the development of a variety of forest maps through the use of orbital sensors in recent years.

In this study, we map the extent of the tropical moist forest of Central Africa at 10m resolution using an advanced Sentienel-2 processing technique. Initial calibration is performed with existing dataset (GFW, TMF) before refining with VHR data. Sentinel-2 data are processed into spatially coherent cloud-free annual composites. Classification into a forest/non forest map is then performed with XGBoost and the addition of ancillary variables, yielding yearly maps of the forest’s extent. This allows forest extent to be monitored with unprecedented resolution, which is crucial for the Congo Basin complex landscape, dominated by small-scale agriculture.