A method to map individual trees in Rwanda using satellite imagery and deep learning is presented in a paper published in Nature Climate Change.

The findings suggest that trees outside of forests make up almost half of the national carbon stock in Rwanda and highlights that improved and more complete tree monitoring in Africa is critical to fulfilling climate commitments.

Trees are important to mitigate climate change. Article 5 of the Paris Agreement requires countries to reduce deforestation, enhance carbon sinks and conserve forests. Many African countries, including Rwanda, regularly evaluate, quantify and monitor forest trees using field inventories.

However, these methods are labour-intensive and fall short of mapping individual trees growing outside of forests, such as in the savanna, woodlands or fragmented agroecosystems. Estimates of carbon benefits are therefore often incomplete, resulting in inadequate natural resource management.

Maurice Mugabowindekwe, Martin Brandt and colleagues used aerial images and deep learning to map the size and carbon stock of individual overstory trees (trees that tower above the surrounding canopy) across Rwanda’s ecosystems. They demonstrate that almost 90% of mapped trees grow in farmlands, savannas and plantations, accounting for nearly half of national aboveground carbon stocks. Natural forests account for only 11% of the tree count, but 51% of carbon stocks, suggesting that conservation, regeneration and sustainable management of natural forests is more effective at mitigating climate change than tree plantations for timber production.