Looking at a forest from above can look a lot like one very large green blanket. But, in this case, each stitch in the blanket is an individual tree contributing to the overall carbon stock of the forest. Countries understand that the number and size of these trees is paramount to future conservation and environmental efforts, but manually mapping the trees of an entire country would be a huge, costly task.
Now, by combining machine learning with aerial and satellite images, researchers at the University of Copenhagen have developed a method to map the carbon stock of individual tress. Working with authorities and researchers in Rwanda, the team has created the first national inventory of tree-level carbon stocks in Rwanda.
"Mapping individual trees and calculating their carbon stocks has traditionally been done in forestry, albeit at a much smaller scale. Basically, what we do equals scaling up these approaches from a very local to a national level," said study author Ankit Kariryaa, a researcher at the University of Copenhagen.
The relationship between the extent of the crown and the total carbon content of a tree varies drastically depending on the size of a tree. For example, one very large tree has a much higher carbon content than a group of trees with the same joint crown extent. If the group was mistaken for one tree, the carbon content would be significantly overestimated.
The researchers understood this critical need for any method to be able to distinguish between individual trees, causing them to turn to a deep neural network for analysis.
The Rwandan and Danish research team trained the machine learning algorithm on a set of 97,500 manually delineated tree crowns, representing the full range of biogeographical conditions across the country. They also used publicly available aerial and satellite images of Rwanda at 0.25 x 0.25 meter resolution. These images were collected from June to August 2008 and 2009.
By the end of the study, over 350 million trees were mapped, including those outside Rwandan forests. In fact, 72% of the mapped trees were in farmlands and savannas, 17% on plantations, and only 11% in natural forests.
Even so, according to the study results, that small proportion of forest trees accounted for 51% of the national carbon stock of Rwanda. The researchers say this is likely because natural forests have very little human disturbance thanks to legislation and policy. Thus, they have a very high carbon content per tree volume.
“This suggests that conservation, regeneration and sustainable management of natural forests is more effective at mitigating climate change than plantation,” said first author Maurice Mugabowindekwe, a Ph.D. researcher at the University of Copenhagen.
Mugabowindekwe is Rwandan, but he says the rationale of using Rwanda for the development of the new method was scientifically based.
“The country has a rich landscape variation including savannas, woodlands, sub-humid and humid forests, shrubland, agro-ecosystem mosaics, and urban tree ecosystems, which are representative of most tropical countries. We wanted to prove the method for all these landscape types,” the geology researcher said.
When the study results were presented to Rwandan authorities in July 2022, they immediately asked Mugabowindekwe and team to update the mapping based on newer aerial images acquired in 2019—a project the Danish team is currently undertaking.
Additionally, the method yielded good results when tested on a handful of other countries, including Tanzania, Burundi, Uganda and Kenya. The way different countries take their forest inventories is not consistent due to different contexts, goals and available datasets. The researchers said they hope this method will establish itself as a standard, thereby enabling better comparisons between countries.
“If you are not able to create an accurate and reliable inventory, there is a risk of lacking a framework to track the impact of landscape restoration,” said Mugabowindekwe. “This could make the conservation and sustainable management of both forests and other tree-dominated landscapes impossible. Therefore, this is science that is likely to have an impact."