NASA maps sequestered carbon of 10 billion trees, thanks to AI

"Our team gathered and analyzed carbon data down to the individual tree level across the vast semi-arid regions of Africa or elsewhere – something that had previously been done only on small, local scales," said Compton Tucker, lead scientist on the project and an Earth scientist at NASA's Goddard Space Flight Center in Greenbelt, Maryland. Past satellite-based estimates of tree carbon in Africa's drylands often mistook shrubs and grasses for trees. "That led to over-predictions of the carbon there."    

Carbon constantly fluctuates between the land, the ocean, the atmosphere, and the back. During the process of photosynthesis, trees tend to remove a greenhouse gas, Carbon Dioxide, from the atmosphere of the Earth and store it in their leaves, branches, trunks, and roots; for this particular reason why increasing tree cover is often suggested as a way to deal with continuously growing carbon emissions.   

Power of Machine Learning and AI 

As per the researchers, anything with a green, leafy crown and an adjacent shadow could be defined as a tree. From this information, the machine learning software could count the trees during supercomputing on the Blue Waters supercomputer at the University of Illinois. When the machine learning results were compared to human assessments, the accuracy rate of the computers was 96.5 percent in measuring the area of tree-crown.   

During the study, the team utilized sophisticated machine learning and AI algorithm to examine around 326,000 commercial satellite images from the GeoEye-1, WorldView-2, QuickBird-2, and WorldView-3 satellites (operated by Maxar Technologies). The researchers acquired the images through NASA's Center for Climate Simulation. They used its ADAPT/Explore Science Cloud to prepare and organize the photos for machine learning.   

Study abstract:

The distribution of dryland trees and their density, cover, size, mass, and carbon content is not well known at sub-continental to continental scales. This information is important for ecological protection, carbon accounting, climate mitigation, and restoration efforts of dryland ecosystems. We assessed more than 9.9 billion trees derived from more than 300,000 satellite images, covering semi-arid sub-Saharan Africa north of the Equator. We attributed to wood, foliage, and root carbon to every tree in the 0–1,000 mm year−1 rainfall zone by coupling field data, machine learning, satellite data, and high-performance computing. Average carbon stocks of individual trees ranged from 0.54 Mg C ha−1 and 63 kg C tree−1 in the arid zone to 3.7 Mg C ha−1 and 98 kg tree−1 in the sub-humid zone. Overall, we estimated the total carbon for our study area to be 0.84 (±19.8%) Pg C. Comparisons with 14 previous TRENDY numerical simulation studies for our area found that the density and carbon stocks of scattered trees have been underestimated by three models and overestimated by 11 models, respectively. This benchmarking can help understand the carbon cycle and address concerns about land degradation. We make available a linked database of wood mass, foliage mass, root mass, and carbon stock of each tree for scientists, policymakers, dryland-restoration practitioners, and farmers, who can use it to estimate farmland tree carbon stocks from tablets or laptops.