NASA uses deep-learning satellite images and deep-learning to capture the carbon sequestered by 9.9 billion trees
NASA-led researchers used artificial intelligence and satellite imagery to map millions of tree crowns at a scale of 50 cm. Images covered a vast area of northern Africa from the Atlantic Ocean to the Red Sea. The researchers used allometric equations, based on tree samples, to convert the images into estimates of wood, foliage and root size.

The new NASA estimate, published in Nature, was shockingly low. The NASA technique uses only the trees in the area. Unlike the usual estimation, which relies on small areas, and then extrapolating upwards. Jules Bayala, Meine van Noordwijk and the NASA team published a News & Views piece in the same journal.

In other areas, the initial assessment was overestimated by a large amount. In earlier satellite attempts, cropland and ground vegetation negatively affected optical images. Radar backscatter was affected by topography, wetlands and irrigated land. This led to a higher estimate of carbon than NASA’s current estimates.

Source:
https://phys.org/news/2023-03-nasa-captures-sequestered-carbon-billion.html