Satellite Tree Enumeration Outside of Forests at the 50-Centimeter Scale

Overview

Only 50% of the Earth’s total carbon dioxide (CO₂) emissions from fossil fuels and deforestation accumulate in our atmosphere; of the other half, 20% is absorbed into the oceans, and 30% goes somewhere on land. We haven’t a clue where on land this CO₂ from the atmosphere is extracted by vegetation via photosynthesis and stored in plant tissue as carbon compounds or as soil carbon. Understanding carbon storage on land is critical for fully understanding the impacts of CO₂ on global warming; we need to ascertain where this carbon is stored and what are the mechanisms that control this biotic carbon storage.

"The team distinguished trees from non-trees, mapped 1.8 billion trees over 1.3 million square kilometers, determined the area of leaves within the tree crowns, and now know every tree’s location to ±5 meters. We are presently determining tree height, and when we combine heights with tree crown areas, that will be an accurate predictor of carbon in the wood of trees over vast semi-arid regions."

Project Details

Our current understanding of the land carbon sink—where carbon is stored—is based on numerical simulation models that run at grid cells sizes of 50 x 50 kilometers (km) or larger. These simulations indicate that a large proportion of the land carbon sink occurs in Earth’s semi-arid regions.

Our team has undertaken research to quantitatively determine the carbon sequestered in trees within a large semi-arid area stretching from the Sahara Desert southward to the humid sub-tropics in West Africa. The first step is to map individual trees, which are important because of carbon residence time in wood. To date, no one has successfully mapped individual trees with satellite data in semi-arid areas.

In our most recent study, we used 50,000 individual commercial satellite images with a 50-centimeter (cm) x-y spatial resolution to map tree crowns across West Africa. The study area includes the arid southern portion of the Sahara Desert, stretching through the semi-arid Sahel Zone and into the humid sub-tropics. In the study, we leveraged the high-performance computing (HPC) resources of both the NASA Center for Climate Simulation (NCCS) at NASA’s Goddard Space Flight Center and the Blue Waters supercomputer located at the National Center for Supercomputing Applications (NCSA) and used artificial intelligence and machine learning (AI/ML) to map the trees and their properties.

Results and Impact

The team distinguished trees from non-trees, mapped 1.8 billion trees over 1.3 million square km, determined the area of leaves within the tree crowns, and now know every tree’s location to ±5 meters. We are presently determining tree height, and when we combine heights with tree crown areas, that will be an accurate predictor of carbon in the wood of trees over vast semi-arid regions.

Why HPC Matters

This ongoing 5-year research project has benefited greatly from substantial HPC resources and staff assistance from both NASA’s NCCS and the former National Science Foundation’s Blue Waters HPC facility, now operated by NCSA at the University of Illinois. Continued access to NCCS and Blue Waters resources will be instrumental to applying our newest techniques to the full study area—a coast-to-coast swath of Sub-Saharan Africa covering 10 million square km (an area larger than the continental United States).

With thousands of 50-cm-resolution images to assemble and analyze, we use a combination of the NCCS Advanced Data Analytics Platform (ADAPT) Science Cloud for image processing and the NCCS Discover supercomputer for ML-based analysis to organize our satellite data for processing. The tree identification AI/ML work gets executed on Blue Waters to identify tree canopies, estimate tree height, and ultimately calculate the amount of carbon stored in those trees. Without support from NCCS and Blue Waters, we would never be able to complete this work.

Related Links

• An Unexpectedly Large Count of Trees in the West African Sahara and Sahel,
• NASA Supercomputing Study Breaks Ground for Tree Mapping, Carbon Research
• NCCS ADAPT Science Cloud
• NCCS Discover Supercomputer
• Blue Waters