An unexpectedly large count of trees in the West African Sahara and Sahel
- Visualizations by:
- Greg Shirah
- View full credits
Visualization showing study region, climate zones, close up of high res satellite data with machine learning-based tree crown regions, counting of trees, and overall tree counts and area
This video is also available on our YouTube channel.
A deep learning algorithm was trained by the scientists to identify trees using very high resolution satellite imagery (0.5m per pixel) from DigitalGlobe. The algorithm identified trees by looking for appropriate colors and shadows cast. A total of 11,128 multispectral images were used to identify trees in this study region. Using this technique 1,837,565,501 trees were identified in the study region with a median tree crown area of 12 square meters.
The visualization starts at a global scale then pushes in to show the study area. To illustrate that this is a dry area, climate zones are shown using annual rainfall averages from 1982-2017 including regions that are:
- hyper-arid (0-150 mm rainfall/year)
- arid (150-300 mm/year)
- semi-arid (300-600 mm/year)
- sub-humid (600-1000 mm/year)
The visualization next shows an area of high resolution imagery of the trees, then overlays the results of the machine learning which are filled regions of tree crowns for each tree in view. The trees are then counted up. The areas of trees are also totaled using the tree crown regions. We then zoom back out to see the entire study area and the total tree count and area.
This is a success story in using deep learning and big data to perform large scale scientific analysis. Now, scientists know the number and size of these trees in the Sahara and the Sahel. They will be able to use this information to calculate their impact to our planet.
Credits
Please give credit for this item to:
NASA's Scientific Visualization Studio
Visualizers
- Greg Shirah (NASA/GSFC) [Lead]
- Helen-Nicole Kostis (USRA)
- Leann Johnson (GST)
- Lori Perkins (NASA/GSFC)
Scientists
- Compton Tucker (NASA/GSFC)
- Jérôme Chave (French National Center for Scientific Research)
- Martin Brandt (University of Copenhagen)
Technical support
- Ian Jones (ADNET)
- Laurence Schuler (ADNET)
Data provider
- Erin Glennie (SSAI)
Papers
This visualization is based on the following papers:Datasets used in this visualization
OpenStreetMap
City lon/lat names and locations
See more visualizations using this data setTree Locations (generated via ML from DigitalGlobe mosaics)
Tree Densities (generated via ML from DigitalGlobe mosaics)
Landsat-8 Band Combination 4-3-2 (Collected with the OLI/TIRS sensor)
Terra and Aqua BMNG (A.K.A. Blue Marble: Next Generation) (Collected with the MODIS sensor)
Credit: The Blue Marble data is courtesy of Reto Stockli (NASA/GSFC).
Dataset can be found at: http://earthobservatory.nasa.gov/Newsroom/BlueMarble/
See more visualizations using this data setRasterized Tree Crowns (generated via ML from DigitalGlobe mosaics)
WorldView-2 © 2010 DigitalGlobe
Note: While we identify the data sets used in these visualizations, we do not store any further details nor the data sets themselves on our site.