Researchers from ETH and the University of Zurich created a world map of treetop heights using satellite imagery and machine learning. It has been shown, for example, that trees over 30 meters tall grow on only 5% of the global land area. The knowledge gained from the world map is valuable for climate protection and nature conservation, as the height of trees is an important indicator of biomass and the amount of carbon stored, as announced by ETH Zurich.
The map is based on 2020 satellite imagery and has a resolution of 10 by 10 meters. To create the world map, the researchers trained their model with laser measurements from the US space agency NASA’s Gedi mission, which provides elevation data for points between 51 degrees north and south latitude, and linked that information to optical images. of two Copernicus Sentinel -2 European Space Agency (ESA) satellites. Through training, the neural network learned to automatically estimate tree heights from hundreds of thousands of satellite images from around the world.
New calculation every five days
From the researchers’ point of view, the new approach offers interesting starting points for research, governments, administrations and NGOs. “Thanks to Sentinel-2, we could recalculate vegetation heights every five days and we would have a monitoring system to observe deforestation,” Nico Lang cited as an example. In his doctoral thesis, he developed the approach that has now been presented in a study that has not yet been peer-reviewed.
Overall, the new method meets ongoing efforts to protect forests and has the potential to advance advances in modeling climate, the global carbon cycle, and biodiversity, the researchers write. (apa)