Remote sensing and neural networks combine to measure tree height | Research & Technology

Scientists at ETH Zurich’s EcoVision Laboratory have developed a deep learning framework to map treetop height globally at high resolution, using publicly available optical satellite imagery as input. The World Canopy Height Map is the first map of its kind. It could become an essential tool for tracking carbon emissions that contribute to climate change and for planning sustainable regional development.

To acquire the data needed for the world’s first canopy height map, EcoVision Lab researchers relied on two sources: NASA’s Global Ecosystem Dynamics Investigation (GEDI) and the Copernicus Sentinel-2 satellites operated by the European Space Agency. GEDI, which has the highest resolution and densest sampling of any lidar ever launched into orbit, performs laser ranging observations of nearly all of Earth’s tropical and temperate forests and provides high-resolution measurements of the 3D structure of the Earth. It provides sparse but well-distributed data on canopy height worldwide. Optical satellite images from the Sentinel-2 satellites capture every location on Earth every five days with a resolution of 10×10 meters/pixel. Sentinel-2 satellites provide dense observations on a global scale, although they cannot measure vertical structures.

By combining GEDI data with Sentinel-2 data, researchers created a probabilistic deep learning model to retrieve canopy height from Sentinel-2 images anywhere on Earth, and to quantify the uncertainty of these estimates with GEDI data.

The researchers prepared a set of convolutional neural networks to map the height of trees on a global scale. Neural networks showed millions of sample images from two Sentinel-2 satellites.

“Since we don’t know what patterns the computer should look for to estimate height, we let it learn the best image filters on its own,” said researcher Nico Lang, who developed the networks.

The neural network algorithm accesses the appropriate tree height response from the spatial laser measurements taken by GEDI. “The GEDI mission provides globally dispersed data on the height of vegetation between latitudes 51° north and south, so the computer sees many different types of vegetation in the process of formation,” said Lang said.

Researchers at ETH Zurich have developed a world map that uses machine learning to infer vegetation heights from high-resolution satellite images. Courtesy of EcoVision Lab.

With the input of optical satellite images and lidar measurements, the algorithm is able to itself acquire the filters for the textural and spectral models. By dragging a 3×3 pixel filter mask over the satellite image, the algorithm obtains information about the brightness patterns in the image. “The trick here is that we stack the image filters,” said Professor Konrad Schindler. “This gives the algorithm contextual information, since each pixel, from the previous convolution layer, already includes information about its neighbors.”

Five neural networks were trained independently of each other, each returning its own estimate of tree height. “If all the models agree, then the answer is clear based on the training data, Lang said. “If the models arrive at different answers, that means there is greater uncertainty in the estimate.”

The models also incorporate uncertainties that may arise from the input data; for example, when a satellite image is blurry, the uncertainty is greater than when the atmospheric conditions are clear.

Once trained, the neural networks only require image data, which means the map can be updated every year with satellite images from Sentinal-2. At the same time, the more data the GEDI mission collects, the denser the reference data for the global canopy height map will be.

Once a neural network has been trained, it can automatically estimate vegetation height from the more than 250,000 images – around 160TB of data – needed for the world map. According to the researchers, calculating the global vegetation height map would take a single powerful computer three years. “Fortunately, we have access to ETH Zurich’s high-performance computing cluster, so we didn’t have to wait three years for the map to be computed,” Lang said.

The global canopy height map can provide insight into carbon emissions, as tree height is a key indicator of biomass and the amount of carbon stored. “About 95% of forest biomass is made up of wood, not leaves. Thus, biomass is strongly correlated with height,” Schindler said. According to the map, only 5% of the world’s landmass is covered in trees taller than 30m, and only 34% of those tall canopies are in protected areas.

The ETH Zurich model will enable consistent and informed global mapping of uncertainties and support continuous monitoring to detect changes and inform decision-making. It can serve ongoing forest conservation efforts and it has the potential to drive advances in climate, carbon and biodiversity modeling.

The world map and underlying source code and models are made available to the public to support conservation efforts. For more information, visit N. Lang et al., “A High-Resolution Canopy Height Model of Earth” (

Comments are closed.