Overlooked Optical Signal: The Role of Leaf Angles

Plants are not just visually stunning; their appearance can reveal a wealth of information about their health and functioning. Even subtle changes in the shades of green in a plant’s leaves can tell us about its biochemical properties, such as the amount of pigment it contains. But the story doesn't end with the colors we can see—wavelengths of light beyond our visual spectrum, like infrared, also offer a window into deeper plant states, such as water or protein content. This is possible because different biochemical compounds absorb light at specific wavelengths, and the portion that isn't absorbed gets reflected. This reflected light forms the basis of monitoring plant properties from Earth observation satellites. Since NASA's Landsat mission in the 1970s, space agencies have been providing remote sensing data on the dynamics of the terrestrial biosphere, enabling scientists to track changes in water content, pigments, and other plant traits over time. 

Vegetation indices derived from the Sentinel-2 satellite mission depict spatio-temporal dynamics in plant properties. The example here shows the normalized difference vegetation index, which is often used to approximate plant productivity. This data was obtained with the package Sentle (Clemens Mosig) and visualized using Lexcube (Maximilian Söchting).                                                                        

One of the most common methods for converting the reflected light recorded by satellites into meaningful plant information is through vegetation indices—simple empirical equations based on combinations between the values of different wavelength regions that translate reflectance data into proxies for biophysical variables. These indices help to approximate various plant traits, such as chlorophyll levels, water content, and overall plant health. However, there’s more to the story. While the biochemical composition of plants changes over time, so does the plants’ three dimensional structure. 

You may have noticed in everyday life that objects appear brighter or darker depending on their orientation relative to a light source. One example is the parallel stripes on football fields, created by the mower's effect on the grass, bending the blades at different angles. Similarly, leaf inclination angles affect how much sunlight is reflected toward satellite sensors. In other words, the changing angles of leaves can alter the reflectance measured from satellites. So, what does this mean for monitoring vegetation from space? Plants aren’t static; they move, and their leaves, in particular, can shift in response to environmental conditions.

An illustrative example of how changes in vertical leaf angles can alter the reflectance of plants. Our analysis reveals that this effect has an imprint on several vegetation indices as measured from satellites.                                                                         

Leaf Movements: From Darwin to Modern Science

The variation of leaf angles has already been studied for quite a while. Already Charles Darwin was intrigued by leaf movements, as described in his 1880 book “The Power of Movement in Plants”. Leaves adjust their angles predictably to optimize sunlight capture, not only across the seasons but even throughout the day. A less predictable factor driving leaf movement is their hydration status. Leaves lose turgor pressure when they are dehydrated due to dry soil or atmosphere, causing them to wilt or droop. This drooping changes the angles of the leaves and, as a result, alters the reflectance signal captured by satellites.

Although this seems intuitive, the effect of leaf angle movement on satellite data could so far not be investigated in depth. Why? Radiative transfer models can predict how leaf angle changes impact satellite measurements, but they require accurate data on leaf angles over time—and measuring leaf angles across large areas has been a significant challenge. Measuring leaf angles manually using a protractor is simply impractical on a large scale, and laser scanning, though promising, can be expensive and weather-sensitive. To overcome these challenges, we conducted a study employing a novel approach to measure the temporal variation of leaf angles, combining time-lapse cameras with deep learning-based computer vision.

Key Findings: Leaf Angles Confound many Vegetation Indices

By recording videos of plants over several weeks, we used the deep learning method AngleCam (Kattenborn et al. 2023) to predict leaf angles and track their changes over time. These leaf angle time series were then fed into radiative transfer models. These models can simulate reflectance data as observed from a satellite. With these simulated reflectance spectra we derived a series of vegetation indices that are used to monitor plant properties and assess if they are influenced by leaf movements.

The results of this procedure reveal that a wide range of 124 frequently used vegetation indices are significantly affected by leaf angle variations. Depending on the vegetation index, we observed that up to 60% of its variability can be attributed to leaf angle dynamics. This includes indices typically used to estimate chlorophyll levels, water content, and overall plant health. Furthermore, the research showed that leaf angle changes are closely tied to soil moisture and environmental factors, but these relationships differ across species and even between individual plants.

From plant videos to satellite data: The workflow included recording plant videos of 10 tree species over a vegetation period. Leaf angle time series were then extracted using AngleCam. These leaf angle time series were then used to model the effect on satellite-based vegetation indices, such as the kNDVI frequently used to track vegetation productivity or leaf area index.

Implications for Satellite-Based Ecosystem Monitoring

So, what does it mean if leaf angles confound the observed signals from space? If leaf angles cause changes in reflectance signals, we may mistakenly attribute these changes to variations in plant properties when, in fact, the properties remain constant. Conversely, leaf angle-induced changes might mask genuine variations in plant health or behavior. Therefore, vegetation indices need to be interpreted with caution, especially when variation in environmental conditions are involved. However, not all vegetation indices are equally strongly affected as some prove relatively robust. Our comparison of 124 different vegetation indices may provide guidance in selecting such robust indices. We also believe that we can maybe learn from the discrepancies across indices about ecosystem states but there is still much research needed to provide best practices guidance. In other words: The variation in leaf angles may not just be a challenge—it could also be an opportunity. Since leaf angles are linked to environmental factors like drought stress, the imprint of leaf angles on reflectance may provide a useful feature for tracking plant responses to environmental changes or climate extremes. As Albert Einstein once said, “In the middle of difficulty lies opportunity.”

References

• Kattenborn, T., Richter, R., Guimarães‐Steinicke, C., Feilhauer, H., & Wirth, C. (2022). AngleCam: Predicting the temporal variation of leaf angle distributions from image series with deep learning. Methods in Ecology and Evolution, 13(11), 2531-2545. https://doi.org/10.1111/2041-210X.13968
• Kattenborn, T. (2023). AngleCam (Version 2023-06-05). https://github.com/tejakattenborn/AngleCAM
• Mosig, C. (2024). Sentle (Version 2024.8.3). https://github.com/cmosig/sentle
• Söchting, M., Mahecha, M. D., Montero, D., & Scheuermann, G. (2023). Lexcube: Interactive visualization of large earth system data cubes. IEEE Computer Graphics and Applications. https://doi.org/10.1109/MCG.2023.3321989
• Söchting, M. (2023). LexCube. https://github.com/msoechting/lexcube