Filling spatial and temporal gaps of satellite-based retrievals of vegetation indices by considering three-dimensional cloud and vegetation effects.

• Applicant: Dr. Kevin Wolf
• Subject Area: Atmospheric Science
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 582922396

Vegetation and forests play a vital role in the Earth’s climate system. However, deforestation and anthropogenic climate change pose a threat to forests worldwide. Therefore, satellite remote sensing (RS) of vegetation is an indispensable tool for monitoring health and estimating carbon fluxes. Vegetation is often monitored using vegetation indices (VIs), which are based on passive satellite RS measurements of spectral reflectance at two or more wavelengths. However, passive RS faces challenges due to clouds, cloud shadows, and cloud scattering. These are known to influence radiative transfer (RT) in the atmosphere and likely affect VIs estimated from cloud-free pixels near clouds. Regions with high cloud cover, such as the tropics, are particularly affected, as are mid-latitude regions with an average cloud cover of around 70%. To recover VIs in cloud-free pixels that are biased by nearby clouds, the three-dimensional (3D) effects of neighbouring cloudy pixels must be investigated. The hypothesis to be tested is that including and correcting pixels affected by neighbouring cloudy pixels will significantly increase the temporal and spatial coverage of satellite-based VI estimates, which then allows to track rapid changes in vegetation. The first work package (WP) will identify regions and vegetation types most susceptible to VI biases using Sentinel-2 satellite observations. In the second WP a coupled atmosphere-vegetation 3D radiative transfer model setup will be developed. The major advancement will be the combination of an atmosphere and a vegetation Monte Carlo radiative transfer model to account for the full 3D cloud-vegetation radiation interactions. The developed model combination will be used in the third WP to simulate various cloud and canopy scenarios, and allows to determine the effect of heterogeneous cloud fields on canopy reflectance and albedo. This will address questions about the spatial extent of VI biases, quantify their magnitude, and determine how they are influenced by cloud macrophysical properties, sun geometry, canopy properties, and satellite spatial resolution. In the fourth WP correction methods will be developed to minimise the identified VI biases. Two approaches will be pursued: a parameter-based correction and a correction using a convolutional neural network. In the final WP, the developed correction methods will be tested on satellite images obtained in the first WP, and the gain in temporal and spacial VI coverage will be quantified.

• DFG Programme: Research Grants
• International Connection: Japan
• Cooperation Partner: Professor Hironobu Iwabuchi, Ph.D.