The terrain effect correction : how it works

Caution, this post contains formulas.

 

=>

Terrain effects are one of the disrupting elements of surface reflectance retrieval. This can be easily checked with the image below.

Gizeh Pyramids as seen by Pleiades soon after its launch. We can assume that the 4 faces of the pyramids have similar surface reflectances, but their illumination varies because of the geometry of the shot. One face is completely in the shade. The correction of the terrain effects should allow to correct these apparent differences in reflectances (except for the face in the shade for which we lack information)

The topographic (or terrain) effects on the observed reflectances are due to several phenomena, illustrated below :

  • the closer the surface is perpendicular to sun direction, the more energy it receives per surface unit (we talk about irradiance). If the surface is parallel to sun direction, it does not receive direct sunlight. We can model it this way :
    • For an horizontal surface : \( E_h= E_0.T_{dir}^\downarrow.cos(\theta_s) \)
Definition of sun zenith angle and sun incidence angle
  • For a sloped surface \( E_i=E_0.T_{dir}^\downarrow.cos(\theta_i) \)
  • \( E_0 \) is the Top of Atmosphere irradiance, and \( T_{dir}^\downarrow \) is the downward direct transmission, i.e. the proportion of the light that reaches directly the surface without being scattered by the atmosphere.
  • Assuming that all the irradiance is direct, the measured reflectance if the surface was horizontal is calculated from the following formula:\( \rho_h=\rho_i \frac{cos(\theta_s)}{cos(\theta_i)}\). However, the above assumption is not true and this formula tends to over correct terrain effects
  • The surfaces also receive a diffuse sun irradiance scattered by the atmosphere. If the surface is not horizontal, a part of the sky is obscured by the slope reducing the diffuse irradiance. Moreover, the diffuse irradiance depends on the amount of aerosols (and clouds) in the atmosphere. In addition, the surrounding terrain can also hide a part of the sky, but we do not take this effect into account here in our modelling. We use the following approximation, which is equivalent to assuming that the slope is alone in a horizontal region.
    • If surface is horizontal, the visible sky fraction is 1, if it is vertical, this fraction is 1/2
    • \( \displaystyle F_{sky}= \frac{1+cos(slope)}{2} \)
  • Finally, the slope can receive light from surrounding surfaces, which become directly visible. In our simplified model, we always assume the entire environment of our slope is flat and,for instance, we do not take the effect of the opposite side in a valley into account :
    • If surface is horizontal, the visible ground fraction is 0, if it is vertical, it is 1/2.
    • \( \displaystyle F_{fround}= \frac{1-cos(slope)}{2} \)
  • Finally, we use the following formula  to compute the reflectance that would be observed if the surface was horizontal \( \rho_{h}\), as a function of the slope (inclined) reflectance \( \rho_{i}\) :

    \( \rho_{h}=\displaystyle \rho_{i}.\frac{T^{\downarrow}}{T_{dir}^{\downarrow}.\frac{cos(\theta_i)}{cos(\theta_s)} + T_{dif}^{\downarrow} F_{sky} + T^{\downarrow} F_{ground} \rho_{env}} \)

    où \( T^{\downarrow}\) is the downward transmission, sum of direct and diffuse irradiances : \( T^{\downarrow}= T_{dir}^{\downarrow}+ T_{dif}^{\downarrow}\), and [/latex]\rho_{env}[/latex] is the average reflectance of the neighbourhood.

    Finally, we can also account for bidirectional reflectance effects, but this correction is tricky since directional effects depend on the surface cover type. See for instance : Dymond, J.R.; Shepherd, J.D. 1999: Correction of the topographic effect in remote sensing. IEEE Trans. Geosci. Remote Sens. 37(5): 2618-2620.

    It is very difficult to validate the correction of directional effects : , we could compare the correction results for satellite overpasses at different times in the day. But all the satellite optical imagers have nearly the same overpass time. A qualitative way of estimating the accuracy is to check that similar land covers on opposite slopes in a valley ( a meadow, a forest) have a similar reflectance after correction. The most suitable points are North-South valleys. Here are some examples of terrain effects correction results.

    Formosat-2 image in the Alps, before (left) and after (right) terrain effect correction. The image on the right seems to have been flattened, and the opposite slope reflectances seem much more alike after correction.

    Finally, an essential part of the method’s accuracy is the availability of a highly accurate digital  elevation model (DEM),  up to now, only the SRTM DEM is available globally, and it only has a 90 meter resolution. Its accuracy is somewhat inadequate and sometimes leaves artefacts if the slope changes are poorly located.

    Plus d'actualités

    Biophysical parameter retrieval from Sentinel-2 images using physics-driven deep learning for PROSAIL inversion

    The results presented here are based on published work: Y. Zérah, S. Valero, and J. Inglada. « Physics-constrained deep learning for biophysical parameter retrieval from sentinel-2 images: Inversion of the prosail model« , in Remote Sensing of Environment, doi: 10.1016/j.rse.2024.114309. This work is part of the PhD of Yoël Zérah, supervised by Jordi Inglada and Silvia Valero. […]

    Copernicus should keep S2A operational after S2C launch

    The launch of Sentinel-2C (S2C) is scheduled on the 4th of September 2024, next week ! After 3 months of commissioning phase, S2C will replace S2A, to fulfill the Sentinel-2 mission together with S2B. S2B will later be replaced by S2D. The current plans are to keep S2A as a redundant satellite, in case something […]

    The end of VENµS imaging phase

    =>  It is with some sadness but also a lot of pride that I remind you that the VENµS operational phase will end at the end of July after 7 years of good work. The current acquisition phase (VM5) will stop on the 12th of July. The remaining weeks will be devoted to a few […]

    Rechercher