After working with radar data for several years now, there is one thing I am not sure about.
In the documentation of SNAP it says about the Terrain Flattening Operator: “This operator removes the radiometric variability associated with topography using the Radiometric Terrain Correction algorithm proposed by Small [1] while leaving the radiometric variability associated with land cover.”
But radiometry is not only dependent on topography. Assuming flat terrain and a homogeneous land cover, then just the increasing incidence angel from near to far range does lead to a decrease in radar backscatter, which may lead to wrong conclusions when trying to interpret the radar backscatter behavior.
Now my two question are:
Does the Terrain Flattering Operator also account for this effect?
As part of the Geometric Correction, a Radiometric Normalization procedure is mentioned in the documentation.Does this account for the above described effect of range vs. incidence angle?
yes. It takes Beta0 as an input (no angular correction applied at all) and removes the impact of both the global (ellipsoid) and local (topography) incidence angle. An explanation is given here: Radiometric & Geometric Correction Workflow - #71 by ABraun
This checkbox is a leftover from ealier stages of the software. The developers decided to leave it in in case someone is using it in operational workflows, but its functionality has been replaced by the Terrain Flattening operator. Additionally, it seems to have little effect on Sentinel-1 data. We had discussed this here, please have a look at lveci’s answers: S1 radiometric correction - #2 by lveci
Thanks for the clarification. I was assuming the same, but even after all these years with radar data I really doubted myself. I really wasn’t sure about this anymore.
The operation compensates for the local contributing scattering area. In the past, people tried to use the angle of incidence as a proxy for that area, but robust terrain flattening dispenses with that “angle” proxy and calculates the area directly from the DEM. So it compensates for what was formerly know as an “angle effect” but it does so by ignoring angles and integrating all the DEM area applicable within each (range, azimuth) sample coordinate.
