I did a few tests because I still wonder about the best pre-processing parameters for Sentinel-1.
I therefore downloaded a Scene from 12.09.2016 which covers Cyprus (GRD and SLC). I chose this area due to its topographic complexity and the little vegetation cover.
I wanted to see the change in the result if different input data and modules are applied. These were the parameters:
- Incidence Angle Image
- GRD > RD Terrain Correction with checkbox “apply radiometric normalization”
- GRD > Calibrate to S0 > RD Terrain Correction
- GRD > Calibrate to B0 > Terrain Flattening > RD Terrain Correction
- SLC > Calibrate to B0 > Deburst > Terrain Flattening > RD Terrain Correction
These are the results:
You can compare them best when you download them and skip through them (by Windows Preview/Slideshow, for example)
radiometricCorrection.zip (1.8 MB)
Incidence Angle Image:
GRD > RD Terrain Correction with checkbox “apply radiometric normalization”:
GRD > Calibrate to S0 > RD Terrain Correction:
GRD > Calibrate to B0 > Terrain Flattening > RD Terrain Correction:
SLC > Calibrate to B0 > Deburst > Terrain Flattening > RD Terrain Correction:
This raises the following questions?
- Does radiometric normalization in the RD Terrain Correction module replace/substitute Calibration? You can’t apply both. The result is however the very same.
- Shouldn’t actually Terrain Flattening somehow replace the topographic/radiometric normalization? The first relies on the illuminated area and the second relies on the incidence angle (see a nice comparison here)
- Why are there slight differences between terrain flattened SLC and GRD data?