radiometric correction is applied to the backscatter intensity. When you derive displacements with interferometry, you are working with the phase. Calibration is not needed, it moreover destroys the complex information required for InSAR. You only need to terrain correct your data in the end.
Really thanks for your reply!
But if I calibrate it and keep the complexity, then it won’t destroy the complex information?
If so, will it be better after calibration?
that is true. But the quality of the phase will not improve.
OK, I understand that! Really thanks for your explanation!!
what is the need for doing terrain flattening on water surfaces?
Shouldn’t filtering be a step right before RDTC and after Terrain flattening. As per your flow the Speckle Filtering sits right after calibration to beta0 and right before terrain flattening
To my understanding, filters should be applied to the original image geometry, so at least before RDTC - here we agree. I’m not sure if it makes a large difference if it is applied on the adjusted radiometry (e.g. in areas of foreslopes and backslopes) or not. It could be that it those areas are better smoothed if the filtering comes afterwards. Is there any reference on this maybe?
It is my intuitive feeling that we filter the corrected radiometry, so right at the end of the radiometric correction and right before the geometric correction. Don’t you think it is not right to filter out the beta0 image before we perform the radiometric terrain flattening?
it does make sense
I’ll update my suggestion above.
I would also say the Orbit file goes before the Thermal Noise Removal or may be it doesn’t matter but that is how it has been suggested most of the times.
well, the orbit information does not change the pixel values so I think the order is not crucially important here. Do you have a reference for that?
You are right. It doesn’t matter. Just thought I should bring into the notice.
Hello Mr. @ABraun ! In SNAP help file, it is written that “GRD products have thermal noise removed” then may I know why is it necessary to remove thermal noise in GRD products again according to your workflow ?
maybe the help is wrong here. I was referring to
If I want to stack calibrated, speckle filtered and terrain corrected RADARSAT-2 products should I co-register them or should I just stack them?
if you want to use interferometry, you skip the calibration to Sigma0 and directly start with the co-registration.
If you simply want to create a stack of backscatter time-series, you calibrate, geocode and then stack them (Create stack). If you want to go sure, you can coregister the geocoded products (Coregistration). This takes more time but especially if you use images of ascending and descending orbits the quality of the alignment of the images is better. Just to go sure you select “geolocation” as initial offset method.
Once you have a stack of multiple images, you can decide if you want to apply a multi-temporal filter. But especially if you are aiming for subtle changes, you need to check that these changes are not elimited by the filter. They are adaptive but there are also reports where they caused the loss of relevant informaiton.
I need to find some trihedral corner reflectors by Sentinel SLC imagery.
So, to do this, is it correct to follow your procedure for SLC data?
Thanks in advance.
yes, the terrain correction is the most important part because it produces geolocated pixels. You can also try the SAR simulation terrain correction which is often even more exact, but you can compare both.
Okay, I try the SAR-Simulation Terrain Correction and then I will compare both.
In your opinion, for find a corner reflector, could be it useful to convert linear to db? If yes, at which step is it okay to do it?
Thanks again @ABraun.
The conversion to dB should always be the last step.