This might look like a bit of a basics-related query, but I’m unsure of a few questions.
The data I use is S1 EW GRD Medium resolution.
One, I want to know if incidence angle correction is necessary for all the scene preprocessing. Because, following this forum, I performed the basic radiometric and geometric corrections (using Ellipsoid method), but I did not find incidence angle normalization as a mandated step anywhere.
So, is it mandatory to do incidence angle normalization?
Two, how do I identify the frequency and incidence angle changes in general of a particular scene in SNAP?
Third, I made sure the row and paths chosen for the study are identical throughout the multiple scenes. So, does it mean the incidence angle would be the same for all these images?
My application: sea-ice studies, and from the previous topics on the forum I got a few points that say, incidence angle normalization is necessary for sea-ice studies. How do I do it in SNAP?
Thanks in advance
As we know, for Sentinel-1 data the incidence angel varies significantly (20º-46º) which means the backscatter intensity from objects is also affected. Incidence angle normalization is required based on what your application is. For instance, When we try to classify sea ice types, incidence angle normalization is important, This is because, the same sea ice type might be located in both near and far range of the incidence angle where backscatter signature (of that same sea ice type) might be different due to the incidence angle. As the incidence angle increases, the backscatter intensity decreases.
Hence, when you perform sea ice classifciation, this can be an issue.
SNAP does not support incidecne angle normalziation. You will have to you a programming language to do that.
I did my master degree on sea ice type classification. You can a look at my thesis here where i describe the equations to do that. Another nice paper you can look at is this
@Gk_burada by the way,
If you are willing to use python to perform incidence angle normalization, I have just added a script i wrote in python to do that. It can be found in my github page.
You just need to digitize some points over the SAR image from near to far range for the algorithm to do the regression analysis and perform angel correction. I have instructions on github
Wow, that’s great of you, many thanks for responding @johngan, and for the attached code.
Having gone through “Classification of sea ice types for the East part of Greenland waters using SENTINEL 1 data”, I have a straightforward question on the incidence angle correction - like, most of my scenes have constant rows and paths, and also constant orbit. does the incidence angle affect such scenes too?
The incidence angle of Sentinel-1 is the same regardless what the satellite orbit or location is. Using a mode such as wide swath or interferometric wide swath, the S-1 swath acquisition is very wide and that affects the brightness of the scene from near to far range.
For oceanographic application where the whole scene is smooth, incidence angle normalization is required
One last question, I have processed huge scenes already following the group instructions.
For some reason, no one ever mentioned about incidence angle correction in mandatory steps.
Is there any quick alternative for the rectification or just re-doing everything is the only solution?
If you have a geocoded incidence angle (ellipsoid) file per path-row as this doesn’t change between acquisitions. The normalization should just be a simple band math operation on the geocoded products - for this specific model mentioned above.
Incidence Angle Correction is very discipline/ application specific step with a specific model. While the model proposed for sea ice is very popular - you will find a ton of literature on why that model doesn’t work for other terrain types - example: https://pubag.nal.usda.gov/download/56285/pdf
As @piyushrpt mentioned, as long as your processed output is calibrated and terrain-corrected and contains the following 3 bands: 1) HH_sigma_nough, 2) HV_sigma_nough, 3) incidence angle then you can apply the equations and perform the angle correction.
No need to re-process them