Incidence angle correction of Sentinel 1

@junlu, @lveci
Hi,everyone,I have read something about"incidence angle" through the forum, and I have some questions that make me confused. Is the effect of the incident angle really corrected when using SNAP for radiometric calibration? According to the official documentation ,as show above,I think the EAP and RSL has been corrected,but the effect of incidence angle not be corrected. Conversely,if corrected,Is there a reference angle for correction? you know that the incidence angle range from 30 to 45. I encountered this problem in the analysis of time series data, there are different angles among the time series data.This problem has troubled me for a long time.
Looking foward to your reply.
Thanks
Chen

@ABraun, @bwbj, Looking forward to your suggestions.Sorry ,I can only aite two people at a time

@johngan,Can you give some advice? thanks

Hi Chen,

I think the EAP and RSL has been corrected, but the effect of incidence angle not be corrected.

This is correct. S1 Level 1 products (GRD and SLC) have been corrected for EAP and RSL in the Instrument Processing Facility (IPF).

In SNAP, when radiometric calibration is performed (i.e. the conversion of digital numbers to radar cross-section (sigma0)) I would assume that the sin(incidence angle) term is used (there is no reference incidence angle).

If you measure, for example, the same forest at difference incidence angles then the measured sigma0 will be different even if there has been no physical change in the forest. Every surface feature will have a sigma0 that is a function of incidence angle. So you need to be careful when comparing the sigma0 of a feature at difference incidence angles.

Hi Chen,
Based on my understanding of Sentinel data system, there are two levels processing steps in term of incident angle: the radiometric correction you mentioned, and normalization which needs a reference angle in the center of wide-swath.

The first level radiometric correction does take incident angle effect to sigma0 into consideration(like the formula you post, there is a gridded look-up-table affiliated with SAR image), although this is not enough to account for all the effect. In other words, the second level of incident angle correction does not work in default SNAP tools.

I guess there are at least two reasons for that:

  1. for some narrow-swath sensing mode, the incident angle does not vary much, only 5-7 degree or so.
  2. to compensate all the incident effect, you need a large homogeneous area across the whole range in scene(like forest peter mentioned), which is not always easy to satisfied.

Hi, Peter
Thanks for your reply.According to my understanding of your reply, there is no incidence angle correction, here is one of the PPT (from David Small, Nuno Miranda, Erich Meier ),pay attention to the formula in the red box, I think that the sin(incidence angle) term is just like the term sin(θ),if you wang to get Gamma naught (not Sigma naught)just replace it with tan(incidence angle).Right? I’m not sure whether i understand it correctly.
In addition, the effect from incidence angle shouldn’t be ignored when comparing the sigma0 of a feature at a different incidence angle. So, it is necessary to correct the incident angle after radiation calibration.Of course ,just a coarse correction.
Chen

Hi,bwbj
Thanks for your reply. As for the two levels processing steps you said,I have different opinions.The first level may not adjust for the effect of the incidence angle, as the pictutre show above. the second level : the normalization in the Range Doppler TC, it uses the following formula:

I think it simply adjusts the influence of changes in the incidence angle caused by the local topographic factors, not the effects of incidence angle in different range distsnce.
Chen

Hi Chen
We have the same point of view for the second level incident angle processing, where the normlization only account local topography effect based on DEM you choose, but not for range effect.
However, for first level, I am not quite sure as well.
Expect more illustrations from experts in the forum.

Expect more illustrations from experts in the forum.:slight_smile:
Thanks.

Gamma Naught is implemented in SNAP by first calibrating to Beta0 and then using the Terrain Flattening.
If I recall correctly there were some changes between its first implementation in NEST and the module today but it still works fine in moderate relief.
Have a look at this topic: S1 radiometric correction

Hi, ABraun Thanks for your reply.
But I’m still confused. Based on my understading of Gamma Naught, the Terrain Flattening will adjusts the influence of changes in the incidence angle caused by the local topographic factors, not the effects of incidence angle in different range distsnce. And it perform better than Normalization in Range doppler TC. I’m not sure whether i understand it correctly. I’d like to normalize SAR data to a fixed incidence angle, :worried:
Look forward to your reply.
Chen

yes, Terrain Flattening only compensates for topographically induced radiometric distortions. Maybe I misunderstood your question, sorry.

Thank you very much @ABraun,I learned a lot in the forum,:grinning:

Hi Chen,

… there is no incidence angle correction…

Just to clarify - a incidence angle correction is required when converting DN to sigma0 but this correction is not applied to S1 products themselves. It needs to be applied by the user or the software they are using.

… I think that the sin(incidence angle) term is just like the term sin(θ).

These two equations are identical.

…if you want to get Gamma naught (not Sigma naught)just replace it with tan(incidence angle).Right?..

Yes you are correct.

In addition, the effect from incidence angle shouldn’t be ignored when comparing the sigma0 of a feature at a different incidence angle. So, it is necessary to correct the incident angle after radiation calibration.

Yes this correct as well.

Peter

totally agree!

I think that doing radiometric terrain flattening is the best option for most users. The fact remains that most targets scatter non-isotropically so they will look different when viewed from differing incidence-angles. The resulting difference in backscattering is signal and not something that should be normalised away, in my opinion. If this causes problems in your analysis you should treat the time-series separately (one series per incidence-angle).

Hi,Peter

That means the S1 product whether SLC product or GRD product, no incidence angle correction is considered(during Calibration).When I read the contents about Calibration in the Help of SNAP, the incidence angle correction is mentioned for ASAR、ERS-1、Radarsat-2、etc,but not for the Sentinel-1.[quote=“peter.meadows, post:17, topic:9639”]
These two equations are identical.
[/quote]
Yes,:grinning:

As for the correction method,I have searched for some literature,could you give me some advice?:slight_smile:

Thanks for your reply and help me solve some confusion.
Best regards
Chen

Hi,mengdahl

As you said,complete angle correction is very difficult, but can we try to do some compensation? I have searched for some papers about this.Maybe I ended up failing.:worried:

Thanks for your reply.
Chen

Please summarize this topic.
How to normalize the incidence angle?

come on… It takes about 5 minutes to read it yourself :slight_smile: