Backscatter normalization in Sentinel-1


I am trying to perform backscatter normalization for Sent-1 images and went through literature; however, I couldn’t derive an exact way to perform it in SNAP. Can anyone please help me with this?
Thank you very much.


Radiometric Calibration to Sigma0 will correct for the global incidence angle (different angle in the near and far range of an image between the incoming signal and the flat earth surface). If you want to reduce the impact of topography on backscatter variation, you first perform Radiometric Calibration to Beta0, followed by Radiometric Terrain Flattening

Please also see here:

1 Like

Thank you very much for your explanation and the link to the correction workflow. My last question is, do we need to calibrate first to Sigma0 and then to Beta nought in a single preprocessing step or they are mutually exclusive? Could you also add when to use Sigma nought and when to use Beta or gamma nought in general?
I am aware that you provided me with the reference materials but still, it is kind of technical for me and quite hard to follow.

You can have both Beta0 and Sigma0 in the calibrated product. But the Terrain Flattening operator only takes Beta0 to convert it into terrain flattened Gamma0.

Now its very clear. Thank you very much.


I have a related question on the preprocessing of SAR Images. I see a lot of studies suggesting conversion to db at the end. Does it really make a difference in improving the signals from the SAR?
Thank you.

There is no general answer on this, please see here:

Thank you, ABraun. Your explanations are very helpful. Could you please tell me one more thing about speckle filtering? Is it true in all the cases that the multitemporal speckle filtering is better in all the cases as compared to a single date? I am working on the estimation of leaf area using Sent 1 where multitemporal SF worsens the model as compared to single date SF. It’s a little confusing since researchers are advocating in favour of multitemporal SF which is not true in my case.

Yes, mult-temporal is not always better, especially for time series which contain subtle or gradual changes. These can ne suppressed by such filters.
Depends on the number of images and the type of information you want to extract.

Thank you very much. Its a great help.