Effect of number of bursts on interferogram

Microwave Toolbox Interferometry
1

Hi,

Maybe someone can shed some light on an issue I’m facing when computing interferograms using Snap 12.

I have the following processing chain described in the TOPS Interferometry tutorial (link):

TopsarSplit->ApplyOrbit->BackGeocoding->ESD->Interferogram->Deburst->MultiLook->Subset->Snaphu Unwrap->Phase2Displacement->Terrain Correction->GeoTiff Export

Now I get different results when I choose different bursts that cover the region of interest.
Using bursts 6&7 (which already cover the ROI fully):
I get a displacement range of -9.81cm to 6.72cm

Using bursts 5-8, the displacement is roughly the same: -9.86cm to 6.88cm

but using bursts 4-9, the displacement is significantly different: -13.40cm to 3.20cm

As far as I can tell, the step that introduces the first difference is the Interferogram.
Is this behaviour expected and which of those ranges is considered more “accurate”?

The picture below shows a section of the ROI and I chose one point in the centre of the dark red spots to illustrate this behaviour.

image
image1810×1354 362 KB

The reason why I am using additional bursts is due to the fact that sometimes subsetting does not cover the whole roi and I am working with an extended region that ensures that I can crop to the right size at the end of my processing chain.

I’m a bit baffled that using more bursts should affect the end result so much…

Kind regards,

Steve

2

Just some extra information.
It looks as if it is just one burst that makes the difference…
The numbers are somewhat consistent (ie reasonably close) when I use different combinations from burst 5 to 9 inclusively. As soon as I add burst 4 to the mix, the resulting displacement changes significantly.

3

InSAR is a relative measurement. A limit of (-10.0, 10.0) can be the same as (-5.0, 15.0). If you want to check for consistency, look for consistency between difference of the same 2 pixels in each of the analysis. Your current method of checking for consistency by looking at limits of the final raster is not appropriate.

4

@piyushrpt
Thank you for the answer! That helped me figure out what I was missing.
I found a GPS sensor in the roi which reported about ~0cm/year vertical movement and using that location as a fixed point, I could adjust the various displacements and they are now reporting roughly the same displacement no matter what the number of bursts is…
So I guess, it’s the snaphu unwrapping that causes this relative shift… It must be using slightly different points to start its unwrapping from, even when I subset to the same region of interest.
Moral of the story: If you want a consistent result, you definitely need a fixed point to calibrate the results to!
Thanks again!