Atmospheric Fringes

I recently processed a pair of SLC IW images to observe some prospective deformations but instead came across some atmospheric fringes (in consultation with InSAR expert) as can be seen in the image below.

Is there a possibility to have some actual deformation since this observation coincides over an area with high coherence?

Which snaphu did you apply?

Did you call it within SNAP, snaphu 1.4.2, ?

Did you apply the default parameters as this,

Or did you apply the latest version snaphu v.2.1?

Also, did you apply goldstein filtering?

Please take a look at this post,

Source of the post

SNAPHU: Statistical-Cost, Network-Flow Algorithm for Phase Unwrapping


In snaphu v.2.1 there are more parameters you could play with.

Also please have a look at the ,

TRAIN - Toolbox for Reducing Atmospheric InSAR Noise

Source :

The urban area does have high coherence but the scene in general looks pretty noisy. Do check out filtering and Snaphu parameters.

To deal with atmosphere you can try to create more interferograms over that time period with different baselines to get a better idea of what you are dealing with. Check out the ESA InSAR tutorial and maybe Hansen’s PhD for a better understanding.

Time series approaches are the best way to deal with atmosphere - you will need to run a PSI (like StaMPS) or SBAS (MintPy) approach.

I processed the pair as per the following flow,

along with the above SNAPHU setting as you mentioned (except the fact that I took 1x1 patch at a time instead of 10x10). I went with the default values for Goldstein phase filtering in the process.

By different baselines you mean one master multiple slave approach to see how the deformation develops over time and correlate it with a natural phenomena that could have possibly resulted in the observation.

as @andretheronsa has mentioned

you could play with the parameters, and as I have mentioned, snanphu v.2.1

You could try up it and then compare the results.

I have the latest version of SNAPHU from,

How does one installs it in a Windows system?

The latest snaphu is not windows version, therefore cygwin is needed to run snaphu through out it, Please take a look at this post,

Source of the post

Seems to be working now

But one question stills remains unanswered.
How do we figure out if a fringe over a high coherence area is an atmospheric fringe or is due to actual displacement?

Please take a look at this post

Source of the post

And this one as well, Source of the post

Thanks for the links @falahfakhri but I would like to know two basic things,

  1. How do we know when we see displacement fringes if they are the result of actual displacement or just atmospheric fringes?
  2. In my case, as again posting below for your reference, the displacement appears to be happening over areas with high coherence so I am wondering why would only areas of high coherence show atmospheric fringes and why can’t it be actual displacement? And if this is due to atmospheric reasons why is there a correlation with the coherence and why don’t we see them in the neighbouring areas including anywhere else in the entire sub-swath at this prominent scale?

In fact I notice all over the image some kind of correlation between high coherence values and fringes. Can someone clarify the trend? This also makes me wonder what else can define the goodness of the fringes other than the coherence map.

that is suspicious indeed. I don’t know if you can determine if some fringes are atmospherically induced or not, but the fact that they are larger than the coherence are makes it quite likely. On the other hand, I have seen wonderful fringes over ice with comparably low coherence.

You can set a low-coherence threshold for the valid pixel expression and see if the remaining pattern over the urban area makes sense. An example is given here Subsidence map in 3d view
Furthermore you can check if there was a rain event during one of both images in weather archies.

Lastly, you can try another image pair and see if the same area is covered by fringes.

It’s not possible to avoid the atmospheric delay of single interferogram, the topographic phase is correlated to the atmospheric phase as well, especially in your depicted case, there is no specific event has occurred which causes highly correlated deformation, therefore it’s not easy to differentiate between both, for more information about how the atmospheric and deformation fringes are appearing, please have a look at this,



Thank you Andreas,
I will try out the method and see if it makes sense in my case.
I tried out InSAR processing on another pair of images located geographically distant from the previous data and this is how it looks

Here we can see the wrapped differential interferogram and the coherence values. The previous pattern can be seen to be again repeating, i.e. fringes in areas of high coherence.

Could you subset only this part and re-process it, in this case I think you could find a clear differences of atmospheric phase and displacement,

What makes you say that?

The unwrapping algorithm, is working over big matrices, and most of the AOI is noisy, that’s why subset small area with high coherence gives better understandable results.

@falahfakhri @ABraun
I am wondering what are the checks and unchecks in the Interferogram box,

In one of the land subsidence videos by RUS there is no mention of subtracting the topographic phase.

In that case it is possible that your area really produces this pattern. Give it a try and continue the processing.

But this area is geographically distant and not the same area as we have been discussing yesterday.