Hello everyone! I need to consult with you about my result! I used Sentinel 1 SLC images from winter 2016 and 2019. Below you can find some screenshots (research area, wrapped phase, unwrapped phase and displacement). Note that I didn’t apply reference points yet so it is only a relative displacement result.
As you can see in the wrapped phase there are some incomplete fringes (red arrows). While the displacement mostly shows a difference in vegetation areas, the incomplete fringes are not only limited to vegetated areas, they are also visible in sand dune area, mudbrick ancient site ( upper red arrow).
After the unwrapping and the displacement proces, these are not visible and the range of displacement also seems improbable to me. I don’t know how SNAPHU is handling these fringes but I would like your opinions on interpretation and processing.
Thank you very much!
Hi. Nice figures you’ve got there
First, when presenting displacement maps or interferograms, always show the coherenec image, with white being = 1 and black being equal to 0.
Second, be careful with vegetated areas.
Third, because SNAPHU does give you a unwrapped phase value does not mean this value is correct. It tries to give you its best estimate but low coherence areas will definitely give you bad displacement estimation.
I am not entirely sure of what you means by incomplete fringe. Can you paraphrase?
Thank you very much for your response!
I totally forgot about the coherence image, I will upload it tomorrow!
What I mean about the fringes is: they are not connected, there is no clear center. There seems to be a change of phase that suddenly stops instead of coming full circle. This is different from the subsidence and earthquake examples I saw.
Is this more clear?
Thank you, I now perfectly understand.
If coherence is very low between two high coherence areas, these two areas are disconnected. You should consider these two areas as independent.
There exists methods using frequency stable scatterers and split-band techniques that allows to retrieve the absolute phase of disconnected areas (split-band assisted phase unwrapping -> https://www.mdpi.com/2072-4292/9/9/879). Note that these methods are quite complex and cannot be performed through SNAP
They could be several things, for example DEM-errors that cause fringes via uncompensated topography.
This the coherence image:
I can check if it is because of DEM by using different DEM,right? If the DEM is the problem, the results should be very different with another DEM. If the result is similar, the DEM is okay.
When I see your coherence map, there is no chance to connect bottom and the top area without prior information, unfortunately.
That can be mathematically expressed. Compute the phase standard deviation using cramer-rao bound. With such a low coherence, you’ll get standard deviation of more than \pi from pixel to pixel, leading to inability of the algorithm to correctly unwrap the phase.