Subsidence map in 3d view

I would be interested in your results and how they are different. I am not sure either :slight_smile:

Once I have some results, I will post them.

1 Like

I’m trying to create a subsidence map using the steps mentioned above.
It looks like that I have a problem. My unwphase images (always) have values which decrease from one side if an image to another. As a result Step1 image [(Unw_Phase * wavelength in mm) / (-4 * PI * cos(rad(incident_angle)))] has the same pattern of its values too. Values here fall from 65 on the north to - 20 on the south of image. I can’t find where my mistake or how to fix it.
My processing steps.

  1. Split
  2. Orbit file
  3. Deburst
  4. Coregistration
  5. Ifg creation
  6. Topophase removal
  7. Goldstein filter)
  8. Unwrap
  9. What ABraun mentioned in the post Sep. 8, 2016.
    I tried all the precessing steps possible: with and without topo, range and azimuth filters, all the options of Unw Export, etc. Despite that I always receive these “angled” unw.phases.
    Please recommend me something. Thanks!

can you please show us the interferogram before unwrapping?
What is your perpendicular baseline?

This is my phase.
Here I have a baseline of 24m since I used 2 closest passes of the same sattelite. But in another case, I have a baseline of 80 m (2 satellites used) and unw looks the same. I started to think about SNAPHU, maybe is it possible that it works wrong?

The impact of topography on your phase is too low. Try another image pair with a higher baseline:

1 Like

Hi ABraun,

Actually I could not make Snaphu Export working when using the phase after masking with coherence threshold.

I used the coherence coefficient (threshold=0.50) to create a mask and used this mask to extract related phase (wrapped). A new phase band was generated. I gave this new phase band a different name (compared with the original phase name) and added it to the original data file. In this way, Snaphu Export worked and after unwrapping process, I noticed that the new phase band I generated was not used since no data gap was found in the unwrapped phase.

Then I tried another method. Deleted the original phase and gave the new phase band the same name as the original one. However Snaphu Export did not work and the reminder is “Wrapped phase band required”.

Snaphu Export did not find the wrapped phase band which was not because of the file name of wrapped phase band. Maybe because of some other information in the original data file. However I just created the new phase band with Band Maths, I did not think I changed much head information of phase band. Maybe the developer of SNAP can help to explain this.


Masking and unwrapping was reported to have issues several times. Try unwrapping the original phase and mask out low coherence areas at a last step.

Would you like to tell me please, what is a vertical resolution of a subsidence map? Does it equal to a wavelength?

It depends on the level of phase-noise (coherence) and can be a small fraction of the wavelength. Please note that the error-sources (atmosphere and ionosphere) are much larger than this.

mengdahl, thank you for your answer. Could you please suggest me something to read about it. Maybe some equations or something…

I’ve been managed to create a nice subsidence map. Thanks to ABraun and his formulas.
These are the processing steps I’ve taken. On a finish of the process, I created a decision point: you may not exclude the low coherence points for having a map without empty places If you can accept possible mistakes caused by insufficient coherence.


Pardon me, Tomcater, but isn’t the RD Terrain Correction supposed to be the last and final step of the processing chain? Can you explain your reasoning why you carried out the band math phase to displacement operation and pixel subtraction after geometric correction?

Thank you!

Hello ABraun,

I myself am also struggling with eliminating a phase ramp in my study area that occurred during unwrapping. This post of your is quite interesting, though I was unable to find any information as to what it might mean. By a “modeled surface” do you mean a more accurate DEM to be used during topographic phase removal or are you referring to a different operation altogether? Any information on this (links to papers included) would be greatly appreciated.

Thank you!

what I meant was to somehow create a raster that shows the overall trend of your ramp. If you can get that, for example by selecting and interpolating points at regularly distributed points you can then subtract it from your unwrapped phase.
My colleague does this because his TanDEM-X data maybe lack of precise orbit state vectors and this creates a ramp that superimposes his interferogram. With basic GIS steps (creating points between fringes, reading out phase values, interpolating points to get a surface of the ramp) he managed to reduce the ramp to a acceptable degree. But it’s just a work-around.

I quickly made a image to illustrate it:

Hi, Momut1
I do it because Local Incidence band is created by Range-Doppler TC. Before it, this band does not exist.

I have a suspicion that Back-Geocoding + ESD without any Subsets provide minimum ramp possible.

Very interesting. Can you perhaps share your colleague’s name and/or results or perhaps a paper he might of written/collaborated on documenting the process? I will try his method, but having a paper to reference is always helpful. As far as I understand from your post he does this when no precise orbits are available for his images. I am curious then if the process works well in the case of applied precise orbits.

By “local incidence band” you mean the incidence angle raster located in the tie-point grids folder of the product?

Vis a vis:

If so, I would disagree, in the example i’ve attached above the product is not terrain corrected and yet has the incidence angle raster.
Might you clarify this for me?

Thank you!

It is not published yet but you can do these steps in any gis. His TanDEM-X data had precise orbits but there was still a ramp indicating that there is a wrong angle estimated between both acquisitions.