The right way to build Earth vertical displacements

Hello everyone! I’m newbie in this forum, so if I’ve just done something wrong please forgive me :slight_smile:

I need to build Earth vertical displacements of big territory for many years as much as possible. I’ve never do this before, so I found many manuals with various advices. In short I was following the next steps:

  1. picked up Sentinel-1B images with Bperp < 50 meters and Btemp around 12 days.
  2. coregistration
  3. interferogram formation
  4. deburst
  5. topophase removal
  6. goldstein filtering
  7. phase unwrapping
  8. phase to displacement
  9. terrain correction
  10. export to TIFF

In all cases I used standard preferences except “export” in “unwrap” - I entered in “row” & “column” 20 instead 10.
In step 8 I used two different solutions. 1) standard function in SNAP; and 2) I created new band with this expression:

(Unw_Phase * 5.55) / (-4 * PI * cos(rad(incident_angle))).

The pictures were become similar but with different values.

Band with formula:

Band after procedure:

So I have many questions.

  1. Did I get the right pictures generally? And it’s something wrong where I did a mistake? What is the right way to build vertical displacements?
  2. May I need to do procedure “multilooking” before filtering?
  3. When I did merge of swaths I got weird picture. Does it means that I cannot merge this swaths into one image? Is it possible or not?
  4. Can I calculate Earth horizontal displacements? All is need - to make these steps with HH polarisation, am I right? If not, how can i make it?

Thank you in advance for your answers!

Statistics of band with formula:


Statistics of band after procedure:


Image of bands with formula after merge and its statistics:


Image of bands after procedure after merge and its statistics:


I’m really sorry that I’ve just flooding my post of screenshots, but I don’t have opportunity to upload many images in one post yet.

I’m hoping for your understanding :slight_smile:

just a few comments.

Contrary to your understanding, the selection of VV or HH polarization does not primarily result in vertical or horizontal displacements. Both measure the change of elevation along the line of sight. There are many topics discussing how to get absolute displacements but this is not possible with a single scene and also not implemented in SNAP.

Your processing steps seem correct but probably the perpendicular baseline is not ideal, also the quality of the orbit files plays a role, especially if the data looks “ramped” like in your case.
Also it would be worth trying a different DEM for the topographic phase removal and compare if the output makes more sense.

Finally, as you are working on very large areas and the final resolution will probably of minor importance I would suggest multi-looking to about 30 meters (the resolution of SRTM 1Sec) after the debursting. Of course this reduces the spatial resolution but the outcome is less prone to unwrapping errors and these kind of ramps could be avoided. I personally think having more realistic displacements over the full area is more important than the size of the final pixels.


I’m sorry for my long response for you - personal circumstances.

@ABraun Thanks a lot for your advices, that’s exactly what I needed =) Yes, you’re right. I need realistic displacements. It’s more important than the size of the pixels.

I did all the things you said. The only thing I’ve changed is multi-looking was done before snaphu import. Idk, but final image was as so rough as early when I did multi-looking after the debursting.

I have 2 questions about your advices:

  1. Am I right when I’ve tried DEM “SRTM 1Sec HGT” at all operations I’ve done?

  2. Am I right when I’ve entered such value in “Number of Range Looks”, that “Mean GR Square Pixel” was equal around 30?

The derived pictures of the same area displacements. I hope the pictures are become more realistic and suitable:

August 2018:

September 2018:

yes, this is the best which is currently freely available.

multi-looking generates square pixels. Depending on the incidence angle of your acquisition, the ratio of azimuth and range pixels may be different.

1 Like

So what the value should I need to enter in Number of Range Looks? It depends of what?

I hope that I don’t bother you so strong =)

the lowest resolution you find tolerable. There is no right or wrong.

1 Like

Ok, I got it. Thanks a lot for your answers, @ABraun! :smile: