Attempting to measure rock glacier displacements using DInSAR

[bro076]

EDIT: Sorry if this post seems longer than necessary, I have just tried to be comprehensive to help others with the same issue!

Hi,
I am trying to use DInSAR based on Sentinel 1 images to look at the displacements of rock glaciers in Kyrgyzstan. I’ve been reading about the methods as much as possible. I read on this forum and in the ESA InSAR principles that the perpendicular baseline should be less than 300 m, for me they are all under 110 m.

From papers that have addressed similar topics such as this and this the temporal baseline is typically around 45 days, but in some cases around 90 days or greater.

Based on this I have tried using the following InSAR pairs:

• 16/08/2016 - 11/08/2017 (360 days)
• 05/06/2016 - 09/09/2016 (96 days)
• 18/07/2017 - 23/08/2017 (36 days)
• 06/07/2017 - 23/08/2017 (48 days)
• 23/07/2016 - 09/09/2016 (48 days)

I then used the following methods, batching 1- 7:

1. Deburst
2. Subset to a very small area for testing
3. Compute Interferogram
4. Topographic Phase Removal
5. Goldstein Phase Filtering
6. Multilook
7. Snaphu Export
8. Unwrap
9. Snaphu Import
10. Phase to Displacement
11. Terrain Correction

All but one of the unwrapped phases show little data, for example this is the unwrapped phase between the 23/07/2016 and 09/09/2016

Here is the most complete unwrapped phase (from 06/07/2016 - 23/08/2017)

Which translates into the following displacements:

While it would be supposed that the valley bottom would be most inactive, I was wondering if what I am seeing instead is elements of the topography still in the data?

So I have the following questions:

1. do these unwrapped phases look OK, or has something gone wrong?
2. are there any settings that should be tweaked in the methods? (since I just used the default values)

Many thanks

[mengdahl]

What are the coherences of those pairs? Probably very low. You need to have high enough coherence so that unwrapping can work well.

In the reference articles L-band data is used, which enables their use of longer temporal baselines (targets tend to decorrelate a lot slower in L-band than with C-band). With Sentinel-1 C-band I would start looking at 12-day pairs and see if any of them shows a signal.

[bro076]

Thank you for the reply. I download some 12 day pairs. Between 16/09/2017 and 28/09/2017 I get this conference

With these displacements (showing where the coherence > 0.5)

It’s hard to really see any proper displacement patterns in this. I will try more 12-day pairs and see if I get any different results.

Thanks

[mengdahl]

Ok so coherence is good at 12-days. If the displacement is slow the signal might be difficult to see. Looking at more data might help.

[bro076]

I downloaded some ALOS data from the offthefly service. The coherence seems good but I seem to be running into problems when unwrapping the phase.

Here is the coherence

the wrapped phase

and the unwrapped phase

Do you have an idea of what could have caused this?

Thanks

[mengdahl]

One cannot meaningfully unwrap noise, and in that pair there are large decorrelated areas (snowfall perhaps?). If the terrain-caused fringes are too dense for unwrapping you could try finding a pair with a shorter baseline.

[bro076]

Towards the centre of the image are some glaciers, but in the valleys there is good coherence. Can that not be used to unwrap the phase? The temporal baseline for my ALOS pair is around a month, so less than the papers I originally linked to that used 46 days.