temoral coherence problem

[yohai magen]

1. Description of the problem / issue

I'm trying to produce a time series of a stack of 26 Sentinel-1 SAR images that span ~ year and a half and processed with GAMMA to 126 interferograms where each one of them has average spatial coherence > 0.9(attached figure 1 ).

when I'm trying to invert the unwrapped interferograms to time series the temporal coherence gives "weird" results base on the chosen reference point

figure 2 shows results of a reference point in the lower left of the images.

and figure 3 a reference point in the upper right of the images.

do you have an idea to the source of this behavior? and what I need to do to fix it?do you have an ide to from where this behavior can come from? and I need to do to fix it?


2. The full command that generated the error AND the full error message

***in config file****
mintpy.reference.yx            = y,x
********************

smallbaselineApp.py ./longvaly.txt --dostep reference_point

smallbaselineApp.py ./longvaly.txt --dostep invert_network



3. System / software info:
Operating system (macOS / Linux / Windows): Linux
Python version: 3.6.10
MintPy version (output of smallbaselineApp.py -v):v1.2.2-32
4. Dataset info, if reporting an issue with a particular dataset:Sentinel-1A/B
Used InSAR processor (ISCE/topsStack, gamma, snap, etc.):gamma
SAR sensor and its imaging mode:Sentien-1A/B IW

[Yunjun Zhang]

Hi Yohai,

The low temporal coherence means that the triplets of the interferometric phase are not closed. With help from Piyush, we can think of the following two possible causes:

1. Did you apply any phase deramping to the interferogram before the network inversion? Independent phase derampping for each interferogram will cause closure issues unless it's the network derampping approach like GIAnT has.

2. Does GAMMA use some sort of polynomial for coregistration? These horizontal patterns almost look like a polynomial overfit during coregistration. For S1 data, I would recommend check interferograms for burst boundaries in the unwrapped phase. If you could show the unwrapPhase.png and coherence.png files here, that would help diagnose. The entire pic folder would be even better of course.

Yunjun

[yohai magen]

 Hi, thanks for the reply and suggestions.

 

1. I’ve tried deramping all the slcs stack with the reference image ramp, after doing so I am still getting the same behavior although with a bit wider strip of temporal coherence(see the temporal coherence figure).

2.  I don't think that I see the boundaries of the bursts. see the unwrapPhase and coherence figures.

 

[yohai magen]

Im posting the figures in separate posts due to file size limit

[yohai magen]

[yohai magen]

[yohai magen]

[Yunjun Zhang]

Hi Yohai,

How about the numTriNonzeroIntAmbiguity.png and timeseries.png file but without masking (view.py --mask no)?

Yunjun

[yohai magen]

here is the timeseries.png.

what is the numTriNonzeroIntAmbiguity.png and how to produce it?

thanks

yohai

[Yunjun Zhang]

Make sure to have latest development version on github, then it should be generated by default in the pic folder.

Yunjun