Displacement velocity too high?

This topic contains 13 replies, has 3 voices, and was last updated by  Yuxiao 1 year, 4 months ago.

  • Author
  • #2936



    I have been experimenting with a stack of 60+ S1 images from Jan 2017 to Dec 2017. I have followed the tutorials, especially the notes by Mr. Timo Balz, and have completed all the steps with the recommended parameters.

    I exported the def. trend to analyse the velocities of the points, and some of them have extreme results (40-50mm per year – dark blue), such as the airport and other sites.

    Is there something wrong with the processing parametres? Could it have something to do with APS?

    • This topic was modified 1 year, 6 months ago by  steffan.
    You must be logged in to view attached files.
  • #2941


    Yes, there is surely something wrong.
    We should check the APS results

    • #2951


      Dear periz,

      I am reprocessing the dataset with 135 images (from 60 previous), I will post the results soon.

      Thank you

  • #2954


    After reprocessing the point of interest with 135 images, the results are the same (extremes of displacement).

    Here are examples:

    Deformation error

    Deformation error city

    Here are APS parameters:

    Aps error

    Maybe this time series can provide some insight:

    Displacement error

    Thank you!

    • This reply was modified 1 year, 5 months ago by  steffan. Reason: Links not working
    • This reply was modified 1 year, 5 months ago by  steffan.
  • #2959


    Hi Steffan,
    sorry for the late answer. We need to find some time to check your data. We need to see things from the beginning.
    For the moment, I can suggest you to check this simple exercise
    (the last tutorial). It gives you an idea of the results you should see step by step

    • #2963


      Hello friends,

      I followed the tutorial by Mr. Yuxiao very closely. Results seem much better than before, but there are still a couple of anomalies.

      First, the resulting points from 0.85 thresh amp. stab:
      Ref Point 0 99 0 96

      Coherence Before:
      Connections Coherence

      Graph Coherence;
      Delaunay Coherence

      Parameters Plot…

      Graph 2

      Graph 1

      Coherence after weighting (0.8; 0.95)
      Cohrence After Weighting 0 8 0 88 0 95

      Coherence after APS removal:
      Connections after APS

      Coherence after Graph Inversion and APS removal:
      Coherence after GI APS

      There appears to be some problems with coherence, but this is a 3 year timeline over a very large (1 subswath) rural area, maybe it is understandable? Also, do these atmospheres look normal?

      Atmos 20151624

      Atmos 20150706

      Final coherence graph:
      Graph Post
      PS Coherence after GI APS

      MASK: Amp. Stab. Index 1-Sigma/Mu > 0.70

      This looks much better now! I believe the culprit for extreme displacement was the refence point… Still there are suspicious PS’s on NW coastal area, lots of orange (-5 mm) subsidence there.


      A close up of the city:


      And finally a subsiding PS (albeit with about 0.5 temporal coherence)


      Can I trust these results or do params need tweaking?

      Thank you for your support.

      • This reply was modified 1 year, 5 months ago by  steffan.
  • #2967


    * numbers are always relative to the particular case. E.g. .85 with 35 images is different than .85 with 90 images. With 90 images, you can take this threshold lower. to get an idea, plot the histograms of these parameters in your area of interest
    * the bad looking APS you plotted above deserves some closer look. probably you got some problems/messages during coregistration? they look like azimuth ramps.
    * generally speaking, to understand what goes on in your dataset, you should generate the MST graph and using that graph you could process multilooked interferograms. So you can see if they are ok or if there are any problems
    * your result is not yet clean enough to be fully trusted

    • #2971


      So for APS estimation with 130+ images I can reduce amp. stab. thresh to much lower levels? I will try something like .75 and see what happens.

      The only coregistration errors I got were S1 Stitching erros, which I inquired in another thread and was told to ignore. I will try coregistering the dataset again and see if there are any errors on those particular looks with bad APS.

      Thank you for your support as always.

  • #2972


    You don’t need to re-coregister the dataset. What you can do is looking at the interferograms to see if there are fringes or other problems. In case of problems you may try to remove those images from the dataset (delete the corresponding lines in the dataset.txt file) and re-coregister them, if you think it is a coregistration problem.

  • #3019


    So I re-cogistered the dataset (I know I didn’t have to, oh well) and increased the number of reference points to 500K. Didn’t notice any errors except S1_Stitching like before. Using Max Area for AOI selection.

    Then I changed to MST graph I processed Interferograms. There still seems to be a lot of problems…

    Ramping in numerous images, here are some examples:


    Also this problem in over 20 images! Seems like a fail in interferogram formation.


    And finally, some interesting results from others… Not sure if movement or just more errors…


    I know you said to remove pairs with problems, but at the moment that would be half my dataset!

    Sorry for the long post, and thank you for the help.

    • This reply was modified 1 year, 5 months ago by  steffan.
    • #3021


      Hello Steffan,

      Thanks for the post.

      The first few azimuth ramps seem very likely to be caused by the initial coregistration accuracy not good enough. In the next paragraph, I will explain how to tackle this problem.

      I am not sure about the ramps on the range direction, though. It might be orbital errors, but we need to check the logs.

      The blank interferogram and the one with serious discontinuities are definitely wrong. At this moment it would be good to send me the site.log (please try to do an initial check and make sure the logs including the coregistration part are there. I need the log to start understanding what is going on).

      Abou the azimuth ramps, in the coregistration option for Sentinel-1 TOPS, if you click “coregistration parameters”, there is going to be an option called “Orbits Only”. This is a newly implemented method for TOPS that should improve the coregistration results. See

      You can email me at


    • #3023


      Dear Yuxiao,

      I have sent you the site log file to the provided email address.

      In the meantime I will check out “orbits only” coregistration.


  • #3024


    Dear friends,

    I have good news and bad news…

    Good news – Using “orbits only” coregistration parameter not only made coregistration much faster but also corrected the failed interferograms from before, excelent!

    Bad news: There is still ramping, mostly in azimuth, and strange results.

    A – First the ramping pairs:


    B – Now the pairs with what appears to be burst discontinuity:


    C – These pairs seem to have phase shift between what I believe is a river (will confirm), which is quite strange. I’m not sure what could cause this… Could it just be an anomaly?


    D – Finally here are two interesting images. Could this be due to loss of coherence? Displacement? Or a problem with interferometry? Fringes are all over the place. It would be great if someone with more experience than me could explain this type of phenomena.


    By the way, I did not delete previous inteferograms from the folder, as I assumed SARPOZ would OVR the old data when reprocessing.

    Thanks for the support as always,


    • This reply was modified 1 year, 4 months ago by  steffan.
    • This reply was modified 1 year, 4 months ago by  steffan.
    • #3027


      Hello Steffan,

      Could you send the log of this new one to me as well? I will start to investigate from there. I will probably ask for more files but I will check see the log first for signs.



You must be logged in to reply to this topic.