How to prepare Sentinel-1 images stack for PSI/SBAS in SNAP

What if the images to be used later for PSI processing (which are filtered to be on the same path number and frame number) aren’t actually lined up, like this

, and there is like a 1 year interval of missing data for such particular frame number that can only be filled with images from different frame the overlaps with the next frame. Will it be valid to use data sets from different frame numbers if my study area lies in the overlap region? or will this cause a drammatic orbital error

Images from the same track can be combined, the different frames of an area do not cause problems (regular intervals are desirable).

I appreciated your reply. That is so much helpful, I can now work on a larger time series since my study area is covered by the overlap between data sets from 2 consecutive frames on the same track/path.

Pardon my little knowledge but I thought the areas lying at the overlap between two acquisitions are prone to produce undesirable range Doppler centroid frequency error that contributes to the total Interferometric phase, that’s why I was only attempting to work on images that are acquired at the exact same frame (even though some images may not be perfectly lined up within a single frame) .

No need to apologize. Many studies combine images of S1A and S1B or acquisitions from different frames. We talked about the effect on baselines here and it shouldn’t matter: S1A-S1B coherence of partially overlapping images - #16 by mengdahl

I am unfortunately not able to answer your question. I just would like to know whether setting the parameter select_reest_gamma_flag to yes, to enable more accurate estimation of PS points, affects the time series plot (displacement vs time chart) where points would appear more in-line and show a regular displacement pattern rather than this noisy plot ?

I was wondering whether this step is extremely necessary or not, because the process is too demanding more than what my machine can handle

I have been informed that the TRAIN toolbox is no longer maintained and therefore the APS removal would be erroneous during StaMPS time series processing, here!

Is there another method to separate that tropospheric phase contribution ?
Otherwise, do you possibly know of some tutorial to apply GACOS correction for InSAR Sentinel 1 Stack ?

Well… that depends on which APS method you will like to use. The ones depending on stocastical models or the ones depending on weather models, etc.

Using GACOS for StaMPS is possible, I believe that are some scripts to do that but I do not have myself yet experience on that.


used the time series filtering linear phase correction method and the results are questionable due to the persistence of topographic pattern indicated by the PS points either before or after the removal of the calculated atmospheric phase, as shown here

and also due to the noisy time series plot which shows random displacement pattern for each PS selected

@mdelgado As you have mentioned before, you have used TRAIN’s basic tropospheric Linear or Power Law correction methods. So, I was wondering if you have you used this GitHub - jgomezdans/get_modis: Downloading MODIS data from the USGS repository script to be called by TRAIN’s Into Matlab after StaMPS step 6 phase unwrapping ? Did it provide you with valid results ?

hello i wonder to know how to process SBAS by SNAP ? Looking forward to your reply, Thanks very much.

it is not officially implemented but some successful attempts ere achieved here: Need someone to help test SBAS method

Hello everyone, i’m working on a quite big area (3 bursts of one subswath is needed) and my memory it’s not big enough to make one year in one process (the error begin already in back geocoding when im using all 32 scenes). With only 10 scenes the tasks are finished with no errors.

So i’m wondering what is the correct way to process them? Is there a way to merge all exports in Stamps? In my head I was thinking that this could be made with patches but after a reading at StaMPS/MTI ManualVersion 4.1b is cleared that the patch division is for azimuth and range, not for temporal divisions. Just to clarify I’ve already processed 10 then more 10 and more and more till complete one year but I can’t get the same PS in all… Some help?

Regarding this parameter (select_reestimate_gamma_flag), the manual and tutorial says

“this step makes a selection of PS based on probability, by comparison to results for data with random phase. This is usually done twice. After the first selection, the temporal coherence of each selected pixel is re-estimated more accurately, by dropping the pixel itself from the estimation of the spatially-correlated phase. Then the selection process is repeated”

I unfortunately couldn’t find the particular paper by Hooper in 2018 that discussed this re-estimation process and I was wondering if you could elaborate why the temporal coherence is re-estimated in Step 3, and why step 2’s calculation if temporal coherence isn’t accurate