I am a GAMMA software user, how should I organize the data so that I can use dolphin

[LDcrystal]

Checked for duplicates

Yes - I've already checked

Describe the bug

When I did [...] action, I noticed [...]

What did you expect?

How should co-registered binary SLC be organized?

Reproducible steps

1.
2.
3.
...
If you have a `dolphin_config.yaml` that you used, you can attach that as well.

Environment

Please paste the output of the following:
`python -c "import dolphin; print(dolphin.show_versions())"`
...

[scottstanie]

I have not ever used Gamma (I don't have access to the software).
Is there a description of what the outputs are after coregistering the SLCs?

I'll point out that currently dolphin can read in any GDAL-compatible rasters. So wherever your coregistered SLC stack is, you can point to it like

dolphin config --slc-files ./path/to/*.slc.bin

Are you able to call gdalinfo <name of slc file>.slc?

A follow up- What satellite are you trying to process? If it's Sentinel-1, does Gamma organize them separately by bursts? By swath? Are they in radar or geo coordinates?

[scottstanie]

thanks! are those RSLCs already resampled so that the stack of images are coregistered? Or is this an intermediate step where you still need to use the lookup table to resample the SLCs?

[songzwgithub]

thanks! are those RSLCs already resampled so that the stack of images are coregistered? Or is this an intermediate step where you still need to use the lookup table to resample the SLCs?

These RSLC images have been co-registered, without multi-looking. Within the GAMMA software, these RSLC products are now ready for subsequent interferogram generation. The lookup table file is prepared for geocoding.

[scottstanie]

I'm finding it difficult to find GAMMA documentation online- is this how you'd interpret the SCOMPLEX data format of theirs? as a flat binary file with real/complex values each as an int16?

data = np.fromfile('rslc_vv/20150110.vv.rslc', dtype=np.int16).reshape(600, 2000, 2)
slc = data[:,:,0] + 1j * data[:,:,1]

[songzwgithub]

I'm finding it difficult to find GAMMA documentation online- is this how you'd interpret the SCOMPLEX data format of theirs? as a flat binary file with real/complex values each as an int16?

data = np.fromfile('rslc_vv/20150110.vv.rslc', dtype=np.int16).reshape(600, 2000, 2)
slc = data[:,:,0] + 1j * data[:,:,1]

I haven't personally used Python to read GAMMA-generated SLC files, but I came across some sample code that might be useful. Hope this helps!

  # read data
  cnt = ny*width
  if is_scomplex:
    scpx = np.fromfile(img, dtype=dt, count=2*cnt)
    scpx = np.reshape(scpx, (scpx.size//2, 2))
    data = np.empty((scpx.size//2,), dtype='c8')
    data = 1j*scpx[...,1];  data += scpx[...,0]
    data = data.astype('>c8')
    del(scpx)
  else:
    data = np.fromfile(img, dtype=dt, count=cnt)
  

[scottstanie]

I was able to run it on your data, though it looks the flat earth phase is still present? that's not something handled by Dolphin

[scottstanie]

You can try this script: https://github.com/scottstanie/dolphin/blob/2ecaf7cdf63254d4bec2da9c4accf522da549fe0/scripts/write-gamma-vrts.py
I'll add it to dolphin at some point along with an FAQ or notebooks:

(mapping-312) staniewi:gamma-data$ python ~/repos/dolphin/scripts/write-gamma-vrts.py --par-files rslc_vv/*par
[vrt] rslc_vv/20150110.vv.rslc.vrt -> 20150110.vv.rslc (CInt16, 2000×600)
[vrt] rslc_vv/20150122.vv.rslc.vrt -> 20150122.vv.rslc (CInt16, 2000×600)
...
[vrt] rslc_vv/20160410.vv.rslc.vrt -> 20160410.vv.rslc (CInt16, 2000×600)
(mapping-312) staniewi:gamma-data$ gdalinfo rslc_vv/20150215.vv.rslc.vrt
Driver: VRT/Virtual Raster
Files: rslc_vv/20150215.vv.rslc.vrt
       rslc_vv/20150215.vv.rslc
Size is 2000, 600
Corner Coordinates:
Upper Left  (    0.0,    0.0)
Lower Left  (    0.0,  600.0)
Upper Right ( 2000.0,    0.0)
Lower Right ( 2000.0,  600.0)
Center      ( 1000.0,  300.0)
Band 1 Block=2000x1 Type=CInt16, ColorInterp=Undefined

then config and run like normal:

(mapping-312) staniewi:gamma-data$ dolphin config --slc-files rslc_vv/*.vrt --strides 5 2

(mapping-312) staniewi:gamma-data$ dolphin run dolphin_config.yaml
[INFO|wrapped_phase|L63] 2025-04-17T17:21:07-0400: Running wrapped phase estimation in /Users/staniewi/repos/dolphin/docs/notebooks/gamma-data
[WARNING|wrapped_phase|L490] 2025-04-17T17:21:07-0400: Could not make nodata mask: /Users/staniewi/repos/dolphin/docs/notebooks/gamma-data/rslc_vv/20160410.vv.rslc.vrt is not a CSLC file
[INFO|wrapped_phase|L107] 2025-04-17T17:21:07-0400: Creating persistent scatterer file /Users/staniewi/repos/dolphin/docs/notebooks/gamma-data/PS/ps_pixels.tif

[scottstanie]

1. Not impossible (it ran though fine!), but I'm assuming you'll want to remove it for analysis? I'm sure GAMMA has a way to remove that during coregistration.
2. I'm not sure how geocoding usually works for GAMMA. I know ASF HyP3 uses Gamma, and it looks like they may have a module for it
3. The phase linking does do multilooking during the covariance estimation, dolphin handles the output pixel spacing using the strides option: In my example, I did dolphin config --strides 5 2 ..., which means the x (column) size is 5x smaller, and the rows (azimuth) are 2x coarser. This makes it run ~10x faster, but it's up to you what output posting you want.
4. Yes - though I haven't used this script in awhile, there's this "prep dolphin" script in the opera science team repo

[songzwgithub]

Thank you again. I will proceed to try out the suggestions you provided.

[saboyce]

Were you able to establish a workflow to use dolphin with gamma files? I am interested in knowing whether the flat earth phase can be removed during coregistration and how the geocoding was handled.

[songzwgithub]

Were you able to establish a workflow to use dolphin with gamma files? I am interested in knowing whether the flat earth phase can be removed during coregistration and how the geocoding was handled.

For flat earth phase removal, you can refer to the log section of the import file in this link:
https://github.com/kanglcn/moraine/blob/main/nbs/Tutorials/CLI/load_data.ipynb

[scottstanie]

Thanks @songzwgithub !
Just to repost: here it looks like he's running an offset calculation, then an phase simulator from each orbit using Gamma comands:

    for i,(date,rslc,rslc_par) in enumerate(zip(dates,rslcs,rslc_pars)):
        off_par = scratch_dir/(reference+'_'+date+'.off')
        create_offset_command = f'create_offset {str(ref_rslc_par)} {str(rslc_par)} {str(off_par)} 1 1 1 0 >> {str(scratch_dir/"gamma.log")}'
        logger.info('run command: ' + create_offset_command)
        os.system(create_offset_command)
        sim_orb = scratch_dir/(reference+'_'+date+'.sim_orb')
        phase_sim_orb_command = f'phase_sim_orb {str(ref_rslc_par)} {str(rslc_par)} {str(off_par)} {str(hgt)} {str(sim_orb)} {str(ref_rslc_par)} - - 1 1 >> {str(scratch_dir/"gamma.log")}'
        os.system(phase_sim_orb_command)
        sim_orbs.append(sim_orb)

Then, the simulated phase is read and an cross multiplied by the SLCS np.exp(1j * sim_orb)*rslc . These should be usable in dolphin once you save the results.

[renierlgv]

Following up on this discussion with a similar problem. How to process other sensors data with dolphin? will it directly ingest coregistered slc from isce2 for example? If so, can you give a bit of a "walk trough" on how/what to do?

[scottstanie]

Yep dolphin should work! anything that is GDAL readable that's passed to the --slc-files will be accepted if it has some date(time) in the name. So far I have had success running it on Sentinel-1, NISAR, ALOS, and Capella SLCs. You'll just need to ensure the wavelength input option gets populated for unfamiliar preprocessors.

@mirzaees made this nice full walkthrough for running dolphin on isce2 results for the Earthscope ISCEPLUS short course

https://github.com/isceplus/2025-isceplus/blob/main/5.3_Intro_to_Dolphin/isce2-dolphin/dolphin-isce2.ipynb

The "deramping" shown above does need to happen before dolphin, as we haven't built in anything for coregistering+flattening SLCs due to the wealth of options out there

[renierlgv]

Thank you. My question aligns with @mirzaees case, ie, ISCE2 processed SLC from other sensors. However, I see no deramping step on Sara's walkthrough, which make me believe that the Orlean data used there is already corrected as ISCE2 does the flattening at the interferogram level, right? meaning that I still have to do it myself using the ISCE2 generated range offset files or one of the options out there. Do you have any one in mind that you prefer/recommend?

[scottstanie]

Oh I see, perhaps it's a good idea to note in the docs that Dolphin accepts any GDAL-readable stack, but you need to coregister and phase-flatten them yourself.

Are you using one of the ISCE2 "contrib" workflows? I believe they all flatten phase during coregistration.

[renierlgv]

Noted. I just realized that the ISCE2 conda installation as from @lijun99 lacks the stack folder within contrib workflows you get when compiling isce2 from source (which somehow get stuck in the first step script when trying to create a slc stack for me), so, I'll carefully check the slc stack scripts and see whether or not they do deramp at slc level as it was my understanding it happens at interferogram level. thanks for the help!!

[lijun99]

I added instructions to install stack processors in isce2 a while ago.

In ISCE2/topsStack, the flat earth phase can be removed either in generateIgram step or in resamp_withCarrier step (the default is latter in topsStack while the former in topsApp.py), search flat.

[renierlgv]

Thank you. I looked into the codes and it seems to be the case also for stripmap stacks (ie, resampleSlc.py) with flatten defaulting to True (flatten=(not inps.noflat)). can you confirm? for tops stacks I dont need to go trough ISCE2 as ISCE3/COMPASS handles it. I appreciate your time and help and accept my apologize if some questions are rather obvious as I'm not familiar (yet) with isce.

[lijun99]

That is correct.