Indeed it is foreseen.
All feedback is welcome. The workflow is clear, but any comments and suggestions for logging or other is more than welcome
1 Like
Hi Jose,
The snap2snap has been downloaded and unzip in linux, the python 27 is existed, first thing I have confused, how to run it from everywhere without to navigating to the folder of the py.scripts, and secondly I got the following error.
can you share the content of the project.conf file?
Had you installed the pathlib python package as in the manual?
Thanks for your response, Yes I did, and has been installed successfully.
Then it sounds like a permission issue. Please ensure the scripts can read and write in the destination folder
For successive steps you should also define the gpt path and the master image (splitted and applied orbit)
Hello snap2stamps community,
I have been using this snap2stamps with success to map volcanic areas and need some advice on data quality. I have noticed that smaller stacks have less temporal decorrelation (higher PS density), and this makes sense. I am wondering if there is a general rule for minimum stack size (e.g. 8 images)? Also, I have been kriging the exported PS maps, and am wondering if anyone has experience doing this? Is it wise to filter the PS results (e.g. remove PS below a certain r2 value), or perhaps the kriging process smooths the data and this is unnecessary? Any thoughts are appreciated
Hi @wharfbanger,
I am glad the snap2stamps package makes life easier! It was developed for that purpose
1)Regarding the stack, I agree that smaller stacks provide higher PS density, but remember than minimum PS analysis with minimum quality requires at least 20 interferograms (I personally suggest >25)
2) Kriging is an interpolation method, so clean and quality points will provide better interpolated maps. noisy points will provide worse. This is my personal opinion, please correct me if I am wrong.
I hope this helps.
Best
Thank you for the fast reply. I have been experimenting with different stack sizes (44, 21, 11 and 7 interferograms), for the same area. The time period is after 2017, so the stacks have 12-day repeats, with very few gaps.
The study area contains a powerful groundwater well (geothermal well) that has caused subsidence, and this is clearly visible in all results - even the 7 stack! Another factor for me is the short but intense dry season in my study area, which eliminates vegetation foliage for a period of about 4 months every year. By selecting shorter stacks that cover only the dry season I think I am avoiding de-correlation from vegetation growth in the wet season. So, it seems that PSInSAR works very well, even for the smaller stacks. Can I ask where your minimum of 20 comes from?
I have been encouraged to use the smaller stacks because the PS density is so much greater.
My logic is this: for the small stacks I have very high PS density (I understand individual PS accuracy is very low). Kriging smooths the inaccuracy by averaging 16 – 20 PS in the search radius. Perhaps this is better than having a larger stack with very low PS density? I have been checking data quality by calculating r2 and p-values for each PS time series, and these provide “quality checking” layers. I don’t know how familiar you are with kriging, but it may be worth mentioning that the variograms for all datasets are beautiful (even the 7 stack), which suggest to me the spatial trends in PS are real.
Snap2Stamps has allowed me to experiment with many different stack sizes and time periods. I think it is a fantastic tool.
I am afraid you will be measuring seasonal effects and not long term displacements using only 7 x12 (84 days < 3months).
the minimum of 20 images come from many studies, when you check literature on PS analysis, as some of the phase noise is estimated, the mathematic approach provides good accuracy once reaching a relevant number which allows to estimates.
I am just wondering if you could use those 7 images per year and put them together, then you may have over 28 images (only covering the dry season). What about this?
I hope this helps
Best
Thanks Jose! I will try the longer dry season stack as you suggest as I had not thought of that. However, I still think my results are conclusive, even for the short stacks. I have uploaded a spreadsheet for you to see my results, and perhaps this will convince you that the short stacks are working 8-). The spreadsheet contains time series with plots for 3 stacks (7, 11 and 21 interferograms). The time series are for PS nearby to the powerful geothermal well (see yellow squares on maps in spreadsheet) and all show a clear and strong subsidence signal (about 40 - 60 mm per year).
Geothermal well time series for different stack sizes.xlsx (2.9 MB)
Also, I do have access to the literature and I have had a good look as you suggested. I have found a few papers where the “20 minimum stack size” is mentioned, but never properly explained. Do you have a good reference? It seems that if you have a strong and smooth subsidence signal (like a geothermal well causing subsidence), then this may be an exception to the “20 stack rule”? Do you agree?
I hope you understand that 20 images is a good practice.
Che k in the literature many of the papers doing PSI, and then see how many images they normally use.
It may be the case using few due to data restrictions/ constraints or availability. But this is not the case for sentinel1
Yes Jose I think I understand what you are saying, that 20 images is good practice. I thank you very much for your comments, and Snap2stamps is a great tool!
1 Like
I am sure that you did and are doing a great job! Look forward to read the full story!
Hi Jose,
I am looking at an Island volcano and get the error at Step 6 of stamps:
“Error using uw grid wrapped (line 84)
Minimum dimension of the resampled grid (27 pixels) is less than the prefilter window size (32)”
Inside SNAP I run TOPSAR-split and select 2 bursts, which I understand is a minimum for snap2stamps. I think the error is because the very small island (3km2) is completely inside a single burst, and is surrounded by a huge area of ocean (only ocean inside the other burst). Is there anyway to process small islands in a large area of ocean?
Thanks,
Mark
what happens if you set the window size to 24, for example?
setparm('unwrap_gold_n_win', 24)
I performed StaMPS on a single burst already (just needs leaving out the ESD operator during the preprpocessing), so this should not be the problem.
1 Like
Try to set up the window size for the unwrapping to 16. Sometimes it solves the issue.
Let us know
must the filter size be of 4, 16, 32, 64…? I haven’t thought of this.
Brilliant! That worked!!
Thanks!!!
1 Like