PSI using SNAP-STaMPS workflow in Windows Subsystem for Linux (WSL)

I haven’t yet resolved this issue. I will try it again. I have an alternative solution that I have got MATLAB for linux and i will try to install it and proceed in WSL.

2 Likes

Dear StaMPS user,

I have processed Sentinel-1 SAR images along with ERS and EnviSAT images during my PhD.
This work were focused on the monitoring of residual mining subsidence over decades and the related paper has been accepted.
I had also a computer with Linux System during this period.

Now, I am doing a Postdoc in another lab, on a similar topic, the potential induced subsidence by CoalBed Methane (CBM) extraction.
But, here people are not really familiar with linux.
Moreover, we have chosen to have an intern to process Sentinel-1 images over the period 2014-2019.
Of course, the internship includes a period to discover and understand InSAR techniques by reading papers (Hooper) and books (Feretti).

So, my question : is it possible to perform the processing of Sentinel-1 SAR images with the SNAP-StaMPS only on Windows ? Or is it mandatory to install a WLS or to perform some steps on a computer with a Linux system ?
From what I understand, the main limiting step regarding this point is the first selection based on the amplitude dispersion of each pixel. Am I right ?

Cheers,
Guillaume Modeste

Some reported in here that they managed to get results in Windows as well, but you have to be careful about the compilation of the scripts and how they are called in the shell, espeically the mt_prep at the beginning. In my experience, Windows always raised additional errors, so it might be easier to install linux as an additional system or on a separate machine. If this is not possible, you can try the instructions from the first post, but you might have to get around some individual issues (with probably less people who can help).

How comfortable are you with linux? For people in positions similar to yours (and with some linux experience) I recommend that they look around for an old, unloved and unwanted windows system that can re-purposed to linux. There are huge advantages to using the same platform as the the developers of mission-critical software are using. If you encounter a problem using the same platform, it is easy for the developers to duplicate the issue (or explain your mistake). Using a different platform increase the difficulties in troubleshooting. In my pre-retirement workplace, heavy lifting is done with linux, but at some point there are data clients who use Windows and will import data into ArcGIS, Matlab, etc. on Windows. A couple issues to consider:

  1. If your workflow relies on a commercial package such as Matlab you need to examine the license policies to see if you will be able to obtain a Matlab license for the linux machine, and

  2. If your organization uses a Windows fileserver, there can be administrative issues with access to Windows network filesystems.

Gnwiii, I understand your point regarding the use of the Linux system.
Unfortunately, I am not that confortable with Linux …
In my current lab, they mainly have computers with OS system, some with Windows and none with Linux …

But thanks for your answers, I let you know about my advancements and results.

Hello everybody. I am new to Stamps, I am trying to PSI from wsl. Followed the instructions in the videos and the pdf.
However I have a problem, “mt_prep__snap: command not found”.
Maybe for you it is something simple, but I am stuck in this step.
Thank you very much to all or your comments, I have learned a lot from you.
imagen

Have you modified and sourced the config file correctly as described in the manual?

The error indicates that this has not worked out well.

Hi guys,

We are at the part B of the snap2stamps processing chain, meaning the creation of the ifg.
There are 20 processor cores and 64 Go of Ram in the computer.
However, it takes more than 1h to realize this step for a full sub swath …
From what I remember, that was not the case when I work on linux.
Is the duration of processing related to the operating system ?

Cheers

have you adjusted the -J-Xmx parameter in the snap.conf file?

Is it the one from the package snap2stamps ?
Oh I just saw the link.

From the monitoring tool, I can see that snap is using more than 30 Go of RAM and from 60 to 80% of the processor while processing …
Despite the use of the consequent ressources, it takes a long time (>2 hours for a single pair) to compute this step while in my memory, it was between 5 and 10 minutes … As indicated in this post : Processing time for stamps export

And before, during my PhD, I worked with SNAP v6 while currently, SNAP v8.
Could the new version be less effective in the processing ?

On average v8 is faster. It is possible that there’s a bug that is currently slowing StaMPS export - @estebanaguilera @esteban.aguilera could you verify?

Hi @ABraun. Thank you very much for answering. I am using STAMPS 4.1 beta.
I have read the manual but can’t find the reference for SNAP.
I also do not have access to the “root” folder as its status is denied.
Is it necessary to work from that direction?
Instead I am using the Software / PSI_insar / INSAR_master_data * directory.

Therefore the command I execute is like this:

Software / PSI_insar / INSAR_master_data $ mt_prep_snap 20170602 ~ / Software / PSI_insar / INSAR_master_data / 0.40

There is some procedure or previous installation for the operation of “mt_prep_snap”

Thank you very much for your time.

this is not related to SNAP, but you have to tell where the StaMPS scripts, snaphu and triangle are located so that these can be used in the processing. This is done by modifying and sourcing the StaMPS config file. It is briefly mentioned in the official manual, but explained at more detail here or here.

Once this is done, the mt_prep_snap command can be executed anywhere, but it makes sense to create a new (empty) directory for this, so you can keep track of the processing and delete it in case you want to restart again without deleting too much of other files.

2 Likes

The situation does not concern the stamps export, this is the coregistration, ifg creation, deburst ans the topophase removal step.
On a windows 10 computer, with 12 Go of RAM and 12 processor cores, it takes 7h30 for a single pair.
On a windows 10 computer, with 64 Go of RAM and 20 processor cores, it takes 6h for the same single pair …

If you do that in a single graph you can improve performance by doing things sequentially with shorter graphs.

Solved.

I finally found the solution.
Thanks to the link provided by @ABraun, I solved the situation.
The problem came from the small cache size, about 1 Go by default.
By setting it to 22 Go, the calculation of part B of the processing chain snap2stamps was performed within 5 minutes.
However, that is strange for me that the maximum RAM was correctly set by default (11 Go for the 12 Go RAM computer and 44 Go for the 64 Go RAM computer) while the cache size is definitively too small by default …

Hello , thank you very much for your comments, I was studying this methodology in order to acquire a Matlab license. But for the moment it will not be possible. What tool could I substitute?
I am studying subsidence deformation.
I appreciate your comments

we discussed two alternatives (pyrate and LiCSBAS) here: 5 Interferograms - what is the next step? - #9 by ABraun

1 Like

Hi guys,

This is just to let know that we have been able to perform the whole processing chain from a computer with a Windows OS.
All the SNAP processing have been performed on Windows.
We have installed a Virtual Machine on Oracle to perform the StaMPS processing.
The first results seem coherent with the German monitoring (BodenBewegungsdienst Deutschland (BBD)), so now we can focus on the interpretation !

Cheers

1 Like