I’m using: S1A_IW_GRDH_1SDV_20180321T055027_20180321T055052_021109_02444F_007A
My processing graph is here:
GRD_Processing_dB.xml (7.6 KB)
Visualizing the resulting VH band in dB and switching of the no-data value to show those pixels as white, results in this over a lake:
So I guess the issue is clear here. The FloorValue is not set in the thermal noise removal, and these pixels become 0.
Guys, honestly, thanks for the support, but this is kind of discouraging for me… It’s taking already many months now to get one proper answer and solution, and it seems nobody is really taking the time to look into this using the information I have provided (which already took a lot of my time).
@lveci, I have provided my graph XML in the earlier post. As you would see in there, I do thermal noise removal before calibration, as I always do …
Apply orbit -> remove border noise -> thermal noise removal -> calibration -> terrain correction
I do get descent output just like you show, but it’s when you zoom in into one of the water bodies that you see the issue of missing pixels in the VH channel. That’s the issue I’ve been talking about since months now, and nobody seems able to really get me an answer.
To be clear once again,
Yes, I have the latest version of the toolbox (6.0.3)
Yes, I do the steps in proper order
Yes, I terrain correct directly in EPSG32631
-> the issue starts after thermal noise removal, when some of these dark pixels become 0, the no-data value.
Really, I cannot provide more detailed information than this.
Thanks.
we are looking into this problem again, sorry for the delay
Another month and 10 days have passed …
Any updates?
And if you do calibration first and drop the border-noise removal…same issue? Or changing to nearest neighbor resampling? Or adding to a very small value to all pixels that are zero by using band-math? There are often many possible workarounds for a given issue…
I am also facing the problem now.
Exporting the image from SNAP, pixels outside the mask is set to 0.
We are having the same issues.
Who at ESA is responsible for this? I’m sorry to say but in my experience both S1 and S2 processing are a complete mess.
I’ve been hunting for a proper fix for many months now. @mengdahl I’m not sure if a workaround like “adding very small value to all pixels” is a scientifically sound way to get around this issue. The other options I need to try out.
Anyways, it remains ridiculous to keep “0” as the no-data value.
In the very low signal areas the pixel values contain only thermal noise instead of backscatter from the target. One can handle these pixels as you like in post-processing. Unfortunately before a fix is out one has to use workarounds.
I’ll try to get an update for this released later today. The problem is with the calibration of the noise floor values introduced by noise removal.
Great Luis, does that mean the problem manifested only if thermal noise removal was done before calibration?
Kristof, could you please try again with the latest 6.0.4 update
Thanks
6.0.4 seems to solve it for one of the scenes where we had the problem.
thank you so much.
Hi, it seems like 6.0.4 is tagged in github, but not available as a maven version:
http://nexus.senbox.net/nexus/content/repositories/releases/org/esa/s1tbx/s1tbx-op-calibration/
Would it be possible to release this version? Or is there another way to obtain it?
Is there also a way to get a Snap installer package with this version? We have a number of machines to update with this patch, and would like to do it in a headless manner.
Thanks!
It seems the deployment to the maven repository was forgotten. I did this now.
Regarding the update in headless mode. It is possible. Please see here: Update SNAP from the command line
An updated installer is not available.
@lveci, @mengdahl, @jdries, @marpet,
I can finally confirm that the long-standing issue mentioned in this thread has been solved. As can be clearly seen in the image below, the original VH image (lower panel) contained these no-data pixels with 0 dB, while the same pixels in the upper panel, after updating to the latest toolbox version, do not contain these no data pixels anymore. Thanks to the developers for fixing this issue.
Yes from 6.0.4 onward it’s possible to place thermal noise removal either before or after calibration, while earlier only the order calibration -> thermal noise removal worked.
Yes, it is possible.
I’ve linked to the documentation page in my post.
Update SNAP from the command line
