I am facing a new issue in our SLSTR processing pipeline, used to generate daily cloud masks.
Snap is up to date, version 9.0.4, on Ubuntu 18.04 LTS.
The pipeline consist in the operators read, subset, reproject, rad2refl, write.
Unfortunately, it seems that the combination of those generates a geographic shift in the final output.
From my tests the following combination of processes work fine:
Currently I’m not aware of such an issue.
I’m ordering the product you have used. But it is an offline product. So, it takes a moment.
Will get back to you when I know more.
I was able to reproduce the shift and I have an idea what the problem is.
Thank you for the good issue report. It helped to identify the issue.
The root cause the SLSTR format and that it has several different geo-location information.
But I don’t want to take this as excuse. We handle it at some points not correctly.
For the each of the different views there is separate geo-information available.
In addition, there is geo-location information based on tie-points for the whole scene.
Unfortunately, these are not very accurate.
For some reason, the tie-points have been used in your error case.
This issue needs to be further investigated: [SIIITBX-434] Incorrectly the tie-point geo-locations are used for SLSTR data - JIRA
You can work around this problem by specifying in the read node the special SLSTR-format.
Add this tag to the read node <formatName>Sen3_SLSTRL1B_500m</formatName>
This ensures a consistent geo-coding.
Attached is an updated version of your graph file. working_read_subset_reproject_rad2refl.xml (3.4 KB)
