Weird striped/blocky patterns in processed Rws and Flagbit 2048

Post Reply
jan.elkassar
Posts: 3
Joined: Sun Sep 30, 2018 6:34 pm
company / institution: Institute of Space Sciences - Freie Universität Berlin

Weird striped/blocky patterns in processed Rws and Flagbit 2048

Post by jan.elkassar »

Hello everyone, hello Francois,
me and a colleague are working with Polymer to perform atmospheric corrections on radiances measured by OLCI on Sentinel 3a and b. So far we are quite happy with the results. However, by looking at some scenes we found that there are weird patterns in the reflectances (i.e. Rw400) and subsequently the "Inconsistency" Flag (2048) sometimes. I guess something similar was found and addressed by the user vishnu in the topic "What is Bitmask 2112" (viewtopic.php?f=7&t=74).

You can see them in the attached images (left: RGB from radiances, middle: Rw400, right: Rw-Spectra for the 4 pins). The scene is from the Arctic, captured on the 22.06.2020 and the Level1 radiances were processed with Polymer version 3.14. The landmask was set to "none", however having a landmask or not did not influence the stripe pattern.
polymer_stripes_example.png
In other scenes the pattern is not so pronounced but still visible. It is also remarkable how different the spectra from the two regions look. Do you think the data are still ok to use? Is there a possible explanation for this behaviour? Maybe I overlooked something in the manual? Have others also observed these patterns?

Thanks,
Jan
You do not have the required permissions to view the files attached to this post.
User avatar
fsteinmetz
Site Admin
Posts: 306
Joined: Fri Sep 07, 2018 1:34 pm
company / institution: Hygeos
Location: Lille, France
Contact:

Re: Weird striped/blocky patterns in processed Rws and Flagbit 2048

Post by fsteinmetz »

Dear Jan,
Thanks for your feedback. Indeed in this case the area appearing in red does has unstable results, similar to what user vishnu identified. Work is underway to try and reduce this effect, but these are complex cases with very turbid waters (Mackenzie river). So it seems that the flag 2048 should be applied as well to remove these cases, I will investigate this but in the meantime I advice you reject pixels according to "bitmask & 1023+2048 != 0" instead of "bitmask & 1023 != 0".
These stripy patterns appear because the result of each pixel is reused as an initialization for next pixel, in the vertical direction. In case of unstability this leads to vertical patterns.
Could you send product name for this image ?
Thanks,
François
jan.elkassar
Posts: 3
Joined: Sun Sep 30, 2018 6:34 pm
company / institution: Institute of Space Sciences - Freie Universität Berlin

Re: Weird striped/blocky patterns in processed Rws and Flagbit 2048

Post by jan.elkassar »

Dear François,
thank you for your quick reply and the info and explanation provided!

I will send you the exact scene name in a private message. I can also provide you links to the processed data if you want.

Kind regards,
Jan
Post Reply