Open dionhaefner opened 5 years ago
Sorry to reply so late, your message arrived during my holiday period and I did not notice it. We have never met this issue so far.
However, there seems to be something weird with the L1C data. The "angle zones" refer to the detectors (a line of S2 detectors has 12 linear detectors) which have participated to the observation of the tile. It is strange to see that an image was observed with detectors [0,3,4,5], as the numbers should be contiguous , probably [2,3,4,5].
I'll chek that.
Did you process with start_maja ? Could you please provide the command line ? Best regards, Olivier
/opt/maja/core/1.0/bin/maja -i /maja-work-root/work_20180724_5c01cdb8-15b9-44bf-a636-b3325459940b -o /maja-out -m L2NOMINAL -ucs /maja-work-root/userconf --TileId 32UNF
Hello Dion,
Sorry you have to wait so long. The error seems to indicate a configuration issue between the the footprint detectors of the product, who are used to compute miniluts, and the angles for each detector, which are available in the metadatas of the product header. Could-you provide us with these files so we can keep on investigating? Thanks for your answer.
Best regards, Bruno
Sure. Which files exactly do you need from the SAFE?
Hi,
again quite a delay, but this time we've got some news: Can you please show us the contents MASKS folder of the Level-2 product just before the date where it failed? I had some time to investigate this week and I am suspecting that the DETFOO masks are missing. This seems to occur irregularly and we don't know yet where this is coming from - It could be a faulty download for example.
Sorry again for the delay!
Kind regards, Peter
Sorry, I'm not quite following - where should I find the MASKS folder? In the MAJA .DBL.DIR
directories?
If you mean the original SAFE, that one does contain the DETFOO files.
To clarify, the MAJA processing chain rolls on happily if I remove this one product, so I doubt it has much to do with the products before or after.
Yes Dion, we would need the DETFOO from the original SAFE. We have found the same issue with another product where some parts of the DETFOO files were incorrect
Thanks, Olivier
Sure. Here's the whole QI_DATA
folder of this product.
Hi, is there any news on that issue ? We still have this problem. At least, is there a way to detect these wrong images before running MAJA ? Thank you very much, Best regards
Dear @olivierhagolle ,
we got the same error, i.e.,
2021-03-20T23:41:46.781548 ImageListWriter 4.3 [000000026659] [E] vns::Business::ERROR: ComputeScatteringCorrectionImageFilter(0x1645cb0): For band id '0' (nb channel='4') ' no miniLUT has been generated for this angle zone (detector) : '2'. Angles zones are [0,3,4,5,6,7,8,] !!!,Check coherency between metadata and input zone mask source. [vnsComputeScatteringCorrectionImageFilter.txx:ThreadedGenerateData:252]
from a MAJA run. What we did find out is that there were two L1C scenes provided by ESA for this date (T34SEJ: 20190703), containing different parts of the same scene:
Hope this may give you an insight for these cases.
Christina
Hi Christina, MAJA handles the case with images cut in two parts (at least, it processes one of the images), but the issue you have found is due to ill-formated Sentinel-2 images, that lack one of the information MAJA needs. Our only solution so far is to remove them frm the stack of images to be processed. Best regards, Olivier
Interesting. This might actually be why we were getting this issue in the first place - the split products might have issues with the metadata for the view angles etc. Those angles are given in the metadata XML files on a 5 km grid with slight variations for each sensor and maybe the production sometimes does not get them right on these products.
From my experience, the same tile published in two parts is actually not a rare glitch, but a pretty frequent pattern. I can see you also discussed this elsewhere a while back @olivierhagolle: https://forum.step.esa.int/t/sentinel-2-weeding-out-bad-products/15041
So, does MAJA stitch the two together or only process one of them? I guess it all depends on the input script that does the bookkeeping and compiling of the lists of input files? And then you need to do the spatial merging afterwards...
I think when I discovered this issue for the first time, I also tried to monkey-patch the products - merging the JPEG2000 files and trying to trick downstream software in believing these were the original products. But I was unable to get the file format and headers right with GDAL - another really nasty implementation detail in the Sentinel-2 product format...
I run MAJA on some S2 scenes, and all work well except this one (
S2A_MSIL1C_20180724T103021_N0206_R108_T32UNF_20180724T142006
). Do you have any idea what could be causing this error?Full output: