[sen2cor 2.11.0][linux/wsl] Strange bright areas after correction

Hello!

I use sen2cor 2.11.0 with bash (Ubuntu 22.04) through wsl.
For L1C product – S2A_MSIL1C_20170815T015651_N0205_R060_T54WXT_20200906T110412.SAFE

Use following batch command with attached GIPP_custom.xml:

for f in /mnt/s/02_Sentinel/*L1C*SAFE; 
do "/home/den/Sen2Cor-02.11.00-Linux64/bin/L2A_Process" 
--resolution={20,10} --processing_baseline "02.05" 
--output_dir "/mnt/s/den/02_Sentinel/" 
--GIP_L2A "/home/den/sen2cor/2.11/cfg/L2A_GIPP_custom.xml" $f; done

L2A_GIPP_custom.xml (8.8 KB)
As I understand, the processing complete successfully, attached log file.
S2A_MSIL2A_20170815T015651_N0205_R060_T54WXT_20230831T144649.SAFE_20230831T144648_report.xml (104.2 KB)

But after AC there are a lot of strange bright areas.


Rock shadow looks like the sun is straight over the scene, but it isn’t.

I’m not experienced user of sen2cor, but could it be a problem with calculation sun angles or dem height?

Some advices how to fix it and is it possible to fix?

Hello @dench

I can’t be certain, but it might be that Sen2Cor has identified some pixels as semi-transparent clouds. Is there anything in the L1C mask?

In these instances, the terrain correction is not applied, and thus artifacts can become apparent.

Cheers

Jan

Hello @Jan

Thank you for reply!

There is nothing in the L1C mask.

Maybe the reason is there are no AUX_CAMS* files?

Hello @dench

The CAMS and ECMWF data was included in PB04.00. Introduced in late January 2022. So your product wouldn’t include it. I would ask @Sen2cor_dev_team if they can help you.

Cheers

Jan

1 Like

Dear @dench,

Thank you for your ticket with all the necessary information provided. The @Sen2cor_dev_team was able to reproduce the strange bright areas you observed. As mentionned by @Jan, those effects are sometimes related to the L2A Scene Classification classes like topographic shadows or clouds on which no terrain correction is performed. However investigations are on-going to find the source of this effect. Maybe it could be related to some DEM characteristics in this area but not precise answer is available at that time.

The @Sen2cor_dev_team will continue its investigations and post them here.

Cheers,
@Sen2cor_dev_team

1 Like

Hi everyone, I think I have a worse example of the same issue, this time with Copernicus products.

The L1Cs look fine, but the L2As seem over-corrected, especially in the N0500 version, see below.

S2B_MSIL1C_20200102T111349_N0208_R137_T30TUN_20200102T114218.SAFE

S2B_MSIL1C_20200102T111349_N0500_R137_T30TUN_20230425T042636.SAFE

S2B_MSIL2A_20200102T111349_N0213_R137_T30TUN_20200102T122616.SAFE

S2B_MSIL2A_20200102T111349_N0500_R137_T30TUN_20230425T201605.SAFE

Cross-posted on https://helpcenter.dataspace.copernicus.eu/hc/en-gb/community/posts/17714274473245-Strange-artifacts-in-Sentinel-2-L2A-products.