SCL in tile overlap not consistent

Optical Toolbox sen2cor
1

I’ve seen Sen2Cor - pixel values of overlapping area and the related threads but they were about the pixel values of the bands.

I have the same problem but in the SCL band after L2 processing. After reading the ATBD I don’t see why there would be such a big difference in the SCL layer but maybe I’m misunderstanding something.

The 2 images I’m looking at are:

  1. S2A_MSIL2A_20170525T112121_N0205_R037_T29SND_20170525T112434
  2. S2A_MSIL2A_20170525T112121_N0205_R037_T29TNE_20170525T112434

SCL of 29TNE:

image
image.png1543×611 90.9 KB

SCL of 29SND:

image
image.png1544×679 75 KB

The colorbar is the same as https://earth.esa.int/web/sentinel/technical-guides/sentinel-2-msi/level-2a/algorithm in but unclassified is orange

image

I’m happy to provide the raw data if required.

2

Dear cpaulik,

are you sure that both L2A data have the same processing baseline? It might be possible that they have different baselines and therefore different versions of sen2cor were used to generate the scene classification. I have downloaded the current L2A of the T29SND tile (S2A_MSIL2A_20170525T112121_N0205_R037_T29SND_20170525T112434) and the scene classification looks exactly like your example of T29TNE.

Cheers,
Jan

3

Dear @JanWev ,

I should have probably put that in my original description.

I have processed the scenes myself using Sen2Cor 2.05.05 . I can do that again to make sure that nothing went wrong in our automated processing chain.

4

I’ve downloaded the level 2 scene for 29SND from scihub and it is closer to my scene of 29TNE but it is not the same.

I’ve also reprocessed the scene locally to make sure everything went OK and I get the same result.

So it seems that Sen2Cor 2.05.05 produces non-consistent results. I will retry with an older Sen2Cor version to check that.