Different results with tiles from one scene


#1

Hello there!
I’m trying to play with sen2cor with 2 Sentinel-2 tiles (S2A_OPER_MSI_L1C_TL_SGS__20160724T085956_20160724T123647_A005681_T37VDF_N02_04_01 and S2A_OPER_MSI_L1C_TL_SGS__20160724T085956_20160724T123647_A005681_T37VEF_N02_04_01). Both tiles have been downloaded from EarthExplorer.
The only thing that these tiles underwent is sen2cor processing with 10 m resolution.
Probably there is something that I’m missing, but I expected these tiles to be corrected to the same result, but instead I got something like you can see at screenshot.

Could someone, please, tell me if this is OK or some preprocessing/postprocessing needs to be done before/after sen2cor.
My aim is to monitor and classify some plants but they appear to be different on these tiles.
Thanks in advance.


#2

Hello!

I think you should check the pixel values in the overlapping area. Sometimes the the images appear with different brightness and contrast depending on the contrast strech applied by the software you are using to visualize them, even though they might have the same pixel values.
If your pixel values are, indeed, different, then you have the same problem as me and I hope somebody can fix it!

The thing with the Sen2Cor process is that each tile is processed separately, and the calculations for the atmospheric correction are done based on the values of each tile. They have an overlap, but the tiles themselves are different. Consequently, the calculations are different for each one.

Sorry I don’t have a solution!


Sen2Cor - pixel values of overlapping area
#3

Hi Pilar!
These are the pixel values in the overlapping area. The values are different, especially in blue channel.


So, I guess we both have the same problem and it’s unsolvable.
Have you come up with a solution of how to get a consistent results with different tiles without sen2cor? My interest is in detecting plants and it’s vital to have a good and consistent results after atmo correction, otherwise the trained classifier will produce waste.


#4

Hello,

I know this post is a little bit old, but is there still no solution/explanation how to solve this kind of problem?


#5

No, it is inherent to the atmospheric correction of sen2cor. No way around. Sorry


#6

But some other atmospheric correction software can do better (even if not perfect too !).
See for instance MAJA: http://www.cesbio.ups-tlse.fr/multitemp/?p=6203

Here is for instance a time series of cloud free monthly syntheses with no visible artefact along tile borders


#7

@OHagolle Hello, can MAJA work in windows system? And how do you think about start-maja


#8

Hi dorbodwolf,
Sorry, but MAJA is for Linux only. Start_maja is a script to help you use MAJA (which is not easy, much less than sen2cor, even if several users managed).

If your zone is not among those we process, the easiest to have a try with Maja is to use the possibility offered by the French collaborative ground segment, PEPS, as explained in http://www.cesbio.ups-tlse.fr/multitemp/?p=14335
You can either use the browser interface, or use a python script. The implementation of MAJA in PEPS is however still a beta version, regularly improved.
Olivier


#9

Hi Olivier,

I will try these and give feedback.

Best
Deyu Tian