L1C-to_L2A Processing Problems for Northern Territories

Optical Toolbox sen2cor
#1

Hi, Everyone!
Can anybody help me understanding what’s wrong with data processing?
Sample scene is S2A_MSIL1C_20160920T044652_N0204_R076_T49WEU_20160920T044755.SAFE

  1. Region is Polar;
  2. Date is in between summer and winter;
  3. Terrain is almost flat.
    Source data views are (Bands 1 & 2):

L1C_Band1_View
L1C_Band1_View.jpg1398×704 427 KB
L1C_Band2_View
L1C_Band2_View.jpg1398×704 443 KB

Data after processing are (Bands 1 & 2):

L2A_Band1_View
L2A_Band1_View.png1398×704 64.6 KB
L2A_Band2_View
L2A_Band2_View.png1398×704 236 KB

Changing winter/summer parameters doesn’t affect the result. As well as other atmospheric correction parameters.

What’s wrong?

By the way, I found that S2Cor v.2.8 generates error when trying to switch off wv correction parameters …

Thanks!

#2

In general SNAP 6 has bug with sen2cor 2.8 according to the Source of the post

But I applied it from the cmd, and successfully implemented and got the results,

image
image.png1104×638 30.9 KB

image
image.jpg1918×916 445 KB

I also applied sen2cor 2.5 to your product and I got same results as you have,

image
image.jpg1905×934 531 KB

As you see applying both sen2cor 2.5 and 2.8 don’t change the result, I’m curious now to know the reason,

The reason of the datastrip might be explained in here,

image
image.png1075×395 53.5 KB

image
image.png1042×183 28.3 KB

Source: https://sentinel.esa.int/documents/247904/685211/Sentinel-2_User_Handbook

But I also now wants to know the reason of this result,

#3

Effect of No-Calibration probably doesn’t depend on territory, date of sensing, orbit, etc.
S2A_MSIL1C_20160920T044652_N0204_R076_T49WEV_20160920T044755.SAFE - Yes, Processed
S2A_MSIL1C_20160920T044652_N0204_R076_T49WEU_20160920T044755.SAFE - Not Processed

Another series of granules with the same effect:
S2A_MSIL1C_20160920T044652_N0204_R076_T49WFV_20160920T044654.SAFE - Yes
S2A_MSIL1C_20160920T044652_N0204_R076_T49WFV_20160920T044755.SAFE - Yes
S2A_MSIL1C_20160920T044652_N0204_R076_T49WFU_20160920T044755.SAFE - No

Example of source bands of the series mentioned above:

Shot1
Shot1.jpg1398×905 507 KB

All granules look similar by statistics …

Probably, Bands 1 & 2 degraded only.

#4

Hi, Everyone!
I think I found the reason for data degradation of scenes from Northern territories.

  1. The latitude of scenes is higher 70 degrees North.
  2. Dates of scenes I use, are of deep autmn (September)
  3. Time of sensing is morning
    In fact, that time is almost nightfall …
    As the result, Source L1C data in both B01 and B02 bands have low intensities, and quite narrow dynamic range. Sure, low intensities are in all bands, but only B01 and B02 degraded after Sen2Cor processing.

Band 01 L1C Source View:

B01_L1C_SRC
B01_L1C_SRC1346×888 496 KB

See the full range of B01 is 0.0001-0.216 and actual data range is 0.14-0.18

The result after Sen2Cor B01 Processing View:

B01_L2A_Default
B01_L2A_Default1346×888 236 KB

See the histogram. About half part has been cut.

To resolve the problem I multiplied both B01 and B02 source bands by 2. Just to increase intensities and broaden the dynamic range.
The result of Sen2Cor B01 processing after multiplyin source data:

B01_L2A_Mult_by_2
B01_L2A_Mult_by_21346×888 588 KB

As you can see, data is not degraded.
I think it is the not right way to escape of data degradation.
But I think that the reason is low L1C source data intensities.
And probably either some round operations or some integer operations occur within Sen2Cor processing flow.
I hope it can be corrected, if so …