Image 2021-05-05 (S2A_MSIL2A_20210505T105031_N0300_R051_T31TCF_20210505T154202.SAFE):
Image 2021-05-07 (S2B_MSIL1C_20210507T103619_N0300_R008_T31TCF_20210507T124404.SAFE):
There is such a big variation between both images with just 2 days of difference. This affects to products like NDVI. Both images were taken at the same hour. Why is this happening? Is there a way to fix it?
Is you did this in QGIS this is not SNAP or S2TBX related.
In such cases it is better to contact the data provider. If this is Open Access Hub (copernicus.eu) then you can mail to eosupport@copernicus.esa.int.
The top is level L2A while the bottom is L1C.
In my experience the L2A is always gives a brighter scene on the ground.
L2A is computed surface reflectance
rather than L1C’s top-of atmosphere radiance.
Good catch. I haven’t seen that one is atmospherically corrected and the other is not.
This of course explains the difference.
Hello
It should also be noted that one image is S2A, and one S2B. Although that will provide less of a change than the L1C vs L2A component identified by @gbrelstaff
It should also be noted that both the radiometry of S2A and S2B acquisitions is routinely monitored by expertise, and both S2A and S2B are within the specifications set down in the Mission Requirements. The radiometric (and geometric) status of the L1C from both S2A and S2B is provided to users via the L1C Data Quality Report and L2A Data Quality Report.
Cheers
Jan
S2 MPC Operations Manager
Thank you for your answers. I didn’t realize I was comparing L1C with L2A. Now I have done the right comparison but they are still very different:
2020-05-05 S2A L1C (S2A_MSIL1C_20210505T105031_N0300_R051_T31TCF_20210505T144122.SAFE):
2020-05-07 S2B L1C (S2B_MSIL1C_20210507T103619_N0300_R008_T31TCF_20210507T124404.SAFE):
2020-05-05 S2A L2A (S2A_MSIL2A_20210505T105031_N0300_R051_T31TCF_20210505T154202.SAFE)
2020-05-07 S2B L2A (S2B_MSIL1C_20210507T103619_N0300_R008_T31TCF_20210507T124404.SAFE)
If I compare the 1st vs 2nd photo and the 3rd vs 4th photo they are too different for just a timelapse of 2 days even though as @Jan commented one is S2A and the other S2B.
Anyways I’m going to have a look at the links that you shared and if I don’t find the problem I’ll send an email to eosupport@copernicus.esa.int.
This seems to be due to a temporary change in the Scene Classification on 5 of May at Illa de Buda.
If you look at the time sequence of the Scene Classification Map online here
and click the next date arrow - the blue (water) changes to Yellow and grey
click again and it goes back to the usual blue.
My guess is the artificial irrigation system on the ground broke down before 12:00 on 05 May 2021 - so the usual blue classification is lost and the usual L2A rendering got radically disturbed,
Hello @gbrelstaff
Interesting. Although the SCL changes, I note that the Moisture Index (20 metre Band ratio) is pretty static.
Cheers
Jan
This type of observed differences seems to be due to sunglint in certain images (Sun reflection on rough water surface or with some Fresnel reflection compound).
The sunglint is highly directional (and very sensitive to surface roughness linked to surface winds): so you might have very different sunglint amount in your image between S2A and S2B where the azimuth angles can be different over your scene (is it a kind of rice field?).
I provide here a small example for a lake (which also undergoes a whiting effect)!
it could also be that the wind changed and causes the differences in the data. To look at wind data of the dates might be worth it.
Yes, it is a rice field.
This is the information about the sensing time and the sun angles for both dates:
2021-05-05 S2A L1C/L2A:
Sensing time: 2021-05-05T10:59:39.620492Z
Zenith angle: 27.0852973971615
Azimuth angle: 151.766784678725
2021-05-07 S2B L1C/L2A:
Sensing time: 2021-05-07T10:49:42.736518Z
Zenith angle: 27.5280040740446
Azimuth angle: 146.577157482994
They seem pretty similiar.
Also, I checked wind data (https://www.wunderground.com/history/daily/es/reus/LERS/date/2021-5-5) and (https://www.wunderground.com/history/daily/es/reus/LERS/date/2021-5-7) and it doesn’t seem very different, although it is likely not very accurate.
If you are just checking possible causes, it would also be interesting to see if there was a precipitation event between both days. In the first image the soil looks drier than in the second one. Artificial irrigation could also be tested. But in my opinion, it looks that your second image has a higher load of aerosols in the atmosphere and the AC is not working perfectly (or in the first one there is an overcorrection).
Actually, you will have to check the relative azimuth:
view_azi - sun_azi
For your example, check RGB image, you can see that for tyhe “sunglinted” image (left) it is the right edge of the swath, and conversely for the other image on the right it is the left edge (sorry for the left/right ambiguities…):
Then you can check the sunglint signal over water with the SWIR bands -water totally absorbing, so your signal comes from surface reflection). This clearly shows that one image is impacted and not the other:
ANd we can now verify with relative azimuth (around 140deg for the sunglint impacted image), around -50 deg for the other:
All the best
Tristan
