Colour anomalies in Sentinel 2A products

[OrbitalEos]

Hi everyone

I am trying to work with atmospherically corrected imagery of Sentinel 2. When I load the image and select an RGB (B4,B3, B2) natural color I am getting strange outputs. I show you an example:

What is the reason behind that issue? Can it be fixed?

Regards

[Jan]

@OrbitalEos
Hi Juan,
It’s not clear to me what you see as “strange outputs”. Are you refering to the variance between detectors over water?

Cheers

Jan

S2 MPC CC Operations Manager

[OrbitalEos]

Hi Jan

Thank you for your answer. I notice oversaturated unnatural colors at land and sea. I have been working with other 2A products and I always had more natural realistic colors when generating RGB composites. Another issue, as you mention, is the radiometric difference among detectors. Is it a problem of detector coefficients? Is there a software to normalize that radiometric difference among detectors?

Thank you

[Jan]

Hi Juan,

I believe (@marpet may confirm) that the view of your RGB is a consequence of the handling of the data by the software. I would suggest that that the high percentage of sea in the visible bands leads to the land being rendered brightly.

Regarding the detector/detector difference, this is a spectal response non-uniformity, and is highlighted in Section 5.1 of the monthly Data Quality Report (DQR) produced by the S2 MPC and available from the Sentinel-2 MSI Document Library.

Cheers

Jan

[marpet]

Yes, it is very likely that the different display is cause by a different distribution of bright and dark pixels in the histogram.
Unfortunately, the images are scene specific and can not easily be compared.
Good constant value ranges for the RGB images are
Red: 0 - 0.35
Green: 0 - 0.30
Blue: 0 - 0.25

If you want to do this automatically you could follow the description I gave recently:

Sentinel 3 olci RGB unit interval

This is not easily doable. But you can use the expressions to do it with pconvert. red = Oa09_radiance < 0 ? 0 : Oa09_radiance > 430 ? 430 : Oa09_radiance This limits the values between 0 and 430. For the blue and green channel do it equivalent. As additional option to pconvert use the -s option to disable the skip histogram values. With -s 0,0 nothing will be skipped. I thinks this will do what you want.

[OrbitalEos]

Thank you both for that information. Limits for RGB values worked quite well, and then adding brightness to the blue channel in order to see better ocean areas ended up in a reasonably natural color. I know that Sentinel 2 was not conceived for Ocean monitoring, but it´s high spatial resolution makes it really useful for certain applications. I know there are some issues with atmospheric correction for ocean applications. Concerning the best available atmospheric correction for ocean monitoring, I have been suggested to test Polymer. What would be your advise about that?

Kind regards

[marpet]

You could try the C2RCC processor. It does water constituents retrieval and an AC over water.