Unexplainable Fragments/Efects in S2 image

Hi Alexey,

Thanks. That’s a lot clearer. Having discussed this with colleagues, I can confirm this is not a data anomaly. In order to provide you with a definitive answer, I shall raise it to our expertise. This may take a few days.



Jan, thanks! It’s not data anomaly surely. I suppose some processing bug (feature). As I see on geometry - neighbor detectors have some overlap in data - so one may choose the manner how to compose them to result image.

Hi Alexey,

Indeed. It will be good to get some clarification on it. From better minds than mine :slight_smile:



I think that all bands are not recorded simultaneously. So the satellite records red band first, then green, blue, ultraviolet, and since it is moving very fast, the green band and blue band get the plane and his contrails from different perspectives.

Hi Wobi!
in any case I can not explain such distance between “airplanes” by document with optic geometry
I can only imagine if plain is at 10 000m height and blue/geen/red detectors are at edge/center/contrEdge of detector assably. But that’s not evident from document.
Other qwestion - what are UFO (in circle) without traces - may be helicopters (at the same height - ??? ) or some optic effect of instrument (or atmosphere mirage).

1 Like

Sentinel 2 altitude is 786 km, which roughly translates to a velocity of 7.5km/s relative to Earth surface.
For a plane at 10 km, that means that it is displaced, roughly, one pixel each second (10m) (ignoring the plane movement, which adds more pixels, and contrails moving with the wind).

It is evident that the plane is displaced more than one pixel, so recording each band must take many seconds.
Those must be small planes, at high speed. Probably military.

That effect is common, you can see plane color ghosts on Google maps/earth.

1 Like

Hi Cipa!

some comments to you. look a rough scheme (some suggestion - 10km is small when consider 800km.)
Displasment l is about 100 m. So you can see at image - distance between blue and red planes. And plane is not so small - pixel size of MSI - 10m. And we see plane as X, not as point.

PS And more over - if we suppose the plane is regular flight over 10 km. Then he has a velocity about 900 km/h = 250 m/s. So it should move far away ))) - so it seems to be simulteneous photo and color splitting of plane is due to detector geometry when we compose onground (surface) image. But the shadow of plane over earth should be only one )))

Hi Alexey,
I only read now your post, and I have to say I like it a lot! Quite puzzling indeed, since the plane (and contrail) effect on the left is predictable, the second plane image is more puzzling if you assume it comes from the same one. But if you consider there is actually a second plane which does not produce contrail (which is possible if not at the same altitude), the puzzle is solved!
To be investigated, I will retrieve the image.
Thanks a lot for sharing.
PS: I have published a post in Linkedin providing elements about such kind of effect. I will use yours as well to illustrate it… when understood! :wink:

Hi Olivier!

I visited linkedin site. I will continue here (it’s specific for Sentinel2). The break of plane trails I confirm with you. It’s due to near (odd\even) detector assamblies registrations. But first image with “color tripling” of plane to be leave surface with about 300 km/h velocity - I can’t explain. It seems to be cautch from the same detector assambly. Why there is time registration difference for red-green-blue spectra I can’t imagine when read documents. So I ask for some more detailed optic geometry, timing. If one have to public.


PS the weak mirror of plane (mirage) - author of topic also asked at top ))). I suppose optic effect… but can not approve. Only one idea - more weak images and just the same geometry…

Hi Alexey,

For sure the current forum is the correct place for such kind of discussion, I agree.

Regarding the slight shift in time between bands and detectors, please refer to: https://sentinel.esa.int/web/sentinel/technical-guides/sentinel-2-msi/msi-instrument

The difference of B/H across bands and across detectors explains why you have a shift between the bands and also the discontinuity as well as inversion of contrails colour combination between consecutive detectors, as illustrated below:

Regarding the image originally posted, my conclusion is that we have two different planes for the three following reasons:

  1. the distance between the RGB components of the planes is different from one to the other, which means that they have a different speed and/or they are at different altitude,
  2. their directions are different,
  3. one has contrails not the other which means they are at different altitude (water vapour content variable in the atmosphere column).

Note finally that this discontinuity between high altitude targets across adjacent detectors also impacts clouds imaging since registration between detectors is performed at ground altitude.

I hope this helps.


Hi Marvin,

Could you please communicate on the acquisition, e.g.:

  • sensing time
  • orbit
  • tile
  • location

I am really interested in retrieving the same acquisition, images showing two planes in the same time are not so frequent.
See also my other comments, links and illustrations.

Thanks and regards,

See also the Data Quality Report with the following link: https://sentinels.copernicus.eu/web/sentinel/missions/sentinel-2/data-quality-report

Hi Olivier!

A agree with you on the interdetector shift (break) in trails (arrow 1 below)

and I suppose that color split is due to sensors displacement inside detector (arrow 2).
But there is no details about it in the document (or I can’t clear for myself). Do you have optic registration path for every band? Does detector is a set of one line (row) of sensors
for every band or it is like brush (for example like modis). We may modify my scheme above to make it easier for understanding color shift. Who may rise a picture with detector assambly? )))

What is B/H ? bold/height? )))


Hi Jan!
now I have a reply from S2 support on my qwestion. Here it is:

Dear Alexey,

Currently the boundaries between detectors are defined independently for each spectral band of Sentinel-2 imagery.
While this configuration optimizes images on-ground, it leads to colouring artefacts over clouds like the one identified by yourself. A trade study between these two competing effects is currently under way and could lead to a redefinition of the detector boundaries.

Best regards,

1 Like

Thanks! That is interesting.