I have some Sentinel satellite imagery.

And I have noticed that it looks like there are different locations of nodata values in each channel's mask. (I read in the tif file with 3 channels in via the rasterio python library)

This seems VERY strange to me. Shouldn't the nodata masks be the exact same across channels?

This file is a Sentinel 2 Infrared Ratio band composite image (3 channels) (the yellow dots are the nodata pixels):

ch1 ch2 ch3

I don't have a background in satellite imagery / GIS though, so perhaps this actually is normal? Could someone explain to me how this could be possible.

Or for satellite data, should I only consider a pixel as nodata if ALL 3 bands have the nodata value? Because at those locations where one channels show nodata and others don't, the values at all 3 channels are for example [3, 100, 255] where 255 is nodata.



It will actually depend on what NO DATA values mean; are they representing missing values during acquisition? Or, as the production of these images requires many processing steps, erroneous values may occur during some of them?

Keep also in mind that the 3 channels must be registered after acquisition, hence you can imagine them are initially almost independent devices, where a lot of complexity is involved (...) in a way. The registration operation is known as "geometric calibration". It is normally follow by a validation step.

Incoming light path on the MSI module

Mirrors positions and alignment

Feel free to read this document from where the two previous images come from:

p. 22 "Level-­‐1 Geometric Quality Requirements"
pp 47-58 "Level--1 Geometric Calibration"
pp 79-88 "Level--1 Geometric Validation"

On the 5 products of the Sentinel-2 mission, namely


only the two last are released to the public.

Detailed explanation

You can find some information regarding NO DATA value in the latest ESA Mission Performance Center (MPC) Data Quality Report, for example for Sentinel-2 L1C product:

Sentinel-2 calibration report - 1


Sentinel-2 calibration report - 1


You may also be interested if this anomaly search engine:

Same report, but for the L2A product:
https://sentinel.esa.int/documents/247904/685211/Sentinel-2-L2A-Data-Quality-Report (there is also some mentions on NO DATA values but not as details as for the L1C product, but it's worth a read).

Other related information:

General information, by topic:

Product types:

Product L1C (description):

Product L2A (description):


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.