Understanding the Sentinel 3 cloud mask band


I have opened the “cloud_in” band from a Sentinel 3 acquisition in Snap. Each pixel in the mask seems to be from a small set of integer values, leading me to believe that the mask needs to be decoded in order to extract quantitative information about the cloud probability for that pixel. However I have not been able to find a resource online that explains how to perform the decoding of the mask values.

Could someone point me toward a document explaining what the values of the cloud mask represent please?

Ultimately I would like to import or generate a cloud probability mask in python. So a document containing explicit instructions of transforming the “cloud_in” pixel value to the probability that a cloud is present at that pixel would be desirable.

Many thanks,

Subsequent to this post, I found this discussion: https://forum.step.esa.int/t/cloud-mask-for-slstr-lavel1b-image/8458/3

I still have not found a resource for explicitly spelling out how to extract the probability of cloud cover for a given pixel but the above link was helpful for getting me closer to that goal.

Hi David,

There is some information about the cloud masks including Bayesian cloud mask probabilities for SLSTR in the handbook here: https://www.eumetsat.int/website/wcm/idc/idcplg?IdcService=GET_FILE&dDocName=PDF_DMT_921927&RevisionSelectionMethod=LatestReleased&Rendition=Web

If you need more detail - you can also drop me a query via ops@eumetsat.int and I can help further.



I am working with Sentinel 3 SL_2_WST products .
For cloud cover I would need to understand the differences between the different bands present in the product, and how these were obtained:.

  • l2p_flags_cloud;.
  • quality_level;.
  • probability_cloud_in;
  • probability_cloud_io.

My aim is to use the most real cloud cover possible, especially without losing information as much as possible in coastal areas.

Thank you