Recently ESA announced processing software upgrade for sentinel 2 imagery Announcement Link
Some snippets from the announcement
Data produced by the updated Processing Baseline 04.00 is expected to be available at the Copernicus Australasia Regional Data Hub in mid January 2022. We recommend users carefully review the changes as there is a high risk they could affect processing streams and the creation of downstream products.
We draw your attention to the ‘Provision of negative radiometric values (implementing an offset)’. The new offsets are being applied to both L1C and L2A imagery, so any use of either product must take account of this.
We do consume sentinel 2 imagery. Is there something which need to be handled from our end or the new processing is already being applied on the supplied products?
Also, are there any other concerns which need to be addressed while consuming sentinel 2 imagery going onwards?
Hi @gunjan,
it depends how you are accessing the data:
if you are using Sentinel Hub, then the new data is harmonized to the old data (by default), whereas you can choose to get original value, if you want so (more here)
if you are accessing the data directly in AWS S3, then the values have essentially changed and you need to handle this by yourself, The data on AWS S3 are exactly the same as the one produces by ESA/Copernicus.
What do you mean by - “values have essentially changed and you need to handle this by yourself” - will level-2 surface reflectance values be different before and after the change? will it affect let’s say NDVI values?
Also, will the change be backpropagated for past images as well?
Yes indeed, surface reflectance is changed. Not sure how exactly this will impact NDVI, would have to do a few experiments.
There is no plan to propagate this to past images, so you need to handle this dynamically.
As mentioned, if you are using Sentinel Hub, you should not notice a difference.
Hi @gmilcinski - I’m trying to figure out how to handle manually the changes in Sentinel-2 Level-2A.
Correct me if I’m wrong but according to this doc- in order to get L2A BOA reflectance values, I should use this formula:
L2A_BOAi = (L2A_DNi + BOA_ADD_OFFSETi) / QUANTIFICATION_VALUEi
To make sure these values BOA_ADD_OFFSET and QUANTIFICATION_VALUE appear in the
MTD_MSIL2A.xml:
So if I download Sentinel-2 Level-2A and if the value of a specific pixel is L2A_DNi=1452, then the BOA reflectance should be converted to - (1452-1000)/10000 = 0.0452?
Also, are BOA_ADD_OFFSET and QUANTIFICATION_VALUE constant or they might change?
@gmilcinski any updates on how this change affects the NDVI? I’m trying to correct the values for all bands according to the formula above but the resulting NDVI values seem higher than expected (compared with the eo-browser).
