Sentinel 2 assets available in Google Earth Engine

Hi Msiklar,
I am afraid you misunderstood my conclusion.

Hi guys,

you have an interesting conversation here.
But there is still no answer on the original question and the last reply was November 2017.

@simonf: do you finally got your “broader consensus” after almost one year?
I’m really intrested in using Level2A data in GEE. Could you please give us an update?

Is there a possibility, at least , to apply an atmospheric correction on an image collection of S2-L1C in GEE?
Would be great to hear from you.

All the best, Florian

Sorry, still no news (also, it’s not possible to run classic processing to SR in EE). But we are well aware of the demand for SR, and continue to investigate options.

Did you check the possibility to use Py6S for atmospheric correction?
The following github project seems to use GEE and Py6S…

Don’t know whether I’d call that consensus, but at least ESA decided to use Sen2Cor for producing the official S2A product, and this product is now also available from the Senitnel data hubs. So probably EE doesn’t need to do any processing itself, but store & provide the L2A the way it does for L1C (unless the huge amount of data - L2A being roughly as big as L1C, ie doubling the required storage - is a problem).

The problem with the L2A data in the sentinel hub is, that not all of the L1C are corrected. There is a pre-selection by ESA/Copernicus (?!). Unfortunately I didn’t get an answer how they select which image is worth to correct and which image is not. But a huge amount of L1C data are not corrected.

See the discussion here:

1 Like

Hi guys
it is good to watch your conversation, it is already 2019 now, I wonder if there is an answer about the SR product in GEE, anything new? also, if we can’t use SR product lunch by GEE, is there an alternative option for us to generate our own SR product on GEE?

Just started ingestion into COPERNICUS/S2_SR. Will add to the EE catalog shortly.

Dear Simon,
this is awesome!!!
I’m really looking forward to use S2 L2A data in GEE! Finally!

Do you already know when the data are accessible for the public?

All the best


The ingested assets are already available in EE, we just have not yet added a catalog entry or ingested everything that we downloaded. It’ll take a few weeks to catch up. A copy on Cloud Storage is also available, it’s just not documented yet.

S2 SR is now in the EE catalog:

I found a little mistake in the GEE documentation. Where the bands are described it is said that the value range is 1 to 11 for the SCL band. But actually it is 0 to 11. 0 is no-data.

Marco - this is actually intentional. I used the value of 0 as the fill value to mask out such pixels in the SCL band. Do you think this would cause issues?

Probably not directly issues, but maybe confusion. As it caused for me.
So you have masked these pixels already. But I’m not sure if I understand this correctly. What does it mean in this case you have masked them out? What happens if one wants to get the value for such a pixel? Which value will he get? Or is it not possible to access such pixels?

I can add a note in the docs that 0 is masked out.

Each EE pixel retrieval operation returns a tuple (pixel_value, mask_value). If the pixel is fully present, mask=1. If the pixel is fully masked out, mask=0. If the pixel is partially masked out (which happens in pyramided layers that combine fully present and fully masked-out pixels at lower levels), 0 < mask < 1.

1 Like

Okay, now I understand. Makes sense.
But a note might be helpful.

It is possible to learn which version of the Sen2Cor processor has been used to generate the 2A level data?

I see that metadata state that the processing baseline is 02.11
Does it mean that Sen2Cor 2.11 has been used? The newest stand alone version is 2.55…

We just download the assets generated by ESA. I think we put all the relevant metadata onto the assets, but if not, let me know.

No one knows with which version of sen2cor the data has been processed. It is almost impossible to figure it out from the metadata shipped with the L2A products.
The latest internal sen2cor version is 2.8, as far as I know. The release of it was planned one month ago. But obviously, there is a delay for some reason.

You might want to adjust the “Dataset availability” ( as L2A data from ESA are only available since mid March 2017 (Europe) and mid December 2018 (global).
Or do you plan to reprocess old data as well?