search

Element84 / earth-search

Earth Search information and issue tracking
https://earth-search.aws.element84.com/v1
27 stars 2 forks source link

S2 duplicate results from 2021, but not in 2022 #25

Closed Kongstad closed 8 months ago

Kongstad commented 8 months ago

Hi,

I am trying to understand a weird issue here and was hoping you could help me figure it out. I appreciate your time.

I have been downloading S2 images over Denmark which are subsequently cut down to farm field polygons from August 2021 to April 2022. I have noticed that there seem to be double data occurances for images captured in 2021, but not in 2022. Below I have attached an example of some field level images: e841

Not only are there double occurances, but there also seem to be differences between the “duplicates”. Some occur darker/brighter and for the snow covered images one seems to be more blue. I have double checked on Sentinel-hub EO browser to understand the tiles and the fields do reside inside an area of 4 tiles overlapping as seen in the image below: e842 But if that was the cause of the problem, there should be 4 occurances and it still does not explain the issue of why it happens in 2021 and not in 2022. When I check Sentinel-hub again for the February 28th occurance, it looks identical: e843 So according to Sentinel-hub the data should be similar.

I then did a second experiment where I compare E84 S2 data with data from Microsoft Planetary Computer and they do not seem to have the same duplicate data. See image below with E84 on the left and MSPC on the right: e84_mspc

What am I missing here? Is it an E84 API issue? Is there a general change in data between 2021/2022 that just is not reflected differently on Sentinel-hub?

Appreciate any input. Thank you for your time.

gadomski commented 8 months ago

Hello, thanks for your question. I don't know exactly why the L2 scenes are duplicated, but we are currently reprocessing all Sentinel-2 scenes -- see this email for more information. It's likely that the problem will be fixed by the reprocessing. I'd suggest waiting until the reprocessing is complete, then re-running your experiment -- if the issue persists after reprocessing, we can dig in deeper then.

Thanks for the issue report! For more information on the reprocessing, you can sign up for our mailing list here.

Kongstad commented 8 months ago

Thank you for this information. Very helpful and I have signed up to your news letter.