Hi,
I am doing some comparison between NDVI from EO browser platform and from calculation offline starting from the band rasters downloaded by https://scihub.copernicus.eu/dhus/#/home (both with L2A level)
I have tried also the custom panel in EO platform with the instructions of this post (NDVI values from SentinelHub different from the ones calculates in Qgis) among others, but I keep getting different values (the differences are not negligible), in particolar higher values in EO browser with respect to the offline calculation.
Could you help me please?
Thanks in advance,
Alessia
1 Like
Hi @goffi ,
It is highly possible that the NDVI imagery you downloaded from EO Browser is the value for visualisation. To extract the actual NDVI values using EO Browser, please follow the instruction given by my colleague here.
1 Like
Hi,
thanks for the fast answer, but my problem is upstream, because the band values do not match between EO browser and SciHub. I am referring to this image S2A_MSIL2A_20220806T100611_N0400_R022_T32TQR_20220806T162658: the digital number and the reflectance (with float32) of the 2 bands (B4 and B8) of the 2 sources are very different (I only expected a factor of 10000 instead). I attach also a screen of values of the raw data of EO browser and SciHub respectively
I would like to understand why you get different values (without any processing).
Thank you very much
1 Like
Hi @goffi and @nitin.lakshmanan ,
The difference could come from the reprojection and resampling.
When downloading imagery with EO Browser, there are limited options for the coordinate reference system (Fig 1). I tested with a random area inside the extent of T32TQR tile from SciHub on EO Browser and it only allowed me to download imagery in EPSG:3857, EPSG:4326 and UTM33N. The animation (Fig 2) below shows the rotation effect caused by the reprojection.
Fig 1
Fig 2
It is also difficult to align the pixels in the imagery downloaded from EO Browser to the original tile, meaning that the imagery download from EO Browser has been resampled. If you select the nearest neighbour for resampling, the difference should be reduced.
Last but not least, an offset is implemented with a band-dependent constant, i.e. BOA_ADD_OFFSET, according to the new processing baseline. The offset can be found in the MTD_MSIL2A.xml file. In your case the formula to covert Reflectance to DN is DN = Reflectance * 10000 + 1000.
EO Browser is designed for visualisation, which is not ideal for getting the source data. If you’re interested in getting the source data, Requests Builder is a better option as it gives you more control on the request. Please see the documentation for more details.
1 Like
Thank you very much for the useful answer!
Is there the possibility to directly download the reflectance instead of the digital number via API on the Scihub site?
1 Like
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.