Alignment with Sentinel 2 grid

I am querying Sentinel 2 Data with a bound box (link to attribute in Python SDK) in WGS 84 (Lat/Lon, EPSG4326) coordinate system. The bound bo

I need to do that as I need to provide the data to another application in EPSG 4326.

What I see is that Sentinelhub is modifiying the pixel grid so that it fits into the queried bound box. So basically it seems to recalculate the data from the original grid to a grid fitting exactly into the bound box.

My question is how this is been done and whether I need to fear inaccuracies because of that?

Page 1 / 1

Hi Victor,

There have already been a couple of posts that might be relevant to your question:

Different statistics results between sentinel hub statistical API to numpy and rasterstats results

Retrieving the sentinel L1C 20m grid - #2 by batic

It is quite a complex question to answer, naturally the pixel grid must be modified if the data is outputted in EPSG:4326 (WGS84) because the native projection of the data is in the local UTM zone.

If you still have some questions surrounding this topic, then please let us know!

Thanks @william.ray for your reply. I have checked the posts but they did not answer my questions.

Sentinelhub apparently is modifying the grid when the request is executed with a Boundbox in EPSG:4326. And still my question is if I can rely on this modification

Yes, there is no reason why should not be able to rely on the modification of the grid. There are also several options on how the data is resampled too using the API detailed here.

I suggest that you make some comparisons between the original data and the data outputted by the Sentinel Hub API. If you find any discrepancies, then do post that in the thread and we can help explain them.

Ok I could not find bigger unexpected discrepancies.

In any case there are cases where it would make sense to be able to query the original data.

Due to that I propose to add a method to the Geometry class to get a BBox aligned with the grid of the different satellite constellations (most important Sentinel 2 of course .