Georeferencing data issues


this is more a question about sentinel-3 SLSTR data as such than about the toolbox (I haven’t used that yet. So far, I just downloaded netcdf data from EUMETSAT’s coda and played around with it in python). Nevertheless, I thought, this might be the best place to get an answer…

I am trying to compare data over a specific region over time (ie derive some kind of (mini-) time series). to keep things least complicated, I chose two scenes with identical relative orbit and frame numbers and from successive orbit cycles, and i am limiting it to infrared nadir data (ie the * files).

comparing the associated elevation data (from pixel-by-pixel, I realised larger differences than I expected - ok, might happen for 1km-x-1km pixels in mountainous regions.
to be on the safe side, I started checking the lat/lon as well as the cartesian coordinate data.
pixel-by-pixel comparison of lat/lon data looked not perfect, but ok (my quick&dirty single-point geofit yielded a one pixel shift).

however, doing the same for cartesian coordinates showed pixel-by-pixel location differences of around 16km, mostly due to systematic(!) differences in the y-coordinate (the x-coordinate had extremes of about +/-1.7km). For safety, I checked the latitude corrdinates once again (as the ones that should correspond to the y-coordinates. in my understanding of geocoordinates at least.), and they show by no means similarly big differences between the two scenes.

So, I wonder what is going on. whether I just make one or several mistakes in thinking here. or whether there are some issues with the data.

Thanks for any hints!

ps. the two datasets I have used in the above are

I’m sorry I can’t really help with this. With SNAP I haven’t seen shifts yet in the data. But honestly, I didn’t dive much into it.
As this seems to be specific to the data you could use the contact form on the Eumetsat webpage. Or try the data from the copernicus openhub. Your products are also present there, just a different processing time and centre.

Hi Marco,

thanks for the feedback and suggestions!

I now also tried the data from the Copernicus Openhub. They (or, at least the cartesian and geodetic nc file contents) are identical to the ones I derived from coda, ie - if there’s no reasoning error from the start on my side - they have the same georeferencing issues.
(But thanks for referring to the Copernicus Hub. Got S2 data from there before and had tried to get the S3 data from there initially as well, but couldn’t find any. After your suggestion, I had a look again and finally recognized that S3 has its own hub there :man_facepalming:. So, additional learning outcome! :+1: Thanks!)

As the issue is hence not limited to Sentinel-3 data from coda/EUMETSAT, I think, EUMETSAT is not the correct contact point for complaints. I’ll contact instead.

Thanks & Best wishes,

1 Like