L2W mis-alignment/projection issue

Discussion of your results and ACOLITE in general
Post Reply
stcarp
Posts: 3
Joined: Tue Nov 17, 2020 3:04 pm

L2W mis-alignment/projection issue

Post by stcarp »

Quinten,

Thanks for making this fantastic tool available. I am having issues with the L2W output from S2 files using Linux. The output appears to be misaligned to the original S2 product. This issue does not occur when there is a merge of two images from different tiles on the same date. I have tried projecting to the correct UTM zone, but with no success. See example
misalgined.JPG
misalgined.JPG (76.08 KiB) Viewed 9716 times
Many thanks,
Stephen
quinten
Posts: 1019
Joined: Tue Mar 03, 2015 8:13 am

Re: L2W mis-alignment/projection issue

Post by quinten »

Hi Stephen

Could you provide your settings files for the results with and without misalignment? Which output files are you using, the NetCDF files?

Are you using the binary version or GitHub code?

Quinten
stcarp
Posts: 3
Joined: Tue Nov 17, 2020 3:04 pm

Re: L2W mis-alignment/projection issue

Post by stcarp »

I am using the binary version outputting netcdf files. The settings file with misalignment is below, the other is with merged files.
exemplar_acolite_settings.txt
(2.48 KiB) Downloaded 586 times

acolite_settings_merged.txt
(2.56 KiB) Downloaded 562 times
quinten
Posts: 1019
Joined: Tue Mar 03, 2015 8:13 am

Re: L2W mis-alignment/projection issue

Post by quinten »

Dear Stephen

I have looked at this issue with the GitHub code and found that the computed pixel coordinates match with the original file - both x/y in metres or lon/lat in degrees. There were two small issues in the GitHub code - none which cause the magnitude of shift shown in your example - that are now hopefully resolved: 1) to save disk space the lat/lon datasets were stored as single precision floats and this caused some very small differences at the sixth decimal place, and 2) there was a 1/4 pixel offset (in Y) in some configurations of the Sentinel-2 processing.

This doesn't help you directly since you are using the binary release, of which I hope to do one in 2021, bringing it in line with the GitHub code. I suggest however you give running the source code a try.

I do have a suspicion that the issues you are finding may have another origin. How did you compare the alignments? Are you performing another conversion step on the NetCDF file, perhaps in importing it in a GIS? Did you try the GeoTIFF outputs from ACOLITE? They are probably better supported by other software packages and should also match the input projection.

Hope this helps!

Quinten
Post Reply