Slight shift when subset / reproject some S3 WFR products using snappy

Optical Toolbox
#1

Hi all,

I’m using snappy to apply in batch subset and reproject functions on Sentinel3 L2 WFR products. Sometimes, this works as expected and results match perfectly well the land mask in my AOI. But some other times, there is a slight shift (not always in the same direction) between the mask and the result.

This first result works well.
Image ID : S3A_OL_2_WFR____20180802T101253_20180802T101553_20180803T172429_0179_034_122_2160_MAR_O_NT_002.SEN3

image
image1109×814 82.3 KB

This one is shifted to the SW.
Image ID : S3A_OL_2_WFR____20180805T103519_20180805T103819_20180806T180342_0179_034_165_2160_MAR_O_NT_002.SEN3

image
image999×820 80 KB

And this one is shifted to the NE.
Image ID : S3A_OL_2_WFR____20180806T100908_20180806T101208_20180807T183411_0179_034_179_2160_MAR_O_NT_002.SEN3

image
image1031×817 83.3 KB

Here are the codes I use for subset and reproject
Sentinel3_reproject.py (2.4 KB)
Sentinel3_subset.py (2.6 KB)

At first, I thought this might be related to the distance to nadir of my AOI within each data. But looking at the whole images, it doesn’t seem to be the case. Is there maybe an option in the reproject function that should be triggered to avoid these slight shifts ?

Thanks for your help!

Olivier

#2

Which type of geo-coding do you use?
If you switch to pixel-based geo-coding the result might be more accurate, but reprojection will take longer.
See here:
What’s the difference between tie-point-based and pixel-based geo-coding for Sentinel-3 data?

2 Likes
#3

Hello Marco,

You’re right, the pixel-based geo-coding option was switched off. Don’t really know why I turned this off at some point.
Many thanks for your help, that solved the problem.
Cheers
Olivier