OLCI level 1 raw data geo-offset along GreatLakes?

Optical Toolbox
#1

Hi I have this issue for a while and been posted on another post and EarthData forum. but still not solved after trying like pixelGeoCoding options tick on/off.
This geolocation offset not consistently happen to all scenes, we notice sometimes the vector boundary aligns with the OLCI level 1 imagery very well, esp in summer months.
could this really caused by raw data, or further software processing issues like tie-point, projection, etc?

thanks,
Chui

----------------details ------------------

I work on Sentinel-3 OLCI level 1 images of the Great Lakes, and notice some raw images (in SNAP, load the xxxx.SEN3/xfdumanifest.xml) seems have geolocation offset when comparing with higher resolution images sensed a few days apart.
the following are two examples from my test at early June 2017. both show offset along the south shore of the peninsula/islands.

site 1 (Lake Ontario)
vector shapefile reference:

image
image1280×686 100 KB

OLCI level 1 RGB composition (S3A_OL_1_EFR____20170611T152704_20170611T153004_20180502T143746_0179_018_339_2160_LR2_R_NT_002
):

image
image1280×694 77.6 KB

NDWI [ (Oa08_radiance - Oa17_radiance)/ (Oa08_radiance + Oa17_radiance) ]:

image
image1280×694 217 KB

you can see in the NDWI map above, the south shore has 1-2 pixel offset comparing to the lake boundary shapefile.

site 2 (Lake Superior)
Landsat-8 true color reference align well with the vector boundary file:
image

OLCI level 1:

image
image1280×694 77.5 KB

OLCI NDWI :

image
image1280×694 217 KB

all file above in my google drive.
here are the data products I used for the test:
OLCI products: (download from ESA:
S3A_OL_1_EFR____20170613T161541_20170613T161841_20180502T203026_0179_018_368_2160_LR2_R_NT_002
S3A_OL_1_EFR____20170611T152704_20170611T153004_20180502T143746_0179_018_339_2160_LR2_R_NT_002
Landsat-8 products: (download from Earthexplorer:
LC08_L1TP_025026_20170609_20200903_02_T1
LC08_L1TP_016029_20170610_20200903_02_T1

Sentinel 3 image offset/shift
#2

Hi Chui Zeng, could you check the google drive link. I have only access to one png image. In addition, could you please add the shoreline file you have used.

Further, did you use the shoreline file (a shapefile I assume) to overlay it in SNAP over the S3 file? If that’s not the case, which shoreline file did you use?

Best regards,
Jan

#3

Hello @chqzeng,

my colleague @JanWev and I had a look at this.
We think that your observations are within the normal accuracy range which is possible for OLCI.
There can be several reasons the shift

  • OLCI mission requirement for geolocation accuracy: 0.5 pixel with an RMS of 0.3 pixel
    Half an OLCI pixel is roughly equal to five L8 pixels. ( S3 Product Notice – S3.PN-OLCI-L1.09)
  • Compared to the high resolution of L8, OLCI contains mixed pixels at the coast. Land/water can often not clearly be differentiated.
  • The products of 2017 you use are processed with an older processing baseline. It is likely that more recent data products will have a higher accuracy.

best wishes