Bad reprojection after running l2gen on OLCI L1B

[andrew.meredith]

Hi,

I installed the latest OCSSW tag (v2026.0) on a linux VM with RHEL8. Reprojecting an OLCI L2 file created by l2gen using the SNAP (v12.0.1) "gpt Reproject" operator gives bad results. Output from same sequence of commands with tag v2025.5 are as expected. Also, direct "gpt Reproject" of the L1B is also as expected.

Here's the commands:

Code: Select all

l2gen l2prod="rhos_560 rhos_620 rhos_665" proc_land  ifile=/data/OLCI/ESA_L1_data/S3A_OL_1_EFR____20260329T153332_20260329T153632_20260329T172342_0179_137_339_2520_MAR_O_NR_004.SEN3/xfdumanifest.xml ofile=S3A_OL_1_EFR____20260329T153332_nosubset.L2.nc 

/opt/snap/bin/gpt -Ds3tbx.reader.olci.pixelGeoCoding=true beam_repro_seadas_l2.xml -t S3A_OL_1_EFR____20260329T153332_nosubset.L2.FL3_seadas.nc -f NetCDF4-CF -Pfile=S3A_OL_1_EFR____20260329T153332_nosubset.L2.nc

Andrew

[OB.DAAC - amscott]

Hello, it appears the issue occurred after using SnapGPT, which is maintained outside of OB.DAAC. Please post your question to the ESA STEP Forum: https://forum.step.esa.int/

[andrew.meredith]

Thanks for the response.

I posted on the SeaDAS forum because the reprojected results differ depending on the l2gen version. The reprojected product generated from a L2 file created with l2gen (msl12 9.11.0-V2025.5) is fine. The reprojected product generated from a L2 file created with l2gen (msl12 10.0.0-V2026.0) is not. I get similar results reprojecting the L2s using the SeaDAS GUI (v9.2) as i do with gpt. Does that still point to a SNAP related issue?

Andrew

[OB.DAAC-EDL - amscott]

Thanks for the additional information. Let me pass this along to a few more people to see if the issue can be replicated.

[OB SeaDAS - knowles]

Are the missing chunks in the level-2 file which you generated or are they happening in the reprojected file?

[andrew.meredith]

I do not seeing any obvious problems with the L2 files when viewing in SeaDAS. The missing chunks show up after reprojecting.

[andrew.meredith]

One difference I see in the L2 from V2026.0 is that the latitude and longitude bands contains no data values whereas the L2 from V2025.5 do not.

Top row show products from V2026 and bottom row V2025.5