GEDI height comparison - gridding problems

[JessicaNeedham]

I'm trying to add vegetation height to the diagnostics package using data from GEDI (downloaded from https://www.earthdata.nasa.gov/data/catalog/ornl-cloud-gedi-highquality-shots-rasters-2339-1). rh-95-a0 is a proxy for canopy height. The relevant FATES output is FATES_MEAN_95PCTILE_HEIGHT - added in this PR NorESMhub/fates#40.

The observations come as a tiff file which I converted to a netcdf with minimal processing in this script: https://github.com/JessicaNeedham/GEDI_FATES_comparisons/blob/main/gedi_tiff_to_netcdf_native_res.ipynb

I added the netcdf to the data repository /datalake/NS9560K/diagnostics/ILAMB-Data/DATA/vegheight/GEDI and updated the ilamb.cfg file here: https://github.com/JessicaNeedham/xesmf_clm_fates_diagnostic/blob/85c4351abc500497691932c0ac402cdc1906e854/tests/test-data/ilamb_CLMFATES.cfg#L77

When I try to run the diagnostics package with this json file (https://github.com/JessicaNeedham/xesmf_clm_fates_diagnostic/blob/gedi/scripts/test_gedi.json) it hangs (currently for more than an hour).

I thought maybe the native resolution was too big, so I tried regridding it to the same resolution as a FATES ne16 run outside of the diagnostics package using reproject_match from the rioxarray library (script here: https://github.com/JessicaNeedham/GEDI_FATES_comparisons/blob/main/gedi_tiff_to_netcdf_regrid.ipynb). This obviously isn't ideal but I wanted to see the results :)

When I point to this file I get the following results:

Something clearly up with the coastlines. The model and observations should be the exact same resolution already so this is a bit weird.

If I borrow some of the plotting functions and plot the comparison outside of the diagnostics package it looks ok

[maritsandstad]

To me it looks like it might not be the coastlines, but that all longitude values are off by some constant factor in how the model interprets the observational data...

[maritsandstad]

Also, PSA that a notebook is not a script...

[JessicaNeedham]

To me it looks like it might not be the coastlines, but that all longitude values are off by some constant factor in how the model interprets the observational data...

Interesting. I will continue to investigate.

[maritsandstad]

Happy to have a more proper look at the data tomorrow.

[JessicaNeedham]

If I replace this line

xesmf_clm_fates_diagnostic/src/xesmf_clm_fates_diagnostic/ilamb_configurations.py

Line 186 in 6ac63d2

output = regridder(outd_gn)

with output = outd_gn then the plot looks correct. So something with the regridder is causing the problem. The lat dimension actually gets smaller by one value because of this line:

xesmf_clm_fates_diagnostic/src/xesmf_clm_fates_diagnostic/plotting_methods.py

Line 234 in 6ac63d2

regrid_target = regrid_target.isel(lat=slice(lat_min, lat_max))

but that doesn't seem to be the problem. If I make that regrid_target = regrid_target.isel(lat=slice(lat_min, (lat_max+1))), I get the same result. The lat and lon dimensions seem identical before and after regridding, so I'm not sure why it suddenly plots differently.

[maritsandstad]

Hmm, interesting, I will closer look. Do you have a test dataset and a test-command I can just reuse @JessicaNeedham ?

[JessicaNeedham]

python run_diagnostic_full_from_terminal.py '/datalake/NS9188K/share/jessien/noresm-fates_ne16_transient_1900_borealseeds.2025-11-19/run' outpath=/datalake/NS9188K/share/jessien/noresm-fates_ne16_transient_1900_borealseeds.2025-11-19/figs weight='/datalake/NS9560K/diagnostics/land_xesmf_diag_data/map_ne16pg3_to_1.9x2.5_nomask_scripgrids_c250425.nc' pamfile=test_gedi.json mute_maps=True mute_trend=True

[JessicaNeedham]

The json file and ilamb.cfg are on my gedi branch https://github.com/JessicaNeedham/xesmf_clm_fates_diagnostic/tree/gedi

[JessicaNeedham]

There are two datasets in the DATA repo vegheight_native_res.nc is the data at native resolution - just converted from tiff to netcdf. vegheight_ne16.nc is the one I regridded to 2 degrees. I just updated ilamb.cfg to point to the 2 degree version on my branch.

[maritsandstad]

... and your branch is...?

[maritsandstad]

Aha, I see now

[maritsandstad]

Did you see the same problem with the original resolution?

[maritsandstad]

Ok, I don't actually know what the problem is, but I have implemented a fix that works for the regridded data. The native resolution just hangs at least for now. Will push some changes somewhere later.

[maritsandstad]

Ok, my fix on top of your stuff is here #71 @JessicaNeedham

[JessicaNeedham]

Ok, my fix on top of your stuff is here #71 @JessicaNeedham

Thanks @maritsandstad !