Issues with loading ICON - daily data

[nchawang]

Greetings ESMValtool community!

So i recently came across this issue when working with daily ICON model outputs, other before that working with hourly, no issues.

When setting my config and recipe to accordingly for daily model outputs files, the esmvaltool ran showing no error. However, the preproc and plots were empty.

I proceeded to check if it was an issue with the recipe I had set, I went back to a simpler preprocessor and diagnostic (extract 500hPA level and spatial plot) checked for both mip: day and mip: Eday when loading the dataset, however alas, I came accross the same thing (no errors, but preproc nc file and plot empty).

I then checked with hourly (mip: E1hr) and monthly (mip: Amon) with the same recipe (changing the user-config and mips to point correctly) which worked. Is something wrong with how i'm writing the recipe for daily icon outputs?

I have attached a zip files which contains the log, log_debug, plot, recipe and nc file for all four cases (same preprocessor and diagnostic):

• mip: E1hr hourly outputs : No issues
• mip: Eday daily outputs : No errors shown - but Emply Plot and preproc file
• mip: day daily outputs : No errors shown - but Emply Plot and preproc file
• mip: Amon montly outputs : No issues

daily-plot.zip

Any help in this, if I've made an error or mistake is much appreciated!

[schlunma]

Hi @nchawang, your recipe looks fine! My guess is that the daily input data is somehow different than the other types of data, but this is really hard to diagnose from a distance.

One thing that might cause this is the regridding. Could you maybe repeat your analysis with a different scheme (i.e., scheme: linear or scheme: nearest) and check if that fixes the plot?

[nchawang]

Hi @schlunma ,
Thankyou for your reply!

I tried changing the regridding scheme and this was what i got

• scheme: linear - No errors raised but nc and preproc file empty
• scheme: nearest - Errors raised

2025-04-08 08:39:04,610 UTC [4040811] ERROR   Failed to run preprocessor function 'regrid' on the data
[<iris 'Cube' of air_temperature / (K) (time: 6; first spatial index for variables stored on an unstructured grid: 81920)>]
loaded from original input file(s)
[LocalFile('/scratch/b/b383276/dailyoutput_chem_var/esmval/AR_CH_RAD/AR_CH_RAD_orig-met2_20240604T000000Z.nc'),
 LocalFile('/scratch/b/b383276/dailyoutput_chem_var/esmval/AR_CH_RAD/AR_CH_RAD_orig-met2_20240605T000000Z.nc'),
 LocalFile('/scratch/b/b383276/dailyoutput_chem_var/esmval/AR_CH_RAD/AR_CH_RAD_orig-met2_20240606T000000Z.nc'),
 LocalFile('/scratch/b/b383276/dailyoutput_chem_var/esmval/AR_CH_RAD/AR_CH_RAD_orig-met2_20240607T000000Z.nc')]
(and 3 further file(s) not shown here; refer to the debug log for a full list)
with function argument(s)
target_grid = '1x1',
scheme = 'nearest'

iris.exceptions.CoordinateNotFoundError: "Expected to find exactly 1 coordinate matching the given 'latitude' coordinate's metadata, but found none."
2025-04-08 08:39:05,276 UTC [4040811] INFO

• I also tried commenting the regrid entirely (diagnostic will fail but preprocessor should run) after it flagged the error for the diagnostic but preproc ran, when I checked the preproc file (in mesh 2d face), all values are missing.
• When I did the same for hourly (diagnostic failed but preproc ran, the variable ta had values when checking the preproc file.

The last two tests does make it look like an input data issue but the model run configuration is exactly the same but for a longer time (1 month) and outputted daily. Also, the yearly run - outputted monthly works just fine; same for the hourly outputted 5 day run).

I've attached the zip files with the details for all 4 cases,
regrid-change.zip

[schlunma]

Hmm, interesting. Then it seems this is not related to regridding. Could you try commenting out the extract_levels preprocessor and then check if the preproc file looks okay? You could also try using different interpolation schemes, e.g., scheme: linear_extrapolate, see https://docs.esmvaltool.org/projects/ESMValCore/en/latest/recipe/preprocessor.html#schemes-for-vertical-interpolation-and-extrapolation.

[nchawang]

I tried what you suggested and It's at extract_levels!

• first commented it and checked the ncfile, and it had values.
• I changed the scheme to linear_extrapolate, and it worked no problems! the preproc file and plots came out no issues. yay!
• checked the nearest : did not work (no errors but empty plot) and nearest_extrapolate: worked

This is surprising, I've check the icon namelists between hourly, monthly and daily output again, and see no difference other then time step outputs and start and end times. I'm wondering if it's related to pfull and phalf values, could you help me point to how ESMValcore is handling the ICON vertical interpolation/extraction?

[schlunma]

Great that it works now! I am not sure though why extrapolation should be necessary to extract the 500 hPa level and why it works for hourly and monthly data...

Here is the code that simply adds the pfull and (if possible) phalf values as (3D) coordinate to the dataset. The actual interpolation to 1D pressure levels is then done in the extract_levels preprocessor here.

[nchawang]

Ah, okay! thanks @schlunma for helping out ! I'll continue to try to look into this, If there is anything I find, I'll update here.