Conservative regridding fails for CMIP6 ICON-ESM-LR data in recipe ``ref/recipe_ref_cre.yml``

[bouweandela]

REF run https://dashboard.climate-ref.org/executions/4003/logs fails with:

2026-03-07 13:30:58,080 UTC [1023542] ERROR   Failed to run preprocessor function 'regrid' on the data
[<iris 'Cube' of TOA Longwave Cloud Radiative Effect / (W m-2) (time: 228; cell index along first dimension: 20480)>]
loaded from original input file(s)
[LocalFile('/mnt/data/climate-ref/2026-03-04/scratch/esmvaltool/cloud-radiative-effects/82b7f05ec92f65df822a28007ba89d94b1e8d0e8/climate_data/CMIP6/CMIP/MPI-M/ICON-ESM-LR/historical/r1i1p1f1/Amon/rlut/gn/v20210215/rlut_Amon_ICON-ESM-LR_historical_r1i1p1f1_gn_199001-199912.nc'),
 LocalFile('/mnt/data/climate-ref/2026-03-04/scratch/esmvaltool/cloud-radiative-effects/82b7f05ec92f65df822a28007ba89d94b1e8d0e8/climate_data/CMIP6/CMIP/MPI-M/ICON-ESM-LR/historical/r1i1p1f1/Amon/rlut/gn/v20210215/rlut_Amon_ICON-ESM-LR_historical_r1i1p1f1_gn_200001-200912.nc'),
 LocalFile('/mnt/data/climate-ref/2026-03-04/scratch/esmvaltool/cloud-radiative-effects/82b7f05ec92f65df822a28007ba89d94b1e8d0e8/climate_data/CMIP6/CMIP/MPI-M/ICON-ESM-LR/historical/r1i1p1f1/Amon/rlut/gn/v20210215/rlut_Amon_ICON-ESM-LR_historical_r1i1p1f1_gn_201001-201412.nc'),
 LocalFile('/mnt/data/climate-ref/2026-03-04/scratch/esmvaltool/cloud-radiative-effects/82b7f05ec92f65df822a28007ba89d94b1e8d0e8/climate_data/CMIP6/CMIP/MPI-M/ICON-ESM-LR/historical/r1i1p1f1/Amon/rlutcs/gn/v20210215/rlutcs_Amon_ICON-ESM-LR_historical_r1i1p1f1_gn_199001-199912.nc')]
(and 2 further file(s) not shown here; refer to the debug log for a full list)
with function argument(s)
target_grid = '1x1',
scheme = {'reference': 'esmf_regrid.schemes:ESMFAreaWeighted'}

The run is using this recipe:

documentation:
  title: Model evaluation with focus on the cloud radiative effect.
  description: 'Plot climatologies and zonal mean profiles of cloud radiative effects
    (sw + lw) for the REF.

    '
  authors:
  - bock_lisa
  - lauer_axel
  maintainer:
  - bock_lisa
datasets:
- project: CMIP6
  activity: CMIP
  dataset: ICON-ESM-LR
  ensemble: r1i1p1f1
  institute: MPI-M
  exp: historical
  grid: gn
  mip: Amon
timerange_for_models:
  timerange: 1996/2014
preprocessors:
  full_climatology:
    climate_statistics:
      period: full
    regrid:
      target_grid: 1x1
      scheme:
        reference: esmf_regrid.schemes:ESMFAreaWeighted
  zonal_mean:
    climate_statistics:
      period: full
    regrid:
      scheme:
        reference: esmf_regrid.schemes:ESMFAreaWeighted
      target_grid: 1x1
    zonal_statistics:
      operator: mean
diagnostics:
  plot_maps:
    description: Plot lwcre climatology maps including reference datasets.
    variables:
      lwcre:
        timerange: 1996/2014
        mip: Amon
        preprocessor: full_climatology
        derive: true
      swcre:
        timerange: 1996/2014
        mip: Amon
        preprocessor: full_climatology
        derive: true
    additional_datasets:
    - dataset: CERES-EBAF
      project: OBS
      type: sat
      version: Ed4.2
      tier: 2
      start_year: 2000
      end_year: 2023
      reference_for_monitor_diags: true
    scripts:
      plot:
        script: monitor/multi_datasets.py
        plot_folder: '{plot_dir}'
        plot_filename: '{plot_type}_{real_name}_{dataset}_{mip}'
        plots:
          map:
            common_cbar: true
            x_pos_stats_avg: -0.1
            fontsize: 10
  plot_profiles:
    description: Plot lwcre zonal mean including reference datasets.
    variables:
      lwcre:
        timerange: 1996/2014
        mip: Amon
        preprocessor: zonal_mean
        derive: true
      swcre:
        timerange: 1996/2014
        mip: Amon
        preprocessor: zonal_mean
        derive: true
    additional_datasets:
    - dataset: CERES-EBAF
      project: OBS
      type: sat
      version: Ed4.2
      tier: 2
      start_year: 2000
      end_year: 2023
      reference_for_monitor_diags: true
    - dataset: ESACCI-CLOUD
      project: OBS
      type: sat
      version: AVHRR-AMPM-fv3.0
      tier: 2
      start_year: 1996
      end_year: 2016
    - dataset: ISCCP-FH
      project: OBS
      type: sat
      version: v0
      tier: 2
      start_year: 1996
      end_year: 2016
    scripts:
      plot:
        script: monitor/multi_datasets.py
        plot_folder: '{plot_dir}'
        plot_filename: '{plot_type}_{real_name}_{dataset}_{mip}'
        plots:
          variable_vs_lat:
            legend_kwargs:
              loc: upper right
            plot_kwargs:
              CERES-EBAF:
                color: black
              ESACCI-CLOUD:
                color: black
                linestyle: dashed
              ISCCP-FH:
                color: black
                linestyle: dotted

Possibly related to #751

[alioacar]

Hi, I came across this issue while looking through the repo. I think the problem is that the CMIP6 ICON-ESM-LR files do have the geometry that is needed for conservative regridding (vertices_latitude and vertices_longitude), but this is not turned into an Iris mesh when the file is loaded. So the cube ends up with mesh=None, and ESMFAreaWeighted seems to go through the rectilinear to rectilinear path, where it fails with TypeError: 1D X coordinate must be a DimCoord.
As a quick test, I manually used vertices_latitude and vertices_longitude as bounds for the lat/lon coordinates and then created a MeshXY from them. With that, conservative regridding worked. So it looks like the data contains the needed geometry, but the mesh is not constructed when loaded.
This looks like a dataset-specific fix. I am currently testing a guarded version that only does this in a very narrow case: unstructured icon-esm-lr files where the latitude/longitude coordinates point to valid bounds variables. If that works well, I can follow up with a PR.