[matschaer]

Adds values to the default unstructured grid dimension names (GRID_PROCESS__data_model__grid__unstructured_dim): ["cell", "points", "values"]. cfgrib and earthkit-data name the dimension of non-regular GRIB grids values (e.g. ICON open data), which until now had to be renamed before interpolation (as in example notebook 04: .rename_dims(values="cell")). values is last, so cell or points still win if a dataset has several of these dimensions.

No clash with the .values attribute of xarray objects. grid-process accesses dimensions by name only (ds[name], dicts, **{dim: ...}), never as attributes, and ds.values / da.values stay the xarray method and numpy property. Checked:

• Synthetic data: all 13 methods of interp2poi and interp2grid give identical results with cell and values, both without a coordinate (as cfgrib gives it) and with an index coordinate.
• Real data: ICON-CH2 T_2M from open data ({'values': 283876}, read with earthkit) gives identical results without renaming for:
  • interp2poi with nearest, idw and linear;
  • interp2grid with nearest and idw to LV95 at 1 km and 5 km.

No breaking change. The only risk: a regular grid that also has an unrelated dimension named values would be taken as unstructured (the unstructured names are checked before y/x). cfgrib does not produce this, and the environment variable still overrides the list.

Example notebook 04 no longer renames the ICON-CH2 dimension values to cell, and is rerun with a new IFS-HRES run (2026-11-12 06 UTC) and ICON-CH2 run (2026-11-11 18 UTC).

Related issue

None.

---

Checklist

• Code is well documented and follows project conventions (linted, formatted)
• CHANGELOG updated (if applicable)
• License header is added to all new files (no new files)
• For a new contributors, their name has been added to the AUTHORS file (not a new contributor)
• No sensitive data or secrets are exposed
• Unit tests added covering the new code (test_get_flat_dim: values detected without a coordinate; test_interp2poi: results with values identical to cell; 262 tests pass, ruff and mypy clean, pylint 10.00)