[matschaer]

Adds four example notebooks to the documentation (new "Examples" page) illustrating grid_process.grid_interp.interp2poi and interp2grid with open data only. They replace the draft examples of the internal gridefix-process branch feature/sphinx-nb-display. Each notebook compares the interpolation methods visually and with a runtime benchmark table (first call vs. subsequent cached calls). On regular grids, nearest and linear are also run with try_reg=False (generic path) and compared with the default xarray path.

Notebook	Data → target	Methods
01 interp2poi wind gusts	ICON-CH2 control VMAX_10M (MeteoSwiss OGD, native unstructured grid) → SwissMetNet stations measuring gusts, vs. observed hourly max gust	linear, idw (k=4, weight_power=2), nearest_alt (r_search=1.7)
02 interp2poi temperature	IFS-HRES 0.1° 2t (ECMWF open data, oldest run, 0–144 h) → all SwissMetNet stations; 14 stations from the Plateau to the peaks plotted vs. 10-min observations	nearest, temp (z_fact=15, k=50, weight_power=0.75), temp_fixgrad (r_search=2)
03 interp2grid precipitation	ERA5-Land 2024 yearly precipitation (Copernicus CDS) → RhiresY 1 km LV95 grid, vs. RhiresY	nearest, idw (k=4, weight_power=2), linear
04 interp2grid temperature	IFS-HRES 0.1° and ICON-CH2 2t, same valid time → swisstopo DHM25/200 grid over Valais (496×833)	nearest, linear, idw (k=4, weight_power=2), temp_fixgrad

Details:

• Notebooks are rendered with myst-nb and nb_execution_mode = "off": they are committed with their outputs, since open forecast data expire within days and notebook 03 needs a CDS account.
• Their dependencies (earthkit-data with the CDS, ECMWF open data and MeteoSwiss OGD sources, earthkit-plots, eccodes-cosmo-resources-python, netcdf4, …) are in a new optional example dependency group (poetry install --with example), so CI does not install them. myst-nb is added to the doc group.
• The POI notebooks adapt the grid_process data model through environment variables (GRID_PROCESS__data_model__poi__coord_from, ..._alt_from) instead of renaming the SwissMetNet metadata columns.
• Model orography for the altitude-aware methods: HRES surface z at step 0 (IFS), HSURF from the horizontal constants (ICON-CH2); destination altitude from DHM25/200.
• Data are downloaded at runtime; no raw data is committed. Requests to the ECMWF open data portal are retried with backoff. Attribution is given in each notebook (MeteoSwiss, ECMWF and Copernicus under CC BY 4.0, swisstopo, Natural Earth).
• The notebooks do not modify the user's earthkit configuration; the Examples page explains how to enable the earthkit cache.
• The notebook code was checked once with ruff, mypy and pylint (converted with nbconvert --to script); no tooling was added for this.

No breaking changes.

---

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 — n/a: documentation files only (notebooks, rst)
• For a new contributors, their name has been added to the AUTHORS file — n/a
• No sensitive data or secrets are exposed
• Unit tests added covering the new code — n/a: no library code changed; docs build checked locally with sphinx-build -W, ruff format --check and pytest pass

🤖 Generated with Claude Code