Add combined NOAA GEFS 0.25 degree forecast and analysis virtual quickstart

[aldenks]

---

One quickstart covering noaa-gefs-forecast-10-day-0-25-degree-virtual, noaa-gefs-analysis-0-25-degree-virtual, noaa-gefs-forecast-16-day-0-5-degree-virtual, and noaa-gefs-forecast-35-day-0-5-degree-virtual, following the multi-dataset shape of ucsb-chc-chirps-analysis.ipynb, the temporary-open pattern of #78, and the AGENTS.md intro/question structure.

Temporary direct open

The datasets are not in the STAC catalog yet, so dynamical_catalog.open(...) cannot resolve them. Cell 2 is the only deviation: an open_virtual(dataset_id, **kwargs) helper opens each Icechunk store directly with anonymous S3 storage plus authorize_virtual_chunk_access for s3://noaa-gefs-pds/. The cell header comment carries complete replacement lines for the three used roots and the two group opens, using group="pressure_level" / group="height_above_mean_sea_level" and ending with the trailing forecast so the repr survives the swap. The unused 35-day root open was removed. The swapped sequence has no undefined names, checked with pyflakes over the concatenated cells with cell 2 replaced by its comment block. Install line stays dynamical-catalog cartopy; no new project dependency.

The four intro links point at the expected catalog pages:

• https://dynamical.org/catalog/noaa-gefs-forecast-10-day-0-25-degree-virtual/
• https://dynamical.org/catalog/noaa-gefs-analysis-0-25-degree-virtual/
• https://dynamical.org/catalog/noaa-gefs-forecast-16-day-0-5-degree-virtual/
• https://dynamical.org/catalog/noaa-gefs-forecast-35-day-0-5-degree-virtual/

What it shows and why

The original sections retain the 0.25-degree datasets' extra fields and 18 UTC cycles. The new sections show what the 0.5-degree products add: a month-long horizon, pressure-level fields, a 16-day tail, and fixed-height aviation fields. No event is reused from the materialized GEFS quickstarts, #78, another notebook on main, or an earlier section in this notebook.

Section	Dataset	Field	Event / question
Map	0.25-degree analysis	snow_thickness_surface (virtual-only)	Storm Filomena over Iberia, 2021-01-09T12
Time series	0.25-degree analysis	temperature_2m + soil_temperature_0_10cm (virtual-only)	February 2021 Texas cold wave at Dallas, 8-22 Feb
How hot did the ensemble expect the July 2022 heat to get?	0.25-degree forecast + analysis	maximum_temperature_2m, all 31 members of the 18 UTC run of 13 July vs analysis	UK 40 C record at Coningsby
Where did the ensemble put Hurricane Ian's winds?	0.25-degree forecast	wind_gust_surface, P(max gust on 28 Sep > 25 m/s), faceted over two 18 UTC runs	Ian landfall, 2022-09-28
Did a month-ahead forecast anticipate the January 2021 wind reversal?	0.5-degree 35-day pressure_level	10 hPa / 60 N zonal-mean wind_u, 31 members every 48 h	5 January 2021 sudden stratospheric warming; two members briefly turn easterly after the observed reversal
What did the long-range ensemble show before Portland's record heat?	0.5-degree 16-day root	temperature_2m, all 31 members at Portland every 24 h through 384 h	June 2021 Pacific Northwest heat wave; 240 h product limit marked
How strong were Storm Ciaran's winds above the Channel Islands?	0.5-degree 16-day height_above_mean_sea_level	member wind-speed profiles at all 8 heights	2023-11-02T00, initialized two days earlier
Animation	0.25-degree analysis	convective_available_potential_energy_surface (virtual-only), 3-hourly, 15 frames	26-27 April 2024 Plains outbreaks

All reads are bounded before .load(). The new SSW cell reads 18 leads x 31 members = 558 global pressure-level fields; the Portland plume reads 17 leads x 31 members = 527 fields; the Ciaran profile reads 2 wind components x 31 members x 8 heights = 496 fields. No second animation was added. The final notebook is 2,087,159 bytes (1.99 MiB).

Facts re-derived from the stores

The 0.25-degree forecast and analysis carry 34 surface/single-level variables at the beginning of the record, growing to 38 as NOAA added fields through October 2022. Each 0.5-degree forecast has 247 root variables (209 more than the 0.25-degree sets), 12 variables on 31 pressure levels, 6 variables on 4 model levels, and temperature plus two wind components on 8 fixed heights.

The 16-day root has 31 members and 105 leads: 3-hourly through 240 hours, then 6-hourly through 384 hours. The 35-day pressure-level group has 31 members and extends through 840 hours. The Portland selection was re-read from 16-day snapshot XKFE06FV2DCHJ8HW6SR0: the run initialized 13 June 2021 spans 18.3-42.5 C at the nearest grid point on 28 June, well beyond the 240-hour line.

How the 0.5-degree datasets slot in

• Intro: names all four products, distinguishes the 34-to-38 variable history of the 0.25-degree products from the 247-variable 0.5-degree roots, and includes one documentation bullet per dataset.
• Cell 2 (open): opens the three used roots plus pressure_level and height_above_mean_sea_level with keyword group=. The header's publication-swap block contains matching dynamical_catalog.open(...) lines for every used object.
• Structure note: states the measured variable/level counts, initialization cadence, and 3-hour/6-hour lead ranges for both new products.
• Sections: adds distinct examples for the 35-day pressure-level group, the 16-day forecast tail, and the fixed-height group. The 16-day section is a full 31-member plume and the fixed-height section preserves all members.
• Community challenge: includes open questions using pressure-level icing, analyzed soil moisture/heat fluxes, and fixed-height wind shear.

Checks (HEAD 6122113)

• uv run .internal/run_notebooks.py noaa-gefs-forecast-analysis-virtual.ipynb - passed in 158.97 s.
• uv run python tests/test_notebook_execution.py noaa-gefs-forecast-analysis-virtual.ipynb - passed before and after isolated execution (OK, all notebooks passed validation).
• uv run .internal/run_notebooks.py --isolated noaa-gefs-forecast-analysis-virtual.ipynb - passed in 117.42 s and printed Every notebook ran against only what its install line names.
• Publication-swap concatenated sequence - pyflakes passed with no diagnostics.
• Final file - 2,087,159 bytes, no error or stream outputs, all non-install code cells executed.

Review history

• 0.25° half: two Claude passes (/tmp/rev79-report.md locally) — fixed the cell-12 xr import after the swap, replaced a non-virtual-only animation, an analysis value quoted as an observation, and the variable-count caveat.
• 0.5° extension: one Claude pass on 497bc33 — replaced the eclipse section (its plot did not support "spread grows beyond 10 days"), corrected the SSW member count, restored the 34→38 caveat, plotted Ciarán members rather than the mean, dropped the unused 35-day root open. Applied at 6122113.