BUG: Migrate Forecasts to UCAR THREDDS

[MateusStano]

Summary

Solving #933

We are now using OpenDAP with UCAR THREDDS:

• Used for GFS, NAM and RAP. This works mostly exacly the same as we were doing with NOMADS. Just now we are getting from a different server

The following models are not available on THREDDS and were deactivated

• GEFS
• HIRESW

Particularly GEFS is very important for Monte Carlo simulations and we should look for alternatives as fast as possible

Observations

The Thredds OpenDap approach took me a long time to find information about. So in between I tried out other methods which will still be important for us to understand for the future (probably for re-implementing GEFS):

• GRIB2 filter via NOMADS: this adds a few new dependencies do RocketPy (which are in total bigger in size than RocketPy), It is quite hard to code but can be done, probably the best way to move forward with re-adding GEFS, but including the added libraries as optional dependencies
• Siphon: python library that would also add a bunch of dependencies, that are even bigger than the Grib2 method would require

THREDDS actually has an archival version of GEFS that goes up to 2020. I think we can find other alternatives, but this could be a last resort option.

Useful Links

• THREDDS catalog: NSF and NCEI
  • There are a few models that we can add with this (HRRR and AIGFS)

[codecov]

Codecov Report

❌ Patch coverage is 68.91892% with 69 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.20%. Comparing base (9cf3dd4) to head (6f9f177).
⚠️ Report is 49 commits behind head on develop.

Files with missing lines	Patch %	Lines
rocketpy/environment/environment.py	61.76%	39 Missing ⚠️
rocketpy/environment/tools.py	71.87%	27 Missing ⚠️
rocketpy/environment/fetchers.py	77.77%	2 Missing ⚠️
rocketpy/environment/weather_model_mapping.py	93.33%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #943      +/-   ##
===========================================
+ Coverage    80.27%   81.20%   +0.92%     
===========================================
  Files          104      107       +3     
  Lines        12769    14039    +1270     
===========================================
+ Hits         10250    11400    +1150     
- Misses        2519     2639     +120     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Gui-FernandesBR]

I will take a look at this one over the weekend

[Copilot]

Pull request overview

This PR updates RocketPy’s forecast data ingestion to use UCAR THREDDS OPeNDAP endpoints (fixing NOMADS breakage) and introduces mapping support for both THREDDS-style and legacy NOMADS-style variable naming, while explicitly disabling unavailable latest-model shortcuts (GEFS/HIRESW).

Changes:

• Switch GFS/NAM/RAP latest-model fetchers to UCAR THREDDS “Best” datasets and add projected-grid handling (Lambert Conformal) in the Environment ingestion path.
• Expand WeatherModelMapping with explicit *_LEGACY dictionaries and add runtime mapping resolution/fallback when dataset variables don’t match the selected mapping.
• Update docs/tests/changelog to reflect THREDDS migration and the temporary unavailability of GEFS/HIRESW shortcuts.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
rocketpy/environment/fetchers.py	Migrates GFS/NAM/RAP URLs to UCAR THREDDS; adds HRRR/AIGFS fetcher helpers.
rocketpy/environment/tools.py	Adds Lambert Conformal projection conversion and rewrites index-finding helpers to work directly with netCDF variables (incl. descending grids).
rocketpy/environment/weather_model_mapping.py	Adds THREDDS-first mappings plus explicit *_LEGACY aliases and case-insensitive lookup.
rocketpy/environment/environment.py	Adds mapping compatibility checks/fallback, projected-grid support for x/y datasets, and disables GEFS/HIRESW shortcuts with explicit errors.
tests/unit/environment/test_environment.py	Adds unit tests for dictionary resolution behavior and legacy alias lookup.
tests/integration/environment/test_environment.py	Updates integration tests to expect errors for disabled GEFS/HIRESW shortcuts.
docs/user/environment/3-further/other_apis.rst	Documents built-in/legacy mapping names and projected-grid projection usage.
docs/user/environment/1-atm-models/standard_atmosphere.rst	Adds an explicit Sphinx label and minor formatting tweak.
docs/user/environment/1-atm-models/soundings.rst	Removes RUC-specific workflow guidance and redirects users to supported alternatives.
docs/user/environment/1-atm-models/forecast.rst	Updates guidance around remote OPeNDAP usage and marks HIRESW shortcut as unavailable.
docs/user/environment/1-atm-models/ensemble.rst	Adds Sphinx label and marks GEFS shortcut as unavailable with guidance for explicit datasets.
CHANGELOG.md	Adds an Unreleased “Fixed” entry for the THREDDS migration.

Comments suppressed due to low confidence (1)

rocketpy/environment/environment.py:2109

• Ensemble datasets still require a valid dictionary["ensemble"] mapping when the file actually has an ensemble dimension: if the mapping omits it (or __resolve_dictionary_for_dataset falls back to a non-ensemble mapping), the code later fails in inverse_dictionary[dim] when building params. Instead of treating missing ensemble as “single member”, detect whether the dataset dimensions include an ensemble axis and raise a clear ValueError (or infer the ensemble dimension name) when the mapping doesn’t define it.

        # Get ensemble data from file
        has_ensemble_dimension = True
        try:
            num_members = len(data.variables[dictionary["ensemble"]][:])
        except KeyError:
            has_ensemble_dimension = False
            num_members = 1

        # Get pressure level data from file
        levels = get_pressure_levels_from_file(data, dictionary)

        inverse_dictionary = {v: k for k, v in dictionary.items()}
        param_dictionary = {
            "time": time_index,
            "ensemble": range(num_members),
            "level": range(len(levels)),
            "latitude": (lat_index - 1, lat_index),
            "longitude": (lon_index - 1, lon_index),
        }

        # Get dimensions
        try:
            dimensions = data.variables[dictionary["geopotential_height"]].dimensions[:]
        except KeyError:
            dimensions = data.variables[dictionary["geopotential"]].dimensions[:]

        # Get params
        params = tuple(param_dictionary[inverse_dictionary[dim]] for dim in dimensions)

[Copilot]

The class docstring indentation under “Mapping families” / “Notes” is inconsistent with the surrounding lines, which will likely render as a literal block when Sphinx/inspect dedents the docstring. Align those section headings/bullets to the same indentation level as the rest of the docstring text.

Suggested change

	Mapping families
	----------------
	- Base names (for example ``GFS``, ``NAM``, ``RAP``) represent the current
	default mappings used by the latest-model shortcuts and THREDDS-style
	datasets.
	- ``*_LEGACY`` names represent older NOMADS-style variable naming
	conventions (for example ``lev``, ``tmpprs``, ``ugrdprs`` and
	``vgrdprs``) and are intended for archived or previously downloaded files.
	Notes
	-----
	- Mappings can also include optional keys such as ``projection`` for
	projected grids and ``ensemble`` for member dimensions.
	- The :meth:`get` method is case-insensitive, so ``"gfs_legacy"`` and
	``"GFS_LEGACY"`` are equivalent.
	Mapping families
	----------------
	- Base names (for example ``GFS``, ``NAM``, ``RAP``) represent the current
	default mappings used by the latest-model shortcuts and THREDDS-style
	datasets.
	- ``*_LEGACY`` names represent older NOMADS-style variable naming
	conventions (for example ``lev``, ``tmpprs``, ``ugrdprs`` and
	``vgrdprs``) and are intended for archived or previously downloaded files.
	Notes
	-----
	- Mappings can also include optional keys such as ``projection`` for
	projected grids and ``ensemble`` for member dimensions.
	- The :meth:`get` method is case-insensitive, so ``"gfs_legacy"`` and
	``"GFS_LEGACY"`` are equivalent.

The base branch was changed.