Skip to content

Supported and validated

This page answers one question:

Which wrfkit workflows have actually been tested?

A feature being present in the code does not automatically mean it has been tested end to end.

Supported path at a glance

Capability Status What that means
WRF 4.8.0 build Validated Clean build completed on Lambda Vector and UGA Sapelo2
WPS 4.7.0 build Validated geogrid, ungrib, and metgrid built successfully
Rootless Nix on Sapelo2 Validated Tested with the current single-node Sapelo2 profile
GFS acquisition and staging Validated Tested with the bundled validation GFS window
Low-resolution validation geography Validated athens-minimal; intended for testing and education
High-resolution mandatory geography acquisition Validated Official package acquisition and staging have been exercised; CI also covers the downloader with a local archive fixture
case.toml -> native namelist rendering Validated + CI guarded Tracked namelists are regenerated in CI and must remain unchanged; the 3 km fixture values are pinned by assertions
WPS -> real -> WRF Validated Bundled real-data workflow has completed on Lambda Vector and Sapelo2
3 km athens-highres on Sapelo2 Validated 241 x 241, 3 km, 18 s case completed prep and run; WRF used 32 ranks with a 4 x 8 decomposition
3 km athens-highres on Lambda Vector Pending end-to-end Config/plan resolve correctly, but the observed 128-rank automatic launch exceeded PRRTE's available slots
Single-node interactive MPI Validated Tested during development
Single-node sbatch MPI Validated Sapelo2 batch run completed successfully
wrfctl plan Validated Tested on Lambda Vector and Sapelo2
wrfctl prep through real.exe Validated Tested on Lambda Vector and Sapelo2; current 3 km case validated on Sapelo2
wrfctl run Validated Success-marker and new-output checks have completed on Lambda Vector and Sapelo2; current 3 km Lambda regression is still pending
Shared reusable data_root Validated on Sapelo2 Geography and GFS were reused from external /work storage
Preparation freshness detection Validated on Sapelo2 Live stale-case detection was exercised during development
Preparation freshness hard guard CI validated; live regression pending High-level run now blocks stale prep by default; --allow-stale-prep is an explicit escape hatch
Multi-node MPI Not validated Do not treat it as a supported research path yet
WRF restart/recovery workflow Not validated Individual namelist fields exist, but the complete workflow is not tested
Scratch-backed case workspace Not implemented Case workspaces currently stay with the repository

What "validated" means

Validated means the workflow was actually executed in the named environment and produced the expected program-level result.

It does not mean:

  • every WRF namelist option has been tested;
  • every HPC cluster will behave like Sapelo2;
  • a validation-case scientific setup is recommended for research;
  • successful execution proves the weather simulation is scientifically valid.

The main boundary for new users

If you want the least surprising path today, stay within:

Linux x86_64
WRF 4.8.0
WPS 4.7.0
normal or rootless Nix
GFS real-data input
one compute node
local MPI or single-node Sapelo2 Slurm

The bundled athens-minimal and athens-highres cases are workflow and regression fixtures. For research, create a new case and redesign its scientific settings.

How to validate your own case

A new research case needs two kinds of checking.

1. Technical checks

Confirm that:

case.toml renders to the tracked/native namelists as expected
geogrid completes
ungrib completes
metgrid completes
real creates wrfinput/wrfbdy
wrf reports SUCCESS COMPLETE WRF
expected wrfout files exist
run logs are archived

If you edit a case after prep, the normal high-level run now stops rather than silently mixing current namelists with older wrfinput/wrfbdy. Re-run prep, or use --allow-stale-prep only when that reuse is deliberate.

Check whether a run succeeded

2. Scientific checks

These depend on your research question. They can include:

  • domain placement and size;
  • forcing quality;
  • physics choices;
  • spin-up;
  • numerical stability;
  • expected spatial and temporal patterns;
  • missing values;
  • comparison with suitable observations or reference data.

wrfkit can help keep the experiment reproducible, but it cannot decide whether those scientific choices are correct.

Detailed regression history

The longer development record, including MPI decomposition tests, preparation manifest behavior, and shared-data regressions, is kept in Validation notes.