WRF Tools v0.5.0 (beta)

Release notes

0.5.0 — 2026-10-15

Beta update focused on connected workflows, explicit scientific conventions,
and reproducible examples. This is not scientific certification.

New and connected workflows

• A catalog of 207 public top-level functions, executable TOML recipes and a
  chaptered showcase that records which functions actually run.
• Selectable diagnostic groups in hindcast reports, with separate domain
  results, maps, wind profiles, precipitation, model TKE when available, and
  explicit messages when sampling or variables are unsuitable.
• Editable examples for weighted horizontal, selected-area, vertical and time
  means; model-level versus common-height averaging is an explicit choice.
• Multi-simulation alignment, scores and comparison plots; observation-file
  loading, time matching and comparison examples.
• Wind animations and actual-data 3-D contour export with a Manim scene.
  Manim remains an optional installation; it is not required for core analysis.
• A single real-data example PDF, API catalog, scientific-method references
  and per-capability execution ledger.

Numerical corrections

Some results intentionally differ from 0.4.0. The
calculation change record gives equations,
references, regression evidence and remaining approximations. Changes include:

• Transverse von Karman spectrum normalization and literal TurbSim constants;
  the declared near-neutral vertical Kaimal variant.
• Odd-length cross-spectra, Fourier resampling/Nyquist handling, spectral bins,
  undefined coherence and invalid sampling settings.
• Q gradient ordering, nonuniform-height derivatives, pressure-consistent
  surface heat-flux conversion and consistent saturation/dew-point inversion.
• Missing-value handling, paired statistical scores, empirical Wasserstein
  distance, precipitation resets and incomplete rotor/interpolation coverage.

These spectrum helpers are not used to calculate the wind/resource statistics
in the retained real-data hindcast PDF. Scientific corrections should not be
interpreted as measured atmospheric improvements.

Compatibility and cleanup

• Superseded demonstration scripts, unused internal configuration code and old
  generated reports are removed, not retained in an archive.
• Use the example guide and
  connected workflows instead of the removed
  single_wrf_all_capabilities.py and audit script.
• Invalid inputs that previously produced misleading results may now raise an
  error or return explicitly undefined values. Check the declared convention
  when comparing results with earlier versions.
• Reports read the package version for new provenance instead of a fixed label.

Verification and limits

• Pre-release checks: 177 tests passed locally, including the runnable showcase
  and numerical identities/library comparisons. Linux/Windows CI is configured
  separately; local results do not certify the remote platform matrix.
• Showcase: 204 functions exercised with labelled synthetic inputs; three
  operations require external prerequisites (map download, wgrib2, OpenFAST).
• The real-data two-domain example ran from a clean output directory without
  manual PDF repair. Its PDF and acceptance record retain their original
  v0.4.0 generation metadata; they are not regenerated for the release label.
• No real-observation validation or complete external turbine-model execution
  is claimed. Hourly model output does not support temporal turbulence
  statistics. See the audit for the full verification boundary.

WRF Tools v0.3.0

Highlights

• Flagship WRF-LES to TurbSim/OpenFAST workflow with BTS writing and round-trip validation.
• Citation-backed scientific-method ledger and in-code methodology references.
• Published Cheynet marine boundary-layer spectra with author-supplied coefficients.
• Explicit native-grid or mass-grid destaggering for variable, point, profile, plane, and CLI extraction.
• Researcher-facing CLI for inspection, validation, concatenation, extraction, spectra, BTS, and OpenFAST output.
• Publication-oriented example plots and an anonymized PDF case report.

Tested on Linux and Windows with Python 3.10 and 3.12.