Tags¶

working

The navigation tree can only put a topic in one place, but topics do not respect it. Tags are the second way in — and to be useful they have to be a small controlled vocabulary, not free-form keywords.

Why controlled

An earlier version of this site had 205 tags, 177 of which appeared on exactly one page. A tag used once cross-links nothing, which is the entire purpose. The vocabulary below is 37 tags, every one on at least two pages and most on four or more.

Specific terms — openfoam, betz, skew-t — are deliberately not tags. Press / and search for those; search is better at specifics than tags ever are.

The vocabulary¶

Pillar¶

pillar-1 … pillar-10, from the Skills Map. The primary competency comes first and matches the badge under the page title; a second appears only where a page genuinely spans two.

Section¶

Mirrors the navigation, for filtering within an area:

aerospace · astrophysics · computer-vision · computing · deep-learning · fluids · foundations · hpc · machine-learning · meteorology · resources · sciml · thermal · toolchains

Cross-cutting concept¶

The tags that earn their place — each one connects pages the navigation tree separates:

Tag	Connects
`boundary-layer`	aerodynamics, wind turbines, RANS wall treatment, the atmospheric surface layer, fire weather
`conservation`	governing equations, finite volume, MHD, N-body
`data-formats`	parallel I/O, satellite products, NetCDF/GRIB, geospatial rasters
`discretization`	numerical methods, finite volume, time integration, NWP cores, N-body
`extrapolation`	the shared weakness of trees, forecasts, surrogates, PINNs and neural operators
`inverse-problems`	adjoints, data assimilation, PINNs
`mesh`	grid quality, meshing tools, graph networks on unstructured grids
`neural-networks`	architectures through to scientific ML and vision
`numerical-stability`	von Neumann analysis, CFL, MHD positivity, learned-closure blow-up
`optimization`	adjoint design optimization, model training
`parallelism`	MPI, GPUs, distributed training
`performance`	roofline, memory traffic, profiling — the whole computing section
`radiation`	atmospheric and astrophysical transfer, the same equation twice
`remote-sensing`	scientific imagery, satellite observation, geospatial tooling
`reproducibility`	provenance, containers, data versioning, figures
`scaling`	nondimensional groups, Amdahl, disk loading, Kolmogorov, Rossby
`spectral-methods`	Fourier analysis, from image sampling to FNO to POD
`statistics`	estimation, UQ, ML evaluation, data assimilation
`surrogates`	reduced-order models and emulators
`turbulence`	closures, the boundary layer, learned subgrid models
`uncertainty`	UQ, ensembles, observation error
`validation`	V&V, grid convergence, ML evaluation, detection metrics
`wildfire`	fire weather and fire detection from orbit

Browse¶

aerospace¶

astrophysics¶

boundary-layer¶

computer-vision¶

computing¶

conservation¶

data-formats¶

deep-learning¶

discretization¶

extrapolation¶

fluids¶

foundations¶

hpc¶

inverse-problems¶

machine-learning¶

mesh¶

meteorology¶

neural-networks¶

numerical-stability¶

optimization¶

parallelism¶

performance¶

pillar-1¶

pillar-10¶

pillar-2¶

pillar-3¶

pillar-4¶

pillar-5¶

pillar-6¶

pillar-7¶

pillar-8¶

pillar-9¶

radiation¶

remote-sensing¶

reproducibility¶

resources¶

scaling¶

sciml¶

spectral-methods¶

statistics¶

surrogates¶

thermal¶

toolchains¶

turbulence¶

uncertainty¶

validation¶

wildfire¶

Adding a tag¶

Don't, unless it will land on three or more pages across at least two sections — and then add it to the list above and to the CONCEPT set in scripts/audit_pages.py, which rejects tags outside the vocabulary. The gate is deliberate: the previous taxonomy decayed because adding a tag was free.