Model Solver
- fully compressible nonhydrostatic
equations with hydrostatic option
- complete coriolis and curvature terms
- two-way horizontal nesting with multiple nests and nest levels
- one-way nesting
- moving nest
- vertical nesting
- hybrid sigma-pressure coordinate (user tunable)
- vertical grid-spacing can vary with height
- map-scale factors for conformal projections:
* polar stereographic
* Lambert-conformal
* Mercator
* latitude-longitude (which can be rotated)
- Arakawa C-grid staggering
- Runge-Kutta 2nd- and 3rd-order timestep options
- scalar-conserving flux form for prognostic variables
- 2nd to 6th order advection options (horizontal and vertical)
- time-split small step for acoustic and gravity-wave modes:
* small step horizontally-explicit, vertically-implicit
* divergence damping option and vertical time off-centering
* external-mode filtering option
- lateral boundary conditions
* idealized cases: periodic, symmetric, and
open radiative
* real cases: specified with relaxation zone
- upper boundary absorbing layer option
* increased diffusion
* Rayleigh relaxation
* implicit gravity-wave damping
- rigid upper lid option
- positive definite and monotonic advection scheme for scalars
(microphysics species, scalars and tke)
- adaptive time stepping
- stochastic kinetic-energy backscatter scheme for perturbing forecasts)
- WENO (Weighted Essentially Non-Oscillatory) advection scheme
- stochastic parameter perturbation
Physics
- microphysics
* Kessler
* WRF Single Moment (WSM) 3, 5, 6 and 7 class
* Lin et al.
* Eta Ferrier
* Thompson
* Thompson aerosol-aware
* Goddard 4-ice
* Morrison 2-moment
* WRF Double Moment (WDM) 5, 6 and 7 class
* Milbrandt-Yau double moment
* Stony-Brook University Lin scheme
* NSSL 2-moment / 1-moment
* CAM 5.1 2-moment
* HUJI full and fast SBM
* P3
* Jensen ISHMAEL
* National Taiwan University (NTU) multi-moments
- cumulus and shallow convetion parameterization
* Kain-Fritsch with shallow convection
* Betts-Miller-Janjic
* Grell-Devenyi ensemble scheme
* New Grell 3D ensemble scheme
* Grell-Freitas ensemble scheme
* Tiedtke
* New Tiedtke
* New SAS (Simplied Arakawa-Schubert) from GFS
* Old SAS (from GFS too)
* Zhang-McFarlane
* Multi-scale Kain-Fritcsh
* Kain-Fritsch cumulus potential with shallow convection
* University of Washington shallow convection
* GRIMS shallow convection
- planetary boundary layer
* Yonsei University (YSU, S. Korea)
* Mellor-Yamada-Janjic
* Asymmetric Convective Model (ACM2)
* Quasi-normal scale elimination/Eddy diffusivity/ mass flux (QNSE-EDMF)
* Level 2.5 and 3 Mellor-Yamada Nakanishi Niino (MYNN) PBL
* Bougeault-Lacarrere PBL
* University of Washington TKE PBL
* Total energy - mass flux (TEMF) scheme
* Grenier-Bretherton-McCaa TKE PBL
* Shin-Hong
* Shanghai Met Service (SMS) 3DTKE (scale-aware)
* TKE-TKE dissipation rate (epsilon)
* MRF
- surface layer
* similarity theory MM5
* Eta or MYJ
* PX
* QNSE
* MYNN
* TEMF
* Revised MM5 scheme
- land-surface
* slab soil model (5-layer thermal diffusion)
* Unified Noah
land-surface model
* Urban
canopy model (works with Noah and Noah-MP LSM)
* Multi-layer
building environment parameterization (BEP, works with Noah, Noah-MP,
and requires BouLac and MYJ PBL)
* Building energy model (BEM, works with Noah and requires BouLac and MYJ PBL)
* RUC LSM
* PX LSM
* Noah-MP
* SSiB
* CLM4
* Use of fractional sea-ice
* CLM4 lake model
* WRF-Hydro
- longwave radiation
* RRTM
* CAM
* RRTMG
* Goddard
* Fu-Liou-Gu
- shortwave radiation
* simple MM5 scheme, with Zaengl radiation/topography (slope and shadowing) effects
* Goddard (old)
* CAM
* RRTMG
* Goddard
* Fu-Liou-Gu
- ocean physics
* single-column mixed layer ocean model
* 3D Price-Weller-Pinkel (PWP) ocean model
- sub-grid turbulence
* constant K diffusion
* 2-D Smagorinsky
* predicted TKE
* nonlinear backscatter and anisotropy (NBA) turbulence option for LES
- land-use categories determine
surface properties
- SST, greenness fraction, seaice and albedo update during long
simulations
- analysis nudging, 3-D and surface,
flux-adjusting surface data nudging
- observation nudging
- spectral nudging using gridded analyses
WRF Diagnostic Options
- Constant pressure/height output
- Time series output at lat/lon or i/j locations
- Model trajectories with diagnostics along them
- Weather and severe weather diagnostics
- Climate diagnostics, daily maximum/minimum, monthly statistics
WRF initialization
- idealized: several cases, 1D, 2D and 3D
- single column model
- real-data using WRF Preprocessing System (WPS) to prepare input data
- digital filter initialization for single domain and for one-way nesting
Other WRF Model Packages
- WRF-Chemistry
- WRF Data Assimilation
Software Architecture
- Hierarchical software architecture that insulates
scientific code (Model Layer) from computer architecture (Driver Layer).
- Multi-level parallelism supporting shared-memory (OpenMP), distributed-memory
(MPI), and hybrid shared/distributed modes of execution.
- Active data registry: defines and manages model state fields,
I/O, nesting, configuration, and numerous other aspects of WRF
through a single file, called the Registry.
- ESMF Time Management, including exact arithmetic for fractional
time steps (no drift); model start, stop, run length and I/O frequencies
are now specified as times and time intervals.
- Mesh refinement:
* Horizontal and vertical refinement
* 1-way and 2-way nesting
* Easy to extend: forcing and feedback of
new fields specified by editing a single Registry file.
* Generalized functionality permits introducing processing before
a nest is called, or after the nested grid has caught up to the
parent's time step.
* support specified and automatic moving nest
using a vortex-following algorithm
* fully parallel
* telescoping
* sibling
* I/O, some physics and dynamics options permit option choices
between domains
* high-resolution terrain and landuse input for moving nests
- Enhanced I/O options:
* NetCDF classic, pnetcdf, NetCDF4
* split-output file option available
* units in netCDF files are conforming to standard
* User definable auxiliary history output streams separately controllable
through the namelist (currently 25)
* Output file names and time-stamps specifiable through namelist
* Special output stream for WRFDA
* runtime input / output selection
- Efficient CPU execution on a range of computing
platforms:
* Single node within docker container (GNU 9)
* Modestly sized simulations run on multi core laptops, Raspberry Pi
* Cloud based WRF system, currently AWS only, single node usage during tutorials,
multiple node timing performance competitive
* Traditional HPC sites
