Overview of WRF

Physics
Jimy Dudhia
NCAR/MMM
 WRF Physics
— Radiation
— Longwave (ra_lw_physics)
— Shortwave (ra_sw_physics)
— Surface
— Surface layer (sf_sfclay_physics)
— Land/water surface (sf_surface_physics)
— PBL (bl_pbl_physics)
— Turbulence/Diffusion (diff_opt, km_opt)
— Cumulus parameterization (cu_physics)
— Microphysics (mp_physics)
Direct Interactions of Parameterizations

Cloud
Detrainment

Convective Cloud Effects

Cloud Fraction

Convective Non-convective
Rain Rain

Surface Downward
T, Qv, Wind SW/LW

Surface Surface
(LH/SH) fluxes emission/albedo
 Radiation
Provides
Atmospheric temperature tendency profile
Surface radiative fluxes
Free Atmosphere Radiation Processes
WRF Longwave Radiation Schemes
(ra_lw_physics)
• Compute clear-sky and cloud upward and downward
radiation fluxes
– Consider IR emission from layers
– Surface emissivity based on land-type
– Flux divergence leads to cooling in a layer
– Downward flux at surface important in land energy budget
– IR radiation generally leads to cooling in clear air (~2K/day),
stronger cooling at cloud tops and warming at cloud base
 Longwave Radiation schemes
ra_lw_physics Scheme Reference Added
1 RRTM Mlawer et al. (1997, JGR) 2000

3 CAM Collins et al. (2004, NCAR Tech. Note) 2006

4 RRTMG Iacono et al. (2008, JGR) 2009

5 New Goddard Chou and Suarez (2001, NASA Tech Memo) 2011

7 FLG (UCLA) Gu et al. (2011, JGR), Fu and Liou (1992, JAS) 2012

31 Held-Suarez 2008

99 GFDL Fels and Schwarzkopf (1981, JGR) 2004

 Longwave Radiation schemes
ra_lw_ Scheme Cores+Chem Microphysics Cloud GHG
physics Interaction Fraction
1 RRTM ARW NMM Qc Qr Qi Qs Qg 1/0 constant
or yearly
GHG
3 CAM ARW Qc Qi Qs Max-rand yearly
overlap CO2 or
GHG
4 RRTMG ARW +Chem(τ) Qc Qr Qi Qs Max-rand constant
overlap or yearly
GHG
5 New Goddard ARW Qc Qr Qi Qs Qg 1/0 constant

7 FLG (UCLA) ARW Qc Qr Qi Qs Qg 1/0 constant

31 Held-Suarez ARW none none none

99 GFDL ARW NMM Qc Qr Qi Qs Max-rand constant

overlap
 Clear Sky: IR-active Gases
— H2O – from model prognostic vapor
— CO2 – well-mixed, specified constant in whole
atmosphere (CAM has yearly values)
— For CAM, RRTM and RRTMG, GHG input file can update CO2, N2O
and CH4

— O3 – schemes have own climatologies

— CAM has monthly, zonal, pressure-level data and RRTMG has this
as an option
— Others use single profiles (Goddard has 5 profiles to choose from)
Radiation Effects in Clear Sky

shortwave longwave
 Spectral Bands
— Schemes divide IR spectrum into bands
dominated by different absorption gases
— Typically 8-16 bands are used
— Computations use look-up tables for each
band
— Tables were generated from results of line-by-line
calculations (LBLRTM models)
 Clouds
— All schemes interact with resolved model cloud fields
allowing for ice and water clouds and precipitating
species
— Some microphysics options pass own particle sizes to RRTMG
radiation: other combinations only use mass info and assume
effective sizes

— Clouds strongly affect IR at all wavelengths (considered

“grey bodies”) and are almost opaque to it
 Cloud Fractions
— Cloud fraction for microphysics clouds
— icloud=1: Xu and Randall method
— icloud=2: simple 1/0 method
— Icloud=3: new Thompson option in V3.7 (improved in V3.8)

— Cloud fraction for unresolved convective clouds

— cu_rad_feedback = .true.
— Only works for GF, G3, GD and KF options
— ZM separately provides cloud fraction to radiation
 Cloud Fraction
— Overlap assumptions needed with multiple layers of
varying fraction
— Random overlap
— Maximum overlap (clouds stacked as much as possible)
— Maximum-random overlap (maximum for neighboring cloudy
layers, random for layers separated by clear air)

— Different WRF schemes may use different cloud

overlapping assumption. For example, RRTMG, CAM use
max-random overlap
 WRF Shortwave Radiation Options
(ra_sw_physics)
— Compute clear-sky and cloudy solar fluxes
— Include annual and diurnal solar cycles
— Most schemes consider downward and upward (reflected)
fluxes
— Dudhia scheme only has downward flux

— Primarily a warming effect in clear sky

— Important component of surface energy balance
 Shortwave Radiation schemes
ra_sw_physic Scheme Reference Added
s
1 Dudhia Dudhia (1989, JAS) 2000

2 Goddard Chou and Suarez (1994, NASA Tech Memo) 2000

3 CAM Collins et a. (2004, NCAR Tech Note) 2006

4 RRTMG Iacono et al. (2008, JGR) 2009

5 New Goddard Chou and Suarez (1999, NASA TM) 2011

7 FLG (UCLA) Gu et al. (2011, JGR), Fu and Liou (1992, JAS) 2012

99 GFDL Fels and Schwarzkopf (1981, JGR) 2004

 Shortwave Radiation
ra_lw_ Scheme Cores+Chem Microphysics Cloud Ozone
physics Interaction Fraction
1 Dudhia ARW NMM + Qc Qr Qi Qs Qg 1/0 none
Chem(PM2.5)
2 GSFC ARW +Chem(τ) Qc Qi 1/0 5 profiles

3 CAM ARW Qc Qi Qs Max-rand Lat/mont

overlap h
4 RRTMG ARW +Chem(τ), Qc Qr Qi Qs Max-rand 1 profile
NMM overlap or
lat/month
5 New Goddard ARW Qc Qr Qi Qs Qg 1/0 5 profiles

7 FLG (UCLA) ARW Qc Qr Qi Qs Qg 1/0 5 profiles

99 GFDL ARW NMM Qc Qr Qi Qs Max-rand Lat/date

overlap
 Clear Sky and Aerosols
— Main gas effect in troposphere is water vapor absorption
(CO2 minor effect)

— Aerosols would be needed for additional scattering (WRF-

Chem interacts with Goddard and RRTMG shortwave)
— Dudhia scheme has tunable scattering
— RRTMG has climatological aerosol input options
— aer_opt=1 Tegen (EC) global monthly climatology
— aer_opt=2 user-specified properties and/or AOD map
— aer_opt=3 Thompson microphysics nuclei (V3.8)
 Ozone
— Ozone heating maintains warm stratosphere
— Important for model tops above about 20 km (50 hPa)
— Usually specified from profiles as with longwave options
— Dudhia scheme has no ozone effect
— CAM, RRTMG have zonal climatology

— CAM, RRTMG, Goddard can also handle trace gases

mainly N2O and CH4 (set constant)
 Spectral Bands
— Many schemes use multiple spectral bands
— As with longwave, bands are ranges of
wavelengths usually dominated by different
gases
— Look-up tables
— Also as with longwave
 Clouds and Cloud Fraction
— Similar considerations to longwave
— Interacts with model resolved clouds and in
some cases cumulus schemes
— Fraction and overlap assumptions
— Cloud albedo reflection
— Surface albedo reflection based on land-
surface type and snow cover
 Slope effects on shortwave
— Available for all shortwave options
— Represents effect of slope on surface solar
flux accounting for diffuse/direct effects
— Two levels of detail (namelist options):
— slope_rad: activates slope effects - may be useful for
complex topography and grid lengths < 2 km.
— topo_shading: shading of neighboring grids by mountains - may
be useful for grid lengths < 1 km.
 radt
Radiation time-step recommendation
— Radiation is too expensive to call every step
— Frequency should resolve cloud-cover changes with
time
— radt=1 minute per km grid size is about right (e.g.
radt=10 for dx=10 km)
— Each domain can have its own value but recommend
using same value on all 2-way nests
 Surface Shortwave Fluxes
— swint_opt=1
— provides a smooth surface downward flux over time (interpolates
between radiation steps using cosine zenith angle and clearness
index)
— This also allows smoother variation of ground variables and fluxes
(eliminates steps in time series)

— Diffuse, direct, and direct normal shortwave components

are output (swddir, swddif, swddni) – aerosols affect
diffuse/direct ratio
Direct Interactions of Parameterizations

Cloud
Detrainment

Convective Cloud Effects

Cloud Fraction

Convective Non-convective
Rain Rain

Surface Downward
T, Qv, Wind SW/LW

Surface Surface
(LH/SH) fluxes emission/albedo
 Surface schemes
Surface layer of atmosphere diagnostics (exchange/transfer coeffs)
Land Surface: Soil temperature /moisture /snow prediction /sea-ice
temperature
Surface Physics Components
Exchange coefficients
for heat and moisture
Atmospheric
Land Surface Model
Surface Layer

Land-surface fluxes
of heat and moisture

Friction stress and

Water-surface fluxes
of heat and moisture

PBL
 Surface Fluxes
— Heat, moisture and momentum

H = ρc p u*θ* E = ρu*q* τ = ρu*u*

kVr kΔθ kΔq

u* = θ* = q* =
ln(zr / z0 ) − ψ m ln(zr / z0h ) − ψ h ln(zr / z0q ) − ψ h

Subscript r is reference level (lowest model level, or 2 m or 10 m)

D refers to difference between surface and reference level value
z0 are the roughness lengths
k is the von Karman constant (0.4)
 Roughness Lengths
— Roughness lengths are a measure of the “initial” length
scale of surface eddies, and generally differ for velocity
and scalars
— Roughness length depends on land-use type
— Some schemes use smaller roughness length for heat
than for momentum
— For water points roughness length is a function of surface
wind speed
 Exchange Coefficient
— Chs is the exchange coefficient for heat, defined such that

H = ρc p ChsΔθ
It is related to the roughness length,
stability function and u* by

ku*
Chs =
€ " z%
ln$ ' − ψ h
# z0 &
 WRF Surface Layer Options
(sf_sfclay_physics)
• Use similarity theory to determine exchange coefficients
and diagnostics of 2m T and q and 10 m winds

• Provide exchange coefficient to land-surface models

• Provide friction velocity to PBL scheme
• Provide surface fluxes over water points
• Schemes have variations in stability functions, roughness
lengths
 Hurricane Options
— Ocean Mixed Layer Model (sf_ocean_physics=1)
— 1-d slab ocean mixed layer (specified initial depth)
— Includes wind-driven ocean mixing for SST cooling feedback
— 3d PWP ocean (Price et al.) (sf_ocean_physics=2)
— 3-d multi-layer (~100) ocean, salinity effects
— Fixed depth
— Alternative surface-layer options for high-wind ocean surface
(isftcflx=1,2)
— Use with sf_sfclay_physics=1
— Modifies Charnock relation to give less surface friction at high winds
(lower Cd)
— Modifies surface enthalpy (Ck, heat/moisture) either with constant
z0q (isftcflx=1), Garratt formulation (option 2)
 Fractional Sea Ice
— fractional_seaice=1 - with input sea-ice fraction data can
partition land/water fluxes within a grid box

— Can be used with nearly all surface-layer schemes

 WRF Land-Surface Model Options
(sf_surface_physics)

— Simple 5-layer soil model

— No vegetation or snow cover prediction, just thermal
diffusion in soil layers

— Noah LSM, NoahMP, RUC LSM, PX LSM,

CLM4, SSiB land-surface models
— Sophisticated vegetation model and snow cover prediction
Land-Surface Model Processes
 Land-Surface Model
— Driven by surface energy and water fluxes
— Predicts soil temperature and soil moisture in layers (4
for Noah and NoahMP, 9 for RUC, 2 for PX and 3 for SSiB,
10 for CLM4)

— Predicts snow water equivalent on ground. May be in

layers (NoahMP, RUC, SSiB,CLM4)

— May predict canopy moisture only (Noah, RUC) or

temperature only (SSiB) or both (NoahMP, CLM4)
 Land-Surface Options
— 5-layer thermal diffusion
— Noah LSM (also with mosaic option in V3.6)
— RUC LSM
— Pleim-Xiu LSM
— NoahMP
— SSiB
— CLM4
 Vegetation and Soil

— Processes include evapotranspiration, root zone and leaf

effects

— Vegetation fraction varies seasonally

— Considers vegetation categories (e.g. cropland, forest
types, etc.)

— Considers soil categories (e.g. sandy, clay, etc.) for

drainage and thermal conductivity
 Snow Cover
— LSMs include fractional snow cover and predict snow
water equivalent development based on precipitation,
sublimation, melting and run-off
— Single-layer snow (Noah, PX)
— Multi-layer snow (RUC, NoahMP, SSiB,CLM4)
— 5-layer option has no snow prediction

— Frozen soil water also predicted (Noah, NoahMP,

RUC,CLM4)
 Urban Effects
— Urban category in LSM is usually adequate for larger-
scale studies

— Or can use an urban model (sf_urban_physics) with Noah

LSM
— Urban Canopy Model
— Building Environment Parameterization (multi-layer model)
— Building Energy Model (adds heating/AC to BEP)
— NUDAPT detailed map data for 40+ US cities
 LSM Tables
— Properties can be changed in text files (tables)
— VEGPARM.TBL used by Noah and RUC for vegetation
category properties
— Albedo, roughness length, emissivity, vegetation properties

— MPTABLE.TBL used by NoahMP

— SOILPARM.TBL used by Noah and RUC for soil properties
— LANDUSE.TBL used by 5-layer model
— URBPARM.TBL used by urban models
 Initializing LSMs
• Noah and RUC LSM require additional fields for
initialization
• Soil temperature
• Soil moisture
• Snow liquid equivalent
• These are in the Grib files, but are not from
observations
• They come from “offline” models driven by
observations (rainfall, radiation, surface temperature,
humidity wind)
 Initializing LSMs
— There are consistent model-derived datasets for Noah
and RUC LSMs
— Eta/GFS/AGRMET/NNRP for Noah (although some
have limited soil levels available)
— RUC for RUC
— But, resolution of mesoscale land-use means there will
be inconsistency in elevation, soil type and vegetation
— The only adjustment for soil temperature (done in
real.exe) is for elevation differences between the original
elevation and model elevation (SOILHGT used)
 Initializing LSMs

— Inconsistency leads to spin-up as adjustments occur in

soil temperature and moisture at the beginning of the
simulation
— This spin-up can only be avoided by running offline model
on the same grid (e.g. HRLDAS for Noah) – may take
months to spin up soil moisture
— Cycling land state between forecasts also helps, but may
propagate errors (e.g in rainfall effect on soil moisture)
 Sub-grid Mosaic
— Default behavior is one dominant vegetation and soil
type per grid cell

— Noah (sf_surface_mosaic) and RUC (mosaic_lu and

mosaic_soil) allow multiple categories within a grid cell
 sst_update=1
Reads lower boundary file periodically to update the
sea-surface temperature (otherwise it is fixed with
time)
— For long-period simulations (a week or more)
— wrflowinp_d0n created by real
— Sea-ice can be updated
— Vegetation fraction update is included
— Allows seasonal change in albedo, emissivity, roughness
length in Noah LSM
— usemonalb=.true. to use monthly albedo input
 Regional Climate Options
• tmn_update=1 - updates deep-soil temperature for
multi-year future-climate runs
— sst_skin=1 - adds diurnal cycle to sea-surface
temperature
• bucket_mm and bucket_J - a more accurate way to
accumulate water and energy for long-run budgets
(see later)
• output_diagnostics=1 – ability to output
max/min/mean/std of surface fields in a specified
period
 Lake Model
— 10-layer lake model from CLM (sf_lake_physics=1)
— We have global bathymetry data for most large lakes
(added from geogrid)

— Also can predict lake ice

— Can be used with any LSM
— WPS preprocessing allows diurnal averaging methods to
initialize lake temperatures where not resolved by SST
analysis (TAVGSFC)
 WRF-Hydro
— Coupling to hydrological model available
— Streamflow prediction, etc.
— Sub-grid tiling to ~100 m grid
— Requires special initialization for hydrological datasets
Direct Interactions of Parameterizations

Cloud
Detrainment

Convective Cloud Effects

Cloud Fraction

Convective Non-convective
Rain Rain

Surface Downward
T, Qv, Wind SW/LW

Surface Surface
(LH/SH) fluxes emission/albedo
Planetary Boundary
Layer
Provides
Boundary layer fluxes (heat, moisture, momentum)
Vertical diffusion in whole column
Planetary Boundary Layer
 WRF PBL Options (bl_pbl_physics)
— Purpose is to distribute surface fluxes with boundary
layer eddy fluxes and allow for PBL growth by
entrainment

— Classes of PBL scheme

— Turbulent kinetic energy prediction (Mellor-Yamada Janjic, MYNN,
Bougeault-Lacarrere, TEMF, QNSE, CAM UW)
— Diagnostic non-local (YSU, GFS, MRF, ACM2)

— Above PBL all these schemes also do vertical diffusion

due to turbulence
 PBL schemes
bl_pbl_p Scheme Reference Added
hysics
1 YSU Hong, Noh and Dudhia (2006, MWR) 2004
2 MYJ Janjic (1994, MWR) 2000
3 GFS Hong and Pan (1996, MWR) 2005
4 QNSE-EDMF Sukoriansky, Galperin and Perov (2005, BLM), Pergaud, 2012
Masson, Malardel et al. (2009, BLM)
5 MYNN2 Nakanishi and Niino (2006, BLM) 2009
6 MYNN3 Nakanishi and Niino (2006, BLM) 2009

7 ACM2 Pleim (2007, JAMC) 2008

8 BouLac Bougeault and Lacarrere (1989, MWR) 2009

9 UW Bretherton and Park (2009, JC) 2011

10 TEMF Angevine, Jiang and Mauritsen (2010, MWR) 2011

11 SH Shin and Hong (2014, MWR) 2015

12 GBM Grenier and Brethertion (2001, MWR) 2013

99 MRF Hong and Pan (1996, MWR) 2000

Nonlocal PBL schemes
Non-local schemes have two main components
 TKE schemes

— Solve for TKE in each column

— Buoyancy and shear production ∂ ∂
Kv θ
— Dissipation ∂z ∂z

— Vertical mixing

— TKE and length-scale are used to determine the Kv for

local vertical mixing

— Schemes differ most in diagnostic length-scale

computations
 Nonlocal Schemes
— Diagnose a PBL top (either stability profile or Richardson
number)
∂ %∂ (
— Specify a K profile K v ' θ + Γ*
∂z & ∂z )
— YSU, MRF, GFS include a non-gradient term (Γ)
— ACM2, TEMF, EDMF include a mass-flux
€ profile, M, which
is an additional updraft flux

∂% ∂ (
'K v θ + M(θ u − θ )*
∂z & ∂ z )
 Vertical Mixing Coefficient
— Several schemes also output exch_h which is Kv for
scalars that is used by WRF-Chem

— WRF can do scalar and tracer vertical mixing with PBL K-

coefficients
— scalar_pblmix=1, tracer_pblmix=1

— PBL schemes themselves only mix limited variables:

momentum, heat, vapor and some specific cloud
variables
 PBL Schemes with Shallow
Convection
— Some PBL schemes include shallow convection as part of
their parameterization

— These use mass-flux approaches either

— through the whole cloud-topped boundary layer (QNSE-EDMF and
TEMF)
— only from cloud base (GBM and UW PBL)

— YSU has top-down mixing option for turbulence driven by

cloud-top radiative cooling which is separate from
bottom-up surface-flux-driven mixing
 PBL schemes
bl_pbl_p Scheme Cores sf_sfclay_p Prognostic Diagnostic Cloud
hysics hysics variables variables mixing

1 YSU ARW NMM 1,91 exch_h QC,QI

2 MYJ ARW NMM 2 TKE_PBL EL_PBL, exch_h QC,QI
3 GFS(hwrf) NMM 3 QC,QI
4 QNSE-EDMF ARW NMM 4 TKE_PBL EL_PBL, exch_h, QC,QI
exch_m
5 MYNN2 ARW 1,2,5,91 QKE Tsq, Qsq, Cov, QC
exch_h, exch_m
6 MYNN3 ARW 1,2,5,91 QKE, Tsq, exch_h, exch_m QC
Qsq, Cov
7 ACM2 ARW 1,7,91 QC,QI
8 BouLac ARW 1,2,91 TKE_PBL EL_PBL, exch_h, QC
exch_m
9 UW ARW 1,2,91 TKE_PBL exch_h, exch_m QC
10 TEMF ARW 10 TE_TEMF *_temf QC, QI
11 SH ARW 1,91 Exch_h QC, QI
12 GBM ARW 1,91 TKE_PBL EL_PBL,exch_h, QC, QI
exch_m
99 MRF ARW NMM 1,91 QC,QI
 PBL Scheme Options
— PBL schemes can be used for most grid sizes when
surface fluxes are present
— Lowest level should be in the surface layer (0.1h)
— Important for surface (2m, 10m) diagnostic interpolation
— With ACM2, GFS and MRF PBL schemes, lowest full
level should be .99 or .995 (not too close to 1)
— TKE schemes can use thinner surface layers
— Assumes that PBL eddies are not resolved
— At grid size dx << 1 km, this assumption breaks
down
— Can use 3d diffusion instead of a PBL scheme (coupled to surface
physics)
— Works best when dx and dz are comparable
 Model Grid Spacing: PBL and LES
Energy
Energy-containing turbulence scale
“l”

Turbulent energy spectrum

in convective PBL

Scale of
motions
100 km 10 km 1 km 100 m

Δ >> l Δ << l
PBL Grey Zone
For coarse grid spacing For fine grid spacing
ü PBL schemes have been ü LES schemes have been
designed for Δ >> l designed for Δ << l
ü All eddies are sub-grid ü All major eddies are resolved
ü 1d column schemes handle ü 3d turbulence schemes handle
sub-grid vertical fluxes sub-grid mixing
 Grey-Zone PBL
— “Grey Zone” is sub-kilometer grids
— PBL and LES assumptions not perfect

— New Shin-Hong PBL based on YSU designed for sub-

kilometer transition scales (200 m – 1 km)
— Nonlocal mixing (gamma) term reduces in strength as grid size
gets smaller and resolved mixing increases

— Other schemes may work in this range but will not have
correctly partitioned resolved/sub-grid energy fractions
 Large-Eddy Simulation

— For grid sizes of up to about 100 m, LES is preferable

— LES treats turbulence three-dimensionally instead of
separate vertical (PBL) and horizontal diffusion schemes

— TKE and 3d Smagorinsky options exist for the sub-grid

turbulence
 Large-Eddy Simulation

— To run LES mode

— Use bl_pbl_physics=0 and diff_opt=2 with km_opt=2 or 3
— This scheme can also use real surface fluxes from the
surface physics (heat, moisture, momentum stress) or
idealized constant values
 LES schemes
Unified horizontal and vertical mixing (for dx~dz).
Typically needed for dx<~200 m. Also use mix_isotropic=1.

bl_pbl_p diff_opt km_opt Scheme Cores sf_sfclay isfflx Prognostic

hysics _physics variables
0 2 2 tke ARW 0,1,2 0,1,2 tke

0 2 3 3d Smagorinsky ARW 0,1,2 0,1,2

Namelist isfflx controls surface flux methods

isfflx sf_sfclay_physics Heat flux Drag Real/Ideal
0 0 From namelist From namelist Ideal
tke_heat_flux tke_drag_coefficient

1 1,2 From LSM/sfclay From sfclay physics Real

physics (HFX, QFX) (UST)

2 1,2 From namelist From sfclay physics Ideal

tke_heat_flux (UST)
 Other Options
— For YSU
— topo_wind=1,2: wind-bias correction methods for terrain effects
— ysu_topdown_pblmix=1: cloud-top cooling-driven mixing
— For MYNN
— Wind-farm model has been added to investigate wind-farm
effects on the environment (extra stress and turbulence
generation)

— Gravity-wave drag can be added for low resolution (> 5

km) runs to represent sub-grid orographic gravity-wave
vertical momentum transport (gwd_opt=1)
— Fog: grav_settling=2 (Katata)
 bldt
— Minutes between boundary layer/LSM calls
— Typical value is 0 (every step)
— CLM LSM is expensive, so may consider bldt in that case
Turbulence/Diffusion
Sub-grid eddy mixing effects on all fields, e.g.

∂ ∂ ∂ ∂ ∂ ∂
Kh θ + Kh θ + Kv θ
∂x ∂x ∂y ∂y ∂z ∂z
Difference between diff_opt 1 and 2

mixing

diff_opt=1
Horizontal diffusion acts along model levels
Simpler numerical method with only neighboring points on the
same model level
Difference between diff_opt 1 and 2

diff_opt=2
Horizontal diffusion acts on horizontal gradients
Numerical method includes vertical correction term
using more grid points
 km_opt
— km_opt selects method for computing K coefficient
— km_opt=1: constant (use khdif and kvdif to specify –
idealized)
— km_opt=2: 3d tke prediction used to compute K
(requires diff_opt=2)
— km_opt=3: 3d Smagorinsky diagnostic K (requires
diff_opt=2)
— km_opt=4: 2d Smagorinsky for horizontal K (to be
used with PBL or kvdif for vertical K )
 sfs_opt
— Sub-filter-scale stress model for LES applications
impacting momentum mixing (Kosovic, Mirocha)
— sfs_opt=0 (default) off
— sfs_opt=1 Nonlinear Backscatter and Anisotropy (NBA) option 1:
using diagnostic stress terms (km_opt=2,3)
— sfs_opt=2 NBA option 2: using tke-based stress terms (km_opt=2
only)
— Also m_opt=1 for added outputs of SGS stresses
 Diffusion Option Choice
— Real-data case with PBL physics on
— Best is diff_opt=1, km_opt=4
— From V3.6 diff_opt=2 can be used with km_opt=4
(was unstable with complex terrain before this version)
— This complements vertical diffusion done by PBL
scheme
— High-resolution real-data cases (~100 m grid)
— No PBL
— diff_opt=2; km_opt=2,3 (tke or Smagorinsky scheme)
 Diffusion Option Choice
— Idealized cloud-resolving (dx =1-3 km) modeling (smooth
or no topography, no surface heat fluxes)
— diff_opt=2; km_opt=2,3
— Complex topography with no PBL scheme
— diff_opt=2 is more accurate for sloped coordinate
surfaces, and prevents diffusion up/down valley sides
but still sometimes unstable with complex terrain
— Note: WRF can run with no diffusion (diff_opt=0)
 diff_6th_opt
— 6th order optional added horizontal diffusion on model levels
— Used as a numerical filter for 2*dx noise
— Suitable for idealized and real-data cases
— Affects all advected variables including scalars

— diff_6th_opt
— 0: none (default)
— 1: on (can produce negative water)
— 2: on and prohibit up-gradient diffusion (better for water conservation)

— diff_6th_factor
— Non-dimensional strength (typical value 0.12, 1.0 corresponds to
complete removal of 2*dx wave in a time-step)
 Upper damping (damp_opt)
Purpose is to prevent unrealistic reflections of waves from
model top. Can be important over high topography.
Options
— 1: Upper level diffusive layer
— 2: Rayleigh damping (idealized only – needs input
sounding)
— 3: w-Rayleigh damping (damps w only)
All options use
— Cosine function of height
— Additional parameters
— zdamp: depth of damping layer
— dampcoef: nondimensional maximum magnitude of damping