User's Guide for the NMM Core of the
Weather Research and Forecast (WRF)
Modeling System Version 3
Table of Contents
The
WRF-NMM is a fully compressible, non-hydrostatic mesoscale model with a
hydrostatic option (Janjic et al. 2001, Janjic 2003a,b). The model uses a
terrain following hybrid sigma-pressure vertical coordinate. The grid
staggering is the Arakawa E-grid. The same time step is used for all
terms. The dynamics conserve a number of
first and second order quantities including energy and enstrophy (Janjic 1984).
The WRF-NMM code contains an initialization program (real_nmm.exe; see Chapter 4) and a numerical integration program (wrf.exe). The WRF-NMM model Version 3.0 supports a variety of capabilities. These include:
WRF-NMM Dynamics in a Nutshell:
Horizontally
propagating fast-waves: Forward-backward
scheme
Vertically
propagating sound waves: Implicit
scheme
Horizontal: Adams-Bashforth scheme
Vertical: Crank-Nicholson scheme
TKE,
water species: Explicit, iterative,
flux-corrected (called every two time steps).
Advection (space) for
T, U, V:
Horizontal: Energy and enstrophy conserving,
quadratic conservative, second order
Vertical: Quadratic conservative, second
order
TKE, Water species: Upstream, flux-corrected, positive definite, conservative
Diffusion in the WRF-NMM is categorized as lateral diffusion and vertical diffusion. The vertical diffusion in the PBL and in the free atmosphere is handled by the surface layer scheme and by the boundary layer parameterization scheme (Janjic 1996a, 1996b, 2002a, 2002b). The lateral diffusion is formulated following the Smagorinsky non-linear approach (Janjic 1990). The control parameter for the lateral diffusion is the square of Smagorinsky constant.
The
horizontal component of divergence is damped (Sadourny 1975). In addition, if applied, the technique for
coupling the elementary subgrids of the E grid (Janjic 1979) damps the
divergent part of flow.
All available WRF System physics package options are listed below. Some of these options have not yet been tested for WRF-NMM. Indication of the options that have been tested, as well as the level of the testing, is included in the discussion below.
It is recommended that the same physics be used in all grids (coarsest and nests). The only exception is that the cumulus parameterization may be activated on coarser grids and turned off on finer grids.
0. No microphysics
1. Kessler scheme: A warm-rain (i.e. no ice)
scheme used commonly in idealized cloud modeling studies (Kessler 1969, Wicker
and Wilhemson 1995).
2. Lin et al. scheme: A sophisticated scheme that has ice, snow and graupel processes, suitable for real-data high-resolution simulations (Lin et al. 1983, Rutledge and Hobbs 1984, Tao et al. 1989, Chen and Sun 2002).
3. WRF Single-Moment (WSM) 3-class simple ice scheme: A simple efficient scheme with ice and snow processes suitable for mesoscale grid sizes (Hong et al. 1998, Hong et al. 2004).
4. WRF Single-Moment (WSM) 5-class scheme. A slightly more sophisticated version of option 3 that allows for mixed-phase processes and super-cooled water (Hong et al. 1998, Hong et al. 2004). (This scheme has been preliminarily tested for WRF-NMM.)
5. Ferrier scheme: A scheme that includes prognostic mixed-phase processes (Ferrier et al. 2002). This scheme was recently changed so that ice saturation is assumed at temperatures colder than -30C rather than -10C as in the original implementation. (This scheme is well tested for WRF-NMM, used operationally at NCEP.)
6. WSM 6-class graupel scheme: A new scheme with ice, snow and graupel processes suitable for high-resolution simulations (Lin et al. 1983, Dudhia 1989, Hong et al. 1998). (This scheme has been preliminarily tested for WRF-NMM.)
7. Goddard microphysics scheme: A scheme with ice, snow and graupel processes suitable for high-resolution simulations.
8. Thompson et al. scheme: A scheme with six classes of moisture species plus number concentration for ice as prognostic variables (Thompson et al. 2004). (This scheme has been preliminarily tested for WRF-NMM.)
10. Morrison double-moment scheme: Double-moment ice, snow, rain and graupel for cloud-resolving simulations.
Longwave Radiation (ra_lw_physics)
1. RRTM scheme: Rapid Radiative Transfer Model. An accurate scheme using look-up tables for efficiency. Accounts for multiple bands, trace gases, and microphysics species (Mlawer et al. 1997). (This scheme has been preliminarily tested for WRF-NMM.)
3. CAM scheme: from the
CAM 3 climate model used in CCSM. Allows for aerosols and trace gases.
99. GFDL scheme: Geophysical Fluid Dynamics Laboratory (GFDL)
longwave. An older version multi-band, transmission table look-up scheme with
carbon dioxide, ozone and water vapor absorptions (Fels and Schwarzkopf 1975,
Schwarzkopf and Fels 1985, Schwarzkopf and Fels 1991). Cloud microphysics
effects are included. (This scheme is well tested for WRF-NMM, used
operationally at NCEP.) Note: If
it is desired to run GFDL with a microphysics scheme other than Ferrier, a
modification to module_ra_gfdleta.F is needed to comment out (!) #define
FERRIER_GFDL.
Shortwave Radiation (ra_sw_physics)
1. Dudhia scheme: Simple downward integration allowing for efficient cloud and clear-sky absorption and scattering (Dudhia 1989). (This scheme has been preliminarily tested for WRF-NMM.)
2.
Goddard Shortwave scheme: Two-stream multi-band scheme
with ozone from climatology and cloud effects (Chou and Suarez 1994).
3. CAM scheme: from the CAM 3 climate model used in CCSM. Allows for aerosols and trace gases.
99.
GFDL scheme: Geophysical Fluid Dynamics Laboratory (GFDL) shortwave. A two
spectral bands, k-distribution scheme with ozone and water vapor as the main
absorbing gases (Lacis and Hansen 1974). Cloud microphysics effects are
included. (This scheme is well-tested for WRF-NMM, used operationally at NCEP.) Note: If it is desired to run
GFDL with a microphysics scheme other than Ferrier, a modification to
module_ra_gfdleta.F is needed to comment out (!) #define FERRIER_GFDL.
Surface Layer
(sf_sfclay_physics)
1. Monin-Obukhov Similarity scheme: Based on Monin-Obukhov with Carslon-Boland viscous sub-layer and standard similarity functions from look-up tables (Skamarock et al. 2005). (This scheme has been preliminarily tested for WRF-NMM.)
2. Janjic Similarity scheme: Based on similarity theory with viscous sublayers both over solid surfaces and water points (Janjic, 1996b, Chen et al. 1997). (This scheme is well tested for WRF-NMM, used operationally at NCEP.)
3.
NCEP Global Forecasting System (GFS) scheme: The
Monin-Obukhov similarity profile relationship is applied to obtain the surface
stress and latent heat fluxes using a formulation based on Miyakoda and Sirutis
(1986) modified for very stable and unstable situations. Land surface
evaporation has three components (direct evaporation from the soil and canopy,
and transpiration from vegetation) following the formulation of Pan and Mahrt
(1987). (This scheme has been preliminarily tested for WRF-NMM.)
7.
Pleim-Xiu surface layer.
Land Surface
(sf_surface_physics)
1. Thermal Diffusion scheme: Soil temperature only scheme, using five layers (Skamarock et al. 2005).
2. Noah Land-Surface Model: Unified NCEP/NCAR/AFWA scheme with soil temperature and moisture in four layers, fractional snow cover and frozen soil physics (Chen and Dudhia, 2001). (This scheme has been well tested for WRF-NMM.)
3. RUC Land-Surface Model: Rapid Update Cycle operational scheme with soil temperature and moisture in six layers, multi-layer snow and frozen soil physics (Smirnova et al. 1997, 2000). (This scheme has been preliminarily tested for WRF-NMM.)
7. Pleim-Xiu Land Surface Model: Two-layer scheme with vegetation and sub-grid tiling.
99. NMM Land Surface Scheme: The NMM LSM package is based in the pre-May 2005 Noah Land Surface Model (LSM) in the operational NAM/Eta with soil temperature and moisture in 4 layers, fractional snow cover and frozen soil physics (Ek et al. 2003) and is quite similar to the unified Noah LSM (option 2 above). (This scheme is well tested for WRF-NMM, used operationally at NCEP.)
Planetary
Boundary Layer (bl_pbl_physics)
1. Yonsei University scheme (YSU): Next generation MRF-PBL. Non-local-K scheme with an explicit entrainment layer and parabolic K profile in unstable mixed layer (Skamarock et al. 2005). (This scheme has been preliminarily tested for WRF-NMM.)
2. Mellor-Yamada-Janjic Scheme: One-dimensional prognostic turbulent kinetic energy scheme with local vertical mixing (Janjic 1990, 1996a, 2002). (This scheme is well-tested for WRF-NMM, used operationally at NCEP.)
3. NCEP Global Forecast System scheme: First-order vertical diffusion scheme of Troen and Mahrt (1986) further described in Hong and Pan (1996). The PBL height is determined using an iterative bulk-Richardson approach working from the ground upward whereupon the profile of the diffusivity coefficient is specified as a cubic function of the PBL height. Coefficient values are obtained by matching the surface-layer fluxes. A counter-gradient flux parameterization is included. (This scheme has been preliminarily tested for WRF-NMM.)
7. ACM PBL: Asymmetric Convective Model with
non-local upward mixing and local downward mixing.
99. MRF scheme: An older version of YSU with implicit treatment of entrainment layer as part of non-local-K mixed layer (Hong and Pan 1996).
Note: Two-meter temperatures are only available when running with MYJ scheme (2).
Cumulus Parameterization (cu_physics)
0. No cumulus parameterization. (Tested for WRF-NMM)
1. Kain-Fritsch scheme: Deep and shallow sub-grid scheme using a mass flux approach with downdrafts and CAPE removal time scale (Kain 2004, Kain and Fritsch 1990, 1993). (This scheme has been preliminarily tested for WRF-NMM.)
2. Betts-Miller-Janjic scheme: Adjustment scheme for deep and shallow convection relaxing towards variable temperature and humidity profiles determined from thermodynamic considerations (Janjic 1994, 2000). (This scheme is well tested for WRF-NMM, used operationally at NCEP.)
3. Grell-Devenyi ensemble scheme: Multi-closure, multi-parameter, ensemble method with typically 144 sub-grid members (Grell and Devenyi 2002). (This scheme has been preliminarily tested for WRF-NMM.)
4. Simplified Arakawa-Schubert scheme: Penetrative convection is simulated following Pan and Wu (1995), which is based on Arakawa and Schubert (1974) as simplified by Grell (1993) and with a saturated downdraft. (This scheme is well tested for WRF-NMM.)
5. Grell 3d ensemble cumulus scheme: Scheme for higher resolution domains allowing for subsidence in neighboring columns.
Below
is a summary of physics options that are well-tested for WRF-NMM and are used
operationally at NCEP:
|
&physics |
Identifying Number |
Physics options |
|
mp_physics (max_dom) |
5 |
Microphysics-Ferrier |
|
ra_lw_physics |
99 |
Long-wave radiation - GFDL (Fels-Schwarzkopf) |
|
ra_sw_physics |
99 |
Short-wave radiation - GFDL (Lacis-Hansen) |
|
sf_sfclay_physics |
2 |
Surface-layer: Janjic scheme |
|
sf_surface_physics |
99 |
Land-surface – NMM LSM |
|
bl_pbl_physics |
2 |
Boundary-layer - Mellor-Yamada-Janjic TKE |
|
cu_physics |
2 |
Cumulus - Betts-Miller-Janjic scheme |
|
num_soil_layers |
4 |
Number of soil layers in land surface model |
Description
of Namelist Variables
The settings in the namelist.input file are used to configure WRF-NMM. This file should be edited to specify: dates, number and size of domains, time step, physics options, and output options. When modifying the namelist.input file, be sure to take into account the following points:
time_step:
The general rule for determining the time step of the coarsest grid
follows from the CFL criterion. If d
is the grid distance between two neighboring points (in diagonal direction on
the WRF-NMM's E-grid), dt is the time step, and c
is the phase speed of the fastest process, the CFL criterion requires that:
(c*dt)/[d/sqrt(2.)] ≤1
This gives: dt ≤ d/[sqrt(2.)*c]
A very simple approach is to use 2.25
x (grid spacing in km) or
about 330 x (angular grid spacing) to obtain an integer number of time steps per hour.
For example: If the grid spacing of the coarsest grid is 12km, then this gives dt=27 s
The following are pre-tested time-steps for WRF-NMM:
|
Approximate Grid Spacing (km) |
DELTA_X (in degrees) |
DELTA_Y (in
degrees) |
Time Step
(seconds) |
|
4 |
0.026726057 |
0.026315789 |
9-10s |
|
8 |
0.053452115 |
0.052631578 |
18s |
|
10 |
0.066666666 |
0.065789474 |
24s |
|
12 |
0.087603306 |
0.075046904 |
25-30s |
|
22 |
0.154069767 |
0.140845070 |
60s |
|
32 |
0.222222222 |
0.205128205 |
90s |
e_we and e_sn: Given WRF-NMM’s E-grid
staggering, the end index in the east-west direction (e_we) and the south-north direction (e_sn) for the coarsest grid need to
be set with care and the e_sn value must be EVEN for WRF-NMM.
When using the WRF Preprocessing System (WPS), the
coarsest grid dimensions should be set as:
e_we (namelist.input) = e_ew
(namelist.wps),
e_sn (namelist.input) = e_sn (namelist.wps).
For example: The parent grid e_we and e_sn
are set up as follows:
namelist.input
e_we = 124,
e_sn = 202,
namelist.wps
e_we = 124,
e_sn = 202,
Other
than what was stated above, there are no additional rules to follow when
choosing e_we and e_sn for nested grids.
dx
and dy: For WRF-NMM, dx and dy are the
horizontal grid spacing in degrees, rather than meters (unit used for WRF-ARW).
Note that dx should be slightly larger than dy due
to the convergence of meridians approaching the poles on the rotated grid. The
grid spacing in namelist.input should have the same values as in namelist.wps.
When
using WPS,
dx (namelist.input) = dx (namelist.wps),
dy (namelist.input) = dy
(namelist.,wps).
When running a
simulation with multiple (N) nests, the namelist should have N
values of dx, dy, e_we, e_sn
separated by commas.
For
more information about the horizontal grid spacing for WRF-NMM, please see Chapter 3, WRF Preprocessing System (WPS)
nio_tasks_per_group: The number of I/O tasks (nio_tasks_per_group) should evenly divide into the number of compute tasks in the J-direction on the grid (that is the value of nproc_y). For example, if there are 6 compute tasks in the J-direction, then nio_tasks_per_group could legitimately be set to 1, 2, 3, or 6. The user needs to use a number large enough that the quilting for a given output time is finished before the next output time is reached. If one had 6 compute tasks in the J-direction (and the number in the I-direction was similar), then one would probably choose either 1 or 2 quilt tasks.
The following table provides an overview of the parameters specified in namelist.input. Note that “namelist.input” is common for both WRF cores (WRF-ARW and WRF-NMM). Most of the parameters are valid for both cores. However, some parameters are only valid for one of the cores. Core specific parameters are noted in the table. In addition, some physics options have not been tested for WRF-NMM. Those options that have been tested are highlighted by indicating whether they have been “fully” or “preliminarily” tested for WRF-NMM.
|
Variable Names |
Value (Example) |
Description |
|
|
&time_control |
|
Time control |
|
|
run_days |
2 |
Run time in days |
|
|
run_hours |
0 |
Run time in hours |
|
|
run_minutes |
00 |
Run time in minutes |
|
|
run_seconds |
00 |
Run time in seconds |
|
|
start_year (max_dom) |
2005 |
Four digit year of starting time |
|
|
start_month (max_dom) |
04 |
Two digit month of starting time |
|
|
start_day (max_dom) |
27 |
Two digit day of starting time |
|
|
start_hour (max_dom) |
00 |
Two digit hour of starting time |
|
|
start_minute (max_dom) |
00 |
Two digit minute of starting time |
|
|
start_second (max_dom) |
00 |
Two digit second of starting time |
|
|
end_year (max_dom) |
2005 |
Four digit year of ending time |
|
|
end_month (max_dom) |
04 |
Two digit month of ending time |
|
|
end_day (max_dom) |
29 |
Two digit day of ending time |
|
|
end_hour (max_dom) |
00 |
Two digit hour of ending time |
|
|
end_minute (max_dom) |
00 |
Two digit minute of ending time |
|
|
end_second (max_dom) |
00 |
Two digit second of ending time Note: All end
times also control when the nest domain integrations end. Note: All start and end times are used by real_nmm.exe. One may use either run_days/run_hours etc. or end_year/month/day/hour etc. to control the length of model integration, but run_days/run_hours takes precedence over the end times. The program real_nmm.exe uses start and end times only. |
|
|
interval_seconds |
10800 |
Time interval between incoming real data, which will be the interval between the lateral boundary condition files. This parameter is only used by real_nmm.exe. |
|
|
history_interval (max_dom) |
60 |
History output file interval in minutes |
|
|
frames_per_outfile (max_dom) |
1 |
Output times per history output file, used to split output files into smaller pieces |
|
|
tstart (max_dom) |
0 |
This flag is only for the WRF-NMM core. Forecast hour at the start of the NMM integration. Set to >0 if restarting a run. |
|
|
restart |
.false. |
Logical indicating whether run is a restart run |
|
|
restart_interval |
60 |
Restart output file interval in minutes |
|
|
io_form_history |
2 |
Format of history file wrfout 2 = netCDF |
|
|
io_form_restart |
2 |
Format of restart file wrfrst 2 = netCDF |
|
|
io_form_inpu | |||