WRF
Model 2003 overview (top)
Overview
The
overall goal of the Weather Research and Forecast (WRF)
project is to develop a next-generation mesoscale forecast
model and data assimilation system that will advance
both the understanding and prediction of mesoscale weather
and accelerate the transfer of research advances into
operations. The model is being developed as a collaborative
effort among the NCAR MMM Division, NOAA's National Centers
for Environmental Prediction (NCEP) and Forecast System
Laboratory (FSL), the Department of Defense's Air Force
Weather Agency (AFWA) and Naval Research Laboratory (NRL),
the Center for the Analysis and Prediction of Storms
(CAPS) at the University of Oklahoma, and the Federal
Aviation Administration (FAA), along with the participation
of a number of university scientists. With this model,
researchers seek to improve the forecast accuracy of
significant weather features across scales ranging from
cloud to synoptic, with priority emphasis on horizontal
grids of one to ten km.
The WRF model is state-of-the-art, transportable, and
efficient in a massively parallel computing environment.
It is designed to be modular, and a single source code
will be maintained that can be configured for both research
and operations. It offers numerous physics options, thus
tapping into the experience of the broad modeling community.
Advanced data assimilation systems are being developed
and tested in tandem with the model. WRF is maintained
and supported as a community mesoscale model to facilitate
wide use in research, particularly in the university
community. Research advances will have a direct path
to operations. With these hallmarks, the WRF model is
unique in the history of numerical weather prediction
in the U.S.
Development efforts on the WRF model continued at a
rapid pace over the course of the past fiscal year, as
the project moves toward scheduled release of the full
research-quality version in December. This progress was
facilitated by real-time forecasting, case-study testing
and analysis, and the development of the WRF data assimilation
system. The basic 3DVAR package underwent testing and
was integrated into operational systems running the MM5.
Both the WRF model and the 3DVAR data assimilation system
were released to the community, and broad use is being
supported through user workshops, tutorials, and consulting
assistance. Testing is underway to determine the configuration
for a high-resolution ensemble implementation for WRF
that will begin running operationally at NCEP beginning
in October 2004. The following sections highlight the
recent progress in the various areas of WRF development.
WRF
Model numerics and dynamic cores (top)
Prototype Dynamical Cores
WRF Model prototype dynamical cores continue
to be developed within the WRF framework to facilitate the
comparative
evaluation of the relative accuracy and efficiency of their
numerical
techniques in a controlled computational environment. Two
completed prototypes are the split-explicit Eulerian models
based on
mass and height vertical coordinates; both are available
as run-time selectable cores within the WRF model framework
and
have been developed principally within MMM. A semi-implicit
semi-Lagrangian prototype is being developed in an effort
led by James Purser (NOAA/National Centers for Environmental
Prediction).
Also, the Nonhydrostatic Mesoscale Model (NMM) model, developed
by Zavisa Janjic (NOAA/NCEP) from the operational hydrostatic
ETA model, is being placed into the WRF framework to allow
comparison of the NCEP model with the existing WRF cores.
WRF Eulerian Mass Coordinate
During the past year, William Skamarock conducted
an extensive evaluation of the resolution capabilities
of the WRF Eulerian
mass-coordinate core (WRF-mass) by computing the kinetic
energy spectra for a large number of forecasts over a range
of model
resolutions. For this purpose, Skamarock generated
spectra from real-time 22-km CONUS forecasts together with
four- and
ten-km grid forecasts conducted in support of the BAMEX field
program.
|
|
| Figure 18: Kinetic
energy spectra computed from observations and from WRF-mass
model forecasts for three days in early
June 2003. Evaluation of the resolution capabilities
of the WRF Eulerian mass-coordinate core (WRF-mass) by
computing
the kinetic energy spectra for a large number of forecasts
over a range of model resolutions, shows that spectra
compare favorably with observations. |
These spectra compare favorably with
observations, as illustrated in Figure 18, which shows the
kinetic energy spectra computed from observations and from
WRF-mass
model forecasts for three days in early June 2003. The WRF-mass
model generally reproduces the observed spectra except at small
scales, where dissipation (damping) in the model becoming noticeable
for scales smaller than six to eight times the grid scale.
Working with Michael Baldwin (NOAA/National Severe Storms Laboratory), Skamarock also
computed spectra for the NMM model (currently being implemented
as a
WRF core within the WRF framework) and the operational NCEP
Eta model. The spectra show that the dissipation becomes noticeable
around 20 dx for the NMM and Eta. These spectra indicate
that the WRF-mass model has significantly greater resolution
capabilities than the operational models in their current configurations.
In addition to characterizing the resolving capabilities of
NWP models, kinetic energy spectra can also be used to examine
the nature of dissipation mechanisms in the models and to evaluate
mesoscale model spin-up. Skamarock investigated the effect
on the kinetic energy spectra of a variety of explicit dissipation
mechanisms that are commonly used in NWP, such as second order
filters and horizontal divergence damping, and demonstrated
that these filters remove significant energy at wavelengths
much longer than the grid scale. Skamarock’s analysis
also reveals the strong generation of mesoscale and cloud-scale
energy in high-resolution forecasts.
 |
| Figure 19: Kinetic
energy spectra for WRF-mass 1-24 h forecast started at
0Z 1 June 2003, averaged over 200-500 mb (10 mb increments). |
Figure 19 shows the kinetic energy
as a function of time in a ten-km BAMEX forecast. The initial
fields used by the model
are interpolated from a larger-scale gridded analysis, and
do not contain significant energy at length scales below approximately
300 km. Fortunately, high-resolution models can quickly generate
the missing mesoscale motions, and the spectra are effectively
spun up by 6 to 12 hours. Thus, even without mesoscale and
cloud-scale data assimilation, NWP models can generate the
small-scale structures, and evidence from the BAMEX simulations
suggests that these structures (in this case convection) are
frequently verified.
WRF Model Numerics
Work on model numerics continued as Skamarock examined
candidate advection schemes for WRF chemistry (WRF-Chem) applications.
The advection schemes in WRF-mass are neither monotonic nor
positive definite (PD), although they are fully mass conserving.
Multidimensional monotonic and PD schemes can be difficult
to construct and can be expensive. Skamarock and
Vincent Larson (University of Wisconsin, Milwaukee) implemented
a PD extension
of the RK3 advection schemes
that has produced significant improvements in test simulations
of moist Large Eddy Simulation (LES). For WRF-Chem applications,
more efficient PD and monotonic schemes are needed; timesplit
schemes have shown promise, but must be carefully coupled
to the mass continuity equation when incorporated into WRF.
WRF computational framework (top)
NMM Core
In a key development last year,
John Michalakes, together with Tom Black (NOAA/NCEP) and
Jon Wolfe (SCD), incorporated
the non-hydrostatic NMM dynamic core as a second option
along with the Eulerian mass-coordinate core. The NMM core
and
the suite of NMM physics were ported to the WRF Advanced
Software Architecture (ASF) and adapted for parallel
efficiency on distributed-memory computer architectures. Both
the
NMM and mass cores will be included in the ensemble configuration
NCEP will use in its first operational WRF implementation.
Grid Nesting Schemes
One-way and two-way grid nesting schemes have been implemented
in WRF by Michalakes, David Gill,
William Skamarock, and Jimy
Dudhia during
the past year. Nesting outside of the WRF model is handled
via
a separate program that generates initial and lateral boundary
conditions from a previous coarse-grid forecast for the subsequent
fine-grid WRF model. Nesting within WRF is being implemented
as two-way interactive domains that are mesh aligned and
coincident (that is, non-rotated and with each point in the
parent domain
corresponding to a point in the nest). Nests may telescope
to arbitrary depth and may be instantiated or destroyed on
the fly. John Smart and Brent Shaw (both from NOAA/FSL) are
working to adapt the Standard Initialization (SI) software
to provide the higher-resolution static fields required for
the nested grids.
Gill, working with Smart
and Shaw, also parallelized the mass core initialization
for real-data cases for distributed
memory (DM) architectures. Although the performance of the
new code on DM machines does not run noticeably faster, it
does allow users to build initial and lateral boundary files
for large domains by accessing aggregated memory from multiple
nodes. This upgrade also enhances the capabilities of the WRF
IO API so new arrays from the SI package may now be included
in WRF through the Registry, with no source changes required.
WRF in High-performance Computing
A major element of the WRF software design is to provide good
performance over a wide range of computing platforms. Through
efforts led by Michalakes,
WRF is now ported to and running routinely on seven of the
top 30 fastest high performance computers
in the world, as listed by the Top 500 Super Computing Sites
organization (as of June 2003, see www.top500.org),
including systems at NCAR, NOAA/NCEP and FSL, the Pittsburgh
Supercomputing
Center, NCSA (University of Illinois), and the Navy Oceanographic
Office. Ports are in progress to the NEC SX/6 system and
the Cray X-1. Recent WRF performance benchmarks in terms
of sustained
performance demonstrate good scaling across a large number
of processors, as shown in Figure 20, below.
|
|
| Figure
20: WRF software provides good
performance over a wide range of computing platforms and
is now ported to, and running routinely on, seven of
the
top 30 fastest high performance computers in the world.
These include systems at NCAR, NOAA/NCEP and
FSL, the Pittsburgh Supercomputing Center, NCSA (University
of
Illinois), and
the Navy Oceanographic Office. |
The understanding and prediction of geophysical systems have
moved beyond the capabilities of single-model simulation systems
into the area of multimodel, multiscale interdisciplinary systems
of interacting coupled models. Michalakes,
in collaboration with Robert Jacob (Argonne National Laboratory),
Matt Bettencourt
(University of Southern Mississippi), Daniel Schaffer (CSU/Cooperative
Institute for Research in the Atmosphere), and Christopher
Moore (NOAA/Pacific Marine Environmental Laboratory), developed
and demonstrated flexible, reusable software infrastructure
for
high-resolution
regional
coupling of WRF with ocean models and ecosystem models for
prediction of hurricane intensification, ecosystem and environmental
modeling, simulation of air quality and chemical dispersion,
and other problems of vital concern. The inter-model coupling
involves grid-to-grid translation, interpolation and other
computational transformations, and efficient communication
between component models running in single-program multiple-data
(SPMD) and multiprogram multidata (MPMD) modes. The implementation
of this coupling for WRF is based on the Model Coupling Toolkit
and the Model Coupling Executable Library.
|
|
| Figure 21: Inter-model
coupling allows understanding and prediction of geophysical
systems. Output from WRF, the Navy Coastal Ocean Model
(NCOM), and the
SWAN wave
model coupled for a simulation of a domain in the Gulf of Mexico
off the Louisiana and Mississippi coastlines. |
Figure 21, above, shows output from
WRF, the Navy Coastal Ocean Model (NCOM), and the SWAN wave
model coupled for a simulation of a domain in the Gulf of Mexico
off the Louisiana and Mississippi coastlines.
Advanced NWP Model Development
Collaboration with the Chinese Academy of Meteorological Sciences
(CAMS) of the Chinese Meteorological Administration increased
during the year as opportunities were explored for cooperation
in
advance NWP model development. Joseph
Klemp, Ying-Hwal Kuo, Michalakes, Skamarock, Dale
Barker, Jimy Dudhia, Jason Knievel, and
Shou-Jun Chen traveled
to Beijing in April to conduct the Second NCAR/CAMS Joint
Workshop on NWP Model Development and discuss
specific collaborative projects. As a result of these discussions,
the CAMS model-development group, led by Jishan Xue and DeHui
Chen (CAMS), adopted the ASF as the software
infrastructure for their new national NWP model called GRAPES.
WRF developers are also participating in the NASA-sponsored
Earth System Modeling Framework (ESMF) project to develop a
common software infrastructure for the development and coupling
of climate and weather models. Components of the WRF ASF, such
as the WRF I/O application program interface (API) are providing
a basis for development of ESMF functionality, and ESMF functionality
such as the ESMF time management utility is in the process
of being integrated into WRF. Michalakes is collaborating with
Nancy Collins and Cecelia Deluca (both of SCD) to integrate
the functionality of WRF and the ESMF. ESMF infrastructure
is now a part of the next scheduled WRF model release, and
additional ESMF infrastructure and superstructure will be incorporated
as they become available.
WRF
model physics (top)
Noah LSM Physics Package ported to WRF
For WRF to fulfill its objective as a scientific research
model, it is essential that it incorporate state-of-the-science
physical packages as they become available, and a number
of these have been ported to WRF over the past year. Jimy
Dudhia and
Fei Chen (RAP) incorporated the new Noah LSM into both MM5
and WRF. The Noah LSM represents development of a unified
version of the land-surface model by Kenneth Mitchell (NCEP),
Michael
Ek (NCEP), and George Gayno (AFWA) for use in future weather
prediction models, as well as in regional climate models
and scientific studies of surface processes. This new version
incorporates
improvements in frozen soil and fractional snow cover representations.
It has already been released to the MM5 community as part
of Version 3.6 and will be released to the WRF community as
part
of Version 2.
New Ice Crystal Physics Package
Dudhia also worked with
visitors Song-You Hong and Jeong-Ock Lim (both from Yonsei
University, South Korea) on a new microphysics
development for WRF. This parameterization incorporates a
new approach to representing ice crystal concentrations in
a single-moment
scheme. The scheme has undergone further development using
results provided from field programs by Andrew
Heymsfield, particularly
with regard to parameterizing ice and snow properties in
a gross sense as a function of temperature. Hong has continued
his collaboration with Dudhia in
developing a new planetary boundary layer scheme (YSU) for
WRF that explicitly represents
PBL top entrainment processes. Preliminary implementations
in WRF for real-time forecast efforts have shown some improvements
in soundings compared to the popular MRF scheme in MM5 and
WRF, and the YSU scheme is likely to replace the MRF scheme
in WRF in the future. Dudhia and Wei
Wang also collaborated
with Georg Grell (NOAA/FSL) to incorporate
a new ensemble cloud parameterization scheme in WRF. These
schemes are now being tested in preparation for release in
the community version.
Regional-scale Drizzle Forecasts
William Hall continued to
work with Roy Rasmussen (RAP) to improve the forecasting
of
freezing drizzle within regional
scale models that include MM5, WRF and RUC. He ported a
suite of microphysical parameterization schemes into the WRF
model
that had been previously developed and tested with the
MM5 model. The hierarchy of microphysical parameterization
schemes
range from more complicated ones that utilize two parameter
functions to represent each modeled species of the water
and ice particle spectrum to simple representations with
fewer variables that run efficiently in regional scale
models. The present research provides a systematic method to
test
the water and ice particle spectral assumptions used by
the lower order schemes against the successful higher order
schemes.
The present hierarchy is an extension of the Reisner scheme
that is currently installed in the MM5, WRF and Rapid Update
Cycle (RUC) models. The Reisner scheme has the same number
of variables as the present simplest approach.
The major problem of the original Reisner method is that it
overpredicted the development
of
snow and subsequently underpredicted freezing drizzle.
The present approach overcomes these difficulties by using
the
results from a successful sophisticated scheme to improve
the spectral approximations used in the lower order scheme.
Regional Climate Initiative
Ying-Hwa Kuo and Dudhia worked
with Ruby Leung (Pacific Northwest National Laboratory) in
developing a regional climate initiative for
the WRF model. This initiative is intended to strengthen
collaborations with SCD and the broader regional-climate research
community
to address both the downscaling (detailed regional climatic
effects) and upscaling (regional influences on global climate)
aspects of regional climate. Leung started some practical
implementations of necessary components in WRF that will allow
a specified
sea-surface temperature variation and a broader lateral boundary
zone suitable for use with climate models. This work is continuing
with implementation of climate-model physics into WRF for
better coupling.
WRF
idealized and case-study testing (top)
WRF Testing of Simulated Squall Lines
In collaboration with George Bryan (ASP), Todd Lane (RAP),
and Matthew Parker (University of Nebraska) Jason
Knievel began
exploring how changes in vertical wind shear alter the water
budgets of simulated squall lines. This ongoing project has
two goals. The first is to assess whether previously published
conclusions about the relationships between wind shear and
rain and cloud in simulated squall lines might depend on
the numerical model used for the simulations, while the second
is to test the WRF Model against other more established models.
These tests have already led Knieval and Bryan to make two
important changes to model code. The first change was in
the
TKE scheme, which uses turbulent kinetic energy to parameterize
unresolved turbulence. The second change was in the Kessler
microphysics scheme, which parameterizes the formation of
clouds and precipitation. The WRF model now produces more
realistic
simulations and runs more efficiently.
The diurnal and semi-diurnal cycles of rainfall frequency
are fundamental to most regional climates. The ability of the
WRF model to simulate these cycles is thus an informative measure
of the model’s forecast skill. Knievel,
along with David Ahijevych and Kevin
Manning,
evaluated the 22-km and 10-km WRF model's diurnal and semi-diurnal
modes of summer rainfall across the conterminous
United States in comparison with the observed patterns obtained
from rain gauges and radars. Preliminary results suggest that
the WRF model simulates the diurnal and semi-diurnal modes
of rainfall frequency well in some places, but not others.
In particular, the WRF model does not capture the predominance
of nocturnal rainfall in the Great Plains
WRF use in Large Eddy Simulation (LES)
In collaboration with by Gino Serafini (University of Rome,
Italy), William Skamarock evaluated
the WRF model for Large Eddy Simulation (LES) applications.
They tested the two canonical problems
for LES codes, which are dry boundary layer growth by free
convection and by vertical wind shear. The formulation for
the prognostic TKE parameterization in WRF was originally
developed for LES applications, and WRF reproduced the correct
boundary
layer growth, turbulent fluxes, and kinetic energy distributions
as documented in published results from existing state-of-the-art
LES models.
WRF experimental real-time forecasting
WRF experimental real-time forecasting supports BAMEX field
operations
WRF real-time forecasts were continued this year for a number
of projects. The overall goals were to evaluate the model
under a variety of weather conditions and to test the model's
robustness.
Specific tests were run this year to evaluate different configurations
of the model, different initial and boundary conditions from
Eta, GFS and RUC analyses, and new physics packages, such
as the Noah LSM and the YSU PBL. A number of configurations
of
the model have been run by Wei
Wang,
and results posted on the Web by James
Bresch (http://rain.mmm.ucar.edu/mm5).
In addition to the configurations run in past years, Wang added
a new 22-km CONUS domain run in February that uses NCEP's GFS
analysis and forecast data to initialize the model. This
run is made once a day from 0000 UTC only. The precipitation
forecasts from this run have been verified with the assistance
of Michael Baldwin (NOAA/NSSL) since June 2003. Because of
limited cases being verified over the last couple of months,
it is
too early to draw general conclusions. But it is apparent that
the runs from the two different initial and boundary conditions
often generate precipitation forecasts that are noticeably
different, especially in the second day of the 48-h forecast.
For three months starting in June, another 22-km CONUS domain
was added to the real-time suite to test the newly implemented
Noah LSM, and LSM initialization data from Eta and AFWA's AGRMET
data. No significant difference was found between the runs
initialized from the Eta and AGRMET LSM data.
To facilitate the NCEP WRF test plan, Wang implemented
a configuration designed to mimic the NCEP High-Resolution
Window domain for
central US that was run twice a day for the month of September
on SCD's IBM/bluesky. The main focus
of the test is the use of 20-km operational RUC data for initial
conditions and an increase in the number of vertical levels
from 35 to 51. With the implementation of a selective constraint
on the maximum vertical motion developed by Jimy
Dudhia, William Skamarock, and Joseph
Klemp,
the WRF model remained stable during the testing period without
any reduction in the time step.
In a groundbreaking real-time forecasting experiment, Morris
Weisman, Jordan Powers, Wang, James Bresch, and Christopher
Davis coordinated and executed real-time
forecasts using the WRF model in support of the BAMEX field
operations. During the
BAMEX field phase from 19 May to 6 July, the four-km WRF with
no convective parameterization was run once a day within a
2000 x 2000 km2 central U.S. domain from a 0000 UTC initialization.
For the purposes of comparison and weather update, a ten-km
WRF domain with the Kain-Fritsch convective parameterization
was also run twice a day. All forecasts ran for 36 h and were
initialized only with the 40-km Eta model analysis. These real-time
WRF runs provided field scientists and forecasters with valuable
information about convective outbreaks that was highly beneficial
in planning daily field operations. Striking differences were
apparent between the four-km forecasts with an explicit treatment
of precipitation and the ten-km forecasts with cumulus parameterization.
|
|
| |
| Movie 3: Groundbreaking
real-time forecasting using the WRF model in support of
the BAMEX
field experiment was beneficial in planning daily field
operations. On the left is the WRF model forecast for June
9 - 10, as compared to actual radar from the same time
period. |
Mouse over image
to begin movie. Alternately, you may download
the animation. |
The four-km forecast realistically
represented different modes of convection observed, such as
the case of 9-10 June shown in Movie 3. In contrast, the ten-km
forecast for this period (not shown) produced widespread regions
of relatively light precipitation instead of well-defined mesoscale
convective systems. Overall, the four-km WRF also showed a
remarkable ability to indicate correctly the number of MCSs
on any given day, their approximate location, and the most
severe mode of convection among them. These model runs were
made using 128 processors on SCD's IBM/bluesky under a special
project account.
In September, Wang configured ten-km real-time WRF model runs
to capture tropical storms near the U.S. coastline. When Hurricane
Isabel strengthened to category 5 in the South Atlantic, Klemp,
Powers, Wang, and Bresch initiated more intensive real-time
forecasts for the hurricane, under a special arrangement with
NCAR Director's Office and SCD. The ten-km WRF runs were extended
from a 36-h run to a 5-day run to forecast hurricane Isabel's
movement. A four-km model was also run a number of times the
following week as the storm approached landfall.
|
|
| Movie 4: Comparison between
the simulated maximum reflectivity (a 48-h loop) from the four-km
WRF (right) with the observed ground-based radar reflectivity
composite (left). The four-km forecast accurately predicted
the location and
timing of landfall (41 h into this forecast) and produced realistic
reflectivity structures in the eyewall and convective rainbands. |
Mouse over image
to begin movie. Alternately, you may download
the animation.
|
Movie 4 shows the comparison between
the simulated maximum reflectivity (a 48-h loop) from the four-km
WRF with the observed ground-based radar reflectivity composite.
The four-km forecast accurately predicted the location and
timing of landfall (41 h into this forecast) and produced realistic
reflectivity structures in the eyewall and convective rainbands,
explicitly representing individual convective cores without
need of a cumulus parameterization.
WRF
model data assimilation (top)
Collaboration on the WRF 3DVAR Data Assimilation System
In a highly collaborative effort Dale
Barker, Yong-Run Guo, Wei Huang, and Qingnong
Xiao worked with other WRF partners
from NOAA/NCEP, AFWA, and NOAA/FSL to further develop and
test the WRF 3DVAR data assimilation system. The system already
contains a variety of capabilities not present in other 3DVAR
systems (e.g., choice of background error covariance model,
portability to a variety of platforms, and efficient parallel
scaling). Barker also played a significant role in the coordination
of this effort and in testing for selected mesoscale and
CONUS
applications.
As noted in the Advanced Data Assimilation Systems for
Community Use section, in June, a combined MM5/WRF three-dimensional
variational (3DVAR) data
assimilation
system
was released to
the research
community following several years of development. Barker coordinated
the release and created a web page contains links to the software,
documentation, etc. The 3DVAR system was released to both MM5
(http://www.mmm.ucar.edu/3dvar)
and WRF (http://www.wrf-model.org/wg4)
communities. An initial tutorial was held at NCAR in conjunction
with the release, attended by 45 researchers from 31 institutions
worldwide. Barker, Guo, Huang, and Syed
Rizvi (visitor from
India) lectured at the one-day tutorial. A second tutorial
was requested by members of the Italian data assimilation research
community, and was presented over two days in L’Aquila,
Italy in July by Barker.
WRF community support (top)
WRF User Statistics
Since the first version of WRF was made available to community
in December 2000, over 1400 users have registered to download
the code. Over half of these prospective users are from foreign
countries. The most recent version (beta-version 1.3) of
the WRF model was released in March 2003, and an updated WRF
SI
(version 1.3.1) was released in June 2003. The WRF User page
has been updated to provide current information about the
modeling system. There are about 600 people who have subscribed
to the
WRF email list. The number of user emails continues to increase.
During FY03, Wei Wang and
other MMM staff answered 680 user emails, which is an 88%
increase over the previous year.
Wang, Joseph Klemp, and Jimy
Dudhia organized
the Fourth WRF Users’ Workshop,
which took place 11-13 June, immediately following the MM5
Users' Workshop. A joint session of the MM5 and WRF Workshops
was held in the afternoon of 11 June, where 30 papers of the
interest to participants were presented. About 136 people from
81 institutions participated. Among them, 25 people were from
foreign countries. The joint session highlighted the growing
use of WRF in a wide variety of research applications.
A two-and-a-half day WRF tutorial was offered 16-17 June,
following the WRF Users' Workshop. Sixty-two people participated
in this tutorial (nine from foreign countries). Dudhia,
John Michalakes, Shu-hua Chen (University
of California, Davis), Gill,
Cindy Bruyere, and Wang, along
with Brent Shaw, John Smart, and Paula McCaslin (all NOAA/FSL),
lectured at this WRF tutorial.
Developmental Testbed
Center and testing for WRF operational implementation (top)
Collaborative Development of New Developmental Testbed Center
The effort to create the Developmental Testbed Center (DTC)
at NCAR began in earnest this year. The DTC concept is that
of a facility where new NWP techniques can be evaluated in
an operations-like setting, without interfering with current
operations. It is intended to provide for a rapid and direct
transfer of new NWP research advances into operational forecasting
and to evaluate strengths and weaknesses of both experimental
and operational systems. The Center will provide a rigorous
environment for modeling system development and stringent
code qualification and should significantly decrease the time
for
getting new modeling capabilities into operations. The focus
will initially be on WRF-based regional NWP and will then
expand to encompass global models. The DTC Facility will be
co-located
with NCAR and will host a strong visitor program to promote
community involvement. Robert
Gall, Christopher Davis, Joseph Klemp, and Jordan
Powers,
together with FSL, NCEP, and AFWA, developed an implementation
plan for the DTC and coordinated the startup activities for
the Center. In the early work performed through the DTC, John
Michalakes, Jimy Dudhia, James Bresch, Wei Wang, and Louisa
Nance addressed a broad range
of WRF system preparation tasks and model runs for the WRF
Test Plan (discussed below). In August, Tim Killeen, NCAR
Director, appointed
Gall to be the director
of the facility.
The WRF Test Plan was formulated by NCEP, NCAR, FSL, and AFWA
as a collaborative effort to evaluate the initial operating
capability for the ensemble-based WRF configuration. NCEP will
begin running operationally in its High-Resolution Domains
in October 2004. The plan defines a course of systematic testing
for the WRF-Mass core and WRF-NMM cores and ensembles to be
constructed from them. The testing requires a large number
of model runs, primarily in the form of simulations over four
month-long periods (called retrospective runs). Tasks included
developing and preparing WRF SI, porting the NCEP NMM to the
WRF software framework, determining the configuration of physics
for the runs, comparing results from various computing platforms,
and performing the retrospective runs. This effort tangibly
advances the goals of the nascent DTC, provides the testing
required to configure the first operational version of WRF
at NCEP, and represents an achievement of cooperation among
NCAR, FSL, NCEP, and AFWA.
MM5 development and community
support (top)
Antarctic Mesoscale Prediction System (AMPS)
Kevin Manning, Jordan Powers, and Michael
Duda continued their
support and advancement of the Antarctic Mesoscale Prediction
System (AMPS). Work this year included the addition of new
top boundary conditions devised by David Bromwich and his
colleagues (Ohio State University, Byrd Polar Research Center).
Use of
the new top boundary condition has significantly reduced
model error, particularly in the upper troposphere where
errors had
been particularly high. Jimy Dudhia added
a set of polar physics to the AMPS in MM5 for the user community
model (V3.6). These
modifications come largely from Bromwich and John Cassano
(University of Colorado) and improve the radiation, surface
physics and sea-ice fraction
representation
in the model for polar regions. Other work, such as expanded
grids and additional computational grids, have increased
the utility of AMPS to forecasters for NSF Operations and
to Antarctic
researchers throughout the international community.
MM5 Performance Results
The distributed-memory parallel implementation of the Penn
State/NCAR Mesoscale Model (MM5) and a standard benchmark
case were made widely available to various high-performance
computing vendors.
John Michalakes has collected
performance results and made them available on-line to the
MM5 community and the high-performance
computing community at large at http://www.mmm.ucar.edu/mm5/mpp/performance .
New contributions to the MM5 Parallel Benchmarks page in
2003 included the new X1 system from Cray and vendors offering
systems that use the new line Intel IA-64 processors on Linux.
MM5 Users Increase
The number of MM5 users increased significantly in 2003. At
the end of this fiscal year, there were 950 users on our mailing
list from about 500 institutions worldwide. This represents
an 18% increase over last year. Over 2390 emails from users
were addressed. Version 3-6 of the MM5 modeling system was
released in December 2002. The new features in this release
included the new Noah LSM, which replaced the OSU LSM in previous
MM5 V3 versions, and modification to several physics packages
for use in polar regions, such as the Antarctic. Two minor
releases, 3-6-1 and 3-6-2, were made in March and August of
2003.
MM5 Users Workshop
The Thirteenth Annual MM5 Users' Workshop was held on 10-11
June with 78 participants (23 from foreign countries) from
45 institutions worldwide. Forty papers were presented at
the workshop. Most papers from the workshop are made available
from the MM5 home page at http://www.mmm.ucar.edu/mm5/workshop/workshop-papers_ws03.html.
The users' workshop is a forum for the MM5 model developers
and users to exchange and discuss new developments and applications
of the model. Special themes for this year's workshop were
MM5 3DVAR and its applications and the transition from the
MM5 modeling system to the WRF modeling system. Two MM5 modeling
system tutorials were offered this year, one in January,
and the other in August. A total of 61 participants (21 from
foreign
countries) from 45 institutions worldwide attended the classes.
Dudhia, David Gill, Manning, Cindy
Bruyere, Dale Barker, Yong-Run
Guo, and Wei Wang lectured
at the tutorials.
MM5 Enhancements
Enhancements to the global MM5 system included improvements
to the graphics, which greatly increased the number of viewers
of our real-time global forecasts, particularly in Europe.
Global forecasts were supplied in real-time to researchers
at RAP to aid in their model visualization projects.
Routine verification of WRF and MM5 was conducted using the
verification system developed by James
Bresch and Duda. The
availability of the verification statistics on teh Internet
allows forecasters and researchers to assess model performance
in real time.
|
