FOR MULTIPLE APPLICATIONS
MODELS ACROSS DISCIPLINES
Hurricane Andrew slammed into the southeast coast of Florida in
1992, its central eye pressure plum-meted unexpectedly to the
third lowest recorded this century for a hurricane at U.S. landfall.
With this sudden pressure drop, Andrew intensified to a maximum
sustained wind speed of 125 knots (145 mph) and gusts in excess
of 150 knots (175 mph). The category-four hurricane led directly
to 26 fatalities and became the most expensive natural disaster
in U.S. history, causing property damage estimated at $25 billion.
Joseph Eastman, Mel Nicholls, and Roger Pielke, Sr., of Colorado
State University (CSU) have been probing the dynamics of Andrew
and other hurricanes in unprecedented detail by running the Regional
Atmospheric Modeling System (RAMS) on NPACIs Blue Horizon.
These and other RAMS simulations are advancing atmospheric science
in a variety of areas. The work is also revealing unexpected relationships
between factors such as ocean warming and hurricane behavior,
as well as land use and climate change.
Simulation of Hurricane Georges at landfall on Puerto Rico
RAMS simulation examines the effects of sea surface temperature
on hurricane intensity and effects of topography on a hurricane
at landfall. The model is in reasonable agreement with recorded
data and captures details of Georgess eye. The simulation
uses a 5-km horizontal mesh with more than 500,000 simulation
cells. Environmental WorkBench (EWB), courtesy of SSESCO,
Inc., generated the graphics. The animation is available
"RAMS could be
seen as a sort of Swiss Army Knife of atmospheric
models," said Eastman, a research associate in the Department
of Atmospheric Science at CSU. Scientists are using this sophisticated
modeling system to simulate events from cloud dynamics, flash
flooding, and hurricanes to large-scale weather systems and climate
studies. Simulations can be as local as wind flow around urban
buildings or extend to global scales. "RAMS is even being
used in prototype, real-time weather forecasting, which it can
do very accurately if enough computer resources are available,"
He was the first to
apply the power of a mesoscale model (covering scales from the
area of a city to a continent) to hurricanes. For the first time,
these simulations were able to resolve features as detailed as
the sloping eye wall of Andrew and the rapid intensification event,
a 42-millibar (mb) drop in central pressure to 922 mb in just
24 hours. The rapidly plunging pressure was a crucial factor in
the hurricanes destructiveness and is a dangerous phenomenon
that conventional forecast models do not predict well.
Perspective view of Hurricane Georges
this rotated view of Figure 1, the vertical is stretched
by a factor of 11 to highlight flow over the mountains and
show hurricane size relative to clouds over the Dominican
Republic. Note developed eye and bump of overshooting cloud
tops, marking area of most intense convection. Graphics
generated with Environmental WorkBench courtesy of SSESCO,
"In our simulations,
we start the model off using the weather data recorded at a certain
time for the actual hurricane, and then let the computer model
of the storm spin up, which takes about six simulated
storm hours," said Eastman. The hurricane simulation is integrated
forward in time for an additional 42 hours. The realism of RAMS
simulations comes from its greater reliance on fundamental physics,
with fewer approximations. "By including properly resolved
microphysics, we obtain better predictions of clouds, rain, hail,
and so forth, leading to a realistic thermodynamic profile,"
said Eastman. "These feedbacks are very important in hurricane
behavior, and you really have to include them."
Eastman recently used
RAMS to model Hurricane Georges. In addition to extensive property
damage, authorities attributed 602 fatalities to the 1998 category-four
hurricane, mainly in the Dominican Republic and Haiti. In the
Dominican Republic, Georges dropped 39 inches of rain in 24 hours,
devastating the island with flash floods and mud slides.
models operate at a horizontal grid spacing of 20 km or greater,
and many important phenomena slip through the cracks," said
Eastman. The computational power of Blue Horizon enabled him to
run RAMS at a 5-km grid spacing for Georges. At that finer resolution
there are 16 grid points in each of the previous larger cells,
which enables the model to resolve more details of the storm (Figures
1 and 2).
Effects of clouds and rain on photosynthesis
simulation couples RAMS to the University of Kansas General
Energy and Mass Transfer Model. Green and yellowish patches
are areas of photosynthesis enhanced by previous precipitation,
while darker blue areas indicate decreased photosynthesis
due to shade and precipitation from passing clouds.
The researchers are
focusing on the effects of topography and vegetation on tropical
cyclones, which is important when these storms make landfall and
do much of their damage. To explore this, Eastman ran RAMS simulations
with and without topography included. "This lets us form
hypotheses, test them, and run the simulation again," he
said. "Were learning more about basic hurricane science,
and at the same time moving toward better hurricane predictions."
Eastman is running
simulations of Georges at an even finer horizontal grid spacing
of 1 km. "This is approaching a simulation that can resolve
large turbulent eddies in the atmosphere," he said. "We
think it will give better predictions of precipitation, which
has always been challenging to model." For simulations of
Hurricane Georges at 5-km grid spacing, RAMS ran on 64 of Blue
Horizons 1,152 processors. The 1-km simulations require
more computation, with 25 points in each previous 5-km cell as
well as requiring shorter time steps. They run on 256 or more
Eastman and his colleagues
have also examined the possible effects of ocean warming on hurricanes.
It has been widely speculated that a sea-surface warming due to
global climate change might produce more intense hurricanes with
higher wind speeds. The researchers were surprised to find that
including a 2°C ocean warming in the simulation did not produce
more intense hurricanes; however, it did result in significantly
increased precipitation. Since excessive rainfall often causes
more damage than high winds, these results raise the possibility
of more destructive hurricanes in a warmer climate.
FOR MULTIPLE APPLICATIONS
RAMS is an easy-to-use
mesoscale model. It has a good parallel interface and runs on
different platforms from Linux and AIX to Windows 2000. Based
on a multiple grid nesting scheme, RAMS provides options that
allow it to be tailored to many different problems.
"RAMS is capable
of modeling not only hurricanes, but also a wide array of other
atmospheric phenomena at a level of detail never before possible,"
said Pielke, a professor in the Department of Atmospheric Science
at CSU. "RAMS simulations are providing important research
insights into the basic dynamics of weather and climate. And they
are also producing information that has practical and predictive
RAMS had its beginnings
at CSU in the early 1980s. Researchers at CSU and the Mission
Research Corporation are refining it in step with greater computing
power, incorporating both more realistic physics and greater flexibility.
The accurate physics
and sophisticated radiation and land-surface parameterizations
included in RAMS require greater computing resources than simpler
models. Initially this limited the model to running at regional
scales, but as the power of computers has grown, the realism of
RAMS has allowed it to scale upward until it is now being tested
in a global version.
MODELS ACROSS DISCIPLINES
As RAMS has become
more scalable and realistic, researchers have combined it with
other models. Eastman is linking the weather and climate modeling
power of RAMS to land cover and biodiversity models in collaboration
with NPACI Earth Systems Science partners at SDSC, the University
of Kansas, and the Long-Term Ecological Research (LTER) program.
With computer scientist
Tony Fountain of SDSC, Eastman has run RAMS in one-day integrations
on a 10-km horizontal grid, using a combined atmosphere-ecosystem
model. "We know that atmosphere and biosphere are interrelated,
but modeling this has been a challenge," said Fountain. "Few
ecosystem models include dynamic atmospheric feedbacks, so were
very interested in the results of this coupled model."
For example, the model
can simulate changes in photosynthesis due to cloud cover or rainfall
(Figure 3). Eastman also is working with Bob Waide, LTER Network
Office executive director at the University of New Mexico. They
are analyzing regional vulnerabilities to such environmental influences
as landscape change. They are working on a yearlong run with RAMS
and a coupled plant model on a 10-km grid that covers the U.S.
In previous runs on a 50-km grid, the researchers found that changes
in land use since European settlement--deforestation and different
vegetation--can make just as much difference in local and regional
climate change as doubling the atmospheric concentration of the
greenhouse gas carbon dioxide. "This result was a big surprise,"
said Eastman. "Everyone tends to assume that global warming,
as characterized by a predicted doubling of carbon dioxide, is
the first and main human influence on climate."
RAMS model output has
also been used to drive a biodiversity model developed at SDSC
called the Genetic Algorithm for Rule-Set Prediction (GARP) in
runs on Blue Horizon. The researchers are trying to understand
how climate change may affect biodiversity in simulations that
span several hundred years. "In these simulations, were
going from 50 km to a much finer 10-km grid," said Eastman.
"This is a very useful step because well begin to resolve
mountain, valley, and other mesoscale circulations." These
atmosphere dynamics are important to local climate, which influences
species occurrence. Resolving such phenomena will help researchers
model biodiversity more reliably.
RAMS is also being
used in coupled atmosphere-ocean simulations, linked to local
pollution models in air quality research and used in other research
applications. "Its amazing what were able to
do now," said Eastman. "Five years ago, I didnt
think these things would be possible, but with the power of Blue
Horizon, its feasible to run high resolution simulations
that have never been done before. This evolution keeps teaching
us new things." PT
Colorado State University
Roger Pielke, Sr.
Colorado State University
Colorado State University
A. Townsend Peterson
University of Kansas
LTERUniversity of New Mexico
J. L., M. B. Coughenour, and R. A. Pielke, 2001: The effects of
CO2 and landscape change using a coupled plant and meteorological
model. Global Change Biology, in press.
Pielke, R.A., Jr. and
R.A. Pielke, Sr., 1997: Hurricanes: Their Nature and Impacts on
Society. John Wiley and Sons, England, 279 pp.
Pielke, R. A., W. R.
Cotton, R. L. Walko, C. J. Tremback, W. A. Lyons, L. D. Grasso,
M. E. Nicholls, M. D. Moran, D. A. Wesley, T. J. Lee, and J. H.
Copeland, 1992: A comprehensive meteorological modeling systemRAMS.
Meteor. Atmos. Phys., 49, 69-91.