|
The Weather Research and Forecasting (WRF) model was used to predict
the course of the historic March
2003 Colorado snowstorm. WRF is used by scientists in a variety
of research efforts, a number of which utilize the Advanced Research
Computing System (ARCS), a set of supercomputers managed by SCD.
SCD also supports research in the following areas:
|
Weather forecasters
may soon do a better job handling snowstorms like the one that paralyzed
Colorado's Front Range in March 2003, according to an analysis by UCAR,
the National Oceanic and Atmospheric Administration (NOAA), and colleagues
in the private sector.
According to the study team, the fine detail from a new generation
of computer forecast models may help forecasters convey important town-to-town
differences in snowfall timing and intensity, like those observed during
the
historic storm on 16–20 March 2003. These new tools—such
as the Weather Research and Forecasting model, to be implemented
later this year by the National Weather Service (NWS)—will help
meteorologists include more detail in public forecasts.
Using computer models to predict the storm
Overall, forecasters did a good job alerting the public to last year's
storm, according to the study's leader, Douglas Wesley (UCAR Cooperative
Program for Operational Meteorology, Education and Training). Computer
models indicated the storm's general approach up to a week in advance.
Then, within two days of its arrival, they showed the potential for
over four feet of snow along Colorado's Front Range. "The accuracy
of these model projections, and the public forecasts issued by the
NWS, were perhaps unprecedented," says Wesley.
But neither human nor virtual forecasters predicted how dramatically
the storm would vary across the Front Range. "We found differences
of several feet of snowfall at similar elevations in a span of 15 miles
or less," says Wesley. Just west of Longmont and Loveland, for
instance, only three to six inches of wet snow accumulated. Meanwhile,
at the same elevation on the south and west sides of Boulder and Denver,
three to four feet of snow piled up.
The huge disparities occurred in part because of subtle temperature
differences. As the massive storm spun in a counterclockwise direction,
says Wesley, "relatively warm air descended from the canyons of
Boulder and Larimer County, while cold northerly surface winds were
blocked by the terrain." This helped keep readings at or just
above freezing downstream from the canyons, limiting the snowfall accumulations
there.
Finer-scale models
In large-scale computer models currently used by the NWS, the atmosphere
is depicted at points separated horizontally by more than 10 miles.
This is much too great a distance to capture the small-scale variations
in temperature and precipitation that shaped the March 2003 snowstorm.
To get a better view, the UCAR-NOAA team put two finer-scale research
models to the task, each with less than 2 miles between grid points.
The team fed actual data from the storm into these two models, then
watched how well the models depicted the storm's evolution.
One of the key elements in heavy Front Range snowfall is upslope
flow, where moist winds blow from east to west. Along with their sharper
depiction of the mountains themselves, the finer-scale computer
models did a better job depicting the upslope flow that slammed
into the Front Range, says Wesley. About a mile above Denver, the standard
computer model showed east-to-west flow of 15-20 miles per hour during
the height of the storm. The two finer-scale models showed speeds of
20-25 mph, closer to the observed speed of 30-35 mph.
"The models' abilities to capture the depth and strength of
the upslope flow are likely critical to their forecasts of low-level
temperatures and precipitation," says Wesley. However, he notes,
both models failed to predict the full extent of the warmer air at
ground level, where light downslope winds kept some areas at or above
freezing, even as howling upslope winds fueled snowfall just above
the surface.
"With the ongoing advances in computing speed that we are currently
seeing," says Wesley, "forecast models with horizontal grid
spacing of less than five miles are becoming a realistic expectation
in a 5- to 10-year timetable. These enhanced models will become
even more important tools for forecasters in complex terrain anywhere
in the world."
For more information, contact Anatta at UCAR Communications (anatta@ucar.edu,
303-497-8604); or Doug Wesley at UCAR/COMET (wesley@ucar.edu,
303-497-8337).
SCD's support of atmospheric research
The Scientific Computing Division operates and manages NCAR's Advanced
Research Computing System (ARCS), which is used by researchers in the
international geosciences community to run the Weather Research and
Forecasting Model and other computationally intensive models such as
the Community Climate System Model, the Penn State/NCAR Mesoscale Model
5, the Model for OZone And Related chemical Tracers (MOZART), and the
Parallel Climate Model.
SCD's goal is to enable the best atmospheric research in the world
by providing and advancing high-performance computing technologies.
SCD offers computing, research data sets, data storage, networking,
and data analysis tools for NCAR users.
For more information on SCD's support of NCAR supercomputers,
contact SCD's Digital Information Group (dig@ucar.edu).
|
