OSU WEATHER FORECASTING TOOL OUTPERFORMS COMPETING MODELS
COLUMBUS, Ohio – The dramatic rescue of a physician from a remote science station at the South Pole two years ago provided researchers here with an opportunity to test how well current weather forecast models actually perform.
The results of this study ranked four current models. They suggest that the refinement of one of them might even support longer Antarctic field seasons and make potential rescues from that distant part of the world much safer.
The study, published in a recent issue of the journal Weather and Forecasting, gave high marks to a computer model developed largely by Ohio State University scientists.
Graduate student Andrew Monaghan and David Bromwich, a professor of geography, were able to retrieve data sets produced by the four forecast models for the 10-day period surrounding the retrieval of Ronald Shemenski, a physician stationed at the U.S.’s Admundsen-Scott South Pole Station in April, 2001.
Shemenski had developed a serious pancreatic infection and had to be airlifted from that remote outpost in an unprededented late season landing. Normally, no flights can land at South Pole at that time of the year because of extreme cold, high winds and near 24-hour darkness. Researchers assigned to South Pole station agree to “winter over” with no expectation of outside help.
The year before, another South Pole-assigned physician, Jerri Nielsen, diagnosed her own case of breast cancer and had to be evacuated early in the Antarctic spring, much sooner than flights normally land at the South Pole Station.
After the Nielsen episode, the National Science Foundation sponsored a scientific workshop at Ohio State to assess and improve current climate models which in turn led to gauging their efficacy in such emergency forecasting situations.
Bromwich and Monaghan looked at data produced by the Antarctic Mesoscale Prediction System (AMPS), the National Centers for Environmental Prediction Aviation Model (AVN), the European Centre for Medium-Range Weather Forecasts (ECMWF) global forecast model and the National Center for Atmospheric Research Global Mesoscale Model.
Aside from AMPS, all of the models were designed for global forecasting. Because of that, they work fairly accurately in the mid-latitude regions but tend to work poorly in polar regions. On the other hand, AMPS uses the Polar MM5, a forecasting model optimized by OSU scientists for use over ice sheets. The researchers and NSF were interested in which model provided the best polar forecasts for the Shemenski rescue period.
“Basically, no one knew how well the four models could perform in this circumstance,” Bromwich said. “We wanted to find out how well they predicted the weather aspects critical to landing an aircraft at the bottom of the world.”
At that time of year, temperatures at the pole can routinely reach minus-50 degrees C and winds can kick up blowing snow. The combination of this spindrift and day-long darkness makes it nearly impossible for pilots to see the horizon so landings become extremely dangerous.
The researchers were looking at the data from the four models to determine which best predicted the conditions that actually had existed near the Pole at the time of the rescue.
“Both the AMPS and ECMWF models did about the same,” Bromwich explained, “Unfortunately, the European model is routinely available only to European states. The American AMPS model is available to anyone.”
Monaghan said that while the forecasting models are designed to predict short-term future weather, they are equally important for understanding long-term climate trends.
“It is very important to keep tuning the Polar MM5 model and make it better,” he said. “Several attempts that used other models to reconstruct climate over the past 40 years turned out to be very weak in the polar regions.
“We’re convinced that if a model doesn’t get the polar regions right, it won’t get the global climate right either.”
Long-term climate studies aside, officials at NSF have voiced repeated interest in extending the limited summer research season, Bromwich said. This may soon be plausible with the recent implementation of AMPS. The model is run twice daily by the National Center for Atmospheric Research and is specifically designed to support NSF’S Antarctic operations.
“Our goal is to get our year-round forecasting skill for the polar regions to a level that matches our forecasting skill for the American Midwest in winter,” Bromwich explained.