UCAR > Communications > Staff Notes > March 2000 Search


March 2000

Could busted forecast be a boon for weather research?

Snowfall amounts varied widely across a narrow corridor in the snowstorm of 24-26 January. For instance, central North Carolina got upwards of 20 inches, while the east and west ends of the state saw no snow at all. (Illustration courtesy NWS.)

The timing couldn't have been worse. Just days after their new IBM supercomputer was formally commissioned, forecasters at the National Weather Service found themselves struggling to predict a fast-moving, unexpectedly intense winter storm. Cities from South Carolina to Pennsylvania got far more snow than expected on 24-25 January, and NWS meteorologists found themselves at the center of a media storm.

Bob Gall and Chris Davis (MMM) traveled to Washington, D.C., last month for a congressional briefing on what happened in January and why. They were joined by Louis Uccellini, head of the National Centers for Environmental Prediction (NCEP), and Lance Bosart (University at Albany, State University of New York). The standing-room-only briefing on 17 February was sponsored by two representatives from North Carolina, Republican Howard Coble and Democrat David Price, and organized by UCAR's Office of Development and Government Affairs.

As it turns out, the U.S. Weather Research Program is already hot on the trail of some of the problems that led to the missed forecasts in January. Bob, who is currently dividing his time between roles as MMM director and USWRP lead scientist, moderated the discussion and closed by making a case for continued USWRP support.

"It's a natural outcome of the discussion," says Chris. "The kinds of things that screwed up the forecast--the reasons the models didn't do very well--were things that the USWRP wants to address."

Among these issues are data assimilation, including starting the model with accurate initial conditions: "You have to get the precipitation right early to get it right later." The 24-hour forecast issued Sunday evening, 23 January, didn't accurately represent widespread showers and thunderstorms across Florida and Georgia, says Chris. Moreover, one of the radiosonde reports from the region that evening was rejected, and the winds were missing from another. "Both of these soundings would likely have been critical for defining the jet stream near the region of convection."

Chris Davis. (Photo by Carlye Calvin.)

Through grants at NCAR and elsewhere, the USWRP is working on better assimilation techniques, Chris says. "The main issue is blending various kinds of observations to produce better initial conditions." Among satellites, radiosondes, and surface observations, "you have a very disparate set of data, and you have to come up with a clever way of combining it." Since the meteorological qualities--temperature, wind, humidity--described by the data are interdependent, "you also have to worry about not messing up the underlying links."

By the morning of the 24th, the models had picked up on the intense precipitation across the Southeast and were now projecting intense rain and snow up the East Coast over the next day. Previous storms this winter had stayed further offshore than models had predicted, but this one pulled closer to the coast, bringing much heavier precipitation to population centers than expected.

According to Uccellini, "The models confined the precipitation to a location east of the major cities until the 18Z runs [1:00 p.m. EST Monday]. So the issue really boiled down to [precipitation amounts] from the models, which were very deficient. This is a major issue that we're trying to resolve within the USWRP."

North Carolina's interest in the briefing was understandable. The official forecast for Raleigh-Durham at 5:00 p.m. Monday was for 1 to 2 inches of snow. By midday Tuesday, the area had received over 20 inches (0.5 meters)--not only their largest single snowfall on record, but the highest total they'd ever received in one month. The outlook for Washington, D.C., was revised Monday night to call for more intense snowfall, but many residents missed the late-night newscast and were surprised by snowfalls ranging from 6 to 20 inches across the D.C. metroplex. The higher amounts were among the biggest for the D.C.-Baltimore area since 1983.

Some parts of this forecast nightmare were simply due to timing, Chris points out. The scenario for D.C. became clear by around 6:00 p.m., a shade too late for early-evening newscasts. Further south, the Carolinas' heavy snow was triggered by a burst of convection, including "thunder snows," that likely altered the track and intensity of the surface cyclone.

Chris sees some parallels to the Superstorm of March 1993 that brought all-time record snows and low pressures to many points in the East. In that case, "There was a big eruption of convection in the Gulf of Mexico, almost like an MCS [mesoscale convective system]. If you shifted everything several hundred miles to the northeast, it could have been similar to [the January 24-25 event]--an initial phase of development that wasn't captured very well." However, by the time the Superstorm reached population centers, models had clued in to the rapid growth and the storm was hailed as remarkably well forecast.

This output shows relative humidity, a sign of precipitation likelihood, as projected in the Eta model issued at 00Z 1/24/00, or about 7:00 p.m. EST Sunday, 23 January. The sequence shows the model's starting point and its 12- and 24-hour forecasts extending to 00Z 1/25/00 (Monday evening). By then (third panel), the Eta was shunting most of the modeled moisture offshore. (Illustration courtesy NWS.)

A key part of the USWRP is development of the Weather Research and Forecast model. This high-resolution software, a collaboration led by NCAR, NCEP, NOAA's Forecast Systems Laboratory, and the University of Oklahoma's Center for the Analysis and Prediction of Storms, will serve laboratory and operational needs at the same time. The WRF should be ready for research use in 2001 and actual forecasting in 2003. When considered across a wide range of storms, model improvements make a real difference, say the experts.

"The models now are obviously better than they were in 1993," says Chris. For the January storm, "You can look at the performance of the NGM [the Nested Grid Model, developed in the 1980s] versus the Eta [the current standard for 48-hour forecasts], and the NGM was terrible." However, in the end, "There's just a whole spectrum of cases out there, and every so often you're going to get one that fouls up everything." •BH

Second USWRP symposium is popular

So many abstracts--more than 70--were received for the USWRP's second science symposium that the meeting has been expanded from two days to three. It'll take place at the David Skaggs NOAA building on Broadway, 27-29 March. The first symposium was held at NCAR last spring. Each year the symposium will alternate between the USWRP's two main foci: quantitative precipitation forecasting (emphasized this year) and hurricane landfall (the main topic in 2001). "This will allow for more oral presentations for each of the foci and keep the meeting concentrated on each major subject," says Bob Gall, USWRP lead scientist. The symposium program and abstracts are linked from the USWRP "What's New" Web page.


In this issue...
Other issues of Staff Notes Monthly


UCAR | NCAR | UOP

Edited by Bob Henson, bhenson@ucar.edu
Prepared for the Web by Jacque Marshall