Weather forecasting difficult to do
Feb 26, 2013 (The Salina Journal - McClatchy-Tribune Information Services via COMTEX) --
If Yogi Berra had been a meteorologist, one of this famous quotes -- "It's hard to make predictions, especially about the future" -- might have been further mangled to include "especially about snow."
Many people around north-central Kansas who peeked outside at the crack of dawn Monday expecting to see a foot or so of new snow were relieved to see only about an inch.
And Salina Superintendent Bill Hall had already called off school Sunday evening, joining several other school superintendents in the area who believed the weather forecasts.
"Typically, we're up at 4 in the morning, checking roads and making the decision," Hall said Monday afternoon.
But in this case, Hall said, the forecasts had been consistent throughout the weekend, predicting 8 or more inches of snow on the ground by Monday morning, driven by winds of 40 mph.
"It's not doing anything now, but that wasn't what the forecast was saying," Hall said.
Another important consideration, he said, was, "What if you start school, and then it gets bad and you can't get the kids home safely Typically, we don't want to let out early."
Schools were also out for the day in Lincoln, Neb. -- which was also missed by the predicted blizzard.
Mark Anderson, an associate professor of earth and atmospheric sciences at the University of Nebraska-Lincoln, said computer models used for weather forecasting are better than ever, but that "precipitation, and especially snow, is the hardest to predict."
"Schools here are closed, and people are asking why," he said. "Even (Sunday), one model for Lincoln was giving us a half-inch, and another was giving 8 inches."
"The models probably do the worst in predicting precipitation," Anderson said. "They're very good at cutoff levels, and flows -- all things that mean a lot to a meteorologist," but not to the average person, who just wants to know if it's going to rain. "How much is it going to rain, and is it going to rain on me -- we can't do that at this point."
Generally, Anderson said, short-term modeling is more accurate than longer-term forecasting, "both spatially and temporally." But all face the same hurdle: "The models cannot represent what the atmosphere does, so at some point, you have to parameterize."
In the United States, the government runs three main computer models: the Global Forecast System, the North American Mesoscale Model and the Rapid Refresh model.
Each works on a different scale, but all essentially break the atmosphere into a series of cubes, similar to how digital photography breaks an image into a series of pixels; the more pixels (or the smaller the cubes) the more accurate the picture or forecast is.
The Global Forecast System, which is used for 10-day forecasts, uses cubes that are 60 kilometers on a side, larger than a typical Kansas county, while the North American Mesoscale Model, considered accurate three days out, uses cubes that are 12 kilometers (about 7.5 miles) on a side, and the newest experimental version of Rapid Refresh bases its calculations on a three-kilometer cube.
Given that the typical thunderstorm is about a kilometer across -- smaller than even the finest resolution model, Anderson said -- it's easy to see why "the models cannot pick out a single storm, but can tell you where they are likely."
The models are a product of data and computer horsepower, Anderson said, noting that the European computer model -- called the Integrated Forecast System -- "does a really good job of implementing satellite data, and we don't as much. One of the budget cuts now is to weather satellites, and that's going to hurt all the models. The more data we put into the models, the better outcome we get."
"The U.S. has a model, the Canadians have a model, the Europeans have a model, and they all give different answers to what will happen on a given day," Anderson said.
U.S. models also slice the atmosphere into 60 layers, while the European model uses 80 layers, Anderson said. He said that getting that data from the upper atmosphere also has shortcomings.
The National Weather Service launches weather balloons from some 300 different locations around the U.S. at 6 a.m. and 6 p.m. each day, Anderson said, and "that's our only vertical profile of the atmosphere."
The only two locations in Kansas are in Topeka and Dodge City, Anderson said, noting that if school superintendents are trying to decide at 4 or 5 a.m. whether to call off school, they're making that decision "about as far from current data as you can get."
Another difference between the accuracy in forecasting the snowstorm this past Thursday and the one that missed Salina Sunday and Monday was forecasters had access to a trove of additional data, Anderson said.
Scientists studying aerosols and pollution in the atmosphere over the Pacific happened to be gathering lots of data the week before the storm, Anderson said, "and they were able to put that data into the model."
Typically, he said, data on atmospheric conditions isn't available until a weather system is over the West Coast and into that weather balloon network.
That additional data, Anderson said, may be partly why one forecast was more accurate than the other.
But even with lots of high-quality data, Anderson said, many variables prove difficult to account for.
"How storms form over the mountains can make the difference between 10 inches of snow and an inch of rain," Anderson said. "The difference is whether it's above or below 31.5 degrees, 5,000 feet up in the atmosphere -- and you want to know that 10 days in advance
And exactly where snow turns to rain is important, Anderson said, because "the heaviest snow is usually about 30 miles north of the rain-snow line -- and the models have a grid of 30 or 40 miles, so the difference between no snow and the heavy snow is in one block."
Additionally, Anderson said, what seem like dramatic differences in snowfall would be considered small differences had it rained.
For example, he said, if .2 of an inch of rain is predicted, and rainfall varies from .1 to .3 inches across the area, people generally think the forecast was accurate -- but a tenth-inch of rain ends up being several inches of snow.
He does note that the computer models are getting better; U.S. models now use 60 vertical layers of atmosphere in their calculations, and "a decade ago, we were in the single digits, so it's greatly improved. ... The accuracy we used to have for five days (in the future), we're now getting for eight days."
"You're seeing that (improvement) with lead times, but we're still not down to, 'Is it going to rain on this side of the block or that side of the block ' " he said. "I think people expect more out of the models than they can produce.
"We can't do long-range modeling of the atmosphere, and I don't see that any time in the near future. But we are getting better."
You can help
In the meantime, Anderson said, ordinary people can help the forecasters become more accurate.
Just last week, he became Nebraska state coordinator for the Community Collaborative Rain, Hail and Snow Network, a project based at Colorado State University that recruits people to send in precipitation data.
Information on participating is available online at cocorahs.org.
"One of the things the models can use is where and how much snow is on the ground now," Anderson said. "That's one problem with the models now -- energy goes into melting snow, and if you have a thicker snow base, it will pick up more energy, and that's hard to account for unless you know how much snow there is."
-- Reporter Mike Strand can be reached at 822-1418 or by email at email@example.com.
___ (c)2013 The Salina Journal (Salina, Kan.) Visit The Salina Journal (Salina,
Kan.) at www.saljournal.com Distributed by MCT Information Services
[ Satellite Spotlight's Homepage ]