small image of a bobblehead doll: Bob Uecker
1. Science of baseball2. Denver's big advantage3. Love that curve ball! 4. Rained out?5. National spot: Mouth cancer

  'talk' balloon coming from Bob, says: 'A Why Files Grand Salami!'

The game's in three hours
You've got your tix. But will you wind up dodging tornadoes? Will you need a rain check?

Once upon a time, you could trust a weather forecast about as much as you could trust a used-car dealer.

That was then. Now, a glut of data from satellites, radars, balloons and weather stations are funneled into computers that grind out a decently-reliable forecast before you can say, "Play ball!"

'2000 monthly graph' shows deaths are highest in July, August, then June.
July's a bad month for weather-related deaths, mainly caused by beastly heat, but also from lightning, flooding and tornadoes. NOAA

As a result, you can actually make plans based on weather forecasts -- at least those that don't venture too far into the future.

The most accurate forecasts, naturally, are called "nowcasts," because they concern the next few hours. Nowcasts rest on rapid data dumps from radar and satellites, says Thomas Achtor, executive director for science at the Space Science and Engineering Center at the University of Wisconsin-Madison. "There are new computers systems that allow us to see the data almost instantaneously." (In an era when every flat surface in an office is covered by a computer, it's bizarre to think that as recently as the 1970s, meteorologists entered data and drew maps by hand. That's slow.)

One major source of meteorological data, the GOES satellites, scan the entire Western Hemisphere every 15 minutes, Achtor adds. If you're really in a rush, individual lines of scanned data can be downloaded immediately.

GOES data, taken in both visible and infrared wavelengths, measures water vapor at different levels of the atmosphere, so scientists can use it to track clouds and watch thunderstorms develop. Advanced radars can see features just 1 kilometer across, almost precise enough to tell you whether it will rain in left field while the dugouts stay high and dry. "If you want see what will happen at Miller Park [home of the Milwaukee Brewers], you can get a pretty good idea," Achtor says.

For forecasting
Short-range forecasting, to about three days out, is a bit more challenging, since it relies on understanding how the atmosphere will change over time. (Nowcasting, in contrast, is more focused on how it will move over time, although critical changes in precipitation, temperature and wind do occur as thunderstorms and tornadoes develop.)

Achtor describes forecast models as mathematical devices with equations designed to predict the motion and physical state of the atmosphere. In an astonishingly simplistic simplification, we can say that weather models have three essential elements:

The "base state," (current temperature, pressure, precipitation, humidity and wind),

A theoretical understanding of atmospheric physics, and

One honking big computer.

You can already see where we're heading. Over the past few decades, each of these elements has been revolutionized. Vast hunks of data -- from satellites, balloons, instruments on commercial airplanes, ocean buoys, and observations from volunteer and paid weather watchers -- is being stuffed into a better theoretical model of atmospheric physics that is running on a V-8 computer.

I predict an accurate forecast
The result is -- or should be -- a better forecast.

Bobblehead Bob reads: Five-day forecasts are as good today as three-day forecasts were 15 years ago It's unusually difficult to get a grip on the growing accuracy of weather forecasts. Sure, the weather-wonk on local TV may guarantee that tomorrow's temperature to within 4 degrees Fahrenheit, although cynics might observe that this actually creates a 9 degree range. But is a forecast accurate if the temperature is right, but wind and precipitation are off the mark? Is it accurate when thunderstorms are forecast, but tornadoes strike -- but only in one part of the viewing area?

Having said that, we'll simply cite the National Oceanic and Atmospheric Administration's nifty history of weather forecasting, which says that five-day forecasts are as good today as three-day forecasts were 15 years ago.

Twisted progress
Certainly, the prediction of severe thunderstorms and the tornadoes they sometimes spawn have both improved markedly since the 1970s when most weather forecasting was best described as el-crudo. Meteorologists scribbled the limited data available to them on paper, made some calculations on a hand calculator, typed their warnings on a typewriter, and then had the data entered on punch cards for broadcast to the affected areas.

Hobbled by all these limitations, one can only be impressed that 63 percent of severe weather watches produced tornadoes or severe thunderstorms in 1973. But by 1996, that number had risen to 90 percent.

Perhaps more important, while only 42 percent of serious tornadoes occurred in watch areas in 1978, 95 percent were inside watch areas in 1995 (see "Improved Accuracy..." in the bibliography). Partly as a result, the death toll from tornadoes fell by more than half between 1960 and 1996.

The rain drain
If you're interested in forecasts of precipitation (rain or snow), however, the story grows murky. "In temperature and wind, forecasts have improved immensely," says Thomas Zapotocny, an associate scientist at the Cooperative Institute for Meteorology Satellite Studies (CIMSS) at University of Wisconsin-Madison, "but in precipitation, they really haven't."

To understand why, remember that precipitation is a binary phenomenon. It's raining, or it's not. "In temperature, you have a range, but precip is yes or no, zero or one," says Zapotocny.

Five graph lines are pretty close together, but overall accuracy declines as rain gets heavier.
Graph shows correspondence between four forecast models and actual precipitation, for March, 2002. Vertical axis shows correspondence -- 1.0 would indicate perfection. "ETA" is the National Weather Service's standard forecasting model. All models have decent accuracy for light rain, but are hopeless for downpours -- the most dangerous kind of rain. National Severe Storms Laboratory

With an approach that's about as realistic as the typical Why File, forecast models predict "rain" or "not rain" in an entire forecast box, which may be from 12 kilometers to 50 or more kilometers on a side. Yet rainfall, especially summer's heavy thunderstorms, tend to be local: They may drown a baseball game here but spare one across town. "It can rain two inches in Middleton [Wis.], but not at the airport [a few miles away]," says Zapotocny.

A third source of problems with rain forecasts, says Steven Ackerman, director of CIMSS, is the origin of precipitation -- inside clouds. "Until recently, we were not able to look inside clouds. We are doing that better with radar, and I'd anticipate we'll get better precip predictions as well."

At this point, however, problems remain, Zapotocny says. "In terms of precip amounts, there's almost a linear decrease in forecast quality from light rain to heavy." You can see that from the nifty graph on this page.

And while light rains can drown a ballgame, it's the heavy ones that cause flooding and major property damage.

Here's a prediction you can bank on. Some baseball players, like others who chew "spit tobacco" will get oral cancer.



  Back More
  The Why Files  

There are 1 2 3 4 5 pages in this feature.
Bibliography | Credits | Feedback | Search

©2002, University of Wisconsin, Board of Regents.