Kevin Myatt is The Roanoke Times' weather columnist.

kevin.myatt
@roanoke.com
981-3341

Weather with Kevin Myatt

Recent columns

• We got graupel, but not on official record
• Moisture could get caught up in cold blast
• Forecast for Weather Journal: Partly print, with frequent Internet
• Column archive

Read the Weather Journal blog

• Sprinkles or flurries possible Tuesday, but maybe something bigger for the weekend?
• For now, it looks like a quiet, mostly mild week ahead for SW Virginia
• Coldest morning of winter so far likely across much of Southwest Virginia; Tuesday precipitation looking doubtful
• Weather Journal blog

#swvawx on Twitter

@KevinMyattWx

• Follow Kevin on Twitter at @KevinMyattWx

Forecast models are not meteorologists dressed in the latest styles strutting down a runway.

Forecast models are computerized depictions of the atmosphere that take a wide array of weather data and project how the atmosphere will change in the future.

"Some of the models focus on changes to fine-scale aspects of the atmosphere at a subregional scale, and are updated hourly," said David Wert, meteorologist-in-chief at the National Weather Service in Blacksburg, in an e-mail. "Other models are more regional or national in scale, focus on the motion of larger atmospheric features, and are updated every three or six hours. Other models are more hemispheric or global in nature, and look at weather changes over many days, weeks, or months, and are updated every 12 hours or daily."

The major models weather service forecasters look at for daily and weekly forecasts are the American Global Forecast System and the European model. The North American model, Canadian, British UKMET and the U.S. Navy's Operational Global Atmospheric Prediction System are among others that are sometimes used.

These models produce new runs on six- or 12-hour increments. Each new run incorporates whatever new weather data is available from ground observations, weather balloons, satellites and aircraft.

Besides the main output known as the operational run, the models also kick out a series of "ensembles" or alternate scenarios, changing variables slightly. These can be helpful in determining if a model's odd-looking operational run is a hiccup or if it's catching on to a trend.

Forecasters look for agreement among various models. They also must use their knowledge of meteorology to see if a model's solution makes sense. Sometimes computers kick out what I call "silicon hallucinations" that look interesting but bear little resemblance to reality.

"Meteorologists are trained to view the initial conditions that are input into each model, see if the input is consistent with reality, and make adjustments accordingly," Wert said.

"Although computer models can make a forecast, they are not always reliable and are highly susceptible to bad or missing input data," Wert said. "As a result, there will always be a need to maintain a human in the forecast process, to check the integrity of the incoming data sets, and to blend model output with personal experience and scientific understanding to make the best forecast."

Weather Journal will be taking a break. It will return on June 5.