Predicting the unpredictable
By Peter Smith
Published: April 18, 2007
The Forecaster
Craig recites forecasts for southern Maine on weather band radio: “Periods of rain. The rain could be heavy at times. High near 39. Windy, with an east wind between 22 and 32 miles per hour, with gusts as high as 55 miles per hour. Chance of precipitation is 100 percent.”
Craig rarely stops. He doesn’t drink coffee. Sometimes, Donna takes over and reads the script. But they are no Sarah Long (WGME), Kevin Mannix (WCSH) or Matt Zidle (WMTW). Craig and Donna never smile. They don’t get excited or frequently use weather idioms like “It’s raining cats and dogs.” And they don’t describe “nor’easters” with the regional vernacular of Downeasters.
Craig and Donna are machines; their souls are federally financed.
The programs are run by the National Ocean and Atmospheric Administration’s National Weather Service, which is a little like the U.S. Postal Service. It’s a governmental organization that provides a service—only its reports don’t cost 41 cents apiece. But like the Postal Service, its agents work in snow, rain, heat and gloom of night toward the completion of their appointed rounds.
On Monday at the regional National Weather Service station, which is just over the North Yarmouth line in Gray, the phones were ringing off the hook. Journalists called the station. Volunteer “spotters” called in reports. One called to say two trees had fallen on his house.
Other times, the station hears from listeners who complain about Craig and Donna’s pronouncements. Bangor has become “bang” and “gore,” they say. Damariscotta has become unrecognizable. Since when is Falmouth, “foul mouth”?
Inside the squat brick building, where about a dozen trained scientists work on computer models of incoming fronts, the temperature and relative humidity are steady. It could be snowing an eighth of an inch an hour, with gusts of wind out of the southwest at 10 to 20 knots (possible for winter in Maine) with small tornadoes forming in the east (very atypical for Maine), but the prediction center remains a constant temperature and relative humidity. A large humidifier on one wall pumps water into the air.
“It’s for the computers,” the Warning Coordination Meteorologist, John Jensenius, said, “not the people.”
Modern weather forecasting equipment is fickle. Changes in relative humidity might affect the station’s processors sending data to a supercomputer in Washington, D.C., computers receiving wind velocity data from the Portland International Jetport, or the computer broadcasting the voice of the National Weather Service data on 162 MHz. Data from the Gray station is available for free: weathermen with $80 haircuts read the text forecasts on TV and Craig and Donna recite the forecasts.
Art Lester is a veteran hydro-meteorological technician. He transferred decades ago from Virginia and still releases weather balloons charting weather in northern New England.
Years ago, he was the voice. He broadcast reports of current surface conditions in Cumberland, swells on Casco Bay, or flood watches along the Saco River—all in his rambling, Southern twang.
But nine years ago, the weather service introduced Paul.
Paul was a software program that read data. Paul announced the weather in Springfield, Ill., western New York, everywhere—without passion, without twang and, more importantly, with a decreased reliance on human meteorological technicians. In 2000, Craig and Donna replaced Paul, who sounded like he was stuck inside a tin can.
And as equipment from weather balloons and radio broadcasting equipment modernizes, sophisticated software and hardware occupy an increasing portion of the nation’s weather forecasting centers. “They’re working on eliminating us altogether,” Lester said.
On Monday, even Lester couldn’t get a weather balloon off the ground. Modern technology is no guarantee.
The increased use of computers and decreased reliance on outdoor field work has resulted in no significant changes in accuracy. Predicting the weather is still unpredictable.
“We’re always wrong,” Lester said.
“No,” Jensenius said, correcting him, “We’re just not right.”
In the midst of high-pressure systems, when predictions are more predictable, Lester still has had some fun with computers.
He once programmed Craig to sing a robotic version of Barry Manilow’s “Copacabana.”