For Entertainment Purposes Only–Part I

In a recent blog, I alluded to the pitfalls of extended range weather forecasting. Despite my own admonition never to trust an outlook beyond three days into the future, I stepped out onto a limb and “predicted” a snowfall for Atlanta five days down the road.

The snow fell, right on cue, although it wasn’t nearly so much as the models initially had suggested, which was 1 to 3 inches. Within the coveted three-day time frame, however, the models had backed off to a trace to an inch.

That worked out just right, with only a dusting (see photo) coating the northern suburb where I live. But it was a disastrous dusting. A quick freeze after a period of wet snow turned roads to sheets of ice overnight last Thursday. The result was not surprising. The annual Wintertime Atlanta Destruction Derby took place Friday morning.

All of this leads nicely into the next excerpt from my unpublished manuscript, INSIDE THE WEATHER CHANNEL. Below begins a series about the perils of extended range weather prediction.

Extended-range Forecasts

There’s a scene in the classic Western movie The Magnificent Seven where a bad guy is making a run for it. One of the good guys, Britt, played by a young, lanky James Coburn, means to stop him. Trouble is, Britt is on foot. The bad guy is on horseback, already a couple of hundred yards away and moving at a full gallop.

No matter to our hero, Britt. He takes a two-handed grip on his revolver, extends his arms, aims carefully and squeezes the trigger. BLAM! A second or two pass. The evildoer, just topping a distant ridge, topples from his horse.

Britt’s sidekick, Chico, played by Horst Bucholz, turns wide-eyed to Britt and says, “That was the greatest shot I’ve ever seen.”

“The worst,” says Britt, staring straight ahead, “I was aiming at the horse.”

And so it goes with extended range weather forecasting. Beyond about five days into the future, meteorologists don’t very often hit the horse. (Relax, equine lovers, it’s a metaphor. I have nothing against horses–other than marching behind them in parades.)


The reason forecasters frequently miss the horse is the same reason Britt probably missed the one he was aiming at. I’m guessing he twitched slightly just an instant before he fired and the barrel of his gun ended up pointed a hair off the correct trajectory. (NOTE: forecasters use numerical models, not .45s.) A tiny error such as that doesn’t matter if the target is only a few feet away. But a shot that’s an inch off the desired trajectory at close range can end up a couple of feet askew by the time it arrives at a target on yon ridge. Result: dead bandido, live horsey. Or, in the case of a meteorologist: busted forecast, embarrassed weatherman.

In popular terminology, and relative to the atmosphere, this is known as the Butterfly Effect. It’s a term that appears to have arisen from the title of a paper written by Professor Edward Lorenz (MIT) in 1972: “Does the Flap of a Butterfly’s Wing in Brazil Set Off a Tornado in Texas?”

The answer is no, but the point of the paper was that small disturbances in the atmosphere at a particular point can cascade over time into large changes at some distant location, changes that wouldn’t have otherwise occurred except for our mythical Brazilian butterfly.

What this means for weather forecasting is that a tiny error in the data jammed into a numerical weather prediction model may not make a big difference in tomorrow’s forecast. But ten days down the road, if not sooner, it can balloon into a huge error. Maybe in the speed of a weather system. Maybe in its strength. Maybe in its location.

Even worse, the data injected into a model may fail altogether to reflect the fateful flap of a figurative butterfly’s wing. Thus, on a given Monday, a forecaster might not have a clue that a major East Coast snowstorm will bury New York City and Boston during the coming weekend. Or the obverse, the model may spit out a bogus storm that never materializes.

Part II will follow in a week or two.

Photo: A winter scene from the north side of Atlanta.
I hadn’t gotten around to taking down my Christmas decorations yet, so I shot this Christmasy picture on January 8.


  1. John Tabellione on April 27, 2011 at 4:36 pm

    I think that, in this case, the Bernard Three-day Theorem could be construed to meana forecast of three or four days of icy patches on the roads, and schools closed for equal amount of time :>)

Leave a Comment