Kevin Myatt's Weather Journal at roanoke.com

Oh, I grew up wanting to be one. When I was a gawky kid in Arkansas, I got a reputation in elementary school for being able to out-forecast the local TV weatherman, who was really just an often-tipsy DJ and bad comedian. I grew up on a steady diet of tornado warnings -- and those of you from the Midwest know about those times when the television is nothing but weathermen for hours pointing at hook echoes while the news reporters chase disappearing subdivisions. I saw SIX tornadoes myself between 1974 and 1999. Darn near drove into one.

But the Lord had other plans for me, as newspaper things just started happening. Like "accidentally" having to take a journalism class my sophomore year of high school when other classes were full. Like having a sportswriting job fall in my lap days after my high school graduation. Like trying my hardest once to leave newspapers behind once and for all only to have a small-town newspaper publisher offer me an editor job at a bank parking lot. The newspaper life is a calling. I could more easily run from twisters than from the ink in my blood.

This career turn was all for the better for me. Though meteorologists have been glamorized in recent years by the movie "Twister" and by being labeled the "sexiest men" of certain cities, the more likely destination for someone with a degree in meteorology is sitting behind a radar screen in a windowless office in Sioux Falls, South Dakota.

Don’t get me wrong: These folks are often heroes, right next to firemen and policemen. If you don’t believe that, compare tornado fatality figures between the last 10 years with any 10-year period before 1960. Those nameless faces looking for clashing colors on a Doppler radar screen are saving lots of lives. But I like to be outside more. That’s probably why I took up hiking. (See my hiking page on roanoke.com)

As a journalist and editor for 14 years, I got to write plenty about weather events and lots of other memorable things, too. But my passion for weather still stalks me, and finally caught up with me here in Roanoke. I got myself labeled as the newsroom weather geek -- or "weather guru" as What’s On Your Mind columnist Tom Angleberger somewhat mystically calls me. Even after I moved to advertising in the fall, I’ve still been an unofficial weather consultant. And now, they’re giving me a column online.

So what the heck is this all about? The truth is, whatever I want it to be. We’re going to talk about the weather -- we’re not going to do anything about it, we’re just going to talk about it. Past, present and future weather. This site will evolve into whatever we want to make it. And largely whatever you want us to make it.

I’m sure I’ll be hearing from some of you as the days go along. There’s something of a weather geek in almost all of us.

As a certified weather geek, I patrol weather sources on the Internet that normal folk fear to tread. Or rather, don’t know to tread. Or rather, don’t’ care to tread. Or rather, don’t know exists. Twice a week, on Tuesdays and Fridays, and more often when weather events warrant, I’ll attempt to boil some of this down for you, with a little fun and flavor.

Kevin Myatt's

WEATHER JOURNAL

WHY CAN'T A COMPUTER FORECAST WEATHER?

Once during a hike, a computer technician friend of mine made a remark something like this:

"With the kinds of computers we have today, they should be able to do a lot better job of weather forecasting than they're doing. You should be able to create a program, feed in the data, and it should tell what the weather is going to do."

Sounds simple enough. I wrote a computer weather forecasting program on my 3,500-byte Commodore Vic 20 for a junior high project in the early 1980s. Enter the wind speed, temperature, humidity and barometer and it spit out a forecast. It was only 56 percent accurate over the course of a month. I just didn't have enough data.

And on a large scale, that's the biggest problem meteorologists have today: They don't have enough data.

Even if you had the perfect weather forecasting program, to be able to accurately forecast weather you would need data for every cubic millimeter of the atmosphere. In reality, weather stations are often 30 or more miles apart or non-existent over the ocean and in remote areas. Satellites and modern radar have allowed us to fill in some of the gaps, but still, when an approaching storm system crosses over an area of sparse human observers and instrumentation hardware, the computer models go crazy trying to forecast something they can't see very clearly.

Forecasters utilize many computer models. A highly trained meteorologist might scan the Global Forecast System, ETA, Canadian, UKMET, European, Navy NOGAPS and Air Force MM5 models in the course of a difficult forecast, each with its own strengths, weaknesses and biases. Some tend to be too cold, others too warm. Some drive systems too far to the south, some lift them too far to the north. One model may show the storm of the century raging up the East Coast in five days, while another may show nothing but high pressure and sunny skies. This last scenario has happened more than once in just the past three weeks.

Each of these models in turn provides multiple "ensembles," or possible solutions to a forecast that vary depending on slight changes. While the main model run may not be showing a certain feature, sometimes the ensembles pick up on something that others are not showing, and this turns out to be closer to correct. And sometimes what the ensembles show is nothing more than a silicon-induced hallucination.

Under ideal circumstances, the models begin converging toward one forecast, moving closer and closer to one another as a given storm system approaches. Rarely do they completely intersect, requiring forecasters to make judgment calls up to the very onset of a storm.

Sometimes, the models wildly diverge from one another at the last minute, leaving forecasters in a quandary. The 1993 superstorm that dumped 18 inches in Roanoke and buried places from Birmingham to Boston was accurately picked up by many forecast models days in advance. The 1996 snowstorm that set Roanoke's 24-hour snowfall record of 22 inches was still not being seen as anything more than a run-of-the-mill snowfall by the National Weather Service's lead model less than a day before it hit.

Computers are no substitute for forecasters' training, experience and instincts. While I may rib the National Weather Service from time to time, I can tell you that the local National Weather Service office in Blacksburg has an excellent handle on the local quirks that throw an added monkey wrench in the forecast: upslope and downslope winds in the mountains, elevation differences, cold air damming in the valleys.

I've learned a lot from their discussions and forecasts over three years about local features that were not a factor in forecasting across the mostly flat plains of eastern Arkansas where I grew up. There, the forecasters had an excellent handle on the dynamics of tornado development, but they frequently fumbled on winter weather situations. It's all about experience and location.

So while there's no doubt that computer models are getting better, they're still no match for eyeballs, neurons and gut instincts. We haven't got the weather figured out, nor shall we in my lifetime. It's too complicated for a computer.

How our New Year's winter began

Previous Journal entries

Talk weather to Kevin. Click here.