Hurricane Computer Model Forecasts

Hurricane National Weather Service When hurricanes threaten you’ll often hear our News 5 meteorologists refer to computer model forecasts. You may have even looked some up yourself on the Internet. Here’s some background instruction and a few cautions concerning “the models.”

by John Nodar

Published: Mon, May 26, 2008 - 9:53 am CST Last Updated: Mon, July 21, 2008 - 5:28 pm CST

When a hurricane enters the Gulf, folks want information. We spoke with shoppers at Bel Air Mall recently and they said they always want to know where a hurricane is going and when it might hit land…especially if it’s coming here!

To answer questions like these, meteorologists turn to the models. No, not Christie Brinkley, Tyra Banks, or Cindy Crawford; we’re talking about computer modeling. Computer model forecasts are simple looking maps that forecast movement and intensity of hurricanes and tropical storms…but don’t let their simple outer skin fool you. The models are actually composed of hundreds or even thousands of complicated math problems. Each equation represents some part of the atmosphere--like air pressure, sea surface temperature, or winds.

The most powerful computers in the world are used for weather forecasting. A Teraflop is one trillion calculations per second. The National Weather Service is using an IBM computer that by next year should be operating in the 100 teraflop range. According to IBM it would take a person with a hand-held calculator more than 80 million years to do the number of calculations a 100 teraflop computer can do in one second.

If computer model forecasts were cookies, the equations are like ingredients. The basic recipe is the same, but one model may use slightly more—say—upper level winds, while another includes more surface pressure…or perhaps add a few chocolate chips. In the end, they’re all still cookies, just different flavors of cookies.

Richard Knabb, Senior Hurricane Specialist at the National Hurricane Center in Miami says, “The role of the human is to sift through what each model is trying to tell you. What are its idiosyncrasies and biases?”

There are lots of different models with funny names like LBAR, UKMET, and NOGAPS. BAMS, for example, stands for Baron Advanced Meteorological Systems. It’s the one you see on First Alert VIPIR. Sometimes everything clicks, like the forecast of Hurricane Dennis in 2005. When you compare the BAMS forecast from two days before landfall with the actual radar data, it matches perfectly. Pretty impressive. But sometimes things don’t come together as well. The forecast for Hurricane Katrina just over a month later--this time four days out—was not even close. BAMS wasn’t alone with a big miss. None of the early models came close to a correct forecast. Two points to remeber…1) model forecasts are good, but not perfect. And 2) the farther out you try to forecast, the less perfect they are. Knabb cautions, “ When you see model output on TV or the Internet it’s information that can satisfy curiosity but its very difficult to discern what the models are really saying.”

Which brings us to spaghetti. No, not pasta. Spaghetti plots use all the different models on one map. These give a sense of confidence when they’re all lined up, as they were for Tropical Storm Barry in 2007. Confidence in a forecast like Subtropical Storm Andrea that same year would be lower, because all the models pointed in different directions. Knabb notes that’s why forecasting isn’t just left to computers: “Humans are still adding value and that’s what we spend out entire forecast process doing….trying to figure out what all these good friends of ours are giving us in terms of advice. But sometimes it’s all different.”

For more information about the models and explanations of their strange names, click here

WKRG.com » News Stream » Hurricane