IE 11 is not supported. For an optimal experience visit our site on another browser.

Cracking the code for hurricane forecasts

Where will the hurricane hit — and how hard? If you're just watching that "skinny black line" on the charts of a storm's predicted track, you're not getting the whole story, experts on hurricane prediction and risk assessment say.

The National Hurricane Center's Web page provides a blizzard of data and graphics about impending storms, including two new tools that gauge the likelihood that a storm will strike a particular spot, and how strong the winds are likely to get.

"As scientifically interesting as all these tools are, we need to realize that they serve a lifesaving function, for people as well as property," said Scott Kiser, tropical cyclone program manager at the National Weather Service.

Cone of uncertainty
The most familiar visual in hurricane forecasting is the track that balloons outward from the storm's current position, three to five days into the future. That balloon is what's known as the "cone of uncertainty." Here's how they come up with the cone:

First, forecasters take data about wind, precipitation, temperature and pressure from a variety of sources on the ground and in the air and feed all those readings into increasingly sophisticated computer models.

Scientists then study the computer-generated numbers and come up with specific forecast points. These predict the storm will hit a certain location at a certain time. "The forecast points are human-generated, and they represent a blend of the specialists' experiences and the various models," Kiser said.

The forecast points, linked together, produce the skinny black line you see on storm-tracking maps.

Next, forecasters add the cone of uncertainty by mathematically computing an error range, based on a 10-year average of prediction errors. Because predictions are more likely to be wrong the longer away they are made, this cone widens around the black line as it gets further away in time.

When you're on the National Hurricane Center's Web page, look for the graphics labeled "3-Day Cone" and "5-Day Cone." The three-day cone goes out three days from the time of the forecast, and historically, a given storm has a 60 to 70 percent chance of staying within the cone during that time period.

The extended five-day forecast has been publicly available for just the last three years, and thus it's based on a smaller pool of averaged data.

All this uncertainty provides a valuable reality check for the general public.

"Don't focus too much on the skinny black line," Kiser warned. Last year, for example, the black line showed Hurricane Charley heading for Tampa, Fla., but the storm took an unexpected turn and made landfall well south of the city, catching residents in the danger zone unaware.

After that bum steer, researchers at the weather service considered getting rid of the black line altogether, but the public feedback persuaded them that the forecast was more useful with the line than without it, Kiser said.

Two relatively recent types of forecasts now supplement the storm track and its cone of uncertainty: the strike probability forecast and an experimental set of wind probability forecasts that is making its debut just this year.

"It gives the users another tool in their toolbox," Kiser said.

Strike probability
All the weather service's charts deal with probabilities rather than certainties. It's easy to miss that fact when you look at the skinny black line, but there's no getting around it when you look at the strike probability chart.

The chart graphically shows the probability that the center of the storm will pass within 75 miles of a given point during a 72-hour period. Look for the "Strike Prob" graphic on the National Hurricane Center's prediction page.

If you're in the potential path of a storm and trying to figure out when it might pass through, the textual version of this data set is even more useful. Click on the text link for "Strike Probabilities" on the center's prediction page.

The text table breaks the probability down into time periods of 12 to 24 hours — and a close reading of the numbers can give you an idea how quickly the storm is likely to approach, no matter which way it turns.

Wind probability
The experimental wind probability charts are a little more complicated to read. For each tropical storm, the weather service provides three maps displaying the likelihood that sustained winds in a given area will exceed a particular level over periods of up to five days.

There are separate sets of maps for over 39 mph (or 34 knots) for a tropical storm, over 58 mph (50 knots) for advisories to mariners, and over 74 mph (64 knots) for hurricane-force winds. Each of the maps is labeled "Wind Prob" on the center's prediction page.

"We just put the products out for public comment, and we have what we think is a wonderful new probabilistic product," Kiser said, with the pride of an auto executive rolling out the latest model. Feedback on the wind charts can be sent via e-mail.

Listen to local advisories
In addition to the Web information, textual hurricane advisories are available via e-mail delivery as well as RSS syndication feeds.

Kiser said residents in potentially affected areas should also check the hurricane local statements from the nearest weather forecast office. Those statements provide more detailed information about hurricane watches (where hurricane conditions are possible, usually within 36 hours) and warnings (where hurricane conditions are expected, usually within 24 hours).

"They can do a very good job providing storm specifics, even about tides and road closures," Kiser said.

Looking beyond all that data, Kiser emphasized that residents in any potential disaster situation should follow the instructions from their local emergency management officials rather than relying too much on the numbers. As Hurricane Charley illustrated, sometimes the numbers can be misinterpreted.

"A professional player in baseball doesn't get up and hit a home run every time," he said. "Katrina was a marvelous forecast, we're very proud of it, but that doesn't mean we're going to forecast storms that well every time. Be prepared."

This is an updated version of a report that was first published Sept. 22, 2005.