Hurricane predictions gain ground
Forecasters believe they can do even better in coming years
By ERIC BERGER HOUSTON CHRONICLE
By steadily improving their forecasts and setting high expectations, hurricane scientists may be getting too good for their own good.
As Hurricane Ike crossed Cuba and approached the Texas coast, forecasts provided by the National Hurricane Center proved to be more accurate than the average set during the last five years.
Yet upon returning to Texas, the center’s director, Bill Read, says he hears all the time that Ike was a bad forecast.
Read, former meteorologist-in-charge of Houston’s National Weather Service office, heard it again at the recent National Hurricane Conference in Austin.
“That storm sure did a dance before it came ashore,” Gov. Rick Perry said.
Hurricanes, ultimately are the product of chaotic weather patterns, do dance. That researchers and computers have been able to make some sense of storms’ meanderings during the last few decades ranks as a significant scientific victory.
Despite the public perception of Ike’s forecast, the hurricane center set all-time records for every one of its track forecasts, from one-day to five-day predictions, in 2008.
The center has done better still in bringing down the average error in its four- and five-day forecasts, which first were made public in 2001.
Average errors are down
The first two years the Miami-based center made five-day forecasts, the average error exceeded 370 miles, or greater than the driving distance between Houston and New Orleans.
Last year the five-day forecast’s average error was just 192 miles.
“Models,” said Read, giving a one-word answer when asked why the hurricane center has steadily improved its forecast accuracy.
Read said our physical understanding of hurricanes, while not perfect, has improved our ability to model them. And ever more powerful supercomputers allow increasingly complex models to be run more quickly.
But it’s not just the models, said Sim Aberson, a research meteorologist for the government’s Atlantic Oceanographic and Meteorological Laboratory.
“It’s more experience from the human forecasters,” he said. “For the first couple of years we just didn’t have very much experience doing four- and five-day forecasts.”
The steady improvement in forecasting has some scientists like Aberson asking what the theoretical limits for track accuracy might be.
A study published a decade ago in Meteorology and Atmospheric Physics by Lance Leslie and others concluded the best forecast computers or humans could make was about 80 nautical miles for the average two-day forecast, and 120 nautical miles for the three-day forecast.
May even get better
But modern hurricane center forecasts are now approaching these average errors, and last year the European computer model substantially surpassed these limits for the Atlantic basin.
“I am not sure we know what the limits are right now,” said Fuqing Zhang, a professor of meteorology at Penn State University. “I think we will continue seeing improvements in track forecast, especially in extended range.”
Even if forecasters may be approaching the lower limits of forecasting at two and three days, the basic laws of physics suggest it is possible to make predictions at longer ranges than even today’s five-day forecast.
While chaos in the natural order erodes forecasts substantially over time, it’s reasonable that better computers eventually can help create seven- or nine-day forecasts that have some predictive skill.
To that end, Read said he plans for the hurricane center to eventually develop a seven-day forecast.