The biggest challenge facing climate models is similar to those facing economic ones - predicting the past is relatively easy, but predicting the future is far more of a challenge. In the United States, predicting hurricanes is less accurate than NCAA tournament pools, forecasters did not come close to predicting 15 in 2005 or 2 in 2013, but a team from the University of Arizona write in an upcoming Weather and Forecasting article that they are 23 percent less error prone - at predicting the past, anyway.

Hurricanes are storms with maximum wind speeds in excess of 73 mph and are among the most damaging natural disasters in the U.S. The Atlantic hurricane season lasts from June 1st to November 30th and when they do hit, property is damaged and lives may be lost, all which could be reduced is predictions were more accurate. Hurricane Katrina was not a particularly dangerous storm before it hit Louisiana, for example, while a 2012 event in the northeast was called a Superstorm though by the time it hit it was not a hurricane at all.  

In both instances, damage to life and property would have been reduced if the forecasts about them had been more realistic.

"Our model is better at predicting the number of seasonal hurricanes in the Atlantic than the other existing models," said first author and University of Arizona graduate student Kyle Davis. "On average, our model has 23 percent less error for predicting hurricanes occurring since 2001." 

The team developed the new model by using data from the 1950 to 2013 hurricane seasons. They tested the new model by seeing if it could "hindcast" the number of hurricanes that occurred each season from 1900 to 1949.

"It performed really well in the period from 1949 to 1900," Davis said. "That's the most convincing test of our model."

Better seasonal predictions can help cities and governments in emergency management planning, said co-author Xubin Zeng, who holds the Agnese N. Haury Chair in Environment and is a professor of atmospheric sciences. 

The difference, they say, is that other forecasting models relied too heavily on El Niño, a three-to-seven-year cycle that affects weather all over the globe. Instead they used the Atlantic Multidecadal Oscillation to judge how much influence El Niño has in a particular year.  The AMO affects ocean temperatures, cycling from colder to warmer and back over a time scale of approximately 40-70 years. The AMO was in a warm phase from the late 1920s to the early 1960s and started cycling back toward warm in the late 1990s. Warmer sea surface temperatures generally generate more hurricanes.

Zeng suggested also including the force of the wind on the ocean - an innovation that, to the best of the team's knowledge, no other statistical model used. Strong winds reduce sea surface temperatures because they mix the ocean layers, thereby bringing cooler, deeper water to the surface.

Compared with the other models, the UA model de-emphasized the role of El Niño when the AMO is in the warm phase, as it has been for the past 15 years.