Quiet hurricane seasons run counter to computer predictions

From an article in The New York Times by Evan Lehmann of Climate Wire:

Recent computer models predicting that more hurricanes will strike U.S. shorelines have vastly overestimated the financial losses suffered by insurance companies, according to a new analysis.

The simulations, called "near-term models" because they predict storm strikes over five years, were launched in 2006 after two vicious seasons of landfall hurricanes, including Katrina, crushed homes and businesses along much of the Gulf Coast and Florida. The models emphasize rising storm risk caused by warmer ocean water.

Now, with only one year left in those first forecasts, the models issued by three different firms have so far overshot the level of damage by tens of billions of dollars. A string of relatively benign hurricane seasons began just as the models were introduced.

Storms caused $13.3 billion in damages between 2006 and 2009, far below even conservative expectations. Near-term models predicted much deeper devastation, ranging from $48.8 billion to $54.6 billion during that same period.

"Four years into the five year projection period, the near term models have not performed well as predictive tools. Hurricane activity changes markedly from year to year, and the active hurricane seasons of 2004 and 2005 have not proven to be harbingers of a continuing trend for 2006 through 2009," says a report by Karen Clark & Co., a risk management firm operated by an early architect of catastrophe models.

Near-term models are something of an experiment. Unlike traditional models, they don't use a century's worth of hurricane data related to frequency, landfall and wind speed. Instead, modelers input information from the warmest -- and most dangerous -- periods of the past. That's when hurricanes tend to strike with a fist.