Newsletter Sign-up

After a brutal season last year, Mother Nature may be brewing some more overdue bad luck for the U.S. coastline this season.

The present season, according to the most recent hurricane forecasts, could be a repeat of last year, if not worse. Insurers facing the eye of the storm (or eyes) can find little comfort in these predictions, as well as little to act on.

The overall prediction for the season's destructive force, measured by the Accumulated Cyclone Energy, or ACE Index, which totals all storms' wind speeds squared, ranges from 120 percent to 190 percent of average, according to Pasch.

This NOAA forecast, along with other preseason midrange predictions from William Gray's team at Colorado State University and Tropical Storm Risk led by the Benfield Hazard Research Centre of the University College London, tends to whip up a tempest of news coverage when released.

But for insurers working in long-term risk, these announcements aren't much more useful for assessing exposure than their local five-day forecast.

"It is a moot point," says Steve Smith, atmospheric physicist and vice president of ReAdvisory, part of Bermuda-based reinsurance intermediary R.K. Carvill Ltd. "There's actually a weak correlation between the predicted number of storms and landfall." And landfall, he states, is what insurers, and their customers on the ground, really care about.

The Southeast, especially Florida, got walloped last year, as four hurricanes slammed into the state and caused at least $23 billion in insured losses. A number of other storms churned harmlessly in the Atlantic.

"If you could say with any level of accuracy that there will be a landfall, say, in the month of August and that's it," says Tim Gardner, managing director and worldwide leader of the Property Specialty Group of Guy Carpenter & Co. Inc., "then a lot of companies could forgo 11 months' worth of premium to protect that event."

If models could be this accurate, then pricing decisions regarding policies would be much more efficient, he says, because buyers could predict the frequency of hurricanes, the severity of any given storm, and where the storm was most likely to come ashore.

(Reinsurers who are pricing the risk, he says, are probably going to be privy to the same information and will not sell a policy at one-twelfth the price of an annual deal just because they might know the month the storm would hit.)

It's also doubtful that any risk manager would be willing to forgo coverage for the other 11 months of the year just because the modelers predict that the storm that will hit landfall in August.And even if the models become that accurate, pricing for the month could escalate dramatically or insurers might just decide not to offer any coverage at all.

Some intrepid academics are attempting to formulate landfall predictions, such as Mark Saunders, lead scientist at TSR and James Elsner, professor of geography at Florida State University. Elsner has a model in review, funded by the National Science Foundation. He hopes his model will be able to gauge landfall probability as early as the January before any given hurricane season, which lasts from June to November.

"We try to focus on what factors in the atmosphere and climate are conducive to bringing the storms along the coast of the United States," Elsner says. "What we're trying to do is be able to, even by the early part of January or February, say with some certainty that the coast is going to be hit with a hurricane this season."

While Elsner and his team still need to peer-review their model, Saunders at TSR has already published his statistical model (in the April 21 issue of Nature). The model forecasts an ACE index for hurricanes that make landfall on the United States. Saunders and his team on June 6 estimated four tropical storms would hit the United States, two of which would develop into hurricanes.

Saunders points out that the model already proved accurate in predicting that the 2004 landfalls would be in the upper quartile of wind energy. "Insurers and others would have reduced their losses in 2004 by acting upon the forecast," Saunders and co-author Adam S. Lea concluded in the Nature piece.

As for 2005, "We continue to predict that U.S. landfalling hurricane activity in 2005 will be in the top one-third of years historically, to 70 percent probability," said Saunders.

BEWARE OF CLIMATIC ALCHEMISTS

These landfall models do have their doubters, nevertheless, who view Saunders and Elsner as climatic alchemists in search of the fool's gold of the modeling world.

"If we were able to forecast storms to the location, month and strength (which I think is highly unlikely) with a significant degree of skill," says Smith, "then it has the potential to radically change the way the catastrophic insurance industry works in a whole host of ways."

But as Smith says of the TSR model, "It's not based on any strong predictors." As for Elsner's work, he adds, "It's interesting science but I'm pretty skeptical that it can generate enough skill to be useful for the insurance industry."

Pasch takes much the same approach to it. "We looked at it (the TSR model)," Pasch says. "But we haven't made an official statement. It doesn't seem to have a large predictive signal."

Phil Klotzbach, lead research associate for Gray's Colorado State team, dismissed Saunders and Elsner bluntly, saying that landfall predictions are "really impossible to do months in advance."

"Hurricanes are steered by day-to-day weather patterns that aren't really predictable outside of a week or two out at most, so there's really no skill in that," he says.

Both NOAA and Colorado State do make stabs at a landfall estimation, but these are simple ranges based on historical averages, this year's hurricane forecast and common sense.

"If we're forecasting an active season," Klotzbach explains, "that means there are more storms out there, and that means that the probability of landfall increases."

Still, with these basic probabilities, neither group dares call the intensity of the hurricanes in 2005 that will make landfall, which TSR attempts to do. Nor do they care to predict how many hurricanes will make landfall or where they are going to strike, which Elsner hopes to do. There are just too many variables involved.

Even English majors, who don't spend much time thinking about quantitative models, might understand the uncertainty. Though many hurricanes could form this season, for instance, few or none might reach the shores of the United States, such as what occurred during active years like 2000, 2001 and 2003. (The landfalls for all three years totaled two.) Or as with Hurricane Andrew in 1992, catastrophe could be wrought by a solitary but monstrous storm during an otherwise below-average year.

Even without reliable landfall predictions, insurers are not entirely vulnerable to catastrophe. The 2005 hurricane forecasts don't have carriers and reinsurers quivering under their beds.

"You've done your homework, you've gone through your risk management guidelines, you've done the modeling on your portfolio, and you've bought reinsurance that you think is applicable and appropriate," says Gardner, "and you can kind of live with it regardless of what the forecasting says. Because it just gets too difficult to manage any other way."

MATTHEW BRODSKY is associate editor of Risk & Insurance®.

August 1, 2005

Copyright 2005© LRP Publications