As the U.S. rang in the new year, tropical storm Zeta was still out there. A month after the official end of 2005's record-shattering hurricane season, the year's last tropical storm continued to churn the dark mid-Atlantic into a frenzy with 65-mile-per-hour winds.
Consider Zeta an omen. Just when anxious coastal residents, climate experts, and harried insurance executives might like to settle down for a winter rest, the first forecasts for the 2006 storm season are in, and it looks like another bad one. "This year's theme: Get used to it," says Andrew Castaldi, head of catastrophe and perils at Swiss Re.
In December weather scientists cautiously scoped out the year ahead in order to help nervous businesses start to prepare. The stakes are high. Last season left property and casualty insurers facing a record $80 billion in property losses, on top of $45 billion in 2004, the previous worst year. It's no surprise, then, that big insurers and companies such as Gap (GPS ) Inc. and Charles Schwab (SCHW ) Corp. increasingly use weather and risk models to guide strategic decisions about what to insure, for how much, and even where to site operations.
Meteorologists at both University College London and Colorado State University expect about 16 tropical storms in the Atlantic this year, including about nine hurricanes, of which four or five will be major storms of category 3, 4, or 5. Two hurricanes are likely to strike the U.S. coast, predicts Mark Saunders, professor of climate prediction at University College London and lead scientist at TropicalStormRisk.com (TSR).
Hurricane prediction, admittedly, is nearly as much an art as a science -- especially this early in the year. Yet forecasting has improved enormously since 1992, when Andrew did $26.5 billion in damage, ending a period of relative calm. In the decade since, scientists have acquired better sensing technology, more advanced numerical weather-prediction systems, and an increasingly sophisticated take on how storms form.
Today's hurricane specialists can also crunch through an unprecedented volume of weather data. With about 150 years of storm-tracking information at their disposal, they have concluded that hurricane frequency and intensity rise and fall in 20-to-30-year cycles. Within that rhythm, scientists also watch secondary signals that can alter year-to-year forecasts. The coming season is likely to see a mild cooling in the equatorial Pacific -- called La Niña -- that correlates with increased Atlantic storm activity, says Gerry Bell, lead seasonal hurricane forecaster at the National Oceanic & Atmospheric Administration (NOAA). Recently rising sea surface temperatures in the mid-Atlantic and Gulf offer another clue that it will be an active year.
New data sources are helping to push the frontiers of hurricane theory further. A constellation of earth-orbiting satellites records everything from storm position to water temperature many times every hour. Planes crisscross storms as they form and grow, taking detailed measurements at multiple positions over weeks at a stretch. En route, they scatter parachute-borne sensor packages, explains Naomi Surgi, NOAA's hurricane modeling leader. These sacrificial machines fall through the storm, relaying back data on humidity, temperature, wind speed, and other vital statistics. "What used to take me years to gather can be downloaded in a few minutes," says Bill Gray, an emeritus professor of atmospheric science at Colorado State and co-author of a closely watched forecast, who started chasing hurricanes in 1953.
DECADES OF DATA
Still, the long-term models remain unspecific about critical variables, including where the storms will actually travel. Global warming is another wrench in the works. Climatologists concur that heat from the ocean drives hurricanes, but they cannot agree about how much it changes the annual outlook. And it's unclear whether rising sea temperatures are part of a longer cycle or if they are caused or aggravated by man-made pollution.
Getting a bead on approaching storm seasons is critical to insurers, which can then adjust their rates, capital reserves, and decisions about what to insure based on storm forecasts for the coming year. A small group of specialized catastrophe-risk modelers are helping big insurers and property owners do just that. If the hurricane scientists make programs that simulate sea and sky, "cat-risk" specialists focus on models of the earth and the buildings where humans live and work.
The top cat-risk modelers -- Boston's AIR Worldwide, London-based EQE International, and Risk Management Solutions in Newark, Calif. -- are closely held think tanks staffed with teams of economists and scientists. Their esoteric specialities range from soil dynamics to wild-fire modeling. Models they spin out illuminate a range of possible losses in the coming year for insurers and big chains such as Gap and Staples (SPLS ), says Bruce Norris, a senior vice-president at commercial insurance broker Hilb Rogal & Hobbs Co. (HRH ) Schwab even uses the data to help site its backup offices in low-risk spots, he adds. J.C. Penney Co. does likewise. After it picks a new store site based on demographic factors, the company's insurers provide local catastrophe risk estimates. In sites where bad weather is likely, Penney ruggedizes its facilities beyond local codes. Its reward? Lower insurance premiums.
The cat-risk modelers' raw ore is decades' worth of insurance returns filed after big disasters hit. After a hurricane, for example, these records capture the nature and severity of wind and flood damage, as well as repair costs across thousands of buildings. "Each hurricane is like a laboratory to fine tune our model" of how damage is done, says Robert Muir-Wood, chief research officer at Risk Management Solutions.
Every new catastrophe enriches the data. Studying hundreds of storms and running hundreds of thousands of scenarios based on different storm intensities, modelers can generate "catalogs" -- a statistical distribution -- of losses insurers could face under different outcomes, from a year of no landfalling storms to a record year like 2005. For cat-risk experts, hurricane forecasts -- like those from Colorado State and TSR -- are among the first variables they feed into the model to produce the next year's loss estimates.
Meanwhile, cat-risk models are also getting more realistic. While past analyses didn't differentiate among blocks of houses, the latest software aims to give each home its own real-world attributes. GPS coordinates, along with building records, can help distinguish a low-lying wood-frame house from a brick one on high ground and adjust damage estimates accordingly. "We're putting more real-world physics into the models," says Jayanta Guin, vice-president for research and modeling at AIR. The new tools won't prevent catastrophic losses, but they put companies in a stronger position to prepare for them.
By Adam Aston, with Joseph Weber in Chicago