How Good Are Hurricane Forecasts?
Last year, they blew.

By Daniel Engber
Posted Thursday, May 24, 2007, at 2:43 PM ET

Government weathermen announced their forecast for this year's hurricane season on Tuesday, predicting 13 to 17 named storms. Last year, they guessed there would be eight to 10 hurricanes, but in the end, we had only five. In an "Explainer" column published in 2005 and reprinted below, Daniel Engber wondered about the accuracy of these forecasts.

The first tropical storm of the Atlantic hurricane season appeared in the Caribbean on Thursday. According to the National Oceanic and Atmospheric Administration, we can expect somewhere between 12 and 15 tropical storms this year. Scientists at Colorado State University predict 15 tropical storms, while a private company based in Britain says there will be around 14. If these predictions are accurate, we're in for an active season—the historical average for storms per year is 9.6. But how reliable are these hurricane forecasts?

Not bad at all. In general, the predictions fall within a storm or two of the observed totals. Last season, though, the forecasters had a bad year. 2004's six intense hurricanes doubled most predictions. The seasonal total of nine hurricanes was also significantly higher than expected. Forecasters blamed the poor predictions on a "year [that] did not behave like any other year we have studied."

Forecasting groups use more than 50 years of data to find correlations between oceanic conditions, weather patterns, and hurricane-season severity. El Niño seasons, for example, tend to have few hurricanes, while warm sea temperatures portend stormy weather. Since 1995, ominous temperature and wind patterns have foreshadowed a particularly long run of bad weather.

A seasonal hurricane forecast typically includes predictions for tropical storms, hurricanes, intense hurricanes, and ACE ("accumulated cyclone energy"). In a tropical storm, winds measure between 33 and 63 knots; hurricanes reach speeds between 63 and 95 knots, and intense hurricanes swirl at more than 95 knots. ACE measures the season's total storm energy—to compute it, meteorologists tally up the squared, maximum wind speeds of all tropical storms at six-hour intervals.

Each group has its own style and methods. The seasoned Colorado State forecaster William Gray uses both statistics and intuition to arrive at exact numbers for each kind of storm. NOAA prefers to make probabilistic statements—will the season have an above-average or below-average number of storms? These statements correspond to ranges of expected storms: This year, they say, we should get seven to nine hurricanes. The British company Tropical Storm Risk and the Hurricane Climate group at Florida State University use their own statistical models to assign probabilities for how likely we are to get different numbers of storms. Tropical Storm Risk then publishes the mean values for these distributions (which turn out not to be whole numbers). In 2005, Tropical Storm Risk is predicting 13.8 storms, 7.8 hurricanes, and 3.5 intense hurricanes.

Good hurricane forecasts are important for the insurance industry, which has a deep financial interest in assessing risk to coastal property. But an insurer doesn't care about storms that never hit the coast, no matter how big they are. There were only four hurricanes during the 1992 season (just as Gray and others had predicted), but one of them happened to be the catastrophic Hurricane Andrew.

Forecasters try to assess how many hurricanes will make landfall each season by looking at how certain climatological features—like the placement of a high-pressure area in the subtropics—might affect the path of tropical storms. Gray's team says there's a 77 percent chance of a major hurricane landfall in the United States this year.                                                                                                                                       Weather