Coning In: New Ways to Tap Old Data Boost Hurricane Forecast Accuracy

Despite advances in weather prediction technology, meteorologists must still qualify any hurricane forecasts with a "cone of uncertainty," which depicts just how far the center of a storm might deviate from its projected track. This year's hurricane season in the Atlantic Ocean, which began June 1, is expected to be less active than those recent years, but with season's peak less than a month away researchers are exploring new data-driven approaches to better warn those potentially in the path of approaching storms.

Cones of uncertainty—also called error cones—arise because forecasts become less precise the further in the future they attempt to project. The National Weather Service's National Hurricane Center (NHC) develops its tropical storm forecasts and the surrounding uncertainty cones by studying the degree to which all storms—regardless of location, intensity or other storm-related characteristics—have deviated from their projected paths over the past five years. The storm's track is represented as a series of circles connected by a line, with each circle representing a 12-hour forecast period. A cone shape forms as the circles grow and merge because of increasing uncertainty the further into the future the storm's path is projected.

NHC forecasts have improved steadily since 2003, when the agency first began issuing five-day warnings to the public. This year, for each 24-hour period, the radius of potential divergence is about 103 kilometers, down from about 110 last year. Despite these improvements, there is still only a 60 to 70 percent chance that through five days, the center of the storm will stay entirely within the uncertainty cone.

The hurricane center is refining its cone model and the software that generates it to better assess storm behavior and reduce uncertainty on a case-by-case basis, says James Franklin, branch chief of the NHC's Hurricane Specialist Unit, which maintains a continuous watch on tropical cyclones and areas of disturbed weather within the North Atlantic and eastern North Pacific basins.

Attention to individual storms provides "a more accurate assessment of risk," Franklin says. "It's an assessment more tuned to the situation at hand, rather than assuming the risk is the same all the time." To make its decisions, the hurricane center relies primarily on the Wind Speed Probability Products, which indicates an area's chances of experiencing hurricane-force and tropical-storm-force winds from any active storm.

Another notable effort to improve tropical storm and hurricane forecasting is underway at EarthRisk Technologies, a San Diego–based company developing what it calls a "dynamically adjusted cone model" for its clients in the energy and insurance industries. Given that not all storms are the same, the company creates these cones using the NHC data on storms of similar intensity and location, rather than on all storms in a region over the past five years.

History suggests that storms forming deeper in the tropics have been forecast more accurately in the past than storms that form in higher latitudes, says Steve Bennett, EarthRisk's chief science and products officer. "Of course this remains to be tested, but we believe that the forecast for a strong hurricane moving through the western Caribbean Sea can often have a narrower uncertainty cone than a weak tropical storm moving through the Gulf of Mexico," he adds. EarthRisk's dynamically adjusted cone forecasts are still in the early stages and will not be available in time for this year's hurricane season.

EarthRisk grew out of a Scripps Institution of Oceanography project a few years ago that sought to predict future extreme temperature fluctuations by studying historic weather data—including sea-level pressure, air temperature at altitudes of three to six kilometers and in the stratosphere (between about 10 and 50 kilometers), along with jet stream positions and velocities—gathered by the National Oceanic and Atmospheric Administration since 1948. Whereas most temperature predictions are reliable only up to one week in advance, EarthRisk claims its forecasts can accurately detect the heat waves and cold snaps that impact energy prices as far as 40 days in advance.

Tropical storm and hurricane prediction are a new endeavor for EarthRisk, and Bennett says his company is not looking to compete with the NHC in terms of forecasting. He adds, "Our customers are looking for ways to assess financial risk related to weather, while the hurricane center is providing information for public safety."

Follow Scientific American on Twitter @SciAm and @SciamBlogs. Visit ScientificAmerican.com for the latest in science, health and technology news.
© 2012 ScientificAmerican.com. All rights reserved.