If you were in the business of supplying heating oil in the Northeast, do you think it would be useful to know if a big winter snowstorm is likely to arrive with sub-zero temperatures in Massachusetts next month? How would fire chiefs in the brushy backcountry of Southern California react if they knew the odds of intense Santa Ana winds would increase dramatically in four weeks?
For all of meteorology’s satellite imaging and computer modeling, John “JP” Plavan and Stephen Bennett say it’s just about impossible to use current weather forecasting models to make more than general predictions about the weather more than two weeks in advance. That might be enough time for fuel oil suppliers to get a few extra shipments into local dealers, they say, but it’s not sufficient to make decisions at the highest levels of a big corporation or government agency.
But what if a weather forecasting model could “estimate” the likelihood of extreme weather events 30 or 40 days in advance? Would it be helpful, for example, to know if the odds a major winter storm would hit a particular region had increased from 33 percent to, say, 66 percent?
This, in a nutshell, is the promise of the innovation under construction at EarthRisk Technologies, a San Diego company that Plavan and Bennett founded less than two years ago. “If you’re Home Depot, you certainly want to have snow shovels in stock if you’re anticipating a big snowstorm,” says Bennett, a career meteorologist who helped create the company’s predictive analytics technology with scientists at U.C. San Diego’s Scripps Institution of Oceanography. Likewise, if you’re the Federal Emergency Management Agency, you’d certainly want to get a 30-day advance warning of the next Hurricane Katrina.
“We fancy ourselves as a software company, not as a weather company,” says Plavan, an investor serving as the company’s founding chairman and CEO. “We provide information that helps our clients make decisions of value.”
Bennett, who is EarthRisk’s chief science and products officer, says the highest value for the company and its customers lies in determining the likelihood of extreme weather—heat waves, cold snaps, and the kinds of storms that trigger destructive events like tornadoes, hurricanes, and flooding. “Extreme events are the ones that have the highest impact,” Bennett says. “The places where the opportunities are to be seized, and the risks managed, are at the extremes.”
Plavan compares the startup’s predictive analytics to counting cards in Blackjack, a technique used by some gamblers to optimize their bets and to guide how they play each hand. “The odds change, depending on the cards already played,” Plavan says. Instead of six decks of cards in the dealer’s shoe, however, EarthRisk calculates the odds for extreme weather events based on correlations between existing weather patterns and historical patterns in a database that encompasses more than 60 years of detailed global weather data.
“The research question that Scripps Oceanography helped us answer is whether there are certain things that the atmosphere does that loads the dice, so to speak, in the way things play out,” Bennett says. “The data patterns have become so complex that it’s too much for a meteorologist—for one brain—to digest.”
Their focus on the statistical risks of extreme weather events also represents a fundamentally different approach from the long-range forecasts now issued by the U.S. Climate Prediction Center, which relies on
