A month before a late July heat wave sent temperatures as high as 108F in Newark, N.J., Stephen Bennett saw it coming. He was watching high pressure far above Canada’s Hudson Bay, strong jet stream winds over northern Europe, and a low bubble of hot air in the upper Midwest. They’re among the signals that historically precede extreme heat in the Northeast.
While typical forecasts use computer models to simulate the weather a week or two out, Bennett draws on 60 years of weather data to identify conditions that could lead to big temperature swings weeks later. The weather events in advance of a hot or cold stretch are like “a series of dominoes that start to fall over,” says Bennett, founder and chief scientist at EarthRisk Technologies. The year-old startup makes software that tries to predict the probability of each domino falling and sells it to energy companies that want to lock in fuel prices before periods of peak demand.
Bennett, 38, has been a weather geek since boyhood, when he watched The Weather Channel instead of cartoons on Saturday mornings. He graduated from the University of South Alabama’s meteorology program in 1995 and began doing radio and TV forecasts. He eventually became director of weather research at hedge fund Citadel Investment Group, working with energy traders.
In 2008, Bennett agreed to take a pay cut to help the Scripps Institution of Oceanography at the University of California at San Diego find commercial uses for its academic research. EarthRisk grew out of that effort, after Bennett spent 18 months mining weather data going back to 1950 for patterns linked to extreme temperatures. He detailed the statistical relationships he discovered in a 6,000-page data catalog in the fall of 2009. “Other people could assemble these catalogs of data,” says Scripps Director Tony Haymet. EarthRisk has figured out how to “look at that data in a new way.”
While still at Scripps, Bennett worked with software design shop Digital-Telepathy to build a program that lets meteorologists navigate the data and discern early signs of extreme weather. Together they launched EarthRisk in July 2010, licensing the methodology from Scripps, and Bennett left the university in June to run EarthRisk full-time. He won’t disclose revenue but says he had clients from Day One. His next project: detecting Atlantic storms such as Hurricane Irene as far as 10 days in advance by analyzing conditions such as ocean temperatures, sea level pressures, and vertical wind shear. Bennett also hopes to make his software intuitive enough to be used by non-meteorologists at insurers and other businesses.
So what’s in store for the fall? In mid-September, Bennett had his eye on cold air over the Midwest and high pressure in the North Atlantic that, he says, raise the risk on the East Coast for a late September chill.