What if, wherever you're living, working or traveling, you could receive an alert before it's about to rain or snow and when the precipitation is about to end? And it's pretty accurate, too?
ClimaCell, a fledgling weather technology company based in Boston, released a mobile app Monday that provides notifications for exact locations in more than 50 countries. It promises "street-by-street, minute-by-minute short-term forecasts."
The app leverages the company's technology innovations, which are making waves in weather business industries.
ClimaCell, founded in 2015, has developed a global network of weather data that marries traditional observations of pressure, temperature, precipitation and wind with information drawn from wireless signals, satellites, connected cars, airplanes, street cameras, drones and other electronic sources. Millions of pieces of weather data can be derived from these technologies. It's what the company describes as the "weather of things."
This mix of data is fed into ClimaCell's forecast models, operated in Boulder, Colo. The company created the NowCast model that gives highly specific, minute-by-minute forecasts out to six hours and a longer-term model, known as CBAM, that produces forecasts out to six days.
These models are designed to provide forecasts to help businesses solve problems in which "extra accuracy" is needed, according to Shimon Elkabetz, ClimaCell's chief executive.
Many of the weather companies operating today, spun up in the 1960s and 1970s, just take model forecasts from different governments, blend them and use statistical techniques to try to make them better. But ClimaCell is creating its forecasts from scratch.
Elkabetz said early results on its accuracy are promising. Compared with government forecasts, "we've been able to improve almost every parameter in every time frame," Elkabetz said.
ClimaCell has also created a software platform that allows its forecasts to be optimized and tuned to customers' needs. Elkabetz said it can generate forecast output for any weather variable of interest, at any location and at different degrees of specificity.
The forecasts are updated or refreshed constantly, which is the "best way" to increase their accuracy, according to Daniel Rothenberg, ClimaCell's chief scientist. "In our U.S. precipitation NowCast, we refresh [the forecast] end to end in under five minutes," he said.
By comparison, the U.S. government model used for short-range precipitation prediction, known as the HRRR (high-resolution rapid refresh model) updates hourly.
"We're blowing the pants off HRRR," said Luke Peffers, ClimaCell senior vice president.
Shawn Milrad, a professor of meteorology at Embry-Riddle Aeronautical University, reviewed information provided by ClimaCell about its models and their performance. He found the computing approach "cutting-edge" and the techniques for bringing data into the model "promising" but said he was skeptical about how much the "weather of things" data would improve model performance. "They probably can't hurt a model, but how much they help I think is still up for debate," he wrote in an e-mail.
The app is available on the AppStore for iOS devices, and an Android version is to be launched in September. The app is free and has no ads.