For decades weather forecasts have been reserved for government agencies like the Met Office or NOAA in the US, using huge supercomputers to crunch satellite, radar and other data.
But a wave of tech startups is opening up weather predictions, thanks to cheap components and cloud computing power, which could make forecasting much more accurate.
“The world observing system is under an enormous amount of pressure and I would go as far to say inadequate to deliver the necessary data for accurate weather forecasts,” says Peter Platzer, the chief executive of Spire.
Spire, founded in 2012, runs its own constellation of satellites – 58 small cubesats in orbit all around the globe monitoring the temperature, pressure and moisture of the air.
It sells that data to big weather agencies including NOAA.
Mr Platzer adds: “For weather, the world is starved for information and that in turn starves weather forecasting models for the much needed input to create accurate forecasts.
“And the reason why we don’t have enough data is that the global observing system is shrinking, is driven by technology that was built at the same time.”
Others are taking a terrestrial approach. ClimaCell, founded in 2015, takes meteorological data from communications networks on the ground, like mobile phone networks, and combines it with radar and satellite data.
“We’re looking at multiple ways to sense the environment using software only,” ClimaCell’s CEO Shimon Elkabetz says, explaining that its sensors can spot rain and snow that radar might miss.
Then it comes up with a six-hour forecast. The idea is to create a bottom up weather report, rather than top down.
Other startups up like Jupiter, Earth Networks, Understory and Riskpulse are combining sensors and data analytics to deliver new types of weather reports.
Two things have driven the boom. First, cheap, capable components – the result of the mobile phone revolution started by the iPhone in 2007.
That’s true for ground-based sensors but also helpful for building satellites that are much smaller than traditional ones.
The CubeSat standard of 10cm3 modules, which emerged in 1999, has also driven satellite innovation.
Second, abundant computing power available on the cloud from giants like Amazon and Google means small companies can tackle weather.
“Meteorological startups are by definition a data intensive startup,” Mr Elkabetz says.
Advances in machine learning may spur further improvements.
According to Mr Platzer: “I firmly believe a step change is happening.
“It’s somewhat absurd that we have built computers that can be tuned to beat humans at every single game there is but we still can’t forecast the weather and prepare people that there’s going to be 14 inches of snow.”