The US startup ClimaCell uses the waves of millions of mobile devices around the world to collect data for its weather forecast. The company says its model delivers 60% more reliability than traditional methods.
The weather is a complex science, and despite many improvements over the years, the accuracy is not always the rendezvous. One of the problems is the lack of data, which distorts the simulations. Meteorologists collect data thanks to radar and use a whole network of sensors in weather stations spread across the globe, and specialized satellites.
However, this mesh is insufficient, and it affects the quality of forecasts. On April 14, meteorologists predicted at most a few inches of snow for the city of Chicago in the United States. In the end, more than 13 centimeters fell, paralyzing the city with more than 700 canceled flights. It was the second snowiest day in the city’s history so late in the season. Only a startup called Climacell had expected the severity of this storm snow.
A network of millions of probes around the world
This company, founded in Boston in 2015, claims to achieve precision in its forecasts far beyond that of other meteorological centers.
To do this, in addition to the usual sources of data, the firm collects data obtained through operators, from millions of smartphones and other mobile devices around the world. By analyzing the quality of the signals emitted by these devices, it can determine the weather conditions on the ground, such as air quality or the level of precipitation.
ClimaCell also combines the images from surveillance cameras to further refine the level of accuracy. This allows him to boast on his website to use 500 million new sensors, and his forecasts would be 60% more reliable than those of other meteorological centers.
Better weather forecast
The startup has opened a research center in the city of Boulder, Colorado, to create a mathematical model to turn mobile phone observations into weather data that can be directly exploited by a simulation.
Thanks to data obtained from millions of devices, ClimaCell gets a much more detailed picture of local conditions. The forecast model can then target a specific region and focus on a type of weather with updates of a frequency that can be determined by the subscriber.
ClimaCell intends to sell its services to weather-affected businesses, such as electricity providers who can predict the performance of wind turbines and solar panels.
The startup is also interested in forecasting natural disasters that can have a significant impact on human lives, but also cost businesses a lot of money. The firm tested its model in Israel over three months during substantial floods. According to Luke Peffers, head of the Boulder research team, ” we did a great job compared to the Israeli weather service’s rain gauges.”