Millions of barometers in free fall through the city

Meteorologists have fed pressure readings from ordinary phones into real weather models. The forecasts got better.

A shower that hits one neighbourhood and skips the next is hard to catch if the nearest official weather station is a good distance away, on a flat field by an airport. But a thousand phones, scattered in pockets, cars and buses in that very neighbourhood, all register the same small pressure drop, at the same time. That was the idea meteorologists started taking seriously, in the years after the barometer arrived in the phone.

An official weather station network is expensive to expand, so it never gets denser than the budget allows. A phone in a pocket costs meteorology nothing — the instrument is already paid for and carried out into the world by its owner, in cars, on bicycles, on buses running through the city all day. The question was never whether there were enough of them. It was whether you could trust what they measured.

From PressureNet to Weather Channel

The first attempts to collect pressure from ordinary phones at scale came from apps like PressureNet and WeatherSignal, built for exactly this purpose: letting people donate their phone's pressure readings to research. PressureNet stopped its collection at the end of 2015. It wasn't long before someone else picked up the baton — in August 2016, The Weather Company started collecting pressure through its own Weather Channel app.

So the data stream never really stopped. It just changed sender.

Did it actually make a difference

The answer is yes, but with an important caveat. Crowdsourced surface pressure from networks like Weather Underground improved short-term forecasts — the first zero to six hours — for exactly the weather phenomena that are too small and too fast for a sparse network of official stations to keep up with: fronts passing through and convergence zones, the lines in the air where wind from different directions meets and is forced upward.

Other researchers have since gone further and tested whether dense sets of smartphone pressure data could improve a simulation of something as violent as a hailstorm. And there has been work specifically on the craft itself — how you collect and process barometric data from phones so it can be used in a numerical weather model at all, without polluting it with noise. It's no longer just a question of whether the idea works. It has become a question of how you do it properly.

Aneroid sends nothing

All of this — donated pressure readings, crowdsourced networks, data that ends up in a weather model somewhere, whose result the researcher never feels in their own body — is the opposite model of what Aneroid does. Aneroid doesn't send a single pressure reading anywhere. No server pools data from your phone together with anyone else's. The readings stay on the phone, in your own pocket, and that's not a caveat. It's the point.

Others need your pressure reading to improve a map you'll never get to see your own piece of. Aneroid doesn't need to send it anywhere to do its job — the analysis of how you specifically react to pressure happens with the data already sitting on your own phone.