The Weather Paywall

There will be heavy rainfall in... (subscribe to read on). That’s how all weather information will be delivered in the future, if private players are allowed to have their way. The chase to control and commodify forecast data is heating up. Competition could lead to greater accuracy in predictions, but it might also corrupt the public service that has so far been free.

- Akshit Sangomla

EARLY THIS year, when information and technology behemoth IBM announced that its Weather Company had created a powerful new global weather forecasting system, it evoked hope and fear at the same time. The company claims that its new Global High-Resolution Atmospheric Forecasting System, or graf, will provide the “most accurate local weather forecasts ever seen worldwide” and can predict something as small as thunderstorms or as fickle as tropical cyclones that keep meteorologists on their toes till making the landfall.

Every day at Weather Company’s office in Brookhaven, in US’ state of Georgia, some 600 meteorologists, data analysts and a supercomputer analyse weather-related information gathered by governments and intergovernmental agencies through their weather stations, sophisticated radars, aeroplanes and spy satellites, and feed those into graf. To ensure that the forecasts reflect localised and near-real-time atmospheric, land and oceanic conditions, the team also harnesses data from some 270,000 personal weather stations (PWS) run by weather enthusiasts across the world and from hundreds of millions of smartphones, whose “pressure” and “location” sensors keep relaying data even as the user is on the move, talking or taking a nap. These data are then processed by the artificial intelligence-powered graf to issue 12 trillion pieces of forecast information for virtually every 3 sq km patch of the globe on an hourly basis. By comparison, the best available government or intergovernmental models have a resolution of 13 sq km and can update forecasts only once in every six hours.

“The information we generate and distribute consists of observations (realtime and historical), short- and longterm, seasonal and impact-based information, such as how crop yields or power generation will be impacted by weather,” says Kevin Petty, director of science and forecast operations at IBM.