IBM aims to use crowdsourced sensor data to improve local weather forecasting globally

At CES 2019, IBM outlined the Global High-Resolution Atmospheric Forecasting System, which uses sensor data, crowdsourced input and supercomputing to improve forecasts.
Written by Larry Dignan, Contributor

IBM is hoping that mobile barometric sensors from individuals opting in, supercomputing ,and the Internet of Things can make weather forecasting more local globally.

Big Blue, which owns The Weather Company, will outline the IBM Global High-Resolution Atmospheric Forecasting System (GRAF). GRAF incorporates IoT data in its weather models via crowdsourcing.

While hyper local weather forecasts are available in the US, Japan, and some parts of Western Europe, many regions in the world lack an accurate picture of weather.

Also: IBM, Nvidia pair up on AI-optimized converged storage system

Mary Glackin, senior vice president of The Weather Company, said the company is "trying to fill in the blanks." She added, "In a place like India, weather stations are kilometers away. We think this can be as significant as bringing satellite data into models."

For instance, the developing world gets forecasts based on global data that are updated every 6 hours and resolutions at 10km to 15km. By using GRAF, IBM said it can offer forecasts for the day ahead that are updated hourly on average and have a 3km resolution.

What's in it for IBM? Weather data impacts everything from the supply chain to manufacturing to a bevy of industries. Weather forecasting focuses on storms and big events, but GRAF aims to forecast the smaller events such as thunderstorms. IBM's move to broaden crowdsourced weather data as a Los Angeles lawsuit alleges that the company misused tracking data from its weather apps. IBM maintains that customers opted in to sharing location and sensor data. 

Ultimately, Glackin said data from automobiles, buildings, and wearables could add data to weather forecasting models. IBM's supercomputers can also make inferences from Internet of Things data. For instance, if connected cars are running windshield wipers in a location, the models could infer there's rain. "We still need a national weather system, but it doesn't have the scale for fine tuned forecasts," said Glacklin. 


Among the key points:

  • The new model can use barometric pressure readings from smartphones and sensors from aircraft. People participating would opt in to sharing data. Weather Channel app users would send pressure readings to IBM with permission.  The system would be available globally in 2019.
  • Incoming data is crunched by supercomputers with 84 nodes of IBM Power Systems AC922 server and 3.5 petabytes of IBM Elastic Storage. The technology rhymes with what is used by the US Department of Energy and its Summit and Sierra supercomputers. Glacklin said IBM's effort is the first time global weather modeling is based on GPUs.
  • Emerging markets will benefit the most from the new weather model.
  • Applications would vary by industry. There's an obvious use case for farming with data aiding decisions about fertilization and irrigation timing. Aviation could use it to avoid turbulence and utilities and insurers could better prepare for storms.

The forecasts would be available on The Weather Channel app and site as well as Weather Underground's app and site.

CES 2019: The best tech, gadgets on show (so far)

Related stories:

More CES 2019 coverage:

Editorial standards