• Subscribe

Weather for all

Speed read
  • New weather model democratizes weather forecasting around the globe
  • Hourly, high-resolution forecasts possible thanks to GPU-accelerated supercomputer
  • Areas vulnerable to extreme weather due to changing climate conditions will get greatest benefit

A river bursts its banks after unexpected heavy rain. Water rushes through city streets, deluging the houses, livestock, and people in its path. Lives and livelihoods are ruined.

But what if those swamped communities had access to timely weather forecasts that could more accurately predict just how many inches of rain were about to fall?

A powerful new weather forecasting tool improves global model resolution by 3x, helping to bring the rest of the world’s forecasts up to the standard once limited to a small number of countries. Courtesy Weather Co.

That’s exactly what a new supercomputer-driven weather forecasting system hopes to do.

Previously, high-level forecasting has been available in the US, Japan, and Western Europe. But a new global modeling system developed by The Weather Company will impact parts of the world that previously did not have access to hour-by-hour, localized weather data.

In particular, these precision forecasts could be life-changing in the parts of Asia, Africa, and South America that are the most vulnerable to extreme weather from climate change.

Zooming in on the weather

Most traditional weather models run on high-performance CPUs. But the new IBM Global High-Resolution Atmospheric Forecasting (GRAF) system is the world’s first global weather model optimized for graphics processing units (GPUs).

Thanks to these powerful chips that break complex problems into separate tasks, IBM GRAF can forecast atmospheric changes down to the thunderstorm level on a global scale.

Another key to the model’s success is collaboration with the National Center for Atmospheric Research (NCAR). The Weather Company chose to join forces with NCAR because the center had already developed the Model for Prediction Across Scales (MPAS), a framework for earth and atmosphere simulations.

“Working with NCAR’s atmospheric and software computational scientists, we ported the MPAS to GPUs,” says Todd Hutchinson, manager of computational meteorological analysis and prediction at The Weather Company. “With high-performance GPUs, we can have more grid points across the world and forecast more precisely because the computer is faster.”

More sensors, better forecasts

The basic process of weather prediction involves collecting observations of the atmosphere  from satellites, radar, weather balloons, and surface weather stations. Forecasters assimilate all of that data, and then project forward in time to generate a picture of future conditions.

<strong>Five times faster.</strong> High-performance GPUs allow more grid points across the world to be modeled, leading to forecasts that are more accurate and more precise. Courtesy IBM.“Most global weather forecasts run by government agencies generate forecasts out to roughly 10, sometimes 14 days,” says Hutchinson. “However, there’s a period of time between now and roughly 15 hours from now—essentially the day ahead—where it is difficult to get a consistent picture of the forecast across the world.”

That fifteen-hour gap is caused by the amount of time it takes longer-range weather models to run. The longer lead time for processing means that the very near future falls into a gap. One that IBM GRAF hopes to close. 

“We generate an analysis, run a forecast forward in time one hour, add new data to that analysis, and go forward another hour,” says Hutchinson. “Eventually, we get to a point where we can go out to 15 hours.”

To ensure that coverage is truly global, IBM GRAF will incorporate previously untapped data collected from airplanes and even smartphones. Including these new sources raises accuracy for regions of the world that lack specialized weather equipment.

Preparing for extreme weather

The bottom line: IBM GRAF improves forecasts at the local level. Current global weather models are updated every six to twelve hours and cover 10-15 square kilometers. By contrast, IBM GRAF updates every hour and forecasts down to 3 kilometers.

<strong>With 12 trillion pieces of forecast information</strong> issued each day, farmers in rural areas will receive more information to prepare for severe weather events that could damage their crops. Weather predictions from IBM GRAF are available now to the public through Apple and the Weather Channel app. Going forward, the model could benefit airline, utility, and insurance companies, as well as retailers, governments, and daily commuters, to name just a few.

Because of its hour-by-hour accuracy, the new system can forecast turbulence, potentially enabling pilots to reroute earlier and reducing airline delays. With a more trustworthy forecast, farmers in India could better prepare for heavy rain. In rural areas of Africa, farmers could move livestock to avoid flooding.

As our climate continues to change and extreme weather events become more common, a more equal access to accurate forecasts could save lives around the world.

Read more:

Join the conversation

Do you have story ideas or something to contribute? Let us know!

Copyright © 2023 Science Node ™  |  Privacy Notice  |  Sitemap

Disclaimer: While Science Node ™ does its best to provide complete and up-to-date information, it does not warrant that the information is error-free and disclaims all liability with respect to results from the use of the information.


We encourage you to republish this article online and in print, it’s free under our creative commons attribution license, but please follow some simple guidelines:
  1. You have to credit our authors.
  2. You have to credit ScienceNode.org — where possible include our logo with a link back to the original article.
  3. You can simply run the first few lines of the article and then add: “Read the full article on ScienceNode.org” containing a link back to the original article.
  4. The easiest way to get the article on your site is to embed the code below.