AI in weather forecasting, prediction and communication
Global High-Resolution Atmospheric Forecasting System (GRAF)
Drive better decision-making with weather predictions that operate 6x faster and at 3x the resolution versus leading competitors.
A global weather modeling system unlike any other
Weather is changing. We all feel it. No industry is untouched. Increasingly volatile, weather drives the decisions we make each day and has an economic impact in the billions each year. Having access to accurate and reliable weather information is increasingly important as the number and severity of extreme weather events climbs. The good news is–from media, to advertising, to aviation and our consumer products–every time you receive a forecast or weather intelligence from The Weather Company, it is informed by GRAF.
Far more accurate than traditional weather models
Proprietary to The Weather Company, GRAF is a physics-based, high precision, rapidly-updating, global weather modeling system that can predict something as small as a thunderstorm virtually anywhere on the planet.
Historically, global weather forecast data has lacked accuracy, which makes it difficult to look to the past for future decision making and to access timely weather data on demand. These inaccuracies are due to models that update infrequently (~every 6-12 hours), can only provide forecast data at ~15 minute intervals, and operate at a resolution too wide to capture many weather phenomena (~10km).
The GRAF global weather modeling system uses high-resolution data and the latest in NVIDIA weather graphics to predict weather activity around the globe. The first weather model to run on a GPU-accelerated supercomputer which enables increased resolution and update frequency, GRAF incorporates extensive atmospheric observations and employs machine-learning (AI) techniques to refine and optimize models.
GRAF in action
A look at India radar, before and after, with GRAF
Five key differentiators
Best recency and only hourly updating model, versus the leading competitor which updates every 6 hours.
Best temporal resolution globally, producing 5 minute forecast data for the US and Europe and 15 minute forecast data globally, versus the leading regional models that provide 15 minute forecast data only.
Leading spatial resolution, operating at a 3.5km resolution for North America and Europe vs. the leading competitor which operates at a 10km resolution.
First operational global weather model to run on a GPU-accelerated supercomputer, enabling rapid assimilation of weather data inputs.
GRAF is a modeling system, not a singular model. The combination of best of the best inputs enables it to be superior to any singular model alone (including those created by AI).
A closer look at GRAF
Fueling the most accurate forecast for all people, everywhere.
Driving better decisions
A government disaster management department can better prepare for severe weather to save lives and maximize relief resources.
An off-shore energy and utility company can receive warnings of quick-forming, localized weather activity—such as thunderstorms—and be able to quickly optimize operations and keep workers safe.
A farmer in India can have a better sense of when rainfall is likely to occur and can make a more informed decision about using expensive fertilizer or pesticides which could be washed away if there was a heavy rainfall.
GRAF keeps getting better
We won’t stop at being the world’s most accurate forecaster1. Our team of over 100 meteorologists is continuously working to improve GRAF, collaborating with organizations like the National Center for Atmospheric Research and the American Meteorological Society to enhance the modeling system.
Looking ahead, we plan to increase our high computing capacity by 3x, improve 72-hour forecasting frequency, and implement JEDI Data Assimilation, which will enable the incorporation of additional, previously untapped data sets (such as satellite radiances, aircraft observations, radiosondes, and pressure sensor readings from smartphone users with consent) to contribute to even greater accuracy and granularity across time and space.