Image of Hurricane Florence on Sept 10 2018. from the International Space Station. (Photo by: … [+]
Artificial Intelligence Forecasting System is “first operational open AI model” for weather forecasting, according to European Centre for Medium-Range Weather Forecasts.
As the United States grapples with the fallout from the abrupt firings at the National Oceanic and Atmospheric Administration (NOAA), Europe has launched the Artificial Intelligence Forecasting System (AIFS).
The European Centre for Medium-Range Weather Forecasts, who released AIFS, claims that this new model outperforms existing “state-of-the-art physics-based models” by as much as 20% in some measures.
ECMRWF, an independent agency based in the UK, claims it is the first fully operational open machine learning forecasting model. The agency tested AIFS with Member States and users to refine it for 18 months before deploying, and among other measures, they found that the machine-learning model could predict hurricane tracks 12 hours further ahead than other models.
AI and Climate
The relationship between AI and climate change is complicated by the fact that contemporary generative models demand higher computing power than prior models and tools—and therefore energy—to train and to run.
“My latest research shows that switching from older standard AI models—trained to do a single task such as question-answering—to the new generative models can use up to 30 times more energy just for answering the exact same set of questions,” wrote Hugging Face AI and Climate Lead Sasha Luccioni in a Dec. 2024 article for WIRED.
To that end, ECMWF claims its AI model is vastly more efficient, using approx. 1,000 times less computing energy than physics-based simulations. Furthermore, it generates forecasts much faster, potentially further cutting supercomputing emissions and costs for meteorology.
ECMWF Director-General Florence Rabier called AIFS a “milestone” that will make high-quality forecasts freely available to all. In addition to ensuring the public can still access critical information, this forecasting could also improve disaster preparedness and renewable energy planning, such as surface solar radiation levels or wind speeds at turbine levels. As such, ECMWF leaders believe it could be used to maximize the function of cleaner energy sources.
Chart depicting how AIFS model was trained.
“It is not only us who are innovating as it is important to remember that, with ECMWF, 35 nations are working together to advance weather science to improve global predictions,” Rabier said in a statement. “This is to help national meteorological agencies in their work to contribute to a safe and thriving society.”
Comparing the Models
AIFS is not currently replacing existing models, but rather running alongside them. It uses the same combination of data—”60 million quality-controlled observations” from satellites, as well as planes, boats, buoys, and other sources—that the traditional model, Integrated Forecasting System (IFS), relies on.
What distinguishes AIFS is that it uses historical data of how weather events have evolved to evaluate initial conditions that occur in real time. IFS, on the other hand, only uses the laws of physics to run simulations.
But ECMWF sees this as a project to develop carefully, and with input from stakeholders.
“We see the AIFS and IFS as complementary, and part of providing a range of products to our user community, who decide what best suits their needs,” said ECMWF’s Director of Forecasts and Services, Florian Pappenberger said in a statement. “Making such a system operational means that it is openly available and has 24/7 support for our meteorological community.”
For more information about ECMWF’s AI weather forecasting developments, visit the AIFS Blog.
