The city's unpredictable weather is in the spotlight during the 2024 Paris Olympics. It rained heavily during the opening ceremony, soaking more than 10,000 competitors, and subsequent rain caused the skateboarding and tennis competitions to be postponed. Now, the Summer Olympics are bracing for a heat wave in the cities of Paris and Marseille, where football and sailing competitions will be held.
Some meteorological services and universities are using this year's Olympics to advance weather forecast research. The teams include a group of researchers at the University of Texas at Austin (UT) who use artificial intelligence tools to aid predictions. The research lab, led by UT Austin professor Dev Niyogi and postdoctoral researcher Manmeet Singh, uses machine learning to focus on weather conditions at smaller “city scales” of 1 km x 1 km or less, from larger models that typically provide weather Get data. “We are trying to incorporate cities and communities within cities into weather forecasts,” Niyogi told the Observer.
In addition to measuring traditional metrics such as rainfall chances, wind speed, temperature and humidity in daily forecasts, UT Austin researchers are using their lab's global building height dataset to pursue more specialized outputs such as thermal comfort, It measures thermal conditions taking into account thermal comfort. The team's work requires significant computing power and is supported by computing systems at the university's Texas Advanced Computing Center.
In addition to about a half-dozen forecasts from organizations such as the national weather services of the United States, United Kingdom, Germany and Sweden, the UT Austin team's daily forecasts are sent to Olympic officials as a demonstration of research supported by the World Meteorological Organization (WMO) part of the plan. Program participants do not provide official forecasts as a supplement to regular forecasts from Météo-France, France's official meteorological association. Still, their results can be shared with organizers, who can use the information for their own needs.
The goal of the program, Niyogi said, is to understand “which methods or approaches work or don't work so that in the future they can help us develop models that may be used in the next Olympics.” We use games as a means of improving understanding – it’s machine learning aiding actual learning.
The emerging potential of artificial intelligence in weather forecasting
With the help of Google-led artificial intelligence (GOOGL), weather forecasting, as a rapidly developing field, is increasingly showing potential. A new model called neural GCM combines artificial intelligence and traditional physics-based models to compare. Traditional weather forecasting models develop forecasts more quickly and, according to a recent study in Nature, are more accurate than purely artificial intelligence models.
In recent years, global technology companies have invested resources in developing artificial intelligence tools for weather forecasting. Last year, Google, Microsoft (MSFT), Nvidia (NVDA) and China's Huawei said their AI weather models performed on par with the highly regarded European model run by the European Center for Medium-Range Weather Forecasts (ECMWF). Confirmation from ECMWF. Earlier this month, Google DeepMind’s prediction system GraphCast correctly predicted that Hurricane Beryl would make landfall in Texas.
The future of weather forecasting will include a mix of artificial intelligence-based models and physics- and dynamics-based models, Niyogi said, with the latter still being the “DNA and backbone” of weather forecasting. “The combination of dynamics-based models and artificial intelligence is the development direction of this field,” he said. “The future won't be either/or, but both.” It will be one versus the other.
