AI fails to predict extreme weather events: University of Geneva study

Di Valerie Linc ·

New study reveals that AI systematically fails to predict extreme weather events, underestimating heat and cold waves

Contesto

In brief - AI fails to predict extreme events - Study by the University of Geneva - Physical models more reliable ## Key facts - What: Study on AI reliability in weather forecasts - When: April 30, 2026 - Where: University of Geneva - Who: Researchers from the University of Geneva - Amount: 3 AI models tested (GraphCast, Pangu-Weather, Fuxi) Artificial intelligence (AI) is revolutionizing the field of meteorology, promising faster and more efficient forecasts. However, a new study led by the University of Geneva and published in the journal Science Advances reveals that machine learning-based models have a significant weakness when it comes to predicting extreme weather events. Researchers compared three of the most advanced AI systems for weather forecasting – GraphCast, Pangu-Weather, and Fuxi – with the reference physical model HRES from the European Centre for Medium-Range Weather Forecasts (ECMWF). The result is clear: when faced with record events, AI systematically fails. ### Problems with extreme predictions According to the study, extreme cold waves are generally predicted to be less intense than they actually turn out to be. Conversely, heat peaks are underestimated, with predicted temperatures lower than the actual ones. Significant deviations are also recorded for strong wind events. Moreover, AI models not only predict the intensity of extreme events too weakly, but they also predict them too rarely compared to what actually happens. ### Intrinsic limits of AI Researchers explain that the problem lies at the very heart of how artificial intelligence works. Models learn from historical data and recognize sequences and correlations that have already occurred in the past. However, record events, by definition, fall outside this range of experience. What...

Dettagli operativi

Implications for cross-border workers in Ticino For cross-border workers living in Italy and working in Switzerland, reliable weather forecasts are crucial for daily planning. Extreme events like heatwaves or storms can significantly impact cross-border traffic, particularly at border crossings such as Brogeda and Gaggiolo. Inaccurate forecasts can lead to delays, inconveniences, and potential safety risks. ### Comparisons with traditional models Traditional physical models, such as HRES from ECMWF, have long been the gold standard for weather forecasting. These models are based on fundamental physical laws and can predict extreme events with greater accuracy than AI models. However, AI models offer advantages in terms of speed and efficiency. Combining both methods could represent the optimal solution to address future challenges. ### Practical scenarios Imagine a scenario where an extreme heatwave is forecast for the weekend. Cross-border workers commuting to Switzerland from Italy might need to plan their journey in advance to avoid heavy traffic and adverse weather conditions. If the forecasts are inaccurate, they could find themselves in uncomfortable or dangerous situations. A hybrid forecasting system could provide more precise and timely information, improving the safety and quality of life for cross-border workers. ### Insights and tools To delve deeper into the implications of weather forecasts for cross-border workers, it is useful to consult tools such as the cost of living calculator and the salary comparator. These tools can help better understand how weather conditions may influence daily expenses and financial planning. ## Useful tools to protect your net income To reduce FX leakage, compare CHF-EUR exchange options and banks for cross-border wor...

Punti chiave

What to do in case of extreme events In case of forecasts of extreme weather events, it is important to follow some guidelines to ensure safety and well-being. Here are some concrete steps: 1. Monitor forecasts: Use reliable sources such as the Swiss and Italian meteorological services to obtain updated information. 2. Plan ahead: If adverse weather conditions are expected, plan your trip in advance to avoid delays and inconveniences. 3. Prepare for emergencies: Have an emergency kit in your car, including water, food, a blanket, and a first aid kit. 4. Stay informed: Subscribe to weather alert services to receive timely notifications of any changes in weather conditions. ### Useful Tools For further information and useful tools, consult the cost of living calculator and the salary comparator. These tools can help you better understand how weather conditions may influence daily expenses and financial planning. ### Conclusion Reliable weather forecasts are fundamental for the safety and well-being of cross-border workers. With the increase in extreme events, it is essential to adopt a hybrid approach that combines the advantages of AI with the robustness of traditional physical models. By using adequate tools and resources, cross-border workers can better face the challenges posed by adverse weather conditions. Source: tvsvizzera.it

Punti chiave

[{"q":"Why does AI fail in predicting extreme weather events?","a":"AI fails in predicting extreme events because it relies on historical data. Record events, by definition, fall outside this range of experience, evading the statistical logic of machine learning."},{"q":"What are the advantages of traditional physical models?","a":"Traditional physical models simulate the evolution of the atmosphere based on immutable natural laws, such as thermodynamics and fluid mechanics, which hold true regardless of whether a certain phenomenon has been observed before."},{"q":"How can cross-border workers prepare for extreme weather events?","a":"Cross-border workers can prepare by monitoring forecasts, planning ahead, getting ready for emergencies, and staying informed through weather alert services."}]

Frequently Asked Questions
Why does AI fail in predicting extreme weather events?
AI fails in predicting extreme events because it relies on historical data. Record events, by definition, fall outside this range of experience, evading the statistical logic of machine learning.
What are the advantages of traditional physical models?
Traditional physical models simulate the evolution of the atmosphere based on immutable natural laws, such as thermodynamics and fluid mechanics, which hold true regardless of whether a certain phenomenon has been observed before.
How can cross-border workers prepare for extreme weather events?
Cross-border workers can prepare by monitoring forecasts, planning ahead, getting ready for emergencies, and staying informed through weather alert services.

Related articles