Temperature anomalies during the 2020 heat wave in Siberia, which broke historical records and caused severe wildfires, among other impacts.
Image Credit: Zhongwei Zhang, KIT
Scientific Frontline: Extended "At a Glance" Summary: AI vs. Physics-Based Weather Models
The Core Concept: AI-driven weather models analyze historical meteorological data to predict weather conditions rapidly, whereas traditional physics-based numerical models simulate atmospheric states utilizing fundamental physical laws. Recent research confirms that while AI models excel at standard forecasting, physics-based models remain significantly more reliable for predicting unobserved, record-breaking extreme weather events.
Key Distinction/Mechanism: Purely data-driven artificial intelligence systems struggle to extrapolate beyond their training sets, causing them to systematically underestimate the intensity and frequency of unprecedented heat, cold, and wind events. Conversely, physics-based numerical models (such as HRES) rely on atmospheric physics, enabling them to calculate robust forecasts even when climatic states venture beyond historical norms.
Major Frameworks/Components:
- Artificial Intelligence Models: Purely data-driven neural networks (e.g., GraphCast, Pangu-Weather, and Fuxi) that utilize historical records to predict future atmospheric patterns.
- Physics-Based Models: Classical high-resolution numerical weather prediction systems (e.g., HRES from the European Centre for Medium-Range Weather Forecasts) grounded in thermodynamics and fluid dynamics.
- Physics-Informed Neural Networks: Proposed hybrid architectures designed to synthesize standard AI pattern recognition with the boundary laws of fundamental physics.
- Extreme Value Statistics: Statistical methodologies recommended to enrich AI training data to better manage severe, record-breaking weather anomalies.
Branch of Science: Meteorology, Climatology, Data Science and Artificial Intelligence, Atmospheric Physics
Future Application: These findings will drive the development of parallel forecasting protocols and hybrid early warning systems for disaster management. By supplementing AI speed with physics-based reliability, agencies can generate timely alerts for rapidly intensifying weather anomalies driven by climate change.
Why It Matters: Accurate forecasting of unprecedented weather extremes is critical for protecting human health, securing infrastructure, and mitigating economic damage. Relying exclusively on AI during unprecedented climatic shifts could result in delayed or completely missed early warnings for catastrophic natural disasters.
Artificial intelligence (AI) has rapidly transformed weather forecasting in recent years. Modern AI models deliver fast and energy-efficient predictions and, under average weather conditions, often achieve accuracy comparable to—or even exceeding—that of classical physics-based numerical models. However, for particularly severe, record-breaking extreme events, AI-based forecasts reach their limits. A new international study led by the Karlsruhe Institute of Technology (KIT) and the University of Geneva demonstrates this vulnerability.
Researchers led by Dr. Zhongwei Zhang at the KIT Institute of Statistics investigated how well modern AI weather models predict extreme heat, cold, and wind events that exceed historical records. They found that under these exceptional weather conditions, the physics-based, high-resolution model HRES of the European Centre for Medium-Range Weather Forecasts consistently outperforms the currently leading AI models.
AI Systematically Underestimates Records
The scientists compared several established AI models—including GraphCast, Pangu-Weather, and Fuxi—with the physics-based reference model HRES. While AI models perform well in overall evaluations across all weather situations, they show consistently larger forecast errors for record-breaking events. In particular, they underestimate both the intensity and the frequency of extreme events. “Our analyses show that AI models generally underestimate the intensity of heat, cold, and wind records,” explains Zhang. “The greater the exceedance of the record in their training data, the larger the underestimation.”
Limitations of Neural Networks Underlying AI Models
The researchers attribute this to a fundamental limitation of pure data-driven models: AI systems learn from historical data and are particularly effective at predicting weather patterns that resemble those observed in the past. By definition, however, record-breaking events lie outside previous experience.
“Neural networks struggle to reliably extrapolate beyond their training domain—that is, to make predictions beyond previously observed values,” says Professor Sebastian Engelke, a full professor at the University of Geneva and former supervisor of Zhang. “Physics-based models such as HRES, by contrast, are based on fundamental laws of physics. This ensures that their forecasts are still reliable when the atmosphere moves into states that have not yet been observed.” Such record-breaking weather situations are occurring more frequently in a rapidly warming climate, with sometimes severe consequences for health, infrastructure, and the economy.
Implications for Early Warning Systems
The findings are particularly relevant for early warning systems and disaster management. A systematic underestimation of extreme events can result in warnings being issued too late—or not at all. The authors of the study therefore emphasize that AI weather models cannot currently replace classical numerical forecasts. “For high-risk applications, one should not rely solely on AI,” states Zhang. Instead, the researchers recommend a parallel use of both approaches, as well as further research into hybrid models and physics-informed neural networks that combine physical knowledge with AI methods.
Prospects for Improved AI Models
At the same time, the study identifies pathways for making AI-based weather forecasts more robust in the future. These include, among other measures, the targeted enrichment of training data with simulated extreme events, new training methods from extreme value statistics, and hybrid modeling approaches. Until then, the central message remains: “AI is a powerful tool for weather forecasting—but for the most extreme and potentially high-impact events, physics-based models remain indispensable,” Engelke concludes.
Additional information: Researchers from ETH Zurich, the Helmholtz Centre for Environmental Research, Technische Universität Dresden, and the University of Geneva were also involved in the study.
Published in journal: Science Advances
Title: Physics-based Weather Models More Reliable Than AI for Extreme Events
Authors: Zhongwei Zhang, Erich Fischer, Jakob Zscheischler, and Sebastian Engelke
Source/Credit: Karlsruhe Institute of Technology
Reference Number: as050426_01
