Extreme weather exposes the limits of AI forecasting

Weather models with artificial intelligence can now keep up with conventional forecasts in many everyday tests. A new study shows that a classic physics-based system still produces more reliable forecasts when weather breaks records.

Researchers examined global weather data from the past few decades to isolate hundreds of thousands of local record-breaking heat, cold and wind events.

The team compared how multiple AI systems and a European high-resolution forecasting model handle these rare extremes up to ten days in advance.

Why extreme weather is difficult to predict

The work was led by Zhongwei Zhang, a statistician at the University of Geneva (UG) who focuses on rare climate events. His team uses advanced statistics to understand risks and uncertainties in weather and climate records.

Scientists call these record-breaking extreme weather events, which reach a new all-time high or low at a specific location and time of year. They are important because communities often plan based on previous records, so events go far beyond…