A new study published in Science demonstrates that artificial intelligence weather forecasting models have achieved superior accuracy compared to conventional physics-based models. The research, led by a team from the University of California, Berkeley, evaluated several leading AI systems, including Google’s GraphCast and Huawei’s Pangu-Weather, over a year-long global test period. The AI models consistently provided more accurate predictions for key atmospheric variables like temperature, pressure, and humidity at various altitudes, particularly for medium-range forecasts of three to ten days. The systems learn from decades of historical weather data to identify patterns, enabling faster and less computationally expensive forecasts. However, the study notes that AI models still face challenges with predicting extreme weather events and rely on traditional models for their initial data. The findings suggest a hybrid approach may be the future of meteorology. Read the full article at https://sciencedaily.com/releases/2024/07/240705150927.htm