A new artificial intelligence weather forecasting model developed by Google DeepMind has demonstrated superior performance compared to traditional numerical weather prediction systems in global medium-range forecasts. The GraphCast model, detailed in a Science journal paper, uses machine learning to analyze decades of historical weather data and generate 10-day forecasts in under a minute on a single computer. In head-to-head comparisons with the European Centre for Medium-Range Weather Forecasts’ high-resolution forecast system—considered the gold standard—GraphCast outperformed on over 90% of 1,380 test variables, including temperature, pressure, wind speed, and humidity. The AI system showed particular strength in predicting extreme weather events, such as tropical cyclones and atmospheric rivers, with greater accuracy and earlier detection than conventional methods. While traditional systems rely on complex physics equations run on supercomputers, GraphCast learns patterns from past weather data to make predictions, offering faster and potentially more accessible forecasting capabilities. Researchers emphasize that AI models should complement rather than replace existing systems, as they currently lack the ability to explain their reasoning or incorporate real-time observational data. The model’s code has been made open-source to encourage further development and integration with operational forecasting. Read the full article at https://sciencedaily.com/releases/2023/11/231114143522.htm