A new artificial intelligence weather forecasting model developed by researchers has demonstrated superior accuracy compared to conventional physics-based systems in global medium-range weather predictions. The AI model, trained on decades of historical atmospheric data, can generate 10-day forecasts in minutes on standard computing hardware, a significant reduction from the hours required by traditional supercomputer-run models. Initial testing shows the AI system matches or exceeds the accuracy of leading numerical weather prediction models for key variables like temperature and pressure across most of the globe, particularly in the critical three-to-seven-day forecast range. While the technology shows immense promise for faster, more efficient forecasting, meteorologists emphasize that AI should complement rather than replace traditional methods, especially for predicting extreme and rare weather events where historical data is limited. The research team plans to continue refining the model and exploring operational integration with existing forecasting infrastructure. Read the full article at: https://sciencedaily.com/releases/2024/05/240507123456.htm