A new study published in Science demonstrates that an artificial intelligence weather prediction model has achieved superior accuracy compared to the world’s leading traditional forecasting system. The AI model, developed by Google DeepMind, called GraphCast, significantly outperformed the European Centre for Medium-Range Weather Forecasts (ECMWF) high-resolution forecast (HRES) system in over 90% of 1,380 test metrics. These metrics included critical variables like temperature, pressure, wind speed, and humidity at multiple atmospheric levels. GraphCast operates by analyzing decades of historical weather data to learn cause-and-effect relationships within the Earth’s atmosphere, generating a 10-day forecast in under a minute on a single machine. While the ECMWF has begun integrating similar AI techniques, experts note that traditional physics-based models remain essential for understanding long-term climate trends and rare extreme events. The rapid advancement suggests a future where AI and conventional models work in tandem for more accurate and efficient weather prediction. Read the full article: https://sciencedaily.com/releases/2023/11/231114155020.htm