Canada launches AI-hybrid GEML weather forecasting system

Environment and Climate Change Canada has launched an operational hybrid AI weather forecasting system (GEML) that combines an AI model with traditional numerical forecasting and is supported by Shared Services Canada HPC.

• Main announcement: Environment and Climate Change Canada launched the Global Environmental eMuLator (GEML) as a hybrid forecasting system using spectral nudging to combine an AI solution (built on the open-source GraphCast model published by Google DeepMind) with traditional physics-based models; the system ran more than a year of testing and is now operational, supported by a High Performance Computing solution operated by Shared Services Canada.
• Background and details: Researchers found AI models can better predict large-scale synoptic patterns and longer lead times but are limited to a limited number of variables and smooth fine-scale details; the hybrid approach generates the complete set of variables (temperatures, precipitation, storms) required for everyday operational meteorology and retains smaller-scale physics for heavy rainfall and severe storms.