Canada Launches Hybrid AI Weather Model for Severe Forecasting

Canada is preparing to launch a new hybrid weather forecasting model that combines artificial intelligence with traditional forecasting methods, a move that could significantly improve how the country predicts severe weather. Environment and Climate Change Canada says the new system will begin rolling out this spring and is designed to give Canadians earlier and more accurate warnings when major weather events are approaching. At a time when the country is experiencing more extreme climate related events, from heat waves and floods to wildfires and winter storms, the agency is presenting this model as a practical way to strengthen public safety and emergency readiness.

The new approach blends the speed and pattern recognition of AI with the proven strengths of physics based forecasting. According to the federal government, artificial intelligence can analyze decades of atmospheric data from across continents in minutes, learning relationships between pressure, wind, and temperature to estimate future weather conditions. But AI alone can miss smaller local details, so the hybrid system also relies on ECCC’s traditional GEM model, which captures factors such as local wind, temperature, and precipitation. The goal is a forecasting tool that is both broader in reach and sharper in local accuracy, especially when severe systems begin to develop.

Officials say the performance gains could be substantial. One of the biggest improvements is that major weather systems such as winter storms, heat waves, and atmospheric rivers may now be detected more than 24 hours earlier, compared with roughly eight hours of lead time under the current setup. ECCC also says the model improves confidence in predicting when weather conditions will begin and in tracking the path a storm is likely to follow. Another major benchmark is that the agency’s six day forecast should now be as accurate as its current five day forecast, a level of progress that would previously have taken years of research and development to achieve.

This announcement is also tied to a broader research strategy already underway inside the department. Scientists and meteorologists have spent the past year testing the hybrid model in parallel with the existing one and using it to see how it would have performed during past storms. At the upcoming EGU General Assembly, ECCC researchers are outlining several related projects that show how deeply AI is being integrated into Canadian forecasting research. These include GEML, or Global Environmental eMuLator, a global AI forecast model based on DeepMind’s GraphCast and fine tuned in house, as well as GDPS SN, an experimental hybrid AI and numerical weather prediction system that improves the operational Global Deterministic Prediction System through spectral nudging.

Researchers also highlighted PARADIS, a fully Canadian, physically inspired AI based weather forecasting model developed by ECCC and its partners. Together, these efforts show that the hybrid model is not a one off experiment but part of a coordinated strategy to bridge physics based science and machine learning. The department says this work is intended to improve forecasting across short term, medium term, and longer range time scales while keeping the process grounded in scientific rigor, operational relevance, and collaboration. It also reflects a growing recognition that adapting to a changing climate will require better tools for detecting dangerous conditions earlier.

Even with these technological advances, ECCC is making clear that human expertise remains essential. Meteorologists will continue to interpret the results, apply judgment, and communicate forecasts and warnings to the public. That matters because weather forecasting is not just about generating data. It is also about turning uncertainty into useful decisions that help people prepare, whether that means changing travel plans, protecting crops, or mobilizing emergency crews before dangerous conditions arrive. If the new hybrid model performs as expected, Canadians may soon get more warning before severe weather strikes and more reliable forecasts in the days leading up to it, giving the country a stronger tool for facing increasingly unpredictable weather.

Works Cited

Bond, Meredith. “Canada’s Weather Agency to Launch Hybrid Forecasting Model That Uses AI.” CityNews Calgary, 9 Apr. 2026, calgary.citynews.ca/2026/04/09/artificial-intelligence-hybrid-forecasting-environment-and-climate-change-canada/. 

“Canada to Launch Hybrid AI Weather Model to Strengthen Forecasting for Severe Weather.” Environment and Climate Change Canada, 9 Apr. 2026, www.canada.ca/en/environment-climate-change/news/2026/04/canada-to-launch-hybrid-ai-weather-model-to-strengthen-forecasting-for-severe-weather.html. 

Diaconescu, Emilia, et al. “Bridging Physics and Machine Learning to Enhance Weather Forecasting at ECCC.” EGU General Assembly 2026, 2026, meetingorganizer.copernicus.org/EGU26/EGU26-7801.html. 

“Severe Thunderstorm Watch Issued.” CochraneNow, 28 June 2022, cochranenow.com/articles/severe-thunderstorm-watch-issued-. 

Tsekouras, Phil. “What Does the Weather Forecast Hold for Toronto After a Rollercoaster Few Days?” CP24, 11 Apr. 2025, www.cp24.com/news/2025/04/11/toronto-to-see-some-sun-this-weekend-after-snowy-friday-morning/.