Meteorologists use a range of forecast models during the hurricane season. With this season wrapping up, experts are evaluating the performance of different models. Google’s new model has surprised them with its exceptional accuracy in predicting hurricane paths and strengths. Brian McNoldy, a meteorologist and senior researcher at the University of Miami, conducted a preliminary analysis and found that Google DeepMind’s Weather Lab outperformed all other models this season. In contrast, the Global Forecast System, America’s primary weather model, performed the worst.

While the National Hurricane Center will provide official data on each model’s performance later on, this initial review hints at a shift in hurricane forecasting. The evident superiority of AI-driven models suggests that it might be time to phase out traditional physics-based models. According to Eric Berger, a meteorologist and space reporter based in Houston, relying more on AI weather models like Google’s is likely the way forward as they continue to evolve and improve.

McNoldy’s analysis includes two charts comparing track forecast accuracy and intensity forecast accuracy for all 13 named storms in the Atlantic Basin this season. The charts show various colored lines representing different forecast models, with the GFS (AVNI) depicted in orange at the top due to its poor performance. Developed by NOAA in the 1980s, the GFS still serves as the primary forecast system for the National Weather Service today.

In contrast to Google’s AI-based approach, the GFS relies on traditional physics and sophisticated supercomputers. The difference in their performance is strikingly evident on the charts, with Google’s model ranking at the bottom, indicating superior accuracy compared to other models—especially the GFS.

Hurricane specialist Michael Lowry highlighted the GFS’s significant forecast error for Melissa in a recent blog post, emphasizing its limitations compared to AI-based models. Lowry praised AI models like DeepMind for their rapid forecasting capabilities and ability to learn and adapt in real-time.

Hurricane Melissa’s impact on the Caribbean underscores how rising sea temperatures are intensifying storms. As climate change fuels more potent hurricanes, accurate forecasting tools become crucial for predicting their paths and strengths.

AI-based models offer a promising solution for adapting to a changing climate. DeepMind’s impressive debut has captured forecasters’ attention and signals a potential shift towards more effective hurricane prediction methods.