The 2025 Atlantic hurricane season concluded on Sunday, fulfilling active year forecasts. The season recorded 13 named storms and three Category 5 hurricanes. For the first time in a decade, no hurricane made landfall in the U.S.
Hurricane Melissa, identified as one of the strongest Atlantic storms on record, struck Jamaica with 185 mph winds, causing widespread damage and multiple fatalities. Prior to Melissa's landfall, forecast models varied in their predictions for its trajectory. Google's DeepMind AI-based hurricane model accurately predicted both Melissa's path and its Category 5 intensity.
James Franklin, a former branch chief at the National Hurricane Center, assessed the performance of forecast models for the season. He indicated that Google's DeepMind model demonstrated high accuracy, referring to it as "the best guidance we saw this year."
Artificial intelligence has been integrated into weather forecasting for some time. Google's DeepMind represents a significant advancement, suggesting a potential evolution beyond traditional physics-based models.
- Physics-based models, such as the Global Forecast System (GFS) developed by NOAA, operate on equations that compute the movement of wind, moisture, and heat within the atmosphere to predict hurricane track and intensity.
- AI models, in contrast, are developed to analyze historical records, identifying subtle patterns and relationships that may not be apparent through human observation.
Engineers at Google collaborated with scientists from the National Hurricane Center and Colorado State University's Cooperative Institute for Research in the Atmosphere (CIRA) to develop their hurricane model. Kate Musgrave, a research scientist at CIRA, noted that while previous AI models were effective at predicting storm tracks, they were less accurate regarding intensity. The Google model improved intensity forecasting by incorporating historical data on hurricane development.
Musgrave suggests that AI modeling could be applied to forecasting other weather phenomena, including tornadoes and cold snaps. She anticipates that advancements in AI models will enable earlier and more precise forecasts of hurricane tracks and intensity, which is critical for public safety and timely evacuations in coastal areas.
The National Hurricane Center integrated the Google DeepMind model into its forecast discussions, particularly during the tracking of Hurricane Melissa. Wallace Hogsett, a science operations officer at the National Hurricane Center, affirmed that AI is expected to become a component of the hurricane forecast process. NOAA and the European Centre for Medium-Range Weather Forecasts are also developing additional AI models.
Franklin noted that AI models can be perceived as a "black box" by forecasters due to the non-transparent nature of their predictive mechanisms. Despite the increasing importance of AI models, Franklin and Musgrave do not anticipate them replacing established physics-based models or the judgment of experienced human forecasters.