DeepMind Says Its AI Can Predict Hurricanes Earlier Than Everyone Else
Its WeatherNext model, which will be open-sourced, can accurately predict both a storm’s track and intensity using lower-resolution weather data. Researchers don't yet fully understand how it does this.
DeepMind's AI model, WeatherNext, has demonstrated the ability to predict hurricanes with an earlier warning than previously possible, giving communities up to a day more lead time to prepare for catastrophic events like Hurricane Melissa. This extra day can be crucial for organizing evacuations, staging supplies, and deploying resources efficiently.
Researchers found that the model can predict cyclones with unprecedented accuracy, providing forecasters with predictions three days out that are as accurate as those from previous models two days out. The model's improved performance is attributed to its ability to capture both global weather patterns and local atmospheric conditions, which are critical for predicting a storm's track and intensity.
While the AI model uses lower-resolution atmospheric data than traditional models, it successfully predicts storm intensity, a challenge that previous AI models struggled with. This unexpected accuracy in predicting storm intensity has puzzled researchers, who are still trying to understand how the model achieves such results. The model generates 1,000 potential scenarios for a developing storm, providing forecasters with a range of possibilities to consider.
Google DeepMind has made the WeatherNext models open-source, inviting the research community to build upon and improve them.
Written by urgent.news from Wired Science's reporting — not their text. Machine-written; read the original for the full account.
