WeatherNext 2 improves cyclone forecasting

Google says its artificial intelligence model WeatherNext 2 achieves state-of-the-art accuracy in predicting how tropical cyclones move, intensify and organize. The findings appear in a study published in the journal Nature.

Predicting how a hurricane or cyclone will evolve remains difficult. Its track can change, its intensity can increase rapidly, and the distribution of its winds determines which areas will experience the greatest impacts.

Every hour counts. A more accurate forecast can give weather services more time to issue alerts and prepare evacuations, shelters and other resources.

What the model achieves

According to Google, WeatherNext 2 delivered state-of-the-art results in three important areas:

• Track: it better anticipates where the cyclone will move.
• Intensity: it more accurately estimates how its strength will change.
• Wind structure: it calculates how winds are distributed around the system.

The company describes the result as equivalent to approximately a decade of meteorological progress concentrated in a single model. That comparison comes from Google and summarizes the scale it attributes to the advance. It is not a guarantee that every forecast will be accurate.

Why it matters to you

Tropical cyclones are among the most destructive weather events on the planet. Better forecasts do not eliminate the risk, but they can help authorities act sooner and with more precise information about the areas that could be affected.

They could also improve infrastructure planning, the deployment of emergency teams and the communication of alerts. In practice, the value is not just knowing that a storm is coming. It is estimating its path, strength and wind reach more accurately.

The model will be open source

Google has announced that it will make WeatherNext 2 available to the global research community. Opening the model allows other teams to study it, compare it with their own systems and explore its use in different climate contexts.

That does not mean the model will replace meteorologists or that official alerts will no longer be necessary. Forecasts must continue to be validated with data, meteorological tools and local knowledge.

The next point to watch will be how WeatherNext 2 performs outside the evaluations presented by Google and how weather centers use its capabilities. If its performance holds up in real-world situations, each additional hour of warning could become a practical tool for reducing damage.