Updated
Updated · Ars Technica · Sep 8
Google Upgrades WeatherNext 3 for 1-Hour Forecasts as Satellite Data Cuts Lag
Updated
Updated · Ars Technica · Sep 8

Google Upgrades WeatherNext 3 for 1-Hour Forecasts as Satellite Data Cuts Lag

3 articles · Updated · Ars Technica · Sep 8

Summary

  • WeatherNext 3 now ingests satellite weather data, letting Google issue hourly forecasts instead of relying on updates tied to six-hour reanalysis snapshots.
  • Nearly all AI weather models have depended on reanalyses—global atmospheric snapshots that can lose detail from raw observations and leave a longer gap between current conditions and new forecasts.
  • Google also raised the model’s spatial resolution and expanded the machine-learning system, while adjusting processes to contain the added computing load.
  • A separate model trained on satellite-based precipitation estimates adds another rainfall forecast stream, underscoring how AI weather systems are moving closer to traditional models’ use of raw data.

Insights

Could Google's new AI weather model completely replace traditional meteorology, or will its struggle with extreme rainfall prove too dangerous?
How will the integration of raw satellite data in WeatherNext 3 change how emergency responders react to fast-developing cyclones?
If AI weather models still struggle with unprecedented extremes, is it safe to integrate them into critical global supply chain decisions?