Skip to content

Opening the datasets behind our AI weather models

Google DeepMind

The live feeds and historical archives behind WeatherNext 3 are openly available on Google Cloud Storage and Google Earth Engine.

Key points

  • SatWeave blends imagery across 17 channels from multiple geostationary satellites into an hourly global mosaic with over a decade of archive data.
  • PARDIG provides global precipitation estimates at 15-minute intervals and roughly 10-kilometer resolution with archives spanning 1999 to 2026.
  • PARDIG is trained on NASA GPM spaceborne radar, SatWeave imagery, and ECMWF atmospheric fields, updating with roughly five hours of latency.
  • Google reports PARDIG improves hourly precipitation detection skill over NASA's IMERG-Final by more than 30% on light-to-moderate precipitation.

Why it matters

Weather and climate researchers gain access to high-frequency global satellite and precipitation feeds that do not depend on dense ground radar networks.

Sources