Today for AI
HOT RADAR
opensourceHEAT 11.0°

AI CURATED HOT STORY · 11/13/2026

Google Open Sources WeatherNext 3 Weather Datasets

1 reports archived1 independent sourcesupdated 11/13/2026, 20:09:02
Synthesis & Latest Updates
1 Sources Cross-Validated

Google DeepMind has publicly released SatWeave and PARDIG, the two foundational datasets behind its WeatherNext 3 weather model, both hosted on Google Cloud Storage under CC BY 4.0. PARDIG uses a spherical-grid Transformer with ~10 km spatial and 15-minute temporal resolution, outperforming NASA IMERG-Final by over 30% on hourly light-to-moderate precipitation detection and remaining reliable in regions lacking ground radar.

LATEST/Google DeepMind has publicly released SatWeave and PARDIG, the two foundational datasets behind its WeatherNext 3 weather model, both hosted on Google Cloud Storage under CC BY 4.0. PARDIG uses a spherical-grid Transformer with ~10 km spatial and 15-minute temporal resolution, outperforming NASA IMERG-Final by over 30% on hourly light-to-moderate precipitation detection and remaining reliable in regions lacking ground radar.

Hourly heat curve

3 fully observed hours
Now
11.0
Peak
11.821:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 1 reports · Latest first
  1. IT之家 · 智能时代T2·78 pts

    Google Open Sources WeatherNext 3 Weather Datasets

    • SatWeave merges five geostationary satellites (GOES-East/West, Meteosat 0, Meteosat IODC, Himawari), updating hourly with over a decade of archives
    • PARDIG trains directly on sparse spaceborne radar observations at 0.1° spatial and 15-minute temporal resolution with ~5-hour latency
    • Evaluation spans US, European and Japanese ground radar networks plus 36,000+ global weather stations, with notably strong extreme-precipitation capture