DeepMind releases SatWeave and PARDIG datasets behind its WeatherNext 3 weather model

Google DeepMind released two foundational datasets, SatWeave and PARDIG, that underpin its recently launched WeatherNext 3 weather model. SatWeave blends imagery from six geostationary satellites into an hourly global mosaic across 17 channels, while PARDIG is an AI-based global precipitation dataset at 15-minute and roughly 10-kilometer resolution with about five hours of latency, which DeepMind says beats NASA's IMERG-Final by more than 30% on hourly detection skill for light-to-moderate precipitation. Live feeds and historical archives, SatWeave spanning over a decade and PARDIG covering 1999 to 2026, are now available on Google Cloud Storage, with PARDIG catalogs also in Google Earth Engine.