New open datasets released to advance weather and climate modeling

Researchers and developers can now access two foundational datasets, SatWeave and PARDIG, which are being released to support the creation of new applications in weather, hydrology, and climate science. These datasets are integral to the recently launched WeatherNext 3 model. SatWeave provides a global mosaic of geostationary satellite imagery across 17 channels, updating hourly, while PARDIG is an AI-based global precipitation dataset trained using sparse observations from the NASA Global Precipitation Measurement (GPM) Core Observatory. SatWeave integrates data from five satellites—GOES-East, GOES-West, Meteosat 0, Meteosat Indian Ocean, and Himawari—into a coherent global map, offering hourly live feeds and a historical archive spanning over a decade. PARDIG, which has a 15-minute temporal resolution and approximately 10-kilometer resolution, outperforms leading baselines like NASA’s IMERG-Final, improving hourly precipitation detection skill by over 30% on light-to-moderate rain. These resources, which include historical PARDIG archives from 1999 to 2026, are now openly available on Google Cloud Storage, with PARDIG catalogs also accessible in Google Earth Engine. The development effort involved teams from Google DeepMind and Google Research, in collaboration with Google Cloud and Google Geo, and acknowledges contributions from organizations including ECMWF, NASA, and NOAA.