On November 13, 2026, Google DeepMind opened two datasets used by its AI weather models. SatWeave combines imagery from geostationary weather satellites, listed on the page as GOES-East, GOES-West, Meteosat 0, Meteosat Indian Ocean and Himawari, across 17 visible, water vapor and infrared channels. It updates hourly and comes with a live feed and a historical archive spanning over a decade.
PARDIG (Precipitation AI Reanalysis: Densely Inferred GPM) is a global precipitation map at 15-minute temporal resolution and roughly 10-kilometer spatial resolution, with average latency of about five hours and historical archives from 1999 to 2026. DeepMind reports it improves hourly precipitation detection skill over IMERG-Final by more than 30% on light-to-moderate precipitation, evaluated against over 36,000 weather stations and radar networks in the US, Europe and Japan.
Both are openly available on Google Cloud Storage, and PARDIG’s historical catalogs are also in Google Earth Engine.
Why it matters: Weather AI progress has been gated by data as much as by models. Publishing the processed inputs lets other groups train and test forecasting and nowcasting models on the same footing as a leading lab.
What it does not show: The page does not state dataset sizes or the license. The 30% skill gain is DeepMind’s own evaluation, and a figure caption mentions six satellites while the text lists five.