# DeepMind opens live data behind its weather AI _Google DeepMind released SatWeave and PARDIG, live and historical data behind WeatherNext 3._ **Published:** 2026-11-15 **Source:** https://www.startuphub.ai/ai-news/ai-research/2026/deepmind-opens-live-data-behind-its-weather-ai --- What does it take to forecast rain where ground radar does not exist? [Google DeepMind](https://deepmind.google/blog/opening-the-datasets-behind-our-ai-weather-models/) says start with open data, and it just released SatWeave and PARDIG to power WeatherNext 3. SatWeave is a global mosaic that stitches geostationary satellites, spacecraft that hover over one spot on Earth, into one hourly view. It merges GOES-East, GOES-West, Meteosat 0-degree, Meteosat Indian Ocean and Himawari across 17 visible, water vapor and infrared channels, according to [DeepMind](https://www.startuphub.ai/ai-news/ai-figures/2026/figure-demis-hassabis-scientific-discovery-vision-2026-05-11). That matters because each satellite uses different projections, formats and 10 to 30 minute scan schedules, so most composites drop channels or update every few hours. That hourly refresh is the point. Models initialized on conventional analysis data wait six hours for a refresh, DeepMind said, while SatWeave lets WeatherNext 3 issue new global predictions every hour. The lab is releasing the live feed plus a historical archive spanning over a decade, and it points solar power and wildfire monitoring teams as early users who must now test satellite-native forecasting instead of waiting for model cycles. DeepMind’s WeatherNext page adds context the announcement skips, saying WeatherNext 3 predicts station-targeted temperature and humidity at 5km resolution and wind at 10km. PARDIG, or Precipitation AI Reanalysis: Densely Inferred GPM, is an AI system that learns rainfall directly from space radar to fill the gaps where gauges and ground radars are thin or missing. It trains on sparse observations from NASA’s GPM Core Observatory and uses SatWeave imagery plus atmospheric fields from ECMWF as inputs, with an architecture similar to WeatherNext 3, according to DeepMind. The output is a global rain map at 15-minute and roughly 10-kilometer resolution with about five hours average latency. Hydrology and flood modelers carry the clearest workflow change. PARDIG improved hourly detection skill over NASA’s IMERG-Final by more than 30% for light-to-moderate rain in DeepMind’s 2016 to 2023 evaluation against radar networks in the US, Europe and Japan plus more than 36,000 weather stations worldwide. It held that edge across 0.2 to 25 mm per hour and proved strongest on heavy extremes, including the July 2025 Texas Hill Country flash floods where DeepMind compared near-real-time PARDIG at 10km against NOAA MRMS radar at 1km and IMERG-Final. PARDIG is not a radar replacement where dense radar exists. Its 10km pixels cannot match MRMS detail and its five-hour delay rules out nowcasting, so teams with local radar keep their stack. What they can now do is pull live SatWeave and PARDIG feeds from Google Cloud Storage, with PARDIG history from 1999 to 2026 also in Earth Engine, and test flood, landslide and drought models in regions that never had that baseline. --- Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory. © StartupHub.ai. All rights reserved. You may not republish this article in full without a license. Search engines and AI research tools may crawl and summarize for reference. Bulk reproduction or model training on this content requires a license. See https://www.startuphub.ai/terms.