~/MACHINE LEAR/google-open-sources-weathernext-3-datasets-satweave-and-pardig

Google Open Sources WeatherNext 3 Datasets SatWeave and PARDIG

Google DeepMind has open-sourced SatWeave, a global geostationary satellite imagery mosaic dataset, and PARDIG, a global precipitation reanalysis dataset used by its WeatherNext 3 meteorological model. Both datasets are now available to researchers worldwide on Google Cloud Storage under a CC BY 4.0 license, featuring real-time data streams and historical archives spanning multiple decades. This release provides researchers and developers globally with high-resolution, real-time meteorological data that can improve weather forecasting, climate modeling, and downstream environmental applications like solar power prediction and wildfire monitoring. By making these foundational datasets openly accessible, Google accelerates innovation in AI-driven earth sciences and operational meteorology. SatWeave harmonizes observations from five geostationary satellites into an hourly updated global mosaic across 17 spectral bands, while PARDIG delivers precipitation data at a 10 km spatial resolution and 15-minute intervals. PARDIG has been shown to outperform baseline products like NASA IMERG-Final, demonstrating superior accuracy even in regions lacking ground radars and rain gauges.

## BACKGROUND

WeatherNext 3 is Google DeepMind's flagship global weather forecasting model, which generates hourly forecasts at high spatial resolution by ingesting live satellite observations. Datasets like SatWeave and PARDIG serve as the crucial observational inputs and training targets needed for AI models to understand complex atmospheric dynamics and precipitation patterns globally.

## REFERENCES

## KEYWORDS

#Machine Learning#Meteorology#Open Source#AI for Science#Climate Modeling

$ subscribe --daily