November 13, 2026 Science
WeatherNext Team

We are releasing PARDIG and SatWeave to help researchers and developers build new meteorological, hydrological, and climate applications.
Understanding and predicting Earth’s weather and water cycle begins with our ability to observe the planet in the present. Satellites and ground-based instruments monitor the Earth continuously and in great detail. However, turning these raw measurements into a coherent global picture and making them rapidly available is a major scientific and engineering challenge.
Today, we are releasing two foundational datasets—SatWeave and PARDIG—to support the global community in building meteorological, hydrological, and climate applications. These two datasets were key to the breakthroughs in our recently announced WeatherNext 3 model. SatWeave is a global mosaic of geostationary satellite imagery that provides the model with a real-time view of developing clouds and storms. PARDIG is an AI-based global precipitation dataset used to train the model’s precipitation forecasting capabilities. We are opening up both historical archives and live feeds for these datasets so that researchers and developers can use them in modeling and downstream applications.
SatWeave: A unified view of Earth from space
SatWeave combines imagery from five satellites covering different regions of the globe: GOES-East and GOES-West, Meteosat 0 and Meteosat Indian Ocean, and Himawari. Because each satellite uses different map projections, file formats, sensors, and scanning schedules (ranging from 10 to 30 minutes), it is difficult to form a coherent global map. Most existing global composites discard incompatible parts, offer four or fewer channels, or update only every few hours.


SatWeave fuses observations from six geostationary satellites into a global mosaic across 17 visible, water vapor, and infrared channels.
SatWeave updates hourly, enabling WeatherNext 3 to generate new global forecasts, whereas traditional models rely on analysis data refreshed every six hours for initialization. Along with this hourly live feed, we are also releasing a historical archive spanning more than a decade.
While we have already used SatWeave for medium-range weather forecasting in WeatherNext 3, we are excited to explore additional forecasting applications—such as solar power prediction and wildfire monitoring—with the broader research and developer community.
PARDIG: High-precision, high-resolution global precipitation data
Precipitation affects everything from food and water supplies to flood and drought risks, yet accurately tracking it on a global scale remains a major challenge. While ground-based radars and dense rain gauge networks provide high-quality data, they are not always available and do not offer good coverage across all regions.
To bridge this gap, we developed PARDIG (Precipitation AI Reanalysis: Densely Inferred GPM). Built on an AI architecture similar to WeatherNext 3, PARDIG uses SatWeave imagery and atmospheric fields from the European Centre for Medium-Range Weather Forecasts (ECMWF) as inputs, and is trained directly on sparse spaceborne radar observations from NASA’s Global Precipitation Measurement (GPM) Core Observatory. The result is a highly accurate global precipitation map with a 15-minute temporal resolution, a spatial resolution of approximately 10 km, and an average update latency of about 5 hours.


Hourly rainfall during the July 2025 flash flood in the Texas Hill Country, comparing ground-based NOAA MRMS radar (left, ~1 km) with near-real-time PARDIG (center, ~10 km) and IMERG-Final (right, ~10 km).
We evaluated PARDIG against three continental radar networks in the United States, Europe, and Japan, as well as over 36,000 weather stations worldwide. Across all analyzed rain rates, PARDIG outperforms leading global baselines, including NASA’s IMERG-Final—a research-grade precipitation estimate vital to countless applications. For light to moderate rain, PARDIG improves hourly precipitation detection skill by more than 30% compared to IMERG-Final. It also proves especially effective at capturing extreme precipitation, which existing global products often struggle to detect.
