Google DeepMind opens PARDIG AI rainfall data to researchers and developers
In brief
- PARDIG estimates rainfall every 15 minutes on a roughly 10-kilometer global grid.
- Historical archive dates back to 1999; a separate dataset covers 2016 to the present.
- DeepMind says detection skill beat NASA's IMERG-Final by more than 30% for light-to-moderate rain.
- Near-real-time data arrives in six-hour batches; Google calls the system experimental.
How PARDIG works
Google said PARDIG learned from precipitation measurements collected by NASA's Global Precipitation Measurement satellite, and that's where it parts ways with older approaches. Conventional systems depend heavily on ground-based weather stations. PARDIG doesn't, so it can generate estimates over oceans and in regions with sparse observation networks.
The data isn't only for outside researchers. It also supports Google's WeatherNext 3 forecasting model, which used PARDIG-generated rainfall information during training.
Researchers get historical rainfall records dating back to 1999, plus a separate dataset covering 2016 to the present. Google released both under a Creative Commons license that permits commercial reuse with attribution.
DeepMind's own benchmarks
DeepMind said PARDIG achieved more than 30% higher hourly precipitation detection skill than NASA's IMERG-Final dataset for light-to-moderate rainfall. That's the company's figure (drawn from its own evaluation and research paper), not an independent test. DeepMind said it checked the system against radar networks in the United States, Europe and Japan, along with observations from more than 36,000 weather stations worldwide.
The research paper also reported improved identification of intense rainfall events, which existing global precipitation datasets can struggle to capture. Google pointed to the July 2025 Texas Hill Country floods as one example, comparing PARDIG's estimates with ground-based radar and NASA's existing rainfall dataset.
The limits Google flagged
It's not a live warning system.
PARDIG estimates rainfall at 15-minute intervals, but Google publishes near-real-time data in six-hour batches, creating an average delay of about five hours between observation and availability. Cloud Storage delivers the estimates with roughly two to eight hours of delay. CryptoSlate described the system as experimental and noted that emergency services can't use the dataset as an immediate substitute for live weather warnings.
Google also acknowledged that PARDIG can underestimate snowfall in mountainous regions and is less reliable in certain polar areas. CryptoSlate suggested the release could still expand access to rainfall data in developing countries and remote regions where weather radar coverage is limited.
Frequently asked questions
What is Google DeepMind's PARDIG system?
PARDIG is an AI-powered global precipitation monitoring system. It combines satellite observations and atmospheric data to estimate rainfall every 15 minutes on a global grid with roughly 10-kilometer resolution. Google said it learned from measurements collected by NASA's Global Precipitation Measurement satellite.
Can PARDIG replace official flood warnings?
No. PARDIG is described as experimental precipitation estimation that doesn't replace official warnings. Near-real-time data is published in six-hour batches, with an average delay of about five hours, so emergency services can't use it as an immediate substitute for live weather warnings.
What PARDIG data can researchers access, and under what license?
Researchers can access historical rainfall records dating back to 1999, plus a separate dataset covering 2016 to the present. Google released both under a Creative Commons license that permits commercial reuse with attribution.


