On November 13, Google DeepMind announced PARDIG, a model that estimates precipitation worldwide, and opened access to its live output and historical archives. Researchers and developers can get both through Google Cloud Storage, and the historical data is also available in Google Earth Engine. PARDIG uses satellite imagery and atmospheric data to estimate precipitation in 15-minute intervals at roughly 10-kilometer resolution. Live estimates arrive about five hours after the observations, on average.
Google DeepMind says PARDIG improved hourly detection of light to moderate precipitation by more than 30% over NASA’s IMERG-Final dataset in its evaluations. The global coverage gives researchers a way to study rainfall in places with few ground radars, including for work on floods, landslides, and droughts.
