Rain is hard to track worldwide: gauges and ground radars cover only parts of the Earth, and rainfall can change quickly. That leaves gaps in the information used to prepare for floods and other hazards.
PARDIG is software that learns patterns in satellite images and atmospheric analyses, which are computer-based descriptions of weather conditions. It learns to match those inputs to precise rain measurements from a radar aboard a satellite, then estimates precipitation across the globe.
PARDIG beats several established precipitation products in tests, with the biggest gains for intense rain. Weather forecasts also improve when a global forecast model is trained using PARDIG estimates, suggesting the maps could help both monitoring and prediction.