Floods are the most deadly natural disaster, affecting over 250 million people worldwide. They are most catastrophic in underdeveloped nations, where roughly 80% of all flood fatalities occur. These countries lack sophisticated weather forecasting systems that can accurately predict where severe flooding is likely to happen. Oftentimes people receive little to no warning before floods strike, and have no time to prepare and evacuate.

But a team of Google engineers are using machine learning to develop a forecasting system that can better predict where floods will travel. Their goal is to extend flood warnings from 1 day to 3, and to make this technology available to any country that needs it.