Just over a week ago, Super Typhoon Ragasa ripped through communities in the Philippines, Taiwan, Hong Kong and southern China. In Taiwan, a lake burst resulted in a deadly wall of water pouring down a mountain, killing at least 14 people, injuring 18 and leaving 33 people unaccounted for.
Now, a new AI flood forecasting tool is giving hope that earlier warnings can be given and lives saved.
More than one-fifth of the world’s population are vulnerable to flood risks. Early warning systems have been shown to reduce deaths by as much as 43%. Vinh Ngoc Tran, a hydrologist at the University of Michigan, and colleagues have developed a machine learning tool which sifts through existing flood models and seeks out and learns from the occasions when flood predictions were poor.
Combining this machine learning with the National Water Model, developed by the US National Oceanic and Atmospheric Administration, the researchers showed that they could boost flood prediction accuracy as much as sixfold. The study is published in AGU Advances.
The AI tool can be adapted to work with other existing flood models and could be a gamechanger for flood prone regions all over the world, providing detailed flood forecasts up to several days in advance and giving communities vital time to prepare.
