About 1,5 billion people around the world, or 20% of the total population, are at risk of serious damage from floods, with losses to the global economy estimated to be around $50 billion per year. So Google's flood prediction AI model is very valuable.
Published in academic journals Nature, Google's AI research can provide early warnings up to 7 days before a flood occurs. Predicting floods is considered relatively difficult because many rivers in the world have not yet installed flow meters, especially in developing countries.
Below is a chart showing the amount of flowmeter data on the vertical axis and each country's GDP on the horizontal axis. It is easy to see that there is a negative correlation between the data available in each country and GDP, meaning that poorer countries have less data that can be used to forecast floods.
Therefore, Google's AI model is launched to predict natural disasters and floods. Led by a team led by Gray Nearing from Google Research, this AI model uses data collected from 5.680 flowmeters around the world from 1980 to 2023. From there, the AI can Expands flood prediction by using historical flood information, thereby predicting floods up to 7 days in advance.
Initial evaluation results show that Google's AI can make flood predictions 5 days in advance with the same accuracy as conventional prediction software of the Global Flood Awareness System (GloFAS), which requires The day of the flood is coming.
This AI allows Google to send flood forecasts in 80 countries around the world, where 460 million people live, and provide forecast information through Search, Maps, Android notifications, Flood Hub... In its article , researchers say: “There is still much room for improvement in global flood forecasting and early warning systems. It is vitally important to the well-being of everyone around the world.”