Google has recently developed an artificial intelligence (AI) model that predicts floods in different locations worldwide up to a week in advance. The aim is to help people create early response plans and prevent severe damage caused by floods. About 1.5 billion people, which is 20% of the world’s population, are at risk of floods that can cause significant losses to the global economy, estimated to be around $50 billion per year.
Predicting floods is challenging as many rivers worldwide, especially in developing countries, do not have flow meters. Google’s AI research has been published in the academic journal Nature, and the AI model can provide early warnings up to seven days before a flood occurs.
A team of researchers led by Gray Nearing from Google Research to train the AI model used data collected from 5,680 flow meters worldwide from 1980 to 2023. The AI model expands flood prediction by using historical flood information, enabling it to predict floods up to seven days in advance.
Initial evaluation results indicate that Google’s AI model can predict floods up to five days in advance with the same accuracy as the conventional prediction software of the Global Flood Awareness System (GloFAS), which requires the day of the flood to be approaching.
There is a negative correlation between the amount of flow meter data available in each country and GDP, which makes it challenging to predict floods in developing countries. Therefore, Google’s AI model can be a valuable tool in predicting natural disasters and floods worldwide.
Google can send flood forecasts in 80 countries worldwide, where 460 million people live, and provide forecast information through Search, Maps, Android notifications, and Flood Hub. The researchers say that global flood forecasting and early warning systems still have much room for improvement, and such systems are vital to the well-being of everyone worldwide.