Google’s AI revolutionises flood forecasting, extending prediction window to 7 days

In a groundbreaking development, Google’s AI has achieved remarkable success in predicting riverine floods up to seven days in advance, offering reliable forecasting services to residents in 80 countries. The findings, published in the esteemed science journal Nature, underscore the transformative potential of machine learning in mitigating one of the world’s most prevalent natural disasters.

This innovative approach has significantly enhanced the accuracy of global flood forecasts, extending the average prediction time from zero to an impressive five days. Leveraging a diverse array of data sources including historical events, river level readings, elevation, and terrain data, Google’s Machine Learning models have undergone rigorous training, enabling them to generate localised maps and conduct hundreds of thousands of simulations for each geographical location.

The comprehensive nature of this methodology has empowered the AI models to accurately anticipate impending floods, even in regions where data availability is limited. Particularly impactful in underrepresented areas like parts of Africa and Asia, where traditional flood forecasting has been hampered by the scarcity of streamflow gauges, Google’s AI has transcended these challenges by harnessing available data and applying machine learning techniques to basins with sparse information.

With a cumulative population of 460 million spanning 80 countries, Google has democratised access to these invaluable flood forecasts through a myriad of platforms including Google Search, Google Maps, Android notifications, and its proprietary Flood Hub web application, operational since 2022.

Looking ahead, Google remains committed to further exploring the potential of Machine Learning in refining flood forecasting models. 

Collaborating closely with academic researchers, the company aims to fine-tune its AI-driven approach with the overarching objective of developing a comprehensive, globally scalable flood forecasting platform that encompasses the entire forecasting process from data acquisition to dissemination of actionable insights.