Google expands AI-driven flood forecasting to cover 700 million people in 100 countries

Google has significantly expanded its AI-based flood forecasting system, now covering a global population of 700 million across 100 countries. This initiative aims to provide riverine flood warnings in more regions worldwide, enhancing preparedness and safety in vulnerable areas. To support research and collaborations, Google has also introduced a new API, granting partners access to real-time flood data and forecasting models.

In a recent blog post, Google detailed its efforts to expand the reach and accuracy of its flood forecasting system. Previously limited to 80 countries with coverage for around 400 million people, the enhanced system now spans 100 countries, benefiting an additional 300 million people. The tech giant attributes this expansion to advancements in AI, which allow it to process more labelled data and refine forecasting models.

The upgraded system now features a “frontier forecasting model” and a new architecture that improves prediction accuracy. This technology can forecast potential floods up to seven days in advance, an improvement over the previous five-day window, enabling communities to prepare sooner and more effectively.

Alongside the expansion, Google has announced new opportunities for researchers and organisations to access its flood forecasting data through two main channels. First, the company is making datasets from its Google Runoff Reanalysis & Reforecast (GRRR) available to partners. Second, it has developed an application programming interface (API) to simplify data access for those looking to integrate flood forecasting insights into their work.

With the API, users can access critical information on expected flood status and hydrology forecasts, even in regions where local data may be limited. Interested researchers and organisations can now join a waitlist to gain access to this data.

Google has also added a new data layer to its Flood Hub, now featuring over 250,000 “virtual gauges” worldwide. These virtual gauges leverage geological and atmospheric data to predict riverine flooding risks. Although Google’s system can model flood risks globally, it only displays verified data in regions with historical records from reliable sensors.

With its extended forecasting capabilities and global data-sharing, Google hopes its flood prediction system will serve as a vital tool for public authorities, helping them develop emergency response plans to protect communities and infrastructure.