Google has announced an expansion of its artificial intelligence (AI)-based flood forecasting system, which will now encompass 100 countries and provide riverine flood forecasts to approximately 700 million people globally. Along with this geographical expansion, the tech company will make its datasets available to researchers and partners to enhance their understanding of the system and leverage its capabilities. Furthermore, Google has introduced a new application programming interface (API) to facilitate seamless access to this data.
Google Enhances Flood Forecasting Models and Extends Coverage
In a blog post, Google elaborated on its plans for the flood forecasting system’s expansion, noting that it now reaches a far broader audience than before—growing from coverage in 80 countries to 100 and increasing the population served from 400 million to 700 million.
The enhancement in coverage was attributed to advancements in research that allowed for access to a greater volume of labeled data essential for training its AI models. The company also implemented a new forecasting model that serves as a foundational input for its advanced forecasting capabilities, along with a fresh model architecture.
Additionally, Google has committed to making its flood forecasting model forecasts accessible to researchers and partners in two ways: through its existing datasets from the Google Runoff Reanalysis & Reforecast (GRRR) and through the new API currently in development.
This API will enable users to tap into Google’s hydrology forecasts and assess expected flood conditions in both urban environments and areas with limited local data. Interested partners and researchers can now register on the waitlist to gain access to the AI-driven forecasting model.
The Flood Hub within the forecasting system has also been bolstered with an additional layer of data that includes 250,000 forecast points generated using what Google refers to as “virtual gauges.” These virtual gauges utilize simulation-based predictions that integrate geological and atmospheric data to estimate the likelihood of riverine flooding. While this system allows for global discharge prediction, it will only display areas where the data can be corroborated by historical information from trustworthy sensors.
Moreover, Google noted that its updated system can deliver flood forecasts with a seven-day accuracy range, an improvement over previous systems that were limited to five-day predictions. The company anticipates that its AI-assisted forecasting capabilities will be invaluable for public authorities in affected regions, aiding them in the development of crisis response plans and enhancing infrastructure and public safety management.
