Globally, climate change is exacerbating the frequency and severity of flooding events. However, artificial intelligence (AI) holds promise in mitigating some of these impacts by being trained to offer precise warnings, even in flood-prone areas lacking traditional water gauges.
A recent paper published in Nature reveals that an operational AI model spanning 80 countries delivers more accurate river flooding predictions than the previously dominant system.
Developed by Google researchers, this model significantly enhances forecasting accuracy, particularly in regions like Africa, where flood gauge infrastructure is limited.
Remarkably, the AI forecasts are freely accessible to the public in real-time, underscoring their potential for widespread application in flood risk management.
Google’s AI modelers endeavored to forecast floods, including exceptionally devastating events, within a river’s watershed, even without traditional gauges.
Also read: Google joins US DOE with $35 million carbon removal credit pledge
In a comprehensive, collaborative effort involving numerous academics and experts from the EU’s renowned global flood forecasting system, known as GloFAS, which currently sets the benchmark, scientists constructed a predictive AI model.
This Google model harnesses various publicly accessible data sources, such as weather predictions, satellite imagery, topographical information, and soil composition.
It then employs AI algorithms to anticipate the areas likely to be impacted by flooding and estimate the water’s depth. The model demonstrated its efficacy through rigorous testing and refinement based on feedback from 5,680 watersheds.
Researchers discovered that with the aid of AI, they could forecast floods five days in advance in river basins lacking gauges with the same precision that GloFAS could only achieve on the day of the event.
Also read: AI could fuel climate misinformation, not solutions, warn environmental groups
Beth Tellman, chief scientist at Floodbase, a company specializing in technologies for flood insurance products in developing nations, concurred that Google’s model exhibited substantial advancement over GloFAS.
“If forecasts can be reliable, they could be used not just for early warning and evacuation to save lives, but to release strategic funding to save lives and property,” Tellman said.
She listed several examples, including “using money to evacuate animals, pile sandbags along the river, harvest rice crops early enough to save them or even stockpile gas and food at cheaper prices before the flood happens and prices spike.”
