Most waterways around the world lack accurate flow gauges, making flood forecasting difficult. Low-income countries suffer disproportionately due to lack of data, while rivers and lakes in higher-income countries are often better gauged and therefore more accurate at predicting floods.
Google's flood forecasting system can now predict floods up to four days in advance in data-poor regions like South America and Africa, while it can predict up to seven days in advance in data-rich regions like Europe and the US.
In 2018, Google introduced an AI-based flood forecasting system for India and Bangladesh, two of the world’s most flood-prone countries. Google has since expanded the number of countries and launched a website called Flood Hub. The system has high predictive power for several countries and helps warn people living there about potential flood risks.
As part of using artificial intelligence to tackle the climate crisis, Google AI has now expanded its flood forecasting capabilities to 80 countries.
From October 2022, with the addition of territories in 60 new countries across Africa, Asia-Pacific, Europe, South and Central America, the Flood Hub platform now covers some of the territories with the highest proportion of their population at risk of flooding and experiencing more extreme weather events, affecting 460 million people globally.
Google is working to expand flood warnings in Search and Maps to get this information to people when they need it most.
To develop the Google AI model, researchers fed it publicly available streamflow data collected by the World Meteorological Organization, as well as previous satellite and ground-based weather forecasts. Flood Hub's AI uses a variety of publicly available data sources. It then combines two models: a hydrological model, which predicts how much water will flow in a river, and a flood inundation model, which predicts which areas will be affected and how deep the water will be.
From 1984 to 2021, the researchers tested their model on more than 5,000 streamflow measurements from around the world and found that Google AI's accurate forecasts were up to four days ahead, while the best traditional forecasting systems could only predict two days ahead.
Source
Comment (0)