Developing countries vulnerable to extreme flooding could soon receive up to five days' advance warning of major rainfall and flooding, giving them valuable time to leave or prepare for the impact of flooding.
That's the main claim of a new report Google released yesterday, which details a new artificial intelligence (AI) system designed by the company to predict both small-scale and extreme flooding much earlier than most existing flood warning systems.
Floods are the most common form of natural disaster worldwide, with nearly 1.5 billion people (or 19 percent) of the world's population directly exposed to the significant risks posed by flooding. Floods are currently estimated to cost around $50 billion in annual economic damage worldwide, but costs are expected to rise as climate change leads to longer and more intense rainfall.
Advanced warnings for major floods can therefore help governments, communities and businesses better prepare for their consequences and limit damage and disruption.
However, most flood forecasting systems currently only provide a day or two of warning, limiting the ability of governments, businesses and communities to respond. Google, on the other hand, claims that its new open-source AI system can provide alerts up to five days in advance with a level of reliability that is “as good or better” than current systems.
Google said improving access to reliable flood forecasts for developing regions that lack monitoring stations is “particularly critical.”
Current forecasting methods are often limited by their reliance on stream meters – or measuring stations along rivers – which are not evenly distributed around the world, meaning unmetered rivers are harder to predict, Google explains.
“For years, accurate flood forecasting at scale was not possible due to the complexity of the problem and the lack of resources and data,” said Yossi Matias, Google's vice president of engineering and research. “Given that only a small percentage of the world's rivers are equipped with flow meters, this posed an additional barrier to safety for people in developing countries, as well as for undeserved and vulnerable communities.”
The new AI model was trained using 5,680 existing gauges to predict daily streamflow in ungauged watersheds over a seven-day forecast period. These were then tested with the leading global flood forecasting software – the Global Food Awareness System (GloFAS) – for both short- and long-term scenarios.
In addition to advanced flood forecasting, Google said its new AI technology has also enabled the FloodHub system to provide real-time river forecasts up to seven days in advance, covering rivers in 80 countries, home to more than 460 million people.
Where possible, Google said it would provide flood forecasts through Google Search, Google Maps and through Android notifications, giving anyone with an internet connection access to flood forecasts and warnings.
“People turn to Google for accurate and useful information during crises so they can protect themselves and their families,” Matias said.
To develop its flood forecasting technology, Google partnered with a number of humanitarian organizations in Africa and India.
Working with Google in Mozambique during Cyclone Freddy in March 2023, aid organization GiveDirectly was able to provide cash transfers to more than 4,100 households affected by the extreme weather conditions based on the American technology giant's flood forecasts.
Google said the pilot provided the company with meaningful insights into the level of geospatial detail and turnaround time it will take for relief organizations to distribute money to the communities that will be affected.
“Our partnership with Google Research allows us to combine new insights from AI-based forecasts with local early warning systems to better reach people on the frontlines of the climate crisis,” said Vera Lummis, senior manager of Digital Innovation at GiveDirectly.
Could you or a colleague be recognized at the first Women in Green Business Awards? You can now submit your nominations for the awards.
#Google #unveils #model #predict #extreme #flooding #days #advance
