Google announced Tuesday that a tool using artificial intelligence to better predict river floods will be expanded to the U.S. and Canada, covering more than 800 North American riverside communities that are home to more than 12 million people. Google calls it Flood Hub, and it’s the latest example of how AI is being used to help adapt to extreme weather events associated with climate change.
“We see tremendous opportunity for AI to solve some of the world’s biggest challenges, and climate change is very much one of those,” Google’s Chief Sustainability Officer, Kate Brandt, told Newsweek in an interview.
At an event in Brussels on Tuesday, Google announced a suite of new and expanded sustainability initiatives and products. Many of them involve the use of AI, such as tools to help city planners find the best places to plant trees and modify rooftops to buffer against city heat, and a partnership with the U.S. Forest Service to use AI to improve maps related to wildfires.
Courtesy of Google Research
Flooding and storms are among the most damaging climate impacts as a warmer atmosphere holds more moisture, fueling extreme precipitation. A recent study of the economic costs of extreme weather due to climate change found that about 75 percent of those costs came from storms and floods. In addition to the loss of life and property, storms and high waters displace millions of people, often the most vulnerable, exposing them to additional health threats.
A UNICEF report on children displaced by extreme weather found that from 2016 to 2021, floods and storms accounted for 95 percent of displacements globally, with nearly 41 million children uprooted. Projections in the same report, based on climate models, show riverine floods have the potential to displace almost 96 million children over the next 30 years.
Brandt said Flood Hub’s engineers use advanced AI, publicly available data sources and satellite imagery, combined with hydrologic models of river flows. The results allow flooding predictions with a longer lead time than was previously available in many instances.
“They kept tuning it to get to being able to predict it further in advance,” she said. “Right now, we’re able to predict it seven days in advance.”
Google’s flooding research effort started in 2018 and this year, Flood Hub started service in 80 countries in parts of Asia, Europe and South America. This August, Google said in a press release, emergency responders in Chile used the tool to improve evacuation alerts and minimize flooding impacts.
Courtesy of Google Research
Dr. Satchit Balsari is an emergency physician who uses AI and big data tools like Flood Hub to make medical systems and communities more climate resilient. Balsari is an Assistant Professor at Harvard’s T.H. Chan School of Public Health, where he co-directs Crisis Ready, a collaboration with the humanitarian organization Direct Relief and several academic groups and technology companies.
Crisis Ready offers data-driven decision making for disaster planning, and Balsari said that goes beyond the immediate effects of a disaster. He’s encouraging physicians and emergency responders to look at the “long tail” of health impacts from floods and other extreme weather events that often contribute to further death and disease.
“What is happening is people will be trapped in their homes and they’re unable to get to needed care, critical medications run out, or power went out,” Balsari told Newsweek, adding that displaced people will have trouble keeping up with vital health care. “We need to identify where these people are, where the vulnerable are, and where the infrastructure is that can support them.”
Balsari said Crisis Ready offers data showing real-time population shifts during extreme weather events, helping hospitals prepare for increased demand from an influx of displaced people. Over the long term, these tools can help emergency planners and other officials think of ways to make communities better able to withstand the type of events that climate scientists warn are becoming more likely in a warming world.
Some of these tools are already showing value but Balsari said much work remains. Many important data sets are still siloed, even within some government agencies, he said, and anything involving health information must be handled with great care.
The use of AI in weather forecasting presents its own unique set of challenges, according to Amy McGovern, who has dual appointments in the schools of meteorology and computer science at Oklahoma University. McGovern directs the Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography, which is funded by the National Science Foundation.
McGovern said AI holds promise to improve weather forecasting models, but the results have to be dependable, and that takes a lot of work with meteorologists and weather modelers.
“Trust is really a process,” she told Newsweek. “You don’t just hand somebody a method and say, ‘Here, trust it.'”
Forecasters work with the AI models to find their strengths and weaknesses, provide feedback and identify the best applications for the new technology. McGovern said her group is working on better predictions for flash floods by teaching weather forecasting models to better anticipate weather extremes.
“For these extreme precipitation events, sometimes they’re happening so extremely that they’ve never been seen before,” she said. “You’ve just never seen rainfall rates like that before.”
Traditional weather models are largely based on historic weather data. But in a changing climate, assumptions drawn from the baseline measurements of a more stable climate may no longer hold true.
“We’re having to train the models to correct that bias to actually give you an event that’s outside of the distribution that anybody’s ever seen before,” she said. “These types of events are starting to be more common.”
Weather extremes are part of what some call the “new normal” of climate change, and McGovern’s work shows that living in such an altered world will likely require something beyond normal intelligence. AI is rapidly emerging to fill that role. The question now is whether we can adapt fast enough to the changes rapidly coming our way.