Can AI help predict and mitigate flooding?

A member of the Canadian Forces carries a roll of plastic as he walks along a dike protecting homes along the Ottawa River, Wednesday May 1, 2019 in Ottawa. THE CANADIAN PRESS/Adrian Wyld

As winter turns to spring and snow begins to melt, experts are turning to new forms of technology to predict and mitigate flooding, a hazardous natural phenomenon that is increasing in its intensity and frequency.

Researchers at the University of Ottawa are using artificial intelligence with the goal of being able to predict flooding sooner to allow authorities and communities to better manage the threat using real-time information.

On Friday, the Rideau Valley Conservation Authority issued the first flood watch of the year in eastern Ontario, with up to 15 mm of rain predicted in the national capital region this weekend. While no significant flooding is expected, excessive snowmelt and warmer than usual temperatures are raising uncertainty over the risk.

“Traditional flood forecasting relies on high historical data and basic hydrological or traditional models. But those don’t always work when weather patterns are changing so rapidly like we observed during these last days,” said Hossein Bonakdari, a professor in the University of Ottawa’s faculty of engineering during an interview on Newstalk 580 CFRA’s Live with Andrew Pinsent.

“That’s why AI is becoming a game changer.”

UOttawa researchers have been using a machine learning technique on the Ottawa River, known as the Real-Time Multi-Steps-Ahead Flood Forecasting Model, that predicts runoff and the overflow of water that occurs when the ground can no longer absorb it.

The technology is able to outperform traditional methods of prevention by extending flood warning times up to five days in advance, Bonakdari said. Teams are also combining satellite data with imagery to track real-time snowmelt and moisture levels, giving emergency planners a heads up for potential risks.

“AI it’s not just about prediction. We can use it like a decision maker,” he said.

He says researchers are learning from other catastrophic events across the world, including the severe flash floods that caused extensive damage and killed hundreds in Valencia, Spain last fall.

“We tried to highlight what happened and what was the reason behind this catastrophic event and transfer this information to Canadians,” he said.

Floods are the most common natural hazard in Canada, happening almost five times as often as wildfires, according to the University of Ottawa. The issue has hit the city of Ottawa hard in recent years, damaging more than 6,000 homes and causing nearly $10 million in damage between two major floods that struck the city in 2017 and 2019.

Bonakdari says the research can serve as an early warning system for other natural disasters, including wildfires, droughts or severe weather. He adds municipalities and authorities need to take the threat more seriously as flooding events increase due to the impacts of climate change by investing in green infrastructure, increasing drainage capacities and taking a flexible approach to urban planning.

“I believe we are a little bit behind from what we would expect,” he said.

“The floods are staying with us, its not just for today or tomorrow… it’s part of a bigger climate trend of how we experience floods. This is something that is in front of us and we need to be careful about that.”