Between November 2019 and February 2020, three successive storms – Ciara, Dennis and Jorge – hit the UK, causing rivers to swell to record levels across the nation and flooding thousands of properties. 

According to the 2017 CCRA report, an estimated 1.8 million people are living in areas of the UK that are at significant risk of coastal, surface or river flooding, and flood damages currently cost the nation around £1.3 billion pounds’ worth of damage a year. 

With global warming, flood risk is on the rise globally, which is why building resilience to flooding is key to climate adaptation. Partnering with the national Environment Agency (EA), we are working to improve how flood risk is mapped and assessed using Artificial Intelligence (AI), so that we can better mitigate and manage future disasters.

Seeing the bigger picture

Imaging plays a pivotal role in the recording of flood events. Historically, pictures were taken by EA staff in the field to document the extent and levels of flood water, and to direct emergency response teams if necessary. 

More recently, images have been captured from a chartered plane to observe the wider picture of flood boundaries from an aerial perspective. Both image types hold valuable insights that can be used for predicting floods and protecting communities on the ground. These can be used to help emergency responders identify areas of high risk and aid researchers in understanding flood behaviour to help prepare for future flood events. It can also help assess the scope of damage after flood water has receded. 

But there are drawbacks for both methods. Images taken from the ground often do not include the GPS local information that identifies where they were taken, and aerial images can contain inaccuracies due to factors like challenging flight conditions and missing metadata.

Harnessing AI tools to mitigate climate risk

To solve these issues, we created two projects using AI tools. The first phase was to automatically detect the location of each image when location data was absent. The second phase was to detect and map flood boundaries from oblique images (aerial photography captured at a 40-45 degree angle with the ground). 

To do this, our data experts developed an algorithm that projected an image to a map, which an AI system then used to match key landmarks to the image beneath. Through the combination of both techniques, we could then build a fuller picture of the extent and scope of the flood, and therefore gain a better insight into the steps needed to lessen the negative impacts of flooding.

“Our work with EA shows how AI can play an integral role in flood prevention, from using data to guide decisions about water flow in cities, to monitoring water levels and building more effective flood defences.” 

Sustaining livelihoods and communities

As a result, these projects automated the process of recording and significantly reduced manual work, allowing engineers more time to design flood prevention solutions. 

From the destruction of homes and infrastructure to the displacement of entire communities, flooding has major consequences for humanity – and as global temperatures rise, the number of extreme rainfall days is expected to increase. 

This is why by working to improve flood prediction and flood monitoring, we can help avoid the worst effects of flooding and inform decision-makers in locating where and how much support is required. 

Our work with EA shows how AI can play an integral role in flood prevention, from using data to guide decisions about water flow in cities, to monitoring water levels and building more effective flood defences. 

With the planet in the grip of a climate emergency, projects like these are essential in helping us better understand how we can harness new technologies to reduce climate risk and mitigate against future emergencies, so that communities and ecosystems can continue to thrive.