🌊 How AI is Changing Flood Prevention

Imagine iconic coastal cities like Shanghai and Mumbai gradually sinking under the relentless rise of sea levels. By 2050, this could be a reality for around 150 million people globally, their homes threatened by the encroaching tides. In the US, flood damage costs are projected to soar by 26 per cent in the next three decades, while the UK faces a similar increase in flood risk. This looming reality reshapes our world.

Researchers have turned to machine learning, analysing hundreds of factors, from people’s insurance choices to local safety rules, to determine the best ways to protect against floods. Their findings? People usually buy flood insurance only after a disaster strikes, leaving less frequently flooded areas more vulnerable. Moreover, high resident turnover in cities often erases the memory of past floods, weakening defences.

However, there’s a silver lining. Proactive community-level policies are stepping up, offering perks like lower insurance premiums to incentivise resilience. These strategies particularly benefit underserved and high-risk areas, potentially bridging the gap in flood risk and protection. Researchers champion this approach as a key to levelling the playing field in flood defence.

This is just one example of machine learning impacting flood prevention. Across the globe, it is transforming how we predict, prepare for, and manage flood risks. From precise forecasting models to analysing terrain for improved floodplain management, machine learning is crucial in protecting communities.

In 2018, Google’s flood forecasting system began providing real-time flood warnings, focusing on large rivers. By 2021, it operated in India and Bangladesh, covering regions home to over 350 million people. During the monsoon season, the system accurately predicted floods and sent warnings to 22 million affected residents.

In Georgia, a nationwide multi-hazard early warning system uses machine learning to forecast severe weather events like hail and windstorms. The model improves its flood predictions by learning from recent local weather observations and a large amount of past data. It uses special techniques to make these predictions more precise, particularly in areas with complex topography.

Machine learning models are transforming flood prediction with their efficiency and cost-effectiveness. They quickly interpret complex flood patterns using historical data, bypassing the need for deep knowledge of physical processes. This rapid, minimal-input approach provides accurate predictions essential in time-sensitive scenarios, marking a significant advancement in outsmarting the unpredictability of floods.

According to ML Sense, traditional flood prediction relies on guessing or monitoring a camera feed attached to some motorways. Employing machine learning allows companies to gather additional data sources: conditions of pipes data, historical weather data and historical flood index data. In the UK, combining these data sources helped to create a week-in-advance flood warning system with 80 per cent more floods predicted.

Steering machine learning in flood prevention is akin to navigating a stormy sea. Every step is laden with challenges. Imagine the complexity: datasets must be robust and unbiased, lest predictions veer off course for rare events like natural disasters. And as new AI technologies emerge, we’re balancing on a tightrope of data privacy, transparency, and managing vast data reservoirs. The crucial task is ensuring AI models don’t wash away socio-economic fairness under the tide of bias.

Cut to the model development phase, where machine learning models are like hungry beasts, needing a feast of training resources that aren’t always easy to come by. And once these models are built, there’s the Herculean task of making their outcomes human-friendly. At the end of the day, it’s people who need to understand and act on these insights. 

In the operational phase, the challenge shifts to clear and effective communication. The model outputs must be translated into a language that resonates with everyone, from the local community folk to the high-stakes emergency managers. Transparency in data and processes here is the beacon of trust, guiding decision-makers through the murky waters to ensure that the AI’s advice isn’t just smart but also relevant and timely.

Elia Kabanov is a science writer covering the past, present and future of technology (@metkere).

Illustration: Elia Kabanov feat. MidJourney.