AI improves flood warnings, but it cannot eliminate the risk of disasters

LONDON, Oct. 15 — When floods hit parts of Europe in September, the scale of the destruction took people by surprise. The intense rainfall should not have occurred as it was predicted by advanced forecasting systems supplemented with artificial intelligence.

But forewarned didn’t mean double-crossed. Although the rains were accurately forecast, the impacts in the flooded areas were not – a fact that highlights the problems associated with increasingly frequent extreme weather.

AI has boosted weather forecasting by using a range of statistical tools to analyze years of historical data and predict patterns, and at a lower cost than traditional numerical weather forecasting.

AI technology can make more specific predictions ahead of events such as urban flooding or in complex terrain such as mountainous areas.

For example, the Google-funded GraphCast, a machine learning-based method trained directly from reanalysis data, was found to outperform traditional models. Reanalysis data is based on past forecasts, repeated with modern forecast models, to provide the most complete picture of past weather and climate.

But there are still gaps in knowledge, in how the information is used and in investments to strengthen data collection models, experts say.

“In some cases and for some variables, AI models can beat physics-based models, but in other cases the other way around,” says Andrew Charlton-Perez, professor of meteorology at the University of Reading in Britain.

One problem is that the effectiveness of an AI model is only as good as the information it feeds. When there is little input data, or when extreme events occur more often at different times of the year or in different regions, weather disasters become more challenging to predict.

“A good use of the AI-based weather forecasts would be to complement and improve our forecasting toolbox, perhaps by allowing us to produce larger ensembles of forecasts that enable accurate assessment and interpretation of the probability of extreme events” , Charlton-Perez added.

COMMUNICATION IS KEY

Since January, the European Center for Medium-Range Weather Forecasts (ECMWF), an independent organization that provides forecasts to European countries four times a day, has been using the Artificial Intelligence/Integrated Forecasting System (AIFS).

This data-driven forecast model quickly makes multiple forecasts and provides long-term forecasts of weather events such as cyclones and heat waves.

ECMWF measurements before the September floods were accurate, experts say.

Thomas Wostal, press officer at meteorological observatory GeoSphere Austria, told Context/the Thomson Reuters Foundation that their numerical models – including the ECMWF forecasts – forecast 300-400 millimeters (11.8-15.7 in) of rain locally, which came true .

But even with accurate forecasts, scientists say communication is crucial, especially in an era when climate change means extreme weather is becoming more common.

“I think what has happened with (the recent floods) … is that it is so rare – a once every 150 to 200 year event – that even if the weather models capture it, there is a fair degree of uncertainty exists,” Shruti said. Nath, a postdoctoral research fellow in weather and climate forecasting at the University of Oxford.

“You have to issue the warning in a communicative way, to the degree of severity that it could potentially have for people, then people could see that the cost of inaction versus the cost of action is actually much greater. So then they would actually (in) more resources,” she said.

EUROPE BEHIND THE CURVE?

Europe faces urgent climate risks that transcend policy and adaptation measures, a report from the European Environment Agency warns.

Extreme heat, drought, forest fires and floods will worsen in Europe even under optimistic scenarios of global warming and affect living conditions across the continent, the EEA says.

After the floods, European Commissioner for Crisis Management Janez Lenarčič said the disaster was not an anomaly.

“These extreme weather events that used to occur once in a lifetime now occur almost annually. The global reality of climate change has penetrated the daily lives of Europeans,” he said.

Some technology entrepreneurs say Europe is not ready.

Jonas Torland, co-founder of Norway-based 7Analytics, which develops models for predicting floods and landslides, said governments and companies in the United States had risk managers more accustomed to assessing environmental hazards, while the authorities in Europe were not ready for it.

“We often see substantial expenditure with minimal data support for informed decision-making,” Torland, whose models are used in the cities of Oslo, Bergen and Kristiansand, told the Thomson Reuters Foundation.

“While AI is a critical part of these models, governments are unfortunately not investing in or purchasing these advanced AI solutions,” he said, adding that he believed governments are “sticking with their old data providers and consultants.”

Data processing is also challenging because these complex AI models must make hourly updates as predictions change.

That requires both a lot of computing power and a lot of time, especially on a smaller scale.

A 1 by 1 meter grid, which 7Analytics uses for its predictions, is 100 times more detailed than a 10 by 10 meter grid, but takes more than 100 times as much time to process.

High computing power also means that huge amounts of energy and water are required, which makes AI models part of the problem because they contribute to the global warming emissions that are causing the climate crisis.

Some major technology companies, such as Microsoft and Google, are exploring the use of nuclear energy to power their massive data storage centers.

Other scientists emphasize that, in addition to refining their forecasting capabilities, authorities must also invest in physical solutions, such as developing areas where flood water can be safely stored and early warning systems.

They must also minimize development in flood-prone areas, given the likelihood of more intense climate change-induced flooding, and fulfill their pledges to limit emissions.

“It is not a matter of data, technology or knowledge. It is a matter of implementation, political will,” Friederike Otto, senior lecturer at Imperial College London, said in an email response to questions.

“As long as the world burns fossil fuels, the root cause of climate change, extreme weather events will continue to intensify, killing people and destroying homes. To curb this trend, we must replace oil, gas and coal with renewable energy.”

Sign up here.

Our Standards: Thomson Reuters Trust Principles.opens a new tab