Extreme Weather: AI-assisted Early Warning

AI can assist early warning systems that predict impacts of extreme weather events such as droughts and heavy rainfall

To the point:

• Particularly severe climate impacts: the latest climate report from the European earth observation service Copernicus shows that Europe is especially vulnerable to extreme events such as heatwaves, droughts, and heavy rainfall as a result of climate change.
• Similar weather extremes, different impacts: droughts, heavy rainfall, and other extreme weather events can have very different impacts in different ecosystems and populated areas.
• Effective early warning with AI: artificial intelligence opens up new possibilities for predicting the damage caused by extreme weather events anywhere in the world using an integrated system with high temporal and spatial resolution and also taking into account communication science and psychological methods.

2024 - another year of weather extremes

2024 once again clearly showed that Europe is especially vulnerable to climate change, and in a number of different ways. Southeast Europe experienced several particularly severe heatwaves in 2024, including the longest since weather records began. There was also a pronounced drought there. In the same year, other parts of Europe experienced the worst floods since 2013, which claimed 335 lives, affected around 413,000 people, and caused damage amounting to around 18 billion euros. In September, storm Boris caused devastating floods in eight countries in Central and Eastern Europe, while record rainfall in October led to a disaster in the Spanish region of Valencia.

Heavy precipitation in 2021: extreme weather with diverse impacts

But extreme weather events can have very diverse impacts. 2021 provides a particularly good demonstration of this fact. In July, persistent heavy rainfall led to a flood disaster that claimed many lives and caused several billion euros worth of damage, particularly in the Ahr valley but also in other parts of Rhineland-Palatinate and North Rhine-Westphalia. A very similar weather situation in Brandenburg in the same year meant that the parched sandy soils were finally supplied with plenty of water again. The example shows that precise weather forecasts are important, but they are not enough to reduce storm damage. This applies not only to heavy rainfall but also to droughts and other extreme meteorological events.

AI-assisted early warning systems to help with prevention

In a study published in Nature Communications, an international team led by Markus Reichstein and Vitus Benson, who conduct research at the Max Planck Institute for Biogeochemistry in Jena, Germany, developed a concept for new early warning systems supported by artificial intelligence. It is intended to help aid organizations and disaster control institutions to reduce damage caused by droughts or heavy rainfall, for example where it could be particularly severe. Organizations are already using extreme weather forecasts to avert disasters by preventive measures or at least to plan operations in disaster areas at an early stage. The improved AI-assisted early warning system enables them to deploy their resources in an even more targeted and efficient manner. "Early warning systems are usually designed for short-term periods of weeks to a few months to enable acute protective measures. But we should also think strategically about early warning systems over longer periods of time - from several years to decades - in order to plan and implement far-reaching preventive measures," explains Reichstein. Early warning systems can provide information on how societies can adapt to the extreme events that are becoming more frequent and more severe as a result of anthropogenic climate change. This can mean that infrastructure is expanded or settlements are relocated to avoid flood damage. However, it can also mean that agriculture and forestry are adapting to the changed climatic conditions and growing more drought-resistant crops than has perhaps been the case for generations.

Integration of communicative and psychological measures

The measures to prevent floods, crop failures, and famines can be very far-reaching and expensive. "We not only have to predict extreme weather events and their possible impacts as accurately as possible," emphasizes Markus Reichstein. "It is just as important to incorporate findings from communication science and psychology so that warnings are understood, taken seriously and translated into effective action - both individually and politically." Accordingly, the early warning concept presented by Reichstein and his colleagues provides for six modules: starting with temporally and spatially high-resolution measurements of extreme weather impacts, through precise weather forecasts and predictions of the ecological and economic impacts, to communication science and psychological methods designed to make warnings as effective as possible. Artificial intelligence can be useful both for predicting the damage caused by extreme weather events and for effective communication through voice, image, and sound.

AI uses examples to learn the impacts extreme weather events have had in different locations

With conventional physical climate models, it is not possible to calculate exactly what impacts a drought or heavy rainfall will have on different locations. This is because many influencing factors play a role. "In order to be able to accurately forecast the impacts of an extreme weather event, soil conditions, vegetation and terrain shape, for example, must be taken into account on a very local scale," says Reichstein. "We can predict with a resolution of 20 meters, i.e. for every field or garden, what damage a drought can cause, for example." This is made possible by the extensive data available from the Copernicus satellites. AI can then learn from the impacts of an event of this kind in a geologically and ecologically comparable area. Markus Reichstein's team is already very good at predicting how droughts will affect different ecosystems. Other research groups have developed algorithms that can predict the impacts of heavy rainfall events.

We need AI that understands causalities

The aim is to create early warning systems that reliably identify the impacts of various extreme weather events around the world, provide effective warnings and ideally also suggest measures to minimize damage. But there are still a few hurdles to overcome before then. This not only concerns the availability of meaningful data and linking of statements about effects on a large and small scale. This also relates to a problem that is currently playing a major role in the further development of artificial intelligence methods in general: the explainability of decisions made by AI. Artificial intelligence derives its results from statistical correlations and not from causal ones. "This March was drier than ever before in Germany. For an AI early warning system to be able to predict something like this, it can't just generalize and say that Germany won't have such extreme droughts in March - it has to refer to the physical causes, such as the general weather situation," says Vitus Benson.

A reliable early warning app for everyone

The team at the Max Planck Institute for Biogeochemistry is developing systems that make smarter recommendations, especially when it comes to forecasting the impacts of extreme weather events. "Early warning messages are often still rather general, especially in the Global South. Our developments in the area of predicting drought impacts are intended to democratize access and make even small-scale information available to everyone," says Benson. The team now wants to use this progress for other extreme weather conditions and then provide an early warning system in an app.

PH