Artificial intelligence helps predict extreme meteorological events and their consequences

dry

Predictable climate stress, Prof. Dr. Markus Reichstein,
Max Planck Institute for Biogeochemistry, Jena


Droughts, heat waves and floods – such extreme weather and climate events are likely to increase and become more severe with climate change. Markus Reichstein, director at the Max Planck Institute for Biogeochemistry in Jena, is working with his team to predict their consequences. He is relying on large amounts of data and artificial intelligence. In this way, he hopes to make societies more robust in the face of climate extremes.

Markus Reichstein likes rosemary. Ideally, the scientist would like to have a bushy shrub of the Mediterranean plant in his garden in Jena. But the plant probably wouldn’t survive that for long. That’s because every few years there is a very severe cold snap – such as in the winter of 2020/21 – that prevents a large rosemary bush from thriving. Markus Reichstein, director at the Max Planck Institute for Biogeochemistry, can still use rosemary to illustrate his research on climate extremes. After all, the sunny, warm and dry climate in Jena is actually ideally suited for the plant, which originates from the Mediterranean region, and even the average temperatures in winter do not pose a problem. But just a few days at temperatures below minus ten degrees Celsius can kill the shrub. Markus Reichstein can use this fact to explain the different dimensions of climate and weather extremes.

Climate extremes refer to longer-lasting exceptional events such as droughts or heat waves. Extreme weather describes short and unusually violent events such as storms or heavy hail. However, there is no uniform definition of what is now classified as extreme: “First, you can look at the meteorological data and determine how frequent or just rare a particular event is at a location – for example, temperatures below minus ten degrees Celsius,” Reichstein explains. “But you can also look at such an event in a different way and analyze how far this value deviates from the mean. Even more exciting, however, is the question of what effects the event has and whether they are equally extreme, that is, unusually strong.” That’s because while native plants can take some very cold winter days without a problem and the effects are not extreme for them, they very much are for rosemary.

The far-reaching effects of climate extremes have been hard to miss in Germany in recent years, and some have been quite unexpected. The hot and very dry summer of 2018 posed major problems for foresters and farmers, leaving damage in German forests that is becoming increasingly visible. The exceptional drought also led to large-scale forest fires in some parts of Germany. These direct consequences were certainly to be expected. What was more surprising, for example, was that the low water level in the Rhine led to supply bottlenecks and that power plants ran out of cooling water. The extreme climatic conditions thus had a serious impact on both people and nature – and were also a major public issue.

For many ecosystem researchers, including Markus Reichstein, however, another extreme event was already a wake-up call: the heat wave in Europe in the summer of 2003. The high temperatures and the increased pollution of the air caused by various climatic effects cost the lives of several tens of thousands of people. In addition to people and the economy, nature also suffered badly from the heat. Using extensive data from measuring stations and remote sensing satellites, researchers were able to analyze the consequences of the heat wave for ecosystems in detail. Markus Reichstein’s research at the time focused on the carbon cycle and, in particular, on the carbon balance between the biogeosphere and the atmosphere. In other words, the scientist looked into the question of how much carbon dioxide, for example, plants and soil take up from the air and release into it. The strong influence that climate extremes have on the global carbon cycle was clearly shown in the data from the heat wave at that time: Normally, plants on the European landmass sequester large amounts of CO2 during their growing season in the summer half-year. This, as opposed to a carbon source, is called a carbon sink. But due to the effects of heat and, in particular, water shortages, plants grew much less in 2003. As a result, not only was the sink over Europe much weaker, European ecosystems actually became sources of carbon dioxide.

This finding startled the research community. Until then, researchers had assumed that human-induced climate change would cause plants in mid- and higher latitudes to sequester more CO2 in the future. This was because global warming and increased carbon dioxide levels in the atmosphere were believed to cause plants to grow earlier in the year and more vigorously. This was expected to slow the rise in carbon dioxide levels from human-caused emissions, and thus climate change. “But the 2003 heat wave was an eye-opening event,” says Markus Reichstein. Because even then, there was every indication that climate change would bring more extremes like droughts and heat in the future. And it showed that such an event is capable of temporarily transforming a carbon sink into a carbon source. If the vegetation takes permanent damage or dies, the effect can be amplified over the years. Reichstein’s team found that the impact of climate extremes on the global carbon cycle is about as large as the total carbon sink on Earth’s landmasses. And if climate extremes increase, atmospheric CO2 levels could continue to rise – a feedback between the atmosphere and biogeosphere that further accelerates climate change.

Of course, how seriously extreme events affect humans and nature also depends on their frequency and severity. Predictions of whether climate change will lead to more extremes therefore help us to estimate what is in store for us. For a long time, such predictions were based on abstract considerations. Detailed predictions were not possible because of insufficient data on rare extreme events in the complex climate and weather system. But from basic thermodynamic considerations, researchers concluded that climate change would make extremes more likely and thus more frequent. This is because the global rise in temperatures means that there is more energy in the Earth system, more water evaporates, and the atmosphere can also absorb more water. So the weather picks up speed – there will be more extremes such as heat waves, heavy rain or storms.

As has been shown in recent years, climate science has been correct in its argumentation. Researchers have indeed succeeded in attributing the increase in extremes to climate change. Their analysis is based on improved global and regional climate models calculated on powerful computers. At the moment, the attribution is most successful for heat waves; for other extreme events such as droughts, heavy rainfall or floods, the statements are less reliable. The researchers do not establish a causal relationship between a specific event and climate change, but calculate how much more likely such an extreme event has become due to climate change. To do this, they compare the probability of an extreme event in a world without man-made climate change with the probability in a world with climate change. This is like rolling two dice many times to compare how often a particular number occurs. One of the dice, that of our real world, has been thrown by climate change. For example, for the devastating bushfires in Australia in 2019 and 2020, an analysis by the World Weather Attribution research initiative found that climate change has increased the risk of such an event by at least 30 percent.

Markus Reichstein wants to go one step further: He doesn’t just want to attribute climate extremes to climate change in retrospect; rather, he wants to predict them as accurately as possible for a region or location. His group is relying on big data and artificial intelligence to first develop a better understanding of climate extremes. In this way, the team hopes to enable a spatially high-resolution forecast of climate extremes and, above all, their effects, thus contributing to an early warning system. To this end, the scientists are bringing together a wide range of data and linking very different types of information. For example, they combine meteorological measurement data with data describing ecosystems. Artificial intelligence methods help them process and merge the data. For example, they can compare temperature and precipitation values with plant activity determined from satellite images and also analyze the carbon dioxide concentration measured near the surface. The researchers generate a world map for all variables that characterize the state of an ecosystem. For example, a drought stress map is created, which is then available for many past points in time, i.e. with high temporal resolution. Along with longitude and latitude, the time slices form the third dimension of this so-called data cube. Ultimately, this allows researchers to assess, for example, how severely drought has damaged and will damage vegetation over time. The particular strength of the approach lies in the fact that spatial relationships are revealed and the temporal development becomes clear. In this way, scientists detect anomalies. For these deviations from the norm that characterize an extreme, they then analyze the various state variables and obtain a multidimensional picture of the complex interplay.

The scientists used a past extreme event to investigate how different factors contribute to the development of a climate extreme and influence its effects: a heat wave that occurred in Russia in 2010. At that time, temperatures there rose to over 38 degrees Celsius and were more than ten degrees Celsius above average for several weeks. At the same time, there was a massive drought – a devastating combination: there were crop failures, forest fires and peat fires. Here, too, tens of thousands of people died, not only because of the high temperatures, but also because of air pollution caused by drought, heat and fires. Reichstein’s team found in the data cube analysis, however, that the consequences for nature were not so clearly negative. That’s because the extreme meteorological event didn’t quite match up spatially and temporally with trends in plant productivity. In the mid-latitudes, which are dominated by agriculture, the expected effect was seen: the hot and dry summer caused the plants to stop growing and wither, and plant productivity collapsed. But in the higher latitudes with extensive forests, the mild spring and unusually hot summer triggered early and strong plant growth. The extreme meteorological event thus had very different effects on ecosystems in different regions.

Artificial intelligence to recognize patterns of extremes
For a robust prediction of climate extremes, it is crucial to learn from as many different such events as possible. After all, it is only from the in-depth analysis of many data that a clear and generalizable picture of the complex interrelationships emerges. This is precisely the strength of artificial intelligence, or more precisely, of machine learning methods, which can recognize patterns in unknown data. Markus Reichstein’s research group is therefore working with Bernhard Schölkopf, director at the Max Planck Institute for Intelligent Systems, and other researchers at the European Laboratory for Learning and Intelligent Systems (Ellis) to further develop machine learning algorithms and use them for Earth system research. With the help of artificial intelligence, the Jena researchers are not only studying the impact of extreme events. The analyses should also improve the understanding of causal relationships, how ecosystems and climate influence each other. Meanwhile, climate extremes that have occurred worldwide in recent decades are piling up in the data cubes. And so the scientists hope that artificial intelligence will detect revealing patterns in the data. This could also link risk factors or indirect consequences to a climate extreme that would otherwise be unlikely to be associated with it. “If we merge the results of these analyses with established climate knowledge and with models, it would be possible in the future to predict the risk of a climate extreme and, above all, its effects to within 20 meters,” Markus Reichstein explains.

With the findings that the geoscientist and his colleagues around the world have since gained on climate extremes, he also wants to make his voice heard in society and politics. He is supported in this by Dorothea Frank at the Max Planck Institute for Biogeochemistry: “We want to raise awareness of the danger posed by the fact that climate change is making extreme weather and climate events increasingly likely,” says the researcher, who is jointly responsible for many projects and initiatives in this context. Because one thing is clear: Even if efforts to slow and stop climate change are successful, climate extremes will initially increase worldwide. The forces of nature will encounter constantly changing social conditions. The Jena researchers therefore want to use findings from various scientific disciplines to better prepare society for the challenges posed by climate change. “Particularly in the case of systemic risks, which arise from the interaction of natural systems with the economy, politics and individuals, it is crucial for understanding to look at the development from a natural science, economic, psychological, sociological and historical perspective,” says Reichstein.

Reliable forecasts help make societies more resilient
Dorothea Frank and Markus Reichstein are currently trying to strengthen scientific exchange on extreme events, disaster risk reduction and governance with the Risk Kan initiative, which brings together numerous international colleagues. Together, they want to develop recommendations for action to deal with systemic risks. After all, extreme climate and weather events are increasingly putting states and societies to the test. While rich and highly developed countries are often still able to avert the worst consequences of extreme events, resulting disasters in developing countries threaten many lives and make humanitarian aid operations necessary. One example is the 2011 drought in East Africa and the resulting famine in countries such as Ethiopia and Somalia. This disaster endangered more than ten million people, cost several hundred thousand lives, and caused nearly one million people to flee their homes. The World Bank estimates that up to 143 million people could become climate refugees by 2050, many of them due to the effects of climate extremes. That’s why it’s necessary to take action now, to take preventive measures and make investments. “The goal must be a sustainable society that is as resilient as possible to climate extremes,” says Markus Reichstein. The possible measures are diverse and depend heavily on the particular location: Near coasts or rivers, higher dams and floodwalls may be needed, while elsewhere new crops that are more drought-resistant must be introduced.

In this regard, robust predictions of the effects of climate extremes, such as those being developed by Reichstein’s team, are helping to make societies more resilient. For example, the team is currently working on a large, EU-funded research project to establish this approach in Africa. This is because an early warning system gives people in an affected region time to prepare for an extreme. Necessary funding could be released in advance to help local people and prevent a disaster. Forecast-based disaster relief is already in use today, but could be greatly expanded in the future and would benefit from reliable and accurate forecasts. Markus Reichstein is convinced of the strength of his data-based research approach and even believes it can be expanded: By using artificial intelligence to analyze climate and ecosystem data as well as socioeconomic data, researchers could also examine the vulnerability of societies to climate extremes. But even if vulnerable societies are identified or an early warning system based on data cubes sounds the alarm, the decisive factor in the end will always be how people respond. The Corona pandemic, of all things, gives Dorothea Frank and Markus Reichstein some encouragement in this regard. “Because this crisis has shown that our society – in Germany and worldwide – is quite capable of acting quickly and decisively,” says Dorothea Frank. “This determination is also needed now to counter the climate crisis and avert the massive impact of extreme events.”

About The Author

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Posts