How Mathematics and Computer Science Help to Tackle the Climate Crisis

Maybritt Schillinger

Understanding and fighting the climate crisis requires the collective effort from many disciplines. The role of fields like politics or economics seems rather evident. The contribution of mathematics and computer science might be less obvious – it was discussed in the Hot Topic panel discussion “Understanding the Climate Crisis” at the 11th Heidelberg Laureate Forum (HLF) that took place in September, 2024.

Three panelists participated: Aglaé Jézéquel from the Laboratoire de météorologie dynamique at the Ecole Normale Superieure in Paris, Jakob Zscheischler, who is head of the Department of Compound Environmental Risks at the Helmholtz Centre for Environmental Research in Leipzig, and Beatrice Ellerhoff, a climate scientist at the German Weather Service.

Mathematical Problems in Environmental Science

The speakers gave an overview of their respective research fields and pointed out where mathematical challenges arise. Extreme event attribution sparks a lot of media interest, as Aglaé Jézéquel explained. In attribution science, researchers consider the likelihood of an extreme weather event in the world today. They compare it with the likelihood in a counterfactual world without anthropocentric climate change. Hence, she often repeats a similar message to the media: Heatwaves are getting more and more likely due to climate change. The field of attribution includes many statistical challenges, for example for very rare and unprecedented events: There is limited observational climate data available, and estimating probabilities for events that are far beyond the previously recorded events is very difficult.

Further, what if the extreme event is not only extreme in one aspect, but in several? This is the field of Jakob Zscheischler. He is particularly interested in co-occurring events, for example a joint drought and heatwave which could cause wildfires or crop failures. Such joint events can pose higher risks for impacts than events where only one aspect is extreme. For a correct risk analysis, scientists need to model and understand the dependency between drought and heatwave occurrences properly.

Finally, physicist Beatrice Ellerhoff investigated temperature fluctuations in her PhD. She had a special interest in comparing statistical properties of temperature time series across different timescales. She now works at the German Weather Service in a very interdisciplinary team. Their goal is to monitor greenhouse gas emissions and better quantify contributions from several sectors such as agriculture or traffic.

Challenges in Applying Machine Learning to Climate

Applying mathematics and computer science to problems in environmental science brings special challenges. For example, applying machine learning has been very successful in other scientific disciplines like weather forecasting. However, it cannot be readily applied to problems in climate science. A machine learning model trained to predict good weather today is not very useful in a much warmer future world with a different state of the atmosphere. In addition, the atmosphere is chaotic and, consequently, random weather fluctuations interfere with the average climate change signal. Thus, it is easier to predict the mean future climate, but modelling its full variability becomes very difficult. Climate is not a pure prediction problem – instead, we also require process understanding.

Climate scientists rely on process-based models to simulate the climate of the future. These models numerically solve physics-based differential equations to compute the climate. However, these models are computationally very costly. Machine learning and computer science more generally can help to speed up these models.

Additionally, some sub-questions in environmental science can be tackled with machine learning. Zscheischler gave one example in his presentation at the beginning: Using explainable machine learning, his team investigated drivers of floods. Firstly, they trained a machine learning model to predict floods from several climate variables, such as snowfall or rainfall. Secondly, via explainability tools, they could identify which climate features contribute most to a given flood. In addition, they analyzed which floods have single or multiple drivers.

Bridging and Communicating Research Across Disciplines

All three panelists believe that bridging disciplines is key to success. However, this might require structural changes. During the discussion, participants mentioned the prominent role of statistics in climate science several times. There is a need for advancing statistical methods in climate science. But young researchers working at the intersection of these disciplines eventually face the decision in which journals to publish. It can be hard to find academic positions later if one works as a statistician that publishes in climate science. Developments from computer science give one hope. The number of positions in data science has increased in recent years, where bridging computer science and applications is possible.

The interdisciplinary nature of environmental research is also visible in the personal background of the panelists. Jézéquel conducted research in both social as well as physical climate science. Her current work is connecting these two fields more and more. She addresses the climate community, but also tries to reflect on the values that her scientific community has. Zscheischler studied mathematics, and decided to move into environmental research by an internship at the Potsdam Institute for Climate Impact Research. Ellerhoff is a physicist and did her Master’s thesis in quantum computing. At that time, she became interested in science communication. That interest eventually lead to the question of how to apply her methodological skills also to other relevant disciplines.

For communicating science, Ellerhoff highlighted the importance of finding one’s own niche within the many options. She prefers writing over other media like videos and is currently working on a book about climate models. Jézéquel is in touch with newspapers and TV in France. She is also curious about the opportunities of connecting art and science. For example, she started a project linking creative writing with climate science. There, students write short stories with a climate focus, assisted by scientists.

Balancing “Usefulness” and Scientific Curiosity

All speakers mentioned their desire to contribute to an important societal problem through their work. However, societal demand and scientific curiosity might not always match. Consequently, researchers need to trade off and find their own balance between these poles. In addition, which science is useful and needed in practice might not always be clear to the scientists. Hence, the panelists called for stronger collaborations between politics, industry and science – enabling a joint initiative to tackle one pressing problem of our time.

The post How Mathematics and Computer Science Help to Tackle the Climate Crisis originally appeared on the HLFF SciLogs blog.