Traditional climate models are highly complex and are often slow to produce results. According to two data scientists at Aalborg University, this creates a significant challenge when attempting to generate timely evidence in climate science.
That is why J. Eduardo Vera-Valdés, Associate Professor in Mathematics-Economics, and Olivia My Töffner Kvist, PhD student in Climate Econometrics, are working to develop climate models that are both faster and simpler. While machine learning plays a role, advanced data processing is central to their work, particularly in handling vast amounts of temperature data in order to produce reliable climate model results more efficiently.
“We look for average patterns across the globe. Our aim is to better understand climate change. We are not producing climate science results as such, but rather faster evidence – evidence that is difficult to dispute and which can be used in public debate when influential sceptics need convincing,” Olivia explains.
The key discipline behind their work is econometrics. Although both researchers are deeply engaged with climate change, their backgrounds are rooted in mathematics and economics. Like many others, they discovered that mathematics – through econometrics – could become a powerful tool for understanding climate developments.
Econometrics combines statistical methods, mathematics, and computer science to analyse economic data, test theories, identify patterns, and forecast trends. In this context, those methods are now being applied to climate data.
From time series to climate modelling
Eduardo, who holds a Bachelor’s degree in Mathematics from CIMAT and a Master’s degree in Economics from CIDE, the Mexican Centre for Research and Higher Education, originally developed mathematical models for other purposes. During his PhD at CREATES, Aarhus University’s former Centre of Excellence in Econometrics, he recognised that these models could also be applied to climate science.
“At CREATES, I worked extensively with time series analysis. Since changes in temperature over time are among the most important indicators of climate change, I began adapting these mathematical models for use in climate research,” Eduardo says.
Concern about the pace of rising global temperatures formed the basis for the Paris Agreement, adopted at COP21 in December 2015. The agreement aims to limit global warming to well below 2°C above pre-industrial levels, while pursuing efforts to keep warming below 1.5°C. Achieving this goal requires a reduction in global greenhouse gas emissions of approximately 50% by 2030 compared with 2010 levels.
“Nobody knows exactly when the 1.5-degree threshold will be crossed. That is why models capable of predicting the pace of climate change are so important – and why we need faster models to document developments. Existing climate models can take months to produce results. Our models also process very large datasets, but they do so much more quickly. Data science is part of everything Eduardo and I do,” says Olivia.
Although their models simplify some aspects of traditional climate models, they still rely on extensive datasets and incorporate multiple variables and considerations.
So, we let Eduardo explain the difference between the existing climate models and the models Olivia and Eduardo are working on:
“While traditional climate models are essential, they are highly complex as they must account for all the physical constraints of the climate system. They are also chaotic in the mathematical sense of the word: tiny changes in initial conditions, like imprecise CO2 emission reports, can cascade into significant deviations in model output. This makes it that projections from these models require large computational power.”
He adds:
“In our work, we aim to abstract as much as possible from the need to model all variables in the system, at all points across the globe, to speed up computations. One projection from the time series models that we use takes a couple of seconds, while traditional climate models can take weeks. The trade-off is that climate models remain the state-of-the-art for long-term projections, while time series models are only reliable for short horizons, say a decade,” Eduardo explains.
Concern about climate awareness
Despite the ambitions established in Paris in 2015, both researchers believe that global progress has been uneven. While improvements have been made in some areas, other developments have continued to place pressure on the climate.
When asked when they believe the 1.5°C threshold may be crossed, Olivia responds cautiously:
“Our estimate would be within this decade – but of course we cannot be certain.”
Both researchers are also concerned about what they see as insufficient awareness of climate issues. This is one of the reasons why Eduardo has repeatedly applied for DDSA funding for events such as the recent Earth Day initiative at Aalborg University.
“More universities in Denmark should organise Earth Day events or similar initiatives. More researchers should engage with climate research, and we would very much like to see more experts in mathematics contributing to climate modelling,” Olivia says.
Olivia is expected to complete her PhD in late August at Aalborg University, where she has pursued her entire academic career. She initially obtained a Bachelor of Science in Mathematics-Economics, before progressing to a Master of Science in the same discipline.
Now, she and Eduardo are preparing a new research paper presenting their latest findings – work they hope will contribute to documenting and understanding critical changes in the Earth’s climate.