AI expands scope of climate science

AI climate science AI Climate science

The climate crisis is becoming more urgent and more serious, with increasingly extreme events happening more often. Ban Ki-moon, former Secretary-General of the United Nations, cited the example of the current heatwave driving record temperatures in Siberia (Russian Federation), the United States and Canada.

 

Globally, climate change is driving conflict and refugee flows, while growing numbers of species are at risk of extinction as ecosystems collapse, he noted at an AI for Good webinar hosted by the International Telecommunication Union (ITU) on 7 July.

 

Nearly 37 per cent of heat exposure around the world between 1991 and 2018 could be attributed to global warming caused by humans, Ban added, citing research published in Nature. Rising sea levels threaten the existence of many small island developing states, as well as some of world’s largest and most important cities, both rich and poor. “Climate threats do not discriminate,” he said.

 

Climate change is one of the most serious challenges we face today, agreed Professor Philip Stier of Atmospheric Physics at the University of Oxford. But climate science is inherently complex, involving massive uncertainties and multiple feedback effects from different changes in our climate and environment.

 

The webinar examined the many exciting capabilities of artificial intelligence (AI) to help climate science see where we really are with global warming and assessing realistic mitigation options.

Untangling the data

While vast Earth observation datasets are at heart of climate science, only a small fraction of such data is currently used to “train” climate models, Stier said. AI could help to draw more value from those data sources.

As one example, AI could extend the analysis of climate datasets over longer timeframes, taking paleo-climate data from past eons into account, added Valérie Masson-Delmotte, Co-Chair of the “Physical Science Basis” working group of the Intergovernmental Panel on Climate Change (IPCC).

“There is a massive amount of historical information that exists, but sometimes it's not readily available or not yet digitalized,” she said.

Furthermore, AI can improve our understanding of feedback effects, narrow down uncertainties, and detect key signals from vast masses of unanalysed data, or “background noise”. With better-quality data and specialized inventories, it could help to unpack more of the interactions involved in climate change and further illuminate how the climate crisis could evolve.

Exploring causality

For years, international attention centred mainly on the effects of carbon dioxide (CO2), although methane gas also drives warming. AI could also help to understand the key role of oceans as heat and CO2 absorbers.

The IPCC studies the causes and impact of climate change on a rigorously scientific basis, Masson-Delmotte said. Following the science, recent IPCC reports, produced with 230 scientists from 160 countries, suggest a global imperative to cut CO2 and other greenhouse gas emissions to net zero.

As policy makers make tough choices on behalf of their citizens, algorithms can heighten space-time resolution and clarify stochastic aspects in climate modelling. Machine learning, for instance, helps to explore causality and capture dynamics at lower spatial and temporal resolution, explained Yoshua Bengio, Director of the Quebec AI Institute, Mila.

But while machine learning can heighten our understanding of the crisis, countries must adopt accurate carbon pricing to reflect the true impact of all our activities, Bengio said. Effective carbon pricing, he noted, would include the cost of uncertainty.

Numerous researchers are eager to help with improving the sustainability of transportation, discovering new production materials, stepping up conservation, and helping people better understand the changes that are coming, Bengio said.

New technologies,” added Masson-Delmotte, “can help build on local knowledge through citizen science and participatory approaches.”

Calls for concerted action

 

The Paris Agreement, signed by 195 nations worldwide, set out a clear game plan to fight rising emissions, spur climate-resilient development and foster greener socio-economic development. Over 60 other countries have set their sights on carbon neutrality by 2050, while the largest market, China, has pledged to become carbon-neutral by 2060.

 

Multilateral action is vital, Ban reiterated. Houlin Zhao, Secretary-General of ITU, called for active partnerships to coordinate ICT innovation and address the global climate challenge.

 

This webinar was the first in a new series of AI for Good expert talks intended to feed into a white paper on the use of AI to accelerate climate science. The next talk in the series will address AI-accelerated Climate Modeling and takes place on Friday, 16 July.