Dr Geoffrey Hinton – often called the godfather of AI – has spoken openly of his concerns that AI may be a more urgent threat to humanity than climate change. But you don’t have to look far to find similarly well-informed voices claiming the existential challenge posed by climate change is unrivalled.
While debate over which constitutes humanity’s greatest threat seems unlikely to end any time soon, there appears to be little argument that the extraordinary power of AI could potentially be utilised to tackle the enormous challenges posed by climate change.
Experts at the University of Edinburgh have had their finger on the pulse of all things AI longer than most, establishing Europe’s first dedicated research group on the topic in 1963. Among its many illustrious alumni is none other than the aforementioned AI pioneer, Dr Hinton.
Edinburgh academics are still at the forefront of AI research today, and a growing number are now using it to better understand humanity’s impact on climate change and to map out ways of adapting to it.
“AI could have transformative effects on efforts to combat climate change,” says Dr Joe Kennedy, of the Global Academy of Agriculture and Food Systems. “We’re seeing already that it can unlock doors to wide-ranging interdisciplinary work, enabling researchers to explore important questions that might otherwise be unanswerable, or that could only be examined very crudely.”
Dramatic shifts in weather patterns and extreme events wrought by climate change pose stark threats to global food security, but the reverse is also true: the foods we eat – specifically, the environmental costs associated with them – are major contributors to climate change. Dr Kennedy’s work with AI is making it possible to reveal the environmental impacts of UK food consumption as a whole for the first time.
Previous research has shown the impacts of individual foods – for instance, it is well known that meat and dairy are the worst culprits – but carrying out in-depth analysis of the environmental footprint of an entire country’s dietary habits based on nationally representative surveys has simply not been possible. Working with Professor Lindsay Jaacks, Dr Kennedy is aiming to do just that.
The researchers are using AI to match two huge, complex sets of data. One is UK Government data on consumption patterns and the nutritional content of different foods, and the other is data from FoodDB, a database holding details of the environmental impact of tens of thousands food products sold in UK supermarkets. Combining these two data sets will reveal the climate impact of UK diets by tallying sales figures on the foods people buy with data on their environmental footprint, from growing or making foods to transporting them to supermarket shelves.
The process of matching these two data sets can be done manually by researchers, but it’s long, painstaking work. Part of what makes it so challenging is that the UK Government data is broad, while the data in FoodDB is very specific. For example, while beef lasagne is logged as one product in the UK Government data, details of more than 100 different beef lasagnes sold in supermarkets is recorded in FoodDB. That can make doing projects like this manually extremely time consuming and expensive – often prohibitively so.
Dr Kennedy and the team worked out that it would take a team of four researchers around nine months and cost close to £300,000. Using AI, the matching took two weeks and cost around £300. “Given the cost and time involved, this project simply couldn’t have been done without AI,” says Dr Kennedy.
As well as greatly speeding up the matching process, using AI makes the technique much more replicable. That means with the right data it could be used to calculate similar impacts for other parts of the world. Drilling down to that level of detail would provide policymakers with key evidence to inform decision-making about encouraging changes to diets and food production practices.
With an average of 50 new satellites going into orbit each week – and even more planned in coming years – enormous volumes of climate data are being relayed back to researchers on Earth every day. Doing meaningful analysis of that data on a large scale by hand, as it were, is nigh on impossible. That is where AI comes in.
“Terrabytes of climate data are produced on a monthly basis, and that is only going to increase,” says Dr Doug Finch of the School of GeoSciences. “AI is an extremely powerful tool that can be used to identify important patterns and trends in the ever-growing amounts of climate data.”
Dr Finch has created an AI model that scours huge swathes of near real-time satellite data to identify emission hotspots. It uses image recognition to pick out key features in the data to pinpoint emissions from places like power stations, cities and pipelines. Using AI means vast amounts of data can be processed rapidly.
The emission he is currently focusing on is methane – the second most abundant man-made greenhouse gas in the atmosphere. Every day, satellites send back dozens of data files on methane emissions covering the entire globe within hours of being recorded. Similar data on other greenhouse gases is also relayed back daily. The ultimate aim is to use AI technology to create a worldwide database of near real-time emission data on all greenhouse gases.
“As AI gets more powerful and research in this area matures, it will become possible to quickly pass up-to-date climate information on to policymakers for immediate action,” says Dr Finch. In this way, he believes AI could help to fight climate change on a worldwide scale.
The headline aim of this year’s United Nations climate change conference, COP28, is to scale up actions that support the Paris Agreement goal of limiting global warming to 1.5 degrees Celsius. It is a monumental ambition but one that we must strive for, says Professor Gabi Hegerl: “I hope we can limit global warming to as near 1.5 degrees as possible because every increment above that makes future extremes worse.”
And she should know. A renowned, award-winning scientist, Professor Hegerl is the University’s Personal Chair in Climate System Science, and a regular contributor to influential reports produced by the Intergovernmental Panel on Climate Change (IPCC). Indeed, Professor Hegerl contributed to IPCC work that shared the 2007 Nobel Peace Prize with Al Gore.
Professor Hegerl explains that if we can better understand how these extremes influence land, oceans, ecosystems and society, we can adapt management methods to them. But there is also only so much adapting we can do, so finding out what those limits are is crucial. She can see clear roles that AI might play in those efforts to both mitigate and adapt to climate change.
“Combining climate modelling and AI will enable us to better understand how extreme events will affect us in the future, how we can better adapt to them and what the limits of adaptation are,” she adds. “AI is a promising technique to bridge the different scales at which factors like the atmosphere, oceans, ice and land interact with each other and how these are built into climate models.”
Climate, weather, environment and society strongly interact and influence each other, but because of the myriad complex ways that they all intersect, pinpointing clear links between them is tricky. Professor Hegerl says AI can address this by uncovering links within the vast sources of climate data now available, helping to explain to what extent greenhouse gas increases lead to worse extremes and cause impacts such as ice loss, crop failures or wildfires.
“We often rely on physical mechanisms and evidence of strong correlations – one factor changing relative to another – but this is not the same as causation – one factor causing another to change,” she explains. “Used responsibly, AI can help disentangle which factors in an extreme event like a heatwave are the active drivers and which consequences are linked to it, rather than just happen at the same time.”
So-called causal networks are AI tools that can be used to better understand the interplay between complex factors. Professor Hegerl’s research group is looking into causal links between fine particles called aerosols – fumes and smoke from human activities are major sources of these – and climate change, and also how to detect extreme events like large-scale heatwaves using satellite data.
“Climate science is enthusiastically embracing AI,” says Professor Hegerl. “Used effectively, it could help answer some of the most difficult – and important – questions in climate science.”
Finding answers to those and other key questions – and finding them quickly – will be critical to the success of efforts to combat and adapt to a world that is continuing to warm rapidly. While AI is by no means a panacea when it comes to climate change – its own environmental footprint is growing increasingly large – the potential solutions it offers are difficult to ignore.
As Dr Kennedy explains, in the fight against climate change, speed is of the essence and we need all the help we can get: “AI will enable us to answer questions in a much faster timeframe, and that is essential because the time window we have to mitigate and adapt to climate change is rapidly closing.”
Picture credits: digital concept image – wayra via Getty; dairy farm – PeopleImages/Getty; Sentinel-5P satellite – ESA/ATG medialab; COP28 delegates – Sean Gallup/Getty; Wildfires – David Gray/Stringer via Getty