Friday 20 September saw over 4 million people participate in climate change strikes worldwide, with 300,000 people in the UK taking to the streets to show their anger around the lack of action from global leaders. Another strike is due to take place on 27 September, bookending the newly dubbed Global Week for Future. Technology
Friday 20 September saw over 4 million people participate in climate change strikes worldwide, with 300,000 people in the UK taking to the streets to show their anger around the lack of action from global leaders. Another strike is due to take place on 27 September, bookending the newly dubbed Global Week for Future.
Technology and climate change have a difficult relationship. As technology has become more pervasive throughout our homes, workplaces and society, the amount of data we create has spiralled, causing data centres across the globe to now have the same size carbon footprint as the aviation industry.
Although some might argue that the only way to decarbonise is to decomputerise, like most of these things, the situation is not quite that black and white. While technology is undoubtedly making the climate crisis worse in on some fronts, artificial intelligence (AI) and machine learning (ML) technologies are now being used across the world to help mitigate some of the impacts that are resulting from this increase in global temperature.
In June 2019, a research paper was published titled Tackling Climate Change with Machine Learning. Co-authored by 22 leading authorities on climate change and artificial intelligence, it outlines the role AI and ML technologies have to play in fight against climate change. The suggested uses within the paper are widespread and varied. However, it ultimately outlines 13 key areas where artificial intelligence would have the best chance of mitigating some of the worst effects of climate change, were it to be deployed.
The 13 fields include transportation, electricity systems, farms, forests, and education, and have been categorised by how long a focus on the area would take to have an impact, to how developed the technology relating to the fields is. Individual solutions within these subsections are labelled either ‘high leverage’, ‘long term’, or ‘high risk’.
Many of the recommendations made within the paper do actually exist already, it’s just that no one is currently conducting them on a scale that would have an impact.
While the list contains some obvious use cases for AI and ML, there are a number of “high leverage” recommendations included in the report which use the technology in more unconventional ways. One proposal in the report focuses on the ways in which we can use machine learning to develop materials that store, harvest, and use energy more efficiently; with the technology being deployed to enable scientists to find, design and evaluate new chemical structures to see if they can be useful.
Another use case suggested by the paper is the deployment of ML to help reduce the life-cycle of fossil fuel emissions, by helping to prevent the leakage of methane from natural gas pipelines and compressor stations. The technology can also be used to reduce emissions from the freight transportation of solid fuels or in the sequestration of CO2.
AI and ML can also be used to create maps from aerial imagery and information retrieved from social media data. The report notes that “accurate and well-annotated maps can inform evacuation planning, retrofitting campaigns, and delivery of relief” and assist in damage assessment efforts by comparing scenes immediately pre- and post-disaster.
The prediction of extreme weather events is another area where the authors of the report believe machine learning could prove invaluable in the mitigation of the consequences of climate change and one where there is already a lot of work being done.
Last year, the University of Cambridge set up The UKRI Centre for Doctoral Training in the Application of Artificial Intelligence to the study of Environmental Risks (AI4ER), a multi-disciplinary research centre that was founded to train researchers “uniquely equipped to develop and apply leading edge computational approaches to address critical global environmental challenges”.
Robert Rouse is a PhD student at Cambridge who is part of the AI4ER. His research focuses on the application of Bayesian machine learning methods to help predict the future risk of urban flooding, working in conjunction with a world-leading environmental engineering company, Mott McDonald, on a range of structural projects and new algorithms which can be implemented into data analysis pipelines.
“The predictions that we have for climate change are predominantly derived from GCM, which stands for General Circulation Model but which most people take to mean Global Climate Model,” Rouse explains to Techworld.
“They output climatic variables at a very coarse resolution, spatially, and with a slight roughness temporarily. With some things, when you’re trying to understand what’s actually going to happen at a local scale, certain questions are always going to be raised and people, at the moment, are quite rightly worried about climate change. They don’t best know how to prepare for that or they don’t know what might happen at a local scale due to the effects of climate change.
“What we’re [in the AI4ER] trying to do is to take the output of those climate models and work with them to create models for impact. We’re all looking at slightly different things, I’m obviously looking at flooding, but other people are looking at heatwaves and mortality, things like that. Basically, we’re taking those climate inputs and producing impact outputs, I think that’s probably the best way of describing it.”
Just this week, the Intergovernmental Panel on Climate Change (IPPC) released a report claiming extreme sea level events will hit once a year by 2050, whether extreme global heating is curtailed or not. Across the globe, almost 2 billion people live in megacities along coastal lines, meaning the future for low-lying coastal communities is extremely bleak.
Rouse is only about halfway through his PhD and says he still has a lot of research to do, however, he’s hopeful that his work and that of his peers will ultimately have a long-term impact.
“The idea is to build up a model that’s transferable,” Rouse says. “Then, we can take it to areas where they may not have the right tools, particularly LEDCs, and we give them this information so that they can understand the risks they face and hopefully mitigate them by putting systems in place.
“The research could also help to inform policy. When you’re talking to someone about the amount of carbon dioxide in the atmosphere, that’s not a tangible. But, if you were to tell someone what the financial damage might be from flooding or from extreme heat, then it becomes a little more real and it may help to inform policy. It may help to bring about political or cultural change.”
His research is being funded by The Royal Commission for the Exhibition of 1851, who annually award research scholarships to “to approximately eight young scientists or engineers of exceptional promise”. For Rouse, one of the important things about being involved with the Royal Commission is the extent to which they can help commercialise his work and “[get] the research to a point where it’s not just going to be a paper on a shelf, but it will be something will be used and actually have, and hopefully have an impact on industry.”
Rouse says that there are around 5 petabytes – or 5,000 terabytes – of climate data available for use in artificial intelligence and machine learning models.
“It’s a burgeoning area of research and there’s lots of people now working with climate data and using the output of these climate models to try and understand what the risk is,” he says.
On a personal level, Rouse is less worried about the general public being complacent about climate change than he is about “certain foreign leaders” having a negative effect.
“I don’t think it’s heathy for individuals to dismiss the work that scientists do. There’s a whole majority of climate scientists who back research proving the fact that climate change is real and is having an effect,” Rouse says.
“There are the discrepancies between the rate at which it’s having an effect and how best to mitigate it but, the fact that we still have people who, despite all the advice they receive, are in denial. And that’s not helpful, particularly when they are in such a high-profile position. I don’t think we should be disregarding the advice.”
However, despite the seemingly optimistic tone of the report, the authors are keen to note that while artificial intelligence will certainly prove “invaluable” in the fight against climate change, it is by no means a “silver bullet”, noting that political action is needed in tandem with technology.
Led by David Rolnick, a postdoctoral fellow at the University of Pennsylvania, the paper states: “technology alone is not enough – technologies that would reduce climate change have been available for years, but have largely not been adopted at scale by society. While we hope that ML will be useful in reducing the costs associated with climate action, humanity also must decide to act.”