Humans are struggling to address big sustainability problems. Maybe machines could do a better job.

Every year, human activity sends around 350 million tons of methane into the atmosphere. Methane traps heat, driving roughly one-third of the global temperature increase since the industrial revolution. And while carbon dioxide emissions are an inevitable byproduct of fossil fuel combustion, much of that methane is the result of bad management. In addition to distributed sources, such as agriculture, a large amount of the powerful greenhouse gas is vented from oil wells, leaks from pipelines or seeps from landfill sites. Stopping those emissions would help to slow global warming, but tracing a fugitive plume back to its source is a complex and labor-intensive activity. It can take multiple data sources and many rounds of analysis to pinpoint the root cause of a methane problem.

Eye in the sky

There might be a quicker way. Peter Joyce, PhD, and his colleagues from the UK National Centre for Earth Observation have shown that artificial intelligence (AI) technology can identify and quantify methane emissions from satellite images. They used weather modeling software to create simulated images of methane plumes of different sizes, then trained an AI system to recognize those images. Once the system knew what it was looking for, they applied the approach to real-world data. The approach proved to be better at finding and evaluating methane plumes than conventional methods. It was much faster too, requiring only around a minute to spot 21 plumes in satellite images covering 900 square kilometers.

Methane spotting is just one of hundreds of challenges we face as we seek to better understand – and mitigate – our impact on the planet. From measuring changes in the extent of sea ice or ocean garbage, to assessing the health of forests, or predicting variations in local air quality, environmental problems are characterized by large volumes of complex data and significant uncertainty. Crunching that data is laborious and difficult for humans, but might be quick and easy for a well-designed AI system.

Beyond helping us to understand the world, artificial intelligence could also help us do a better job of running it. It is not only able to monitor and predict developments, but also to make recommendations and decisions. And the world urgently needs better decisions. In 2015, the United Nations agreed its 2030 Agenda for Sustainable Development, a list of 17 goals and 169 targets designed to improve health and education, reduce inequality, and spur economic growth while tackling climate change and working to preserve oceans and forests. At a 2023 summit marking the midpoint of the agenda’s planned implementation period, UN Secretary General António Guterres warned that the world is on track to meet only 15 percent of those targets, with “many going into reverse.”

In a widely cited 2021 article in the journal Nature Communications, a team at KTH Royal Institute of Technology in Stockholm, Sweden, evaluated the likely impact of AI on the Sustainable Development Targets. They found that AI had the potential to act as an enabler in efforts to meet 134 of those targets, almost four-fifths of them.
“We identified many areas where AI could allow you to do things that just weren’t possible before,” says Associate Professor Ricardo Vinuesa, one of the study’s lead authors. “For example, you can use AI to analyze satellite data and identify regions where crops are failing and the risk of extreme poverty is increasing, then try to coordinate actions to help those areas.” Other areas with significant potential include the use of AI tools to optimize the complex interactions between supply and demand in energy networks dominated by renewable electricity, or the management of transport flows in digitally integrated smart cities.

Waste not

That’s the theory. What’s happening in practice? While AI chatbots are making headlines around the world, other systems are getting on with less glamourous jobs. Several countries like Canada, the USA, Germany, the UK, Japan or South Korea have begun to adopt AI technologies to optimize waste management, for example. AI can do a notable job in sorting and separating waste more consistently. Furthermore, smart waste collection systems can intelligently plan routes and schedules for refuse collection: Fill level sensors in the bins monitor the amount of waste in the bins and show whether a collection is necessary. This reduces emissions and traffic congestion caused by collection vehicles. In Punggol Digital District, a new business and research park in Singapore, one of the world’s first large-scale smart energy grids is under construction. Buildings in the district will be able to communicate with the system and intelligently adapt their power consumption in response to changing conditions.

Finish telecommunication equipment maker Nokia, meanwhile, has developed an AI tool to improve the energy efficiency of mobile data networks. The system analyzes and forecasts demand on radio towers and computer systems, automatically shutting down equipment at times when data traffic is low. The company claims that its approach can reduce energy consumption and CO₂ emissions in mobile networks by up to thirty percent. That’s two to five times better than the savings operators achieve when they use manual methods to manage energy consumption.

The chemical industry is pursuing AI-powered efficiency too. BASF has been gradually digitizing its Verbund integrated production sites for decades, for example. The construction of the company’s new facility in Zhanjiang, China, is giving it the chance to build a site that is designed to be digital from the start. “AI contributes to greater resource and energy efficiency and thus to sustainable operation,” says Dr. Andreas Wernsdörfer, head of Digitalization of Production & Technology at BASF. “This will help to make Zhanjiang the chemical Verbund site with the lowest carbon footprint in the world.” To achieve this goal, BASF collects large amounts of real-time site data to create transparency across the entire Verbund. “In the future our digital tools will ensure full traceability of when and where a renewable material was deployed throughout the value chain. AI algorithms will then be used to simulate which measures the site operations teams can take to contribute to higher carbon reduction, like, for example, exact process and control settings in the production facilities,” says Wernsdörfer.

Down in the weeds

Elsewhere, farmers are using data and AI to improve yields while reducing carbon emissions and saving supplies. ONE SMART SPRAY by BASF Digital Farming and Bosch, for example, uses image recognition technology to automatically recognize weeds growing in row crops. The weed management system works in milliseconds, targeting and delivering herbicide precisely where it is needed. This allows maximum herbicide savings.

Agricultural AI is helping farmers see the bigger picture too. Heinrich Esser is from the sixth generation of his family running a farm in the small town of Vettweiß-Kelz in the West of Germany. The land produces cereals, potatoes, and specialty crops such as asparagus and strawberries.

“It is very important to us to cultivate the farm and our fields in a way that future generations can also make a living from it: We live not only from nature, but also with nature,“ says Esser. That means pursuing strategies that optimize yields while controlling the use of water, fertilizer and crop protection products. At 150 kg of CO₂ equivalent per metric ton of wheat, Esser’s crop has around half the carbon footprint of the average German farm.

But he wants to do even better. In 2022, his farm became a test site for the new “ClimatePartner Agriculture” project of BASF and Raiffeisen Waren-Zentrale Rhein-Main, one of the biggest agricultural wholesalers in Germany. “Over the next years, trial plots on our farm will test different cultivation strategies aimed at cutting carbon emissions in wheat production by 30 percent,” he explains.

The approach has digital and AI technologies at its heart. BASF’s agronomic decision support platform xarvio^® FIELD MANAGER, which is available as an app, provides timely and precise recommendations to optimize crop production. This includes what to do with the crop, how much to do, and when to do it, providing advice for sowing, fertilization, and crop protection use. The “Climate Partner Agriculture” project is now being extended to other farms.

All good?

Speaking from his office in Stockholm, Ricardo Vinuesa wants to make it clear that the AI revolution brings sustainability risks as well as benefits. “We found that only 35 percent of the UN’s sustainable development targets might be negatively affected by AI,” he says. “But even one target negatively affected is something to worry about, because the cascading consequences can be unpredictable.”

Many of the sustainability risks stem from the choices people make in their application of AI, he explains. AI optimization of industrial production could be designed to reduce costs while ignoring the resulting pollution, for example. Inequality of access could also create challenges, he adds. AI technologies might exacerbate the gap between rich and poor, as the benefits of new tools accrue to the wealthier regions, industries and companies that can afford to develop and deploy them.

There’s also a risk that people put too much faith in the ability of these technologies to solve tough sustainability problems. When analysts at consultancy PWC assessed the use of AI in environmental applications, they found opportunities to boost global GDP by 3.1 to 4.4 percent while reducing carbon emissions by 1.5 to 4 percent. That’s nowhere enough to take the world to its net zero target.

Even leading AIs are cautious about their own potential. “AI can be a powerful tool to aid human efforts in environmental conservation and restoration, but it must be part of a broader strategy that includes policy change, education, and international cooperation,” says large language model ChatGPT. Vinuesa agrees. He is convinced: “A better understanding of the capabilities and limitations of AI – by industry, governments and the wider public – will be key to exploit its potential.”