Using AI and Machine Learning to Boost the Fight Against Climate Change

Advances in artificial intelligence (AI) and machine learning can increase the chances of reducing carbon emissions through carbon capture or geoengineering projects.

Ranveer Chandra, general manager of industry research and chief technology officer (CTO) of agribusiness at Microsoft, has spent his career researching climate and agritech solutions. Microsoft, in this regard, continues its own commitment to reducing carbon emissions. Chandra says AI can help “full-solution” geoengineering projects become more affordable, focused, and transparent.

Using AI, researchers can better estimate the locations and impacts of ocean or solar geoengineering to decide on the most appropriate and effective approach. AI can also replace expensive simulations or complement process-based models, especially with new and promising technologies such as Python-based Causal ML.

Facing an existential crisis

Chandra warns, however, that there are still huge challenges: “The first is the cost of these solutions, to be implemented at scale,” he explains. IT Professional. “Second, there is limited understanding of the total impact – primary and secondary – of any of these geoengineering solutions.”

Carbon capture and storage (CCS) is generally seen as more promising, but it’s “still a very expensive way to remove carbon,” he adds. “With AI, we are able to do large-scale seismic modeling at a speed of more than 1,500 times existing approaches that use partial differential equations and simulations.”

This means better modeling of carbon flows and better planning of CCS operations; Microsoft is working with various partners, including Nvidia, on developing AI and machine learning approaches that will propel CCS projects.

Current and planned projects could sequester a combined amount of approximately 40 megatons of CO2 per year. But to keep temperature rises to 1.5C above pre-industrial levels requires a hundred times the storage capacity, he warns. “This is an existential problem, so we need to help invent techniques to mitigate climate change,” Chandra offers.

Professor Ted Shepherd, Grantham Chair in Climate Science at the University of Reading, notes that simulated physics-based models are generally more reliable in determining cause and effect than purely model-based approaches. data.

“Data science, as generally understood, is good for finding effective solutions in situations with lots of data to explore, but climate intervention strategies are, by their very nature, out of sample – c i.e. not represented in the existing data,” says Berger. “Data science methods tend to break down when applied to out-of-sample problems.”

Innovations in causal AI – which identifies the underlying causes of a behavior or event that predictive modeling fails – can help reduce the risk of undesirable outcomes by getting a much better picture of cause-to-cause relationships. effect, says Shepherd.

Microsoft research projects drive advances in CCS

Northern Lights: Work with the Norwegian government, Equinor, Shell and Total to standardize and scale CCS, for a storage reservoir in the North Sea from 2024.

KarbonVision: Using a computer vision approach to map geological faults from seismic data, reducing the processing time needed to detect potential CO2 leak paths.

Q-FNO for 3D stream: Develop scalable and industry-relevant 3D simulations for CO2 fluxes and storage, involving typically complex and computationally high coupled PDEs.

Sequoia: Towards clusterless supercomputing on Azure, by building a more manageable distributed programming infrastructure on top of existing Azure HPC services.

Hyper-waves: Using a cloud-based, fault-tolerant framework for large-scale 3D seismic imaging, with large-scale Docker, Kubernetes, and Dask parallel containerized seismic workloads on Azure.

Innovating to get out of a climate crisis

Beyond CCS, more ambitious concepts include stratospheric aerosol injection (SAI) and marine cloud brightening (MCB). The idea is to beam sunlight back into space to reduce global warming, either by spraying reflective particles into the stratosphere or “seeding” clouds with crystallized salt out of the oceans, respectively.

Dr. Vitali Avagyan, data scientist at TurinTech, notes that AI could help predict CCS plant failures in real time, as well as compare different complex decarbonization strategies.

“The rapid growth of environmental data from sensors, weather and climate models makes it difficult to interpret them quickly,” says Avagyan. “AI can help measure the collective impacts of CCS on entire energy systems.”

Shepherd notes that SAI seems “very doable”, but no one knows exactly how it would play out. What if, for example, the establishment of an SAI in one country causes, for example, problems in another country? A failure of the South Asian monsoon would be a disaster, for example, suggesting governance issues.

Dr Timothy Farewell, Chief Scientist at Dye & Durham, emphasizes the need for robust assessment, filtering and cleaning, as well as a solid understanding of the interactions and processes involved.

“Some blind AI or machine learning models will seek to extrapolate beyond the range of training data to more extreme conditions, which will lead to serious accuracy issues,” confirms Farewell.

Jim Haywood, professor of atmospheric science at the University of Exeter, also tells IT Professional that greater knowledge of the physical sciences is still needed to manage the risks and opportunities of SAI and MCB in particular.

Shepherd agrees, adding: “We don’t really have faith in the regional aspects of climate change. You have a lot of factors; you need a very structured way of doing it, and that’s where the ‘AI comes into play.’

Moreover, applying even well-established laws of physics to an atmospheric simulation means breaking down an Earth system into grid boxes – but the equations needed should be continuous in space, rather than discrete. A typical grid box for a climate model might be 50 km2 horizontally and 1 km vertically, Shepherd says. “It’s quite rude. You don’t represent much,” he confirms. “There are uncertainties. That makes the model much, much more computationally expensive.”

Ideally, a “set” of executions, many processes at very high spatial resolution, is also needed to work out all possible realizations, says Shepherd. “And different scientists will argue differently, for different trade-offs, as well as some processes that aren’t fundamentally understood — such as mixed-phase clouds of ice and liquid.”

Featured Resources

Your key to digital differentiation and competence

DBaaS transformation: database services suitable for application modernization, cloud-native innovation and data-driven strategies

Free download

An analysis of the European cyber threat landscape

2022 human risk assessment

Free download

CIAM Buyer’s Guide

Finding the right CIAM solution to capture and retain customers, fuel business growth, and keep customers safe

Free download

The CIO Imperative: Leading the Digital Future

Reinvent how you differentiate yourself through technology

Free download

Sherry J. Basler