How Machine Learning Can Help Us Stop Climate Change
Successfully analyzing and addressing climate change involves wrangling an entire world of data. Now, a new paper — “Tackling Climate Change with Machine Learning” — from more than 20 machine learning experts across 16 organizations is shining a spotlight on the many critical roles that machine learning can play in fighting back against the climate crisis.
“Climate change is one of the greatest challenges facing humanity,” the paper reads, “and we, as machine learning experts, may wonder how we can help.” The researchers hail from a wide range of institutions – Harvard, MIT, Cornell, Stanford, Google, Microsoft and others are represented – with each of a dozen researchers contributing a distinct section to the paper.
This article highlights major areas where the researchers say that ML may have applications for mitigation – the prevention of climate change. A subsequent article will highlight the intersections of ML and adaptation to climate change.
Electricity Systems & ML
Electricity systems are responsible for about 25% of human-caused greenhouse gas emissions, explains Prita L. Donti of Carnegie Mellon University. She goes on to outline how ML can be used to aid in generation and demand forecasting – a grid that learns from behavior can offer better short-term demand forecasts, reducing reliance on polluting standby plants and allowing for better forecasts of renewable energy generation – which is often intermittent and difficult to predict – helping solar, wind and hydroelectric plants integrate into the grid.
Donti also highlights how ML could allow storage and smart devices to respond to electricity prices with agility; accelerate materials science for moonshots like solar fuels and nuclear fusion; and optimize existing renewable technology – for instance, by predicting how to rotate solar panels for optimal sunlight. Further, she posits that ML could help reduce emissions from existing fossil fuel operations by reducing methane leakage from natural gas pipelines.
Transportation & ML
Lynn H. Kaack – of ETH Zurich – delves into the role of ML in transportation, which constitutes another quarter of mankind’s greenhouse gas emissions. ML, she argues, could decrease transportation activity by using traffic forecasting to optimize public transit, road development, shipping routes and urban planning generally. Further, ML could help optimize engines for efficiency and is key to the development of autonomous vehicles – which are better at reducing their fuel consumption than human drivers. Kaack also highlights how the intersection of electric vehicles and the electricity grid will offer strong opportunities for ML applications.
Buildings, Cities, & ML
Haack returns – along with Nikola Nolojevic-Dupont of the Mercator Research Institute on Global Commons and Climate Change – to discuss ML’s role in the sustainability of buildings and cities. ML, they say, could be used to forecast building energy demand and optimize design and operation strategies, integrating more fully with the electric grid in a manner not dissimilar to electric cars. ML could, for instance, help heating and cooling systems dynamically adapt depending on whether or not a room was occupied. They also highlight how ML could help estimate building energy use using external data – such as footprint, material, roof type, etc. – allowing for better, more climate-friendly urban planning in areas with limited building data. Cities can also apply ML outside of buildings – for instance, by using ML to improve public lighting systems based on foot traffic.
Industry & ML
“The availability of large quantities of data, combined with affordable cloud-based storage and computing,” posits Anna Waldman-Brown of MIT, “indicates that industry may be an excellent place for ML to make a positive climate impact.” Waldman-Brown begins by discussing supply chains, suggesting that ML could be used to predict supply and demand, identify lower-carbon products and optimize shipping routes. Like others, she discusses material development as a major application for ML, highlighting the fact that 9% of global greenhouse gas emissions come from cement and steel production. She also stresses the overlap with building management, outlining how factories and warehouses could benefit greatly from ML-enhanced operation.
Farms, Forests, & ML
“Our current economy encourages practices that are freeing large amounts of this sequestered carbon through deforestation and unsustainable agriculture,” writes Alexandre Lacoste of Element AI. “The large scale of this problem allows for a similar scale of positive impact, and ML will play an important role in many of these solutions. Lacoste explains that satellite-informed ML can be used to estimate sequestered carbon, monitor the health of forests, and more. For agriculture – which constitutes 14% of greenhouse gas emissions – he highlights the growing role of ML-enabled precision agriculture, which could engage in predictive and dynamic weeding, crop yield prediction, fertilization and other tasks.
Carbon Dioxide Removal & ML
The final mitigation category – written by Andrew S. Ross (Harvard University) and Evan D. Sherwin (Carnegie Mellon University) – focuses on removing CO2 from the atmosphere. “Even if we could cut emissions to zero today,” they write, “we would still face significant climate consequences from greenhouse gases already in the atmosphere.” Ross and Sherwin discuss how ML-accelerated material development could lead to the availability of materials with a great ability to store and retain CO2 or identify existing storage locations using a variation of the ML-enabled technology currently used to identify oil and gas deposits.
The researchers hope their work will prove valuable to a number of audiences – from engineers and entrepreneurs to local and national governments. Furthermore, they stress the need for collaboration for these ML-oriented advances to become reality. “All of the problems we highlight in this paper,” they write, “require collaboration across fields.” They recommend working with experts in relevant domains, as well as developing accessible code on popular platforms and using popular programming languages.
The authors stress the immediacy of the problem, but also urge caution.
“We emphasize that machine learning is not a silver bullet,” reads the paper. “The applications we highlight are impactful, but no one solution will ‘fix’ climate change. There are also many areas of action where ML is inapplicable, and we omit these entirely. Furthermore, 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.”
To read the researchers’ full paper or see more information about their efforts to integrate ML into climate change mitigation and adaptation, visit https://www.climatechange.ai/.