Climate Change AI - Interactive Summaries

Electricity Systems

Forecasting supply and demand

High Leverage

The supply and demand of power must both be forecast ahead of time to inform electricity planning and scheduling. ML can help make these forecasts more accurate, improve temporal and spatial resolution, and quantify uncertainty.

Improving scheduling and flexible demand

Scheduling algorithms on the power grid have trouble handling large quantities of solar, wind, and other time-varying electricity sources. ML can help improve electricity scheduling algorithms, control storage and flexible demand, and design real-time electricity prices that reduce CO2 emissions.

Accelerating materials science

High Leverage Long-Term

Designing new materials is important for many applications, including energy storage via fuels and batteries. ML can help suggest promising materials to try, thereby speeding up the materials discovery process.

Additional applications (for variable sources)

There are many additional ways to facilitate the adoption of solar panels, wind turbines, and other low-carbon electricity generators. ML can help efficiently operate these systems, detect the locations of solar panels, and monitor the electric grid in places with rooftop solar panels.

Managing existing technologies

Many controllable low-carbon technologies, such as geothermal, nuclear fission, and (in some cases) dam-based hydropower, are already commercially available. ML can help plan where these technologies should be deployed and maintain power plants that are already operating.

Accelerating fusion science

High Leverage Long-Term

Nuclear fusion has the potential to produce safe, carbon-free electricity, but such reactors continue to consume more energy than they produce. ML can help suggest parameters for physical experiments and also model the behavior of plasma inside reactors.

Reducing life-cycle fossil fuel emissions

High Leverage Uncertain Impact

In addition to the unavoidable climate impacts of burning fossil fuels, natural gas pipelines leak methane, a powerful greenhouse gas. ML can help detect and prevent these leaks until society stops burning natural gas.

Reducing system waste

As electricity gets transported from generators to consumers, some of it gets lost as resistive heat on electricity lines. ML can help prevent avoidable losses through predictive maintenance.

Modeling emissions

Reducing the emissions associated with electricity use requires understanding what the emissions on the electric grid actually are at any given moment. ML can help estimate and forecast emissions, and potentially model the uncertainty in these estimates.

Improving clean energy access

Improving access to clean electricity can address climate change while benefiting society, resulting in decreased emissions from wood-fired stoves, diesel generators, and other sources. ML can help manage rural microgrids and gather data to aid energy access policy.

Approaching low-data settings

High Leverage

In many places, data on electricity systems are not widely collected or shared; but addressing these systems is essential for both equity and impact. ML can help gather data in some cases, and ML techniques designed with low-data settings in mind can have outsized impact.

Transportation

Understanding transportation data

Many areas of transportation lack data, and decision-makers often plan infrastructure and policy based on uncertain information. ML can help provide relevant information about today’s transportation systems by interpreting data from sensors and satellite imagery.

Modeling demand

High Leverage

Infrastructure design shapes what kinds of transportation options people choose -- both for freight and personal transit. ML can help inform infrastructure decisions by modeling current transportation usage and forecasting future demand.

Shared mobility

Uncertain Impact

It is unclear whether shared mobility (such as ride-sharing services) will increase or decrease emissions. ML can help model the emissions impacts of shared mobility.

Freight routing and consolidation

High Leverage

Intelligently routing and consolidating freight shipments can dramatically reduce emissions. ML can help optimize freight transportation by gathering information about suppliers, freight demand, and transit delays, and then by using this information to create optimal transportation schedules.

Alternatives to transport

Uncertain Impact

Technologies such as additive manufacturing or virtual communication may help replace or reduce transportation demand. ML can help develop and improve both of these technologies.

Designing for efficiency

There are many ways to reduce the amount of energy that a vehicle uses. ML can help design and operate engines to improve efficiency, provide proxies for complex aerodynamics models used in vehicle design, and improve processes for manufacturing lighter vehicle parts.

Autonomous vehicles

Uncertain Impact

Autonomous vehicles may increase the efficiency of transportation, though they may also increase driving (and therefore emissions). ML can help develop drones to more efficiently deliver freight, synchronize autonomous fleets to drive efficiently, and improve other applications specifically aimed at reducing energy consumption.

Electric vehicles

High Leverage

Electric vehicles (EVs) can play a key role in decarbonizing transport. ML can help improve EVs by modeling and scheduling charging, integrating EVs into the electric grid, and developing and optimizing EV batteries.

Alternative fuels

Long-Term

Aviation, long-distance road transportation, and ocean shipping require fuels with high energy density and are thus not conducive to electrification. ML can help accelerate the discovery of low-carbon fuels.

Passenger preferences

Understanding passengers’ travel preferences can inform long-term transportation planning. ML can help predict or infer passengers’ travel choices, as well as design better travel surveys.

Enabling low-carbon options

High Leverage

Good low-carbon transportation should be reliable, respond to user demand, and be integrated with other transit options. ML can help by predicting arrival times and demand, optimizing the locations of shared bikes and electric scooters, and predicting maintenance needs for rail.

Buildings & Cities

Modeling building energy

Making buildings more energy-efficient will require understanding the increasing amounts of data produced by meters and home energy monitors. ML can help model energy demand within buildings, quantify the energy used by individual appliances, and evaluate the effectiveness of energy efficiency measures.

Smart buildings

High Leverage

Buildings can decrease their carbon footprint using intelligent appliances and control systems. ML can help control heating and cooling systems more efficiently, adapt lighting and temperature to occupancy patterns, and change energy use in response to carbon-sensitive energy prices.

Modeling energy use across buildings

Energy use data is not available for most buildings, limiting the information available for city planners deciding which buildings to retrofit. ML can help predict energy use for buildings without data, based on other buildings within the city or district.

Gathering infrastructure data

High Leverage

Many places have almost no energy consumption data, which can make it difficult to design targeted mitigation strategies. ML can help infer attributes of buildings, such as volume or construction material, that are crucial to predicting energy use and efficiency potential.

Data for smart cities

High Leverage

Cities have increasingly large and diverse datasets at their disposal to improve efficiency and decrease emissions. ML can help analyze and infer information from these datasets via text-mining and other methods, preprocess large amounts of raw data (such as real-time traffic data), and integrate heterogeneous data sources.

Low-emissions infrastructure

Urban planning and infrastructure can be designed to make cities more sustainable and foster low-carbon lifestyles. ML can help model and coordinate the interaction between different infrastructure systems (such as buildings and transportation), and also assess how well solutions may transfer between different cities and regions.

Industry

Reducing overproduction

Uncertain Impact

The production, shipment, and climate-controlled warehousing of excess goods are major contributors to industrial greenhouse gas emissions. ML can help reduce overproduction by improving demand forecasting.

Recommender systems

Consumers and purchasing firms in search of climate-friendly options may have difficulty interpreting information about products’ life-cycle greenhouse gas emissions. ML can help identify the cleanest options, given relevant data.

Reducing food waste

High Leverage

Globally, society loses or wastes 1.3 billion metric tons of food each year. ML can help reduce food waste by optimizing delivery routes, forecasting demand, improving refrigeration systems, and identifying spoiled produce.

Climate-friendly construction

High Leverage Long-Term

The production of cement and steel accounts for over 10% of all global GHG emissions. ML can help reduce the need for these materials by developing techniques that require less raw material and accelerating the discovery of carbon-friendly replacements.

Climate-friendly chemicals

High Leverage Long-Term

Ammonia production for fertilizer use relies upon natural gas to heat up and catalyze the reaction, and accounts for around 2% of global energy consumption. ML can help facilitate cleaner ammonia production by accelerating the discovery of new electrocatalysts or proton conductors for use in the production process.

Adaptive control

High Leverage

Industrial HVAC systems and other control mechanisms can be extremely energy- and emissions-intensive. ML can help reduce GHG emissions from such systems via adaptive control.

Predictive maintenance

Effective maintenance of industrial equipment can increase energy efficiency and also cut emissions directly by reducing leaks of methane and other greenhouse gases. ML can help in the creation of “digital twin” models that simulate wear and tear, as well as possible interventions, before a system fails in real life.

Using cleaner electricity

High Leverage

Industrial processes such as cement crushing and powder-coating can be extremely energy- and emissions-intensive. ML can help optimize the timing of these processes to use cleaner electricity (e.g. wind and solar) when it is available.

Farms & Forests

Remote sensing of emissions

High Leverage

Greenhouse gases interact with sunlight, by definition, which means they can be picked up by hyperspectral cameras. ML can help track greenhouse gases from satellite or aerial imagery.

Precision agriculture

High Leverage Uncertain Impact

Typical industrial agriculture releases CO2 into the atmosphere by disrupting natural soil chemistry and biodiversity, and also requires chemicals that are emissions-intensive both to produce and to use. ML can help monitor emissions; reduce the need for chemicals by pinpointing pests, diseases, and weeds; and change agricultural paradigms by controlling physical robots.

Monitoring peatlands

High Leverage

Peatlands (a type of wetland ecosystem) cover only 3% of the Earth’s land area, yet hold twice the total carbon in all the world’s forests. ML can help monitor peatlands, guarding them against drainage or fire, both of which release carbon.

Estimating carbon stock

High Leverage

Modeling (and pricing) carbon stored in forests requires us to assess how much is being sequestered or released across the planet. ML can help identify tree species and heights from satellite and aerial imagery.

Automating afforestation

Long-Term Uncertain Impact

Planting trees in well-chosen locations can help sequester carbon while (if done with care) improving local ecology. ML can help locate appropriate planting sites, monitor plant health, assess weeds, analyze trends, and even control tree-planting drones.

Managing forest fires

Large forest fires release CO2 into the atmosphere, besides potentially harming people and property, while small fires can be beneficial by preventing large ones. ML can help firefighters better understand and control forest fires by forecasting droughts, estimating moisture in the forest canopy, and predicting the progression of fires.

Reducing deforestation

High Leverage

Forests store a large amount of CO2 in biomass and soil, but much of this CO2 is released when forests are destroyed. ML can help track illegal deforestation using imagery or even audio signals, and also provide tools for sustainable harvesting and forest management.

CO2 Removal

Direct air capture

Long-Term

To some extent, CO2 can be removed from the air using direct air capture (DAC) technologies. ML can help speed up the search for corrosion-resistant materials and sorbents (CO2 “sponges”) that allow DAC technologies to capture CO2 more efficiently and inexpensively.

Sequestering CO2

High Leverage Long-Term Uncertain Impact

Once CO2 is captured from power plant exhaust, industrial processes, or ambient air, it will ultimately be released back into the atmosphere unless it is permanently sequestered. ML can help identify promising sequestration locations as well as monitor these locations to prevent leaks.

Climate Prediction

Data for climate models

There are many opportunities to integrate data-driven insights into climate models. ML can help calibrate data-gathering sensors and use satellite imagery to generate relevant data (such as on crop cover or pollutant sources).

Clouds and aerosols

High Leverage

Clouds are the largest source of uncertainty in current climate models. ML can help provide computationally inexpensive representations of cloud physics for inclusion in climate models.

Ice sheets and sea level rise

High Leverage

Ice sheet dynamics and sea level rise are important but uncertain components of climate models. ML can help infer critical information such as sea ice reflectivity and ocean heat mixing from satellite data.

Working with climate models

Climate models can be extremely complex, and climate predictions are often made using the outputs of 20+ climate models. ML can help identify and leverage relationships between variables within climate models, and intelligently combine the outputs of multiple climate models.

Storm tracking

Identifying extreme events in climate model outputs can inform scientific understanding of where and when these events may occur. ML can help classify, detect, and track climate-related extreme events such as hurricanes in climate model outputs.

Local forecasts

High Leverage

Storms, droughts, fires, floods, and other extreme events are expected to become stronger and more frequent as climate change progresses. ML can help forecast extreme events and infer climate change risks at a local level, enabling more informed decisions about infrastructure, asset valuation and disaster response planning.

Societal Impacts

Monitoring ecosystems

High Leverage

Climate change poses catastrophic threats to global ecosystems, endangering the food supply and other natural resources on which humanity depends. ML can help model ecosystem effects and track changes in real time via remote sensing.

Monitoring biodiversity

High Leverage

Responding to ecological impacts requires understanding the changing abundance and distribution of individual species. ML can help work with data created by networks of citizen scientists and can also monitor species directly via camera trap data and aerial imagery.

Designing infrastructure

Long-Term

Infrastructure such as roads, power grids, and water mains must be designed to account for the increasing frequency and severity of extreme weather events. ML can help pinpoint vulnerable areas, provide localized predictions, and incorporate historical or proxy data to identify what infrastructure is needed.

Maintaining infrastructure

High Leverage

Under the increased threat of extreme events, society must efficiently allocate limited resources to effectively maintain infrastructure. ML can help predict where failures are most likely to occur and pinpoint failures if they do occur.

Food security

High Leverage

Climate change is expected to decrease crop yields in many locations, due to drought and desertification. ML can help monitor the real-time risk of food shortages, forecast longer-term crop yields, and improve the resilience of food supply chains to help society better manage food security threats.

Resilient livelihoods

Farmers and other individuals whose livelihoods depend directly on ecological systems are at especially high risk from climate change. ML can help increase resilience by facilitating equipment and information sharing, better target social interventions by monitoring economic health, and improve online employment matching and skill training systems.

Supporting displaced people

Long-Term Uncertain Impact

Climate change will likely have large-scale economic and socio-political consequences, which can lead to mass human migration. ML can help provide support to migrants by monitoring and predicting large-scale migration patterns, though care must be taken to ensure such monitoring is not used to exploit migrants.

Assessing health risks

Climate change is expected to exacerbate existing health hazards, by increasing the frequency and severity of heatwaves, expanding the range of vector borne diseases, and reducing air quality. ML can monitor exposures and drive strategies to protect vulnerable populations.

Managing epidemics

Climate change may increase the likelihood of disease epidemics. ML can help model the spread of disease and provide scalable tools for diagnosis to help society better respond to epidemics.

Disaster response

High Leverage

Storms, droughts, fires, floods, and other disasters will continue to become stronger and more frequent as climate change progresses. ML can help inform evacuation plans, infrastructure retrofits, and relief efforts by improving maps and retrieving information from social media.

Solar Geoengineering

Understanding and improving aerosols: design

Long-Term

Many solar geoengineering proposals rely on injecting aerosol particles into the atmosphere to partially reflect sunlight. ML can (speculatively) accelerate the search for new aerosols that are chemically nonreactive but still reflective, cheap, and easy to keep aloft.

Understanding and improving aerosols: modeling

Many solar geoengineering proposals rely on injecting aerosol particles into the atmosphere to partially reflect sunlight, but their physics is not fully understood. ML can help speed up physical models and quantify the uncertainty of predictions.

Engineering a planetary control system

High Leverage Long-Term Uncertain Impact

Controlling a geoengineering system comes with a multitude of challenges and a host of possible side effects, many of which could be catastrophic. Speculatively, ML can help fine-tune geoengineering interventions by suggesting control actions and emulating the complex dynamical systems involved.

Modeling impacts

Long-Term

It remains unclear what consequences will result from geoengineering proposals such as injecting aerosols into the stratosphere. ML can help model the impact of aerosols on human health, the effect of diminished light on agriculture, and other potential consequences of solar geoengineering.

Individual Action

Understanding personal carbon footprint

Individuals and households constantly make decisions that affect their carbon footprint, and many wish to reduce their impact. ML can help quantify the climate impact of consumer products and actions, estimate the benefits resulting from personal behavior change, provide appliance-level residential energy use data, identify households with high potential for efficiency gain, and optimize appliances to operate when low-carbon electricity is available.

Facilitating behavior change

High Leverage

Many individuals are eager to contribute to climate change solutions, and engaging them can be highly impactful. ML can help effectively inform people and provide them constructive opportunities by modeling consumer behavior and simplifying information on climate-relevant laws and policies.

Collective Decisions

Modeling social interactions

When designing climate change strategies, it is critical to understand how organizations and individuals act and interact in response to different incentives and constraints. ML can help model social interactions by integrating data-driven insights into agent-based models and by providing new tools such as multi-agent reinforcement learning.

Gathering data

High Leverage

When creating policies, decision-makers must often negotiate fundamental uncertainties in the underlying data. ML can help alleviate some of this uncertainty by extracting information from satellite imagery, sensors, social media posts, policy documents, and other sources (as detailed elsewhere in the paper).

Assessing policy options

Decision-makers often construct mathematical models to help them assess or trade off between different policy alternatives. ML can help provide new techniques for working with integrated assessment models, multi-objective optimization, and other models commonly used by decision-makers.

Evaluating policy effects

High Leverage

When creating new policies, decision-makers may wish to understand previous policies and analyze how these policies performed. ML can help on both fronts by analyzing the text of existing policies and by performing causal inference on historical data.

Predicting carbon prices

Carbon pricing and other market-based measures can incentivize the reduction of greenhouse gas emissions. ML can help predict prices in carbon markets and analyze the main drivers of these prices.

Non-carbon markets

Market design can influence greenhouse gas emissions even in settings where such emissions are not directly penalized, for instance by enabling more renewables in electricity markets or encouraging more efficient shipping of goods. ML can help analyze market behavior, set dynamic prices, and solve auctions with the goal of encouraging climate-beneficial behavior in the markets where it is applied.

Assessing market effects

When designing market-based strategies, it is necessary to understand how effectively each strategy will reduce emissions, as well as how the underlying socio-technical system may be affected. ML can help assess the outcomes of market-based strategies to ensure they are effective and equitable.

Education

AI in Education

Long-Term

In addition to being universally beneficial, education can improve the resilience of communities to climate change, especially in developing countries. ML can help enable personalized and scalable tools for education.

Learning about climate

Education can empower individuals to adopt more sustainable lifestyles. ML can help educate the public about climate change through conversational agents and adaptive learning techniques.

Finance

Climate investment

There is wide demand for investment in companies with low carbon footprint or those that actively address the climate crisis, both for reasons of societal benefit and because these are widely expected to be good investments in the long term. ML can help, potentially, both in designing portfolios and in timing investments.

Climate analytics

High Leverage

Climate change will seriously impact a wide swath of global companies, and quantifying the expected financial impacts can incentivize investors and companies to act. ML can help forecast prices in carbon markets, identify climate risks and investment opportunities from corporate disclosures, and quantify the monetary impact of climate change on supply and demand.