Bioretention is an adapted landscape feature that provides onsite storage and infiltration of collected stormwater runoff. Stormwater runoff is directed from surfaces to a shallow depression that allows runoff to pond prior to infiltration in an area that is planted with water-tolerant vegetation. As runoff accumulates, it will pond and slowly travel through a filter bed (pictured on the right) where it either infiltrates into the ground or is discharged via an underdrain. Small-scale bioretention areas are often referred to as rain gardens.
Training can help to better equip staff to assess green infrastructure proposals. For example, EPA offers a Green Infrastructure Webcast Series. EPA and other federal agencies and nongovernmental organizations have formed the Green Infrastructure Collaborative, a network to help communities more easily implement green infrastructure.
This ordinance can help local jurisdictions incorporate climate change projections or green infrastructure incentives into local legislation. For example, the City of Seattle developed a citywide model ordinance for stormwater management using green infrastructure.
Update or use existing tools including the EPA's National Stormwater Calculator, the Center for Neighborhood Technology's Green Values National Stormwater Management Calculator and The Value of Green Infrastructure guide.
Assess whether green infrastructure could be included as a control measure in Municipal Separate Storm Sewer Systems (MS4s). MS4s transport stormwater runoff that is often discharged into water bodies. Since 1999, even small MS4s within and outside urbanized areas have been required to obtain National Pollutant Discharge Elimination System permit coverage. Jurisdictions with MS4s can include green infrastructure as a control measure. EPA published a factsheet that discusses how green infrastructure can be integrated into stormwater permits and provides examples of communities that have done so.
Coordinate across federal, state, local, and tribal agencies Engage the full suite of agencies and departments, particularly at the federal level, that affect or could be affected by solutions to address changing climate conditions in stormwater management. Consider involving, for example, FEMA, the Army Corps of Engineers, Departments of Transportation, Parks and Recreation, and State Departments of Ecology or Natural Resources. Also encourage a "no wrong door policy" (i.e., that data and information is shared across web portals and resources are shared across agencies). Seven federal agencies have come together with nongovernmental organizations and private-sector entities to support the Green Infrastructure Collaborative, a network to help communities more easily implement green infrastructure.
,
,
Assemble existing data sets with information such as historic land use, planned development, topography, and location of floodplains. They are often sufficient to support a near-term conversation about how stormwater management may need to change to accommodate changes in climate. Land use has a tremendous effect on climate change impacts on stormwater management; managers can incorporate land use change maps into planning discussions. EPA's Integrated Climate and Land Use Scenarios (ICLUS) project can serve as a resource. Consider updates to data management practices to facilitate use of the best and most recent data.
Use resources to show historical and future trend lines to understand future climate changes, techniques that use historic data, such as analogue events or other sensitivity and threshold information in the historic record, can be used as illustrations (e.g., see the IPCC [Intergovernmental Panel on Climate Change] report Climate Change 2001: Working Group II: Impacts, Adaptation, and Vulnerability, Section 3.5. EPA's SWC and SWMM-CAT provide regional downscaled climate projections. EPA is also developing a web application for visualizing and downloading climate model output.
Use regional climate change, population demographics, transportation demand, and related projections to understand where community assets could be vulnerable.
Read more:
Align land use, hazard mitigation, transportation, capital improvement, and other plans so all plans are working toward the same goals.
Read more:
Create a list of desired development elements in more-vulnerable areas, and encourage or require developers to implement a certain number of them.
Read more:
Evaluate development incentives to see if they encourage development in particularly vulnerable areas.
Read more:Conduct a safe growth audit.
Read more:Improve public education about the risks of developing in sensitive areas.
Read more:Assess and address the needs of people who might be particularly vulnerable and/or are likely to be most affected, especially if they live in higher-risk areas.
Read more:
Use scenario planning to inform local planning and policies.
Read more:
Incorporate fiscal impact analysis into development review, and make sure it includes costs related to climate change impacts.
Read more:
Encourage on-site renewable energy generation and storage.
Read more:
Incorporate into capital projects features that enhance resilience and bring multiple other benefits.
Read more:
Plan for post-disaster redevelopment before a disaster strikes.
Read more:
Pilot a sustainable streetscape program with green infrastructure features.
Read more:
Help private property owners better manage stormwater through education and incentives.
Read more:
Design open space in flood plains for multiple amenities.
Read more:
Require new development or redevelopment to capture and infiltrate the first 1 or 1.5 inches of rain.
Read more:
Update any Clean Water Act Section 402 National Pollution Discharge Elimination System permits to consider climate change.
Read more:
Restrict development in areas buffering water bodies or wetlands.
Read more:
Adopt green and complete streets design standards.
Read more:
Acquire properties at risk of flooding, use the land for infiltration, and help the property owners resettle in the community.
Read more:
Enter a community-based public-private partnership to install and maintain green infrastructure.
Read more:
Create an overarching framework for water-related initiatives.
Read more:
Establish elevation requirements with design guidelines for streets and infrastructure.
Read more:
Adopt a site plan requirement that requires all new development to retain all stormwater on-site.
Read more:
Add projected sea level rise to flood zone hazard maps that are based exclusively on historical events.
Read more:
Designate and protect "transition zones" near tidal marshes.
Read more:
Change the definition of "normal high water" for land adjacent to tidal waters to change regulatory setbacks.
Read more:
Incorporate sea level rise impacts into all future land use planning and regulations using projections rather than past trends.
Read more:
Strengthen building codes in coastal zones by requiring additional adaptation strategies.
Read more:
Modify the steep-slope ordinance to account for slopes exposed to increased moisture due to changes in precipitation and sea level rise.
Read more:
Create an overlay district where flood regulations and standards would apply, or establish context-sensitive shoreline classifications with appropriate standards.
Read more:
Design for disassembly and adaptability in buildings.
Read more:
Designate and protect working waterfronts.
Read more:
Implement rolling development restrictions.
Read more:
Begin planning for managed retreat from the shoreline.
Read more:
Offer financial or procedural incentives to use passive survivability.
Read more:
Map "hot spots," and conduct pilot programs in these places to reduce heat.
Read more:
Offer incentives to plant and protect trees.
Read more:
Include reducing heat island effects as an objective in complete streets projects.
Read more:
Require or encourage green or reflective roofs on new buildings with little or no roof slope.
Read more:
Revise the zoning ordinance to allow urban agriculture.
Read more:
Require shade trees in all municipal projects and private parking lots.
Read more:
Adopt a tree canopy or urban forest master plan and implementing ordinances.
Read more:
Require or offer incentives for using cool paving in municipal capital improvement projects.
Read more:
Amend site plan requirements and design guidelines to encourage light or permeable paving, shade, green alleys, vegetation, and tree canopy.
Read more:
Adopt an energy conservation code to establish minimum requirements for energy efficiency in buildings, or adopt a stretch or reach code.
Read more:
Incorporate passive survivability into the building code.
Read more:
Recommend the use of drought-tolerant plants or xeriscaping as part of water conservation, landscaping, and water waste ordinances.
Read more:
Promote the use of WaterSense-rated plumbing fixtures through incentives.
Read more:
Implement a water impact fee that reflects each property's consumption.
Read more:
Offer rebates or other incentives to encourage drought-tolerant plants, residential rainwater harvesting, water-efficient fixtures, or other water-saving practices.
Read more:
Mandate graywater-ready residential development.
Read more:
Adopt a citywide policy promoting water recycling for nonpotable uses.
Read more:
Require use of water-efficient fixtures through the building code.
Read more:
Enact a building energy and water benchmarking ordinance.
Read more:
Enact a water conservation or water waste ordinance to restrict the type of landscaping on new development and public properties.
Read more:
Mandate rainwater harvesting for all new commercial construction.
Read more:
Integrate water resource management with land use plans to make sure the community has enough water for planned growth.
Read more:
Incorporate wildfire scenario planning into local planning.
Read more:
Strengthen requirements for building and roof materials to be fire-resistant and green.
Read more:
Require sites for new emergency facilities to be outside of high-risk areas, well-connected, and easy to access.
Read more:
Require new developments to submit a fire protection plan during site plan review.
Read more:
Adopt wildfire hazard overlay districts with development regulations based on factors like slope, structure, and fuel hazards.
Read more:
Require subdivisions to have a highly connected street network with multiple connection points to the external street network.
Read more:
Acquire and maintain open space between dense forested areas and residential development.
Read more:
The adaptation strategies provided on this site are intended to inform and assist communities in identifying potential alternatives. They are illustrative and are presented to help communities consider possible ways to address anticipated current and future threats resulting from the changing climate. In particular, it is important to note:
share this!
August 29, 2024
This article has been reviewed according to Science X's editorial process and policies . Editors have highlighted the following attributes while ensuring the content's credibility:
fact-checked
trusted source
written by researcher(s)
by Steven Lam and Gloria Novović, The Conversation
Canada's first National Adaptation Strategy urges Canadians to consider climate change impacts in their everyday decisions.
The strategy calls such an approach "climate mainstreaming." The approach states that: "as climate impacts become more severe and frequent, and the costs mount, incorporating adaptation considerations in health, social, environmental, infrastructure and economic decision-making is critical to ensure that our collective efforts keep pace."
Similar statements are outlined in the 2023 press release of the Intergovernmental Panel on Climate Change (IPCC). The IPCC Chair Hoesung Lee stated, "mainstreaming effective and equitable climate action will not only reduce losses and damages for nature and people, it will also provide wider benefits."
Global greenhouse gas emissions need to be cut 43% from 2019 to 2030 to limit global warming to 1.5 C. At the 2023 United Nations climate conference (COP28) in Dubai, parties were deemed off track in meeting their Paris Agreement goals.
A rapid and meaningful expansion of climate mainstreaming—the integration of climate considerations into all development programs and policies—is vital for addressing the urgent global climate crisis.
Mainstreaming climate considerations ensures that responses to climate change are systemically embedded in all policies and actions, rather than treated as a separate issue. This integration allows for more comprehensive and cost-effective interventions by addressing multiple issues at once.
For instance, within an ongoing program focused on improving food safety in informal, outdoor markets through enhanced hygienic practices, mainstreaming might entail additional activities related to climate adaptation, such as raising awareness among food vendors about the importance of refrigeration during heat waves to prevent bacterial growth .
Failing to mainstream climate considerations can hinder climate action as well as result in maladaptation, which occurs when well-intentioned development actions inadvertently increase climate impacts. For example, seawalls can protect people and property from damage in the short term. However, if they are not part of a long-term plan that can adapt to changing conditions, they can trap communities in risky situations and increase their exposure to climate risks over time .
While attention to climate mainstreaming calls for the prioritization of climate considerations across all policy arenas, progress remains slow and uneven due primarily to an institutional resistance to change. Climate action is often seen as the responsibility of a single sector rather than the collective, and incremental changes are inferior to transformative ones.
Furthermore, climate mainstreaming is often narrowly interpreted as simply the addition of climate to existing structures and initiatives. Often derisively dubbed a "just add climate and stir" approach.
To help address these preconceptions, our research has explored how climate mainstreaming challenges resemble similar decades-long struggles to mainstream gender equality across international and national public policy agendas. The question we have asked is: what can climate mainstreaming learn from gender mainstreaming?
The longer history of gender mainstreaming, including institutional investments dating back to the 1990s, offers lessons about policy and institutional bottlenecks of mainstreaming. These lessons can help tackle political and institutional challenges of climate mainstreaming. The UN system, with clear gender and climate mainstreaming targets, offers a suitable arena for analysis.
In a new study published in 2024 , we reviewed documents of United Nations agencies working in the food and agriculture sector, which is strongly impacted by climate change. We found varying degrees of gender and climate mainstreaming across selected UN agencies.
Key areas where climate mainstreaming fell short compared to gender mainstreaming included: strategic planning, leadership, organizational culture and accountability.
Our review showed ways to improve climate mainstreaming. Here are three actions governments, development partners and industries can take now:
Accountable and integrated climate justice interventions are prerequisites for a more sustainable and resilient future. Financing is another.
While mainstreaming is important, it is nothing without adequate financing. The 2015 Paris Agreement requires high-income countries to contribute $100 billion annually . However, this goal has not been met , and the existing funds are unevenly distributed.
Historically disadvantaged countries are the least responsible for yet the most impacted by climate crisis . These countries are largely left to balance development and climate action investments in a generally unjust international financial system.
In 2022, official development assistance reached US$204 billion , but this still left nearly half of the humanitarian requirements unmet. Rich countries spent only 0.36% of their total income on aid—slightly up from 0.33% in 2021, but still much lower than the 0.7% promised back in 1970.
With the financing to back it up, a climate mainstreaming perspective may just be the solution to addressing both global development and climate goals.
Provided by The Conversation
Explore further
Feedback to editors
55 minutes ago
2 hours ago
Relevant physicsforums posts, alaska - pedersen glacier: landslide triggered tsunami.
Aug 23, 2024
The secrets of prof. verschure's rosetta stones, shiveluch volcano erupts on kamchatka peninsula.
Aug 18, 2024
Aug 14, 2024
Aug 13, 2024
More from Earth Sciences
May 2, 2024
Aug 1, 2024
Jun 26, 2024
Jun 24, 2020
Mar 5, 2024
Jun 20, 2024
7 hours ago
10 hours ago
5 hours ago
Aug 28, 2024
Use this form if you have come across a typo, inaccuracy or would like to send an edit request for the content on this page. For general inquiries, please use our contact form . For general feedback, use the public comments section below (please adhere to guidelines ).
Please select the most appropriate category to facilitate processing of your request
Thank you for taking time to provide your feedback to the editors.
Your feedback is important to us. However, we do not guarantee individual replies due to the high volume of messages.
Your email address is used only to let the recipient know who sent the email. Neither your address nor the recipient's address will be used for any other purpose. The information you enter will appear in your e-mail message and is not retained by Phys.org in any form.
Get weekly and/or daily updates delivered to your inbox. You can unsubscribe at any time and we'll never share your details to third parties.
More information Privacy policy
We keep our content available to everyone. Consider supporting Science X's mission by getting a premium account.
The world can take a major step to meeting the goals of the Paris Climate Accord by focusing on 63 cases where climate policies have had the most impact, new research has revealed. The findings have been published today in Science .
Our results inform contentious policy debates in three main ways. First, we show evidence for the effectiveness of policy mixes. Second our findings highlight that successful policy mixes vary across sectors and that policy-makers should focus on sector-specific best practices. Third our results stress that effective policies vary with economic development. Study co-author Dr Moritz Schwarz , an Associate at the Climate Econometrics Programme at the University of Oxford
The study, led by Climate Econometricians at the University of Oxford, the Potsdam Institute for Climate Impact Research (PIK), and the Mercator Research Institute on Global Commons and Climate Change (MCC), analysed 1,500 observed policies documented in a novel, high quality, OECD climate policy database for effectiveness. It is the first time a global dataset of policies has been compared and ranked in this way.
Using a methodology developed by Climate Econometrics at The Institute for New Economic Thinking at the Oxford Martin School (INET Oxford), the researchers measured ‘emission breaks’ that followed policy interventions. The break detection methodology, called indicator saturation estimation, developed at Climate Econometrics, allows break indicators for all possible dates to be examined objectively using a variant of machine learning.
The results were sobering: Across four sectors, 41 countries, two decades and 1,500 policies, only 63 successful policy interventions with large effects were identified, which reduced total emissions between 0.6 and 1.8 Gt CO2.
However, the authors say the good news is that policymakers can learn from the 63 effective cases where climate policies had led to meaningful reductions to get back on track.
The researchers have made the data available to policy-makers across the world, and have produced a sector by sector, country by country data visualisation in a dashboard .
Overall, the Team concluded:
Scaling up good practice policies identified in this study to other sectors and other parts of the world can in the short term be a powerful climate mitigation strategy…The dashboard that we make available to policy-makers provides an accessible platform to conduct country-by-country, sector-by-sector comparisons and to find a suitable policy mix for different situations. Study co-author Professor Felix Pretis , Co-Director of the Climate Econometrics Programme at Nuffield College, University of Oxford
Study co-author Ebba Mark , researcher at the Calleva Project at INET Oxford, said the world needed to get back on track to meeting the Paris Climate Accord targets. ‘Meeting the Paris Climate objectives necessitates decisive policy action as we are still falling short - data from the UN estimates that there remains a median emissions gap of 23 billion tonnes of CO2 equivalent by 2030 . It is now clear that the persistence of this emissions gap is not only attributable to an ambition gap but also to a gap in the real versus expected outcomes of implemented policies. The 63 success stories identified in this study provide key information about how we can bridge the emissions gap more meaningfully going forward.'
What works: Examples from the UK and USA
The country by country analysis showed that the UK has made very successful progress in the electricity sector, with two adjacent breaks detected following the mid-2013 introduction of a carbon price floor that imposed a minimum price for UK power producers. However, the study did not find in other UK sectors any major emission reductions following a policy intervention beyond what would be expected based on long-term economic and population dynamics.
The US has managed to reduce carbon emissions in the transport sector following actions taken in the aftermath of the financial crisis. While successful policy implementation in the transport sector is generally difficult and hence can be viewed as a positive example for the climate policy globally, the lack of any further climate policy successes in other sectors points to huge remaining challenges in the power sector or industry.
Dr Anupama Sen , Head of Policy Engagement at the Oxford Smith School of Enterprise and the Environment said: ' In more than 80% of investments the total lifetime cost of a clean technology is considerably lower than that of a fossil technology. While the new UK government’s policies are moving in the right direction, they need to go further and faster to unlock these lower costs. New Oxford research now provides evidence that an optimal mix of policies can achieve this, and rapidly lower a country’s emissions.'
Further analysis can be found in INET Oxford’s accompanying Insight brief .
The study ‘Climate policies that achieved major emission reductions: Global evidence from two decades’ has been published in Science .
You can view all news or browse by category
Aug. 27, 2024
Jon A. Krosnick and Bo MacInnis
Publication
Reading time
In Climate Insights 2024: American Understanding of Climate Change, we showed that huge majorities of Americans believe that the earth has been warming, that the warming has been caused by human activity, that warming poses a significant threat to the nation and the world—especially to future generations—and that governments, businesses, and individuals should be taking steps to address it.
In this report, we turn to specific federal government opportunities to reduce future greenhouse gas emissions, often referred to as climate change mitigation. Policies to accomplish this goal fall into several categories, including:
This 2024 survey asked Americans about their opinions on a wide array of such policies, which allows us not only to assess current opinions, but to track changes in those opinions over the past two decades through comparisons with responses to comparable questions asked in earlier national surveys.
Click here to explore the report's findings using our interactive data tool.
In 2024, we asked for the first time whether Americans prefer using “carrots” to reduce emissions or “sticks.” The former entails offering incentives to reward companies for achieving desired outcomes, and the latter involves penalizing companies that fail to reach desired goals. 59 percent of Americans prefer a carrot approach in which government lowers taxes for companies that reduce emissions, and 35 percent prefer a stick approach such that government raises taxes on companies that do not reduce emissions (see Figure 1).
Over the past decades, a consistently large majority of Americans has wanted the government to reduce greenhouse gas emissions by US businesses. In 2024, 74 percent of Americans endorse this mitigation policy principle (see Figure 2). This number is not significantly different from the 77 percent seen in 2020 and is about the same as it has been since 1997 when this series of surveys was launched.
Taxing imported emissions.
In 2024, we asked about import taxes tied to emissions; respondents were asked whether they would favor taxing foreign companies for importing products that put out more greenhouse gases than a comparable US product. A huge majority of Americans, 84 percent , favor the special tax (see Figure 3).
In 2024, we asked whether the federal government should spend money to help people who lose jobs due to a transition from fossil-based electricity generation to electricity generated from renewable sources. 78 percent of Americans favor the government paying those people to learn to do other kinds of work.
In 2024, we asked whether the federal government should spend money to close off abandoned oil wells that emit greenhouse gases; 76 percent of Americans favor the government spending money to fill these old wells.
Huge numbers of Americans favor government effort to shift electricity generation away from fossil fuels and toward renewable energy sources.
In 2024, 72 percent of Americans believe that the US government should offer tax breaks to utilities in exchange for making more electricity from water, wind, and solar sources. However, this is a statistically significant decline from the 85 percent seen in 2020, and a record low since 2006 (see Figure 6).
13 percentage points fewer Americans believe that the US government should offer tax breaks to utilities in exchange for making more electricity from renewable sources in 2024 than 2020.
Slight changes to the question wording yielded similar results: 76 percent of Americans in 2024 favor either mandates or tax breaks for utilities to reduce greenhouse gas emissions from power plants. This number is not significantly different from the 82 percent seen in 2020 and is about the same as it has been since 2009. Before 2009, this proportion was slightly higher: 86 percent and 88 percent in 2006 and 2007, respectively (see Figure 7).
About two-thirds of Americans favor government efforts through tax breaks or mandates to improve the energy efficiency of various consumer products (see Figure 8).
Specifically, 62 percent of Americans in 2024 favor increasing the fuel efficiency of automobiles, a statistically significant drop from the 72 percent seen in 2020.
68 percent favor increasing the energy efficiency of appliances, similar to the 71 percent observed in 2020.
69 percent favor increasing the energy efficiency of new buildings, a statistically significant decline from the 76 percent in 2020.
In 2024, 63 percent of Americans favor reducing emissions by sequestering (i.e., capturing and storing) carbon released by burning coal. This level of support has been steady over the past 15 years (see Figure 9).
84% of respondents favor import taxes tied to emissions, making it the most popular policy we surveyed.
Reducing subsidies for fossil fuels.
In 2024, we asked for the first time whether the federal government should continue its long-standing practice of offering subsidies to oil and natural gas companies by reducing their taxes.
61 percent of Americans favor ending government reduction of oil companies’ taxes, and 37 percent believe these subsidies should continue.
42 percent of Americans favor ending government reduction of natural gas companies’ taxes, and 56 percent believe that these subsidies should continue.
When asked whether companies should be charged a tax for every ton of greenhouse gases they emit, 54 percent of respondents were in favor in 2024, a statistically significant decline from the 66 percent observed in 2020 (figure 10).
Although economists generally assert that a carbon tax incentivizes companies to reduce emissions (Baumol and Oates, 1971; Climate Leadership Council, 2019; Marron and Toder, 2014; Montgomery, 1972; World Bank, 2017), a carbon tax does not guarantee that such emissions reductions will happen.
A cap-and-trade or cap-and-dividend policy, on the other hand, are alternative policies in which a government sets a limit, or ‘cap,’ on emissions. The cap is imposed by government-issued permits that limit emissions. The government gives, sells, or auctions the permits to companies, creating an opportunity to generate revenue. A cap-and-dividend program would return this revenue to consumers through a rebate.
The logic in asking this question about cap and trade is to assess whether more Americans would favor a greenhouse gas tax if assured that it would result in emissions reductions. However, we show cap-and-trade and cap-and-dividend policies are not notably more popular than straightforward taxes.
In 2024, 52 percent of Americans favor a cap-and-dividend policy, a statistically significant decline from the 63 percent observed in 2020 (see FIgure 11).
In 2024, we asked respondents whether the federal government should spend money to help people install solar panels on houses and apartment buildings. Respondents were randomly assigned to be asked one of four versions of the question. Two versions asked about the government paying all of the installation costs, and the other two versions asked about the government paying some of the costs.
For half of each group (chosen randomly), the question was preceded by this introduction:
“Solar panels can generate electricity when the sun is shining, and that electricity can be stored in batteries to be used when the sun is not out. However, companies that make electricity cannot install enough solar panels to make all of the electricity needed in the country. People can put solar panels on the roofs of many houses and apartment buildings so much more of America’s electricity can be made from the sun. But it is expensive to do this, and most people cannot afford to pay that amount of money.”
Among people who did not hear the introduction, 51 percent favor the government paying some of the cost, and 42 percent favor the government paying all of the costs.
Among people who did hear the introduction, 77 percent favor the government paying some of the cost, and 74 percent favor the government paying all of the costs.
In 2024, we asked whether the federal government should expedite the process of granting permits to build new power plants that make electricity from sources other than coal and petroleum. 52 percent of Americans favor expediting this process.
Nuclear power tax breaks.
Although nuclear power does not directly emit greenhouse gases, tax breaks for the construction of new nuclear power plants are among the least popular policies asked about in 2024. 47 percent of Americans favor this policy; however, it is notable that this is a statistically significant increase from the 37 percent observed in 2020 (see FIgure 12).
In 2024, 46 percent of Americans—a record low—think the government should require or give tax breaks to companies to build all-electric vehicles, a statistically significant decline from the 60 percent observed in 2015 when this question was last asked (see Figure 13).
The least popular policies impose new taxes on consumers to incentivize them to consume less fossil fuel. Few Americans favor increasing taxes on retail gasoline and electricity purchases for this purpose. 15 percent approve increasing taxes on electricity, a statistically significant decline from the 28 percent observed in 2020. Likewise, 28 percent approve increasing taxes on gasoline, a statistically significant decline from the 41 percent observed in 2020 (see Figure 14).
Perceived effect on the economy.
Implementing many policies to reduce greenhouse gas emissions will cost consumers and companies in the short term. Implementing such policies may also increase the cost of American-made goods and services relative to the costs of those goods and services produced elsewhere. This has led some observers to urge caution about implementing greenhouse gas emissions reduction policies (e.g., Cassidy, 2023; Gross, 2021), because they may result in undesirable economic side effects.
However, this argument does not appear to have taken hold with the majority of Americans. For example, only 36 percent of Americans in 2024 believe that taking action to address global warming will hurt the US economy, about the same as was observed in 2013 (30 percent), though this is a statistically significant increase from the 29 percent observed in 2020 (see FIgure 15). Likewise, in 2024, 34 percent believed that these efforts would hurt their state economy, a statistically significant increase from the 24 percent observed in 2020.
More Americans believe that climate action will help the economy. 44 percent of Americans believe this in 2024, about the same as was observed in 2020 (48 percent) and 15 years ago (46 percent) (see Figure 15). 39 percent of Americans believe that efforts to reduce global warming will help their state economy in 2024, a statistically significant decrease from 46 percent in 2020.
A similar picture emerged regarding beliefs about how climate action will affect job availability. Only 27 percent of Americans believe that efforts to reduce emissions will reduce the number of jobs in the nation—the same as was observed in 2020 (see Figure 16). And in 2024, 35 percent of Americans believe that climate change action will increase the number of jobs in the country, similar to the 39 percent observed in 2020.
28 percent of Americans believe that climate change action will reduce the number of jobs in their state, similar to the 23 percent observed in 2020 (see Figure 16). And 32 percent of Americans in 2024 believe that climate action will increase the number of jobs in their state, about the same as the 35 percent observed in 2020.
27 percent of Americans believe that efforts to reduce emissions will reduce the number of jobs in the nation
In 2024, we asked respondents about the likely impact of mitigation efforts on their own personal economic situation. A majority of Americans believe that they will have the same amount of money regardless of mitigation efforts (54 percent). 36 percent believe their wealth will decrease, a statistically significant increase over the 20 percent observed in 2020 (see Figure 17). But 8 percent believe that climate change mitigation will increase their wealth, similar to the 10 percent observed in 2020.
Likewise, a majority (64 percent) believe that mitigation efforts will have no impact on their chance of getting a good-paying job. 17 percent believe that mitigation efforts will make them less likely to get a good-paying job, a statistically significant increase from the 12 percent observed in 2020 (see Figure 17). 17 percent believe that mitigation efforts will increase their ability to get a good-paying job, similar to the 16 percent observed in 2020.
Are the many policy preferences outlined in this report just talk, or do they inspire action in the voting booth? We turn to that question next and describe the findings from a test.
Respondents were read a statement by a hypothetical candidate running for a seat in the US Senate and were asked whether hearing that statement makes the respondents more likely to vote for the candidate, less likely to vote for the candidate, or has no impact.
One statement expressed “green views” that summarized opinions expressed by majorities of Americans:
“I believe that global warming has been happening for the past 100 years, mainly because we have been burning fossil fuels and putting out greenhouse gases. Now is the time for us to be using new forms of energy that are made in America and will be renewable forever. We can manufacture better cars that use less gasoline and build better appliances that use less electricity. We need to transform the outdated ways of generating energy into new ones that create jobs and entire industries and stop the damage we’ve been doing to the environment.”
The other statement proposed expanding production of energy from traditionally used fossil fuels:
“The science on global warming is a hoax and is an attempt to perpetrate a fraud on the American people. I don’t buy into the whole man-caused global warming mantra. We must spend no effort to deal with something that is not a problem at all. We should not invest in windmills and solar panels as alternative energy sources. Instead, we should continue to focus on our traditional sources of energy: coal, oil, and natural gas. We should expand energy production in our country, including continuing to mine our coal and doing more drilling for oil here at home.”
Hearing a green view makes 57 percent of Americans more likely to vote for the candidate, a statistically significant decline from the 65 percent observed in 2020 (see Figure 18). Democrats are significantly more likely to be attracted to a “green” candidate (83 percent) than are Independents (56 percent) and Republicans (23 percent) (see Figure 19).
Hearing the green statement makes only 18 percent of Americans less likely to vote for the candidate (see Figure 18).
Hearing the candidate make a “not-green” statement makes only 21 percent more likely to vote for the candidate (see Figure 18). 43 percent of Republicans are more likely to support the candidate, compared to 20 percent of Independents and 5 percent of Democrats (see Figure 19).
Hearing the “not-green” statement makes 63 percent of Americans less likely to vote for the candidate, similar to the 66 percent observed in 2020 (see FIgure 18). This proportion is greatest among Democrats (88 percent), smaller among Independents (62 percent), and still smaller among Republicans (29 percent) (see Figure 19).
Because almost all Republican citizens vote for Republican candidates and almost all Democratic citizens vote for Democratic candidates, the greatest impact of candidate statements in shaping election outcomes is among Independents. Among them, the same pattern appears that appears among all Americans: taking a “green” position helped a candidate and taking a “not-green” position hurt the candidate.
Taken together, these results point to climate change mitigation policies that may be pursued in the future with widespread public support (such as efforts to reduce emissions from power plants). Furthermore, these results also identify a few policy directions that are well received by few Americans, despite being plausible in theory and in practice (like taxes on electricity and gasoline).
For decades, one school of thought commonly followed by some scholars and policymakers is that economic growth and environmental protection are incompatible, and that any efforts to grow the economy must, of necessity, take resources away from helping the environment. Such a presumption creates an “either the economy or the environment” mindset. This mindset has been reinforced by survey questions asking Americans (e.g., Mildenberger & Leiserowitz, 2017), for example: “With which one of these statements about the environment and the economy do you most agree? ‘Protection of the environment should be given priority, even at the risk of curbing economic growth.’ Or: ‘Economic growth should be given priority, even if the environment suffers to some extent’” (Gallup, 2024).
If Americans do perceive this trade-off as inevitable, the COVID-19 pandemic and associated economic crisis a few years ago might have tilted them away from environmental protection generally and away from efforts to mitigate climate change; the subsequent recovery might have increased support for such efforts. The present study refutes that notion resoundingly. In fact, we see small changes in the opposite direction: slightly less public support for some emissions-reducing policies than four years ago.
Few people believe that taking steps to reduce emissions will hurt the national economy, their state’s economy, or their personal finances, and more Americans believe that such policies will improve these economic outcomes.
Finally, we saw that the policy positions candidates take on this issue are likely to influence the votes of many Americans. Thus, policymakers and their challengers have opportunities to use these issues to help assemble the coalitions needed to accomplish electoral victories. By taking “green” positions, candidates gain considerably more votes than they lost.
Click here to view the full report with citations.
PDF — 5.2 MB
Click here to read more about the methodology and survey questions.
PDF — 716.3 KB
Click here to read the report press release.
Jon A. Krosnick
University Fellow
Jon A. Krosnick is an RFF university fellow and Stanford University-based social psychologist who does research on attitude formation, change, and effects, on the psychology of political behavior, and on survey research methods.
Bo MacInnis
Lecturer, Political Psychology Research, Stanford University
Report — Jul 15, 2024
Climate Insights 2024: American Understanding of Climate Change
This first report in the 2024 Climate Insights report series by researchers at Stanford University and RFF examines American public opinion on issues related to climate change.
Policy Leadership Series — Oct 8, 2024
The Future of Environmental and Energy Justice: A Policy Leadership Series Event with Shalanda Baker
Press Release — Aug 27, 2024
Most Americans Approve of Government-Led Climate Action, but Support for Some Policies is Cooling
A new survey of American opinions on climate has found that Americans widely support government action to tackle climate change—but support for some policies has dropped since 2020.
In Focus — Aug 27, 2024
In Focus: American Opinions on Climate Policies
This video shares insights about a new survey that dives into American opinions on federal climate policy, with RFF University Fellow and Stanford University Professor Jon A. Krosnick.
We use cookies to provide you with a better service. Carry on browsing if you're happy with this, or find out more.
An official website of the United States government
The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.
The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.
Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .
Maria maddalena tortorella.
1 School of Agricultural, Forestry, Food and Environmental Sciences, University of Basilicata, 85100 Potenza, Italy; [email protected] (M.V.); [email protected] (M.C.); [email protected] (S.R.)
2 Institute of Methodologies for Environmental Analysis-National Research Council of Italy (CNR-IMAA) C.da S. Loja, 85050 Tito Scalo (PZ), Italy; [email protected] (S.D.L.); [email protected] (C.C.); [email protected] (F.P.); [email protected] (M.S.)
Carmelina cosmi, patrícia fortes.
3 CENSE (Center for Environmental and Sustainability Research), NOVA School of Science and Technology, NOVA University Lisbon, 2829-516 Caparica, Portugal; [email protected]
Mario cozzi, filomena pietrapertosa, monica salvia, severino romano.
The European Union’s 2030 climate and energy policy and the 2030 Agenda for Sustainable Development underline the commitment to mitigate climate change and reduce its impacts by supporting sustainable use of resources. This commitment has become stricter in light of the ambitious climate neutrality target set by the European Green Deal for 2050. Water, Energy and Food are the key variables of the “Nexus Thinking” which face the sustainability challenge with a multi-sectoral approach. The aim of the paper is to show the methodological path toward the implementation of an integrated modeling platform based on the Nexus approach and consolidated energy system analysis methods to represent the agri-food system in a circular economy perspective (from the use of water, energy, biomass, and land to food production). The final aim is to support decision-making connected to climate change mitigation. The IEA-The Integrated MARKAL-EFOM System (TIMES) model generator was used to build up the Basilicata Water, Energy and Food model (TIMES-WEF model), which allows users a comprehensive evaluation of the impacts of climate change on the Basilicata agri-food system in terms of land use, yields and water availability and a critical comparison of these indicators in different scenarios. The paper focuses on the construction of the model’s Reference Energy and Material System of the TIMES model, which integrates water and agricultural commodities into the energy framework, and on the results obtained through the calibration of the model β version to statistical data on agricultural activities.
The 2030 Agenda for Sustainable Development identifies 17 Sustainable Development Goals (SDGs) to be achieved by 2030, which aim to encourage a change in the current development model regarding the environmental, economic, and social dimensions [ 1 ]. This document represents one of most important global agreement that highlights an integrated and multi-sectoral vision of the different dimensions of sustainable development. It represents an important reference for the nexus approach, addressing the risks and changes associated with the reduced availability of water, energy, and food, in a growing World’s population context (8.6 billion by 2030 and 9.8 billion by 2050) [ 2 ]. In addition, the Paris Agreement commits the signatory parties to reduce drastically their greenhouse gas (GHG) emissions and to take urgent actions to combat climate change and its impacts, supporting a transformation of anthropogenic activities toward more sustainable trajectories [ 3 ]. As concerns Europe, the need to make production and consumption patterns more sustainable is also emphasized by the European Green Deal, which aims to make Europe climate-neutral in 2050. In view of this ambitious goal, the political objectives and targets set for the period 2021 to 2030 in the EU’s 2030 climate and energy policy framework will be made more ambitious to effectively support the transition to a climate-neutral economy [ 4 ] and to implement the Paris agreement commitments.
To reach the policy objectives, it is necessary to implement coordinated actions that can guarantee economic growth and at the same time a drastic reduction of GHG emissions, to mitigate climate change and support environmental protection.
The “Nexus Thinking” and its multi-sector approach are therefore crucial to respond to the sustainability challenge for an effective management of resources in compliance with the SDGs, the Paris Agreement, and the European climate neutrality goal [ 5 ]. Concepts of the Nexus Thinking were successfully applied in developing countries [ 6 , 7 ], and currently also to countries with a more advanced economy [ 8 , 9 , 10 ]. Since 2011, when it was first brought to the attention of the institutions in the opening report of the World Economic Forum [ 11 ], the Nexus approach took on a central role as a method for understanding and modeling the complex interactions between the different resource systems (energy, water, food) [ 12 ]. It has therefore been successfully used both by academics and international organizations (e.g., The Food and Agriculture Organization of the United Nations (FAO), The United Nations Economic Commission for Europe (UNECE), World Wide Fund for Nature (WWF), etc.). In particular, the Nexus approach aims at managing efficiently water, energy, and food systems as a whole, minimizing potential conflicts and strengthening intersectoral integration in order to guarantee a secure and sustainable use of resources [ 13 ]. In fact, the complex interconnections between energy, water and food are difficult to represent due to the numerous variables and phenomena involved. Water affects food production (e.g., crops, livestock) as well as energy production (e.g., hydropower, cooling water). Energy affects food production (e.g., energy for chemical and mineral fertilizers, transportation, and food storage) and also water supply (e.g., water preparation, desalinization, pumping). Agriculture, a major player in food production, is a major user of water (over 70% of all water consumption globally [ 14 ]) and energy (about 30% of the total energy demand). It also affects the water sector through land degradation, changes in runoff and disruption of groundwater discharge. On the other hand, the area available for agricultural activities must also compete for a share of electricity generation from fossil and renewable sources, both in terms of the area required for the installation of power plants and the impact of the related activities (e.g., mining, dams and water flow management, biofuel production, etc.). In particular, cultivation of biofuels, which has a high profitability per hectare and, in many cases, benefits of public incentives, causes an excessive exploitation of territories, generating, indirectly, a potential pressure on the prices of food crops and increasing the competition between land use and water consumption to produce biofuels or food. Rulli et al. [ 15 ] estimated that the worldwide production of biofuel exploits 4% of the land and water used for agriculture, corresponding to an area sufficient to feed about 280 million people if used to grow food.
In this already complex situation, it is necessary to consider the negative effects of climate change that as known, affects the availability of resources and land, with significant impacts on the water-energy-food system. A recent Intergovernmental Panel on Climate Change (IPCC) Report [ 16 ] highlights how climate change increases the rate and extent of ongoing land degradation through two main factors: increased frequency, intensity of heavy rainfall and extreme high-temperature events. Furthermore, global warming will make soil degradation processes more severe in the various geographical areas due to an increased frequency of floods, droughts, cyclones and hurricanes, forest fires and sea levels rise.
The greatest risk in using the nexus approach without considering the data of climate models is to overlook the possible effects of climate change on the balance between the resources involved in the water-energy-food cycle and within their interactions [ 17 ].
In fact, both agriculture and energy production are vulnerable to changes in meteorological parameters and to the occurrence of extreme events such as drought and floods [ 18 ]. Temperatures and precipitation levels can strongly influence the availability of water, and consequently the production of energy and food. Crop productivity may increase in Northern Europe due to prolonged thawing period opening the possibility for new crops cultivation [ 19 ]. In Southern Europe, and in particular in the Mediterranean basin, crop productivity is negatively affected by droughts, heat waves, reduced water availability and other related phenomena such as pest and disease epidemics [ 20 ].
From this perspective, the adoption of a holistic approach that allows representing the interrelationships between the three sectors (water, energy, and food) is a priority to encourage a sustainable and efficient use of resources, reduce risks and define effective integrated policies.
Numerous examples in the literature underline the increasing importance of an integrated approach to water-food-energy challenges and different models have been used to assess WEF interactions ranging from economic to technological, and geographic information systems (GIS) tools. In particular, Endo et al. [ 21 ] analyzed the water, energy, and food nexus by reviewing 37 projects across different world regions, to highlight the current state of art by investigating the nexus keywords and stakeholders to characterize the specific nexus type.
Haji et al. [ 22 ] used a ‘Node’ methodology based on GIS-based approaches, the Analytical Hierarchy Process and resource assessment to evaluate the critical factors that increase the risk in open field farms and to improve water and energy efficiency. In Pakdel et al. [ 23 ] a multi-objective optimization of energy and water management based on GAMS software is performed to minimize both energy and freshwater use, introducing the concept of “energy hubs” and validating the interdependency of energy and water structures. In Lee et al. [ 24 ] a nexus approach is used to analyze the water-food-energy interconnections and their economic implications in the sugar industry in India. Nie et al. [ 25 ] focus on the agricultural system and use a multi-objective procedure and a comprehensive WEF index to select optimal land allocation strategies that can limit stresses in the water-energy-food nexus. Chiodi et al. [ 26 ] integrated the energy and agriculture systems into the IEA-ETSAP methodology to individuate the GHG reduction strategies for Ireland.
Many gaps still need to be filled in the operational application of the nexus concept for decision-making. However, Weitz et al. [ 27 ] point out that despite the still open questions, “a nexus approach promotes policy coherence through identifying optimal policy mixes and governance arrangements across the water, energy and food sectors”.
These considerations and the relevant examples discussed above, have strengthened our motivation to develop a modeling platform focused on the agricultural system that integrates the nexus concept in a framework typically used to support decision-making when different competing goals must be achieved. The outcome of this research, The Integrated MARKAL-EFOM System (TIMES)-WEF model, will be validated in selected areas of Mediterranean Europe, in order to evaluate the robustness of solutions at different spatial scales, and to perform a joint assessment of the effects of climate change and agricultural policies. The scenario analysis will focus on two IPCC pathways (RCP4.5 and RCP8.5) and multi-level (European Union, national and local) agricultural, energy and climate policies to evaluate their effects in terms of availability of land, water and energy as well as other parameters of interest for the agri-food system (e.g., fertilizers, pesticides, etc.).
The paper is structured as follows: Section 2 presents a literature review of the Nexus modeling approach and its application to the development of the TIMES-WEF model; Section 3 describes data requirement, technical assumptions, the preprocessing procedure for the implementation of the model data input and the calibration to the statistical base year data of the TIMES-WEF model for agriculture; finally, Section 4 concludes with the main outcomes and future development of the model.
2.1. state of art.
The WEF Nexus currently represents the most advanced methodological and operational approach to address the complexity of sustainable development. It aims to overcome the silo vision and evaluate the interdependencies and management of the different sectors (energy, water, food) as an integrated process. In this way, it is possible to highlight how actions in a sector can influence the management of resources, thus avoiding unwanted consequences and exploiting existing potential synergies.
The conceptualization of the WEF Nexus has become increasingly complex, incorporating a plurality of factors and dimensions e.g., environmental, economic, political, and social. Conversely, there has been a slow development of analytical approaches [ 28 , 29 ] and limited use of modeling tools to assess the correlations between water, energy and food and support an integrated decision-making process [ 30 ].
Table 1 reports a summary of main methods used in the literature to address the WEF Nexus (are included in the table also the methods that analyze only some of its components).
Summary of main methods used in the literature to address the Nexus.
Type | Brief Description | Examples |
---|---|---|
Computable General Equilibrium (CGE) | Used for long-term simulations, CGE models analyze the economic implications of policies (e.g., CO tax), assuming that all markets are in equilibrium and not considering the technological details | GEM-E3 [ ], GTAP [ ] and IMPACT [ ]. |
Econometrics models | Oriented to test economic theory through empirical evidence, they currently include open and growth-based macro econometric models, with trend/analysis of time series data on a higher level of aggregation. Their main limitation lies in the strong dependence on data | E3ME [ ] and IREDSS [ ]. |
Input-output models | Suitable for short-term assessment of policies, as they can only provide a static image of the economic structure based on historical data illustrating sectoral production techniques describing the total flow of goods and services of an economic system in terms of production, added value and specific technical input/output coefficients. | [ ]. |
Partial Equilibrium/Optimization | Used to support the decision-making process by providing policy makers with detailed information on technologies and resources on both the demand and supply sides. Partial equilibrium models are characterized by a high technology detail both in the supply and demand side and define the optimal set of technological choices to achieve multiple objectives at the minimum feasible cost in relation to predefined exogenous constraints. | POLES [ ]; MARKAL/TIMES [ , , , , , , , , , , , , ]. |
Simulation | They provide a descriptive and quantitative image of energy conversion and demand based on drivers and technical data exogenously, in order to model the decision-making process. | LEAP [ ] and BUENAS [ ]. |
GIS-based tools | Mathematical models for the representation of georeferenceable variables. They are used to transfer on a larger scale the assessments of the consumption of energy flows or other resources referred to the local scale | [ ] |
However, many applications only focus on dual sector interactions, for example water-food or water-energy, thus implementing a fragmented vision of the WEF Nexus [ 42 ], or provide a narrow perspective of the interactions between water, energy and food, with a limited ability to capture multi-sectoral interconnections and interdependencies between different systems [ 43 , 44 ].
Few studies are based on innovative methods to quantify the connections and interactions between sectors, in order to better describe the systems included in the WEF Nexus. These are modeling platforms that can support the integration of sectoral models, creating flexible tools that can accommodate new modeling inputs or extensions. In this way, decision support tools are created to combine physical models with scenario analysis, allowing decision makers to compare the impact of different policies or actions on the analyzed system [ 45 , 46 , 47 ].
Based on the above considerations, the proven validity of the Nexus approach to tackle a multi-objective problem such as an integrated management of energy, water, and soil resources [ 48 ] guided our idea to develop an integrated decision support platform, based on existing consolidated models used for policy analysis.
Taking inspiration from the Irish model [ 26 ], the TIMES energy models’ generator [ 38 ] has been used to model the agri-food system within an energy system analysis approach based on technical engineering and economic analysis to ensure a sustainable management of agricultural resources.
The TIMES (The Integrated MARKAL-EFOM System [ 38 ], (developed by the Energy Technology Systems Analysis Program (ETSAP) of the International Energy Agency (IEA), an autonomous intergovernmental organization born with the 1973–1974 oil crisis and based in Paris, France) is a bottom-up model generator, which uses linear-programming to compute a least-cost energy system, optimized according to several user exogenous constraints, over medium to long-term time horizons [ 38 ]. It is widely used to represent local, national, and multiregional energy systems and to perform scenario analysis, exploring possible energy futures in relation to environmental and technical constraints, such as policy measures.
The TIMES models are driven by the end-use sector demands (Industry, Residential, Commercial, Transport and Agriculture).
The energy system configuration is optimised to provide the least-cost solution that corresponds to the best allocation of resources and technologies, which fulfil end-use demands and the scenarios’ constraints at the minimum total discounted cost of the system. The optimization of the reference scenario provides the baseline for the comparison of solutions in the alternative scenario analysis.
The TIMES model structure is usually described through the Reference Energy and Materials System (REMS), which describes the entire supply–demand chain, providing an accurate representation of energy flows from supply/conversion technologies to demand processes. It allows representing all the components related to energy production and use, including emissions and materials. The supply chain describes the extraction import/export and secondary production of primary resources (typically energy and materials) whereas the demand chain represents in detail the commodity flows through the network of real or dummy technologies (or processes), (e.g., mining processes, import processes, energy transformation plants, end-use devices). Any item produced or consumed by a certain technology is called “commodity” (e.g., energy carriers, energy services, materials, money flows and emissions).
The key inputs to the TIMES model deals with all specific data that characterize the system under focus: energy demand, primary energy supply (availability of present and future sources), techno-economic factors (technology development and associated costs), environmental variables (e.g., GHG emission factors), and other policy parameters.
This research takes the standard TIMES modeling framework as a starting point to develop a novel model that focus on the agri-food system through a WEF nexus perspective and that can be merged into the general energy modeling framework, exploiting all conversion processes and the end-use sectors related to agriculture.
The TIMES-WEF model represents an innovative application of the water-energy-food nexus approach into the ETSAP-TIMES framework where land use is chosen as an independent driving parameter to connect soil availability with input/output commodities. In particular, a land use-driven model allows evaluating directly the effects of climate change and energy-environmental policies in terms of use of resources (energy, water, and land use), agricultural productivity, highlighting the synergies among the different sectors. The land use demand on the whole time horizon represents therefore the “end-use demand” of the energy and material model to be fulfilled at the minimum feasible cost in compliance with the exogenous constraints on resources. The general objective is to ensure an optimal management of the territory, i.e., able to improve the use of endogenous resources, increase the resilience of the agri-food sector to climatic events and facilitate the implementation of agricultural, energy and environmental policies.
The analytical structure reported in the flowchart represented in Figure 1 was therefore designed to represent the agri-food system and characterize its data input.
TIMES-WEF flowchart, inspired by Chiodi, 2016 [ 26 ].
In this modeling approach, the Used Agriculture Area (UAA) and the Forestry Area (FA) represent the output commodities. More precisely, the UAA refers to the total area (hectares) used for agriculture, which includes arable land, permanent meadows, permanent crops, and vegetable gardens used by farms with reference to Eurostat data [ 49 ]. The FA, instead, is representative of the hectares of surface area covered by forests or the canopy of the forest or open woods.
The agricultural and forestry activities are modeled as end-use processes (dummy processes) with associated input and output commodities, operating costs, and other key parameters characterizing these practices. New elements such as water, fertilizers, pesticides, and CO 2 capture from forestry were included among the input commodities of a standard TIMES model (energy vectors and materials). Biomass residuals from agriculture and forestry, greenhouse gas emissions from both the combustion processes and agricultural activities are modeled as process outputs.
The characterization of agricultural activities was defined according to the standard classification of farming classes used in the main European and national databases. Specifically, 10 categories were considered: arable crops, horticulture, viticulture, olive growing, fruit growing, herbivores livestock, granivorous livestock, polyculture, mixed livestock and mixed (that include livestock and crops).
Each agricultural activity was represented by two processes in series: (i) The first process consumes water, energy (electricity, diesel, natural gas), pesticides and/or fertilizers and produces crops (expressed in ton) or cattle (expressed in livestock unit (LSU)); (ii) The second process converts the productions of the first process in hectares of used agricultural/livestock area through a yield parameter. The sum of the outputs of the second processes of all ten categories of agricultural activities provides the demand for end use, i.e., the Used Agriculture Area. The only exception is represented by mixed activities that were modeled as a single process. It has energy and water consumption as input and hectares used as output.
Forests play a multifunctional role, contributing to the protection of biodiversity and the environment (through carbon sequestration) and to the economy (through the production of biomass as an energy resource). They are also particularly affected by climate change (droughts, forest fires, etc.), which reduce their carbon sequestration power and bioenergy resource potential.
The detailed structure of the TIMES-WEF module is described by the flowchart in Figure 2 that shows the flows of commodities through processes, from resource mining to end-use demands (i.e., UAA and FA).
The TIMES-WEF Reference Energy and Materials System (REMS).
The analyzed time horizon covers a period of 50 years, from 2010, the model base year, to 2060 and was divided into time intervals of 5-years each (model’s time slices), considering 2030 as a milestone. This long-term time horizon makes it possible to evaluate the effects of the strategies for achieving the targets of the 2030 Agenda beyond the year 2030 and to trace the path toward the Energy Roadmap 2050 through a scenario analysis. Furthermore, from a modeling point of view, the choice of a time horizon that goes beyond 2030 derives from the need to harmonize the timing of the TIMES-WEF to the TIMES Basilicata energy system module [ 50 ] to allow its integration and increase the reliability of the solutions in a long-term perspective. Therefore, in the future stages, the integration of the TIMES-WEF agri-food model into the whole energy system model, will make possible to have a comprehensive perspective of the synergies and competition between the two sectors in a circular economy perspective.
The TIMES-WEF module data input is made up of a set of an Excel spreadsheet, structured around three type of file:
To test the applicability and consistency of the integrated approach developed through the TIMES-WEF modeling module, the TIMES model was customized according to the Basilicata region data as a pilot case. Basilicata is a small region located in the South of Italy bordering with Campania on the West, Apulia on the North and East, and Calabria on the South. The region covers 10,073 square kilometers and has a population of 562,869 inhabitants [ 51 ] with a quite low regional population density (57.8 inhabitants per km 2 ).
Basilicata is a relevant case study mainly due to the urgent threat that the effects of climate crisis represent for its territory. In fact, 55% of the Region is at risk of desertification [ 52 ] endangering the future of the agricultural sector, which plays a crucial role in the local economy. Therefore, for this Region it is a priority to define policy mitigation and adaptation actions to increase the resilience of the territory and the agriculture sector, identifying sustainable pathways of local resources. Due to the wide morphological difference, mainly in elevation, the Region is characterized by a varied climate that ranges from the continental one in the internal areas to the Mediterranean one of the coastal areas. There are six distinct soil and climatic zonas (Ionian, Bradanica, Northern Apennines, North Western Apennines and South Western Apennines, Tyrrhenian) in which climate deeply influences the type of agriculture, in particular in the internal and non-irrigated areas [ 53 ].
The territory is mainly mountainous (47%) and hilly (45%) with a modest flat percentage (8%). The total agricultural area is 716,838 hectares accounting for about 70% of the regional surface area. The forest area according to the Regional Forest Charter is 355,409 hectares, characterized by variegated species in both environmental and vegetation terms which make the regional territory a mosaic landscape.
As happens in several Mediterranean countries under the pressure of contingent factors (climate change, changes in land use, over-exploitation of resources, etc.), large areas of Basilicata are particularly susceptible to land degradation [ 54 , 55 , 56 ]. This situation highlights the importance of a more sustainable management of resources and forests to improve the resilience of the territory and guarantee the functioning of a key sector such as agriculture [ 57 ]. Therefore, it is a priority for this region to define comprehensive mitigation and adaptation actions that increase the resilience of the territory and to identify sustainable pathways in the use of local resources aimed at improving the agricultural sector. Agriculture is, in fact, one of main activities of the Basilicata economy [ 58 ], along with industry (manufacturing, automotive, and especially oil extraction) and services, where the importance of tourism is increasing. Despite its small contribution compared to other sectors, with an added value of 3% in 2019, the agricultural sector has a significant weight in terms of exports and employment, registering for the latter, a positive trend in recent years reaching an increase of 7% in 2019, in contrast with the other sectors [ 59 ]. Therefore, the high level of specialization achieved in the agri-food sector (characterized by the production of a wide range of high quality food (most of which included in national list of Protected Designation of Origin (PDO), Protected Geographical Indication (PGI), and Controlled and Guaranteed Designation of Origin (DOCG, the Italian acronym) marks of traditional food and wine products) could be a potential strength to increase the competitiveness of the entire regional system. From this perspective, the 2014–2020 Rural Development Program played a key role by providing a valuable financial support (680 M€) to encourage innovation in this sector in order to improve its economic and sustainability environmental performance [ 60 ].
As regard energy resources, the Basilicata region hosts the Europe’s largest onshore oil and gas field, with an annual production of 48,550,554,911 BBOE of oil and 1,493,816,334 Smc of natural gas [ 61 ]. Furthermore, Basilicata achieved remarkable targets in the production of renewable electricity, reaching and encompassing the goals set by the Regional Environmental Energy Plan (PIEAR) for 2020, namely 981 MW of onshore wind (60% of the total renewable capacity), 359 MW of solar-photovoltaic (20%), 50 MW of biomass (15%) and 48 MW of hydroelectric (5%). The regional authorities, through various policy measures, have also encouraged the use of biomass to produce thermal energy. In 2017, 45% of the regional energy requirement (thermal and electric) was met by renewable sources, also in this case anticipating the objective of the national legislation (33%) expected by 2020 [ 62 ].
The information necessary to implement the TIMES-WEF module for the Basilicata case study was collected by elaborating the national and European statistical sources of data (FADN, RICA and ISTAT), the Regional Environmental Energy Plan–PIEAR and other local sources (e.g., irrigation water, fertilizers and agricultural diesel prices for which there are different values depending on the region) ( Table 2 ).
List of data source and type of information provided.
Data Source | Type of Information Provided |
---|---|
ISTAT Agricultural Census [ ] | Hectares of used agricultural area by type of farming; hectares of total agricultural area; annual water consumptions. |
Regional Environmental Energy Plan–PIEAR [ ] | Energy demand of Agriculture (diesel oil, electricity, and natural gas). |
Annual RICA Survey (National source of European Farm Accountancy Data Network (FADN)) [ ] | Annual micro data on surveyed farm: energy consumption (diesel, natural gas, and electricity), use of fertilizers, production (crops and cattle), fixed and variable production costs. |
Local Chambers of Commerce [ , ] | Prices applied on a local scale to fertilizers and agricultural diesel. |
Local Reclamation Consortium [ ] | Agricultural water prices. |
Ministry of Economic Development [ ] | Natural gas prices. |
Energy Service System Operator (GSE) [ ] | Electricity prices. |
The scenario analysis has two main objectives in this research: (I) evaluating the effects of climate change in terms of land use variations and resources availability (water and land); and (II) assessing the consequences of EU, national and regional environmental and agricultural policies on the entire agri-food system of Basilicata region.
A reference Business-as-Usual (BaU) scenario was set to provide a benchmark for comparing the alternative scenarios solutions. It represents the development of the system under the policies in force, both energy (through Regional Environmental Energy Policy Plan-PIEAR), and agricultural. The reference scenario shows the “status quo” evolution of the Basilicata Region agricultural system in term of land use availability, resources, technologies, and policy in place. The land use demand by category is projected along the analyzed time horizon (2010–2060) using appropriate statistical techniques.
The selection of scenarios considers the phenomena in progress and the most urgent challenges to be faced in the Mediterranean Europe region for mitigation and adaptation to climate change by focusing on the agri-food sector.
In fact, this area is significantly affected by climate change, especially agriculture. The intensity and frequency of extreme weather event (in particular, heat waves, flooding, wildfires) accelerate the degradation of agricultural land and cause a substantial decrease in yields, with an estimated average loss of 3.24 ton/ha per year compared to 2010 (reference year) [ 71 ]. As for the consequences in terms of WEF Nexus, droughts will favor a greater demand for water, with an increase predicted by the Representative Concentration Pathways (RCP) 4.5 and 8.5 climate scenarios between 4% and 18% by the end of the century. This will inevitably reduce the availability of water for irrigation, undermining the suitability of the land for rain crop production [ 72 ].
Indeed, through more sustainable agricultural and forestry practices it is possible to reduce the environmental impact of this sector, strengthen the carbon capture capacity of soils and forests and protect biodiversity.
From this perspective, the agriculture and forestry sector has been included for the first time in the greenhouse gas emission reduction targets set by the European Union (EU) for 2030 [ 73 ] and has been at the center of two fundamental strategies of the Green Deal, “Farm to fork” and “Biodiversity 2030” [ 74 ].
As highlighted by Nikolakopoulou [ 75 ] “food related targets run throughout the Sustainable Development Goals (SDG) and they are often interconnected”. In particular, a sustainable agriculture system, in line with the SDGs, should be more resilient to climate risks, protect the environment and deliver healthy and affordable products to fulfill the demand of an increasing population. The new European Green Deal package of measures will have a strong impact on future planning, reinforcing the orientation already taken by the Common Agricultural Policy (CAP), aimed at favoring a progressive abandonment of intensive agriculture in favor of more sustainable cultivation techniques that preserve the soil quality and fertility by reducing the use of fertilizers and pesticides.
Taking into account the complex environmental and policy framework, two alternative classes of scenarios have been defined as follows:
The climate scenarios parameters have been selected coherently with the IPCC 2019 report [ 16 ]. Moreover, data of water availability and consumption was taken from the Inter-Sectoral Impact Model Intercomparison Project (ISI-MIP) [ 76 ], for the RCP 4.5 and RCP 8.5 climate scenarios.
The scenario analysis will identify sustainable development paths for the agricultural system taking into account the general objectives of the EU Green Deal and the “From Farm to Fork” strategy. In this framework, the contribution of the various alternatives to achieving the SDGs relating to energy, climate and sustainable production and consumption models will be assessed (namely SDG 7 “Clean and accessible energy”, SDG12: “Responsible consumption and production” and SDG 13 “Climate action” [ 77 ]).
3.1. data input construction.
Statistical data pre-processing is a fundamental step for the implementation of the model’s data input.
Following the specifications of the TIMES-WEF model, the agricultural system has been characterized on the basis of the agricultural area (driving parameter), dividing, for the reference year, the total agricultural area in the region (502,197 hectares) into the shares of the various types of farming that characterize the local agricultural system ( Figure 3 ).
Share of land use by type of farming in Basilicata region.
The estimations were based on the data of the Agriculture Census for 2010 [ 63 ], the latest available survey of the National Statistical Institute, integrated with the information of the annual sample survey of Agricultural Accounting Information Network for Italy (RICA) [ 65 ], included in the European Network of Farm Accounting Data (FADN) [ 78 ].
The FADN dataset represents the main information source of the European Commission on the Member States to assess the impact of the proposed changes to the CAP through the simulation of different scenarios on corporate sustainability (economic, environmental, social and innovations).
The RICA data were therefore preliminarily processed to represent the agricultural system of the Basilicata region in relation to the 10 categories considered in the model. Further elaborations were also required to model in detail the features of specific sub-categories.
After determining the land use demand by type of crop, the subsequent step was to estimate the energy consumption by type of farming in order to determine the energy balance. This was done considering the Basilicata Regional Environmental Energy Policy Plan (PIEAR) data for Agriculture ( Table 3 ), which was disaggregated through a weighting procedure based on RICA information on the annual energy expenditures for diesel, natural gas and electricity incurred by local farms.
Energy Consumption in Agriculture for the Basilicata Region (PIEAR data).
Energy Fuel | PJ |
---|---|
Electricity | 0.227 |
Natural gas | 0.037 |
Diesel | 1.741 |
In addition, a further disaggregation was necessary for permanent crops, represented in the TIMES-WEF model by the following categories: viticulture, fruit and olive growing. The energy consumption of each category was estimated by a weighting procedure based on the used agricultural area. Table 4 summarizes the percentage breakdown of the estimated energy consumption for each type of farming.
Percentage breakdown of energy consumption.
Type of Farming | Diesel (PJ) | Natural Gas | Electricity |
---|---|---|---|
Arable Crops | 46.0% | 58.2% | 30.5% |
Horticulture | 2.6% | 0.3% | 3.2% |
Viticulture-PermCrops_1 | 2.9% | 6.0% | 1.9% |
Fruit growing-PermCrops_2 | 6.9% | 14.2% | 4.6% |
Olive growing-PermCrops_3 | 5.5% | 11.4% | 3.7% |
Herbivores Livestock | 10.7% | 9.6% | 18.6% |
Granivorous Livestock | 0.3% | 0% | 0.6% |
Polyculture | 14.1% | 0.4% | 9.5% |
Mixed Livestock | 3.5% | 0% | 8.0% |
Mixed | 7.5% | 0% | 19.3% |
The costs of the energy carriers for the reference year were estimated considering different sources of data. In particular, the electricity price (59 Euro/MWh, to 16.4 MEuro/PJ) was estimated on the basis of the National Energy Services Operator data [ 70 ], diesel price for agriculture (0.60 Euro/liter, 16.2 MEuro/PJ) using the local Chambers of Commerce data [ 66 ] and natural gas price (0.66 Euro/liter, 25.3 MEuro/PJ) using the Ministry of Economic Development data [ 69 ].
Similarly, the share of water consumption for the ten agricultural activities was calculated using the aggregate data provided by the National Institute of Statistics, assuming a sales price of 0.47 euro per cubic meter [ 68 ]. The estimated values by type of crop are shown in Table 5 .
Water consumption by type of farming.
Type of Farming | M /ton | M /LSU | M /ha |
---|---|---|---|
Arable Crops | 15 | ||
Horticulture | 40 | ||
Viticulture-PermCrops_1 | 48 | ||
Fruit growing-PermCrops_2 | 272 | ||
Olive growing-PermCrops_3 | 172 | ||
Herbivores Livestock | 70 | ||
Granivores-Livestock | 0 | ||
Polyculture | 156 | ||
Mixed Livestock | 436 | ||
Mixed | 144 |
The three most important fertilizers used in agricultural practices, potassium (K), phosphorus (P), and nitrogen (N), were also included in the model data input, estimating the tons of fertilizers used in the base year 2010 [ 65 ] starting from the hectares of agricultural area for each type of crop and the tons of product ( Table 6 and Table 7 ).
Consumption of fertilizers per hectare by type of crop.
Type of Farming | N (ton/ha) | P (ton/ha) | K (ton/ha) |
---|---|---|---|
Arable Crops | 0.00157 | 0.04424 | 0.00510 |
Horticulture | 0.01340 | 0.04316 | 0.01210 |
Viticulture-PermCrops_1 | 0.03110 | 0.05621 | 0.02763 |
Fruit growing-PermCrops_2 | 0.03351 | 0.06969 | 0.02525 |
Olive growing-PermCrops_3 | 0.03047 | 0.03175 | 0.02323 |
Polyculture | 0.01900 | 0.05988 | 0.02074 |
Consumption of fertilizers per ton of product and by type of crop.
Type of Farming | N (ton of N/ton of crop) | P (ton of P/ton of crop) | K (ton of K/ton of crop) |
---|---|---|---|
Arable Crops | 0.0005 | 0.0153 | 0.0018 |
Horticulture | 0.0236 | 0.0760 | 0.0213 |
Viticulture-PermCrops_1 | 0.2219 | 0.4011 | 0.1971 |
Fruit growing-PermCrops_2 | 0.0506 | 0.1052 | 0.0381 |
Olive growing-PermCrops_3 | 0.1471 | 0.1533 | 0.1121 |
Polyculture | 0.0227 | 0.0714 | 0.0247 |
The average costs of fertilizers were estimated based on the Turin Chamber of Commerce data [ 67 ] assuming a constant value at national level. Specifically, 359 Euro/ton for Nitrogen, 328 Euro/ton for Phosphorus and 498 Euro/ton for Potassium were considered.
Once characterized the input commodities (diesel, natural gas, electricity, water, and fertilizers), it was necessary to characterize the production processes as fictitious technologies (dummy processes) estimating the operating costs, since the technical parameters are not of interest in this approach.
The agricultural production (expressed in tons) and the number of cattle (expressed in in livestock unit-LSU) for 2010 for each type of agricultural activity were estimated by the RICA database. Furthermore, the fixed and variable operating costs incurred by the farms were estimated considering their annual budgets ( Table 8 ).
Production and operating costs of agricultural activities.
Type of Farming | Unit of Measure | Production | Fixed Costs (Euro/ton or Euro/LSU) | Variable Costs (Euro/ton or Euro/LSU) |
---|---|---|---|---|
Arable Crops | ton | 749,387 | 34 | 89 |
Horticulture | ton | 146,729 | 55 | 930 |
Viticulture-PermCrops_1 | ton | 36,229 | 92 | 206 |
Fruit growing-PermCrops_2 | ton | 171,283 | 88 | 155 |
Olive growing-PermCrops_3 | ton | 53,555 | 250 | 545 |
Herbivores Livestock | LSU | 127,693 | 323 | 589 |
Granivorous Livestock | LSU | 20,096 | 255 | 699 |
Polyculture | ton | 21,6746 | 61 | 168 |
Mixed Livestock | LSU | 7691 | 277 | 1267 |
Another aspect of the setup of the model’s data input concerned waste production that was estimated considering the statistical data relating to the production of straws and manure by crops and livestock type [ 79 ], harmonizing these data to the categories of the TIMES–WEF model ( Table 9 ).
Waste production by crops and livestock type.
Type of Farming | Straw Production | Manure Production | |
---|---|---|---|
Ton/ha | Ton/ton of Production | Ton/LSU | |
Arable Crops | 1.99 | 0.6 | |
Horticulture | 0 | 0 | |
Viticulture-PermCrops_1 | 2.15 | 0.65 | |
Fruit growing-PermCrops_2 | 2.20 | 0.35 | |
Olive growing-PermCrops_3 | 2.16 | 0.87 | |
Herbivores Livestock | 0.53 | ||
Granivores-Livestock | 0.03 | ||
Polyculture | 2.13 | 0.26 | |
Mixed Livestock | 0.1 |
In addition, the emissions of the main greenhouse gases (CO 2 , CH 4 and N 2 O) were estimated by considering the emission factors provided by the United Nations Framework Convention on Climate Change (UNFCCC) database [ 80 ]. In the base year calibration of the β version of the TIMES-WEF model, only the energy combustion processes were initially considered. The emissions from processes will be added subsequently.
Once the data input for the base year has been implemented, the subsequent fundamental step concerns model calibration to statistics [ 81 ]. This is essential to validate the modeling approach and to refine the initial data. The main results of the TIMES-WEF calibration are shown in this section. Figure 4 shows the energy consumption by type of farming. Diesel oil is the most used fuel (1.74 PJ), (in particular from arable crops (0.8 PJ), polyculture (0.25 PJ) and herbivores livestock (0.19 PJ)) followed by electricity (0.23 PJ) and natural gas (0.04 PJ). Granivorous livestock, mixed livestock and mixed farming categories do not use natural gas.
Energy consumption for each type of farming. Base year (2010).
Figure 5 shows water consumption of the 10 agricultural activities. As expected, fruit growing and polyculture have the highest consumption, 47 Mm 3 and 34 Mm 3 respectively, as they are typical “water demanding” categories. Arable crops consumption is about 11 Mm 3 , while granivorous livestock consumption is negligible.
Water consumption by farm typology. Base year (2010).
As concern the use of fertilizers ( Figure 6 ), fruit growing shows the highest consumption of both Phosphorus (18.02 kton) and Nitrogen (8.66 kton), while viticulture shows the highest consumption of Potassium (14.53 kton). Arable crops, which represent an energy-intensive category covering the highest percentage of territory, show a high consumption of Phosphorous and moderate consumption of Nitrogen (0.41 kton) and Potassium (1.32 kton). On the other hand, olive growing shows an almost similar consumption for all three fertilizers (7.88 kton for Nitrogen, 8.21 kton for Phosphorus and 6.01 kton Potassium).
Fertilizers consumption for each type farm. Base year (2010).
As regards waste production ( Figure 7 ) arable crops provide the highest values of dry matter (442.46 kton). Herbivorous livestock (64.66 kton of manure), fruit growing (59.27 kton of dry matter), polyculture (56.37 kton of dry matter) and olive growing (46.59 kton of dry matter) mixed livestock and granivorous produce a very low amount of waste (respectively 0.22 kton and 0.06 kton of manure) also produce small quantities while waste production from horticulture is negligible.
Waste production by farm category. Base year (2010).
In the calibration runs, CO 2eq emissions from combustion processes are entirely determined by fossil fuel consumption for agricultural activities, with a total amount of CO 2eq of 131 ton for the base year, the highest contribution being provided by arable crops (61 ton of CO 2eq ) while the lowest by granivorous livestock (0.3 ton of CO 2eq ).
Figure 8 provides an overview of the distribution of the different commodities (energy consumption, water consumption, fertilizers consumption, waste production and CO 2eq emission production) by farming activity.
( a ) Consumption and emissions distribution by type of agricultural activity ( b ) Weight of agricultural activities in resource consumption and emission production.
In the two polar maps of Figure 8 , energy and fertilizer consumption were aggregated. Arable crops represent the category with the highest energy consumption (44%), associated CO 2eq emissions and waste production (64%). Fruit growing has the greatest weight in both water (37%) and fertilizer consumption (23%). Polyculture is the second category for water consumption (27%) and viticulture for fertilizer consumption (21%). Granivorous show the lowest energy consumption (0.3%), associated CO 2eq emissions (0.3%) and waste production (0.1%).
The calibration of the β version of the TIMES-WEF model highlights the differences between the various agricultural activities in terms of resources used and output, confirming the validity of the modeling approach and preparing a solid reference structure for the implementation of the scenarios.
This paper presents the fundamental steps for the implementation of a modeling platform that integrates the water-energy-food nexus approach into the energy system modeling framework, with the aim to investigate the impacts of climate change and EU policies on the agricultural system.
In fact, agriculture plays a crucial role in achieving sustainable development goals and climate targets. The rational use of soil, water and energy is essential to ensure the well-being of the population and adequate food production, while the adoption of sustainable agricultural practices can help mitigating the effects of climate crisis and supporting adaptation to extreme weather events.
The Nexus approach, increasingly used for an integrated vision of the main challenges of sustainability, allows highlighting the interdependencies between the three key variables Water-Energy-Food and to understand how they are influenced by climate change and the policies in place. The study goes beyond the state of the art [ 26 ] by designing, implementing, and validating a modeling approach based on energy system analysis, widely used in policy assessment.
A broad scientific debate is underway about the choice of modeling tools and indicators for defining policy strategies and measuring progress toward the achievement of sustainable goals. In fact, the selection of indicators can deeply influence the results of the monitoring [ 82 ] and the implementation of SDGs, requiring broader qualitative analyses [ 83 ].
In this context, the use of a modeling framework based on the IEA-ETSAP methodology, explicitly designed for long-term energy-environment analyses, to design the least-cost pathways for a sustainable development, allows ensuring transparency in the basic assumptions and a high detail in the identification of the possible strategies.
The integration of the nexus approach in a comprehensive partial equilibrium model based on the ETSAP-TIMES structure makes possible to set up a robust platform to identify the optimal allocation of energy and material resources in compliance with the EU strategic policy targets and to explore possible alternatives, measuring their effectiveness in terms of economic, energy and environmental indicators.
An innovative modeling approach based on land use as driving variable was adopted to develop the TIMES-WEF model whose database includes non-energy resources (water, fertilizers, pesticides) among the input commodities while food and biomass residuals represent the output commodities. The β version of the model was customized and calibrated on the agriculture system of the Basilicata Region, to enable its straightforward integration with the former TIMES Basilicata energy model [ 50 ] in a circular economy perspective.
Starting from the official European Union Classification [ 84 ], ten end-use categories representing the entire agri-food system were identified (arable crops, horticulture, viticulture, fruit growing, growing olive growing, polyculture, livestock herbivorous, granivorous livestock, mixed and mixed livestock) and the model database was set up accordingly.
The modeling approach was validated by calibrating the model β version and evaluating the congruence of the results to the statistics.
The next steps will concern, first, the optimization of the model under the Reference Scenario, representing a “status quo” development along the time horizon and, second, a scenario analysis to identify suited roadmaps for a sustainable development of the regional agricultural system.
The proposed innovative modeling framework applied to a regional energy and agriculture system can contribute to supporting decision makers in a complex governance of a system in which conflicting objectives must be faced. In particular, a thorough analysis of the water-energy-food system will allow enhancing the role of the agriculture and forestry sectors to achieve the national CO 2eq reduction and RES targets, evaluating the effectiveness of different mitigation strategies.
At the same time, the assessment of the CAP at the local scale will provide insights to coordinate the implementation of policies at local and national level with a long-term perspective.
The authors are grateful to anonymous reviewers for their helpful comments and suggestions that resulted in an improved quality of the manuscript and gratefully acknowledge the Portuguese Foundation for Science and Technology (FCT) for the support provided to CENSE through the strategic projects UIDB/04085/2020.
Conceptualization, M.M.T., S.D.L., C.C., P.F., M.V., M.C., F.P., M.S. and S.R.; methodology, M.M.T., S.D.L., C.C., P.F., M.V., M.C., F.P. and M.S., software, M.M.T. and S.D.L.; validation, M.M.T. and S.D.L.; formal analysis, M.M.T., S.D.L., C.C.; data curation, M.M.T., S.D.L., P.F. and M.V.; writing—original draft preparation, M.M.T., S.D.L. and C.C.; writing—review and editing, M.M.T., S.D.L., C.C., P.F., M.V., M.C., F.P., M.S. and S.R.; supervision, C.C., P.F., M.V. and S.R.; project administration, S.R.; funding acquisition S.R. All authors have read and agreed to the published version of the manuscript.
This research was funded by the regulation “Dottorati Innovativi con Specializzazione in Tecnologie Abilitanti in Industria 4.0”, funded by the Basilicata Region-Industry 4.0 Strategy, period 2017–2020.
The authors declare no conflict of interest.
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Vasilis dakos, chris a. boulton, joshua e. buxton, jesse f. abrams, beatriz arellano-nava, david i. armstrong mckay, sebastian bathiany, lana blaschke, niklas boers, daniel dylewsky, carlos lópez-martínez, isobel parry, paul ritchie, bregje van der bolt, larissa van der laan, els weinans.
Tipping points characterize the situation when a system experiences abrupt, rapid, and sometimes irreversible changes in response to only a gradual change in environmental conditions. Given that such events are in most cases undesirable, numerous approaches have been proposed to identify if a system is approaching a tipping point. Such approaches have been termed early warning signals and represent a set of methods for identifying statistical changes in the underlying behaviour of a system across time or space that would be indicative of an approaching tipping point. Although the idea of early warnings for a class of tipping points is not new, in the last 2 decades, the topic has generated an enormous amount of interest, mainly theoretical. At the same time, the unprecedented amount of data originating from remote sensing systems, field measurements, surveys, and simulated data, coupled with innovative models and cutting-edge computing, has made possible the development of a multitude of tools and approaches for detecting tipping points in a variety of scientific fields. However, we miss a complete picture of where, how, and which early warnings have been used so far in real-world case studies. Here we review the literature of the last 20 years to show how the use of these indicators has spread from ecology and climate to many other disciplines. We document what metrics have been used; their success; and the field, system, and tipping points involved. We find that, despite acknowledged limitations and challenges, in the majority of the case studies we reviewed, the performance of most early warnings was positive in detecting tipping points. Overall, the generality of the approaches employed – the fact that most early warnings can in theory be observed in many dynamical systems – explains the continuous multitude and diversification in their application across scientific domains.
Dakos, V., Boulton, C. A., Buxton, J. E., Abrams, J. F., Arellano-Nava, B., Armstrong McKay, D. I., Bathiany, S., Blaschke, L., Boers, N., Dylewsky, D., López-Martínez, C., Parry, I., Ritchie, P., van der Bolt, B., van der Laan, L., Weinans, E., and Kéfi, S.: Tipping point detection and early warnings in climate, ecological, and human systems, Earth Syst. Dynam., 15, 1117–1135, https://doi.org/10.5194/esd-15-1117-2024, 2024.
Tipping points characterize a situation when a system experiences abrupt, rapid, and sometimes irreversible changes. Such shifts occur when a threshold is crossed and the system transitions from its current state to a contrasting one (van Nes et al., 2016). Given that tipping points are associated with abrupt, rapid, and sometimes irreversible changes, numerous approaches have been proposed to identify if a system is getting closer to such a point. These approaches are often referred to as early warning signals (EWSs), and they represent a powerful generic tool for anticipating tipping points in a variety of systems (Scheffer et al., 2009). The general mechanism behind EWSs is that, as a dynamical system approaches a tipping point, it becomes slower at recovering from small perturbations (Wissel, 1984), and this critical slowing down (CSD) of the system leaves signatures in the temporal and spatial dynamics of the system (Drake et al., 2020). EWSs rely on identifying exactly such changes in the underlying behaviour of a system across time or space prior to a tipping point.
Table 1 Available software tools for the estimation of early warnings with temporal and spatial datasets.
Download Print Version | Download XLSX
Early, after their introduction in the literature, it became clear that EWSs did not allow the anticipation of all types of tipping points in advance (Hastings and Wysham, 2010) and that they are not unique to tipping point responses but also occur when systems are undergoing smoother transitions (Kéfi et al., 2013). These realizations imply that some shifts (typically referred to as abrupt shifts or regime shifts) may require alternative or additional signals (Boettiger et al., 2013; Dakos et al., 2015). Thus, a rich research programme has been triggered in the theory behind tipping point anticipation and the development of tools (Table 1 ) for quantifying changes in dynamical patterns of system behaviours that could be used as early warnings preceding tipping points and abrupt shifts in general. Different terms have been used to describe the great variety of metrics proposed in the literature, like “early warning systems” (Lenton, 2013b), “observation-based early warning signals” (Boers, 2021), “statistical stability indicators” (Bathiany et al., 2016), “critical slowing down (CSD) indicators” (Tang et al., 2022), “leading indicators” (Carpenter et al., 2008), “resilience indicators” (Dakos et al., 2015), “generic indicators” (Scheffer et al., 2015), “dynamical indicators of resilience” (DiOR) (Scheffer et al., 2018), “indicators of transitions” (Clements and Ozgul, 2018), and “universal early warning signals” (Dylewsky et al., 2023). In the rest of the paper, we will use the term “early warnings” to refer to this whole family of indicators.
Whatever the term used, while early warnings are well grounded in theory, the challenge remains to apply them to real-world systems. A number of review and synthesis papers have summarized the theoretical aspects of early warnings and provided partial accounts of their empirical applications (Alberto et al., 2021; Bestelmeyer et al., 2011; Dakos and Kéfi, 2022; Lenton, 2011, 2013b; Litzow and Hunsicker, 2016; Nijp et al., 2019; Scheffer et al., 2012a, 2015). However, although the utility of early warnings has led to early warnings proliferating beyond ecology and climate and being applied across a variety of scientific domains, we miss a complete picture of where, how, and which early warnings have been used so far in real-world case studies.
Here, after summarizing the basics of the theory underlying early warnings and giving an overview of their taxonomy, we review the literature for the use of early warnings in empirical studies across all scientific fields. We document what metrics have been used; their success; and the field, system, and tipping points involved. We then classify this information in order to provide an overview of the progress, limitations, and opportunities in the empirical application of early warnings after 15 years of research on the topic.
The theory behind tipping point anticipation is mostly based on destabilizing stable fixed equilibrium points. In such cases, there are three ways that a tipping point may theoretically occur (Lenton, 2013a). A system may undergo a bifurcation when a parameter (or multiple parameters) in the system changes beyond a critical threshold and the stability of the state the system occupies is lost, thus causing the system to shift to an alternative state (bifurcation tipping or B-tipping). Noise-induced tipping can occur when a system is shifted outside its stable basin of attraction by some form of stochastic forcing (N-tipping). A third class, known as rate-induced tipping (R-tipping), occurs when a parameter rapidly changes and the system is no longer able to track its stable state (Ashwin et al., 2011). Tipping points also occur through phase transitions, a long-studied set of emergent phenomena in physics which resemble the characteristics of the B-tipping described above (Hagstrom and Levin, 2021; Sole et al., 1996).
The majority of the early warnings discussed below are primarily developed to detect cases where there is a gradual approach towards a bifurcation-tipping event causing a loss of system state stability. Rate-induced tipping could also show early warning (Ritchie and Sieber, 2016). Noise-induced tipping is likely to occur unpredictably; therefore early warnings are less expected. In a realistic scenario with constant stochasticity and conditions gradually changing, tipping is commonly a combination of a movement towards bifurcation and noise pushing the system before the bifurcation actually occurs. In such cases, noise-induced tipping becomes more likely, as it is easier for the system to leave its current basin of attraction when it is closer to the bifurcation, and this increase in the probability of tipping can be identified through particular early warnings (Sect. 2.2 ).
Although most of the theory behind early warning signals is related to saddle-node (or fold) bifurcations, other types of bifurcations have also been considered, like transcritical, pitchfork, or Hopf bifurcations (more general codimension-1 bifurcations; Kuznetsov, 1995). Such bifurcations are smooth (also called continuous) in contrast to the abrupt (i.e. discontinuous) fold bifurcations associated with tipping points, yet it has been shown that similar early warning signals can be applied for them (Boettiger et al., 2013; Kéfi et al., 2013). A full list of bifurcation types (discontinuous and continuous) and their relationship to CSD can be found in Thompson and Sieber (2011). In this paper, the early warnings considered are mostly developed in the context of the discontinuous fold bifurcation.
Table 2 A taxonomy of early warnings depending on whether or not the warning is based on critical slowing down (CSD). CSD-based early warnings are mostly associated with bifurcation tipping (B-tipping), while non-CSD-based ones are associated both with B-tipping and noise-induced tipping (N-tipping; see also Sect. 2.1 ). A second dichotomy is based on whether the warning is a statistical metric based on the dynamical patterns of the system or whether it is based on a process model that is as simple as possible. In parentheses is the type of data (temporal and/or spatial) used to estimate the early warning.
We hereafter present a representative (but not complete) overview of the early warnings mostly used, both theoretically and empirically. These signals can be classified in different ways depending, for instance, on the type of mechanism or tipping point (e.g. CSD-based, non-CSD-based), the type of data used (e.g. temporal, spatial, trait, abundance data), and the approach employed (e.g. analysing patterns, fitting models, network methods). In Table 2 , we suggest a taxonomy of early warnings based on the mechanism and the approach used. We then present their basics without going into the details. A full description and the methods to estimate them can be found elsewhere (Clements and Ozgul, 2018; Dakos et al., 2012; Génin et al., 2018; Kéfi et al., 2014; Lenton, 2011; Scheffer et al., 2015) and in dedicated software packages (Table 1 ).
The early warnings most used are based on searching for evidence of “critical slowing down” (CSD) in the system. Essentially, as the system is forced towards a tipping point, the state it currently occupies starts to lose its stability, and the restoring feedbacks that “pull” the system back to that state after it is perturbed start to degrade. This causes the system to respond more sluggishly to these perturbations and thus slow down (Wissel, 1984). Figure 1 shows this concept visually using the “ball in the well” analogy. When the system is more stable, represented by the well with steeper sides, the recovery is faster as the ball (representing the state of the system) returns faster. A system close to tipping, represented by a shallower well, has a slower recovery as the ball takes longer to return. Eventually, the restoring feedbacks of the system may become so weak that the stability of the current state may be lost, and the system may transition to a new stable state. Mathematically, CSD occurs as the leading eigenvalue of the system approaches 0 from below. However, in reality, we do not have the equations that govern the system's dynamics, and, as such, we have to estimate the occurrence of CSD with methods that aim to infer CSD mostly from the patterns of the system dynamics or by fitting very simple and generic process-based models (Table 2 ).
Figure 1 Using the “ball in the well” analogy to compare a system that is far from tipping (left) and a system that is close to tipping (right). The system that is further away from tipping recovers faster from perturbations, with the steeper sides of the well describing the stronger restoring feedbacks of the system. Closer to tipping, the sides of the well are shallower, such that the system will take longer to return from the same perturbation because the restoring feedbacks are weaker.
Using statistical techniques makes it possible to detect CSD based on the dynamical patterns a system is generating. The most direct way to detect CSD is to consider the rate at which a system returns to its initial state following a perturbation (return rate or return time). A resilient system with strong restoring feedbacks will return to its initial state faster than one which is near to a tipping point (Wissel, 1984). However, this method requires the occurrence of a well-defined perturbation and clear knowledge of when the equilibrium state of the system has been reached, neither of which are always clearly defined in the real world.
Figure 2 A comparison of the lag-1 autocorrelation (AR(1)) for a system that is far from tipping (blue), getting close to tipping (purple), and close to tipping (red). As the time series approaches tipping (top row), there is no correlation between subsequent values of the time series in the blue part of the time series (far from tipping). However, closer to tipping, in the purple and then red regions of the time series, there are correlations and thus higher AR(1) values. In the time series itself there are clear deviations towards the end compared to the beginning, suggesting that CSD occurs as the tipping point approaches. The early warnings are calculated on a moving window (coloured regions in the bottom plot). Here, AR(1) is shown at the end of the window used to calculate it, with examples shown as coloured points to match those windows on the detrended time series.
As the system approaches a tipping point and its recovery slows down, each time step X ( t ) is more correlated to the previous time step X ( t −1) (as shown in Fig. 2 ). This can be measured with lag-1 autocorrelation, or AR(1), which tends towards 1 as a system experiences CSD prior to tipping to an alternate state. Visually, this can be viewed by observing a scatterplot of a time series of the system against the time series lagged by 1 time point (Fig. 2 ). When the system is far from tipping (top row of Fig. 2 ), there is no relationship between the current state and the state at the previous point in time (low AR(1)). As the system approaches the tipping point, CSD means that there is a strong correlation between the system state now and at the previous point in time (and thus a higher AR(1)). Larger deviations in the red section of the time series can be seen, further showing this slowing down and increase in AR(1).
Similarly, as the system struggles to return to its initial state as resilience is lost, the variance of the system is also expected to increase, as the system can sample more of the “state space” (all the possible states the system can be in) due to the shallower well. However, this is often recorded alongside an increase in AR(1) because other factors can lead to a change in variance, such as how the system is forced externally.
Spatial analogues of the temporal variance and temporal AR(1) exist, too, with a similar underlying theory to the one for the temporal ones: as a system approaches a tipping point and responds more sluggishly to external perturbations and samples more of the state space, it is expected that there will be a higher spatial autocorrelation (Dakos et al., 2010) and spatial variance (Guttal and Jayaprakash, 2009).
Just like AR(1) and variance, all other CSD-based early warnings aim at detecting characteristic changes in the dynamical patterns of the system either by directly estimating a statistical property (e.g. spectral reddening) or by fitting a statistical model (e.g. detrended fluctuation analysis) (Table 2 ). A parallel approach involves more complex methods to predict the movement towards tipping points that involve the use of simple process-based models. One such example is that of using a generalized model that integrates knowledge about the system into a model, which may allow us to estimate changes in the leading eigenvalue of the system once minimal model assumptions have been made (Lade and Gross, 2012).
CSD-based early warnings rely on the assumption that the system state shows only small deviations around the equilibrium state of the system. However, this assumption does not hold in the presence of strong stochasticity. In other cases, either CSD is hard to measure or more idiosyncratic metrics have been suggested to act as alternatives to CSD-based warnings. Below, we outline a few of the most representative non-CSD-based early warnings (Table 2 ).
As the current equilibrium state of the system is losing resilience and the probability to shift to an alternative equilibrium increases, the temporal distribution of states of the system is expected to become increasingly skewed toward the alternative state. This can be quantified by the skewness of the system. The skewness may increase or decrease, depending on whether the alternative equilibrium is larger or smaller than the current equilibrium (Guttal and Jayaprakash, 2008). Similarly to the change in skewness observed between the two states with temporal data, it is also possible to observe this change in skewness in the spatial domain (Guttal and Jayaprakash, 2009).
“Flickering” is the situation where strong stochasticity can “push” a system temporarily into the basin of attraction of the alternative state before returning to the current state with increasing likelihood as the system is approaching tipping (Dakos et al., 2013; Wang et al., 2012). Flickering can be measured either by a simple increase in variance (Dakos et al., 2013) or by more complex statistical approaches, e.g. quickest detection method (Carpenter et al., 2014) and heteroscedasticity (Seekell et al., 2012; Seekell and Dakos, 2015).
Information about a system at multiple sampling points through time or multiple locations across space can allow us to reconstruct a “stability landscape” of the system – or potential, which gives an idea of the most frequent states of the systems observed in systems experiencing different environmental conditions and history (Livina et al., 2010). Multimodality in such a landscape for a given set of environmental conditions suggests that the system could exhibit alternative stable states for that range of conditions (Abis and Brovkin, 2019; Hirota et al., 2011; Scheffer et al., 2012b; Staver et al., 2011), although seasonality patterns should be accounted for to reduce misinterpretation of externally forced “states”.
A number of ecosystems have a clear spatial structure which is self-organized (e.g. drylands, peatlands, salt marshes, mussel beds; Rietkerk et al., 2008). Theoretical models have shown that the size and shape of the spatial patterns change in a consistent way along stress gradients, and, as such, they are good indicators of ecosystem degradation (von Hardenberg et al., 2001; Kéfi et al., 2007; Rietkerk et al., 2004). Probably one of the most studied examples is the case of dryland ecosystems, where changes in the shape of regular patterns (Rietkerk et al., 2004) and in the patch size distribution (Kéfi et al., 2007) could inform us about the stress experienced by the ecosystem (Dakos et al., 2011).
An alternative approach to pattern-based early warnings is based on fitting process-based models on the time series of a system prior to a tipping point. This approach mainly consists of fitting the simplest dynamical model with a tipping point (i.e. a saddle-node normal form) (Ditlevsen and Ditlevsen, 2023) and testing its likelihood compared to a model without a tipping point (Boettiger and Hastings, 2012b) or of fitting threshold models assuming simple autoregressive state-space models (Ives and Dakos, 2012; Laitinen et al., 2021).
A novel way to detect tipping points involves monitoring structural change properties (e.g. connectivity, node centrality) in network systems (i.e. a network of interacting components) like spatially connected sites, interacting actors, or species in a community (Cavaliere et al., 2016; Mayfield et al., 2020; Yin et al., 2016). Alternatively, the temporal correlation between components in multivariate systems has been used to construct an interaction network and analyse its structural properties (Tirabassi et al., 2014).
Another idiosyncratic approach involves monitoring changes in the statistical moments of fitness-related traits (e.g. body size) (Clements and Ozgul, 2018). Such trait changes have been found in populations under stress where there are changes in the traits of individuals (i.e. decreasing mean and increasing variance in body size) (Clements and Ozgul, 2016; Spanbauer et al., 2016). These trait-based and the above-mentioned structural-based signals are case-specific and idiosyncratic to the details of the system, as there is no universal mechanism that would generate an expected pattern related to the approach of tipping points.
We performed a (non-exhaustive) literature review on the empirical (not theoretical) use of warning signals. We first did a topic search (TS) that included title, abstract, and keywords in the Web of Science for the period from 1 January 2004 to 1 April 2023 with the following terms. TS = ((“tipping point ∗ ” or “tipping” or “catastrophic bifurcation ∗ ” or “catastrophic shift ∗ ” or “regime shift ∗ ” or “abrupt shift ∗ ” or “critical transition ∗ ”) and (“early warning ∗ ” or “early warning ∗ ” or “warning sign ∗ ” OR “resilience indicator ∗ ” or “leading indicator ∗ ” OR “precursor ∗ ”)). We selected the year 2004 as the starting date of our search, despite the fact that CSD was known in ecology much earlier (Wissel, 1984) and that signatures of catastrophic bifurcations were theoretically described for dynamical systems (Gilmore, 1981). Our choice was driven by the fact that 2004 is around the time of the first studies in climate (Held and Kleinen, 2004; Kleinen et al., 2003) and ecology (Carpenter and Brock, 2006) where the theoretical idea of using CSD as a warning signal emerged, while a few years later the first review on early warnings on critical transitions was published (Scheffer et al., 2009). Within this time period, our topic search returned 887 unique publications. For completeness, we also ran the same topic search before 2004 going back to 1960, and we retrieved 11 publications of which only 1 was related to bifurcations. Clearly, we might have missed relevant records with the TS we selected. For example, had we also included the term “phase transition ∗ ”, we would have retrieved 3916 records. We decided not to include this term, as it pertains to a specific and rich field of physics, but with our TS we are confident to have a rather complete overview of the tipping point (and related terms) literature.
We screened all 887 publications to select only the ones where there was an empirical application of early warnings (i.e. an indicator was measured on real data to signal the occurrence of a tipping point). This screening led to 229 papers that we classified as ones that included at least one empirical application of early warnings. For each paper, we collected the following information: “domain” (e.g. climate, ecology), “system” (e.g. Arctic sea ice, fisheries, mental depression), the “tipping point” described, data source (e.g. lab experiment, field survey, remote-sensed datasets, social data), “data type” (i.e. temporal, spatial, spatiotemporal), “indicator” (i.e. the specific warning signal(s) used), and “performance” (whether the performance of the early warning was reported in the paper as positive, negative, mixed (in the case of multiple signals used or multiple datasets analysed), or inconclusive). To facilitate the analysis, we regrouped the “data source” and “indicator” categories into broader groups (see Sect. S1 in the Supplement). We also created two extra categories: we classified systems under a specific “field”, and we introduced an “indicator type” based on whether the early warning was CSD-based or non-CSD based. We then excluded the running year 2023 and summarized results in terms of unique publications using simple statistics and alluvial plots in R (4.3.1).
Figure 3 Evolution of studies applying early warnings in empirical datasets. The total 229 papers we identified through our literature review between 2004 and 2022 were classified within five main scientific domains (ecology, health, climate, social sciences, and physical sciences). The dotted white line shows the cumulative number of papers.
We were able to classify the total 229 papers published from 2004 to 2023 into five main domains: ecology, climate, health, social sciences, and physical sciences. We found that empirical papers first appeared in 2007 in the domains of ecology and climate but that the first papers in health, social sciences, and physical sciences were only published after 2010 and 2011 (Fig. 3 ). This change may be associated with the highly cited review by Scheffer et al. (2009) that introduced (and popularized) the terms “early warning signals” and “critical transitions”. Since then, the number of empirical studies has quickly increased but has remained dominated by ecology (43.6 % of the papers overall), followed by health (22.6 %), climate (14.6 %), social sciences (12 %), and physical sciences (7.6 %) – showing the diversification of the uses of early warning (Fig. 3 ).
The higher number of publications in the health domain compared to the climate domain is unexpected. We found a large number of studies in the medical field (Fig. S1 in the Supplement) that form a distinct group on the emergence of human diseases, such as cancer (Liu et al., 2020), which use a non-CSD context-specific early warning (“dynamic_network_biomarkers”; see also Sect. 3.3 ). Zooming into within each domain, we observed that the most ecological studies are on terrestrial and freshwater fields (Fig. S1), namely on drylands and in forest and lake ecosystems (Table S1 in the Supplement). The majority of climate studies are on past climate transitions and modern records (Fig. S1, Table S2), while the social studies are split between societal shifts (like in politics, social behaviour, and transport) and finance transitions (Fig. S1, Table S4). Lastly, studies on physical sciences appear more heterogeneous, including tipping points in materials, power systems, and even astronomy (Fig. S1, Table S5).
Figure 4 Alluvial plot connecting scientific domains, data sources, and data types. Colours indicate the data source used for the estimation of early warnings. The size of the boxes in each column represents the proportion of each category. The figure is “read” from the middle column ( b , “data source”) to either the right ( c , “data type”) or the left ( a , “scientific domain”). The thickness of the lines is proportional to the studies of a given data source that belong to a certain domain (from the “data source” column, b , to the “domain” column, a ) or is proportional to the specific data type used (from the “data source” column, b , to the “data type” column, c ). For example, for the “data source” field experiment (light green), all studies using field experiments belong to the “ecology” domain, while field experiments are split into three types of data (spatial, spatiotemporal, and temporal). field_exp: field experiment; lab_exp: lab experiment; palaeo: palaeo-reconstructed data; remote-sensed: data through remote sensing; social_data: financial data and from social media; survey: data from surveys (field, lab, social).
Across scientific domains, the vast majority of early warnings were analysed on temporal data (77.7 %), while the spatial data were used in only 8 % of all studies (Fig. 4 ) only pertaining to ecology (Fig. S2). Survey data made up the majority of the data sources (43.8 %; including field surveys, social survey data, data from weather stations or other monitoring devices, medical data from hospitalization records to electroencephalograms (EEGs)), followed by data from lab experiments (20.7 %), remote sensing (12 %), palaeo-reconstructions (10 %), and field experiments (7 %). This partitioning can mostly be explained by our classification, meaning that we have grouped together a heterogeneity of data sources (e.g. field surveys, historical climate data, social study surveys, hospitalization records; Sect. S1). However, it also reflects the availability of each data source (e.g. most survey and palaeo data were readily available and reanalysed in the context of tipping points) or the difficulty in their acquisition (e.g. field experiments are harder to execute compared to lab experiments). Looking at how data sources are used across domains, ecology is the only domain where all kinds of data sources have been used. What is also interesting to note is that two sources of data are increasingly used: survey data and remote-sensed data (Fig. S3). Specifically, the latter were the latest to be used (2011) but show a consistent rising pattern over the last years, mainly due to the fact that satellite products span a long enough time period by now ( ∼ 20 years) to allow the estimation of early warnings.
A closer look at studies using remote-sensed products reveals a focus on the analysis of temporal early warning in land environments, mainly forests (e.g Boulton et al., 2022; Majumder et al., 2019; Saatchi et al., 2021) and drylands (e.g Veldhuis et al., 2022; Uden et al., 2019; Wu et al., 2023), but this also extends to the cryosphere, focusing on the analysis of the Arctic and the Antarctic ice sheets (AlMomani and Bollt, 2021; Carstensen and Weydmann, 2012). The spatial resolution of remote-sensed data has also been exploited for the identification of spatial early warning, especially regarding desertification (Berdugo et al., 2017) and vegetation analyses (Majumder et al., 2019). Overall, we found that the use of remote sensing products offers two distinct yet complementary approaches to detect early warning: high-level products, which correspond to physical variables, for instance sea surface temperature (SST) (Wu et al., 2015), or different types of indices like the normalized difference vegetation index (NDVI) (Liu et al., 2019) and low-level products or direct sensor observables.
Figure 5 Number of papers of the 21 early warnings used more than once in our literature review. Each bar is partitioned into the five scientific domains.
We recorded 65 different early warnings after reclassifying some into the same group (for example, variance, coefficient of variation, and standard deviation were reclassified as “variance”; Sect. S1). As expected, the majority were CSD-based warnings (74.9 %), while 25.1 % were non-CSD-based ones. Out of the 65 reclassified early warnings, only 21 were used more than once (the remaining 44 early warnings were used only once; Figs. 5 , S5). Variance and autocorrelation were the dominantly used early warnings across all domains, followed by skewness (Fig. 5 ). Besides these 3 early warnings, the remaining 18 were used selectively within particular domains. The most striking are “spatial variance” (only used in ecological studies) and “dynamic network biomarkers” (only used in health studies; see also Sect. 3.1 ). Within domains (Fig. S4), ecology is the domain with the highest heterogeneity in the early warnings (18 out of the 21 used more than once), followed by health (10) and climate (7).
Figure 6 Alluvial plot connecting scientific domains, the performance of the early warnings, and the type of early warning (CSD-based vs non-CSD-based). Colours indicate the performance. The size of the boxes in each column represents the proportion of each category. The figure is “read” from the middle column (“performance”, b ) to either the right (“early warning type”, c ) or the left (“scientific domain”, a ). The thickness of the lines is proportional to the performance that belongs to a certain domain (from the “performance” column, b , to the “domain” column, a ) or is proportional to the type of early warning (from the “performance” column, b , to the “early warning type” column, c ). For example, for the “performance” mixed (blue), studies with mixed performance were done with both CSD-based and non-CSD-based warnings (“early warning type” column, c ), while the CSD-based mixed were found in all domains and the non-CSD-based were split among climate, ecology, and social sciences (“scientific domain” column, a ). “Positive” performance indicates there was a warning identified, “negative” indicates no warning was identified, and “mixed” indicates positive and negative performances when tested in multiple datasets or when testing more than one early warning in the same dataset. For “inconclusive”, the results could indicate neither a positive nor a negative warning.
Figure 7 Early warnings (used > 1) and their performance. “Positive” performance indicates there was a warning identified, “negative” indicates no warning was identified, and “mixed” indicates positive and negative performances when tested in multiple datasets or when testing more than one early warning in the same dataset. For “inconclusive”, the results could indicate neither a positive nor a negative warning.
Of all studies, 67.8 % reported a positive performance of all 65 reclassified early warnings across all domains (Fig. 6 ). Only 3.4 % of the studies reported negative performances (i.e. no expected warning or opposite to expected warning). Studies in ecology reported the most negative results, followed by climate and health, with none reported for physical or social sciences. The performance in the rest of the studies was either mixed (i.e. positive or negative in studies which analysed multiple early warnings or datasets; 24.7 %) or inconclusive (i.e. a statistically weak result; 4.8 %). This is an impressively positively skewed result potentially reflecting the known bias in publishing significant results (Fanelli, 2012) or in post hoc analysis where a tipping point has already been documented and early warnings have been applied in hindsight (Boettiger and Hastings, 2012a). Interestingly, all the negative results included CSD-based warnings, while for non-CSD warnings only a small fraction reported inconclusive or mixed results (Fig. 6 ). This difference could be attributed to the fact that non-CSD warnings are at times idiosyncratically developed for the specific system under study compared to the more generic CSD-based warnings. Indeed, focusing on the 21 early warnings that were used more than once (Sect. 3.3 ), such system-specific indicators (such as the “dynamical_network_biomarkers”) always had a positive result (Fig. 7 ). Overall, the least-used warnings were associated with a positive performance, whereas the most-used ones (like variance, autocorrelation, skewness, and power spectrum) showed all types of performance. There was no early warning that had predominantly negative or mixed results except for kurtosis (Fig. 7 ). There was no particular difference in the performance across domains of the early warnings used more than once (Fig. S5).
The idea of early warnings based on CSD is relatively old. In the book Catastrophe Theory for Scientists and Engineers (1981), Gilmore already talked about “catastrophe flags” for indicators of CSD (Gilmore, 1981). After an early ecological paper (Wissel, 1984), the topic generated an enormous amount of interest, mainly theoretical, 20 years later, with the first empirical tests being on past climate tipping points (Dakos et al., 2008; Livina and Lenton, 2007). Our review of the literature of the last 2 decades shows how the use of these indicators has since spread to many other disciplines. Indeed, the generality of the approach – the fact that CSD can be observed in many dynamical systems, independent of the details of the underlying dynamical equations – created an opportunity for testing their validity on many systems and explains the enthusiasm they generated and the diversification of applications which followed.
Our literature overview suggests that the 65 early warning signals identified successfully detected a tipping point almost 70 % of the times they were used. This is an impressively positive result which should nonetheless be treated with caution. Firstly, in many cases, the empirical studies were conducted on systems that are either relatively simple or under controlled lab conditions, making them mostly proof-of-principle demonstrations (Dai et al., 2012; Veraart et al., 2012). Secondly, most empirical studies were restricted to hindsight application, meaning that an a priori knowledge of a tipping point may introduce bias towards detecting CSD indicators (Boettiger and Hastings, 2012a; Spears et al., 2017). Thirdly, the documented publication bias against negative or insignificant results (Fanelli, 2012; Franco et al., 2014) probably applies in the case of the early warning research given the attention this specific topic has attracted in recent years. One important aspect that we have not considered in the comparative analysis of the reviewed literature is the fact that each paper uses different statistical methods, different hypothesis-testing approaches (like surrogate data and Bayesian and frequentist p-values), and different significance levels to conclude on the identification of an early warning or not. To what extent such differences may even induce p-hacking is unclear, but it needs to be acknowledged in future work.
However, these considerations do not reduce the value and prospect of early warning research. For instance, one of their biggest values lies in the possible detection of an approaching tipping point. A number of studies have demonstrated the potential proximity of tipping points in modern climate data using early warnings (Boers, 2021; Boulton and Lenton, 2015; Ditlevsen and Ditlevsen, 2023). It has also become increasingly clear that early warnings can be very useful for comparing the resilience of similar ecosystems across space (e.g. (Forzieri et al., 2022; Lenton et al., 2022; Verbesselt et al., 2016) to provide an approximate estimate of resilience that can help prioritize management. In this way, with “resilience maps”, rather than speculating about the proximity to a potential threshold, we can rank situations at a given moment and place.
There are a number of conceptual, operational, and methodological challenges that can blur the detection of early warnings on real data (Dakos et al., 2015). For these reasons, some studies have highlighted their failure at detecting early warnings on data (Burthe et al., 2016), while others have raised caution about their uninformed use without knowing more about the system's drivers and underlying mechanisms (Boettiger et al., 2013). Our review can be used as a starting point in trying to understand when and why early warnings can fail or not by looking at how they have been applied within the domains we have identified, as their performance seems to be idiosyncratic to the data type and case study used. In what follows, we discuss more generally some of the most important challenges related to the use of early warnings.
The detection of early warnings relies on the assumption that the system is approaching a transition gradually. A system should be externally forced on a slow timescale towards the tipping point while experiencing perturbations on a shorter timescale such that CSD-based early warning signals can be estimated. In theory, it is generally assumed that the short-term noise is independent and identically distributed with a mean of zero. This is unlikely to be the case with climate systems experiencing extreme weather events, for example, which are likely becoming more prevalent with the changing climate. There have been three “1-in-100-year” droughts in the Amazon rainforest since 2005 (Erfanian et al., 2017; Lewis et al., 2011) which clearly alter the signals observed. For cases like these, it is worth measuring early warnings on the drivers themselves. If these show early warnings, then it is likely that signals observed in the system itself are being driven by changes in forcing rather than by a gradual movement towards tipping. However, early warnings of the drivers as a false-positive check make sense only in the case where the drivers are independent from the system variable. For instance, in the case of the Amazon, early warnings of rainfall can be seen as indicators of the Amazon tipping itself because of the strong moisture recycling feedback present rather than an external factor inducing early warnings on Amazon vegetation dynamics.
Things get even more complicated when more than one driver is acting on the system. In most cases, the assumption is that there is a single driver with a monotonic directional change towards the tipping point or that there are multiple drivers which all have the same effect and directionality. However, it has been shown experimentally that, in the presence of multiple drivers, contradictory early warnings may arise even if both drivers would produce similar patterns in early warnings acting in isolation (Dai et al., 2015).
When monitoring a system, longer time series are desirable to detect the upcoming tipping point. For instance, the best-case studies found in this literature review from remote-sensed products, which have been available since ∼ 1972, have approximately 50-year-long time series. However, due to sensor degradation and upgrades, it can be challenging to get a long time series from a single sensor, and products are often created from combined data sources. This can interfere with most of the early warnings if this merging changes the signal-to-noise ratio (SNR) across time (Smith et al., 2023). For example, newer sensors will measure with a greater radiometric accuracy, increasing the SNR and in turn “erroneously” increasing the AR(1) as far as an early warning is concerned. This increase in SNR will also decrease variance, thus allowing the user to check for anticorrelation between AR(1) and variance to see if the early warnings are being influenced or not.
As well as questions around data availability and noise behaviour, the inherent timescale of the system being studied can hinder our ability to detect tipping points. While tipping is by definition a “fast” process, for slow-moving systems like the thermohaline circulation (AMOC), this tipping event occurs over decades; therefore, it could be difficult to detect that the tipping point has been passed using early warnings. Another example of this is the Amazon rainforest, where, at least in modelled vegetation, there is a slow response of the forest based on the climate change that it has been subjected to (Jones et al., 2009). It could take decades for dieback to occur even under a constant climate such that a tipping point could be passed long before it is actually realized (Hughes et al., 2013). This “committed response” has been explored in a number of GCM experiments (Boulton et al., 2017; Jones et al., 2009), but it is unclear how early warnings would be affected by this (van Der Bolt et al., 2021).
The generic and universal character of most (but in particular CSD-based) early warnings comes at a price of these warnings not being specific to abrupt and irreversible tipping points. Instead they can also be used to detect smooth and reversible transitions (Kéfi et al., 2013). This limitation suggests that we need additional indicators, in particular system-specific indicators (Boettiger et al., 2013). In the case of spatially structured ecosystems such as drylands, studies have shown that temporal early warnings could fail (Dakos et al., 2011), in which case the use of the changes in the patterns themselves could provide a good alternative (Kéfi et al., 2007; Rietkerk et al., 2004). In the same way, specific indicators have been developed in health sciences for the monitoring of disease emergence (Table S3).
System-specific early warnings may also be a better prospect, where understanding processes in the system can help us to monitor its resilience in novel ways (Boulton et al., 2013). However, the original idea behind the development of early warnings was based on the premise that this knowledge is missing or insufficient; thus a pattern-based approach could be more informative (Scheffer et al., 2009). Therefore, the challenge is to strike the right level of system-specific warnings and to combine them with the generic ones. For instance, trait-based (Clements and Ozgul, 2016) and function-based (Hu et al., 2022) warnings have been recently suggested as complementary to the existing generic warning signals. A first step towards that direction could be to map the 65 classified early warnings we reviewed on a gradient of generic to system-specific indicators.
Most early warnings are well-tailored for unidimensional systems, meaning systems described by a single observable (e.g. vegetation cover). However, real dynamical systems are typically high-dimensional, and the quantification of early warnings in those multivariate systems presents challenges. For instance, two different variables may give conflicting information or obscure a clear signal (Boerlijst et al., 2013; Weinans et al., 2021). In theory, one expects that the variables directly involved in the interactions to cause a tipping point are the best to monitor (Carpenter et al., 2014). However, it is challenging to know from which variable(s) to measure early warnings in a multivariate system (Dakos, 2018).
Two main approaches in the analysis of multivariate systems have recently been developed. One relies on conceiving the system as a network, where the nodes are the variables and whose dynamics are followed through time, and evaluating changes in the structure of the network. For instance, as the system moves towards a tipping point, changes in degree distributions of such a network are representative for an approaching tipping point (Lu et al., 2021). Recent research explores a complementary approach where causal links are calculated instead of correlation links and where the strength of the causal link works as the indicator of resilience (Nowack et al., 2020).
Alternatively, dimension-reduction techniques can capture overall network dynamics into a representative statistic. For instance, principal component analysis (PCA; often referred to as empirical orthogonal function (EOF) in climate science) can be used to get directions of change (Held and Kleinen, 2004; Weinans et al., 2019). Data can be projected onto the leading principal component, effectively yielding a univariate time series on which the univariate early warning can be applied (Bathiany et al., 2013; Boulton and Lenton, 2015; Held and Kleinen, 2004). This analysis does not make any a priori assumptions about the interactions between the different network nodes and is therefore quite flexible in its use. However, it requires large amounts of high-quality data to yield accurate results. The underlying assumption is that, as the system approaches the tipping point, the dynamics become more correlated, leading to a high explained variance of a PCA and clear directionality in the dynamics (Lever et al., 2020).
A more peculiar challenge in the application of early warnings is their ability to detect cascading tipping points, where a tipping point in one system has a knock-on effect on another system, causing that to also tip (Klose et al., 2020; van de Leemput et al., 2018; Saade et al., 2023). Unless these systems are linked in such a way that early warnings can be observed in both systems, the cascade is likely to present as a shock to the second system such that it would be unpredictable whilst monitoring it in isolation (Bathiany et al., 2013). For systems where tipping in one system causes the connected system partially towards a tipping point (known as a “two-phase transition”), a stepwise jump in early warnings in the second system can be detected. For coupled systems where the tipping in the second system happens instantaneously (a “joint cascade”) or soon after the tipping in the first system (a “domino cascade”), early warnings are unlikely to be detectable (Klose et al., 2021).
These challenges associated with the use of early warnings are also accompanied by a number of opportunities to improve their detection in real data. Below we outline a few of the most promising ones.
Although there exists a multitude of early warnings (CSD-based and non-CSD-based; generic and system-specific; and on spatial, structural, and temporal data), few studies have compared in a systematic way how these warnings behave one against the other or across different systems (Dakos et al., 2011; Veldhuis et al., 2022). Apart from the CSD-based warnings where their relationships are mathematically known (Kuehn, 2012), we simply do not know what similar information early warnings provide. Understanding the interrelationships between all types of the early warnings most used will be crucial to improve their use for detecting tipping points. Composite metrics – where multiple early warnings are combined (Drake and Griffen, 2010), abundance-based with trait-based warnings are compared (Clements and Ozgul, 2016), or machine learning has been used to train models of multiple warnings as predictors (Brett and Rohani, 2020) – have been suggested to improve the significance and detectability of approaching tipping points. Given the increasing capacity to monitor the multivariate aspects of most systems (discussed in Sect. 4.2.3 ) and the increasing availability of such data (see Sect. 4.3.2 ), we are not far from estimating multiple early warnings on multiple dimensions of a system. The next step is to develop meaningful ways to best combine them for detecting tipping points.
Over the last decade, data from long-term databases and remote sensing have grown to become the primary sources for capturing temporal and spatial early warnings for tipping points. Especially for remote sensing data, this coincides with the expansion of freely available Earth observation datasets combined with access to cloud-based systems which provide the computational power to process this increase in data (Gorelick et al., 2017). A primary focus has been on the temporal analysis of optical imagery from satellites such as the Moderate Resolution Imaging Spectroradiometer (MODIS) (Liu et al., 2019; Majumder et al., 2019) or from the AVHRR (Lenton et al., 2022). Additionally, the vegetation optical depth (VOD) derived from microwave passive radiometers (Moesinger et al., 2020) has been employed to analyse early warnings, with temporal records since the late 1970s (Boulton et al., 2022; Smith et al., 2023). Overall, the continued growth of remotely sensed datasets is likely to drive further temporal early warning research, while the emergence of new satellite sensors with enhanced spatial resolutions (in the order of metres) will also enable an improved analysis of spatial early warning at large scales. However, such development requires a profound understanding of the acquisition systems to effectively control and account for parameters that may impact the extraction of early warnings.
The success of neural networks for time series classification problems has inspired the development of machine learning (ML) techniques for early warning detection. There is a natural synergy to this approach in that the same CSD phenomena manifest across a wide range of systems approaching tipping points, so the notoriously data-intensive task of training a neural network can be accomplished using plentiful synthetic data and still produce a result which can plausibly be applied to empirical data.
Deep learning models which combine convolutional layers have been shown to outperform methods using statistical CSD-based warnings (e.g. variance, AR(1)) in a variety of both real and simulated case studies (Bury et al., 2021; Deb et al., 2022). Furthermore, these models have exhibited success in inferring the type of upcoming bifurcation from observed pre-transition dynamics and have performed well in extensions to phase transitions on spatiotemporal lattices (Dylewsky et al., 2023). Other ML techniques can also tell us something about how far systems are from tipping. For example, random forest models could be used to determine the factors that determine autocorrelation in forest areas on a global scale and thus how close to tipping these forest areas could be based on driving variables (Forzieri et al., 2022), or they can help us determine the factors the influence the occurrence of tipping points (Berdugo et al., 2022). However, one should always bear in mind that ML will be as good as their training sets. Testing these approaches on existing datasets will help understand their potential along with testing them in cases of noise- or rate-tipping. Taking into consideration their limitations (Lapeyrolerie and Boettiger, 2022), combining ML techniques with “traditional” early warnings could become promising for monitoring systems that may be approaching tipping points.
The unprecedented amount of data originating from remote sensing systems, field measurements, surveys, and simulated data, coupled with innovative models and cutting-edge computing, has made possible the development of a multitude of tools and approaches for detecting tipping points in a variety of scientific fields. Early warnings can tell us that “something” important may be about to happen, but they do not tell us what precisely that “something” may be and when exactly it will happen (Dakos et al., 2015). The next step is to test the real potential of early warnings as preventive and management tools in anticipating natural and human-induced changes to come.
Literature review data can be found in Sect. S3 in the Supplement.
The supplement related to this article is available online at: https://doi.org/10.5194/esd-15-1117-2024-supplement .
VD, CAB, JEB, and SK developed research and drafted the paper. VD led the literature review analysed by VD, CAB, JEB, JFA, DIAM, DD, CLM, BvdB, LvdL, EW, and SK. VD and SK ran the analyses. All authors commented on the final text.
The contact author has declared that none of the authors has any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors.
This article is part of the special issue “Tipping points in the Anthropocene”. It is a result of the “Tipping Points: From Climate Crisis to Positive Transformation” international conference hosted by the Global Systems Institute (GSI) and University of Exeter (12–14 September 2022), as well as the associated creation of a Tipping Points Research Alliance by GSI and the Potsdam Institute for Climate Research, Exeter, Great Britain, 12–14 September 2022.
Chris A. Boulton, Joshua E. Buxton, and David I. Armstrong McKay were supported by the Bezos Earth Fund via the Global Tipping Points Report project. Carlos López-Martínez was supported by INTERACT project PID2020-114623RB-C32 funded by MCIN/AEI/10.13039/501100011033.
This paper was edited by Jonathan Donges and reviewed by Juan Rocha and two anonymous referees.
Abis, B. and Brovkin, V.: Alternative tree-cover states of the boreal ecosystem: A conceptual model, Global Ecol. Biogeogr., 28, 612–627, https://doi.org/10.1111/geb.12880 , 2019.
Alberto, J., Stelzer, A., Padric, J., Adrian, R., and Willem, B.: Early warning signals of regime shifts for aquatic systems: Can experiments help to bridge the gap between theory and real-world application?, Ecol. Complex., 47, 100944, https://doi.org/10.1016/j.ecocom.2021.100944 , 2021.
AlMomani, A. A. and Bollt, E.: An early warning sign of critical transition in the Antarctic ice sheet – a data-driven tool for a spatiotemporal tipping point, Nonlin. Processes Geophys., 28, 153–166, https://doi.org/10.5194/npg-28-153-2021 , 2021.
Ashwin, P., Wieczorek, S., Vitolo, R., and Cox, P.: Tipping points in open systems: bifurcation, noise-induced and rate-dependent examples in the climate system, Philos. T. R. Soc. A, 13, 20, https://doi.org/10.1098/rsta.2011.0306 , 2011.
Bathiany, S., Claussen, M., and Fraedrich, K.: Detecting hotspots of atmosphere–vegetation interaction via slowing down – Part 1: A stochastic approach, Earth Syst. Dynam., 4, 63–78, https://doi.org/10.5194/esd-4-63-2013 , 2013.
Bathiany, S., van der Bolt, B., Williamson, M. S., Lenton, T. M., Scheffer, M., van Nes, E. H., and Notz, D.: Statistical indicators of Arctic sea-ice stability – prospects and limitations, The Cryosphere, 10, 1631–1645, https://doi.org/10.5194/tc-10-1631-2016 , 2016.
Berdugo, M., Kéfi, S., Soliveres, S., and Maestre, F. T.: Plant spatial patterns identify alternative ecosystem multifunctionality states in global drylands, Nat. Ecol. Evol., 1, 0003, https://doi.org/10.1038/s41559-016-0003 , 2017.
Berdugo, M., Gaitán, J. J., Delgado-Baquerizo, M., Crowther, T. W., and Dakos, V.: Prevalence and drivers of abrupt vegetation shifts in global drylands, P. Natl. Acad. Sci. USA, 119, 1–10, https://doi.org/10.1073/pnas.2123393119 , 2022.
Bestelmeyer, B. T., A. M. Ellison, Fraser, W. R., Gorman, K. B., Holbrook, S. J., Laney, C. M., Ohman, M. D., Peters, D. P. C., Pillsbury, F. C., Rassweiler, A., Schmitt, R. J., and Sharm, S.: Analysis of abrupt transitions in ecological systems, Ecosphere, 2, 129, https://doi.org/10.1890/ES11-00216.1 , 2011.
Boerlijst, M. C., Oudman, T., and de Roos, A. M.: Catastrophic Collapse Can Occur without Early Warning: Examples of Silent Catastrophes in Structured Ecological Models, PLOS ONE, 8, e62033, https://doi.org/10.1371/journal.pone.0062033 , 2013.
Boers, N.: Observation-based early-warning signals for a collapse of the Atlantic Meridional Overturning Circulation, Nat. Clim. Change, 11, 680–688, https://doi.org/10.1038/s41558-021-01097-4 , 2021.
Boettiger, C. and Hastings, A.: Early warning signals and the prosecutor's fallacy, P. R. Soc. B, 279, 4734–4739, https://doi.org/10.1098/rspb.2012.2085 , 2012a.
Boettiger, C. and Hastings, A.: Quantifying limits to detection of early warning for critical transitions, J. R. Soc. Interface, 9, 2527–39, https://doi.org/10.1098/rsif.2012.0125 , 2012b.
Boettiger, C., Ross, N., and Hastings, A.: Early warning signals: the charted and uncharted territories, Theor. Ecol., 6, 255–264, https://doi.org/10.1007/s12080-013-0192-6 , 2013.
Boulton, C. A. and Lenton, T. M.: Slowing down of North Pacific climate variability and its implications for abrupt ecosystem change, P. Natl. Acad. Sci. USA, 112, 11496–11501, https://doi.org/10.1073/pnas.1501781112 , 2015.
Boulton, C. a., Good, P., and Lenton, T. M.: Early warning signals of simulated Amazon rainforest dieback, Theor. Ecol., 6, 373–384, https://doi.org/10.1007/s12080-013-0191-7 , 2013.
Boulton, C. A., Booth, B. B. B., and Good, P.: Exploring uncertainty of Amazon dieback in a perturbed parameter Earth system ensemble, Glob. Change Biol., 23, 5032–5044, https://doi.org/10.1111/gcb.13733 , 2017.
Boulton, C. A., Lenton, T. M., and Boers, N.: Pronounced loss of Amazon rainforest resilience since the early 2000s, Nat. Clim. Change, 12, 271–278, https://doi.org/10.1038/s41558-022-01287-8 , 2022.
Brett, T. S. and Rohani, P.: Dynamical footprints enable detection of disease emergence, PLOS Biol., 18, e3000697, https://doi.org/10.1371/journal.pbio.3000697 , 2020.
Burthe, S. J., Henrys, P. A., Mackay, E. B., Spears, B. M., Campbell, R., Carvalho, L., Dudley, B., Gunn, I. D. M., Johns, D. G., Maberly, S. C., May, L., Newell, M. A., Wanless, S., Winfield, I. J., Thackeray, S. J., and Daunt, F.: Do early warning indicators consistently predict nonlinear change in long-term ecological data?, 53, 666–676, https://doi.org/10.1111/1365-2664.12519 , 2016.
Bury, T. M.: ewstools: A Python package for early warning signals of bifurcations in time series data, J. Open Source Softw., 8, 8–11, https://doi.org/10.21105/joss.05038 , 2023.
Bury, T. M., Sujith, R. I., Pavithran, I., Scheffer, M., Lenton, T. M., Anand, M., and Bauch, C. T.: Deep learning for early warning signals of regime shifts, P. Natl. Acad. Sci. USA, 450, e2106140118, https://doi.org/10.1073/pnas.2106140118 , 2021.
Carpenter, S. R. and Brock, W. A.: Rising variance: a leading indicator of ecological transition, Ecol. Lett., 9, 311–318, https://doi.org/10.1111/j.1461-0248.2005.00877.x , 2006.
Carpenter, S. R., Brock, W. A., Cole, J. J., Kitchell, J. F., and Pace, M. L.: Leading indicators of trophic cascades, Ecol. Lett., 11, 128–38, https://doi.org/10.1111/j.1461-0248.2007.01131.x , 2008.
Carpenter, S. R., Brock, W. A., Cole, J. J., and Pace, M. L.: A new approach for rapid detection of nearby thresholds in ecosystem time series, Oikos, 123, 290–297, https://doi.org/10.1111/j.1600-0706.2013.00539.x , 2014.
Carstensen, J. and Weydmann, A.: Tipping Points in the Arctic: Eyeballing or Statistical Significance?, AMBIO, 41, 34–43, https://doi.org/10.1007/s13280-011-0223-8 , 2012.
Cavaliere, M., Yang, G., Danos, V., and Dakos, V.: Detecting the Collapse of Cooperation in Evolving Networks, Sci. Rep.-UK, 6, 30845, https://doi.org/10.1038/srep30845 , 2016.
Clements, C. F. and Ozgul, A.: Including trait-based early warning signals helps predict population collapse, Nat. Commun., 7, 1–8, https://doi.org/10.1038/ncomms10984 , 2016.
Clements, C. F. and Ozgul, A.: Indicators of transitions in biological systems, Ecol. Lett., 21, 905–919, https://doi.org/10.1111/ele.12948 , 2018.
Dai, L., Vorselen, D., Korolev, K. S., and Gore, J.: Generic Indicators for Loss of Resilience Before a Tipping Point Leading to Population Collapse, Science, 336, 1175–1177, https://doi.org/10.1126/science.1219805 , 2012.
Dai, L., Korolev, K. S., and Gore, J.: Relation between stability and resilience determines the performance of early warning signa ls under different environmental drivers, P. Natl. Acad. Sci. USA, 112, 201418415, https://doi.org/10.1073/pnas.1418415112 , 2015.
Dakos, V.: Identifying best-indicator species for abrupt transitions in multispecies communities, Ecol. Indic., 94, 494–502, https://doi.org/10.1016/j.ecolind.2017.10.024 , 2018.
Dakos, V. and Kéfi, S.: Ecological resilience: what to measure and how, Environ. Res. Lett., 17, 043003, https://doi.org/10.1088/1748-9326/ac5767 , 2022.
Dakos, V., Scheffer, M., van Nes, E. H., Brovkin, V., Petoukhov, V., and Held, H.: Slowing down as an early warning signal for abrupt climate change., P. Natl. Acad. Sci. USA, 105, 14308–12, https://doi.org/10.1073/pnas.0802430105 , 2008.
Dakos, V., van Nes, E. H., Donangelo, R., Fort, H., and Scheffer, M.: Spatial correlation as leading indicator of catastrophic shifts, Theor. Ecol., 3, 163–174, https://doi.org/10.1007/s12080-009-0060-6 , 2010.
Dakos, V., Kéfi, S., Rietkerk, M., van Nes, E. H., and Scheffer, M.: Slowing down in spatially patterned ecosystems at the brink of collapse, Am. Nat., 177, E153–166, https://doi.org/10.1086/659945 , 2011.
Dakos, V., Carpenter, S. R., Brock, W. A., Ellison, A. M., Guttal, V., Ives, A. R., Kéfi, S., Livina, V., Seekell, D. A., van Nes, E. H., and Scheffer, M.: Methods for Detecting Early Warnings of Critical Transitions in Time Series Illustrated Using Simulated Ecological Data, PLoS ONE, 7, e41010, https://doi.org/10.1371/journal.pone.0041010 , 2012.
Dakos, V., van Nes, E. H., and Scheffer, M.: Flickering as an early warning signal, Theor. Ecol., 6, 309–317, https://doi.org/10.1007/s12080-013-0186-4 , 2013.
Dakos, V., Carpenter, S. R., Nes, E. H. V., and Scheffer, M.: Resilience indicators: prospects and limitations for early warnings of regime shifts, Philos. T. R. Soc. B, 370, 20130263, https://doi.org/10.1098/rstb.2013.0263 , 2015.
Deb, S., Sidheekh, S., Clements, C. F., Krishnan, N. C., and Dutta, P. S.: Machine learning methods trained on simple models can predict critical transitions in complex natural systems, Roy. Soc. Open Sci., 9, 211475, https://doi.org/10.1098/rsos.211475 , 2022.
Ditlevsen, P. and Ditlevsen, S.: Warning of a forthcoming collapse of the Atlantic meridional overturning circulation, Nat. Commun., 14, 4254, https://doi.org/10.1038/s41467-023-39810-w , 2023.
Drake, J. M. and Griffen, B. D.: Early warning signals of extinction in deteriorating environments, Nature, 467, 456–459, https://doi.org/10.1038/nature09389 , 2010.
Drake, J. M., O'Regan, S. M., Dakos, V., Kéfi, S., Rohani, P., Drake, J. M., O'Regan, S. M., Dakos, V., Kéfi, S., and Rohani, P.: Alternative stable states, tipping points, and early warning signals of ecological transitions, in: Theoretical Ecology: Concepts and Applications, edited by: McCann, K. S. and Gellner, G., Oxford University Press, 263–284, https://doi.org/10.1093/oso/9780198824282.003.0015 , 2020.
Dylewsky, D., Lenton, T. M., Scheffer, M., Bury, T. M., Fletcher, C. G., Anand, M., and Bauch, C. T.: Universal early warning signals of phase transitions in climate systems, J. R. Soc. Interface, 20, 20220562, https://doi.org/10.1098/rsif.2022.0562 , 2023.
Erfanian, A., Wang, G., and Fomenko, L.: Unprecedented drought over tropical South America in 2016: significantly under-predicted by tropical SST, Sci. Rep.-UK, 7, 5811, https://doi.org/10.1038/s41598-017-05373-2 , 2017.
Fanelli, D.: Negative results are disappearing from most disciplines and countries, Scientometrics, 90, 891–904, https://doi.org/10.1007/s11192-011-0494-7 , 2012.
Forzieri, G., Dakos, V., McDowell, N. G., Ramdane, A., and Cescatti, A.: Emerging signals of declining forest resilience under climate change, Nature, 608, 534–539, https://doi.org/10.1038/s41586-022-04959-9 , 2022.
Franco, A., Malhotra, N., and Simonovits, G.: Publication bias in the social sciences: Unlocking the file drawer, Science, 345, 1502–1505, https://doi.org/10.1126/science.1255484 , 2014.
Génin, A., Majumder, S., Sankaran, S., Danet, A., Guttal, V., Schneider, F. D., and Kéfi, S.: Monitoring ecosystem degradation using spatial data and the R package spatialwarnings, Methods Ecol. Evol., 9, 2067–2075, https://doi.org/10.1111/2041-210X.13058 , 2018.
Gilmore, R.: Catastrophe theory for scientists and engineers, John Wiley & Sons, ISBN 0-486-67539-4, 1981.
Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., and Moore, R.: Google Earth Engine: Planetary-scale geospatial analysis for everyone, Remote Sens. Environ., 202, 18–27, https://doi.org/10.1016/j.rse.2017.06.031 , 2017.
Guttal, V. and Jayaprakash, C.: Changing skewness: an early warning signal of regime shifts in ecosystems, Ecol. Lett., 11, 450–460, https://doi.org/10.1111/j.1461-0248.2008.01160.x , 2008.
Guttal, V. and Jayaprakash, C.: Spatial variance and spatial skewness: leading indicators of regime shifts in spatial ecological systems, Theor. Ecol., 2, 3–12, https://doi.org/10.1007/s12080-008-0033-1 , 2009.
Hagstrom, G. and Levin, S.: Phase Transitions and the Theory of Early Warning Indicators for Critical Transitions, arXiv [preprint], https://doi.org/10.48550/arXiv.2110.12287 , 23 October 2021.
Hastings, A. and Wysham, D. B.: Regime shifts in ecological systems can occur with no warning, Ecol. Lett., 13, 464–472, https://doi.org/10.1111/j.1461-0248.2010.01439.x , 2010.
Held, H. and Kleinen, T.: Detection of climate system bifurcations by degenerate fingerprinting, Geophys. Res. Lett., 31, 1–4, https://doi.org/10.1029/2004GL020972 , 2004.
Hirota, M., Holmgren, M., van Nes, E. H., and Scheffer, M.: Global resilience of tropical forest and savanna to critical transitions, Science, 334, 232–235, 2011.
Hu, Z., Dakos, V., and Rietkerk, M.: Using functional indicators to detect state changes in terrestrial ecosystems, Trends Ecol. Evol., 37, 1036–1045, https://doi.org/10.1016/j.tree.2022.07.011 , 2022.
Hughes, T. P., Linares, C., Dakos, V., van de Leemput, I. A., and van Nes, E. H.: Living dangerously on borrowed time during slow, unrecognized regime shifts, Trends Ecol. Evol., 28, 149–55, https://doi.org/10.1016/j.tree.2012.08.022 , 2013.
Ives, A. R. and Dakos, V.: Detecting dynamical changes in nonlinear time series using locally linear state-space models, Ecosphere, 3, 1–15, https://doi.org/10.1890/ES11-00347.1 , 2012.
Jones, C., Lowe, J., Liddicoat, S., and Betts, R.: Committed terrestrial ecosystem changes due to climate change, Nat. Geosci., 2, 484–487, https://doi.org/10.1038/ngeo555 , 2009.
Kéfi, S., Rietkerk, M., Alados, C. L., Pueyo, Y., Papanastasis, V. P., ElAich, A., de Ruiter, P. C., Alados, L., and Ruite r, P. C. D.: Spatial vegetation patterns and imminent desertification in Mediterranean arid ecosystems, Nature, 449, 213–217, https://doi.org/10.1038/nature06111 , 2007.
Kéfi, S., Dakos, V., Scheffer, M., Van Nes, E. H., and Rietkerk, M.: Early warning signals also precede non-catastrophic transitions, Oikos, 122, 641–648, 2013.
Kéfi, S., Guttal, V., Brock, W. A., Carpenter, S. R., Ellison, A. M., Livina, V. N., Seekell, D. A., Scheffer, M., van Nes, E. H., and Dakos, V.: Early Warning Signals of Ecological Transitions: Methods for Spatial Patterns, PLOS ONE, 9, e92097, https://doi.org/10.1371/journal.pone.0092097 , 2014.
Kleinen, T., Held, H., Petschel-held, G., and Held, Á. H.: The potential role of spectral properties in detecting thresholds in the Earth system: application to the thermohaline circulation, Ocean Dynam., 53, 53–63, https://doi.org/10.1007/s10236-002-0023-6 , 2003.
Klose, A. K., Karle, V., Winkelmann, R., and Donges, J. F.: Emergence of cascading dynamics in interacting tipping elements of ecology and climate: Cascading dynamics in tipping elements, Roy. Soc. Open Sci., 7, 200599, https://doi.org/10.1098/rsos.200599 , 2020.
Klose, A. K., Wunderling, N., Winkelmann, R., and Donges, J. F.: What do we mean, “tipping cascade”?, Environ. Res. Lett., 16, 1–11, https://doi.org/10.1088/1748-9326/ac3955 , 2021.
Kuehn, C.: A Mathematical Framework for Critical Transitions: Normal Forms, Variance and Applications, J. Nonlinear Sci., 23, 457–510, https://doi.org/10.1007/s00332-012-9158-x , 2012.
Kuznetsov, Y. A.: Elements of Applied Bifurcation Theory, Springer-Verlag, New York, https://doi.org/10.1007/978-1-4757-3978-7 , 591 pp., 1995.
Lade, S. J. and Gross, T.: Early warning signals for critical transitions: A generalized modeling approach, PLOS Comput. Biol., 8, e1002360, https://doi.org/10.1371/journal.pcbi.1002360 , 2012.
Laitinen, V., Dakos, V., and Lahti, L.: Probabilistic early warning signals, Ecol. Evol., 1–14, https://doi.org/10.1002/ece3.8123 , 2021.
Lapeyrolerie, M. and Boettiger, C.: Limits to ecological forecasting: Estimating uncertainty for critical transitions with deep learning, Meth. Ecol. Evol., 14, 785–798, https://doi.org/10.1111/2041-210X.14013 , 2022.
Lenton, T. M.: Early warning of climate tipping points, Nat. Clim. Change, 1, 201–209, 2011.
Lenton, T. M.: Environmental Tipping Points, 38, 1–29, https://doi.org/10.1146/annurev-environ-102511-084654 , 2013a.
Lenton, T. M.: What early warning systems are there for environmental shocks?, Environ. Sci. Policy, 27, S60–S75, https://doi.org/10.1016/j.envsci.2012.06.011 , 2013b.
Lenton, T. M., Buxton, J. E., Armstrong McKay, D. I., Abrams, J. F., Boulton, C. A., Lees, K., Powell, T. W. R., Boers, N., Cunliffe, A. M., and Dakos, V.: A resilience sensing system for the biosphere, Philos. T. R. Soc. B, 377, 20210383, https://doi.org/10.1098/rstb.2021.0383 , 2022.
Lever, J. J., van de Leemput, I. A., Weinans, E., Quax, R., Dakos, V., van Nes, E. H., Bascompte, J., and Scheffer, M.: Foreseeing the future of mutualistic communities beyond collapse, Ecol. Lett., 23, 2–15, https://doi.org/10.1111/ele.13401 , 2020.
Lewis, S. L., Brando, P. M., Phillips, O. L., Van Der Heijden, G. M. F., and Nepstad, D.: The 2010 Amazon Drought, Science, 331, 554–554, https://doi.org/10.1126/science.1200807 , 2011.
Litzow, M. A. and Hunsicker, M. E.: Early warning signals, nonlinearity, and signs of hysteresis in real ecosystems, Ecosphere, 7, e01614, https://doi.org/10.1002/ecs2.1614 , 2016.
Liu, L., Shao, Z., Lv, J., Xu, F., Ren, S., Jin, Q., Yang, J., Ma, W., Xie, H., Zhang, D., and Chen, X.: Identification of Early Warning Signals at the Critical Transition Point of Colorectal Cancer Based on Dynamic Network Analysis, Front. Bioeng. Biotechnol., 8, 530, https://doi.org/10.3389/fbioe.2020.00530 , 2020.
Liu, Y., Kumar, M., Katul, G. G., and Porporato, A.: Reduced resilience as an early warning signal of forest mortality, Nat. Clim. Change, 9, 880–885, https://doi.org/10.1038/s41558-019-0583-9 , 2019.
Livina, V. N. and Lenton, T. M.: A modified method for detecting incipient bifurcations in a dynamical system, Geophys. Res. Lett., 34, 1–5, https://doi.org/10.1029/2006GL028672 , 2007.
Livina, V. N., Kwasniok, F., and Lenton, T. M.: Potential analysis reveals changing number of climate states during the last 60 kyr , Clim. Past, 6, 77–82, https://doi.org/10.5194/cp-6-77-2010 , 2010.
Lu, Z., Yuan, N., Yang, Q., Ma, Z., and Kurths, J.: Early Warning of the Pacific Decadal Oscillation Phase Transition Using Complex Network Analysis, Geophys. Res. Lett., 48, e2020GL091674, https://doi.org/10.1029/2020GL091674 , 2021.
Majumder, S., Tamma, K., Ramaswamy, S., and Guttal, V.: Inferring critical thresholds of ecosystem transitions from spatial data, Ecology, 100, e02722, https://doi.org/10.1002/ecy.2722 , 2019.
Mayfield, R. J., Langdon, P. G., Doncaster, C. P., Dearing, J. A., Wang, R., Nazarova, L. B., Medeiros, A. S., and Brooks, S. J.: Metrics of structural change as indicators of chironomid community stability in high latitude lakes, Quaternary Sci. Rev., 249, 106594, https://doi.org/10.1016/j.quascirev.2020.106594 , 2020.
Moesinger, L., Dorigo, W., de Jeu, R., van der Schalie, R., Scanlon, T., Teubner, I., and Forkel, M.: The global long-term microwave Vegetation Optical Depth Climate Archive (VODCA), Earth Syst. Sci. Data, 12, 177–196, https://doi.org/10.5194/essd-12-177-2020 , 2020.
Nijp, J. J., Temme, A. J. A. M., van Voorn, G. A. K., Kooistra, L., Hengeveld, G. M., Soons, M. B., Teuling, A. J., and Wallinga, J.: Spatial early warning signals for impending regime shifts: A practical framework for application in real-world landscapes, Glob. Change Biol., 25, 1905–1921, https://doi.org/10.1111/gcb.14591 , 2019.
Nowack, P., Runge, J., Eyring, V., and Haigh, J. D.: Causal networks for climate model evaluation and constrained projections, Nat. Commun., 11, 1415, https://doi.org/10.1038/s41467-020-15195-y , 2020.
O'Brien, D. A., Deb, S., Sidheekh, S., Krishnan, N. C., Sharathi Dutta, P., and Clements, C. F.: EWSmethods: an R package to forecast tipping points at the community level using early warning signals, resilience measures, and machine learning models, Ecography, 2023, e06674, https://doi.org/10.1111/ecog.06674 , 2023.
Rietkerk, M., Dekker, S. C., de Ruiter, P. C., and van de Koppel, J.: Self-Organized Patchiness and Catastrophic Shifts in Ecosystems, Science, 305, 1926–1929, https://doi.org/10.1126/science.1101867 , 2004.
Rietkerk, M., van de Koppel, J., and Koppel, J. V. D.: Regular pattern formation in real ecosystems, Trends Ecol. Evol., 23, 169–175, https://doi.org/10.1016/j.tree.2007.10.013 , 2008.
Ritchie, P. and Sieber, J.: Early-warning indicators for rate-induced tipping, Chaos, 26, 093116, https://doi.org/10.1063/1.4963012 , 2016.
Saade, C., Fronhofer, E. A., Pichon, B., and Kéfi, S.: Landscape Structure Affects Metapopulation-Scale Tipping Points, Am. Nat., 202, E17–E30, https://doi.org/10.1086/724550 , 2023.
Saatchi, S., Longo, M., Xu, L., Yang, Y., Abe, H., André, M., Aukema, J. E., Carvalhais, N., Cadillo-Quiroz, H., Cerbu, G. A., Chernela, J. M., Covey, K., Sánchez-Clavijo, L. M., Cubillos, I. V., Davies, S. J., De Sy, V., De Vleeschouwer, F., Duque, A., Sybille Durieux, A. M., De Avila Fernandes, K., Fernandez, L. E., Gammino, V., Garrity, D. P., Gibbs, D. A., Gibbon, L., Gowae, G. Y., Hansen, M., Lee Harris, N., Healey, S. P., Hilton, R. G., Johnson, C. M., Kankeu, R. S., Laporte-Goetz, N. T., Lee, H., Lovejoy, T., Lowman, M., Lumbuenamo, R., Malhi, Y., Albert Martinez, J.-M. M., Nobre, C., Pellegrini, A., Radachowsky, J., Román, F., Russell, D., Sheil, D., Smith, T. B., Spencer, R. G. M., Stolle, F., Tata, H. L., Torres, D. del C., Tshimanga, R. M., Vargas, R., Venter, M., West, J., Widayati, A., Wilson, S. N., Brumby, S., and Elmore, A. C.: Detecting vulnerability of humid tropical forests to multiple stressors, One Earth, 4, 988–1003, https://doi.org/10.1016/j.oneear.2021.06.002 , 2021.
Scheffer, M., Bascompte, J., Brock, W. A., Brovkin, V., Carpenter, S. R., Dakos, V., Held, H., van Nes, E. H., Rietkerk, M., and Sugihara, G.: Early-warning signals for critical transitions, Nature, 461, 53–59, https://doi.org/10.1038/nature08227 , 2009.
Scheffer, M., Carpenter, S. R., Lenton, T. M., Bascompte, J., Brock, W., Dakos, V., van de Koppel, J., van de Leemput, I. A., Levin, S. A., van Nes, E. H., Pascual, M., and Vandermeer, J.: Anticipating Critical Transitions, Science, 338, 344–348, https://doi.org/10.1126/science.1225244 , 2012a.
Scheffer, M., Hirota, M., Holmgren, M., Van Nes, E. H., and Chapin, F. S.: Thresholds for boreal biome transitions., P. Natl. Acad. Sci. USA, 109, 21384–9, https://doi.org/10.1073/pnas.1219844110 , 2012b.
Scheffer, M., Carpenter, S. R., Dakos, V., and van Nes, E. H.: Generic Indicators of Ecological Resilience: Inferring the Chance of a Critical Transition, Annu. Rev. Ecol. Evol. S., 46, 145–167, https://doi.org/10.1146/annurev-ecolsys-112414-054242 , 2015.
Scheffer, M., Bolhuis, J. E., Borsboom, D., Buchman, T. G., Gijzel, S. M. W., Goulson, D., Kammenga, J. E., Kemp, B., van de Leemput, I. A., Levin, S., Martin, C. M., Melis, R. J. F., van Nes, E. H., Romero, L. M., and Olde Rikkert, M. G. M.: Quantifying resilience of humans and other animals, P. Natl. Acad. Sci. USA, 115, 201810630, https://doi.org/10.1073/pnas.1810630115 , 2018.
Seekell, D. A. and Dakos, V.: Heteroskedasticity as a leading indicator of desertification in spatially explicit data, Ecol. Evol., 5, 2185–2192, https://doi.org/10.1002/ece3.1510 , 2015.
Seekell, D. A., Carpenter, S. R., Cline, T. J., and Pace, M. L.: Conditional Heteroskedasticity Forecasts Regime Shift in a Whole-Ecosystem Experiment, Ecosystems, 15, 741–747, https://doi.org/10.1007/s10021-012-9542-2 , 2012.
Smith, T., Zotta, R.-M., Boulton, C. A., Lenton, T. M., Dorigo, W., and Boers, N.: Reliability of resilience estimation based on multi-instrument time series, Earth Syst. Dynam., 14, 173–183, https://doi.org/10.5194/esd-14-173-2023 , 2023.
Sole, R. V., Manrubia, S. C., Luque, B., Delgado, J., Bascompte, J., Solé, R. V., Manrubia, S. C., Luque, B., Delgado, J., and Bascompte, J.: Phase transitions and complex systems, Complexity, 1, 13–26, 1996.
Spanbauer, T. L., Allen, C. R., Angeler, D. G., Eason, T., Fritz, S. C., Garmestani, A. S., Nash, K. L., Stone, J. R., Stow, C. A., and Sundstrom, S. M.: Body size distributions signal a regime shift in a lake ecosystem, P. R. Soc. B, 283, https://doi.org/10.1098/rspb.2016.0249 , 2016.
Spears, B. M., Futter, M. N., Jeppesen, E., Huser, B. J., Ives, S., Davidson, T. A., Adrian, R., Angeler, D. G., Burthe, S. J., Carvalho, L., Daunt, F., Gsell, A. S., Hessen, D. O., Janssen, A. B. G., Mackay, E. B., May, L., Moorhouse, H., Olsen, S., Søndergaard, M., Woods, H., and Thackeray, S. J.: Ecological resilience in lakes and the conjunction fallacy, Nat. Ecol. Evol., 1, 1616–1624, https://doi.org/10.1038/s41559-017-0333-1 , 2017.
Staver, A. C., Archibald, S., and Levin, S. A.: The Global Extent and Determinants of Savanna and Forest as Alternative Biome States, Science, 334, 230–232, https://doi.org/10.1126/science.1210465 , 2011.
Tang, Y., Zhu, X., He, C., Hu, J., and Fan, J.: Critical slowing down theory provides early warning signals for sandstone failure, Front. Earth Sci., 10, 934498, https://doi.org/10.3389/feart.2022.934498 , 2022.
Thompson, J. M. T. and Sieber, J.: Predicting climate tipping as a noisy bifurcation: a review, Int. J. Bifurcat Chaos, 21, 399–423, https://doi.org/10.1142/s0218127411028519 , 2011.
Tirabassi, G., Viebahn, J., Dakos, V., Dijkstra, H. A., Masoller, C., Rietkerk, M., and Dekker, S. C.: Interaction network based early-warning indicators of vegetation transitions, Ecol. Complex., 19, 148–157, https://doi.org/10.1016/j.ecocom.2014.06.004 , 2014.
Uden, D. R., Twidwell, D., Allen, C. R., Jones, M. O., Naugle, D. E., Maestas, J. D., Allred, B. W., and Wallace, K. J.: Spatial Imaging and Screening for Regime Shifts, Front. Ecol. Evol., 7, 1–16, https://doi.org/10.3389/fevo.2019.00407 , 2019.
van de Leemput, I. A., Dakos, V., Scheffer, M., and van Nes, E. H.: Slow Recovery from Local Disturbances as an Indicator for Loss of Ecosystem Resilience, Ecosystems, 21, 141–152, https://doi.org/10.1007/s10021-017-0154-8 , 2018.
van Der Bolt, B., van Nes, E. H., and Scheffer, M.: No warning for slow transitions, J. R. Soc. Interface, 18, 20200935, https://doi.org/10.1098/rsif.2020.0935 , 2021.
van Nes, E. H., Arani, B. M. S., Staal, A., van der Bolt, B., Flores, B. M., Bathiany, S., and Scheffer, M.: What Do You Mean, `Tipping Point'?, Trends Ecol. Evol., 31, 902–904, https://doi.org/10.1016/j.tree.2016.09.011 , 2016.
Veldhuis, M. P., Martinez-Garcia, R., Deblauwe, V., and Dakos, V.: Remotely-sensed slowing down in spatially patterned dryland ecosystems, Ecography, 2022, e06139, https://doi.org/10.1111/ecog.06139 , 2022.
Veraart, A. J. A. J., Faassen, E. J. E. J., Dakos, V., Van Nes, E. H. E. H., Lürling, M., Scheffer, M., and Lurling, M.: Recovery rates reflect distance to a tipping point in a living system, Nature, 481, 357–359, https://doi.org/10.1038/nature10723 , 2012.
Verbesselt, J., Umlauf, N., Hirota, M., Holmgren, M., Van Nes, E. H., Herold, M., Zeileis, A., and Scheffer, M.: Remotely sen sed resilience of tropical forests, Nat. Clim. Change, 6, 1028–1031, https://doi.org/10.1038/nclimate3108 , 2016.
von Hardenberg, J., Meron, E., Shachak, M., Zarmi, Y., and Hardenberg, J. V.: Diversity of vegetation patterns and desertification, Phys. Rev. Lett., 8719, 198101, https://doi.org/10.1103/PhysRevLett.87.198101 , 2001.
Wang, R., Dearing, J. A., Langdon, P. G., Zhang, E., Yang, X., Dakos, V., and Scheffer, M.: Flickering gives early warning signals of a critical transition to a eutrophic lake state, Nature, 492, 419–422, https://doi.org/10.1038/nature11655 , 2012.
Weinans, E., Lever, J. J., Bathiany, S., Quax, R., Bascompte, J., Van Nes, E. H., Scheffer, M., and Van De Leemput, I. A.: Finding the direction of lowest resilience in multivariate complex systems, J. R. Soc. Interface, 16, https://doi.org/10.1098/rsif.2019.0629 , 2019.
Weinans, E., Quax, R., van Nes, E. H., and Leemput, I. A. van de: Evaluating the performance of multivariate indicators of resilience loss, Sci. Rep.-UK, 11, 9148, https://doi.org/10.1038/s41598-021-87839-y , 2021.
Wissel, C.: A universal law of the characteristic return time near thresholds, Oecologia, 65, 101–107, 1984.
Wu, H., Hou, W., Yan, P.-C., Zhang, Z.-S., and Wang, K.: A study of the early warning signals of abrupt change in the Pacific decadal oscillation, Chinese Phys. B, 24, 089201, https://doi.org/10.1088/1674-1056/24/8/089201 , 2015.
Wu, S., Liu, L., Li, D., Zhang, W., Liu, K., Shen, J., and Zhang, L.: Global desert expansion during the 21st century: Patterns, predictors and signals, Land Degrad. Dev., 34, 377–388, https://doi.org/10.1002/ldr.4466 , 2023.
Yin, Z., Dekker, S. C., Rietkerk, M., van den Hurk, B. J. J. M., and Dijkstra, H. A.: Network based early warning indicators of vegetation changes in a land–atmosphere model, Ecol. Complex., 26, 68–78, https://doi.org/10.1016/j.ecocom.2016.02.004 , 2016.
You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.
All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .
Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.
Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.
Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.
Original Submission Date Received: .
Find support for a specific problem in the support section of our website.
Please let us know what you think of our products and services.
Visit our dedicated information section to learn more about MDPI.
Removal of atmospheric methane by increasing hydroxyl radicals via a water vapor enhancement strategy.
1.1. importance of methane mitigation for climate, 1.2. relationship between methane, atmospheric oh, and water vapor, 1.2.1. relationship between methane and atmospheric oh, 1.2.2. relationship between atmospheric oh and water vapor, 2. effects of oh on methane concentration, 3. greenhouse effects of water vapor, 4. comparison of radiative forcing due to methane and water vapor variations, 5. discussion, 6. conclusions, author contributions, institutional review board statement, informed consent statement, data availability statement, conflicts of interest.
Click here to enlarge figure
Initial Concentrations | Initial Emission Rates | Reaction Rate Coefficients | |
---|---|---|---|
CH (N) | 1975 ppb | 395 Tg yr | |
CH (S) | 1875 ppb | 167 Tg yr | |
CO (N) | 120 ppb | 910 Tg yr | |
CO (S) | 60 ppb | 420 Tg yr | |
OH (Global mean) | 1.09 × 10 molecules cm | 251.2 Tmol yr | = 0.432 s |
60° N to 90° N | 60° S to 60° N | 60° S to 90° S | ||||
---|---|---|---|---|---|---|
Longwave | Shortwave | Longwave | Shortwave | Longwave | Shortwave | |
Linear kernel | −0.0032 | 0.0016 | −0.1047 | 0.0272 | −0.0024 | 0.0016 |
Logarithmic kernel | −0.3370 | 0.1694 | −11.0373 | 2.8357 | −0.2593 | 0.1645 |
The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
Liu, Y.; Yao, X.; Zhou, L.; Ming, T.; Li, W.; de Richter, R. Removal of Atmospheric Methane by Increasing Hydroxyl Radicals via a Water Vapor Enhancement Strategy. Atmosphere 2024 , 15 , 1046. https://doi.org/10.3390/atmos15091046
Liu Y, Yao X, Zhou L, Ming T, Li W, de Richter R. Removal of Atmospheric Methane by Increasing Hydroxyl Radicals via a Water Vapor Enhancement Strategy. Atmosphere . 2024; 15(9):1046. https://doi.org/10.3390/atmos15091046
Liu, Yang, Xiaokun Yao, Li Zhou, Tingzhen Ming, Wei Li, and Renaud de Richter. 2024. "Removal of Atmospheric Methane by Increasing Hydroxyl Radicals via a Water Vapor Enhancement Strategy" Atmosphere 15, no. 9: 1046. https://doi.org/10.3390/atmos15091046
Further information, mdpi initiatives, follow mdpi.
Subscribe to receive issue release notifications and newsletters from MDPI journals
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Nature volume 632 , pages 1060–1066 ( 2024 ) Cite this article
54k Accesses
3 Citations
678 Altmetric
Metrics details
General circulation models (GCMs) are the foundation of weather and climate prediction 1 , 2 . GCMs are physics-based simulators that combine a numerical solver for large-scale dynamics with tuned representations for small-scale processes such as cloud formation. Recently, machine-learning models trained on reanalysis data have achieved comparable or better skill than GCMs for deterministic weather forecasting 3 , 4 . However, these models have not demonstrated improved ensemble forecasts, or shown sufficient stability for long-term weather and climate simulations. Here we present a GCM that combines a differentiable solver for atmospheric dynamics with machine-learning components and show that it can generate forecasts of deterministic weather, ensemble weather and climate on par with the best machine-learning and physics-based methods. NeuralGCM is competitive with machine-learning models for one- to ten-day forecasts, and with the European Centre for Medium-Range Weather Forecasts ensemble prediction for one- to fifteen-day forecasts. With prescribed sea surface temperature, NeuralGCM can accurately track climate metrics for multiple decades, and climate forecasts with 140-kilometre resolution show emergent phenomena such as realistic frequency and trajectories of tropical cyclones. For both weather and climate, our approach offers orders of magnitude computational savings over conventional GCMs, although our model does not extrapolate to substantially different future climates. Our results show that end-to-end deep learning is compatible with tasks performed by conventional GCMs and can enhance the large-scale physical simulations that are essential for understanding and predicting the Earth system.
Solving the equations for Earth’s atmosphere with general circulation models (GCMs) is the basis of weather and climate prediction 1 , 2 . Over the past 70 years, GCMs have been steadily improved with better numerical methods and more detailed physical models, while exploiting faster computers to run at higher resolution. Inside GCMs, the unresolved physical processes such as clouds, radiation and precipitation are represented by semi-empirical parameterizations. Tuning GCMs to match historical data remains a manual process 5 , and GCMs retain many persistent errors and biases 6 , 7 , 8 . The difficulty of reducing uncertainty in long-term climate projections 9 and estimating distributions of extreme weather events 10 presents major challenges for climate mitigation and adaptation 11 .
Recent advances in machine learning have presented an alternative for weather forecasting 3 , 4 , 12 , 13 . These models rely solely on machine-learning techniques, using roughly 40 years of historical data from the European Center for Medium-Range Weather Forecasts (ECMWF) reanalysis v5 (ERA5) 14 for model training and forecast initialization. Machine-learning methods have been remarkably successful, demonstrating state-of-the-art deterministic forecasts for 1- to 10-day weather prediction at a fraction of the computational cost of traditional models 3 , 4 . Machine-learning atmospheric models also require considerably less code, for example GraphCast 3 has 5,417 lines versus 376,578 lines for the National Oceanic and Atmospheric Administration’s FV3 atmospheric model 15 (see Supplementary Information section A for details).
Nevertheless, machine-learning approaches have noteworthy limitations compared with GCMs. Existing machine-learning models have focused on deterministic prediction, and surpass deterministic numerical weather prediction in terms of the aggregate metrics for which they are trained 3 , 4 . However, they do not produce calibrated uncertainty estimates 4 , which is essential for useful weather forecasts 1 . Deterministic machine-learning models using a mean-squared-error loss are rewarded for averaging over uncertainty, producing unrealistically blurry predictions when optimized for multi-day forecasts 3 , 13 . Unlike physical models, machine-learning models misrepresent derived (diagnostic) variables such as geostrophic wind 16 . Furthermore, although there has been some success in using machine-learning approaches on longer timescales 17 , 18 , these models have not demonstrated the ability to outperform existing GCMs.
Hybrid models that combine GCMs with machine learning are appealing because they build on the interpretability, extensibility and successful track record of traditional atmospheric models 19 , 20 . In the hybrid model approach, a machine-learning component replaces or corrects the traditional physical parameterizations of a GCM. Until now, the machine-learning component in such models has been trained ‘offline’, by learning parameterizations independently of their interaction with dynamics. These components are then inserted into an existing GCM. The lack of coupling between machine-learning components and the governing equations during training potentially causes serious problems, such as instability and climate drift 21 . So far, hybrid models have mostly been limited to idealized scenarios such as aquaplanets 22 , 23 . Under realistic conditions, machine-learning corrections have reduced some biases of very coarse GCMs 24 , 25 , 26 , but performance remains considerably worse than state-of-the-art models.
Here we present NeuralGCM, a fully differentiable hybrid GCM of Earth’s atmosphere. NeuralGCM is trained on forecasting up to 5-day weather trajectories sampled from ERA5. Differentiability enables end-to-end ‘online training’ 27 , with machine-learning components optimized in the context of interactions with the governing equations for large-scale dynamics, which we find enables accurate and stable forecasts. NeuralGCM produces physically consistent forecasts with accuracy comparable to best-in-class models across a range of timescales, from 1- to 15-day weather to decadal climate prediction.
A schematic of NeuralGCM is shown in Fig. 1 . The two key components of NeuralGCM are a differentiable dynamical core for solving the discretized governing dynamical equations and a learned physics module that parameterizes physical processes with a neural network, described in full detail in Methods , Supplementary Information sections B and C , and Supplementary Table 1 . The dynamical core simulates large-scale fluid motion and thermodynamics under the influence of gravity and the Coriolis force. The learned physics module (Supplementary Fig. 1 ) predicts the effect of unresolved processes, such as cloud formation, radiative transport, precipitation and subgrid-scale dynamics, on the simulated fields using a neural network.
a , Overall model structure, showing how forcings F t , noise z t (for stochastic models) and inputs y t are encoded into the model state x t . The model state is fed into the dynamical core, and alongside forcings and noise into the learned physics module. This produces tendencies (rates of change) used by an implicit–explicit ordinary differential equation (ODE) solver to advance the state in time. The new model state x t +1 can then be fed back into another time step, or decoded into model predictions. b , The learned physics module, which feeds data for individual columns of the atmosphere into a neural network used to produce physics tendencies in that vertical column.
The differentiable dynamical core in NeuralGCM allows an end-to-end training approach, whereby we advance the model multiple time steps before employing stochastic gradient descent to minimize discrepancies between model predictions and reanalysis (Supplementary Information section G.2 ). We gradually increase the rollout length from 6 hours to 5 days (Supplementary Information section G and Supplementary Table 5 ), which we found to be critical because our models are not accurate for multi-day prediction or stable for long rollouts early in training (Supplementary Information section H.6.2 and Supplementary Fig. 23 ). The extended back-propagation through hundreds of simulation steps enables our neural networks to take into account interactions between the learned physics and the dynamical core. We train deterministic and stochastic NeuralGCM models, each of which uses a distinct training protocol, described in full detail in Methods and Supplementary Table 4 .
We train a range of NeuralGCM models at horizontal resolutions with grid spacing of 2.8°, 1.4° and 0.7° (Supplementary Fig. 7 ). We evaluate the performance of NeuralGCM at a range of timescales appropriate for weather forecasting and climate simulation. For weather, we compare against the best-in-class conventional physics-based weather models, ECMWF’s high-resolution model (ECMWF-HRES) and ensemble prediction system (ECMWF-ENS), and two of the recent machine-learning-based approaches, GraphCast 3 and Pangu 4 . For climate, we compare against a global cloud-resolving model and Atmospheric Model Intercomparison Project (AMIP) runs.
Our evaluation set-up focuses on quantifying accuracy and physical consistency, following WeatherBench2 12 . We regrid all forecasts to a 1.5° grid using conservative regridding, and average over all 732 forecasts made at noon and midnight UTC in the year 2020, which was held-out from training data for all machine-learning models. NeuralGCM, GraphCast and Pangu compare with ERA5 as the ground truth, whereas ECMWF-ENS and ECMWF-HRES compare with the ECMWF operational analysis (that is, HRES at 0-hour lead time), to avoid penalizing the operational forecasts for different biases than ERA5.
We use ECMWF’s ensemble (ENS) model as a reference baseline as it achieves the best performance across the majority of lead times 12 . We assess accuracy using (1) root-mean-squared error (RMSE), (2) root-mean-squared bias (RMSB), (3) continuous ranked probability score (CRPS) and (4) spread-skill ratio, with the results shown in Fig. 2 . We provide more in-depth evaluations including scorecards, metrics for additional variables and levels and maps in Extended Data Figs. 1 and 2 , Supplementary Information section H and Supplementary Figs. 9 – 22 .
a , c , RMSE ( a ) and RMSB ( c ) for ECMWF-ENS, ECMWF-HRES, NeuralGCM-0.7°, NeuralGCM-ENS, GraphCast 3 and Pangu 4 on headline WeatherBench2 variables, as a percentage of the error of ECMWF-ENS. Deterministic and stochastic models are shown in solid and dashed lines respectively. e , g , CRPS relative to ECMWF-ENS ( e ) and spread-skill ratio for the ENS and NeuralGCM-ENS models ( g ). b , d , f , h , Spatial distributions of RMSE ( b ), bias ( d ), CRPS ( f ) and spread-skill ratio ( h ) for NeuralGCM-ENS and ECMWF-ENS models for 10-day forecasts of specific humidity at 700 hPa. Spatial plots of RMSE and CRPS show skill relative to a probabilistic climatology 12 with an ensemble member for each of the years 1990–2019. The grey areas indicate regions where climatological surface pressure on average is below 700 hPa.
Deterministic models that produce a single weather forecast for given initial conditions can be compared effectively using RMSE skill at short lead times. For the first 1–3 days, depending on the atmospheric variable, RMSE is minimized by forecasts that accurately track the evolution of weather patterns. At this timescale we find that NeuralGCM-0.7° and GraphCast achieve best results, with slight variations across different variables (Fig. 2a ). At longer lead times, RMSE rapidly increases owing to chaotic divergence of nearby weather trajectories, making RMSE less informative for deterministic models. RMSB calculates persistent errors over time, which provides an indication of how models would perform at much longer lead times. Here NeuralGCM models also compare favourably against previous approaches (Fig. 2c ), with notably much less bias for specific humidity in the tropics (Fig. 2d ).
Ensembles are essential for capturing intrinsic uncertainty of weather forecasts, especially at longer lead times. Beyond about 7 days, the ensemble means of ECMWF-ENS and NeuralGCM-ENS forecasts have considerably lower RMSE than the deterministic models, indicating that these models better capture the average of possible weather. A better metric for ensemble models is CRPS, which is a proper scoring rule that is sensitive to full marginal probability distributions 28 . Our stochastic model (NeuralGCM-ENS) running at 1.4° resolution has lower error compared with ECMWF-ENS across almost all variables, lead times and vertical levels for ensemble-mean RMSE, RSMB and CRPS (Fig. 2a,c,e and Supplementary Information section H ), with similar spatial patterns of skill (Fig. 2b,f ). Like ECMWF-ENS, NeuralGCM-ENS has a spread-skill ratio of approximately one (Fig. 2d ), which is a necessary condition for calibrated forecasts 29 .
An important characteristic of forecasts is their resemblance to realistic weather patterns. Figure 3 shows a case study that illustrates the performance of NeuralGCM on three types of important weather phenomenon: tropical cyclones, atmospheric rivers and the Intertropical Convergence Zone. Figure 3a shows that all the machine-learning models make significantly blurrier forecasts than the source data ERA5 and physics-based ECMWF-HRES forecast, but NeuralCGM-0.7° outperforms the pure machine-learning models, despite its coarser resolution (0.7° versus 0.25° for GraphCast and Pangu). Blurry forecasts correspond to physically inconsistent atmospheric conditions and misrepresent extreme weather. Similar trends hold for other derived variables of meteorological interest (Supplementary Information section H.2 ). Ensemble-mean predictions, from both NeuralGCM and ECMWF, are closer to ERA5 in an average sense, and thus are inherently smooth at long lead times. In contrast, as shown in Fig. 3 and in Supplementary Information section H.3 , individual realizations from the ECMWF and NeuralGCM ensembles remain sharp, even at long lead times. Like ECMWF-ENS, NeuralGCM-ENS produces a statistically representative range of future weather scenarios for each weather phenomenon, despite its eight-times-coarser resolution.
All forecasts are initialized at 2020-08-22T12z, chosen to highlight Hurricane Laura, the most damaging Atlantic hurricane of 2020. a , Specific humidity at 700 hPa for 1-day, 5-day and 10-day forecasts over North America and the Northeast Pacific Ocean from ERA5 14 , ECMWF-HRES, NeuralGCM-0.7°, ECMWF-ENS (mean), NeuralGCM-ENS (mean), GraphCast 3 and Pangu 4 . b , Forecasts from individual ensemble members from ECMWF-ENS and NeuralGCM-ENS over regions of interest, including predicted tracks of Hurricane Laura from each of the 50 ensemble members (Supplementary Information section I.2 ). The track from ERA5 is plotted in black.
We can quantify the blurriness of different forecast models via their power spectra. Supplementary Figs. 17 and 18 show that the power spectra of NeuralCGM-0.7° is consistently closer to ERA5 than the other machine-learning forecast methods, but is still blurrier than ECMWF’s physical forecasts. The spectra of NeuralGCM forecasts is also roughly constant over the forecast period, in stark contrast to GraphCast, which worsens with lead time. The spectrum of NeuralGCM becomes more accurate with increased resolution (Supplementary Fig. 22 ), which suggests the potential for further improvements of NeuralGCM models trained at higher resolutions.
In NeuralGCM, advection is handled by the dynamical core, while the machine-learning parameterization models local processes within vertical columns of the atmosphere. Thus, unlike pure machine-learning methods, local sources and sinks can be isolated from tendencies owing to horizontal transport and other resolved dynamics (Supplementary Fig. 3 ). This makes our results more interpretable and facilitates the diagnosis of the water budget. Specifically, we diagnose precipitation minus evaporation (Supplementary Information section H.5 ) rather than directly predicting these as in machine-learning-based approaches 3 . For short weather forecasts, the mean of precipitation minus evaporation has a realistic spatial distribution that is very close to ERA5 data (Extended Data Fig. 4c–e ). The precipitation-minus-evaporation rate distribution of NeuralGCM-0.7° closely matches the ERA5 distribution in the extratropics (Extended Data Fig. 4b ), although it underestimates extreme events in the tropics (Extended Data Fig. 4a ). It is noted that the current version of NeuralGCM directly predicts tendencies for an atmospheric column, and thus cannot distinguish between precipitation and evaporation.
We examined the extent to which NeuralGCM, GraphCast and ECMWF-HRES capture the geostrophic wind balance, the near-equilibrium between the dominant forces that drive large-scale dynamics in the mid-latitudes 30 . A recent study 16 highlighted that Pangu misrepresents the vertical structure of the geostrophic and ageostrophic winds and noted a deterioration at longer lead times. Similarly, we observe that GraphCast shows an error that worsens with lead time. In contrast, NeuralGCM more accurately depicts the vertical structure of the geostrophic and ageostrophic winds, as well as their ratio, compared with GraphCast across various rollouts, when compared against ERA5 data (Extended Data Fig. 3 ). However, ECMWF-HRES still shows a slightly closer alignment to ERA5 data than NeuralGCM does. Within NeuralGCM, the representation of the geostrophic wind’s vertical structure only slightly degrades in the initial few days, showing no noticeable changes thereafter, particularly beyond day 5.
Physically consistent weather models should still perform well for weather conditions for which they were not trained. We expect that NeuralGCM may generalize better than machine-learning-only atmospheric models, because NeuralGCM employs neural networks that act locally in space, on individual vertical columns of the atmosphere. To explore this hypothesis, we compare versions of NeuralCGM-0.7° and GraphCast trained to 2017 on 5 years of weather forecasts beyond the training period (2018–2022) in Supplementary Fig. 36 . Unlike GraphCast, NeuralGCM does not show a clear trend of increasing error when initialized further into the future from the training data. To extend this test beyond 5 years, we trained a NeuralGCM-2.8° model using only data before 2000, and tested its skill for over 21 unseen years (Supplementary Fig. 35 ).
Although our deterministic NeuralGCM models are trained to predict weather up to 3 days ahead, they are generally capable of simulating the atmosphere far beyond medium-range weather timescales. For extended climate simulations, we prescribe historical sea surface temperature (SST) and sea-ice concentration. These simulations feature many emergent phenomena of the atmosphere on timescales from months to decades.
For climate simulations with NeuralGCM, we use 2.8° and 1.4° deterministic models, which are relatively inexpensive to train (Supplementary Information section G.7 ) and allow us to explore a larger parameter space to find stable models. Previous studies found that running extended simulations with hybrid models is challenging due to numerical instabilities and climate drift 21 . To quantify stability in our selected models, we run multiple initial conditions and report how many of them finish without instability.
To assess the capability of NeuralGCM to simulate various aspects of the seasonal cycle, we run 2-year simulations with NeuralGCM-1.4°. for 37 different initial conditions spaced every 10 days for the year 2019. Out of these 37 initial conditions, 35 successfully complete the full 2 years without instability; for case studies of instability, see Supplementary Information section H.7 , and Supplementary Figs. 26 and 27 . We compare results from NeuralGCM-1.4° for 2020 with ERA5 data and with outputs from the X-SHiELD global cloud-resolving model, which is coupled to an ocean model nudged towards reanalysis 31 . This X-SHiELD run has been used as a target for training machine-learning climate models 24 . For comparison, we evaluate models after regridding predictions to 1.4° resolution. This comparison slightly favours NeuralGCM because NeuralGCM was tuned to match ERA5, but the discrepancy between ERA5 and the actual atmosphere is small relative to model error.
Figure 4a shows the temporal variation of the global mean temperature to 2020, as captured by 35 simulations from NeuralGCM, in comparison with the ERA5 reanalysis and standard climatology benchmarks. The seasonality and variability of the global mean temperature from NeuralGCM are quantitatively similar to those observed in ERA5. The ensemble-mean temperature RMSE for NeuralGCM stands at 0.16 K when benchmarked against ERA5, which is a significant improvement over the climatology’s RMSE of 0.45 K. We find that NeuralGCM accurately simulates the seasonal cycle, as evidenced by metrics such as the annual cycle of the global precipitable water (Supplementary Fig. 30a ) and global total kinetic energy (Supplementary Fig. 30b ). Furthermore, the model captures essential atmospheric dynamics, including the Hadley circulation and the zonal-mean zonal wind (Supplementary Fig. 28 ), as well as the spatial patterns of eddy kinetic energy in different seasons (Supplementary Fig. 31 ), and the distinctive seasonal behaviours of monsoon circulation (Supplementary Fig. 29 ; additional details are provided in Supplementary Information section I.1 ).
a , Global mean temperature for ERA5 14 (orange), 1990–2019 climatology (black) and NeuralGCM-1.4° (blue) for 2020 using 35 simulations initialized every 10 days during 2019 (thick line, ensemble mean; thin lines, different initial conditions). b , Yearly global mean temperature for ERA5 (orange), mean over 22 CMIP6 AMIP experiments 34 (violet; model details are in Supplementary Information section I.3 ) and NeuralGCM-2.8° for 22 AMIP-like simulations with prescribed SST initialized every 10 days during 1980 (thick line, ensemble mean; thin lines, different initial conditions). c , The RMSB of the 850-hPa temperature averaged between 1981 and 2014 for 22 NeuralGCM-2.8° AMIP runs (labelled NGCM), 22 CMIP6 AMIP experiments (labelled AMIP) and debiased 22 CMIP6 AMIP experiments (labelled AMIP*; bias was removed by removing the 850-hPa global temperature bias). In the box plots, the red line represents the median. The box delineates the first to third quartiles; the whiskers extend to 1.5 times the interquartile range (Q1 − 1.5IQR and Q3 + 1.5IQR), and outliers are shown as individual dots. d , Vertical profiles of tropical (20° S–20° N) temperature trends for 1981–2014. Orange, ERA5; black dots, Radiosonde Observation Correction using Reanalyses (RAOBCORE) 41 ; blue dots, mean trends for NeuralGCM; purple dots, mean trends from CMIP6 AMIP runs (grey and black whiskers, 25th and 75th percentiles for NeuralGCM and CMIP6 AMIP runs, respectively). e – g , Tropical cyclone tracks for ERA5 ( e ), NeuralGCM-1.4° ( f ) and X-SHiELD 31 ( g ). h – k , Mean precipitable water for ERA5 ( h ) and the precipitable water bias in NeuralGCM-1.4° ( i ), initialized 90 days before mid-January 2020 similarly to X-SHiELD, X-SHiELD ( j ) and climatology ( k ; averaged between 1990 and 2019). In d – i , quantities are calculated between mid-January 2020 and mid-January 2021 and all models were regridded to a 256 × 128 Gaussian grid before computation and tracking.
Next, we compare the annual biases of a single NeuralGCM realization with a single realization of X-SHiELD (the only one available), both initiated in mid-October 2019. We consider 19 January 2020 to 17 January 2021, the time frame for which X-SHiELD data are available. Global cloud-resolving models, such as X-SHiELD, are considered state of the art, especially for simulating the hydrological cycle, owing to their resolution being capable of resolving deep convection 32 . The annual bias in precipitable water for NeuralGCM (RMSE of 1.09 mm) is substantially smaller than the biases of both X-SHiELD (RMSE of 1.74 mm) and climatology (RMSE of 1.36 mm; Fig. 4i–k ). Moreover, NeuralGCM shows a lower temperature bias in the upper and lower troposphere than X-SHiELD (Extended Data Fig. 6 ). We also indirectly compare precipitation bias in X-SHiELD with precipitation-minus-evaporation bias in NeuralGCM-1.4°, which shows slightly larger bias and grid-scale artefacts for NeuralGCM (Extended Data Fig. 5 ).
Finally, to assess the capability of NeuralGCM to generate tropical cyclones in an annual model integration, we use the tropical cyclone tracker TempestExtremes 33 , as described in Supplementary Information section I.2 , Supplementary Fig. 34 and Supplementary Table 6 . Figure 4e–g shows that NeuralGCM, even at a coarse resolution of 1.4°, produces realistic trajectories and counts of tropical cyclone (83 versus 86 in ERA5 for the corresponding period), whereas X-SHiELD, when regridded to 1.4° resolution, substantially underestimates the tropical cyclone count (40). Additional statistical analyses of tropical cyclones can be found in Extended Data Figs. 7 and 8 .
To assess the capability of NeuralGCM to simulate historical temperature trends, we conduct AMIP-like simulations over a duration of 40 years with NeuralGCM-2.8°. Out of 37 different runs with initial conditions spaced every 10 days during the year 1980, 22 simulations were stable for the entire 40-year period, and our analysis focuses on these results. We compare with 22 simulations run with prescribed SST from the Coupled Model Intercomparison Project Phase 6 (CMIP6) 34 , listed in Supplementary Information section I.3 .
We find that all 40-year simulations of NeuralGCM, as well as the mean of the 22 AMIP runs, accurately capture the global warming trends observed in ERA5 data (Fig. 4b ). There is a strong correlation in the year-to-year temperature trends with ERA5 data, suggesting that NeuralGCM effectively captures the impact of SST forcing on climate. When comparing spatial biases averaged over 1981–2014, we find that all 22 NeuralGCM-2.8° runs have smaller bias than the CMIP6 AMIP runs, and this result remains even when removing the global temperature bias in CMIP6 AMIP runs (Fig. 4c and Supplementary Figs. 32 and 33 ).
Next, we investigated the vertical structure of tropical warming trends, which climate models tend to overestimate in the upper troposphere 35 . As shown in Fig. 4d , the trends, calculated by linear regression, of NeuralGCM are closer to ERA5 than those of AMIP runs. In particular, the bias in the upper troposphere is reduced. However, NeuralGCM does show a wider spread in its predictions than the AMIP runs, even at levels near the surface where temperatures are typically more constrained by prescribed SST.
Lastly, we evaluated NeuralGCM’s capability to generalize to unseen warmer climates by conducting AMIP simulations with increased SST (Supplementary Information section I.4.2 ). We find that NeuralGCM shows some of the robust features of climate warming response to modest SST increases (+1 K and +2 K); however, for more substantial SST increases (+4 K), NeuralGCM’s response diverges from expectations (Supplementary Fig. 37 ). In addition, AMIP simulations with increased SST show climate drift, underscoring NeuralGCM’s limitations in this context (Supplementary Fig. 38 ).
NeuralGCM is a differentiable hybrid atmospheric model that combines the strengths of traditional GCMs with machine learning for weather forecasting and climate simulation. To our knowledge, NeuralGCM is the first machine-learning-based model to make accurate ensemble weather forecasts, with better CRPS than state-of-the-art physics-based models. It is also, to our knowledge, the first hybrid model that achieves comparable spatial bias to global cloud-resolving models, can simulate realistic tropical cyclone tracks and can run AMIP-like simulations with realistic historical temperature trends. Overall, NeuralGCM demonstrates that incorporating machine learning is a viable alternative to building increasingly detailed physical models 32 for improving GCMs.
Compared with traditional GCMs with similar skill, NeuralGCM is computationally efficient and low complexity. NeuralGCM runs at 8- to 40-times-coarser horizontal resolution than ECMWF’s Integrated Forecasting System and global cloud-resolving models, which enables 3 to 5 orders of magnitude savings in computational resources. For example, NeuralGCM-1.4° simulates 70,000 simulation days in 24 hours using a single tensor-processing-unit versus 19 simulated days on 13,824 central-processing-unit cores with X-SHiELD (Extended Data Table 1 ). This can be leveraged for previously impractical tasks such as large ensemble forecasting. NeuralGCM’s dynamical core uses global spectral methods 36 , and learned physics is parameterized with fully connected neural networks acting on single vertical columns. Substantial headroom exists to pursue higher accuracy using advanced numerical methods and machine-learning architectures.
Our results provide strong evidence for the disputed hypothesis 37 , 38 , 39 that learning to predict short-term weather is an effective way to tune parameterizations for climate. NeuralGCM models trained on 72-hour forecasts are capable of realistic multi-year simulation. When provided with historical SSTs, they capture essential atmospheric dynamics such as seasonal circulation, monsoons and tropical cyclones. However, we will probably need alternative training strategies 38 , 39 to learn important processes for climate with subtle impacts on weather timescales, such as a cloud feedback.
The NeuralGCM approach is compatible with incorporating either more physics or more machine learning, as required for operational weather forecasts and climate simulations. For weather forecasting, we expect that end-to-end learning 40 with observational data will allow for better and more relevant predictions, including key variables such as precipitation. Such models could include neural networks acting as corrections to traditional data assimilation and model diagnostics. For climate projection, NeuralGCM will need to be reformulated to enable coupling with other Earth-system components (for example, ocean and land), and integrating data on the atmospheric chemical composition (for example, greenhouse gases and aerosols). There are also research challenges common to current machine-learning-based climate models 19 , including the capability to simulate unprecedented climates (that is, generalization), adhering to physical constraints, and resolving numerical instabilities and climate drift. NeuralGCM’s flexibility to incorporate physics-based models (for example, radiation) offers a promising avenue to address these challenges.
Models based on physical laws and empirical relationships are ubiquitous in science. We believe the differentiable hybrid modelling approach of NeuralGCM has the potential to transform simulation for a wide range of applications, such as materials discovery, protein folding and multiphysics engineering design.
NeuralGCM combines components of the numerical solver and flexible neural network parameterizations. Simulation in time is carried out in a coordinate system suitable for solving the dynamical equations of the atmosphere, describing large-scale fluid motion and thermodynamics under the influence of gravity and the Coriolis force.
Our differentiable dynamical core is implemented in JAX, a library for high-performance code in Python that supports automatic differentiation 42 . The dynamical core solves the hydrostatic primitive equations with moisture, using a horizontal pseudo-spectral discretization and vertical sigma coordinates 36 , 43 . We evolve seven prognostic variables: vorticity and divergence of horizontal wind, temperature, surface pressure, and three water species (specific humidity, and specific ice and liquid cloud water content).
Our learned physics module uses the single-column approach of GCMs 2 , whereby information from only a single atmospheric column is used to predict the impact of unresolved processes occurring within that column. These effects are predicted using a fully connected neural network with residual connections, with weights shared across all atmospheric columns (Supplementary Information section C.4 ).
The inputs to the neural network include the prognostic variables in the atmospheric column, total incident solar radiation, sea-ice concentration and SST (Supplementary Information section C.1 ). We also provide horizontal gradients of the prognostic variables, which we found improves performance 44 . All inputs are standardized to have zero mean and unit variance using statistics precomputed during model initialization. The outputs are the prognostic variable tendencies scaled by the fixed unconditional standard deviation of the target field (Supplementary Information section C.5 ).
To interface between ERA5 14 data stored in pressure coordinates and the sigma coordinate system of our dynamical core, we introduce encoder and decoder components (Supplementary Information section D ). These components perform linear interpolation between pressure levels and sigma coordinate levels. We additionally introduce learned corrections to both encoder and decoder steps (Supplementary Figs. 4–6 ), using the same column-based neural network architecture as the learned physics module. Importantly, the encoder enables us to eliminate the gravity waves from initialization shock 45 , which otherwise contaminate forecasts.
Figure 1a shows the sequence of steps that NeuralGCM takes to make a forecast. First, it encodes ERA5 data at t = t 0 on pressure levels to initial conditions on sigma coordinates. To perform a time step, the dynamical core and learned physics (Fig. 1b ) then compute tendencies, which are integrated in time using an implicit–explicit ordinary differential equation solver 46 (Supplementary Information section E and Supplementary Table 2 ). This is repeated to advance the model from t = t 0 to t = t final . Finally, the decoder converts predictions back to pressure levels.
The time-step size of the ODE solver (Supplementary Table 3 ) is limited by the Courant–Friedrichs–Lewy condition on dynamics, and can be small relative to the timescale of atmospheric change. Evaluating learned physics is approximately 1.5 times as expensive as a time step of the dynamical core. Accordingly, following the typical practice for GCMs, we hold learned physics tendencies constant for multiple ODE time steps to reduce computational expense, typically corresponding to 30 minutes of simulation time.
We train deterministic NeuralGCM models using a combination of three loss functions (Supplementary Information section G.4 ) to encourage accuracy and sharpness while penalizing bias. During the main training phase, all losses are defined in a spherical harmonics basis. We use a standard mean squared error loss for prompting accuracy, modified to progressively filter out contributions from higher total wavenumbers at longer lead times (Supplementary Fig. 8 ). This filtering approach tackles the ‘double penalty problem’ 47 as it prevents the model from being penalized for predicting high-wavenumber features in incorrect locations at later times, especially beyond the predictability horizon. A second loss term encourages the spectrum to match the training data using squared loss on the total wavenumber spectrum of prognostic variables. These first two losses are evaluated on both sigma and pressure levels. Finally, a third loss term discourages bias by adding mean squared error on the batch-averaged mean amplitude of each spherical harmonic coefficient. For analysis of the impact that various loss functions have, refer to Supplementary Information section H.6.1 , and Supplementary Figs. 23 and 24 . The combined action of the three training losses allow the resulting models trained on 3-day rollouts to remain stable during years-to-decades-long climate simulations. Before final evaluations, we perform additional fine-tuning of just the decoder component on short rollouts of 24 hours (Supplementary Information section G.5 ).
Stochastic NeuralGCM models incorporate inherent randomness in the form of additional random fields passed as inputs to neural network components. Our stochastic loss is based on the CRPS 28 , 48 , 49 . CRPS consists of mean absolute error that encourages accuracy, balanced by a similar term that encourages ensemble spread. For each variable we use a sum of CRPS in grid space and CRPS in the spherical harmonic basis below a maximum cut-off wavenumber (Supplementary Information section G.6 ). We compute CRPS on rollout lengths from 6 hours to 5 days. As illustrated in Fig. 1 , we inject noise to the learned encoder and the learned physics module by sampling from Gaussian random fields with learned spatial and temporal correlation (Supplementary Information section C.2 and Supplementary Fig. 2 ). For training, we generate two ensemble members per forecast, which suffices for an unbiased estimate of CRPS.
For training and evaluating the NeuralGCM models, we used the publicly available ERA5 dataset 14 , originally downloaded from https://cds.climate.copernicus.eu/ and available via Google Cloud Storage in Zarr format at gs://gcp-public-data-arco-era5/ar/full_37-1h-0p25deg-chunk-1.zarr-v3. To compare NeuralGCM with operational and data-driven weather models, we used forecast datasets distributed as part of WeatherBench2 12 at https://weatherbench2.readthedocs.io/en/latest/data-guide.html , to which we have added NeuralGCM forecasts for 2020. To compare NeuralGCM with atmospheric models in climate settings, we used CMIP6 data available at https://catalog.pangeo.io/browse/master/climate/ , as well as X-SHiELD 24 outputs available on Google Cloud storage in a ‘requester pays’ bucket at gs://ai2cm-public-requester-pays/C3072-to-C384-res-diagnostics. The Radiosonde Observation Correction using Reanalyses (RAOBCORE) V1.9 that was used as reference tropical temperature trends was downloaded from https://webdata.wolke.img.univie.ac.at/haimberger/v1.9/ . Base maps use freely available data from https://www.naturalearthdata.com/downloads/ .
The NeuralGCM code base is separated into two open source projects: Dinosaur and NeuralGCM, both publicly available on GitHub at https://github.com/google-research/dinosaur (ref. 50 ) and https://github.com/google-research/neuralgcm (ref. 51 ). The Dinosaur package implements a differentiable dynamical core used by NeuralGCM, whereas the NeuralGCM package provides machine-learning models and checkpoints of trained models. Evaluation code for NeuralGCM weather forecasts is included in WeatherBench2 12 , available at https://github.com/google-research/weatherbench2 (ref. 52 ).
Bauer, P., Thorpe, A. & Brunet, G. The quiet revolution of numerical weather prediction. Nature 525 , 47–55 (2015).
Article ADS CAS PubMed Google Scholar
Balaji, V. et al. Are general circulation models obsolete? Proc. Natl Acad. Sci. USA 119 , e2202075119 (2022).
Article CAS PubMed PubMed Central Google Scholar
Lam, R. et al. Learning skillful medium-range global weather forecasting. Science 382 , 1416–1421 (2023).
Article ADS MathSciNet CAS PubMed Google Scholar
Bi, K. et al. Accurate medium-range global weather forecasting with 3D neural networks. Nature 619 , 533–538 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Hourdin, F. et al. The art and science of climate model tuning. Bull. Am. Meteorol. Soc. 98 , 589–602 (2017).
Article ADS Google Scholar
Bony, S. & Dufresne, J.-L. Marine boundary layer clouds at the heart of tropical cloud feedback uncertainties in climate models. Geophys. Res. Lett. 32 , L20806 (2005).
Webb, M. J., Lambert, F. H. & Gregory, J. M. Origins of differences in climate sensitivity, forcing and feedback in climate models. Clim. Dyn. 40 , 677–707 (2013).
Article Google Scholar
Sherwood, S. C., Bony, S. & Dufresne, J.-L. Spread in model climate sensitivity traced to atmospheric convective mixing. Nature 505 , 37–42 (2014).
Article ADS PubMed Google Scholar
Palmer, T. & Stevens, B. The scientific challenge of understanding and estimating climate change. Proc. Natl Acad. Sci. USA 116 , 24390–24395 (2019).
Fischer, E. M., Beyerle, U. & Knutti, R. Robust spatially aggregated projections of climate extremes. Nat. Clim. Change 3 , 1033–1038 (2013).
Field, C. B. Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation: Special Report of the Intergovernmental Panel on Climate Change (Cambridge Univ. Press, 2012).
Rasp, S. et al. WeatherBench 2: A benchmark for the next generation of data-driven global weather models. J. Adv. Model. Earth Syst. 16 , e2023MS004019 (2024).
Keisler, R. Forecasting global weather with graph neural networks. Preprint at https://arxiv.org/abs/2202.07575 (2022).
Hersbach, H. et al. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 146 , 1999–2049 (2020).
Zhou, L. et al. Toward convective-scale prediction within the next generation global prediction system. Bull. Am. Meteorol. Soc. 100 , 1225–1243 (2019).
Bonavita, M. On some limitations of current machine learning weather prediction models. Geophys. Res. Lett. 51 , e2023GL107377 (2024).
Weyn, J. A., Durran, D. R. & Caruana, R. Improving data-driven global weather prediction using deep convolutional neural networks on a cubed sphere. J. Adv. Model. Earth Syst. 12 , e2020MS002109 (2020).
Watt-Meyer, O. et al. ACE: a fast, skillful learned global atmospheric model for climate prediction. Preprint at https://arxiv.org/abs/2310.02074 (2023).
Bretherton, C. S. Old dog, new trick: reservoir computing advances machine learning for climate modeling. Geophys. Res. Lett. 50 , e2023GL104174 (2023).
Reichstein, M. et al. Deep learning and process understanding for data-driven Earth system science. Nature 566 , 195–204 (2019).
Brenowitz, N. D. & Bretherton, C. S. Spatially extended tests of a neural network parametrization trained by coarse-graining. J. Adv. Model. Earth Syst. 11 , 2728–2744 (2019).
Rasp, S., Pritchard, M. S. & Gentine, P. Deep learning to represent subgrid processes in climate models. Proc. Natl Acad. Sci. USA 115 , 9684–9689 (2018).
Yuval, J. & O’Gorman, P. A. Stable machine-learning parameterization of subgrid processes for climate modeling at a range of resolutions. Nat. Commun. 11 , 3295 (2020).
Kwa, A. et al. Machine-learned climate model corrections from a global storm-resolving model: performance across the annual cycle. J. Adv. Model. Earth Syst. 15 , e2022MS003400 (2023).
Arcomano, T., Szunyogh, I., Wikner, A., Hunt, B. R. & Ott, E. A hybrid atmospheric model incorporating machine learning can capture dynamical processes not captured by its physics-based component. Geophys. Res. Lett. 50 , e2022GL102649 (2023).
Han, Y., Zhang, G. J. & Wang, Y. An ensemble of neural networks for moist physics processes, its generalizability and stable integration. J. Adv. Model. Earth Syst. 15 , e2022MS003508 (2023).
Gelbrecht, M., White, A., Bathiany, S. & Boers, N. Differentiable programming for Earth system modeling. Geosci. Model Dev. 16 , 3123–3135 (2023).
Gneiting, T. & Raftery, A. E. Strictly proper scoring rules, prediction, and estimation. J. Am. Stat. Assoc. 102 , 359–378 (2007).
Article MathSciNet CAS Google Scholar
Fortin, V., Abaza, M., Anctil, F. & Turcotte, R. Why should ensemble spread match the RMSE of the ensemble mean? J. Hydrometeorol. 15 , 1708–1713 (2014).
Holton, J. R. An introduction to Dynamic Meteorology 5th edn (Elsevier, 2004).
Cheng, K.-Y. et al. Impact of warmer sea surface temperature on the global pattern of intense convection: insights from a global storm resolving model. Geophys. Res. Lett. 49 , e2022GL099796 (2022).
Stevens, B. et al. DYAMOND: the dynamics of the atmospheric general circulation modeled on non-hydrostatic domains. Prog. Earth Planet. Sci. 6 , 61 (2019).
Ullrich, P. A. et al. TempestExtremes v2.1: a community framework for feature detection, tracking, and analysis in large datasets. Geosc. Model Dev. 14 , 5023–5048 (2021).
Eyring, V. et al. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev. 9 , 1937–1958 (2016).
Mitchell, D. M., Lo, Y. E., Seviour, W. J., Haimberger, L. & Polvani, L. M. The vertical profile of recent tropical temperature trends: persistent model biases in the context of internal variability. Environ. Res. Lett. 15 , 1040b4 (2020).
Bourke, W. A multi-level spectral model. I. Formulation and hemispheric integrations. Mon. Weather Rev. 102 , 687–701 (1974).
Ruiz, J. J., Pulido, M. & Miyoshi, T. Estimating model parameters with ensemble-based data assimilation: a review. J. Meteorol. Soc. Jpn Ser. II 91 , 79–99 (2013).
Schneider, T., Lan, S., Stuart, A. & Teixeira, J. Earth system modeling 2.0: a blueprint for models that learn from observations and targeted high-resolution simulations. Geophys. Res. Lett. 44 , 12–396 (2017).
Schneider, T., Leung, L. R. & Wills, R. C. J. Opinion: Optimizing climate models with process knowledge, resolution, and artificial intelligence. Atmos. Chem. Phys. 24 , 7041–7062 (2024).
Sutskever, I., Vinyals, O. & Le, Q. V. Sequence to sequence learning with neural networks. Adv. Neural Inf. Process. Syst. 27 , 3104–3112 (2014).
Haimberger, L., Tavolato, C. & Sperka, S. Toward elimination of the warm bias in historic radiosonde temperature records—some new results from a comprehensive intercomparison of upper-air data. J. Clim. 21 , 4587–4606 (2008).
Bradbury, J. et al. JAX: composable transformations of Python+NumPy programs. GitHub http://github.com/google/jax (2018).
Durran, D. R. Numerical Methods for Fluid Dynamics: With Applications to Geophysics Vol. 32, 2nd edn (Springer, 2010).
Wang, P., Yuval, J. & O’Gorman, P. A. Non-local parameterization of atmospheric subgrid processes with neural networks. J. Adv. Model. Earth Syst. 14 , e2022MS002984 (2022).
Daley, R. Normal mode initialization. Rev. Geophys. 19 , 450–468 (1981).
Whitaker, J. S. & Kar, S. K. Implicit–explicit Runge–Kutta methods for fast–slow wave problems. Mon. Weather Rev. 141 , 3426–3434 (2013).
Gilleland, E., Ahijevych, D., Brown, B. G., Casati, B. & Ebert, E. E. Intercomparison of spatial forecast verification methods. Weather Forecast. 24 , 1416–1430 (2009).
Rasp, S. & Lerch, S. Neural networks for postprocessing ensemble weather forecasts. Month. Weather Rev. 146 , 3885–3900 (2018).
Pacchiardi, L., Adewoyin, R., Dueben, P. & Dutta, R. Probabilistic forecasting with generative networks via scoring rule minimization. J. Mach. Learn. Res. 25 , 1–64 (2024).
Smith, J. A., Kochkov, D., Norgaard, P., Yuval, J. & Hoyer, S. google-research/dinosaur: 1.0.0. Zenodo https://doi.org/10.5281/zenodo.11376145 (2024).
Kochkov, D. et al. google-research/neuralgcm: 1.0.0. Zenodo https://doi.org/10.5281/zenodo.11376143 (2024).
Rasp, S. et al. google-research/weatherbench2: v0.2.0. Zenodo https://doi.org/10.5281/zenodo.11376271 (2023).
Download references
We thank A. Kwa, A. Merose and K. Shah for assistance with data acquisition and handling; L. Zepeda-Núñez for feedback on the paper; and J. Anderson, C. Van Arsdale, R. Chemke, G. Dresdner, J. Gilmer, J. Hickey, N. Lutsko, G. Nearing, A. Paszke, J. Platt, S. Ponda, M. Pritchard, D. Rothenberg, F. Sha, T. Schneider and O. Voicu for discussions.
These authors contributed equally: Dmitrii Kochkov, Janni Yuval, Ian Langmore, Peter Norgaard, Jamie Smith, Stephan Hoyer
Google Research, Mountain View, CA, USA
Dmitrii Kochkov, Janni Yuval, Ian Langmore, Peter Norgaard, Jamie Smith, Griffin Mooers, James Lottes, Stephan Rasp, Michael P. Brenner & Stephan Hoyer
Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA, USA
Milan Klöwer
European Centre for Medium-Range Weather Forecasts, Reading, UK
Peter Düben & Sam Hatfield
Google DeepMind, London, UK
Peter Battaglia, Alvaro Sanchez-Gonzalez & Matthew Willson
School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA
Michael P. Brenner
You can also search for this author in PubMed Google Scholar
D.K., J.Y., I.L., P.N., J.S. and S. Hoyer contributed equally to this work. D.K., J.Y., I.L., P.N., J.S., G.M., J.L. and S. Hoyer wrote the code. D.K., J.Y., I.L., P.N., G.M. and S. Hoyer trained models and analysed the data. M.P.B. and S. Hoyer managed and oversaw the research project. M.K., S.R., P.D., S. Hatfield, P.B. and M.P.B. contributed technical advice and ideas. M.W. ran experiments with GraphCast for comparison with NeuralGCM. A.S.-G. assisted with data preparation. D.K., J.Y., I.L., P.N. and S. Hoyer wrote the paper. All authors gave feedback and contributed to editing the paper.
Correspondence to Dmitrii Kochkov , Janni Yuval or Stephan Hoyer .
Competing interests.
D.K., J.Y., I.L., P.N., J.S., J.L., S.R., P.B., A.S.-G., M.W., M.P.B. and S. Hoyer are employees of Google. S. Hoyer, D.K., I.L., J.Y., G.M., P.N., J.S. and M.B. have filed international patent application PCT/US2023/035420 in the name of Google LLC, currently pending, relating to neural general circulation models.
Peer review information.
Nature thanks Karthik Kashinath and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data fig. 1 maps of bias for neuralgcm-ens and ecmwf-ens forecasts..
Bias is averaged over all forecasts initialized in 2020.
Spread-skill ratio is averaged over all forecasts initialized in 2020.
Vertical profiles of the extratropical intensity (averaged between latitude 30°–70° in both hemispheres) and over all forecasts initialized in 2020 of (a,d,g) geostrophic wind, (b,e,h) ageostrophic wind and (c,f,i) the ratio of the intensity of ageostrophic wind over geostrophic wind for ERA5 (black continuous line in all panels), (a,b,c) NeuralGCM-0.7°, (d,e,f) GraphCast and (g,h,i) ECMWF-HRES at lead times of 1 day, 5 days and 10 days.
(a) Tropical (latitudes −20° to 20°) precipitation minus evaporation (P minus E) rate distribution, (b) Extratropical (latitudes 30° to 70° in both hemispheres) P minus E, (c) mean P minus E for 2020 ERA5 14 and (d) NeuralGCM-0.7° (calculated from the third day of forecasts and averaged over all forecasts initialized in 2020), (e) the bias between NeuralGCM-0.7° and ERA5, (f-g) Snapshot of daily precipitation minus evaporation for 2020-01-04 for (f) NeuralGCM-0.7° (forecast initialized on 2020-01-02) and (g) ERA5.
Mean precipitation calculated between 2020-01-19 and 2021-01-17 for (a) ERA5 14 (c) X-SHiELD 31 and the biases in (e) X-SHiELD and (g) climatology (ERA5 data averaged over 1990-2019). Mean precipitation minus evaporation calculated between 2020-01-19 and 2021-01-17 for (b) ERA5 (d) NeuralGCM-1.4° (initialized in October 18th 2019) and the biases in (f) NeuralGCM-1.4° and (h) climatology (data averaged over 1990–2019).
Mean temperature between 2020-01-19 to 2020-01-17 for (a) ERA5 at 200hPa and (b) 850hPa. (c,d) the bias in the temperature for NeuralGCM-1.4°, (e,f) the bias in X-SHiELD and (g,h) the bias in climatology (calculated from 1990–2019). NeuralGCM-1.4° was initialized in 18th of October (similar to X-SHiELD).
(a) Tropical Cyclone (TC) density from ERA5 14 data spanning 1987–2020. (b) TC density from NeuralGCM-1.4° for 2020, generated using 34 different initial conditions all initialized in 2019. (c) Box plot depicting the annual number of TCs across different regions, based on ERA5 data (1987–2020), NeuralGCM-1.4° for 2020 (34 initial conditions), and orange markers show ERA5 for 2020. In the box plots, the red line represents the median; the box delineates the first to third quartiles; the whiskers extend to 1.5 times the interquartile range (Q1 − 1.5IQR and Q3 + 1.5IQR), and outliers are shown as individual dots. Each year is defined from January 19th to January 17th of the following year, aligning with data availability from X-SHiELD. For NeuralGCM simulations, the 3 initial conditions starting in January 2019 exclude data for January 17th, 2021, as these runs spanned only two years.
Number of Tropical Cyclones (TCs) as a function of maximum wind speed at 850hPa across different regions, based on ERA5 data (1987–2020; in orange), and NeuralGCM-1.4° for 2020 (34 initial conditions; in blue). Each year is defined from January 19th to January 17th of the following year, aligning with data availability from X-SHiELD. For NeuralGCM simulations, the 3 initial conditions starting in January 2019 exclude data for January 17th, 2021, as these runs spanned only two years.
Supplementary information.
Supplementary Information (38 figures, 6 tables): (A) Lines of code in atmospheric models; (B) Dynamical core of NeuralGCM; (C) Learned physics of NeuralGCM; (D) Encoder and decoder of NeuralGCM; (E) Time integration; (F) Evaluation metrics; (G) Training; (H) Additional weather evaluations; (I) Additional climate evaluations.
Rights and permissions.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .
Reprints and permissions
Cite this article.
Kochkov, D., Yuval, J., Langmore, I. et al. Neural general circulation models for weather and climate. Nature 632 , 1060–1066 (2024). https://doi.org/10.1038/s41586-024-07744-y
Download citation
Received : 13 November 2023
Accepted : 15 June 2024
Published : 22 July 2024
Issue Date : 29 August 2024
DOI : https://doi.org/10.1038/s41586-024-07744-y
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative
Google ai predicts long-term climate trends and weather — in minutes.
Nature (2024)
By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.
COMMENTS
Compendium on methods and tools to evaluate impacts of, vulnerability and adaptation to climate change. Final draft. 1. Introduction. 1.1 Focus and scope of the compendium. This compendium is organized in a way that allows existing adaptation analysis and decision frameworks and tools to be catalogued in manner that is clear and easy to use and ...
The proposed methodology can be modified for application to other sectors of CI; moreover, is a valuable resource in academic curricula and ongoing research initiatives.
With the emerging necessity to adapt to climate change, countries and communities are starting to design and implement adaptation initiatives of various types, scales, and coverage. These initiatives seek to manage anticipated climate change risks at the national, sub-national, local/community levels.
The analysis is based on a review of the status of national programs for M&E of CC adaptation and the implementation in those countries of the project "Adaptation to Climate Change Impacts in ...
Compendium on methods and tools to evaluate impacts of, vulnerability and adaptation to climate change. Final draft. Stakeholder approaches in general emphasize the importance of ensuring that the decisions to be analyzed, how they are analyzed, and the actions taken as a result of this analysis are driven by those who are affected by climate ...
Adaptation to climate change must be ramped up urgently. This Comment proposes three avenues to transform ambition to action: improve tracking of actions and progress, upscale investment ...
The climate change vulnerability assessment and adaptation planning methodology is now part of the framework of the Indo-Burma Ramsar Regional Initiative (IBRRI), established under the Mekong WET project, and funded by the German International Climate Initiative.
Climate change is defined as the shift in climate patterns mainly caused by greenhouse gas emissions from natural systems and human activities. So far, anthropogenic activities have caused about 1.0 °C of global warming above the pre-industrial level and this is likely to reach 1.5 °C between 2030 and 2052 if the current emission rates persist. In 2018, the world encountered 315 cases of ...
The report "The Adaptation Principles: A Guide for Designing Strategies for Climate Change Adaptation and Resilience" lays out 6 principles - as well as 26 actions, 12 tool boxes and 111 indicators - to help ministries of finance approach adaptation challenges.
Climate change and human activities are having increasing impacts on the global water cycle, particularly on streamflow. Current methods for quantifying these impacts are numerous and have their merits and limitations. There is a lack of a guide to help researchers select one or more appropriate methods for attribution analysis.
The analysis of building energy consumption showed that climate change can have a significant impact on shifting electricity demands [ 8 ], for example an increase in temperature due to climate change leads to an increase in cooling demand.
Approaches in which ML methods are combined and integrated into classical climate models, so called hybrid models (Fig. 2), have been proposed to be able to address many of the long-standing ...
Achieving the Sustainable Development Goals Accelerated and equitable climate action in mitigating and adapting to climate change impacts is critical to sustainable development. Some response options can absorb and store carbon and, at the same time, help communities limit the impacts associated with climate change.
The goal of this systematic literature review is to summarize the peer-reviewed literature on proposed solutions to climate change in the last 20 years (2002-2022), and to propose a framework for a unified approach to solving this climate change crisis.
Besides, the global tourism industry is devastated as climate change impacts unfavorable tourism spots. The methodology investigates hypothetical scenarios of climate variability and attempts to describe the quality of evidence to facilitate readers' careful, critical engagement.
This contribution extends a methodology previously presented to project temperature-related excess mortality in climate change scenarios. 5, 9 The proposed framework is illustrated through a hands-on tutorial, structured in well-differentiated steps that cover each of the methodologic issues and the essential elements.
This technical note describes the State of Climate Action 2023's methodology for identifying sectors that must transform, translating these transformations into global mitigation targets primarily for 2030 and 2050 and selecting indicators with datasets to monitor annual change.
Anticipating the Effects of applications Climate Change need to be developed. d global impacts. SAR, improvements have included greater emphasis on the of process-oriented models and ransient climate change Methods o
Regarding climate events, the main factors responsible for SW quality changes are intense rainfalls, prolonged periods of drought, extreme high temperatures and floods. These conditions are becoming increasingly frequent in many countries and are considered to be associated with climate change.
On June 21, 2022, EPA proposed amendments to specific provisions of the GHGRP to improve the quality of the data collected under the program by addressing changes in industry practices; adopting improved calculation and monitoring methods; and collecting new data to understand new source categories or new emissions sources for specific sectors.
The methodology described in this report enables councils to establish how much the current flood risk could alter under climate change, based on the expected changes in rainfall intensity during storm events.
Climate change can make it more difficult for communities to provide drinking water and wastewater services, protect water quality, and maintain healthy aquatic environments. The Adaptation Strategies below offer possible ways to address anticipated climate risks to water management. Water Utility. Water Quality. Ecosystem Protection.
The strategy calls such an approach "climate mainstreaming." The approach states that: "as climate impacts become more severe and frequent, and the costs mount, incorporating adaptation ...
The study, led by Climate Econometricians at the University of Oxford, the Potsdam Institute for Climate Impact Research (PIK), and the Mercator Research Institute on Global Commons and Climate Change (MCC), analysed 1,500 observed policies documented in a novel, high quality, OECD climate policy database for effectiveness.
In Climate Insights 2024: American Understanding of Climate Change, we showed that huge majorities of Americans believe that the earth has been warming, that the warming has been caused by human activity, that warming poses a significant threat to the nation and the world—especially to future generations—and that governments, businesses, and individuals should be taking steps to address it.
A recent Intergovernmental Panel on Climate Change (IPCC) Report [ 16] highlights how climate change increases the rate and extent of ongoing land degradation through two main factors: increased frequency, intensity of heavy rainfall and extreme high-temperature events.
B. Proposed Methodology for the Completion of Services ... Climate Change Units and the Design of MRV System in the Line Ministries 1 Background: Egypt is highly vulnerable to climate change that poses several numerous threats to its economic, social and environmental sustainability. Furthermore, the rapid population growth and ambitious
Abstract. Tipping points characterize the situation when a system experiences abrupt, rapid, and sometimes irreversible changes in response to only a gradual change in environmental conditions. Given that such events are in most cases undesirable, numerous approaches have been proposed to identify if a system is approaching a tipping point. Such approaches have been termed early warning ...
Methane is the second largest contributor to global surface air temperature rise. Reducing atmospheric methane will mitigate climate change and improve air quality. Since the main sink of methane is the hydroxyl radical (OH) in the atmosphere, increasing OH concentration will accelerate the methane oxidation process and reduce methane concentration. Because the primary source of OH is the ...
General circulation models (GCMs) are the foundation of weather and climate prediction1,2. GCMs are physics-based simulators that combine a numerical solver for large-scale dynamics with tuned ...