Extreme event attribution

Purpose

A key element of attribution science is framing: the careful posing of questions that science is better able to answer.^{[p. xviii]} For example, science can better answer "whether human influence increased the frequency or intensity of a particular event" than "whether humans caused this event".^{[p. xviii]} As of 2016, most attribution studies considered classes of events similar to a particular observed event, but other studies perform the more challenging task of estimating how much climate change affected a specific event's intensity or frequency.^{[p. 1]} Results of such studies allow scientists and journalists to make statements such as, "this weather event was made at least n times more likely by human-caused climate change" or "this heatwave was made m degrees hotter than it would have been in a world without global warming" or "this event was effectively impossible without climate change".

Climate Central explained that "attribution studies calculate whether, and by how much, climate change affected the intensity, frequency or impact of extremes" such as wildfires, droughts, rainfall, and typhoons. German climatologist Friederike Otto said that attribution science aims to answer the question, "did climate change play a role" in specific extreme events "within the news time frame – so within two weeks of the event".

A further function of attribution science is to assess the impact of extreme events on human vulnerabilities and society's ability to adapt. Attribution science gives people a more concrete description of what climate change means to their environment, provides scientific evidence to the public discourse, and enables targeted adaptation based on expected increases in specific types of extreme events. People, communities, and businesses can use a local-scale perspective to anticipate future extremes at their specific location.

Methods

In what is regarded as a "breakthrough" paper (published in 2004), researchers compared simulations accounting for climate change with scenarios in which human-caused global warming did not exist. The breakthrough was not to ask whether climate change "caused" the event, but how much it might have affected the risk of it occurring at all. It's not a yes/no question.

Preliminarily, what constitutes an "extreme" event can be expressed in terms of percentile among observed historical events, of standard deviations about norms of previous events, of a likelihood of occurrence in a given time period (example: "1% chance per year"), or of expected frequencies (called a return period or recurrence interval, such as "100-year drought").

Event attribution approaches can be generally divided into two classes:^{[p. 3]} a first approach uses observational records to determine the change in probability or magnitude of events over time. A second approach uses model simulations to compare events in two worlds—a first "factual" world with human-caused greenhouse gas emissions and a second "counterfactual" world without such emissions—and attributing differences to human influence.

German climatologist Friederike Otto explained that attribution scientists use climate models to simulate what is possible under current climate conditions, and compute the likelihood of this particular event occurring given the forces affecting today's climate. Then, the preindustrial world is simulated by removing anthropogenic warming forces from the climate models or by performing statistical modeling on observations of the late 19th and early 20th century. Finally, differences in intensity or in frequency between the two simulations are deemed to be caused by climate change. Scientists use methods that have already been peer reviewed, allowing studies of new events to be published within a "news cycle" time frame.

Along similar lines, atmospheric scientist Xubin Zeng said that attribution studies generally proceed in four steps: (1) measuring the magnitude and frequency of a given event based on observed data, (2) running computer models to compare with and verify observation data, (3) running the same models on a baseline having no climate change, and (4) using statistics to compare and analyze the differences between the second and third steps.

Results from multiple attribution studies may differ from each other, based on differing definitions of what constitutes an "extreme" event, on the period of the historical record being analyzed, on the spatial resolution of the climate models, and so forth. Model outputs differ based on the event feature being analyzed (peak wind vs. average wind; rainfall amount vs. duration; etc.).

Findings

Climate Central's November 2024 review of more than 600 climate attribution studies covered almost 750 extreme weather events and trends. The review found that climate change made 74% of events more likely or severe, and that about 9% were made less likely or severe.

Climate change was found to affect the intensity and frequency of extreme weather events differently: for example, the 2010 Russia heat wave was made far more likely but not more intense. Thus, some attribution studies simply "ask different questions" than other attribution studies.

Improved understanding of relevant drivers and their model representation—such as atmospheric dynamics, atmospheric and soil moisture, and surface cover—improves heatwave prediction and projection. Heatwaves are more accurately modeled than other extreme events because long-term temperature data is more reliable than other events' datasets. Rainfall is also relatively straightforward, but droughts, snowstorms, tropical storms and wildfires are more complicated.

Applications and influence

Attribution science can inform choices about risk management and adaptation strategies as studied by insurers, government officials and policy makers, land and resource managers, zoning and infrastructure planners and engineers, litigators, and emergency managers.^{[p. 22]} For example, this science can guide decisions on whether to rebuild or to relocate after a certain extreme event^{[p. 1]}

Attribution science has been cited in climate change litigation against companies for causing, and governments for not addressing, climate change. Extreme event attribution (EEA) was applied to 213 historical heatwaves from 2000–2023, in order to attribute events to specific actors. Specifically, "carbon majors" (fossil fuel and cement producers) were found to have enabled 16–53 heatwaves that would have been virtually impossible in a preindustrial climate.

In late 2025, Copernicus Climate Change Service announced formation of a constantly operating attribution office aiming to provide two assessments a month, each within a week of an extreme weather event. The assessments can be used to help set government policy, assist insurance company risk assessments, and inform climate litigation.

Examples

Various organizations recognize attribution studies. The Sabin Center for Climate Change Law at Columbia University maintains searchable databases with categories: climate change attribution, extreme event attribution, impact attribution, source attribution, and attribution in the courts. Carbon Brief contains an interactive map including more than 600 studies covering almost 750 extreme weather events and trends. The American Meteorological Society publishes annual special collections, Explaining Extreme Events of [year] from a Climate Perspective.

Illustrative examples include:

• 2003 European heatwave: "it is very likely (confidence level >90%) that human influence has at least doubled the risk of a heatwave exceeding this threshold magnitude"
• 2004-2024 — intensified the ten deadliest extreme weather events during that time period, contributing to the deaths of at least 570,000 people
• October 2012 — Hurricane Sandy: flooded an area 27 square miles (70 km²) larger
• August 2016 — Louisiana floods: 40% more likely
• February 2017 — US winter heat wave: three times more likely
• August 2017 — Hurricane Harvey: three times more likely and 15% more intense
• 2017-2018 Tasman Sea marine heatwave: overall intensity was "virtually impossible" without anthropogenic forcing
• 1950–2020: 25–38% reduction in global abundance of tropical birds attributed to human-caused intensification of heat extremes
• 2019–20 Australian bushfire season: at least 30% more likely, and heat extremes more likely by at least a factor of 2
• 2020 Siberia heatwave: >500 times as likely as similar events in 1900; or 600 times ("minimum") to 99,000 times ("best estimate")
• 2021 Western North America heat wave: 150 times more likely
• 2022 United Kingdom heatwave: return period of 100 years for 2-day average temperatures, and 1 in 1000 years for single-day maximum temperature over one region of the UK; made the overall event at least 10 times more likely
• 2022: Worldwide review of likelihood and intensity of heat extremes: tens of thousands of deaths directly attributable
• 2022: Review of rainfall events in North Atlantic basin: events, combined, attributed $500 billion in damages
• May 2023 through May 2024: days with extreme heat (with vs without global warming): Suriname (182 vs 24 days), Ecuador (180 vs 10 days), Guyana (174 vs 33 days), El Salvador (163 vs 15 days), Panama (149 vs 12 days)
• 2023 and 2024 (various heat events): probability ratios (PRs) in Marshall Islands and Micronesia (PR=35). Indonesia and Philippines (each 24 and 29). Central America (13). Spain and Portugal (8). Southern South America (7).
• 2000–2023: of 213 historical heatwaves considered, fossil fuel and cement producers enabled 16–53 heatwaves that would have been virtually impossible in a preindustrial climate.
• 2024: 44% of the loss in Florida of a Category 4 Hurricane Helene, and 45% of the loss in Florida of a Category 5 Hurricane Milton, are attributable to climate change. ClimaMeter analyses show that storms like Hurricane Helene and Hurricane Milton are up to 20% wetter in the present climate.
• 2026: The 2026 Chilean wildfires were made three times more likely (2.5x in Patagonia) by climate change.