Attribution science is a field of climate research that determines whether, and to what extent, human-caused climate change made a specific extreme weather event more likely or more intense. It works by comparing the real world, where global temperatures have risen by about 1.3°C due to fossil fuel emissions, against a hypothetical world where that warming never happened. The gap between those two scenarios reveals climate change’s fingerprint on individual disasters.
How Attribution Studies Work
Every attribution study follows the same basic sequence, whether it’s analyzing a heatwave, a flood, or a wildfire season. First, researchers select an event based on its real-world impact: how many people were affected, how much damage occurred, and whether governments declared a state of emergency. Then they define the event precisely, setting geographic boundaries, a time window, and the best measurement to capture it, such as peak temperature or total rainfall over a specific number of days.
With the event defined, the analysis splits into two tracks. The first is observational. Scientists pull data from weather stations and satellites to determine how often events of similar magnitude have occurred historically and whether they’ve been getting more frequent or intense over recent decades. This gives a real-world trend line.
The second track uses climate models. Researchers run simulations of two worlds: today’s climate with 1.3°C of warming, and a cooler hypothetical climate where greenhouse gas emissions from fossil fuels never accumulated. By comparing how likely the event is in each simulated world, they can isolate the role of climate change from natural variability. The results from both tracks are then combined in a synthesis step, producing a final estimate of how much climate change shifted the event’s likelihood and intensity, along with a 95% confidence interval.
The Probability Comparison at Its Core
The central question in any attribution study is straightforward: how much did climate change increase the probability of this type of event? Scientists calculate the probability of the event in the actual, warming-affected world and compare it to the probability in a counterfactual world without human influence. The ratio between those two numbers tells you whether climate change doubled the odds, tripled them, or made the event essentially impossible without warming.
This concept borrows from epidemiology, where researchers have long used a metric called the population attributable fraction. In medicine, this measures what share of a disease burden would disappear if you removed a specific risk factor. In climate science, the “risk factor” is the extra heat energy trapped by greenhouse gases. The logic is the same: calculate the difference between what happened and what would have happened in the absence of the risk, and that difference is what you can attribute to the cause.
Fingerprinting: Separating Human and Natural Signals
Attribution science relies on a technique called fingerprinting to distinguish human-caused warming from natural climate fluctuations like volcanic eruptions or solar cycles. The method works by comparing model-simulated patterns of change against what’s actually been observed. Scientists use a regression-based approach where they map the expected “fingerprint” of anthropogenic warming (its spatial patterns across regions, its timing, its effect on temperature, precipitation, and humidity) and then check whether observations match that fingerprint or better fit natural variability.
This approach has two key strengths. It captures both the thermodynamic changes caused by warming, like more moisture in the atmosphere, and shifts in atmospheric circulation patterns. And it evaluates spatial patterns alongside temporal ones, making it harder to confuse a regional natural fluctuation with a global human-caused trend. Recent work has applied fingerprinting to fire weather extremes, detecting the anthropogenic signal in temperature, precipitation, and humidity changes that drive wildfire conditions.
Confidence Varies by Event Type
Not all extreme weather events are equally easy to attribute. Heatwaves sit at the top of the confidence scale because temperature is the most direct consequence of a warming atmosphere, and long observational records make statistical detection relatively straightforward. The 2021 Pacific Northwest heatwave is a landmark example. Temperatures in parts of the U.S. and Canada shattered records by margins that seemed physically implausible. An attribution study found the event was at least 150 times less likely in a world without human-caused climate change, leading researchers to describe it as “virtually impossible” without warming.
Heavy rainfall events also have reasonably high attribution confidence because a warmer atmosphere holds more moisture, a relationship that’s well understood physically and well documented observationally. Droughts are harder, since they involve complex interactions between temperature, soil moisture, and precipitation patterns. Tropical cyclone intensity and severe thunderstorms sit further down the confidence ladder. Their relationship to large-scale climate conditions is understood in broad terms, but the events themselves are rare enough and variable enough that long observational records needed for confident statistical detection often don’t exist.
From Science to Courtrooms
Attribution science has moved well beyond academic journals. Its methods are now central to a growing wave of climate litigation. Researchers at Dartmouth College developed a three-step framework that traces economic damages from specific companies’ emissions through to real-world harms. The chain works like this: first, calculate how much a company’s cumulative emissions contributed to global temperature rise. Second, identify where extreme heat caused economic losses in specific regions. Third, calculate a “but for” scenario, asking what the damages would have been if those emissions had never occurred. The difference is the company’s attributable share of the harm.
A 2025 study using this framework found that extreme heat linked to emissions from 111 major fossil fuel companies cost the global economy $28 trillion between 1991 and 2020. Of that, $9 trillion (about 32%) traced back to just five companies: BP, ExxonMobil, Chevron, Saudi Aramco, and Gazprom. One critical nuance: a company responsible for 10% of emissions isn’t necessarily responsible for only 10% of damages, because climate impacts don’t scale linearly with emissions in every region.
In the United States, 34 pending lawsuits brought by states and cities against fossil fuel companies could eventually use attribution evidence to argue their cases on the merits. California is currently suing five major oil companies for misleading the public about climate risks and seeking monetary damages. Legal hurdles around jurisdiction and court authority have slowed these cases, but as Michael Gerrard of Columbia University’s Sabin Center for Climate Change Law has noted, if any of these cases clear those procedural barriers, attribution research could prove highly valuable to plaintiffs.
Where Attribution Science Hits Its Limits
The field’s biggest constraint is data. Attribution studies depend on long, reliable observational records to establish how rare an event truly is, and those records are far thinner in the Southern Hemisphere, across much of Africa, and in parts of Asia and South America. When weather station coverage is sparse, confidence intervals widen, and definitive statements become harder to make.
Model resolution is another challenge. Global climate models operate on grids that can be tens of kilometers wide, which is too coarse to capture highly localized events like individual thunderstorms or tornadoes. Higher-resolution regional models help, but they’re computationally expensive and not available for every part of the world. There’s also the fundamental difficulty of studying events that are, by definition, rare. The statistical power to detect a change in something that happens once every few decades requires either very long records or very large signals, and for some event types, neither is available yet.
Despite these constraints, the field has advanced rapidly. The World Weather Attribution initiative now publishes peer-reviewed analyses within days or weeks of major disasters, giving the public near-real-time answers to the question that increasingly follows every extreme weather event: did climate change do this?