The Next Frontier of Attribution Science Has Arrived

Europe’s record-breaking heat wave in 2025 would have been a third of a degree Celsius cooler if not for emissions released just since the Paris Climate Agreement was signed in 2015, researchers found in a new study published Tuesday by the American Geophysical Union’s Geophysical Research Letters.

The research marks a step forward for attribution science, which has traditionally worked to tie extreme events such as heat waves and floods to climate change writ large. Now, using artificial intelligence trained on climate models, researchers have managed to link extreme weather to a specific subset of emissions.

“Not only does every little bit of emissions count, but the amount of emissions released since 2015 significantly increased the temperature of [the 2025 European] heat wave,” Jared Trok, the study’s lead author and a PhD student at the Stanford Doerr School of Sustainability, told me. “Before this paper” — which found 99-in-100 odds that human-caused emissions since 2015 increased the severity of the 2025 heat wave — “we couldn’t really make a claim to that extent.”

Though the record-breaking 2026 heat wave fell outside the scope of the study, the 2025 heat wave was no joke either — temperatures crested 115 degrees Fahrenheit in Spain and Portugal, and more than 16,000 died across the continent. Trok’s findings about a relationship between the past decade of emissions and intensified heat also held true for Europe’s hottest week in every year since at least 2021.

While a third of a degree Celsius might not sound like a lot — “it’s smaller than our ability to actually sense,” Trok acknowledged — there’s a growing body of scientific literature that suggests even incremental increases in temperature can be deadly. “It’s nonlinear,” Trok added. “For every additional increment of temperature, the impacts on heat-related mortality are even larger than the previous increment.” Though Trok and his colleagues did not look at mortality specifically, the reasoning indicates dozens if not hundreds of people could have died due to that fraction of a degree.

The study highlights the advances in the specificity and speed of attribution science, which a quarter of a century ago struggled to distinguish the influence of all historical emissions on any individual event. But it also suggests something grim: The past decade also overlaps with the biggest global efforts toward decarbonization. “Even if the decarbonization goals are achieved, these results as well as others suggest near certainty that the extremes, particularly extreme heat, will continue to intensify,” Noah Suresh Diffenbaugh, a Stanford climate scientist and the paper’s senior author, told me.

Paired with a separate commentary also published today by the U.S. Climate Collection, a joint project of AGU and the American Meteorological Society, the research adds an urgent underline to the need for research like Trok’s to be incorporated into state and local policymaking. Many of the institutions that existed to do so in the U.S., however, have collapsed or been actively dismantled by the second Trump administration.

The Climate Collection formed in the void that followed the forced breakup of the sixth National Climate Assessment (and is made up of many of its authors), and argues that the NCA did more than just good rigorous science — it also helped translate that research into a reliable springboard for policymakers.

The U.S. Climate Collection aims to compile an open-access collection of research papers that “lays the groundwork for future national and subnational assessments of climate risks and solutions in the United States.”

The group’s first paper serves as “a call to our colleagues to meet that need and the charge that has been given to us by society to produce the science” necessary for policymakers and other groups to “make better decisions,” Melissa Kenney, one of the commentary’s lead authors and director of research and knowledge initiatives at the University of Minnesota’s Institute on the Environment, told me.

In the past, the formal NCAs have helped inform everything from New Hampshire flood risk management plans to city- and state-level climate policies, the Climate Collection writes in their commentary. (They also set expectations: The last NCA required the involvement of 500 authors, 250 technical contributors, and synthesized more than 8,200 studies, meaning the Collective likely couldn’t replicate the rigor and scope even if it wanted to.) The Climate Collection specifically singles out attribution as an area of priority.

“Compounding extremes and cascading climate risks are increasingly overwhelming our legacy policies and infrastructure,” Kenney said, adding that “being able to understand the impact of these compounding extremes is really critical in a number of communities to be able to make smart, multi-decadal decisions like infrastructure choices.”

But as Trok’s research shows, even assumptions about the climate of 2015 are out of date. Investments in adaptation are a small fraction of the total dollars spent addressing climate change, and as Diffenbaugh stressed, the new paper is just the latest “of a number of studies that highlight that we can expect further acceleration of impacts from extreme events.”

The U.S. Climate Collection doesn’t intend to fill the gap left by the collapse of NCA6 (nor could it, its authors point out, given that it’s a self-organized volunteer group). But its call for synthesis papers of smaller scopes could give policymakers grounds to make decisions pulled from rigorous, peer-reviewed research as the world changes all around us. “These types of assessment reports are one of our greatest professional obligations as scientists,” Kenney said. “Most people will not go and read hundreds of scientific papers to be able to understand what we know and what we still need to know.”

“But,” she added, “there’s a real need for us to be able to provide the information” — before it becomes old news, too.