Human Fingerprints on Climate Are Clearer Now

IN 1967, two researchers ran a simple climate model and turned the carbon dioxide dial up, from 150 parts per million to 300 and then 600. The lowest layer of the atmosphere warmed while the layer above it cooled, a two-way signature that scientists have been hunting for in real measurements ever since. A new review by five climate scientists argues that the evidence for human influence on Earth’s climate has grown markedly stronger since 2009, the year the U.S. Environmental Protection Agency determined that greenhouse gases endanger public health. The authors say the case was already compelling then, and that a July 2025 report from the U.S. Department of Energy, which questioned it, fails to reflect the weight of evidence.

The review’s subject is fingerprinting: rather than asking only how much the planet has warmed, researchers ask where, when and in what combination things changed, because human and natural influences leave different patterns. Benjamin Santer at the University of East Anglia has collaborated on such work for more than 30 years, and his early research fed into the 1995 conclusion of the Intergovernmental Panel on Climate Change that a human influence on global climate was discernible.

Why would one gas do both jobs? More carbon dioxide lets the lower atmosphere hold on to heat, while the stratosphere above it radiates more of its own heat out to space, and human-caused loss of ozone (the stratosphere’s main heat source) adds to the chill. Syukuro Manabe and Richard Wetherald ran that 1967 experiment, and Manabe later shared the 2021 Nobel Prize in Physics partly for models of this kind. Weather balloon and satellite data then showed the same warming-below, cooling-above pattern in the real atmosphere.

Beyond Temperature Alone

One concern about the 1995 finding was its reliance on temperature. So analysts went looking elsewhere, and before 2009 they had found human patterns in sea-level pressure, rainfall, water vapor, humidity and Arctic sea ice.

After 2009 the list kept growing: salinity, snow cover, cloud cover, drought, the timing of streamflow in western U.S. rivers, and even the seasonal cycle itself, with fingerprints showing up in seasonal temperatures from the lower atmosphere down to the upper few hundred meters of the ocean. Dozens of research groups monitor these variables independently, using instruments in space, in the air, at sea and on the ground. The authors lean on that agreement (they call it consilience, a convergence of separate lines of evidence) throughout.

“The climate science that’s been published since the 2009 Endangerment Finding is now stronger than in 2009, the evidence is incontrovertible,” says Santer. Susan Solomon at the Massachusetts Institute of Technology adds that many independently monitored parts of the climate have all changed in physically consistent ways, from the top of the stratosphere to the depths of the ocean.

Time itself helped. The 10 warmest years in multiple surface temperature records all fall after 2009, which the authors say makes the human signal stand out more clearly against the noise of natural variability. Those 16 extra years of records also undercut talk of a pause in warming: natural swings can produce a cooling spell lasting a few years to a decade or more while greenhouse gases keep adding heat, and as the record lengthened, the dip was swamped. Large ensembles of model runs, many with more than 20 members, help separate the human signal from that background. And the upper ocean has warmed in every major basin at once since 1960, which natural swings, the authors say, cannot manage, since they can warm or cool one basin for decades but not the whole global ocean together. Better satellite records now reach 25 to 50 kilometers up, right where Manabe and Wetherald’s model predicted a large cooling.

A Review With Its Own Limits

It helps to know what sort of paper this is. It is a narrative review: the authors survey published fingerprint studies instead of running new analyses, and their paper states that no simulation output or satellite data were used. The release files the work under data and statistical analysis, which it is not. The paper also describes no procedure for choosing which studies to include, and many of the studies it cites are the authors’ own. None of that makes the cited evidence wrong; it does mean the review is a summary from inside the field, not an outside audit of it.

Extreme events are where the authors concede the most ground. Heat waves and heavy rain in some regions can now be tied to human influence with numbers attached, and the 2021 Pacific Northwest heat event looks nearly impossible without global warming. Gabi Hegerl at the University of Edinburgh says methods now exist to work out how greenhouse gases changed the frequency and intensity of such events. The paper adds what the release leaves out: significant scientific challenges remain, especially for events whose links to changes in average conditions are poorly understood.

Science Meets a Legal Finding

The release speaks of a move to revoke the Endangerment Finding, but the paper says the agency finalized the rescission on 12 February 2026 and that it is unclear whether the action will survive legal challenge. Santer adds that the review reinforces the need for policies limiting greenhouse gas emissions, which is a policy position, not a finding. In its last pages the paper turns openly political, calling the congressional investigation of a National Academies report a repudiation of decades of research and conflict-of-interest accusations against that report baseless. Those are the authors’ judgments. The Department of Energy’s side appears only as they describe it, and they answer the report largely by citing other critics.

Santer’s own worry is closer to home: that rigorous attribution science is itself endangered, because it connects fossil fuel burning to harm. Whether courts weigh the fingerprints is another matter, and the review does not claim to settle it; it asks only that the evidence inform fact-based decisions.

FAQ