Climate Models vs. Reality: Why Heatwaves Keep Outrunning the Forecast

How much hotter can the hottest days get before climate models catch up? Researchers who compared six decades of observed temperatures against dozens of climate model simulations found that in several regions of the world, the most extreme heat is intensifying far faster than the models expected, and by a wide margin.

The gap is largest in some of the most crowded places on Earth. Western Europe stands out as the clearest example: the hottest days of the year there are now warming roughly twice as fast as an average summer day, a trend behind the record-shattering heatwaves of 2003, 2010, 2018 and 2022, the last of which coincided with more than 60,000 heat related deaths across the continent in a single season. Similar hotspots turned up in central China, the Arabian Peninsula, eastern Australia, Japan and Korea, southern South America, and even high latitude regions of Canada and Greenland.

Climate models are missing the sharpest extremes by a factor of four. Using reanalysis data going back to 1958, the team tracked the difference between the very hottest days of each year and the merely warm ones, a measure they call tail widening. Averaged across the globe, models underestimate the areas where that widening exceeds half a degree Celsius per decade by roughly four times compared with what reanalysis data actually shows, and they capture only about a third of the statistically significant hot spots that show up in the observational record.

In some regions, reality has moved completely outside what any model predicted. For southern South America, the Arabian Peninsula and Arctic Canada, the observed warming trend fell outside the full range produced by 49 different model runs, not just at the edge of it. Higher resolution simulations and versions forced with real historical ocean temperatures barely narrowed that gap, which suggests the problem is not simply a lack of computing power.

The researchers are careful to frame this as models struggling to capture a genuine physical trend, not proof of a single tidy cause. They point to a tangle of interacting factors, drying soil that amplifies heat, stalled high pressure systems that trap it in place, and shifting patterns of aerosol pollution, as candidates that current models likely underrepresent. Untangling how much each one contributes is still an open question, and the paper stops short of claiming any one mechanism explains the whole pattern.

The stakes reach well beyond an uncomfortable summer. Many climate targets, including the 1.5 degree Celsius threshold in the Paris Agreement, were set using the average projection across many models. If the most damaging extremes are already arriving faster than that average suggests, then relying on it alone risks underestimating exactly the kind of event most likely to strain hospitals, power grids and water supplies.

The study, led by Kai Kornhuber alongside Samuel Bartusek, Richard Seager, Hans Joachim Schellnhuber and Mingfang Ting, was published in the Proceedings of the National Academy of Sciences (DOI: 10.1073/pnas.2411258121).