It's not a paradox, it's a pattern and a recent paper described the mechanics of how it works. The short answer is that carbon dioxide (CO2) reacts differently to wavelengths of light. Closer to earth, CO2 traps it but it makes the stratosphere better at radiating, which cools it—but because it becomes colder, the Earth system ends up losing less heat to space overall, strengthening warming below.
As predicted in Nobel Prize-winning work by American physicist Professor Syukuro Manabe in the 1960s, in the lower atmosphere, CO2 molecules trap heat that would otherwise escape into space but starting at about 20 miles up, CO2 molecules absorb infrared energy from below and radiate some of that stratosphere energy into space. Accurate temperature readings began to occur in the 1980s and show that the stratosphere has cooled by 2 degrees Celsius, which is an order of magnitude more than would be natural in the CO2 levels of 300 years ago.

Increases in CO2 cool the global-mean stratosphere.
The new paper seeks to turn it from modeling and estimates into a quantitative theory for CO2-induced stratospheric cooling. Their process used an iterative method of identifying the key processes, assigning mathematical values to them, comparing the results of their models to comprehensive simulations and real-world data, tweaking their equations and repeating.
Predicting the past doesn't great, economists have done it for decades without success predicting real-world behavior, but in science predicting is necessary to try and predict the future. The authors write that how CO2 molecules interact with longwave - infrared - light, can cause changes to cooling more than others. They outline a “Goldilocks zone” that is most efficient and say that as CO2 accumulates in the atmosphere, that zone expands.
They included ozone and water vapor but say those have little influence compared with CO2.
Citation: Cohen, S., Pincus, R.&Polvani, L.M. Stratospheric cooling and amplification of radiative forcing with rising carbon dioxide. Nat. Geosci. (2026). https://www.nature.com/articles/s41561-026-01965-8





Comments