Our Best Climate Models Are Never Finished

Why Models Matter and Always Fall Short

Models quietly shape almost everything we do. A model, as Cathy O’Neil reminds us in Weapons of Math Destruction, is “nothing more than an abstract representation of some process.” By necessity, it is a simplification—a “toy version” of the world that includes certain features while omitting others. This makes models practical, but also incomplete. Every model has “blind spots” that reflect the priorities and values of the people who build it, which means models are never neutral descriptions of reality.

Physicist Carlo Rovelli offers a helpful distinction: scientific theories can be extraordinarily effective while still being limited. Newton’s laws, for example, work beautifully at everyday scales but fail near light speed or intense gravity. “Wrong,” in this sense, does not mean useless; it means valid only within a certain domain. Climate models, like all models, guide our understanding even as they inevitably distort parts of the picture.

A new study in PNAS seeks to correct a distortion in our understanding of the atmospheric impacts of climate change. 

The Too Long: Didn’t Read Summary

Climate change is causing nitrous oxide, a potent greenhouse gas and ozone-depleting substance, to break down in the atmosphere more quickly than previously thought, introducing significant uncertainty into climate projections for the rest of the 21st century.

Stratospheric Photochemistry: Where N₂O Meets Its End

Nitrous oxide (N₂O), the third most important long-lived greenhouse gas after carbon dioxide and methane, is removed efficiently once it reaches the upper stratosphere. It enters this region through a slow, global circulation pattern that lifts air upward in the tropics and carries it downward at higher latitudes. Once N₂O reaches the middle and upper stratosphere, sunlight-provided ultraviolet light and chemical reactions (photolysis) destroy 90% of stratospheric N₂O.

When N₂O breaks apart, it produces nitrogen oxides (NO_x), a family of compounds that catalyze ozone destruction. That chemistry is why N₂O is considered both an ozone-depleting substance and a greenhouse gas. Crucially, the balance between NOx that actively destroys ozone and NOx that is neutralized or transported away depends strongly on temperature.

This temperature sensitivity has direct implications for climate projections. Rising CO₂ cools the stratosphere, slowing key chemical reactions and reducing the overall abundance of NOx. With less NOx available, ozone destruction becomes less efficient. Paradoxically, a cooler stratosphere can therefore weaken N₂O’s ozone-depleting impact over time.

The Complications Beneath the Surface

One might think that measuring the “lifetime” of an atmospheric component would be straightforward, however there are definitional and methodological uncertainties. First, our current methods for measuring the “lifetime” of atmospheric components do not fully account for chemical changes that depend on circulation and temperature. Second, the researchers point to other definitional ambiguities. The lifetime, based upon where these gases are produced, varies – the 99% emitted from the Earth’s surface may live for more than 100 years, while those generated in the stratosphere last perhaps a decade. And because the peak losses occur in late winter, February and March, how one accounts for leap years in climate calculations can introduce significant variation. 

Overall, atmospheric N₂O concentrations continue to rise, but its rate of destruction in the stratosphere appears to be increasing even faster. The result is a modestly shorter atmospheric lifetime. Satellite data support this interpretation, yet the trend is difficult to pin down because natural variability is large and the statistics are delicate. As the authors note, signals that appear “likely” can vanish when stricter confidence thresholds are applied—an outcome that complicates long-term climate modeling.

Cooling Stratosphere, Rising Ozone, and Unexpected Feedback Loops

Leaving aside the controversial question of its source, one of the clearest large-scale atmospheric changes is the rise in carbon dioxide levels. A first-order estimate suggests several degrees of cooling by the end of this century, right in the altitude range where most N₂O loss occurs.

As the stratosphere cools, catalytic ozone loss slows, allowing ozone concentrations to recover. Because ozone regulates how much ultraviolet radiation penetrates the stratosphere, higher ozone levels feed back on N₂O chemistry, further reducing ozone destruction. At the same time, climate models and observations suggest that warming at lower altitudes accelerates upward air transport in the tropics, delivering more N₂O into the region where it is destroyed and shortening its lifetime even further.

While we have a good understanding of the underlying NO_x chemistry, long-term predictions depend strongly on how models represent circulation changes and coupled feedbacks. The researchers note that the uncertainty introduced by these stratospheric processes may significantly alter our current emissions scenarios. The Shared Socioeconomic Pathways, climate scenarios from the United Nations’ Intergovernmental Panel on Climate Change (IPCC), when accounting for this updated modeling, would reduce projected N₂O from a “high-emission” to “moderate-emission” scenario without any real-world change in emissions.

“Stratospheric chemistry and dynamics present uncertainties in projecting N2O that are as large as uncertainties across different emissions scenarios. We need to incorporate these effects into the models used for international climate assessments.” 

- Michael Prather, Ph.D., UC Irvine professor of Earth system science

Models Evolve

In the end, models are unavoidable. We cannot understand something as vast and intricate as Earth’s atmosphere and climate without building simplified representations. But the lesson from nitrous oxide is that even well-established climate processes can shift once overlooked chemistry, circulation, and feedback loops are accounted for. The goal is not to demand perfect predictions from imperfect tools, but to recognize that models improve precisely because they are revised. Models guide science and policy best when we remain alert to their blind spots, honest about uncertainty, and ready to update our conclusions as the real world pushes back against our equations.

Source: Projecting nitrous oxide over the 21st century, uncertainty related to stratospheric loss PNAS DOI: 10.1073/pnas.2524123123