In Climate Models, Undoing Emissions Does Not Undo What They Did to the Land

Suppose the world one day takes more carbon dioxide out of the air than it puts in. Emissions fall to zero, then go past it. The air thins, the planet stops warming and slowly starts to cool. What do the forests and the soil do then?

The tidy answer is that they run backward. Land absorbed a large share of what industry put into the air, so it seems fair to expect it to give that carbon back in roughly the order it was taken. A peer-reviewed paper published Wednesday in the open-access journal Earth System Dynamics finds the answer is untidier than that (inside a set of computer simulations, at least).

The study was led by Abigail Swann, a climate scientist at the University of Washington. She and colleagues at modeling centers around the world compared ten Earth system models, climate models that track carbon. Every model was given the same invented emissions history, so any difference in the answer came from the model and not from the question. That is the purpose of the exercise. Uncertainty about what land does with carbon dominates the uncertainty in how much carbon the planet takes up, and that feeds straight into estimates of future warming.

The emissions history is deliberately plain. It comes from flat10, a shared protocol built so that different models can be compared directly. Carbon goes into the air at a steady 10 billion metric tons a year, close to today's rate, for a century. Then, in the branch that matters here, emissions ease down to zero and keep falling until the same amount is being taken back out of the air each year. That continues until every metric ton ever emitted has been removed. The removal has no geography: it happens everywhere at once and evenly, which is not how any real method would work. The authors say so themselves. The design, they write, "does not capture the co-occurring impacts of any particular carbon dioxide removal method."

The land does not simply run backward

While emissions are still positive, every model puts carbon on land, and most of it goes into plants rather than soil. That much is expected. Extra carbon dioxide in the air lets plants grow faster, and it takes time for that carbon to reach the ground and stay there.

The interesting part comes later. When the models are pushed into negative emissions and run until the books balance, the land does not end up where it started. In the tropics, every one of them has lost carbon, mostly from plants. In the middle and high latitudes, most of them have gained it, and mostly in soil. That is a statement about the simulations. Nobody has measured a planet under negative emissions.

The explanation the authors offer is easy enough to picture. As carbon dioxide comes back out of the air, the growth boost it gave plants fades. The heat does not fade with it. Warm ground keeps breaking down dead material and releasing carbon, and hot places stay hard on plants. Farther from the equator the warmth helps more than it hurts over this stretch of time. The extra growth ends up in soil, which holds on to it for longer.

One more thing about the tropics. The models leave out several ways that heat harms plants: damage to the machinery of growth, higher death rates, pests and disease. The authors think the losses they see in the tropics are, if anything, an underestimate of what prolonged heat would do to the carbon stored there.

None of that is unanimous, and the paper does not claim it is. It opens by conceding "many differences in simulated land carbon pools and fluxes across models" and reports only "some consistent behavior" on top of them. The models do not even agree on how much carbon the land held before industry started. The largest store in the group is about three times the smallest.

Plants keep time and soil does not

The second finding is about when, not where, and it is measured from the moment emissions cross zero rather than from any calendar year. Across the models, carbon in plants peaks about ten years after that moment, and they cluster closely around it. That is close enough agreement to suggest the models handle the growth and death of plants in much the same way. Soil is another matter. Peak soil carbon arrives on average 33 years after zero, but the models spread from nine years before it to 126 years after.

That spread is not a technicality. Soil sets the slowest clock in the land carbon cycle, and slow clocks are what decide how much more the world warms after emissions stop. If the models cannot agree on when soil responds, they cannot pin that number down.

What this says about taking carbon back out

The authors' own summary is careful. Tropical carbon, they write, is "most likely to be both gained and subsequently lost," with "possible implications for carbon dioxide removal efforts." Those hedges are theirs and they matter. What the simulations describe is a general property of a modeled Earth under an idealized removal, not a verdict on any scheme that anyone has proposed or built.

And the sharpest practical point in the paper is not about the tropics at all. It is about soil. These models agree most about the part of the land carbon cycle that sits above ground and can be watched, and least about the part beneath it that runs the longest. Until that changes, the authors write, working out how much extra warming follows the end of emissions "remains a challenge," and they call the need to understand soil carbon urgent.