Does Soil Carbon Saturate? What the Ceiling Is, Where It Sits, and What Happens If You Keep Adding

Somewhere in the second half of most soil carbon meetings, someone says it: yes, but soils saturate. The room nods. The discussion moves on, and everyone leaves believing they agreed about something. In my experience they rarely did, because that one sentence is doing the work of at least four different claims, and only one of them is saturation in the sense the literature means.

The confusion matters commercially. If a soil has stopped gaining carbon because it has run out of surplus input, adding more input will start it gaining again. If it has stopped because the mineral surfaces that hold carbon are full, adding more input will not, and the extra carbon will go somewhere far less durable. Those two situations look identical on a monitoring chart and imply completely different things about the next twenty years of a project.

A companion piece, our benchmark for soil carbon removal rates, took the time axis: how fast a hectare can gain carbon, and why the first year is not the thirtieth. This one takes the capacity axis. How full is this soil, full of what, and according to whom.

Four different claims wearing one word

The first claim is an observation: this field has stopped gaining carbon. That is a measurement, and given how small soil carbon changes are relative to the stock, it is usually a measurement that cannot distinguish a genuine plateau from a slow gain hidden in sampling noise. The second is a flux statement: inputs and losses have come into balance, so the stock is at a steady state. The third is the technical one: the soil has exhausted its physical capacity to stabilise carbon, so no achievable input would raise the stock further. The fourth is economic: further gains are possible but no longer worth what they cost.

Only the third is saturation. The other three are routinely called saturation in project documents, and each of them has a different remedy. A measurement plateau is answered with better sampling. A steady state is answered with more input. A capacity limit is answered by accepting that this field is the wrong place to buy removals. An economic limit is answered by a different price. Conflating them produces the most common failure I see in soil carbon planning, which is a project that walks away from a field with plenty of headroom because someone said the word.

A plateau is not a ceiling

Take the simplest possible soil carbon model: carbon enters at some rate, and leaves in proportion to how much is present. Raise the input and the stock climbs, but the losses climb with it, until the two balance and the stock stops changing. That soil has plateaued. It has no capacity limit whatsoever. The flattening curve is not evidence of a ceiling, it is the expected behaviour of a system with no ceiling at all, and this is the single most important thing to understand about the word.

Which means a flat line on its own tells you nothing about which of the two you are looking at. The diagnostic is not another year of monitoring. It is a change in input. Raise the carbon going in, and a soil at equilibrium will move to a new and higher equilibrium, while a soil at capacity will not. Long-term experiments with several input levels are valuable for precisely this reason, and it is why the original test of the saturation hypothesis was built on sites where the input rate itself was the treatment​4.

The number worth reading in that tool is how much of the extra input the ceiling swallows. Without a capacity term, doubling the input doubles the eventual stock. With one, most of the additional carbon you paid to put in never appears, because retention efficiency in this formulation falls in proportion to how full the soil already is. That is the standard way the effect is written, and it has an underappreciated consequence: the ceiling starts costing you long before you get anywhere near it.

Note also how long the approach takes. Even with no capacity limit at all, a soil with a forty-five-year turnover time needs the better part of a century to settle after a change in management. Almost every field measurement ever made was taken from a soil still in transit, which is worth remembering the next time someone describes a landscape as being at equilibrium. A global synthesis of 1,741 paired experiments put the time to a new steady state at 28 to 73 years under fertiliser plus straw and 26 to 117 years under fertiliser plus manure, against the twenty-year default many accounting frameworks assume​38.

This is not a modelling curiosity. The most famous soil carbon plateau in the world is the farmyard manure plot on Broadbalk at Rothamsted, which has received 35 tonnes of manure a hectare every year since 1843. Its carbon stock climbed from 23 to roughly 75 tonnes per hectare and levelled off: after about eighty to a hundred years the rate of change was no longer statistically distinguishable from zero, against a fitted asymptote near 79 tonnes​39. That plot is routinely offered as proof that soils saturate.

It proves nothing of the sort, for a reason that is almost too simple to notice. Broadbalk has only ever run one manure rate. An experiment with a single input level can show you that a soil stopped changing over time. It cannot show you what would have happened had the input been larger, and that second thing is the entire definition of saturation. The recent consensus statements are explicit on this point: a capacity claim requires carbon to stop responding to an increase in inputs, not merely to stop changing with time​40, and any number of steady states exist for the same soil under different management​13.

Where an experiment does vary the input, the plateau tends to move. At Woburn, a three-year grass ley rotation settled near 1.3 per cent organic carbon and stayed there from the mid-1960s onward, looking for all the world like a ceiling. On the same soil at the same site, the eight-year ley reached 1.40 to 1.42 per cent and was still climbing after thirty years​41. The first plateau was an equilibrium at that input rate. Raising the input raised the plateau, which is exactly what a ceiling would not have permitted.

Only one fraction is supposed to fill up

Saturation is not a claim about soil organic matter in general. It is a claim about one part of it. The working division in current soil science splits organic matter into two forms that behave so differently that averaging them destroys the information: particulate organic matter, which is fragmented plant and animal residue sitting free in the pore space, and mineral-associated organic matter, which is largely microbial residue bound to clay and silt surfaces​6. The two differ in origin, in nutrient content, in vulnerability, and, critically here, in whether they have a ceiling.

Mineral-associated carbon is the persistent fraction. Globally it holds around 975 gigatonnes of carbon in the top metre against roughly 330 in the particulate pool, and its mean turnover time is about 129 years against 23​9. It is also the fraction with a plausible physical limit, because mineral surface area is finite and a soil only has so much clay and silt. Particulate carbon has no such constraint. It is limited by input and by decomposition, not by surface area, which is why the framework that separated these pools concluded that mineral-associated carbon saturates while particulate carbon can in principle accumulate indefinitely​7.

It is worth saying why mineral surfaces earn this much attention. For most of the twentieth century, organic matter was thought to persist because certain molecules were intrinsically hard to break down. That view has been abandoned. Persistence is now understood as a property of the surroundings rather than of the molecule: what protects carbon is where it sits, what it is bound to, and whether decomposers can physically reach it​3334. Once persistence is a matter of binding rather than chemistry, a finite quantity of binding surface becomes the obvious candidate for a limit, and the saturation question follows directly.

That asymmetry is the whole reason saturation is interesting rather than academic. If the fraction that fills up is also the fraction that lasts, then a soil approaching its mineral capacity is not simply gaining carbon more slowly. It is gaining a different and more fragile kind of carbon. The pool that keeps accepting input is the one that a warm year, a tillage pass, or a change of tenant can undo.

The published ceilings disagree with each other

The founding observation is nearly thirty years old. Across grassland and arable soils from several continents, the carbon held in the fine fraction scaled with the proportion of particles below twenty micrometres, which suggested that the fine fraction had a finite protective capacity set by texture​1. That relationship became an equation, the equation became a saturation deficit, and the deficit became the basis for estimating how much carbon a given soil could still take. A formal framework followed, dividing organic matter into pools with and without capacity limits and proposing saturation as a property of each​2. The approach is still in active use: a Swiss study of cropland and grassland monitoring sites fitted its own version of the relationship and put croplands at 62 per cent of the mineral-associated capacity of comparable permanent grassland​19, and a thirty-six-year subtropical no-till trial found the capacity to stabilise further carbon in the top few centimetres had fallen by 90 to 97 per cent while the layer from 10 to 20 centimetres still held most of the remaining headroom​35.

The trouble starts when you ask what the number is. Successive studies have refitted the same conceptual relationship on different datasets with different fractionation cutoffs and different statistical treatments, and they do not agree. Applied to one soil, the published capacity estimates span close to a factor of two, which is larger than most of the management effects anyone is trying to detect.

Some of that spread is real disagreement and some is definitional. A capacity fitted through the mean of a dataset answers a different question from one fitted through its upper envelope: the first describes how much carbon soils of this texture typically hold, the second how much the most carbon-rich of them manage. The size of that distinction is easy to underestimate. Applied to the same 342 observations, a boundary line through the upper ninetieth percentile gave a maximum loading of 78 grams of carbon per kilogram of fine fraction, while ordinary linear regression on the identical data gave 33​3. Recent work argues explicitly for the upper-envelope approach and finds that once you use it, large datasets from the United States and Europe both show a limit, at roughly 54 and 65 grams per kilogram of fine mineral material​18. The method is not a technicality, it is most of the argument.

One more trap sits underneath all of this, and it catches careful people. Capacity is sometimes quoted per kilogram of fine fraction and sometimes per kilogram of whole soil, and the two differ by however much silt and clay the soil happens to contain. Comparing a figure of one kind against a figure of the other is a routine error in this literature and has been flagged as such​40. Before you compare two capacity numbers, check they share a denominator.

The serious case that there is no ceiling

This is where an article written five years ago would have stopped. It cannot now, because the concept came under sustained attack in 2023 and the attack was good. Using the German Agricultural Soil Inventory, a study selected 189 samples spanning soil carbon contents from 5 to 118 grams per kilogram and clay contents from 3 to 77 per cent, and looked for the upper limit. It did not find one in any texture class. The proportion of carbon held in the mineral-associated fraction stayed remarkably stable across the entire range of bulk carbon contents, which is not what a saturating pool should do​11.

The proposed mechanism is that organic matter does not need bare mineral surface. It can bind to organic matter that is already bound, stacking outward in layers rather than competing for a finite monolayer. Direct isotope tracing supports this: newly added carbon preferentially attached to mineral surfaces already coated with native organic matter rather than to the clean surfaces still available, and did so in soils that were nowhere near their nominal capacity​15. If carbon accumulates in stacks, then surface area stops being the binding constraint. Supporting observations have accumulated from several directions: across an arable clay gradient from 5 to 37 per cent, bulk carbon showed no systematic increase with clay content, and the low-clay soils simply carried thicker organic coatings on the surface they had​16, while a Mexican survey spanning a wide range of carbon and clay found a strictly linear relationship with no upper limit in non-volcanic soils​17.

Then there is the brute-force test, and it is the closest thing we have to a direct answer to the question of what happens if you simply keep adding. Four long-term field experiments were pushed to organic inputs of up to 20 tonnes of carbon per hectare per year, roughly five times anything a working farm would sustain. Soil carbon stocks rose linearly with input at three of the four sites, with no sign of a plateau. Carbon in the fine silt and clay fraction increased by 17 per cent on average, and carbon in the sand-sized fraction by 146 per cent​14.

Read those last two numbers together, because they are the most useful pair of figures in this entire debate. Under absurd inputs the soil did keep gaining carbon, and the great majority of the gain went into the coarse, unprotected fraction. Whether you call that saturation depends on which question you are asking. The bulk stock did not saturate. The fraction that persists very nearly did.

How both sides can be right

The exchange that followed was unusually direct: a formal letter defending the concept, arguing that the absence of a hard monolayer limit does not abolish capacity, and that the multilayer stacking mechanism has its own ceiling because outer layers are held more weakly​12. That last point has since acquired evidence. Nitrogen-rich organic matter retained through organic-to-organic association decomposes considerably faster than material bound directly to mineral surfaces, which means the stacked carbon is real but not durable, and the functional limit reappears one level down​22.

A second reconciliation is that the constraint may not be mineralogical at all. Apparent saturation can emerge purely from ecological limits on microbial populations, since mineral-associated carbon is mostly microbial residue and there is a ceiling on how much microbial biomass a soil supports regardless of how much substrate arrives​21. On that account the curve bends for biological reasons and the mineral surfaces are innocent. Recent experimental work pushes in a related direction, finding that new mineral-associated carbon forms faster both where the saturation deficit is large and where existing mineral-associated carbon is already abundant, a combination that no simple surface-filling model predicts​23.

It also helps that at least one experiment has caught both halves of the mechanism in a single dataset. On a manure trial running more than thirty years, soil carbon rose with every increase in application rate up to 120 tonnes per hectare per year and then stopped: the plots receiving 180 tonnes stored no more than those receiving 120. Fractionation showed why. The mineral-associated and microaggregate pools saturated while particulate carbon carried on rising linearly​42. That is the whole argument of this article, observed in one field.

The synthesis now emerging is that the useful concept is an effective capacity rather than a theoretical one: not the maximum carbon the mineralogy could conceivably hold, but the level beyond which real soils in a given climate and land use are not observed to go​20. A 2025 review of twenty-five years of the debate lands in much the same place, arguing that the disagreements are largely about measurement and definition rather than about whether carbon accrual slows as soils fill​13. Others have proposed separating a theoretical saturation, the most the mineralogy could conceivably hold, from an apparent one that real soils actually exhibit​36, or moving the unit of analysis from the mineral surface to the whole ecosystem, on the argument that what limits carbon in practice is the supply of inputs a system can generate rather than the surface available to receive them​31.

My own reading is that the behaviour is well established and the number is not. Efficiency of carbon retention declines as soils fill, which was the original claim and has survived twenty-five years of testing in both field comparisons and controlled incubations, where the same litter addition yielded more stabilised carbon and less respired carbon in soils further from their limit​45. Where exactly the efficiency reaches zero is unknown, probably varies by soil, and may not be a useful question.

Keep adding and the carbon lands in the fragile pool

Three things, in order of how often they are ignored. The first, as above, is that the carbon increasingly arrives in the unprotected fraction. This is not a hypothetical: across 7,219 soil samples from six continents, particulate carbon rose 21.8 per cent between 2000 and 2022 while mineral-associated carbon fell 5.3 per cent. Bulk soil carbon went up 11 per cent, and the ratio of unprotected to protected carbon rose by 29 per cent​25. The world is gaining soil carbon and simultaneously making it less stable, and a bulk-carbon monitoring programme would report only the good half of that sentence.

A practical warning light exists. Across twenty-five central European long-term experiments, soils began to approach mineral-associated saturation once the ratio of particulate to mineral-associated carbon exceeded about 0.35, and the authors proposed that threshold as a management target to avoid organic over-fertilisation​24. It is a crude index and it covaries with texture, but it is calculable from a single fractionation and it is far better than nothing.

The second thing is that stabilising carbon costs nutrients. Stable fine-fraction organic matter has a near-constant elemental composition, with carbon, nitrogen, phosphorus and sulphur in a mass ratio close to 10,000 to 833 to 200 to 143​26. Carbon cannot be humified without them. In incubations, supplying those nutrients alongside straw doubled net humification efficiency from 7 to 15 per cent, and in a five-year field trial the same crop residue input produced a gain of 5.5 tonnes of carbon per hectare with supplementary nutrients and a loss of 3.2 tonnes without, a difference of 8.7 tonnes from nutrient balance alone​27.

Those nitrogen figures are the reason soil carbon sequestration has been described as presenting a nitrogen dilemma: the nutrient locked into new organic matter has to come from somewhere, and if it comes from fertiliser it carries its own manufacturing emissions and its own nitrous oxide losses​28. Nitrogen addition also skews the split in the unhelpful direction. Across 803 paired observations, nitrogen enrichment raised particulate carbon by 16.4 per cent and mineral-associated carbon by only 3.7 per cent, lowering the ratio of protected to unprotected carbon by around 10 per cent​29.

That framing has been contested, and the objection is a good one. The nitrogen is only an additional bill if carbon sequestration is treated as an extra effort bolted onto farming. If instead it is the by-product of management that raises productivity, much of the nutrient is already in the system and cycling, and the calculation double-counts a cost the farm was paying anyway​37. The same authors concede the narrower point cleanly: if you assume the nutrients can only come from manufactured fertiliser and atmospheric deposition, their own sequestration estimate is an overestimate. Both halves of that are worth carrying. The stoichiometry is not negotiable, but where the nutrient comes from very much is.

The third thing is the one that has nothing to do with soil science. Carbon added as imported manure or compost was already carbon somewhere else, and moving it does not remove anything from the atmosphere. That deduction belongs in the accounting rather than in the ceiling, but it bites hardest on exactly the high-input strategies that people reach for when a soil stops responding.

Saturation should choose your fields, not discount them

Saturation is most useful as a selection criterion rather than a discount factor. The emptier a soil, the more of each added tonne it keeps, so the same practice deployed on a depleted field and a rich one buys very different quantities of removal. That argues for targeting degraded land, and it argues against the intuition that the best soils are the best opportunity. The best soils are usually the ones with the least room left.

It also complicates the second crediting period. A field that has delivered twenty years of accrual is, by construction, closer to whatever ceiling it has than when it started, and its next twenty years will be slower. Renewal pricing built on the first period's observed rate will be wrong in a predictable direction. The same logic applies to the measurement plan, because a shrinking annual increment eventually disappears into the detection limit, which is the point at which a project stops being able to prove what it is selling. We have written separately about the signal-to-noise problem and about how many samples a given change requires.

And it should change what gets built, not only what gets bought. If the mineral fraction is near full, the remaining levers are the ones that add carbon in other forms or in other places: deeper in the profile, where mineral-associated carbon sits at a far lower fraction of its capacity than in topsoil​8, or as pyrogenic carbon, whose persistence has nothing to do with mineral surfaces at all and which we cover in the biochar accounting piece. Managing both pools deliberately, rather than optimising the mineral one alone, is the direction the field is moving​30.

What to ask when someone says a soil will saturate

Six questions, in the order that resolves the argument fastest.

1. Which of the four claims is this? Stopped gaining, at equilibrium, at capacity, or no longer economic. They have different remedies.
2. Saturation of which fraction? Bulk soil carbon and mineral-associated carbon behave differently, and the answer is meaningless without saying which.
3. Where does the capacity number come from? Name the equation, its calibration dataset, and its fractionation cutoff. If it cannot be named, it is an intuition.
4. Was it fitted to the mean or to the upper envelope? The two answer different questions and can disagree about whether a limit exists at all.
5. Has anyone actually measured the fractions here? A saturation deficit calculated from bulk carbon and a texture class is a prediction, not an observation.
6. If the input were doubled, what would happen? If nobody can answer, the plateau has not been diagnosed, only observed.

None of this means saturation is a myth. Accrual really does slow as soils fill, the efficiency of retention really does depend on the deficit, and accounting that ignores the effect overstates the mitigation contribution of soil carbon substantially​32. What it means is that saturation has been doing rhetorical work far beyond what the evidence supports, in both directions: used to dismiss fields that have ample headroom, and quietly ignored in projections that assume a rate holds for thirty years. The concept is sound. The confident numbers attached to it are not.

Poin-poin utama