Almost every gram of a tree came out of the air. What the soil supplied is roughly one per cent of it.
A tree is made of air. Roughly half the dry mass of wood is carbon, and effectively all of it came from carbon dioxide. What matters for the climate is not how much a tree captures but how long it holds it — and most of it comes back.
Two things about trees and carbon are worth getting straight, and almost every popular account gets one or both of them wrong. The first is where the mass comes from: a tree builds itself from carbon dioxide and water, and the soil contributes about one per cent of the dry weight in minerals. Van Helmont demonstrated the negative half of that in the 1640s by weighing a pot of soil before and after growing a willow in it, and it took another century and a half to identify air as the missing input. The second is what happens next. Carbon captured by a forest is not filed away; it is in transit. Leaves return theirs in a year or two, fine roots not much longer, wood in decades to centuries, and soil carbon can persist for millennia. A forest is a carbon sink only while uptake exceeds return, which is a fact about a particular year rather than a property of trees, and the Amazon plots that showed the sink falling by a third did so because trees were dying sooner rather than growing less. This page carries no single figure for how much carbon dioxide a tree absorbs, because no such figure exists.
Developed record · 87% complete · reviewed 2026-08-30
What this page covers
The carbon cycle as it runs through trees and forests: uptake, storage, return and residence time. Not a page about climate policy, and not a carbon calculator.
Quick facts
Carbon in dry wood
Roughly 50% by mass, from carbon dioxide
From the soil
About 1% — minerals, not bulk
Do old trees slow down?
No — most add mass faster as they get larger
CO₂ per tree per year
No general figure exists; it varies by more than tenfold
The single most counter-intuitive true thing about a tree.
The short answer
Where does a tree’s mass actually come from?
Overwhelmingly from carbon dioxide in the air. Roughly half the dry weight of wood is carbon that entered through the leaves as gas; hydrogen comes from water, and the soil supplies about one per cent, as minerals.
The intuition that a tree eats soil is very old and was tested very early. In the 1640s Van Helmont grew a willow for five years in a covered pot of pre-dried, weighed soil, adding nothing but water. The tree gained about 74 kilograms; the soil lost about 60 grams. He drew the wrong conclusion — that the mass came from the water — because air was not something he could weigh, and identifying carbon dioxide as the real source took another century and a half. But the negative result held completely: whatever a tree is built from, it is not the ground it stands in.
A tree is built mostly out of air — the carbon in wood came from carbon dioxide
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
The dry mass of a plant is predominantly carbon, hydrogen and oxygen. Carbon enters exclusively as atmospheric carbon dioxide fixed in photosynthesis; hydrogen and oxygen derive from water. Soil supplies mineral nutrients required in comparatively small quantities, not the bulk of the material.
Who this applies to
plants generally, trees most strikingly
Studied in
Plantae
You may have heard
“Trees get their mass from the soil”
Van Helmont tested this in the 1640s by growing a willow in weighed soil for five years. The tree gained 74 kg; the soil lost about 60 g. He concluded the mass came from water, which was half right — most of the dry mass is carbon, and it came out of the air.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
Established by the carbon-fixation work and by elemental analysis of biomass. Van Helmont’s seventeenth-century willow experiment showed the soil loses almost no mass, though he drew the wrong conclusion from it.
How far it can be extended
Elemental composition of plant biomass is general, and the carbon source is established by isotope tracing.
Caveats
Soil nutrients are essential; the claim concerns bulk mass, not necessity.
A large fraction of a living tree’s mass is water, which does come through the roots.
Mineral nutrient limitation constrains growth even though it contributes little mass.
Still unanswered
How will rising atmospheric carbon dioxide alter growth where nutrients limit it?
The water transport that must accompany carbon fixation, and the ceiling it imposes.
Diagram
Where the atoms in a plank came from
Percentages are typical of dry wood and vary with species and tissue. The one per cent from the soil is the entire mineral contribution.
The same explanation in words
A flow diagram with three inputs on the left. Air supplies carbon dioxide. Soil water supplies hydrogen and oxygen. Soil minerals supply nitrogen, potassium and other nutrients, drawn with a much thinner arrow. All three feed into photosynthesis, which builds sugars from carbon dioxide and water; those sugars become cellulose and lignin, which is wood. Below the flow, a horizontal bar shows the composition of dry wood by weight: carbon at about 50 per cent, oxygen at about 43 per cent, hydrogen at about 6 per cent, and a sliver at the end marking minerals from the soil at about 1 per cent. The carbon and most of the oxygen entered as carbon dioxide through the leaves; the hydrogen came from water taken up by the roots. The soil’s contribution to the bulk of a tree is the sliver.
How we know
The willow, the pot of soil, and the five-year weigh-in
Does a growing tree eat the soil it stands in?
A willow shoot weighing about 2.3 kilograms was planted in a pot holding about 90 kilograms of soil that had been dried and weighed. The pot was covered to keep windblown dust out. For five years the tree received nothing but water. At the end, the tree was pulled up, shaken free of soil and weighed, and the soil was dried and weighed again.
What happened
The willow had gained roughly 74 kilograms. The soil had lost about 60 grams — a rounding error against the mass of the tree. Whatever the tree was made of, it had not come out of the pot.
What it shows
A tree’s mass does not come from the soil. That was the question, the measurement was good enough to answer it, and the answer has never been overturned.
What it does not show
It does not show where the mass did come from. Van Helmont concluded it was the water, because water was the only input he had added, and that is half right: hydrogen in wood comes from water. The larger part — the carbon and much of the oxygen — comes from carbon dioxide in the air, which he could not weigh and had no reason to suspect. It also cannot resolve the small mineral contribution, which is real but too light for his balance.
The controls — what makes this evidence rather than a story
The soil was dried before and after, so the comparison is of dry mass rather than of how wet the pot happened to be.
The pot was covered, excluding dust and debris as an unmeasured input.
Only water was added, so any input other than water and air was deliberately excluded.
A willow gained about 74 kilograms in five years while its pot of soil lost around 60 grams.
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Van Helmont’s willow experiment, reported in 1648, recorded a mass gain of roughly 74 kg in a tree grown for five years in a covered pot of pre-dried, weighed soil, against a soil mass loss of about 60 g — establishing that the added mass did not come from the soil, while leaving its actual source unidentified.
Who this applies to
A single willow in a single pot; the conclusion generalises, the measurement does not.
Studied in
Salix
You may have heard
“Van Helmont proved plants are made of water.”
He showed the soil was not consumed, which was the question he asked and the part that held. His own explanation was wrong, and identifying air as the real source took another century and a half.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
A single historical measurement, but one whose negative conclusion — the soil was not consumed — has never been contradicted and is now explained by the chemistry of photosynthesis.
How far it can be extended
The result has been reproduced in principle by every subsequent measurement of where plant mass comes from, and the underlying mechanism is now known.
Caveats
Van Helmont concluded the mass came from water, which is only partly right: hydrogen does, but most of the mass is carbon and oxygen from carbon dioxide.
One plant, no replication, and a balance too coarse to resolve the small mineral uptake that did occur.
The most-asked question here, and the reason NatureHQ does not answer it with a number.
There is no such thing as the amount of carbon dioxide a tree absorbs in a year.
Well supported
Good evidence backs this, though some details remain open.
Annual carbon uptake by an individual tree varies by more than an order of magnitude with species, size, age, growing conditions and site, and the widely circulated per-tree figures are means over highly skewed distributions in which a small minority of large trees accounts for most of the uptake.
Who this applies to
Any statement of the form "a tree absorbs X kilograms of carbon dioxide a year", however sourced.
Studied in
Tracheophyta
You may have heard
“A mature tree absorbs about 22 kilograms of carbon dioxide a year.”
Figures of this kind descend from averages over urban tree surveys, where uptake per tree ranged over more than an order of magnitude. Quoted without species, size, age and site, the number is not a simplification of that work — it is a claim the work does not make.
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
Both premises are well established: uptake scales strongly with size, and measured populations of trees are dominated in mass by a small number of large individuals. Together they make a single representative figure impossible in principle rather than merely unavailable.
How far it can be extended
The skew is a property of how tree mass scales with size, which holds wherever trees grow.
Caveats
This is not an argument that trees take up little carbon, or that estimation is impossible. Estimates are made routinely and well — with the species, size, site and method attached.
Population-level figures for a city or a country are meaningful in a way that a per-tree figure is not.
The source of much of the per-tree figure literature, and explicit that its averages sit on a distribution in which a minority of large trees holds most of the carbon.
Shows uptake rising continuously with size, which is the reason a mean across a size distribution describes almost no individual tree.
Figures of the form "a mature tree absorbs about 22 kilograms of carbon dioxide a year" descend from urban forest surveys in which uptake per tree ranged over more than an order of magnitude, and in which a small minority of large trees accounted for most of the total. The average across such a distribution is a useful accounting device for a city and a poor description of any individual tree. A young birch in a lawn and a two-hundred-year-old oak on good ground are not doing the same thing to within a factor of two, and the difference between them is most of the answer.
Size dominates. Uptake scales with how much surface a tree is adding wood to, and a large tree can add as much mass in a year as a mid-sized tree contains.
Species matters, through wood density as well as growth rate: two trees of the same volume can differ substantially in carbon.
Site matters. Water, light, nutrients and competition set how much of a species’ potential is realised.
The year matters. A drought year and a good year differ, and the sink is a balance rather than a constant.
Method matters. Mass is estimated from diameter with allometric equations whose error is largest for the biggest trees.
None of this means the quantity cannot be estimated. It is estimated routinely and well — for a stand, a city or a country, with the species mix, the size distribution, the site and the method attached. What cannot be done is to strip those away and keep the number.
Words used here
Allometric equation
A formula estimating a tree’s mass from something easy to measure, usually trunk diameter. Fitted to felled trees, and least reliable for the largest.
Sequestration
Carbon taken out of the atmosphere and held. The word is often used as though holding were permanent; residence time is what decides that.
Big old trees do not stop absorbing carbon. Most of them speed up.
Well supported
Good evidence backs this, though some details remain open.
In repeat measurements of 673,046 individually tagged trees across 403 species, absolute above-ground mass increment increased continuously with tree size for the large majority of species, even though relative growth rate declined.
Who this applies to
Individual trees across tropical and temperate forests; a statement about trees, not about stands.
Studied in
Tracheophyta
You may have heard
“Old trees stop absorbing carbon, so it is better to cut them and plant new ones.”
The first half is wrong at the level of the tree and the second half ignores the stock. A large tree adds carbon faster than a small one, and felling it releases centuries of accumulated carbon that a replacement takes centuries to rebuild.
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
A very large, global, repeat-measurement dataset with a clear result analysed within species. The main uncertainty is in the allometric conversion from diameter to mass, which is weakest for the very largest trees.
How far it can be extended
The pattern was tested within each of 403 species separately across both tropical and temperate plots, so it is not an artefact of comparing species.
Caveats
Relative growth does slow: a large tree adds a smaller fraction of itself each year. The absolute quantity is what matters for carbon.
An individual result does not scale automatically to a forest, where mortality and decay run against growth.
A minority of species do slow in absolute terms.
Still unanswered
How far the acceleration continues in the very largest trees, where the mass equations are least well tested.
Finds old stands still taking up carbon rather than sitting in balance, which is the stand-level counterpart of the individual-level result.
How we know
Measuring the same 600,000 trees again
Do big old trees slow down, or do they add mass faster than young ones?
Long-running forest plots around the world carry individually numbered trees whose trunk diameter is re-measured every few years. Those repeat measurements were assembled for hundreds of thousands of trees across 403 species, converted to mass using species-appropriate equations, and the mass gained per tree per year was plotted against the size of the tree — separately for each species, so that fast-growing and slow-growing species could not create the pattern between them.
What happened
For the great majority of species, mass gained per year kept increasing with tree size. Some of the largest trees added as much mass in a single year as an entire mid-sized tree contains.
What it shows
Large old trees are not coasting. At the level of the individual tree, absolute carbon accumulation accelerates with size, which contradicts the common assumption that growth tails off with age.
What it does not show
It does not show that an old forest absorbs more carbon than a young one. A stand contains mortality and decay as well as growth, and an old stand has fewer, larger trees; the stand-level balance is a different measurement with a different answer. Mass is also estimated from diameter using equations that are least well tested on exactly the largest trees the result depends on.
The controls — what makes this evidence rather than a story
Each species was analysed on its own, so the trend is within species rather than an artefact of comparing different ones.
Trees were individually tagged and re-measured, so growth is observed rather than inferred from a snapshot.
Both tropical and temperate plots were included, spanning very different growth conditions.
The confusion behind the question is between relative and absolute growth. A sapling doubles in size; a giant does not come close. But a giant adding a millimetre to an enormous circumference is adding far more wood than a sapling doubling itself, and it is the absolute quantity that carbon accounting cares about. Both facts are true, and only one of them is about carbon.
The individual result does not settle the stand-level question, and it is worth being careful about that. An old forest has fewer, larger trees, more deadwood and more decomposition, so its net balance is a different measurement — one that finds old stands still taking up carbon on average, more slowly, while holding a far larger accumulated stock. It is the stock that decides what is lost when such a forest is cleared.
Old forests are still absorbing carbon, and they are holding centuries of it already.
Well supported
Good evidence backs this, though some details remain open.
Compiled flux measurements from boreal and temperate stands over about 150 years old found net carbon uptake continuing rather than reaching balance, alongside very large accumulated stocks in wood and soil that are released if the stand is disturbed.
Who this applies to
Old boreal and temperate forest stands; tropical old growth is under-represented in the data.
Studied in
Tracheophyta
You may have heard
“An old forest is carbon-neutral, so there is no carbon reason to protect it.”
Even where uptake has slowed, the stock is what matters. An old forest holds centuries of accumulated carbon in wood and soil, and disturbing it returns that carbon on a timescale of years rather than centuries.
Why we rate it this way, and what the caveats are
Well supportedModerate confidence
A substantial compilation with a clear average result, but flux measurements at a limited number of sites are extrapolated widely, and the variation between stands is large.
How far it can be extended
The measurements are concentrated in the northern hemisphere, and stands that are net sources exist within the same dataset.
Caveats
The stock is the larger number, and the more consequential one: it took centuries to build and can be released in a season.
Averages conceal stands that are net sources, particularly after disturbance.
Still unanswered
How tropical old-growth forests behave, where flux measurements remain sparse.
Capture is the easy half. Residence time is the half that decides anything.
Most of the carbon a forest captures goes back into the air. What matters is how long it stays.
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Carbon fixed by photosynthesis is returned to the atmosphere by plant respiration, by decomposition of litter and deadwood, and by disturbance, on timescales ranging from months for leaves to decades or centuries for wood and soil; net storage is the residual of these fluxes rather than a quantity captured.
Who this applies to
Forest carbon cycling generally.
Studied in
Tracheophyta
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
The pathways and their approximate magnitudes are well measured by several independent methods, and the deadwood flux in particular now has a global field experiment behind it.
How far it can be extended
The pathways are the same everywhere; only their relative rates differ with climate and vegetation.
Caveats
Soil carbon can persist for centuries to millennia and is in many forests the largest single store.
Returning to the air is not the same as being lost from the system: much of it is taken up again by the same forest.
Still unanswered
How much soil carbon in forests is genuinely long-lived, which remains the least well constrained term in forest carbon budgets.
Quantifies the global deadwood decay flux and shows that most of the carbon released from dead wood returns to the atmosphere rather than entering long-term storage.
Sets out the disturbance pathways — fire, drought, insect outbreak, clearance — that shorten residence time abruptly.
Diagram
How long carbon stays, and by which route it leaves
Every arrow is a flux. A forest is a sink only while the downward arrows outweigh the upward ones.
The same explanation in words
A flow diagram of forest carbon with the atmosphere as a band across the top. An arrow runs down from the atmosphere into living biomass, labelled as photosynthesis taking carbon in; a second arrow runs back up, labelled as respiration returning much of it. The living-biomass box lists residence times: leaves months to a year, fine roots months to years, and wood decades to centuries — with a note that wood holds most of the mass and most of the residence time. An arrow labelled death leads to a second box, litter and deadwood: leaf litter a year or two, fallen trunks decades, with fungi and insects taking it apart so that some carbon returns to the air and some enters the soil. Arrows from both boxes lead down into soil carbon, which persists for decades to millennia and is in many forests the largest single store, and the one most easily lost when the ground is disturbed. A dashed arrow runs from the deadwood box straight back to the atmosphere, labelled fire, harvest and clearance: centuries returned in hours. The closing note states that sequestration is the balance of every arrow rather than the length of any one of them, and that a forest is a sink only while uptake exceeds return — a fact about a particular year rather than about a species.
This is why "a forest absorbs carbon" and "a forest stores carbon" are different statements with different consequences. The absorbing is a rate and it can go to zero or turn negative in a bad year. The store is an accumulation of centuries, sitting in wood and — more of it than most people expect — in soil, and disturbance can return a great deal of it very quickly.
The world’s forests take up an enormous quantity of carbon, and clearance cancels much of it.
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Global forest carbon accounting for 1990–2007 found established forests taking up carbon at a large and sustained rate, with tropical deforestation releasing a comparable quantity, so the net forest sink was substantially smaller than the gross uptake.
Who this applies to
Global forest carbon budgets over recent decades.
Studied in
Tracheophyta
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
Forest carbon accounting is a mature field with independent methods — inventories, plots, atmospheric measurement — that broadly agree on the size of the sink, even where they disagree on its attribution.
How far it can be extended
Built from national inventories and plot networks covering the world’s forest area, with the largest uncertainty in the tropics.
Caveats
A sink is a balance between uptake and loss in a given period, not a property a forest possesses.
Tropical inventory coverage is the weakest part of the accounting and the largest single uncertainty.
Still unanswered
How much further the tropical sink will weaken, and whether temperate and boreal uptake will offset it.
Shows the sink is not a fixed property: in Amazonia it fell by about a third over three decades, driven by rising mortality rather than falling growth.
Sets out the disturbance risks to carbon already stored, which is what makes the sink conditional rather than permanent.
The Amazon plot network is the clearest illustration that the answer is no. Over three decades of re-measuring tagged trees, growth rose — and mortality rose faster. Trees were growing more and dying sooner, and because the second effect was larger, the net sink fell by about a third from its peak. Any account that measures uptake alone would have reported the change with the wrong sign.
The same logic applies to a forest after a fire, a plantation after a harvest, or a peat forest that has been drained: all of them can release far more than they take up. Nothing about being a forest guarantees the direction of the flow.
How much carbon tree planting could remove is genuinely disputed, and the famous figure was challenged hard.
Contested
Researchers actively disagree, and the disagreement is substantive.
A 2019 estimate that global tree restoration could store on the order of 200 gigatonnes of carbon was met with formal technical comments disputing the restorable area, the carbon accounting and the classification of naturally open grassy biomes as degraded, and the authors issued a correction to the carbon figure.
Who this applies to
Global estimates of carbon removal achievable by increasing tree cover.
Studied in
Tracheophyta
You may have heard
“Planting a trillion trees would cancel most human emissions.”
The paper behind that framing corrected its own carbon figure, and its restoration map counted ancient grasslands and savannas as land awaiting forest. Increasing tree cover helps; the arithmetic that made it sound sufficient did not survive review.
Why we rate it this way, and what the caveats are
ContestedModerate confidence
That increasing tree cover removes carbon is not in dispute. The magnitude is, substantially, and the disagreement is about method and about what counts as land available for trees rather than about the direction of the effect.
How far it can be extended
The dispute concerns a global estimate and the methods used to produce it.
Caveats
None of this is an argument against restoring forest where forest belongs; it is an argument against a single global number.
Storage in new trees takes decades to accumulate, while emissions occur now.
Where researchers disagree
The original analysis and its critics disagree on how much land is genuinely available, with estimates of restorable area differing by a large factor.
Treating grasslands and savannas as candidates for tree cover is disputed on ecological grounds: they are ancient systems holding much of their carbon below ground, and planting trees in them can reduce biodiversity and, in some cases, net carbon.
Still unanswered
How much land could support additional tree cover without displacing food production or damaging open ecosystems.
How permanent that storage would be under intensifying disturbance.
Argues that much of the mapped area is ancient grassland and savanna rather than degraded forest, and that the carbon benefit was substantially overstated.
Shows that carbon stored in new forest is exposed to fire, drought and insect risk that accounting schemes tend to under-weight.
This is the clearest example on the site of a striking figure travelling much further than the objections to it. A 2019 global analysis put the restoration potential at a very large quantity of carbon; formal technical comments in the same journal disputed the restorable area, disputed the carbon accounting, and objected that a large share of the mapped land was ancient grassland and savanna rather than degraded forest. The authors corrected the carbon figure. The corrected version and the objections travelled nowhere near as far as the original.
Climatic suitability for tree cover is not the same as land being available, appropriate or ecologically empty.
Ancient grasslands hold much of their carbon below ground, and planting trees into them can reduce both biodiversity and, in places, net carbon.
New forest accumulates carbon over decades, while emissions occur now.
Stored carbon is exposed to fire, drought and clearance, which crediting schemes tend to under-weight.
None of this is an argument against restoring forest where forest belongs. It is an argument against a single global number.
How much forest soil carbon is genuinely long-lived?
Why it matters: Soil is often the largest pool in a forest and is the least well constrained term in every carbon budget, so the uncertainty propagates into everything built on those budgets.
What would settle it: Long-term measurements of soil carbon turnover across forest types, which are slow, expensive and rare.
Will the tropical sink keep weakening?
Why it matters: Amazonian uptake has fallen by about a third from its peak, driven by mortality. Whether that continues determines a large part of the land carbon budget.
How much land could carry additional tree cover without doing harm?
Why it matters: The disagreement between the 2019 restoration estimate and its critics is largely about this, and it is a question about ecosystems and land use rather than about trees.
How complete this record is, and what it is still missing
NatureHQ publishes its own gaps. This record is at 87% completeness against what we would call a finished subject, and was last reviewed on 2026-08-30. It carries 9 claims and answers 5 mapped search questions.
1 high-priority search intent(s) not yet covered
Carbon markets, offsetting schemes and their governance are deliberately out of scope and heavily searched.
Wood products as a carbon store — how long a timber building holds its carbon — is a genuine gap.
Peatland and mangrove carbon are much larger stores per hectare than most forests and are not covered.