Global Carbon Budget 2025 · 2015–2024 averages

The carbon that stays

The Mauna Loa curve does not measure what we emit. It measures what is left over after the oceans and vegetation have taken their share. Every year they absorb about half of it, for free.

The units on this page. GtCO₂ = billions of tonnes of CO₂. ppm = molecules of CO₂ per million molecules of air. Sink = whatever takes CO₂ out of the air: the ocean and vegetation. CO₂e = all greenhouse gases converted into the amount of CO₂ that would warm the same; used where methane and nitrous oxide count, as in the section on wars. Averages for 2015–2024 from the Global Carbon Budget 2025.

40.9GtCO₂ / year Emitted: fossil fuels plus land use
11.8GtCO₂ / year Absorbed by the ocean
8.7GtCO₂ / year Absorbed by vegetation
20.4GtCO₂ / year · 2.6 ppm Stays in the air: what the curve measures
ppm

The Mauna Loa curve

Monthly mean CO₂ concentration at the Mauna Loa observatory, in Hawaii, from 1959 to today. The yearly wobble is the breathing of the northern hemisphere's vegetation: peak in May, trough in late September.

CO₂ is rising, and faster and faster

observed

Mauna Loa, 1959–today. Thin line: monthly means; thick line: annual mean; the dots mark reference values.

How to read it. The vertical axis starts at 305 ppm, not at zero: that way you see the slope, which is the information. The sawtooth is the season, not noise.
Go deeper: the speed by decade, and how warm it was
Growth rate by decade, and how warm it was
Periodppm / year°C above 1850–1900Note
1960–19690.86+0.16The starting decade of the comparison
1970–19791.28+0.23
1980–19891.61+0.43
1990–19991.51+0.56The only decade that slowed: the Pinatubo eruption (1991), the Soviet collapse
2000–20091.97+0.74
2010–20192.42+0.97
2016–20252.56+1.18≈ 3 times the 1960s
ppm

The last time the air was like this

The bubbles trapped in Antarctic ice are real air from eight hundred thousand years ago, and it can be measured. In all that time CO₂ swung between 170 and 300 parts per million, back and forth, with the ice ages. Never once above.

Eight hundred thousand years of breathing, and then us

observed

CO₂ concentration from Antarctic ice cores up to 1958, then the annual means from Mauna Loa. The horizontal axis is linear: each pixel is worth about a thousand years, so the whole industrial era fits in less than one pixel. The vertical line on the right is not a drawing error.

The sawtooth is the ice ages, eight cycles of about a hundred thousand years each: CO₂ fell to 180 ppm as the ice advanced and climbed back to 280 as it retreated, always together with temperature. The vertical line on the right is us: the same distance covered in two hundred years instead of ten thousand, and climbing past the ceiling of the whole record.
Go deeper: what the world was like the last time CO₂ was this high
Palaeoclimate reference periods, IPCC AR6 values relative to 1850–1900
PeriodCO₂ ppmTemperatureSea level
Today427+1.3 °Creference
Last glacial maximum · 20 thousand years ago≈190−5 to −7 °C−120 m
Last interglacial · 125 thousand years ago≈280+0.5 to +1.5 °C+5 to +10 m
Warm Pliocene · 3.3 million years ago360–420+2.5 to +4 °C+5 to +25 m
Miocene optimum · 15 million years ago≈480+5 to +10 °C—

The Pliocene is not a forecast

It is the row that matters, because the CO₂ back then is the CO₂ of today: between 360 and 420 ppm, for a long stretch. The world that came with it was 2.5–4 °C warmer and had the sea between 5 and 25 metres higher. But those values are equilibrium states, reached over millennia. The Greenland and Antarctic ice sheets take thousands of years to respond, and the Pliocene tells you where the planet ends up if CO₂ stays there, not what happens by 2100. Projections for this century are another matter and give tens of centimetres, not tens of metres.

Why the ice ages are not an objection

«CO₂ has changed on its own before» is true, and you can see it clearly in the chart. Those cycles start from changes in the Earth's orbit, which by themselves move little energy: it is the CO₂ released by warming oceans that amplifies them into an ice age. The difference with today is the cause and the speed: those transitions added 100 ppm in about ten thousand years, we have added 140 in two hundred.

Three rulers, three answers

Ice cores are direct measurements of the air and reach eight hundred thousand years. To go further you need indirect indicators, such as boron in the shells of foraminifera: the IPCC, in 2021, concludes from these that current CO₂ is the highest in at least two million years. The Cenozoic synthesis published in Science in 2023 revised the reconstruction of the last 66 million years and places the last period with CO₂ steadily at these levels around fourteen million years ago. The number that often circulates, «three million years», comes from earlier reconstructions.

ppm / year

The last ten years, up close

On the long curve a single year is indistinguishable. Here it is not: and you can see that the years when CO₂ rose the most, 2016, 2023 and 2024, are the same years when the planet was warmest.

The warm years are the years it rises most

observed

The last ten years, three panels on the same time axis. Top: monthly means from Mauna Loa (thin line), the underlying rise once the seasonal breathing of plants is removed (thick line) and the annual means (dots). Middle: how much CO₂ grew in each year, from 1 January to 31 December; the solid bars are the years above average. Bottom: global temperature in the same year.

The years when CO₂ grows most are the warm years: 2016, 2023, 2024. It is not the temperature that produces the extra CO₂: it is El Niño, the periodic warming of the Pacific, which warms the planet and dries the tropics at the same time, so vegetation absorbs less. Once El Niño ended, growth in 2025 was back to 2.2 ppm.
°C

So what? What those ppm actually do

So far we have been counting molecules. More CO₂ in the air means less heat escaping to space: this is the same chart as before, measured in degrees.

The planet has already warmed by about 1.3 degrees

observed

How much warmer each year was than the 1850–1900 average, land and oceans together (NOAAGlobalTemp). It is the same baseline the 1.5 °C goal refers to.

One year above 1.5 °C is not the threshold crossed. The limit is judged on underlying warming, that is on an average of several years: here it is the last ten, today 1.18 °C. The Paris Agreement does not prescribe how many years, and the IPCC uses twenty. A single warm year can be pushed up by El Niño, as in 2024; a cooler year takes nothing back, as in 2025: it is the same post-El Niño return seen in CO₂, and the underlying trend has not changed. The budget in the data box is how much we can still emit for a 50% chance of staying below 1.5 °C: 170 GtCO₂ from the start of 2026, minus the months already gone.2025 is cooler than 2024, and that is not good news. It is the same post-El Niño return seen in CO₂: the underlying trend has not changed.
Go deeper: the thermometers disagree, and CO₂ is not alone

A tenth of a degree of disagreement between thermometers

This page uses NOAAGlobalTemp. Other authoritative series — ERA5 from Copernicus, HadCRUT from the Met Office, Berkeley Earth — give values between 1.5 and 1.6 °C for 2024 (ERA5 1.60; the World Meteorological Organization synthesis 1.55). The same NOAA series, in its previous version, gave 1.46: the 2025 revision lowered recent values by a few hundredths. The difference comes from how each one covers the areas without stations, above all the Arctic, and from how it estimates the pre-industrial period, for which measurements are few. When you read «we have passed 1.5 °C», it is worth asking with which ruler.

And carbon dioxide is not alone

CO₂ 425.6 ppm NOAA global mean, +52% above pre-industrial; Mauna Loa, further north, reads slightly higher
Methane · CH₄ 1936 ppb ppb = parts per billion · +168% above pre-industrial (722)
Nitrous oxide · N₂O 338.9 ppb +26% above pre-industrial (270)

Methane is far rarer but, molecule for molecule, it traps much more heat than CO₂ in the short term; in exchange it breaks down within about ten years, while a fraction of CO₂ stays for centuries. That is why cutting methane works fast and cutting CO₂ works forever: both are needed, and they are not interchangeable.

W / m²

Why it warms: the energy account

Underneath the degrees there is a simpler account: how much energy arrives from the Sun, how much leaves again towards space as infrared. Greenhouse gases make the way out harder, the account opens up, and the planet accumulates — almost all of it in the sea — until the two sides balance again.

Of 240 watts absorbed, just over one stays in

observed and estimated

Average fluxes at the top of the atmosphere, in watts per square metre of the Earth's surface. Above, the fluxes from IPCC AR6 (figure 7.2, early-century conditions); below, the imbalance magnified and split by destination: 1.12 W/m² on average between 2013 and 2025, partition 1971–2020.

The imbalance is not measured by subtracting 239 from 240. Each flux on its own is known to within a few watts per square metre: a difference worth one would vanish inside that uncertainty. It is measured from the other end, by counting the heat that piles up — and since almost all of it piles up in the sea, the most reliable thermometer of global warming is a float in the Pacific, not one in the garden. The red line is drawn to scale.The red line is drawn to scale. The imbalance really is that fraction of the fluxes: just over one watt in two hundred and forty. And it is not measured by subtracting the two fluxes, which are known to within a few watts: it is measured by counting the heat that piles up, almost all of it in the ocean.
Go deeper: how a gas holds heat back, and where the energy goes

How a gas holds heat back

The Earth's surface emits an average of 398 watts per square metre of infrared radiation. Only 239 reach space: the 159 in between are the greenhouse effect, and they are what keeps the surface at about 15 °C instead of the −18 °C of a bare planet re-emitting the 240 watts it absorbs. Greenhouse gases do not reflect heat like a mirror: they absorb it and re-emit it in every direction, downwards included, and they do so from ever higher altitudes as their concentration grows.

This is the mechanism, and it is subtler than it is usually told. Higher up it is colder, and a colder body radiates less: adding CO₂ means that the infrared which actually leaves for space comes from colder layers, so less of it leaves. The account opens up. At that point the planet — surface, atmosphere, ocean — warms until the outgoing infrared balances the absorbed sunlight again. The imbalance measures how far that balance still is: it is the share of warming already decided and not yet delivered.

Why the air thermometer only measures the tip

Between 1971 and 2020 the Earth system piled up 381 ZJ of heat, with an uncertainty of 61. The atmosphere took 5 of them: very few, and they were enough to warm the air by about a degree, because air has almost no heat capacity. The ocean took about 341, nearly seventy times as many, and warmed barely at all: in the top two thousand metres, where the measurements are dense, by just over a tenth of a degree. It is the same phenomenon read on two instruments of wildly different sensitivity: the big number is in the sea, the visible number is in the air.

Where the heat piled up between 1971 and 2020 ended up (von Schuckmann et al. 2023)
DestinationZJShare
Ocean, all depths≈34189%
Land21 ± 26%
Melting ice14 ± 44%
Atmosphere5 ± 11%
Total381 ± 61100%

The ocean row is obtained by difference: the study publishes the total and the three smaller rows. IPCC AR6, over 1971–2018 and with a slightly different method, gives 91% to the ocean, 5% to land, 3% to ice and 1% to the atmosphere: the two partitions say the same thing.

If it had all gone into the air

An arithmetic exercise, not a scenario. The atmosphere weighs 5.15·10¹⁸ kg and its specific heat is about 1005 joules per kilogram per degree: 5.2·10²¹ joules per degree in all, that is 5.2 ZJ. The 381 ZJ piled up since 1971, had they gone there alone, would have warmed it by more than seventy degrees. It did not happen, and it could not have — air that hot would radiate far more and the budget would close by itself — but it gives the measure of what the oceans are absorbing on our behalf, and of how exposed anyone is who judges warming from air temperature alone.

The same sum for water: a square metre of ocean one metre deep is 1027 kg of water at 3990 joules per kilogram per degree, and the surface of the oceans is 3.62·10¹⁴ m². It follows that a single metre of ocean is worth 1.5 ZJ per degree, and that the whole atmosphere is equivalent to three and a half metres of water.

The imbalance is growing

Earth's energy imbalance, in watts per square metre (Forster et al., Indicators of Global Climate Change 2025)
PeriodW/m²90% range
1976–19950.40−0.03 to 0.84
2006–20251.040.82 to 1.25
2013–20251.120.78 to 1.46

More than doubled in thirty years. Why is still under discussion: rising greenhouse gases play a part, but so does the fall in aerosols — the dust that dirtied the air and reflected a little sunlight — and a decline in low cloud that in recent years has let more Sun be absorbed. The jump in 2023 and 2024 is still being investigated.

A note worth reading twice: this is the quantity the Paris Agreement does not name. The negotiation is about degrees, which are a delayed and noisy effect; the imbalance is the cause, it moves first and makes less noise. The authors of the 2023 study have been asking for years that it enter the Agreement's global stocktake.

GtCO₂ / year

Where what we emit ends up

The global carbon budget has two sides. On the left what we put in, on the right the three possible destinations. Only the third one shows up in the curve.

Half of what we emit disappears into ocean and vegetation

observed and estimated

Global carbon budget, annual average 2015–2024, in billions of tonnes of CO₂. Fossil fuels net of the small share that cement reabsorbs. 1 ppm = 7.78 GtCO₂.

The whiskers under the blocks are the uncertainty. The land sink is worth 8.7 GtCO₂ but is known to within ±2.9, a third of it; we know the ocean twice as well. Over the decade average the budget balances; in a single year it does not, and 2024 is the most glaring case.
The budget in figures, 2015–2024 average
Global Carbon Budget 2025, 2015–2024 averages converted into GtCO₂
ItemGtCO₂/yearppm equiv.Share
Fossil fuels and cement (net)35.94.6188%
Land use5.00.6412%
Total emitted40.95.26100%
Ocean sink11.81.5229%
Land sink8.71.1221%
Total absorbed20.52.6350%
Stays in the atmosphere20.42.6250%
Imbalance (decade average)≈0——
ppm

The work the sinks have already done

One way to see what ocean and vegetation are worth: draw where the curve would be if they had absorbed nothing.

Without the sinks we would be at 574 ppm, not 427

estimate

The dashed curve adds to the 1959 value all the emissions accumulated since then. The gap between the two lines is the CO₂ that oceans and vegetation have taken out of the air, for free, in sixty-six years.

  • Observed at Mauna Loa
  • If no sink had absorbed anything
  • What the sinks removed
147 ppm of difference. That is more than the concentration has risen in the whole observed period. Put another way: of the roughly 258 ppm we have put in since 1959, we see 111 in the air — the rest is in the ocean and in vegetation. The calculation assumes all emissions had stayed in the atmosphere; it is an accounting comparison, not a simulation of the climate that would have followed.
ppm / year

Why 2024 depends on where you look

In 2024 emissions grew by about 1%. At Mauna Loa CO₂ rose as much as the year before; in the global mean it made the biggest jump since the series began, in 1980. This is not a contradiction: they are two different instruments, and a single year is not enough to judge.

The same year, measured two ways

observed

Annual growth rate of CO₂. Mauna Loa is a single observatory in the Pacific; the global mean is computed from stations far from cities and industry, combined so that each band of the planet weighs as much as its surface.

  • Mauna Loa
  • Global mean
In 2023 Mauna Loa was running faster than the planet, in 2024 the opposite. A single observatory depends on how the winds bring it air: excellent for the long trend, less so for judging a single year. When someone quotes «2024 did X ppm», they should say which of the two measures they are using.
Go deeper: the 2024 that does not add up, and how it ended in 2025

For 2024 the budget does not add up: summing what ended in the air, in the sea and in vegetation gives more than what appears to have been emitted, with a gap of 6 GtCO₂, the most negative in the series. The Global Carbon Project publishes it as an item of its own. In 2025 growth fell to 2.23 ppm at Mauna Loa and 2.06 in the global mean, and the preliminary Global Carbon Budget estimate indicates that the land sink is back to its former strength. The anomaly was a swing, not a break: building a thesis on 2024 alone would be unwise in either direction.

GtCO₂ · t/person · %

Who emits, in three different readings

These are three distinct questions and they give three distinct answers. All three are needed: who emits most today, who emits most per person, and who filled the tub.

A third of emissions comes from China. Per person and cumulatively, the ranking changes

observed

Emissions from fossil fuels and cement, excluding land use (Global Carbon Budget 2025 via Our World in Data). Each panel has its own scale: compare positions, not lengths across panels.

In 2024

billions of tonnes of CO₂ and share of the world total

China
12.2931.8%
United States
4.9012.7%
India
3.198.3%
EU 27
2.436.3%
Russia
1.784.6%
Air and seainternational
1.193.1%
Japan
0.962.5%
Italy
0.300.8%
Rest of the worldabout 190 countries
11.5529.9%

Per person, 2024

tonnes of CO₂ per inhabitant per year

United States
14.2
Russia
12.3
China
8.7
Japan
7.8
EU 27
5.4
Italy
5.1
India
2.2

Cumulative 1850–2024

% of all fossil carbon ever emitted

United States
23.5%
EU 27
16.3%
China
15.4%
Russia
6.6%
India
3.6%
Italy
1.4%
World total 38.6 GtCO₂ in 2024: China alone accounts for almost a third, Italy for less than 1%. The cumulative figure is the reading that matters for the climate: the CO₂ emitted in 1950 is still up there. Which number to quote: 38.6 is fossil fuels alone for one year; the 40.9 at the top of the page is the 2015–2024 average including land use.
Go deeper: from which fuels and from which uses
Where fossil CO₂ comes from · world 2023 (fuels: Global Carbon Budget; end uses: IEA estimates)
End uses follow the IEA classification: electricity is counted where it is produced, not where it is consumed. With the economic-sector breakdown used by the IPCC the same emissions give different percentages, because industry and buildings take back their share of electricity.
By fuelShareBy end useShare
Coal41%Electricity and heat40%
Oil32%Transport23%
Natural gas21%Industry (direct)19%
Cement4%Buildings9%
Other, including gas flared at wells2%More9%
change 2000–2024

Is it «just China»?

No for the level, largely yes for the growth. China accounts for a third of today's total. Of the increase over the last twenty-five years it accounts for about two thirds, because in the meantime Europe and the United States have come down.

Since 2000 China explains two thirds of the increase

observed

How much each one's contribution changed from 2000 to 2024, in billions of tonnes of CO₂ per year.

  • Emissions increased
  • Emissions decreased
China
+8.6
Rest of the world and international transport
+4.5
India
+2.2
Russia
+0.3
Japan
−0.3
United States
−1.1
European Union
−1.2
World net balance +13.1 GtCO₂, from 25.5 to 38.6. China alone covers about two thirds of that net increase.
% / year

Is China slowing down?

Yes, and by a lot: the pace has gone from +3.9% a year in 2023 to −0.3% in 2025, and since the peak of March 2024 Chinese emissions have hovered around zero. But stopping is not coming down, and China has stopped at the top. Meanwhile, in 2025, those who were coming down stopped coming down.

From +3.9% a year to zero, in three years

observed and estimated

Change in CO₂ emissions from fossil fuels and cement against the same period a year earlier. Up to 2024 it is the Global Carbon Budget 2025 inventory; 2025 and the 2026 quarters are estimates from China's official monthly energy data. Each panel has its own scale.

  • Emissions increased
  • Emissions decreased

China, year by year

change on the matching period, %

2021
+3.6%
2022
+3.8%
2023
+3.9%
2024
+1.0%
2025
−0.3%
2026first quarter
+2.0%
2026second quarter
−1.0%

The other big emitters, same years

annual average 2019–2024, %

India
+4.1%
China
+2.8%
Russia
+0.9%
World
+0.8%
United States
−1.3%
Japan
−2.7%
EU 27
−3.5%
In the five years to 2024 China added 1.58 billion tonnes a year: more than the entire world increase, which was 1.51. Everything else almost cancels out: as much as India and Russia went up, the European Union, the United States and Japan came down. Then China's pace collapsed and 2025 closed below zero. It had happened before, in 2015 and 2016, but energy demand was slowing then as well: this time it was not.
Go deeper: why it stopped, why it can start again, and the exact numbers

Why it stopped. In 2025 Chinese solar grew by 43% and wind by 14%: 490 TWh of extra clean electricity in a single year, enough on their own to cover the whole increase in electricity demand. Then came the road: in the second quarter of 2026 China’s oil consumption fell by 9%, and by 16% in transport, while travel was still increasing. It is the first time that what brought Chinese emissions down was oil and not coal.

Why it can start again. Part of that fall is not structure but shock: with the Strait of Hormuz crisis prices rose and in one quarter China cut crude imports by 32%. And the plateau is so flat that it takes little to pass the old peak: in the first quarter of 2026 emissions were already back at +2%. It is a race between energy demand growing and renewables growing, and so far the outcome has been decided one quarter at a time.

The largest emitters over the last five years · CO₂ from fossil fuels and cement, billions of tonnes
Territorial emissions gross of cement carbonation, from the Global Carbon Budget 2025 database. The annual average is compounded over the five years between the two endpoints: the pandemic year 2020 sits inside the period, but not at its ends, and does not inflate the result. The 2025 column is the projection published by the Global Carbon Budget, which does not publish a separate entry for Russia and Japan; for China the year-end estimate from monthly energy data gives −0.3% instead, just below the bottom of the projection's range, which ran from −0.1% to +0.9%.
Country20192024ChangeAnnual average2025 (projection)
China10.7112.29+14.7%+2.8%+0.4%
United States5.244.90−6.3%−1.3%+2.5%
India2.613.19+22.3%+4.1%+1.1%
EU 272.902.43−16.5%−3.5%−0.1%
Russia1.701.78+4.5%+0.9%—
Japan1.100.96−12.7%−2.7%—
World37.0938.60+4.1%+0.8%+1.0%
MtCO₂e

How much the wars weigh

Two counter-intuitive answers. Armies in peacetime emit far more than the wars they fight. And inside a war the weapons weigh almost nothing: what counts is what they hit, and what then has to be rebuilt. Together they are worth about 5% of the world's greenhouse gases. These are the least solid estimates on this page, and they are floors, not measurements.

Weapons are 1%: the rest is what they hit

estimate

Greenhouse gases in millions of tonnes of CO₂ equivalent. On the left the total for each case, over different periods indicated under each item; on the right where that total comes from in the war in Iran, the most recent breakdown available. Each panel has its own scale.

How much it weighs in total

mind the period: it changes from item to item

The world's armiesevery year, wars excluded
2 750
Ukrainefour years, reconstruction included
311
Italyone year, for comparison
300
Israel and Gazaconflict and reconstruction
33
Iranfirst fourteen days
5.1

What it is made of

preliminary estimate war in Iran, 28 February – 14 March 2026

Buildings destroyed
2.40
Oil burned
1.88
Fuel for vehicles
0.53
Vehicles destroyed
0.17
Missiles and drones
0.06
In two weeks of war in Iran the missiles and drones are worth 55 thousand tonnes of CO₂e, the buildings hit and the oil burned four million three hundred thousand: almost eighty times as much. That is why you do not do the accounting of a war by counting munitions. Mind what these totals include: in Ukraine 73 of the 311 million tonnes are the reconstruction still to come, in Gaza 31 out of 33, and military operations alone in Gaza are worth 1.3. They are not in the atmosphere yet: they are attributed to the war that will make them necessary.
Go deeper: where the emissions of a war come from
Four years of war in Ukraine, 24 February 2022 – 23 February 2026
From whatMtCO₂eShareNote
Fighting114.137%Fuel for vehicles and aircraft, munitions
Reconstruction73.323%Cement and steel to rebuild what is destroyed
Fires70.323%Woods and fields burned: 1.39 million hectares in 2025 alone
Civil aviation29.89%Flights diverted around the closed airspace
Energy infrastructure18.66%Plants hit, including methane leaks
Refugees5.32%Displacement of millions of people
Total over four years311.4100%The fourth year alone is worth 75

In Ukraine too the weapons are the small part

Fighting is little more than a third of the total. The rest is what the war drags behind it: rebuilding, failing to put out the fires, diverting aircraft, moving people. Reconstruction alone weighs almost as much as the fighting, and grows as the destruction piles up. At 185 dollars a tonne, the climate damage Ukraine intends to claim exceeds 57 billion.

Israel and Gaza: 93% is the aftermath

The same pattern, even sharper. Military operations are worth a little over 1.3 MtCO₂e up to January 2025; the total reaches 33.2 counting the barriers built before and the reconstruction to come: almost 450 thousand flats, over 3,000 kilometres of roads, schools, hospitals, water systems. If the number looks small next to Ukraine it is because the territory is tiny and the count of operations is narrow: it covers flights and vehicles, not every building that burned.

Iran: what was hit, and what the count is missing

Around thirty oil processing and storage plants in Iran, plus refineries and tankers outside the country: that is where the 1.88 million tonnes of the «oil burned» item come from. The count covers the first two weeks and has not yet been through a scientific journal: we have marked it as a preliminary estimate in the chart. Take it for the order of magnitude and for the proportions between items, not for the exact figure.

The Syrian paradox

During the war the emissions declared by Syria fell, because oil production collapsed by more than a quarter. Meanwhile the damage was piling up elsewhere: putting the destroyed houses back up alone is worth about 22 MtCO₂, and it will arrive in the years of reconstruction. In official accounts a war can look like a climate policy, which is another reason why the share alongside is a floor.

What no count includes

Even the best accounts measure one thing only, greenhouse gases. Left out are the mines that make fields unfarmable, the heavy metals in the soil — lead, zinc, chromium, arsenic around Sumy —, the PFAS and asbestos lifted from the rubble, the dioxins of the fires, the crude that ended up in the Persian Gulf. That is not climate, it is a different damage, and none of the percentages on this page include it. Vegetation fires are left out too: in the sixteen countries with ongoing or recent conflicts, in 2020 alone, 1,456 MtCO₂e went into the air. Not all of it caused by war, but attributed to nobody.

A precedent: the Kuwait wells

In 1991, retreating from Kuwait, the Iraqi army set fire to hundreds of oil wells. Those fires alone were worth more than 2% of all the fossil CO₂ emitted in the world that year: more than four years of war in Ukraine have produced. It is the limiting case that explains why hitting energy, rather than people, is what weighs most on the climate.

How those two percentages are built

The denominator is the world total of all greenhouse gases in 2024, 57.7 GtCO₂e, from the UNEP emissions gap report. The 4.8% is the 2,750 MtCO₂e of the armies: the authors of the estimate write 5.5%, computed on an older and lower world total, and the substance does not change. The 0.2% is the sum of what is attributed to each year of war: 78 MtCO₂e for Ukraine, four years divided by four, and 27 for the conflict between Israel and Gaza. The two weeks in Iran are worth another 5 MtCO₂e, a thousandth of the world total, and cannot be annualised.

Does the armies' 5.5% include the wars? No

It is the question that number drags behind it, and the answer changes what the sum means. The 4.8% is the apparatus that stands there anyway: fuel for vehicles, ships and aircraft, the bases, and above all the industry that builds and supplies them. Of that total only a fifth, 1% of world emissions, is fuel burned by the armed forces; the rest is the supply chain. The authors of the estimate write that it is conservative precisely because it leaves out the emissions of the fighting: what happens when the shooting starts is not in there.

The wars' 0.2% counts exactly that: fuel spent in battle, munitions, destroyed equipment to replace, burned buildings, reconstruction. The two figures measure different things, which is why they can be added without counting anything twice. A grey zone does exist — an army at war burns more than an army in barracks, and that extra could show up on both sides — but it is small next to the range of the first figure, which runs from 2.8 to 6.1%.

And military spending is rising

Every 100 billion dollars of extra military spending is worth about 32 MtCO₂e a year. The increase in NATO spending between 2019 and 2024 has already added around sixty, and the announced increases would bring another 132: as much as the fossil emissions of a country like Chile, every year. The figures have a wide margin, the sign does not.

1997–2012

Why Kyoto did not bend the curve

The protocol was not designed to stop the growth. And even if it had worked perfectly, the concentration would have risen anyway.

Between the signature and the end of the first period, emissions rose by 43%

observed

World fossil emissions 1990–2024, in billions of tonnes of CO₂. The band marks the 2008–2012 commitment period.

The tub fills up anyway. These are fossil fuels alone, year by year. Kyoto bound a subset of countries to cut by 5%; between 1997 and 2012 the rest of the world grew by over 60%, and the countries without commitments doubled. The only visible drop is 2020, and no treaty deserves the credit.
Go deeper: the chronology of Kyoto, and what it left behind
  • 1997
    Signature of the protocolTarget: minus 5.2% relative to 1990 for industrialised countries only, by 2008–2012. China, India and Brazil have no commitments.
  • 2001
    The United States does not ratifyThe largest emitter in the world at the time stays out. The treaty covers less than a third of global emissions, a share falling year by year.
  • 2005
    Entry into forceIn the meantime global emissions have already risen from 24.4 to 29.6 GtCO₂ compared with the year of the signature.
  • 2011
    Canada withdrawsTo avoid the penalties foreseen for missing the target.
  • 2012
    End of the first periodThe bound countries meet the target, but most of the fall comes from the post-1991 collapse of ex-Soviet industry and from offshoring, not from climate policy.
To be said anyway

Kyoto was not useless. It created the national inventories, the measurement and verification systems and the carbon markets, including the European emissions trading scheme (ETS). It is the infrastructure on which the Paris Agreement was then built. What it did not do is reverse or halt the growth of world emissions, and indeed it did not stop the rise of CO₂ in the atmosphere: precisely in the Kyoto years the concentration kept accelerating, from 1.5 to 2.0 ppm a year between the 1990s and the following decade. These are two separate accounts, and they should be kept separate. What would bend the curve: net zero emissions, not reduced emissions. That is the goal the Paris Agreement later set for mid-century.

in brief

Seven things to take away

If one screen of this page is all that stays with you, let it be this one. The numbers below update together with those in the previous sections.

  1. CO₂ in the air is at 427.35 ppm, 53% above the pre-industrial level, and it is rising by about 2.6 ppm a year: three times faster than in the 1960s.
  2. Every year we emit about 41 billion tonnes; half disappears into ocean and vegetation. The curve measures only the other half. Without the sinks we would be at 574 ppm instead of 427.
  3. The planet warmed by +1.29 °C in the last year, +1.18 on average over the last decade. The Paris limit, 1.5 °C, is judged on an average of several years, not on a single one: it has not been passed yet, and about 140 GtCO₂ remain for a 50% chance of staying below it, less than four years of emissions at the current rate.
  4. Warming is not made of air: about 90% of the surplus energy has ended up in the ocean. The planet keeps just over one watt per square metre more than it gives back to space, and since 1971 it has piled up 381 ZJ of heat. The atmosphere took 1% of it: air temperature is the thin tip, and the real thermometer is at sea.
  5. The warm years are the years when CO₂ rises most. In 2023 and 2024 El Niño weakened vegetation and growth reached 3.3 ppm. A single year makes no trend, in either direction.
  6. Who emits depends on the question. Today China emits a third of the total, but an American emits one and a half times a Chinese person each, and over a hundred and seventy years the United States and Europe filled the tub more than anyone. Cutting slows the rise; to stop it you need net zero emissions.
  7. China has stopped accelerating, but a plateau is not a descent. Chinese emissions have been flat for two years at the highest level ever reached and in 2025 they fell by 0.3%; in the same year US emissions grew by 2.5% and the world total by 1.0%, another record.
notes

Sources, conversions and uncertainties

Each title links to the source, in a new tab: the original document, not an article about it.

Go deeper: conversions and two measures not to be confused

Conversions used on this page

1 ppm = 2.124 GtC = 7.78 GtCO₂
1 GtC = 3.664 GtCO₂

Two measures that should not be confused

The NOAA growth rate is the increase between 1 January and 31 December, computed on the thick line of the charts, the one without the seasonal breathing. The difference between annual means is another thing and gives different numbers: in 2024, 3.33 ppm the first, 3.53 the second. Also, Mauna Loa is not the planet: it is an observatory at 3,400 metres in the Pacific, while the NOAA global mean is computed from marine stations far from cities and industry, where the air is well mixed, combined so that each latitude band weighs as much as its surface. In 2024 the two measures give 3.33 and 3.76 ppm.

The GtC (billions of tonnes of carbon) that appear in scientific publications convert into GtCO₂ by multiplying by 3.664: this page always uses GtCO₂.

What has changed on this page: the revision log

Climate science corrects itself, and this page does the same. The most recent entries are at the top.

  • 16 September 2026
    English moved to the root, Italian moved to /it/. This page is read mostly outside Italy, and the main address now says so. The old /en/ redirects to the root, so links already out there keep working. The change also fixed a flaw that had always been there: before, anyone arriving with a language we do not have — German, Portuguese, Japanese — was sent to /en/, that is, to a page they had not asked for. Now the server redirects only to the four languages that actually exist, and everyone else stays where they landed. The hand-written source is still Italian: everything else is generated from it, English included.
  • 15 September 2026
    The page is now in French too, and the sources are clickable. Five languages: Italian, English, Spanish, French and Chinese, and anyone arriving with a French browser now lands there by itself. The translation is complete — text, captions, tables, chart labels and status messages — and French, like Italian and Spanish, uses the comma for decimals: 40,9 GtCO₂, not 40.9. In the sources section, meanwhile, the eighteen titles now lead to the original document instead of merely naming it.
  • 15 September 2026
    Added the section on energy: why CO₂ warms, and where the heat goes. The page went from molecules to degrees without saying what lies in between. In between there is an account: of the 340 watts per square metre that arrive from the Sun, one hundred are reflected and two hundred and forty absorbed, and of those two hundred and forty, two hundred and thirty-nine leave again for space. What is left — just over a watt — is the whole of the warming under way, and it has not returned to zero for half a century. With the detail that changes the scale of the problem: of that heat the ocean takes about 90% and the atmosphere 1%, so judging warming from air temperature alone means looking at one per cent of what is happening. The section also says something usually taken for granted: the imbalance is not obtained by subtracting the two fluxes, which are known to within a few watts, but by counting the heat that piles up at sea.
  • 15 September 2026
    Added the section on the present: who is actually slowing down. The page said how much each country emits and how much it has grown since 2000, but it did not answer the question that comes straight after: is the direction changing now? For China yes, and fast — +3.9% in 2023, +1.0% in 2024, −0.3% in 2025, twenty-one months flat or falling since the peak of March 2024 — and it is the first time this has happened without a slowdown in energy demand behind it. The rest of the section is the uncomfortable part: stopping is not coming down, in 2025 the United States grew by 2.5% and the European Union stalled, and the world total set another record.
  • 14 September 2026
    A false sentence about the wars corrected. It said that «neither of the two figures includes the fighting itself»: true for the armies' 4.8%, not for the wars' 0.2%, which counts the fighting very much indeed. In its place there is now the answer to the question that number drags behind it: the armies' 5.5% does not include the wars, it is the apparatus that stands there anyway, and of that total only a fifth is fuel burned by the armed forces, the rest is the industry that supplies them. That is why the two figures can be added without counting anything twice.
  • 14 September 2026
    Cross-checked against an investigation in Corriere della Sera. A reader pointed to Milena Gabanelli's Dataroom on the emissions of wars: every figure matches the ones on this page, because they come from the same studies. While we were at it, what was missing has been added: of the 61 conflicts active in 2024 only two have a carbon count, Lebanon joins the list of those that do not, and a note says that even the best accounts measure greenhouse gases alone, not the mines, the heavy metals or the oil at sea. One figure from the article was not taken up: the «59 wars, the highest number since 1945» are in fact 61 active conflicts, while 59 is the 2008 value of a different indicator.
  • 14 September 2026
    The share of the world total, which was missing. The section on wars gave absolute figures and comparisons, but did not answer the most direct question: how much do armies and wars weigh on one year of emissions. Now it does, with the denominator stated: 57.7 billion tonnes of greenhouse gases in 2024, from the UNEP report. Five squares out of a hundred, of which 4.8 the military apparatus and 0.2 the two wars for which a count exists.
  • 14 September 2026
    The page is in four languages — Italian, English, Spanish and Chinese — and links to its own repository: corrections and reports are welcome.
  • 13 September 2026
    The eight hundred thousand years added. The page was missing the long scale: now there is the series from Antarctic ice cores, where you can see that CO₂ never went above 300 ppm in eight glacial cycles, and that no human being has ever breathed today's air. With the right numbers: the record is not «three million years», a figure from superseded reconstructions, but at least two million according to the IPCC and about fourteen according to the 2023 synthesis. The Pliocene, the last epoch with this much CO₂, was +2.5 to +4 °C with the sea between 5 and 25 metres higher, not 35: and those were equilibria reached over millennia, not forecasts for 2100.
  • 13 September 2026
    Every figure declares what kind of number it shows. Observed, observed and estimated, estimate, preliminary estimate: four labels, explained below. While we were at it, some over-firm wordings were corrected. The 1.5 °C limit is not «measured on a decade average»: the Agreement does not prescribe how many years, we use ten and the IPCC twenty. The remaining budget is not a date but a 50% probability. Of Kyoto we no longer say that the effect is «essentially nil» but that it did not reverse the growth of emissions, which is what the data show. «Cutting is not enough» has become «cutting slows the rise, stopping it requires net zero emissions». In the section on wars it is now explicit that part of the totals is future reconstruction, not yet emitted.
  • 13 September 2026
    The section on wars added. An answer to the most frequent question of all was missing. The first version compared only the totals; this one also shows what they are made of, because that is where the surprise is: in the first fourteen days of war in Iran the missiles are worth 1% and the buildings hit together with the burned oil 84%. The sources are four and independent of each other. The totals remain floors: armies do not report and for Syria, Yemen and Sudan no count exists at all.
  • 13 September 2026
    The decade budget numbers and the budget counter corrected. An independent review re-downloaded the Global Carbon Budget 2025 CSV and found that the 2015–2024 averages on the page had been transcribed wrongly: land use 5.0 and not 5.6, ocean 11.8 and not 11.6, vegetation 8.7 and not 9.2, atmosphere 20.4 and not 19.7. The «unexplained imbalance» of 1.2 GtCO₂ over the decade does not exist: it is almost nil, while in 2024 alone it is −6 GtCO₂. The remaining budget for 1.5 °C must be counted from the start of 2026, not 2025: the counter went from 98 to about 140 GtCO₂. Country data also updated to 2024.
  • September 2026
    The revision notes about the original presentation removed. The page was born to check an existing slide («425 ppm in 2020», «3 times faster», «decade after decade») and annotated its errors under the charts. Now the page is the presentation: the charts are correct, and the notes were correcting something the reader no longer sees. What was true in them has stayed in the captions and in the table by decade.
  • September 2026
    The zoom on the last ten years added. On the curve since 1959 the 2023–2024 jump is invisible: the new chart shows the recent months and, below, the growth of each single year. This one too is redrawn with the NOAA data read live.
  • September 2026
    Updated to the Global Carbon Budget 2025. The land sink was revised from 11.7 to 8.7 GtCO₂ a year and the share staying in the atmosphere rose from 45% to about 50%: the largest revision, due both to new data and to the shift of the reference decade.
  • September 2026
    A section that mixed two series corrected. The previous version spoke of a 25% jump in the 2024 growth. At Mauna Loa that jump is not there (3.32 against 3.33 ppm); it is in the global mean (2.70 against 3.76). They were two different datasets quoted as if they were one.
  • September 2026
    The budget imbalance made visible. In a single year sources and sinks, measured with independent methods, never add up exactly: the disagreement is published as an item of its own.
  • September 2026
    The conversion factor corrected from 7.82 to 7.78 GtCO₂ per ppm, which was not consistent with the value in carbon declared next to it.

What is solid and what is not

Every figure carries a label saying what kind of number it is. observed are published measurements and inventories: the NOAA series, the country emissions of the Global Carbon Budget. observed and estimated is the global budget, where the fossil sources are inventories but the two sinks are reconstructed with models, which is the reason for the uncertainty whiskers under the blocks. estimate is a calculation or a model: the counterfactual without sinks, the accounting of wars. preliminary estimate is work that has not yet been through a scientific journal, and on this page it appears only once.

Fossil emissions are known to within about 5%. The ocean sink to within about 15%. The land sink and land use are the worst constrained items, with uncertainties that in single years exceed 50% of the value. The budget never closes exactly: the residual imbalance is published as an item of its own, and in 2024 it reached −6 GtCO₂ (−1.7 GtC), the most negative value in the series.

The cumulative figures for 1850–2024 concern fossil carbon only. Including historical land use the ranking changes significantly.

All values are rounded for readability. Before quoting one, check the exact figure at the primary source indicated above: the NOAA series and the Global Carbon Budget are revised with every edition.